What Cannabinoid Receptors Do Inside Immune Cells, and Why That Is Not the Same as Fighting Infection

The CB2 receptor was found in immune tissue before anyone looked for it there for therapeutic reasons. Munro and colleagues cloned it in 1993 from macrophages in the marginal zone of the spleen, explicitly because CB1 did not explain the non-psychoactive effects of cannabinoids. Human quantification two years later showed CB2 expression in immune tissue at 10 to 100 times the level of CB1.

What CB2 does not do is recognize pathogens. That job belongs to pattern recognition receptors, principally the Toll-like receptor family, which detect conserved molecular signatures on bacteria, fungi and viruses. Cannabinoid receptors modulate what happens after that recognition occurs. The distinction sounds academic and is the whole clinical story.

This question reaches the clinic in concrete forms. A transplant recipient asks whether cannabis will interfere with immunosuppression. A patient with well-controlled HIV asks whether daily use is affecting their counts. A patient reads that cannabidiol has antibacterial activity in a dish and wonders whether to take it for a sinus infection.

Those three questions have three different answers, and only one of them has decent human data behind it. Lumping them together under the heading of cannabis and the immune system is how patients end up making decisions the evidence does not support.

Innate immunity begins with pattern recognition. Toll-like receptors and related sensors detect pathogen-associated molecular patterns, meaning conserved structures such as bacterial lipopolysaccharide or viral nucleic acid that do not appear on host cells. Engagement of those sensors triggers intracellular cascades, principally nuclear factor kappa B and the interferon regulatory pathways, which turn on inflammatory gene expression.

Cannabinoid receptors are not in that detection layer. CB1 and CB2 are G protein-coupled receptors that signal through inhibitory G proteins to adenylyl cyclase and mitogen-activated protein kinase. They do not bind bacterial or viral structures and they do not identify pathogens. What they can do is change the size and shape of the response once a pattern recognition receptor has already fired.

This is the single most common error in consumer writing on this topic. Tables that pair a cannabinoid receptor with a bacterium or a virus, as if the receptor sees the organism, describe a biology that does not exist. The real relationship is one step downstream, and it is a relationship of modulation, not recognition.

S. Galiegue and colleagues quantified cannabinoid receptor transcripts across human immune tissue and blood cell subpopulations in the European Journal of Biochemistry in 1995. CB2 was abundant in immune tissue at 10 to 100 times the CB1 level, with spleen and tonsil carrying as much CB2 messenger RNA as the central nervous system carries CB1. Among blood cells the order was B lymphocytes, then natural killer cells, then a clear drop to monocytes, neutrophils, CD8 T cells and CD4 T cells.

CB2 expression is also inducible. Guy Cabral and LaToya Griffin-Thomas, reviewing the field in Expert Reviews in Molecular Medicine in 2009, emphasized that receptor availability rises during inflammation, which means a baseline expression map understates what a drug could act on in inflamed tissue.

The functional consequence shows up most clearly in knockout animals. Caroline Turcotte, Marie-Renee Blanchet, Michel Laviolette and Nicolas Flamand at Universite Laval reviewed this literature in Cellular and Molecular Life Sciences in 2016 and found the recurring result to be that mice lacking CB2 have an exacerbated inflammatory phenotype. The reasonable reading is that CB2 signaling acts as a brake on immune activation. It is a brake in mice. Whether it is a usable therapeutic brake in humans is a different claim, and the human trial record does not yet support it.

Ewa Kozela and colleagues at the Weizmann Institute ran one of the cleaner experiments on this question, published in the Journal of Biological Chemistry in 2010. They activated BV-2 mouse microglial cells with lipopolysaccharide, a Toll-like receptor 4 agonist, and measured what THC and cannabidiol did to the resulting inflammatory output.

Both compounds reduced production and release of interleukin 1 beta, interleukin 6 and interferon beta. They did so through partly different routes: cannabidiol reduced activity of the nuclear factor kappa B pathway and increased activation of the STAT3 transcription factor, while THC did neither, and both reduced activation of STAT1.

Then comes the finding that reframes the whole subject. The anti-inflammatory action did not appear to involve CB1, CB2 or the abnormal-cannabidiol-sensitive receptors. A large part of what cannabinoids do to innate immune signaling happens through mechanisms that are not cannabinoid receptor mediated at all. Any argument that runs from CB2 expression directly to a cannabis product’s anti-inflammatory effect has skipped the step where the experiment says the receptor was not required.

The most direct animal evidence on infection comes from the University of South Florida group. C. A. Newton, Thomas Klein and Herman Friedman reported in Infection and Immunity in 1994 that BALB/c mice given intravenous THC at 4 mg per kilogram one day before a sublethal Legionella pneumophila infection had increased mortality after a later challenge. The mechanism tracked with a shift away from Th1 activity: reduced interferon gamma, reduced Legionella-specific lymphoproliferation, elevated Th2-associated antibody and depressed Th1-associated antibody.

A follow-up in 1997 compared other cannabinoids. Cannabinol and cannabidiol at 8 mg per kilogram did not change mortality; at 16 mg per kilogram they produced a slight to moderate increase. The synthetic full agonist CP 55,940 at 6 mg per kilogram killed roughly half the animals, matching THC. So the effect tracked with potent cannabinoid receptor agonism rather than with cannabinoids in general.

Evidence in the opposite direction exists and deserves equal airtime. Patricia Molina and colleagues at LSU Health Sciences Center gave rhesus macaques chronic intramuscular THC at 0.32 mg per kilogram twice daily starting 28 days before inoculation with simian immunodeficiency virus, reporting in AIDS Research and Human Retroviruses in 2011 that THC decreased early mortality and was associated with attenuated plasma and cerebrospinal fluid viral load. Two careful animal studies, two opposite directions, different pathogens, different dosing, different species.

One randomized controlled trial is the anchor. Donald Abrams and colleagues at the University of California San Francisco enrolled 67 patients with HIV-1 infection into a 21-day inpatient trial at San Francisco General Hospital, randomizing them to smoked cannabis at 3.95 percent THC, oral dronabinol 2.5 mg, or placebo, three times daily. The results were published in Annals of Internal Medicine in 2003.

Sixty-two patients were evaluable. Adjusted for baseline variables, the estimated average effect on change in log viral load from baseline to day 21 was minus 0.07 for cannabis and minus 0.04 for dronabinol, with confidence intervals crossing zero in both cases. Neither CD4 nor CD8 counts appeared adversely affected, and protease inhibitor levels were not meaningfully changed.

That is a short trial in a specific population, and it is the best human immunological safety evidence the field has. It does not license the claim that cannabis is immunologically inert over years, and it does argue against the strongest version of the immunosuppression worry over weeks. There is no comparable trial in transplant recipients, in patients on biologics, or in people with neutropenia, which is exactly where clinicians most want one.

Separate the questions. Whether a cannabinoid kills bacteria in a dish is a chemistry question with its own literature, and cannabigerol activity against methicillin-resistant Staphylococcus aureus is covered separately in our review of cannabis and antibiotic resistance. Whether cannabis helps you recover from an infection is a clinical question with essentially no human evidence.

Whether cannabis makes you more likely to get infected is a third question. The animal data suggest that potent cannabinoid receptor agonism at high parenteral doses can impair cell-mediated defense against an intracellular bacterium. Extrapolating a 4 mg per kilogram intravenous injection in a mouse to a person taking a 5 mg edible is not a defensible step, and the one human randomized trial that looked for immunological harm over three weeks did not find it.

The practical clinical position: cannabis is not a treatment for infection, there is no good reason to expect it to help you clear one, patients on immunosuppressive therapy should have cannabis use documented and discussed because of drug interaction risk rather than because of receptor biology, and anyone with a fever should be evaluated rather than self-treated.

The receptor biology is well established. CB2 expression in human immune tissue was quantified by concordant methods in a single careful 1995 study and has been repeatedly confirmed. The finding that CB2-deficient mice show exaggerated inflammation is consistent across many models. Read that layer with reasonable confidence.

Everything above that layer weakens fast. The infection work is small-animal, uses parenteral dosing far above human exposure, and points in both directions depending on pathogen and species. The human layer consists of one 21-day randomized trial in 67 people with HIV. For the question most patients are asking, whether cannabis affects their ability to fight off an infection, the honest grade is insufficient evidence.

Dose translation is the central weakness in the animal infection literature. Intravenous THC at 4 mg per kilogram in a mouse is not a model of a person using cannabis. Allometric scaling, route differences and the use of a synthetic full agonist rather than plant THC all widen the gap between what those experiments show and what a patient does.

The immortalized cell line problem applies to the signaling work. BV-2 cells are a useful and imperfect proxy for primary microglia, and lipopolysaccharide stimulation is a clean laboratory stand-in for a messy clinical infection. These experiments establish that cannabinoids alter inflammatory signaling. They do not establish what that alteration does inside an infected person.

None of this evidence shows that cannabis or any cannabinoid treats a bacterial, fungal or viral infection in humans. No randomized trial has tested a cannabis product against an infectious disease endpoint. Laboratory antimicrobial activity, where it exists, has not translated into a demonstrated clinical effect.

The evidence also does not show that cannabis meaningfully raises infection risk in ordinary use. The one human randomized trial designed to look for immunological harm over three weeks did not find it. Absence of evidence in both directions is the accurate summary, and it should be stated as such rather than resolved toward whichever conclusion the reader prefers.

CB2 has been an attractive drug target for three decades precisely because it offered immune modulation without psychoactivity. That promise has not converted. Selective CB2 agonists have repeatedly shown activity in animal models of inflammation and pain and have repeatedly underperformed in human trials, which is a pattern the field should take more seriously than it does.

Part of the explanation may be the Kozela finding. If a meaningful share of the anti-inflammatory effect of plant cannabinoids runs through non-cannabinoid targets, then a compound engineered for clean CB2 selectivity has been optimized away from the mechanism that was producing the effect.

The question I hear most is some version of whether cannabis is going to knock down a patient’s immune system. What I tell them is that the receptor data everyone is citing were generated in mice given injected doses no person takes, and that the one controlled human study, in people with HIV over three weeks, found nothing alarming. That is a genuinely reassuring answer and it is a narrow one.

Where I do get concerned is the patient who reaches for cannabidiol instead of calling about a fever, or the transplant patient who starts a high-dose product without telling anyone. The risk there is not immunological. It is drug interaction and delayed diagnosis, and both are far more likely to hurt someone than anything happening at a CB2 receptor.

CB2 receptors are densely expressed on immune cells and modulate inflammatory responses, but they do not recognize pathogens and they are not the pathway by which anyone should expect cannabis to treat an infection. Animal infection data conflict. The single human randomized trial found no adverse immunological effect over 21 days. Cannabis is not a treatment for infection, and a patient with signs of one needs evaluation, not a cannabinoid.

Hold three layers apart. Receptor expression on immune cells is established. Modulation of inflammatory signaling is established in cell and animal systems, frequently through mechanisms that bypass the cannabinoid receptors. Effect on human infection outcomes is unstudied. Any source that moves from the first layer to the third without pausing at the second is selling something.