A Rare Class of Plant Molecule Turned Up in Cannabis Leaves, and Nobody Has Tested What It Does

Using a two-dimensional liquid chromatography method coupled to high resolution mass spectrometry, researchers tentatively identified 79 phenolic compounds across inflorescence and leaf samples from three commercial cannabis cultivars. Sixteen of those, falling into four structural classes, were tentatively identified as flavoalkaloids, a rare hybrid of flavonoid and alkaloid chemistry not previously reported in cannabis.

The flavoalkaloids appeared mainly in the leaf extracts of one cultivar out of three. That single detail is arguably more clinically useful than the discovery itself, because it quantifies how far the chemistry of one cannabis plant can sit from another.

Cannabis product selection in clinical practice runs almost entirely on cannabinoid numbers. A patient is told a flower is 22 percent THC with a given terpene profile, and that is treated as a reasonably complete description of what they are about to take.

It is not close to complete. The phenolic fraction has received little attention compared with cannabinoids and terpenoids, and this analysis suggests why: separating these compounds from one another requires instrumentation that most laboratories do not run. What cannot be measured routinely tends to be treated as though it were not there.

The paper is Comprehensive two-dimensional liquid chromatographic analysis of Cannabis phenolics and first evidence of flavoalkaloids in Cannabis, by Magriet Muller and André de Villiers of the Department of Chemistry and Polymer Science at the University of Stellenbosch. It appeared in Journal of Chromatography A on May 5, 2025, volume 1754, article 466023.

The method is an on-line comprehensive two-dimensional separation, pairing hydrophilic interaction chromatography in the first dimension with reversed-phase chromatography in the second, coupled to high resolution mass spectrometry. Conditions were set using an automated optimization program, and the resulting separation reached a peak capacity above 3,000 with high orthogonality between the two dimensions.

Those numbers matter more than they look. Peak capacity is roughly the number of distinct compounds a method can resolve, and running two genuinely different separation mechanisms in sequence multiplies resolving power rather than adding to it. This is the difference between seeing a broad hump and seeing the individual compounds inside it.

Diode array and high resolution mass spectrometry data supported tentative identification of 79 compounds across inflorescence and leaf samples of the three cultivars, mainly flavone and hydroxycinnamic acid derivatives. The Stellenbosch announcement in August 2025 reported that 25 of these had not previously been described in cannabis.

Three C-glycosylated flavones and several of their O-glycosylated derivatives set one cultivar apart from the other two. In that same cultivar, flavoalkaloids were detected, mainly in the leaf extracts. Four classes of flavoalkaloids comprising 16 flavone derivatives were tentatively identified.

The caveat the authors state plainly deserves equal billing. The structures of the alkaloid portions could not be unambiguously assigned from the mass spectrometry data available. The compound class is identified with confidence. The individual structures are not.

Flavonoids are among the most familiar plant compounds in medicine and nutrition, present in berries, citrus, tea, and a long list of herbs, and studied extensively for antioxidant and anti-inflammatory properties. Alkaloids are the other great class of plant bioactives, the family that includes caffeine, morphine, and nicotine.

A flavoalkaloid joins the two, carrying an alkaloid fragment attached to a flavonoid core in a single molecule. As Muller described the significance, most plants contain highly complex mixtures of phenolic compounds, and while flavonoids occur widely in the plant kingdom, the flavoalkaloids are very rare in nature. They are known from a handful of species, tea among them, but they are not something a chemist expects to find.

Their absence from the cannabis literature until now appears to be partly an instrumentation story rather than a biology story. De Villiers put it directly: the performance of two-dimensional liquid chromatography allowed separation of the flavoalkaloids from the much more abundant flavonoids, which is why these rare compounds could be detected in cannabis for the first time.

In cultivation and in processing, fan leaves are discarded. They carry less THC and CBD than flower, extraction targeting cannabinoids has no reason to include them, and the economics of the industry push everything toward the resinous parts of the plant.

De Villiers framed the implication of finding rare compounds concentrated there: the analysis again highlights the medicinal potential of cannabis plant material currently regarded as waste. The method the group used had been developed and tested first on rooibos tea, grapes and wine, and was then applied to cannabis.

That framing is worth taking at exactly its stated weight. Potential is not activity. What has been shown is that discarded plant material contains a chemical class worth examining, which is a reason to fund the examination and not a reason to consume the leaf.

The most practically transferable finding is the variability. The flavoalkaloids showed up mainly in one cultivar of three. The distinguishing C-glycosylated flavones did the same. Across a small sample of commercially grown plants, the phenolic profiles diverged substantially.

This is the part that bears on clinical practice, and it cuts in a direction that is uncomfortable for the field. If three commercial cultivars differ this much in a chemical fraction nobody measures, then two products carrying the same THC and CBD percentages may not be interchangeable, and there is currently no way for a clinician or a patient to know.

It is also a caution against generalizing from three plants. Three cultivars grown in one place under one set of conditions cannot establish what the range looks like across the species, and growing conditions, harvest timing, drying, and curing all alter phenolic content in plants generally.

As analytical chemistry, this is careful work. The two-dimensional separation is the right tool for a matrix this crowded, the peak capacity and orthogonality figures are reported rather than asserted, and the group-type separation between phenolic acids and flavonoids gives the identifications structural logic beyond a mass match.

As evidence for anything clinical, it is a starting point and nothing more. Every identification is described as tentative, no compound was isolated and confirmed against a reference standard, and no biological assay of any kind was performed. The paper establishes presence. Presence is the weakest link in the chain that ends at a therapeutic claim.

Three cultivars from one country, grown under one set of conditions, is a small foundation for statements about cannabis as a species. Phenolic content in plants responds strongly to light, water stress, soil, harvest timing, and post-harvest handling, so some of the variation between these cultivars may reflect how they were grown rather than what they are.

Tentative identification also means something specific in analytical chemistry. It means the mass, fragmentation pattern, and elution behavior are consistent with a proposed structure, not that the structure has been confirmed by isolation and independent characterization. The authors were explicit that the alkaloid portions could not be assigned unambiguously, and any summary that drops that qualifier is misreporting the paper.

The study does not show that flavoalkaloids have any biological effect in humans, animals, or cells. It does not show that cannabis leaf is medicinally useful, that full-spectrum preparations outperform isolates, or that these compounds contribute to any clinical outcome patients experience.

It also does not establish the concentrations involved. Detecting a compound with a highly sensitive method says nothing about whether the amount present could be pharmacologically relevant at any realistic exposure, and that question is unaddressed here.

Cannabis chemistry has been mapped outward in stages: cannabinoids first, then terpenes, then flavonoids as the entourage conversation matured. Each stage was limited by what the available instruments could resolve, and each stage produced claims that ran ahead of the data before settling. Flavoalkaloids are the current edge of that expansion and can be expected to follow the same arc.

The history of plant medicine offers a steadying comparison. Willow bark was used for centuries before salicylic acid was isolated, and the opium poppy’s chemistry extends well past morphine. Identifying what is in a plant has reliably preceded understanding what any of it does, often by a long interval, and the interval is where most overclaiming happens.

The most common misunderstanding I encounter is that cannabis is THC and CBD. That framing drives product selection, clinical advice, and regulatory thinking, and it is incomplete in a way that keeps getting more measurable. This paper adds to that.

What I keep coming back to is the variability, not the discovery. One of three cultivars carried these compounds in its leaves. If that degree of difference exists in a fraction nobody tests for, then the confidence with which patients are told two products are equivalent because their cannabinoid numbers match is not earned. I see the consequences of that in clinic: patients who respond well to one preparation and poorly to another with a nearly identical label.

I want to be careful about what I am claiming. I am not saying flavoalkaloids explain that. Nobody knows whether these compounds do anything at all, and the honest position is that the chemistry is real and the pharmacology is entirely unstudied. What this supports is the advice I already give, which is to stay with a consistent, traceable source once something works, because the thing that made it work may not be on the label.

Flavoalkaloids have been detected in cannabis for the first time, tentatively, mainly in the leaf material of one of three cultivars analyzed. No biological activity was tested and none should be inferred. The usable clinical takeaway is about variability: cannabis plants differ chemically in ways that current product labeling does not capture.

Take away the measurement and the variation. A rigorous method found a rare compound class in a part of the plant routinely discarded, and found it in one cultivar and not the others. Leave behind any sentence that attaches a health benefit to those compounds, because the study that would support such a sentence has not been done.