The Entourage Effect Grew Up: What Five Years of Receptor Studies Actually Measured

Terpenes do things at CB1 and CB2 in a dish. Here is who measured it, how, and what it still cannot tell you about a jar on a shelf.

By Cannible Editorial

Cannabis terpenes — the volatile molecules that give a cultivar its smell — have been measured acting at CB1 and CB2, the two cannabinoid receptors that THC acts on. That is a real change. For most of the time the entourage effect has been sold to consumers, the receptor-level claim behind it was inference. Since 2019 it has been an experiment, and the experiments have been done.

Read the next sentence in the same breath as the last one, because the two only mean something together. Every one of those experiments was run in cultured cells, in frog eggs expressing human receptors, or in mice. Across the whole body of work assembled here — six papers from four research groups — the number of human participants is zero. And the laboratories do not agree with each other: one independent group reported activation, two unrelated independent groups looked for it and found none, and the largest run of positive papers comes from a single laboratory using a single apparatus, several of whose authors are employees of a cannabis manufacturer.

So the honest confidence level is low, and it is low in a specific way. It is not that nobody has measured anything. It is that what has been measured is contested at the bench, has never been tested in a person, and does not connect any number printed on a menu to anything a reader will feel. The entourage effect is the reason terpene percentages appear on those menus at all — which is why the gap between what was measured and what is implied is worth an article. This piece reports laboratory findings. It is not medical advice, and nothing in it describes an effect you can expect from a product.

When we say the entourage effect grew up, that is Cannible's reading of one specific change and not a finding in any of these papers: the kind of argument available for it moved from deduction to measurement. It did not move from measurement to prediction.

What the entourage effect was before anyone measured it

The most useful description of the pre-2021 state of the field comes from the researchers who then went and tested it. Writing in Scientific Reports in 2021, the University of Arizona group characterised the existing case as "deductive reasoning arguments, some clinical suggestions, and a few pre-clinical investigations", adding that "there is also skepticism within the literature."

That is the baseline. Not a body of receptor evidence that later work confirmed, and not a myth that later work debunked — an argument from plausibility, with dissent already in it. Everything below is measured against that starting point, and this article does not narrate any origin story for the term beyond what that paper states.

What the evidence actually consists of

Before the findings, the shape of the record. Six papers bear directly on the question, from four research groups. Asked the same questions in the same order, they line up like this.

Three things follow from the listing that no individual paper states, and they are the reason the rest of this article is organised the way it is. Counting papers gives four positive and two negative. Counting independent groups gives one positive and two negative. And every row of the human-participant column reads zero.

What one independent laboratory measured in mice and cells

The load-bearing independent positive result is the 2021 Arizona paper. It reported that the terpenes α-humulene, geraniol, linalool and β-pinene "produced cannabinoid tetrad behaviors in mice, suggesting cannabimimetic activity" — the tetrad being the standard four-behaviour screen used to identify cannabinoid-like activity in rodents. In parallel cell work, "in vitro experiments showed that all terpenes activated the CB1R, while some activated other targets," and the terpenes were selectively additive with the synthetic cannabinoid agonist WIN55,212.

Both halves of that sentence are the finding, and the species is not a footnote to it. This is a mouse behaviour result and a cultured-cell result. Nobody has run the tetrad, or anything else in this literature, in a person. What was measured in mice is whether the animals behaved as they do under a cannabinoid; whether a terpene produces a subjective high in a human being has not been tested by any study registered here.

The concentrations are part of the finding

The paper's own account of how much terpene it took is the single most useful number in this literature for a reader trying to interpret a label, and it is routinely dropped from summaries of the study.

The authors describe "the generally high concentrations of terpene needed to see activation," and quantify it: "> 10 μM, or up to 500 μM depending on the terpene, was needed to see activation." In the animals the requirement was lower but not small — "the doses needed in vivo were not as extreme, producing full responses in most assays for most terpenes at 200 mg/kg."

Those are laboratory quantities in laboratory units, and this article will not convert them into anything human-scaled. No arithmetic on this page will tell you what fraction of a gram, an inhalation or a serving they correspond to, because the conversion has not been done in a study and doing it here would be invention. What the numbers do establish is a boundary condition on the whole positive result: whatever is happening, it happened at concentrations the researchers themselves flagged as high. That same fact turns up again below, as one of the reasons the laboratories disagree.

The authors are not certain it is CB1 doing the work

The Arizona group did not present direct CB1 agonism as settled, and an article that reports their result without their caveat has reported half of it. They observed that some of the mouse behaviours "could be blocked by cannabinoid or adenosine receptor antagonists, suggesting a mixed mechanism of action" — an adenosine-receptor antagonist blocking an effect is not what a clean CB1 story predicts.

They then set out two alternatives to direct CB1 agonism in their own words: "(1) direct modulation of membrane dynamics, shifting CB1 activation equilibrium to favor the activated receptor; and (2) terpene modulation of endocannabinoid synthesis and/or degradation." And they noted that these alternatives "are further supported by the generally weak competition binding observed at the CB1 by terpenes."

These are the authors' hypotheses about their own data, not an objection raised here. Their standing matters for how the headline claim should be read: the strongest independent positive result in this literature is a result whose producers are openly unsure which target produced it.

Three more papers, one laboratory, one commercial sponsor

The remaining positive papers — 2023, and two carrying 2026 issue dates — form a sequence. The first reported that sixteen individual cannabis terpenes activated CB1 in "a controlled in-vitro heterologous expression system", at "about 10-50% of the activation by THC alone." The second extended the work to CB2, reporting "dose-dependent responses" for many terpenes "reaching a maximal response of about 10-60 % the activation elicited by THC," and concluding that the results "support the role of cannabis terpenes as partial agonists at CB1R and CB2R." The third applied isobolographic analysis to THC–terpene combinations and reported that it "suggested additive THC-terpene interactions for some terpenes, and synergistic interactions for others, including borneol, limonene, sabinene, terpineol, α-pinene and ocimene at CB1R, and β-caryophyllene and linalool at CB2R."

Three disclosures belong with those findings, and each belongs here rather than in a footnote.

The commercial interest. The papers' own competing-interest declarations state that several authors are employees of the Bazelet group. The 2023 declaration reads: "NR, AME, DBZ and DHS are employees of the Bazelet group, a medical cannabis manufacturer in Israel." The CB2 paper declares three Bazelet employees among its authors; so does the isobolographic paper. That is disclosed by the researchers, in the papers, and it is reported here on their authority.

The apparatus. These measurements were made in Xenopus oocytes — frog egg cells engineered to express a receptor of interest. The system is a standard and legitimate way to ask whether a molecule activates a receptor. It is a measuring instrument, not a person, not a tissue, and not a nervous system. A response in an oocyte establishes that a receptor can be activated under those conditions and nothing at all about what a body does with the molecule afterwards.

These three papers are not independent replication of each other, and they do not claim to be. The CB2 paper states the continuity itself: "We have previously reported results of activating cannabinoid receptor type 1 (CB1R) by several terpenes that are most common in cannabis. Here we employed the same Xenopus oocytes functional heterologous expression system to complement the CB1R data." One author group, one apparatus, one commercial sponsor, three papers. Counted honestly, that is one line of evidence reported three times, not three confirmations — and the distinction is the whole difference between a literature that is converging and a literature that is accumulating.

One further limit applies to this section specifically, and Cannible states it rather than working around it. The full texts of these three papers could not be obtained; the publisher's site returned an access error to every attempt. What was read is the publisher-supplied abstract and the publisher-supplied competing-interest declaration on each paper's US National Library of Medicine record. The findings and disclosures above are quoted from those. Effect sizes beyond the percentages quoted, the concentrations used, and whatever limitations the authors state inside the papers are unread, and nothing in this article should be taken as an appraisal of their methods.

A conclusion shaped like a product

The 2023 paper does not stop at its measurement. It concludes: "As the most effective terpenes are not necessarily the most abundant ones in the cannabis plant, reaching 'whole plant' or 'full spectrum' composition is not necessarily an advantage. For enhanced therapeutic effects, desired compositions are attainable by enriching extracts with selected terpenes."

That is an argument for a product category, published by a paper on which four of the authors declare employment by a cannabis manufacturer. Cannible is not adjudicating whether the argument is right, and is certainly not repeating it as guidance — the underlying measurement is an oocyte result, and no oocyte result supports a therapeutic claim of any kind. What is being reported is the conjunction: a commercially interested author group published a receptor finding whose stated implication is that a particular way of making product is preferable. A reader deciding how much weight to give the finding is entitled to know that.

Two unrelated laboratories looked and found nothing

The negative results are not a caveat to the positive ones. They are two independent groups, on two continents, asking the same question first, and they belong at full weight.

In 2019 a group working at Macquarie University and the University of Sydney tested six of the most common cannabis terpenes — α-pinene, β-pinene, β-caryophyllene, linalool, limonene and β-myrcene — at human CB1 and CB2 receptors expressed in AtT20 cells, at concentrations up to 30–100 μM. Their reported result was flatly negative: "None of the six of the most common terpenes in Cannabis directly activated CB1 or CB2, or modulated the signaling of the phytocannabinoid agonist Δ9-THC." Their conclusion was correspondingly specific: "these results suggest that if a phytocannabinoid-terpenoid entourage effect exists, it is not at the CB1 or CB2 receptor level." They did not conclude that terpenes do nothing — they proposed that "it seems more likely that they may act at different molecular target(s) in the neuronal circuits important for the behavioral effect of Cannabis."

In 2020 a group at the University of Otago tested five terpenes, alone and in mixtures, using radioligand binding alongside functional assays. Their finding: "With the possible exception of a weak interaction of β-caryophyllene with CB2, no data were produced to support the hypothesis that any of the five terpenes tested (either alone or in mixtures) have direct interactions with CB1 or CB2" — and "similarly, terpene functional effects were also not detected, either alone or in combination with Δ9-tetrahydrocannabinol, cannabidiol, or 2-arachidonoylglycerol." They read their own work as cumulative: "this study adds to the evidence that the putative entourage effect cannot be explained by direct effects at CB1 or CB2."

Both papers carry declared interests of their own, and they are reported to the same standard as the positive ones. Two authors on the 2019 paper are inventors on patents involving cannabinoid therapeutics and one acts as a consultant to a therapeutics company; one author on the 2020 paper is affiliated with Soma Group. In each case the declared interest sits on the side of cannabinoid therapeutics, which is the direction their results did not go.

Why the results disagree, and who says so

The clearest account of the disagreement comes from the Arizona group, and it should be read as what it is: a hypothesis offered by one side about why the other side's instruments may have missed something, not a refutation and not an independent adjudication.

They point to four differences. The cell line used. The single signalling output measured — the 2019 study read one downstream channel, and a receptor can be activated in ways that a single readout does not see, which those authors also acknowledge. Whether terpenes were tested in mixtures or alone. And, most concretely, concentration: their discussion contrasts a "maximum concentration of 10 μM, while we observed receptor activation at higher concentrations."

That last point does real work, and it does not reach equally far. The 10 μM ceiling belongs to the 2020 Otago study, and against the Arizona threshold of more than 10 μM and up to 500 μM, a 10 μM ceiling is a genuine candidate explanation for a null result. It fits the 2019 Sydney study far less well: that group's own abstract records testing up to 30–100 μM and still measuring nothing. A concentration argument that explains one null result and not the other has explained one null result.

What none of these papers contains is the experiment that would settle it — the same terpenes, at the same concentrations, in more than one system, run by groups with nothing riding on the answer.

What none of this tells you

Everything above is a statement about receptors in laboratory systems. None of it is a statement about a person, and the distance between those two things is where almost every consumer-facing version of this science goes wrong.

Receptor activation in an expression system is not an effect. Additivity in an oocyte is not a feeling. A tetrad score in a mouse is not a high. And no study registered in this article connects any terpene percentage — on a label, on a certificate of analysis, on a menu — to any outcome in any human being, because no study registered in this article involved a human being.

That is not a hedge added for safety. It is the boundary the authors themselves work inside: the isobolographic paper's own language is that its findings "suggest that cannabis terpenes may act as both partial orthosteric agonists and allosteric modulators," which is the vocabulary of a proposal. One of the papers uses the phrase "clinical effect level" to describe a concentration; that is the authors' framing of a quantity, not a demonstrated clinical effect, and it should not travel any further than the sentence it was written in.

What would change this conclusion

The open questions are specific, and each of them is answerable.

Which target is actually responsible remains unresolved by the strongest independent positive study's own account, with membrane dynamics and endocannabinoid turnover still open alongside direct CB1 agonism. Whether the concentrations that produce activation in a dish are reachable in a living person has not been established anywhere in this record. Whether a positive result in one apparatus survives a different apparatus is precisely what is in dispute, and the only way to answer it is for a group with no stake in the outcome to run the oocyte findings somewhere other than oocytes. And whether any of this has ever produced an effect in a human being is not an open question so much as an unopened one — there is no human study here to argue about.

The scale of what has not been looked at is worth stating plainly. A 2026 analysis of 28 samples across six cultivars identified 227 volatile terpenes in cannabis, "including 88 monoterpenes and 139 sesquiterpenes." That paper is plant chemistry and says nothing about receptors or effects; it is cited here for the count and nothing else. But set the count beside the literature above — sixteen terpenes in the largest positive study, six and five in the negative ones — and the proportion of this plant's chemistry that has ever been put in front of a cannabinoid receptor is small.

A decade of marketing arrived at the entourage effect before the measurements did. The measurements have now started, and what they have produced so far is one independent positive result its own authors cannot fully explain, three papers from one commercially interested laboratory on one apparatus, and two independent laboratories that found nothing. That is a live scientific question. It is not yet a reason to buy anything.

Sources

View on Cannible →