日本語版

Topic hub — 2026-09-08

Fermentation and Bitter Peptides

Fermentation acts on both sides of bitter peptides: it can create them, and it can break them down. When microbes or koji enzymes break down protein-bearing foods, low-molecular-weight peptides rich in hydrophobic amino acids are formed, and these become the main carriers of bitterness. The Q-rule, which tries to predict from hydrophobicity alone which sequences turn bitter, broadly holds for cheese peptides, while counterexamples have been reported for soy and shrimp peptides. In cheese, changing the combination of bacterial strain and enzyme can either strengthen or weaken bitterness within the same fermentation framework, and in sake, both rice-protein-derived peptides and amino acids stand side by side as explanations for bitterness. Here, we lay out, separately and to the extent confirmed, the conditions under which peptides turn bitter, the process by which cheese writes bitterness and then erases it, the two accounts of sake's bitterness, receptor activation, and debittering processes.

What this page establishes

• That the formation of low-molecular-weight peptides rich in hydrophobic amino acids is the main cause of bitterness in protein hydrolysates has been confirmed across multiple studies.
• A rule that predicts bitterness from mean hydrophobicity broadly holds for cheese peptides, while counterexamples have been reported for soy and shrimp peptides.
• In cheese, multiple experiments have confirmed that the type of enzyme carried by the bacterial strain can either strengthen or weaken bitterness.
• Whether sake's bitterness derives from peptides or from amino acids has not been narrowed down to either side; there are only separate reports supporting each.

When do peptides turn bitter — the Q-rule and its counterexamples

Enzymatic hydrolysis of protein often produces bitterness. A review states that this bitterness is mainly caused by the formation of low-molecular-weight peptides made of hydrophobic amino acids. The relationship between a peptide's molecular weight and its bitterness is not uniform: one report found bitterness peaking at a molecular weight of about 3000 to 4000 Da (one hydrolysate) and 2000 Da (the other), with bitterness weakest in both hydrolysates in the fraction below 1000 Da.

The Q-rule attempts to put this empirical rule into numbers: it predicts that a peptide with mean hydrophobicity Q above 1400 cal/mol and a molecular weight in the range of 100 to 6000 is bitter. However, in a test on soy protein hydrolysates, the hydrophobicity data based on Q-values did not support this prediction; bitterness showed a nonmonotonic dependence on molecular weight, peaking at a mean molecular weight of about 4000 Da (one hydrolysate) and 2000 Da (the other), and the fraction below 1000 Da had the weakest bitterness in both hydrolysates.

In the earlier work the review cites, as Ney's (1971) original statement, whether a peptide with a Q-value between 1300 and 1400 cal/mol is bitter or not is not decided; only the two extremes are fixed — above 1400 cal/mol it is a bitter peptide, and below 1300 cal/mol it is not.

In the earlier work the review cites, a study examining peptides derived from the autolysis products of shrimp heads found that hydrophobic amino acid residues at the C-terminus contributed more strongly to bitterness than those at the N-terminus, and the bitter components were concentrated in the 3 to 5 kDa molecular-weight fraction. The fraction with lower hydrophobicity showed, conversely, stronger bitterness — a result running opposite to the Q-rule's premise that higher hydrophobicity means more bitterness.

Cheese — bacteria write it, then erase it

In the water-soluble fraction of aged Gouda, a total of 16 bitter peptides have been identified as casein breakdown products. 11 are newly reported β-casein-derived peptides (including 6 derived from β-CN(57-69) and 2 from β-CN(198-206), among others), and 6 are derived from αs1-casein. In aged sharp Cheddar, the 5 peptides with the highest bitterness intensity have all been identified as derived from β-casein.

A reconstitution model in which the 5 β-casein-derived peptides were added together at matched concentrations to young mild Cheddar reproduced bitterness intensity with no significant difference from real sharp Cheddar (reported within the scope of the abstract). Of the 5, the one identified as the main contributor was GPVRGPFPIIV, whose amount had swelled to 28.7 times as ripening progressed.

Cheese bitterness can also be rewritten by the specificity of the extracellular proteinase carried by the starter strain used. Cheese made with a strain lacking lactocepin stayed at a bitterness score of 1.8 (on a 0 to 9 scale) at 6 months, whereas multiple lactocepin-carrying strains with different specificities raised it to 3.5 to 4.7, with group h the most bitter.

Fresh cheese made from concentrated milk fermented with Lactococcus lactis strains carrying the extracellular proteinase PrtP showed bitterness scores of 3.9 and 4.1 (on a 0 to 5 scale), significantly more bitter than the 1.7 and 2.8 scores of strains without PrtP.

There is also a report on the other side — making bitterness itself less likely to occur by changing the combination of starter strains. Reduced-fat Cheddar made with a compound starter centered on SK11 as the acid-producing strain had markedly lower bitterness intensity than cheese made with a commercial Cheddar starter (reported within the scope of the abstract).

A review that addresses cheese bitterness as a whole treats bitterness as an unavoidable consequence of the proteolysis that proceeds during ripening, and organizes the correspondence between how far proteolysis has advanced and the amount of bitter peptide that accumulates, together with isolation and detection methods (reported within the scope of the abstract).

Miso, soy sauce, sake — soybean and rice fermentation

Miso and soy sauce are listed as entities under this topic, but no claim confirming bitter peptides in miso or soy sauce themselves could be found. For sake, on the other hand, there are multiple reports that discuss the origin of its bitterness systematically. What follows is centered on claims about sake.

In sake, the acidic subunit of glutelin, a steamed-rice protein, is digested in the moromi (mash) by koji-mold-derived enzymes, generating a characteristic oligopeptide bearing a pyroglutamyl group at its N-terminus; this has been shown to possibly be involved in sake's taste.

The 5 purified and isolated peptides (peaks 9,13,17,18,20) showed detection thresholds of 0.02 to 0.18 mM and recognition thresholds of 0.08 to 0.66 mM, and their taste characteristics also differed by peak, such as "bitterness, unpleasant aftertaste" and "bitterness, egumi." The index obtained by dividing the concentration in sake by the recognition threshold ranged from 6 to 60.

On the amino-acid side of the explanation, in a 3-point discrimination test in which arginine was added to pure synthetic sake at 1500 ppm — 10 times the concentration in commercial sake — 23 of 25 questions were answered correctly (significant at the 0.1% level), and in a 3-point preference test, comments of "bitterness, sweetness" were reported significantly more often for the arginine-added sample, at the 0.1% significance level.

A separate review statement attributes sake's distinctive bitterness to valine, isoleucine, leucine, proline, and arginine — the peptide account and the amino-acid account describe the same phenomenon at different layers.

As for the empirical rule that "sake made from rice with a high protein level has more zatsumi (off-taste)," this study's author discusses that it can be linked to protein level through the generation of glutelin-derived bitter peptides.

In the terminology system that defines sake's standard reference samples, it is recorded that, in storing pasteurized (heat-sterilized) sake, bitterness and stimulant taste (smooth) increase as a taste change, while astringency and stimulant taste (rough) decrease. The entry for unpasteurized sake (before pasteurization) records no increase or decrease in bitterness.

Bitter peptides at the receptor — cell activation versus bitterness on the tongue

The tasteless dipeptide Gly-Gly did not move the receptor even at a high concentration of 100 mM, but the bitter dipeptides Gly-Phe and Gly-Leu moved hTAS2R1 most strongly at 25 mM. When the chain lengthens by one to become a tripeptide, sensitivity rises further, and Phe-Phe-Phe in particular responded at the micromolar scale. What is being measured here is the cellular response in a heterologous-expression system, not the bitterness itself as felt on the tongue.

Rather than a single purified peptide, applying the whole fraction from a tryptic digest of casein (2.5 mg/mL, with bitterness equivalent to 0.125 mM quinine) does not narrow the responding receptors to just one. hTAS2R4, hTAS2R14, hTAS2R16, and hTAS2R1 all responded, with hTAS2R1 and hTAS2R4 showing the strongest responses. However, this fraction is a mixture of many structurally different peptides, and the experimenters themselves acknowledge that nonbitter components are likely mixed in.

Recognition thresholds measured on the tongue line up on a different scale altogether. Among the 12 synthesized and purified peptides, those showing clear bitterness had recognition thresholds of 0.05 to 6.0 mmol/L, with even the lowest, the decapeptide YPFPGPIHNS, at 0.05 mmol/L. Whereas cell-level EC50 values are spoken of on the micromolar order, recognition thresholds felt on the tongue are spoken of on the millimolar order; both use the same word "bitterness," but they are measuring different layers.

Debittering processes — enzymes, adsorption, masking

Methods for reducing the bitterness of protein hydrolysates are broadly organized into three groups: selective separation (activated-carbon treatment, alcohol extraction, isoelectric precipitation, silica gel chromatography, hydrophobic-interaction chromatography), masking that simply covers the bitterness itself, and biological methods such as further hydrolysis by aminopeptidases, alkaline/neutral proteases, and carboxypeptidases, or Lactobacillus starter adjuncts (reported within the scope of the abstract).

Microbial aminopeptidases are reviewed as enzymes that reduce bitterness by selectively cleaving, from the N-terminal side, hydrophobic amino acids exposed at peptide termini. Commercial aminopeptidases are summarized as being used not only for debittering hydrolysates but also for deep hydrolysis of protein and the production of seasonings, cheese, and bioactive peptides (reported within the scope of the abstract).

In milk protein hydrolysates used as a cheese-flavoring raw material, a relationship has been shown in which strengthening hydrolysis with a protease that specifically targets certain amino acid residues lowers the bitterness sensory score (reported within the scope of the abstract). However, the bitterness of the whole mixture could not be explained by simply adding up the bitterness of individual peptides; the analysis found that how much peptide existed in a particular molecular-weight band mattered more.

In a comparison of conditions for hydrolyzing sesame protein with the three enzymes Alcalase, trypsin, and Flavourzyme, it is reported that increasing the proportion of Flavourzyme lowered the bitterness value measured by an E-tongue from 5.98 under Alcalase alone to 5.08 under the Alcalase-Flavourzyme condition.

For sake too, attempts to reduce bitter peptides with enzymes have been reported. In the earlier work the review cites, adding a peptidase preparation with high acid carboxypeptidase activity to sake is reported to have lowered bitter peptides and increased (pGlu)L-ethyl, which carries a pyroglutamyl group. A separate abstract in the same review also states that an enzyme showing carboxypeptidase-type activity "contributes to reducing bitter peptides while also being involved in generating dipeptides with poor taste," showing that operations that reduce bitterness can carry a quality trade-off as well.

Frequently asked

What is the difference between bitter and nonbitter peptides?

Most bitter peptides are rich in hydrophobic amino acids, and having a hydrophobic amino acid at the terminus is considered important for bitterness to appear. A rule that puts this empirical pattern into numbers predicts that a peptide with mean hydrophobicity above 1400 cal/mol and a molecular weight in the range of 100 to 6000 is bitter. However, a test on soy protein hydrolysates did not support this prediction, showing a nonmonotonic relationship in which bitterness peaked at molecular weights of about 4000 Da and 2000 Da. For peptides derived from shrimp autolysates, an opposite result has also been reported, in which the fraction with lower hydrophobicity showed, conversely, stronger bitterness. The range over which bitterness can be decided by numbers alone is limited, depending on the food and the molecular-weight band.

Why can cheese sometimes turn bitter?

There is a review that holds cheese bitterness to arise as a result of the proteolysis that proceeds during ripening. From the water-soluble fraction of aged cheese, 16 bitter peptides have been identified as casein breakdown products, and in aged sharp Cheddar as well, the 5 peptides with the strongest bitterness have been identified. Of these, the peptide identified as the main contributor is reported to have increased in amount to 28.7 times as ripening progressed. A reconstitution model that returned the same 5 peptides to young Cheddar is also reported to have shown no significant difference in bitterness intensity from real sharp Cheddar. These are data from specific ripening conditions and cheese types, and whether they can be generalized to all cheeses has not been confirmed.

Can choosing the right bacterial strain prevent cheese from turning bitter?

Within the scope of what has been reported, the choice of bacterial strain can either strengthen or weaken bitterness. Cheese made with a strain lacking an extracellular proteinase stayed at a bitterness score of 1.8 at 6 months, while multiple strains carrying the enzyme with different specificities raised it to 3.5 to 4.7. On the other hand, reduced-fat Cheddar made with a compound starter centered on an acid-producing strain is reported to have had markedly lower bitterness intensity than cheese made with a commercial starter. However, these are results under specific experimental conditions, and even the authors do not go so far as to say they apply directly to commercial mixed starters in general.

Is sake's bitterness caused by amino acids or by peptides?

There are reports supporting both. On the peptide side, research has shown that a subunit of steamed-rice protein is digested in the moromi by koji-mold enzymes, generating a characteristic oligopeptide that has been shown to possibly be involved in the taste. The 5 purified peptides show detection thresholds of 0.02 to 0.18 mM and recognition thresholds of 0.08 to 0.66 mM. On the amino-acid side, in a discrimination test with arginine added at 1500 ppm — 10 times the level in commercial sake — 23 of 25 questions were answered correctly, and comments of bitterness and sweetness were reported significantly more often. Another review attributes the bitterness to several amino acids, and which side predominates has not been settled.

What is not known

• Why the bitterness of the whole mixture cannot be explained by simply adding up the bitterness of individual peptides has not yet been sufficiently measured.
• We have not been able to reach a source that confirms bitter peptides in miso or soy sauce themselves.
• How far the increase or decrease in bitterness from strain selection is reproduced in actual cheesemaking using commercial mixed starters has not been measured.
• No source has been found that directly compares whether sake's peptide account or amino-acid account contributes more to bitterness.

Read in the encyclopedia

The encyclopedia entries below are in Japanese.

• bitter peptides — an umbrella term for hydrophobic oligopeptides released from protein by fermentation, ripening, or enzymatic hydrolysis
• sake bitter peptides from rice glutelin — a group of oligopeptides derived from the acidic subunit of glutelin, a steamed-rice protein
• Ney Q-rule — Ney's empirical rule predicting bitterness from mean hydrophobicity
• cheese — a ripened dairy product in which casein breakdown drives both the formation and breakdown of bitter peptides
• sake — a brewed beverage made from rice, rice koji, and water by multiple parallel fermentation
• miso — a seasoning made by fermenting and ripening soybeans with koji mold
• soy sauce (shoyu) — a liquid seasoning made by fermenting and ripening soybeans, wheat, and salt
• Basque burnt cheesecake — a cheesecake finished by charring its surface at high heat
• akebi peel with meat miso — a dish of grilled akebi peel topped with meat miso
• takenoko miso-ni (bamboo shoot simmered in miso) — a Toyama regional dish of bamboo shoots simmered in kombu, miso, and bonito dashi

Go deeper in the research notes

• Cheese Bitterness: Bacteria Write It, Then Erase It — follows the story from casein-derived peptide identification to the increase or decrease of bitterness through strain selection
• Fermentation Creates Bitterness, Then Breaks It Down — Miso, Soy Sauce, Sake — crosses through where bitterness is born and where it disappears at each stage of fermentation
• Who's to Blame for Sake's Bitterness? — The Peptide Account and the Amino-Acid Account — follows where the glutelin-peptide account and the amino-acid account overlap and diverge

Source ledger

The primary sources and limitations of the claims listed here can be checked from the source link inside each card. The full list is gathered in the Source ledger.