Informatic prediction of Cheddar cheese flavor pathway changes due to sodium substitution


  • Balasubramanian Ganesan,

    Corresponding author
    1. Dairy Technology and Innovation Laboratory, Western Dairy Center, Department of Nutrition, Dietetics, and Food Sciences, Utah State University, Logan, UT, USA
    • Correspondence: Balasubramanian Ganesan, Dairy Technology and Innovation Laboratory, Western Dairy Center, Department of Nutrition, Dietetics, and Food Sciences, 8700 Old Main Hill, Logan, UT 84322, USA. Tel.: +1 435 7603765;

      fax: +1 435 7972178;


    Search for more papers by this author
  • Kelly Brown

    1. Dairy Technology and Innovation Laboratory, Western Dairy Center, Department of Nutrition, Dietetics, and Food Sciences, Utah State University, Logan, UT, USA
    Search for more papers by this author


Increased interest in reduced and low sodium dairy foods generates flavor issues for cheeses. Sodium is partly replaced with potassium or calcium to sustain the salty flavor perception, but the other cations may also alter metabolic routes and the resulting flavor development in aged cheeses. The effect of some cations on selected metabolic enzyme activity and on lactic acid bacterial physiology and enzymology has been documented. Potassium, for example, is an activator of 40 enzymes and inhibits 25 enzymes. Currently, we can visualize the effects of these cations only as lists inside metabolic databases such as MetaCyc. By visualizing the impact of these activating and inhibitory activities as biochemical pathways inside a metabolic database, we can understand their relevance, predict, and eventually dictate the aging process of cheeses with cations that replace sodium. As examples, we reconstructed new metabolic databases that illustrate the effect of potassium on flavor-related enzymes as microbial pathways. After metabolic reconstruction and analysis, we found that 153 pathways of lactic acid bacteria are affected due to enzymes likely to be activated or inactivated by potassium. These pathways are primarily linked to sugar metabolism, acid production, and amino acid biosynthesis and degradation that relate to Cheddar cheese flavor.


Cheese and yoghurt are classical examples of fermentative microbial biotechnology products. The flavor components of aged cheeses such as Cheddar, Gouda, and Parmesan are mainly small molecules derived from enzymatic reactions that break down protein and lipids during cheese aging (Fox & Wallace, 1997). Especially for Cheddar cheese, the component balance theory suggests that an overall balance of the flavors contributed by individual compounds is responsible for flavor perception (Mülder, 1952). The development of cheese flavors is strongly associated with metabolism by added or adventitious lactic acid bacteria (Ganesan & Weimer, 2007). To foster research progress in this arena, we currently host a Pathway Tools webserver with pathway genome databases (PGDB) of lactic acid bacteria that illustrate their diverse metabolic capabilities (

Due to health implications, consumers increasingly prefer reduced and low sodium foods. However, this generates novel issues for dairy food manufacture as it alters manufacturing processes (McMahon, 2010). Recently, manufacturing protocols for reduced and low sodium Cheddar cheese have been developed (McMahon et al., 2012). Up to 66% sodium reduction was successfully achieved while consistently producing cheeses of comparable moisture, pH, and fat content. However the flavor acceptance of the reduced sodium cheeses was lower than regular sodium cheese, with the low sodium cheese becoming unacceptable at 6 months of age (McMahon, 2010). The lower flavor acceptance of low sodium cheese may be due to variations in bacterial metabolism, which is usually controlled by salt addition.

An alternative to sodium reduction is salt substitution, wherein the sodium cation can be replaced with potassium and calcium (Grummer et al., 2013). But salt substitution also alters cheese flavor. For example, calcium substitution beyond 10% total salt causes bitterness in cheese (Fitzgerald & Buckley, 1985). Other attempts are also currently underway to use potassium and admixtures of salts for sodium reduction by cation replacement (Grummer et al., 2012). Earlier studies by Reddy & Marth (1993ab, 1995ab) suggest that no significant differences in microbial levels or cheese properties are caused by equal substitution of sodium with potassium but those results do not concur with other studies that included a higher proportion of sodium than potassium (Fitzgerald & Buckley, 1985; McMahon et al., 2012). Some studies observe that acceptable sodium-substituted cheeses can be manufactured by controlling the proportion of sodium substitution (Gomes et al., 2011). However, the overall flavor impact of these substitutions in Cheddar cheese is yet to be fully understood.

Certain cations are known to alter cheese flavor-related enzymatic functions or the transcription of related genes in lactic acid bacteria, especially the starter lactococci. Potassium, ammonium, magnesium, manganese, and cobalt cations activate the glycolytic enzyme pyruvate kinase in lactococci (Crow & Pritchard, 1977). Hence, potassium and magnesium addition to cheese is likely to improve rates of glycolysis and lead to more lactic acid production. Notably, Crow & Pritchard (1977) found that sodium or lithium cations only weakly activated pyruvate kinase. Potassium appears to be involved in cytoplasmic pH regulation in lactococci, which in turn regulates the cells' ability to transport glutamate (Poolman et al., 1987), which is essential for cheese flavor development via metabolism of other amino acids (Ganesan & Weimer, 2007). Hence, potassium addition to cheese may spur flavor development via early induction and acceleration of amino acid metabolism. A lactococcal keto acid decarboxylase is stimulated by magnesium, manganese, and calcium, and additionally by sodium (de la Plaza et al., 2004). However, potassium's ability to stimulate this enzyme's activity was not tested in that study. Lactococcal transcriptional studies by Xie et al. (2004) under different stress conditions found that salt (sodium chloride) stress represses transcription of arginine metabolism and fatty acid biosynthesis and induced enolase transcription. Among pathways related to cheese flavor, compound metabolism altered specifically by salt stress, sugar transport and aromatic amino acid and arginine synthesis was repressed and branched chain amino acid synthesis was induced, whereas genes of sugar and peptide metabolism were selectively induced or repressed.

Nonstarter lactic acid bacteria consist chiefly of the genus Lactobacillus, members of which are also modulated by different cations. For example, Lactobacillus pentosus growth is inhibited less by potassium chloride than by sodium chloride (Bautista-Gallego et al., 2008). In another study, Camien & Dunn (1958) showed that 52% of 1000 screened strains of L. casei, an important cheese adjunct bacterium (Trepanier et al., 1992), were inhibited by potassium. These observations have important implications for cheese flavor, because lactobacilli are regarded by some groups as essential microbes for producing diverse flavors (Peterson & Marshall, 1990; Trepanier et al., 1991). Any cation substitutions that alter populations of lactobacilli would also alter the flavor outcomes.

Metabolic pathways of lactobacilli are also selectively modulated by potassium. For example, d-hydroxy-fatty acid metabolism of L. casei is restricted by potassium chloride, but this effect is alleviated by the presence of sodium chloride (Camien & Dunn, 1957). Hence, potassium addition may not alter L. casei levels in cheese as long as sodium is included. A wine-derived L. brevis strain produces a diol dehydratase that converts 2,3-butanediol to methyl-ethyl ketone and then to 2-butanol, a reaction that is not noted in cheese matrices. But the reaction is relevant because 2,3-butanediol is produced by lactococci and is a chief flavor component of Cheddar cheese (El-Atar et al., 2000; Gogus et al., 2006). This dehydratase is activated by ammonium and potassium and inhibited by lithium, sodium, magnesium, and manganese (Radler & Zorg, 1986). This enzyme also reacts with 1,2-propanediol and glycerol, and hence may produce more diverse ketones and alcohols, compound classes important for cheese flavor (Ganesan & Weimer, 2007). The overall impact of such alterations on cheese flavor is yet to be determined.

Salt in the form of NaCl is added to Cheddar cheese specifically to restrict microbial proliferation and limit unwanted bacteria and metabolic pathways from creating unexpected flavors (Weimer et al., 1997). Once sodium is reduced or replaced, the metabolic routes and the resulting flavors are likely to be altered as well. Sodium stress in the lactic acid bacteria only causes a general stress response (Xie et al., 2004), reduces transcription of genes related to sugar and pyruvate metabolism, and induces glutamate biosynthesis. The levels of glutamate are considered by some groups as beneficial for cheese flavor for providing amino-groups for aminotransferases (Yvon et al., 1997; Tanous et al., 2002), whereas others identify its contribution to umami flavor perception (Drake et al., 2007). Notably, aminopeptidases are important for casein breakdown in cheese, and are controlled by sodium levels (Weimer et al., 1997). On the other hand, multiple studies have revealed that potassium is an activator of over 40 enzymes but inhibits nearly 25 enzymes (relevant information and literature about these activities are included in metabolic databases such as metacyc; Similarly, multiple studies have determined that calcium activates nearly 29 enzymes but also is an inhibitor of c. 55 enzymes ( Stu-dies that pertain to these cations' role in enzymatic regulation, however, are not based on cheese or similar matrices nor does the metacyc database include information pertinent to the lactic acid bacterial enzymes for these cations. Although the effects listed in metacyc are manifest at the enzyme level, the impact of these ions on bacterial metabolism is yet to be determined. However, the ability of these cations to modulate multiple classes of enzymes needs to be better understood in the context of altering cheese flavor.

Currently, we can only visualize the effects of these cations as lists via the Pathway Tools web servers (example links provided above). However, the metabolic impact of these activating and inhibitory activities has not been visualized in the context of the pathways and products thereof. By visualizing the pathways and products that are likely to be affected beneficially or detrimentally, we can predict and eventually direct the aging process of cheeses with substituted salts.

The objectives of this study were to create initial templates of pathway databases containing enzymes and their related metabolic pathways altered by the cation potassium, developing a workflow to update the databases frequently, and to collate sources of literature that describe/propose metabolic routes or chemical reactions for cheese flavor compounds that are altered by various ions.

Materials and methods

pathway tools software (Karp et al., 2010) provided by SRI International (version 13, Menlo Park, CA) is hosted as a webserver at procyc ( and contains metabolic databases for 11 lactic acid bacteria that have been sequenced previously (Makarova et al., 2006). procyc was used as the base platform for building two metabolic databases, one for enzymes activated by potassium (potasacticyc; and another for enzymes inhibited by potassium (potasinacticyc; Relevant gene sequences for the enzymes of interest present in lactic acid bacteria PGDBs such as the genera Lactococcus, Streptococcus, and Lactobacillus were downloaded from procyc and concatenated into a single large sequence to create a ‘chromosome’ and the positions of the genes were specified accordingly. A ‘chromosome’ is an essential input component as a genetic element for constructing PGDBs. We created this artificial form of input solely to provide template information for visualizing metabolic networks by enzymology and do not propose its use as a gene analysis tool. The initial templates were created within the software using the Pathologic feature (Karp et al., 2011) that automatically constructs a database using inputs of gene sequences and EC numbers. These databases were then easily modified by migrating pathways from the parent lactic acid bacteria PGDBs into the newly constructed databases using features available within the software. The newly initiated PGDB's ‘input’ folder was then provided annotation and gene sequence files. Once the files were created, the Pathologic suite was applied to complete PGDB creation. Further, links to literature sources were added to the enzyme page of the newly created databases and directly linked to abstract pages at the PubMed website ( A comparison of potasacticyc and potasinacticyc was done on the procyc webserver by the ‘Comparative Analysis’ tool to determine which pathways were modified by potassium (Table 1). All procedures to extract information from PGDBs used in this study were performed as described in the pathway tools user manual (SRI International).

Table 1. Comparison of pathways altered by potassium that likely affect cheese flavora
Pathway classPathway potasinacticyc potasacticyc
  1. a

    Lines highlighted in bold show total pathways altered and not cheese-flavor-related pathways alone. Only pathways relevant to flavor formation in cheese are listed and thus, do not add up to the total numbers.

Biosynthesis (total pathways)   10 69
Biosynthesis – amino acid biosynthesisS-adenosyl-l-methionine cycle ×
Arginine biosynthesis II (acetyl cycle) ×
Arginine biosynthesis III ×
Asparagine biosynthesis II ×
Aspartate biosynthesis ×
Glutamate degradation II ×
Homoserine biosynthesis××
Isoleucine biosynthesis II ×
Leucine biosynthesis ×
Lysine biosynthesis I××
Lysine biosynthesis II ×
Lysine biosynthesis III ×
Lysine biosynthesis VI ×
Methionine biosynthesis II ×
Ornithine biosynthesis ×
Threonine biosynthesis from homoserine ×
Biosynthesis – cofactors, prosthetic groups, electron carriers BiosynthesisGlutathione biosynthesis ×
Biosynthesis – other biosynthesis2-methylbutyrate biosynthesis ×
Degradation/utilization/assimilation (total pathways)   5 47
Degradation/utilization/assimilation – amino acid degradationAsparagine degradation I ×
Glutamate degradation II ×
Glutamate degradation VI (to pyruvate) ×
Isoleucine degradation I ×
l-cysteine degradation II ×
Lysine degradation IV× 
Lysine fermentation to acetate and butyrate ×
Methionine degradation I (to homocysteine) ×
Tryptophan degradation II (via pyruvate) ×
Degradation/utilization/assimilation – C1 compound utilization and assimilation3-hydroxypropionate cycle ×
3-hydroxypropionate/4-hydroxybutyrate cycle ×
Degradation/utilization/assimilation – Carbohydrate degradationLactose and galactose degradation I ×
Degradation – Carboxylate degradationAcetate formation from acetyl-CoA I ×
Generation of precursor metabolites and energyLysine fermentation to acetate and butyrate ×
Mixed acid fermentation ×
Pyruvate fermentation to acetate I ×
Pyruvate fermentation to acetate II ×
Pyruvate fermentation to ethanol II××
Pyruvate fermentation to lactate××
Total   20 133

Results and discussion

Using the above-described approach, we were able to successfully create the PGDBs potasinacticyc (; Fig. 1) and potasacticyc (; Fig. 2). Cumulatively, these databases showed that at least 20 and 133 pathways of lactic bacteria are likely to be inactivated and activated by potassium, respectively. Of the 153 pathways, 10 in potasacticyc and 69 in potasinacticyc belong to the ‘Biosynthesis’ class and 5 in potasacticyc and 47 in potasinacticyc are part of the ‘degradation/utilization/assimilation’ pathways.

Figure 1.

Overview of the metabolic database potasinacticyc, which contains enzymes inhibited by potassium.

Figure 2.

Overview of the metabolic database potasacticyc, which contains enzymes activated by potassium.

A selection of the altered pathways (Table 1) belongs to those involved in cheese flavor generation. Primary steps to cheese flavor generation relate to carbohydrate fermentation to pyruvate (Fox & Wallace, 1997) and further metabolism of pyruvate, of which some are likely to be exclusively activated (e.g. mixed acid fermentation and pyruvate fermentation to acetate, Table 1), whereas others are likely to be both activated and inactivated (e.g. pyruvate fermentation to ethanol and lactate, Table 1). The pathways that belong in both categories likely share enzymes, one of which is activated and one inactivated by potassium, which may lead to accumulation of certain metabolic intermediates and alter their routes.

From a cheese flavor perspective, lactate is usually the end product of lactose metabolism by lactic acid bacteria, and tends to accumulate in cheese (Fox et al., 1990). Although this observation usually implies that it is seldom reused or recycled for energy, Pediococcus is capable of oxidizing lactate, if present in sufficient numbers (Thomas et al., 1985), as is L. plantarum (Murphy et al., 1985). The diversion of pyruvate to acetate due to potassium addition may be beneficial to lactic acid bacteria that can utilize acetate to produce other flavor compounds (Thomas, 1987). However, the flavor desirability of the alternate metabolic end products in cheeses also needs to be addressed.

Volatile sulfur compounds are associated with positive cheese flavor and are generated by sulfur-containing amino acid metabolism by lactic acid bacteria (Weimer & Dias, 2005). Here, we found that cysteine degradation is predicted to be activated by potassium (Table 1), suggesting that potassium addition to cheese may induce beneficial flavors. Likewise, the biosynthesis of branched chain fatty acid 2-methylbutyrate is also predicted to be induced by potassium (Table 1). Branched chain fatty acids also correlate with good cheese flavor (Ganesan & Weimer, 2007). Lysine degradation generates acetyl-CoA but is predicted to be inhibited by potassium. Notably, some pathways relate to assimilation of other minerals such as magnesium and iron (data not shown), suggesting that the effect of potassium substitution will be also linked to other cation substitutions currently under study. These predicted changes allow us to track the hypothesis that potassium addition is beneficial for Cheddar cheese flavor.

To validate the utility of these databases we qualitatively compared our predictions with the available literature on potassium replacement in cheeses and flavor changes. Grummer et al. (2013) and Karagozlu et al. (2008) found that partial substitutions of potassium led to cheeses with lower pH, which is due to higher amounts of lactic acid. The latter study showed that lactic acid levels increased in white pickle cheese by 12.7%. This corresponds with the predicted enhancement of pyruvate-to-lactate fermentation in potasacticyc, and also with increased pyruvate kinase activity of lactococci with potassium addition (Crow & Pritchard, 1977). Another study (Ayyash & Shah, 2010) showed that potassium substitution in Halloumi cheese induced acetic acid production, which is also predicted in potasacticyc. Increase in acetic acid and lactic acid with higher levels of potassium substituting sodium are also reported in Feta cheese (Karimi et al., 2012). Karagozlu et al. (2008) also observed increased release of tyrosine in white pickle cheese; this was not, however, predicted by our approach. A number of additional studies about potassium substitution in cheeses are available (Horita et al., 2011; Ayyash et al., 2012ab; Grummer et al., 2012) but unfortunately they do not address flavor compound generation from non-glycolytic metabolism.

Some groups have started to address salt cation substitution issues on a larger scale with multiple levels of potassium substitution ranging from 25% to 75% and partial replacement of potassium also with smaller levels of magnesium and calcium (Kang et al., 2012; McMahon et al., 2012). The preliminary observations from these studies include a faster pH drop after curd salting with potassium-containing salt mixtures that indicates rapid glycolysis and increase in lactic acid production, predicted by higher pyruvate kinase activity in lactococci and by potasacticyc. Kang et al. (2012) also observed that taste panelists detected higher intensities of sulfur flavor in cheeses with potassium, suggesting that sulfur compound levels increase with potassium levels, which is also inferred from potasacticyc.

Qualitatively, we find that many predictions made by potasacticyc and potasinacticyc relate well to or are inferred from experimentally manufactured cheeses, which highlights the strength of our approach. Uncorrelated observations such as higher tyrosine levels indicate the complexity of the cheese flavor formation process and strengthen the need for more extensive studies and subsequent curation of these databases. We hope that quantitative information from salt cation substitution studies will help us in the future to correlate our predictions numerically. Incorporation of new datasets from gene expression and metabolome profiles will add to these databases the molecular and genetic evidence essential to define the ability of cations to alter bacterial metabolism. Biotechnological production of food flavors in different liquid (yoghurt and wine) and solid (bread and cheese) matrices with altered cation contents can use this approach to understand the feasibility of maintaining beneficial attributes of the fermentation processes. This approach of constructing databases can also be extended to similar data that attempt to define the effects of antimicrobial and pro-microbial molecules such as prebiotics, and is thus applicable to current and future microbiome studies from different ecological niches.


The study was funded by Dairy Management Inc. and administered by the Dairy Research Institute (Rosemont, IL).