Topic hub — 2026-09-19
When bitterness and harm are the same compound
In most bitter foods, the bitterness and everything else ride on different molecules. The bitterness of beer and its alcohol, the bitterness of coffee and its caffeine, can be moved independently inside the same drink. This page collects the cases where they cannot. Amygdalin in almonds and apricot kernels; the steroidal glycoalkaloids of potato and tomato. What the tongue picks up and what matters when the amount rises are carried by one and the same molecule.
That narrows the question to one. Do domestication, ripening and processing remove the bitterness alone, or do they remove both together? What follows lays out only the claims that still hold their measurements, in the order compound, amount, breeding, process.
On this page (8 sections)
What this page establishes
- Both amygdalin and alpha-tomatine have bitterness and harm attributed to the same molecule. An operation that removes the bitterness is not handling two separable things.
- In potato, the correlation between sensory bitterness ratings and total tuber glycoalkaloid content is reported at r=0.93, with the 7 clones above 14 mg/100g rated bitter. Bitterness moves with the amount.
- The sweet almond kernel came from a single substitution (Leu to Phe) in the dimerisation domain of the bHLH2 transcription factor. The mutation that removed the bitterness is what made domestication possible.
- Ripe tomatoes lose their bitterness because an enzyme active during ripening hydroxylates alpha-tomatine into something else; in lines where that enzyme is silenced, the compounds accumulate instead.
- For apricot kernels, debittering processes have been studied, while raw kernels are measured against a separate yardstick, an acute reference dose of 20 µg/kg body weight. The debittering work and the dose figure do not sit in the same source.
When one compound carries both
The amygdalin that wild almonds accumulate is described in the literature itself as a bitter and toxic cyanogenic diglucoside. That is an attribution in words rather than a bitterness measurement, but it shows plainly that the two properties are tied to one compound name.
Scope: Wild almond and bitter-kernel lineages.
Note: Not psychophysical bitterness measurements, but standard descriptions in plant science literature.
Show data
compound-amygdalin —has_sensory_attribute→ sensation-bitterApricot kernels involve the same molecule. The statement that it converts to cyanide after eating, and the statement that it accounts for 2–5% of the kernel, come from different sources. By content it is not the most abundant component.
Scope: Raw apricot kernels.
Show data
ingredient-bitter-apricot-kernel —contains→ compound-amygdalinScope: Layer: Food. Concentration description in the background explanation of research on ultrasound-assisted debittering of bitter-variety apricot kernels (prior research citation).
Show data
ingredient-bitter-apricot-kernel —contains→ compound-amygdalinIn the nightshades the compound changes. Alpha-tomatine is the compound for which high content and bitterness were linked long ago in wild cherry tomatoes from the Rio Mayo valley in Peru, and recent genome work has identified 9 bitter-tasting accessions.
Scope: Fully ripe fruit of wild cherry tomato (Solanum lycopersicum var. cerasiforme) from Peru's Río Mayo valley.
Note: Cárdenas et al. 2019 presents this association in its text as a citation of earlier work by Charles Rick and colleagues. It was not a new, independent human sensory test conducted by Cárdenas et al. themselves.
Show data
compound-alpha-tomatine —has_sensory_attribute→ sensation-bitterIt is not confined to the fruit. The abstract states that it accumulates in all tissues and at particularly high levels in leaves and immature green fruits. Choosing which part to eat is already choosing how much bitterness.
Scope: Layer: food. The background description of α-tomatine tissue distribution and taste in a recombinant-protein study.
Note: Only the abstract was checked; the full text was not reached. The abstract gives no specific concentration values. Earlier claims concern the GAME31 mechanism by which ripe fruit loses bitterness; this claim adds the association of high concentration and bitterness in different tissues, immature fruit and leaves.
Show data
ingredient-tomato —contains→ compound-alpha-tomatineDoes bitterness move with the amount
If both properties ride on the same molecule, the tongue's rating and the analytical value ought to move together. In potato that has been measured directly. Of 13 clones, the 7 whose total glycoalkaloid content exceeded 14 mg/100g were rated bitter by a sensory panel, and the correlation between content and rating was r=0.93.
Scope: 13 clones of Solanum tuberosum.
Note: Sensory-panel assessment by the flavor-profile method. Total phenol content was reported to have no significant relation to bitterness or burning sensation.
Show data
ingredient-potato —has_bitterness_correlated_with→ "Total tuber glycoalkaloid content. Of 13 clones, the 7 exceeding 14 mg/100 g were rated bitter by a sensory panel; the correlation between content and bitterness rating was highly significant (r=0.93)."That figure comes from an abstract, though, and the panel size and scale could not be confirmed. A high correlation cannot be read as the tongue standing in for the instrument. As the next section shows, the people managing the cultivar record drew their line on measured content, not on taste.
What domestication and breeding removed
How the almond became food answers this page's question most directly. Map-based cloning in a reference genome and a taste-segregating F1 population showed that the bHLH2 transcription factor controls transcription of PdCYP79D16 and PdCYP71AN24. A single substitution (Leu to Phe) in its dimerisation domain halts that transcription and produces the sweet-kernel trait. The mutation that removed the bitterness is what allowed domestication.
Scope: Map-based cloning of the almond reference genome and an F1 population segregating for flavor.
Show data
compound-amygdalin —has_biosynthesis_controlled_by→ "The bHLH2 transcription factor. A point mutation in its dimerization domain (Leu→Phe) stopped transcription of PdCYP79D16 and PdCYP71AN24, producing the sweet kernel (bitterness-loss) trait and enabling almond domestication."Potato has had the same thing done by modern means. Silencing the GAME4 gene is described as capable of reducing steroidal glycoalkaloid content in leaves and tubers by up to 74-fold. That review does not connect the content directly to sensory bitterness.
Scope: Solanum tuberosum.
Show data
ingredient-potato —has_debittering_most_likely_due_to→ "Up to a 74-fold reduction in steroidal glycoalkaloid (SGA) content in leaves and tubers through silencing of the GAME4 gene."A case of getting the removal wrong is on record too. 'Lenape' in the United States and 'Magnum Bonum' in Sweden were both widely grown before being withdrawn from the market over high tuber content. That the decision was made on content rather than on taste is worth reading next to the correlation in the previous section.
Scope: History of Solanum tuberosum cultivars.
Note: The source review records only the withdrawal, without giving precise content values or withdrawal years. A quantitative primary source (Zitnak & Johnston 1970, among others) was located bibliographically on Crossref, but its abstract was not read and it is not included.
Show data
ingredient-potato —has_history→ "'Lenape' in the United States and 'Magnum Bonum' in Sweden were both withdrawn from the market after having been widely cultivated because of high steroidal glycoalkaloid content in their tubers."What ripening and processing remove
Even without changing the cultivar by breeding, the fruit removes its own bitterness as it ripens. GAME31, a 2-oxoglutarate-dependent dioxygenase, hydroxylates alpha-tomatine to hydroxytomatine, the first step of a route that converts it into non-bitter esculeoside A and related compounds in the ripe fruit. In lines where the enzyme was silenced, glycoalkaloids normally present only in traces in ripe fruit rose by up to 22-fold against the wild type.
Scope: Tomato, Solanum lycopersicum (cultivated varieties including cv. M82)
Note: In lines where GAME31 was silenced, SGA in ripe fruit increased up to 22-fold relative to the wild type, confirming through reverse genetics that loss of this enzyme can bring bitterness back.
Show data
ingredient-tomato —has_debittering_most_likely_due_to→ "Hydroxylation of the bitter steroidal glycoalkaloid alpha-tomatine to hydroxytomatine by GAME31 (a 2-oxoglutarate-dependent dioxygenase). During ripening, a pathway starting from this reaction converts it into non-bitter esculeoside A and related compounds, which removes the bitterness of the ripe fruit."A parallel route through a different enzyme sits inside the same ripening. Sl23DOX hydroxylates at position 23 and the product isomerises spontaneously to neorickiioside B. This is not a tidy picture in which stopping one pathway brings the bitterness back.
Scope: Layer: food. Biochemical analysis with recombinant Sl23DOX protein and confirmation in Sl23DOX-overexpressing and silenced transgenic tomato plants.
Show data
ingredient-tomato —is_processed_by→ "Sl23DOX, an α-tomatine 23-hydroxylase and a 2-oxoglutarate-dependent dioxygenase, hydroxylates position 23. Its product spontaneously isomerizes to neorickiioside B, forming an intermediate in α-tomatine metabolism during ripening. This is a parallel metabolic pathway involving an enzyme distinct from GAME31."Stopping the ripening itself has also been measured. In red-ripe fruit of mutants lacking the ripening regulators NOR and RIN, the bitter glycoalkaloids increased significantly while non-bitter esculeoside A fell. Turning red and losing the bitterness are under the same control.
Scope: Layer: food. Comparison of red-ripe fruit from NOR-KO and RIN-KO tomato mutants with wild type; glycoside quantification by LC-MS and GAME gene-expression analysis by RT-qPCR.
Show data
compound-alpha-tomatine —occurs_in→ ingredient-tomatoOn the processing side, an ultrasonically accelerated debittering technique for apricot kernels has been examined. It shortens the treatment time against conventional methods while substantially improving beta-glucosidase activity, colour and texture, with little effect on oil and protein content. No human rating of bitterness intensity was carried out; the assessment stops at physical and biochemical indicators.
Scope: Layer: Food. An experimental study on bitter apricot kernels comparing physical properties before and after debittering with UV-Vis spectroscopy, fluorescence spectroscopy, circular dichroism, electrophoresis, environmental scanning electron microscopy, thermal analysis, and other methods.
Note: These are main experimental findings of the study itself, checked in its full text. They add an account of debittering as a processing technology, separate from EFSA’s ARfD assessment.
Show data
ingredient-bitter-apricot-kernel —can_be_debittered_by→ "Ultrasonically accelerated debitterizing. It shortens processing time compared with conventional methods and substantially improves β-glucosidase activity, color, and texture, while having little effect on oil and protein contents."Alongside the debittering work, raw kernels carry a separate yardstick of amount. The acute reference dose for cyanide is 20 µg/kg body weight; for adults, up to 3 small raw kernels (370 mg) stays below it, while more than 3, or less than half of a large kernel, may exceed it, and for toddlers 1 small kernel may exceed it. The debittering research and this statement of amount do not meet inside one paper.
Scope: Consumption of raw apricot kernels.
Note: This is from an official press announcement. The underlying scientific opinion (EFSA Journal 2016;14(4):4424) has not been read, and this is not an intake recommendation. A correction changed the wording for large kernels from 「1個超」 (in translation: "more than 1 kernel") to 「半分未満」 (in translation: "less than half a kernel").
Show data
ingredient-bitter-apricot-kernel —can_exceed_acute_reference_dose_at→ "The acute reference dose (ARfD) for cyanide is 20 µg/kg body weight. EFSA states that up to 3 small raw apricot kernels (370 mg) do not exceed the adult ARfD, but more than 3 small kernels or less than half a large kernel may exceed it. 1 small kernel may exceed it for a young child."In almonds, the bitterness that was not fully removed survives as a difference between cultivars. Across 6 raw cultivars, bitterness and bitter-almond flavour differed by cultivar and were strongest in Molar and Pegarinhos. Across the same 6 cultivars roasted at 138 degC for 33 minutes, the cultivar difference persisted, strongest in Pegarinhos and Refêgo.
Scope: 6 cultivars harvested in Portugal in 2019. A trained panel of 12 rated raw samples on a 1-5 point intensity scale
Note: This is a comparison among cultivars, not a bitterness ranking for edible almonds in general. The authors' speculation that this is due to benzaldehyde is not adopted here.
Show data
ingredient-almond —has_sensory_response→ "Among 6 cultivars of raw almonds, bitterness and bitter-almond flavor differed by cultivar, and were strongest in Molar and Pegarinhos"Scope: Unpeeled kernels of the same 6 cultivars were roasted at 138°C for 33 minutes, and a trained panel of 12 rated them on a 1-5 point scale in a session separate from the raw samples
Note: This is a cultivar comparison within a single treatment, not a result directly comparing bitterness between raw and roasted samples. The causal component is not established.
Show data
ingredient-almond —has_sensory_response→ "Among 6 cultivars of almonds roasted at 138°C for 33 minutes, bitterness and bitter-almond flavor also differed by cultivar, and were strongest in Pegarinhos and Refêgo"Roasting temperature itself acts on bitterness. Comparing 150 to 190 degC, the 190 degC samples showed intense burnt notes, and consumer liking was negatively linked to pyrazines, bitterness, astringency and burnt flavour. That this runs in the opposite direction from cacao, where higher roasting temperature lowered perceived bitterness, can be read against thecacao hub.
Scope: Layer: food. Spanish 'Vairo' almonds grown under regulated deficit irrigation, roasted in a convection oven. 35 volatile compounds (GC analysis), a descriptive sensory profile, and a consumer preference test.
Note: Based on an abstract quotation via a search engine and Europe PMC (the full text was not confirmed on either Wiley or PubMed, both blocked with a 403). Roasting at 190°C is described as giving a deep color and burnt notes, 150°C as under-roasted, and 170°C for 10 minutes was concluded to be the optimum condition. Consumers' overall liking is reported to have shown a positive correlation with certain volatile compounds and with roasted/nutty sensory attributes, while showing a negative correlation with pyrazines, bitterness, astringency, woodiness, and burnt flavor. Statistical values (correlation coefficients, p-values) were not confirmed. An example of a roast-degree series for a nut (almond) other than coffee.
Show data
ingredient-almond —has_perceived_bitterness_affected_by→ "Roasting temperature (150-190°C)"What was measured and did not move
Measurements that produced no difference are given the same weight as those that did.
For 6 cultivars blanched for 30 seconds in boiling water, skinned, and roasted at 138 degC for 33 minutes, the only sensory attributes that differed by cultivar were the two about hardness at first chew, and bitterness and bitter-almond flavour carried no significance marks. It looks as though removing the skin erased the cultivar difference, but this is a comparison within one treatment group, not a direct comparison between treatments.
Scope: 30 seconds in boiling water, skin removed by hand, room-temperature drying, roasting at 138°C for 33 minutes. A trained panel of 12 rated the samples on a 1-5 point scale
Note: This is a result showing that, within this treatment group, bitterness did not distinguish the cultivars. It is not a result directly testing whether bitterness fell relative to raw or unpeeled-roasted samples.
Show data
ingredient-almond —has_sensory_response→ "Among 6 cultivars that were blanched for 30 seconds to remove the skin and then roasted at 138°C for 33 minutes, the only sensory attributes showing a cultivar difference were the 2 hardness-related attributes"On the growing side nothing moved. In raw 'Vairo' almonds under 4 irrigation regimes, bitterness intensity ran 0.5 to 0.6 and astringency 0.5 to 0.7, with no significant differences. This is a single cultivar, a single year and a single orchard.
Scope: A single orchard in Spain, harvested in 2017. A trained panel of 10 rated each sample 3 times on a 0-10 point scale
Note: This is a single-cultivar, single-year, single-orchard result, and it is not generalized to say that irrigation has no effect on almond bitterness. Whether the ± on irrigation volume denotes standard deviation or standard error is not specified in the text, so it is not included here.
Show data
ingredient-almond —has_sensory_response→ "In raw almonds of the 'Vairo' cultivar, there was no significant difference in bitterness intensity (0.5-0.6) or astringency intensity (0.5-0.7) across the 4 irrigation conditions"Questions people ask
Does "bitter means dangerous" hold here?
Within the scope of this page there is one case where amount and bitterness moved together. In potato the correlation between total glycoalkaloid content and bitterness rating was r=0.93, and the 7 clones above 14 mg/100g were rated bitter. But the correlation comes from an abstract whose panel size and scale could not be confirmed, and the decision to withdraw 'Lenape' and 'Magnum Bonum' was made on content. Nothing on this page records a line drawn on taste alone.
Are sweet almonds and bitter almonds different species?
They are not. A single substitution (Leu to Phe) in the dimerisation domain of the bHLH2 transcription factor is reported to halt transcription of the two genes involved in amygdalin synthesis and to produce the sweet-kernel trait. One difference at one position separates edible from inedible.
Why do tomatoes stop being bitter as they ripen?
Because an enzyme converts the bitter compound into a different one during ripening. GAME31 hydroxylates alpha-tomatine, opening the route to non-bitter esculeoside A and related compounds in the ripe fruit. In lines where that enzyme is stopped, the glycoalkaloids rise by up to 22-fold even in ripe fruit, and in fruit lacking the ripening regulators NOR and RIN the bitter side rises while esculeoside A falls. Colouring up and losing the bitterness are not separate.
Is removing the bitterness by processing enough?
For apricot kernels an ultrasonic debittering method has been studied, but that study carried out no human rating of bitterness intensity and stops at indicators such as beta-glucosidase activity. Separately, raw kernels carry an acute reference dose of 20 µg/kg body weight. The two sit in different sources, and this page does not add them together.
What is not known
- Within the scope of this page, the bitterness of both amygdalin and alpha-tomatine is attributed from the literature and from compound identification rather than from human sensory measurement. No intensity value can be placed here.
- The potato correlation of r=0.93 is a statement in an abstract whose panel size and scale could not be confirmed. How far that correlation holds across conditions is not known.
- The statement that silencing GAME4 lowers content by up to 74-fold is not tied to sensory bitterness. How far the tongue registers that drop is not on this page.
- The debittering process for apricot kernels and the statement of amount for raw kernels come from different sources. What happens to the amount after processing cannot be said here.
- The disappearance of cultivar differences in blanched roasted almonds is a result within one treatment group, not a direct comparison of skin on against skin off. How much of the difference the skin carries is not known.
- For the withdrawal of 'Lenape' and 'Magnum Bonum', the years and content figures are absent from the review, and the primary sources have not been read.
Read in the encyclopedia
- Amygdalin — the molecule to which both properties are attributed, accumulated by wild almonds and bitter-kernel lines
- Almond — the seed made sweet by one mutation; cultivar and roasting temperature still act on its bitterness
- Bitter apricot kernel — debittering is studied while raw kernels carry a yardstick of amount
- Alpha-tomatine — the nightshade side; accumulates in all tissues, highest in leaves and immature fruit
- Tomato — ripening enzymes remove the bitterness; stop the enzyme and it stays
- Potato — the only entry here where rating and content were correlated directly, and where the withdrawals are recorded
All entries in this hub (6)
Every encyclopedia entry that belongs to this hub, grouped by category. Entries without an English version open in Japanese.
ingredientalmond · bitter apricot kernel · potato · tomato
compoundamygdalin · α-tomatine
Go deeper in the research notes
- Bitterness is plural — why the compound carrying the bitterness differs from subject to subject
- A world map of debittering — techniques for taking bitterness out, by region and by process
- Cooking edits bitterness — adding, removing, designing and suppressing, in four quadrants
- Does roasting make bitterness? — almond roasting set beside cacao and coffee
Source ledger
The primary sources and limitations for the claims on this page can be checked from the source links inside each card. The full list is on the Japanesesource ledger.
How to cite this page
Aoki, M. (NIGAMI LAB) (2026). When bitterness and harm are the same compound. https://nigamilab.com/en/topics/toxic-bitter-seeds/
Each claim card has a stable anchor (#clm-…). The dataset is CC BY 4.0; see the data page. Dataset DOI: 10.5281/zenodo.22739101