Opioid receptors (mu, delta, kappa, NOP)

Opium is ancient; the receptor is recent. A stereospecific binding site for opiate drugs was demonstrated in nervous tissue in 1973, using a radiolabelled opiate that active opiates displaced and their inactive mirror images did not ^[2]. Stereoselectivity is the tell: a membrane that distinguishes the left-handed from the right-handed version of one molecule is using a protein pocket, not a general effect on fat. Two years later the answer to why a brain carries such a site arrived as two pentapeptides pulled from pig brain, Met-enkephalin and Leu-enkephalin, which behaved like opiate agonists in bioassay ^[3].

The idea that there is more than one opioid receptor came from pharmacology, not cloning. Work in the chronic spinal dog in the mid 1970s found that morphine-like and nalorphine-like drugs produced distinguishable syndromes, and the types were named after the drug that produced each: mu after morphine, kappa after ketocyclazocine, sigma after SKF-10047 ^[4]. Two names survived. That sigma site is not an opioid receptor at all and is filed separately; see sigma receptors, and treat it as a reminder that a receptor named from a drug effect is a hypothesis rather than a protein.

The receptors are one piece of a larger circuit, the endogenous opioid system: the four receptors, the opioid peptides, and the enzymes (prohormone convertases and carboxypeptidase E) that cut those peptides out of their precursors ^[5]. The tidy version most people carry is that beta-endorphin leans mu, enkephalins delta, dynorphins kappa. It is a useful map, but these peptides are promiscuous; they bind all three classical receptors at different affinities, and the peptidases that trim them can turn one active fragment into another ^[5].

So why does one receptor class produce effects ranging from bliss to misery? Location and wiring rather than chemistry. Mu receptors sit densely in reward circuits like the ventral tegmental area, so activating them feels good; kappa receptors and their dynorphin ligand are engaged by stress and lower the rewarding value of experience, which reads as dysphoria ^[10]. Same family, same inhibitory G protein, opposite emotional sign.

All four are Gi/o-coupled GPCRs, so activating any of them tends to inhibit adenylyl cyclase, lower cAMP and make neurons less excitable ^[1]. What differs is where they sit and what that quieting does to behaviour.

Mu carries almost the whole clinical opioid package on its own, and there is an unusually clean experiment proving it. Mice engineered without the mu receptor gene lose morphine analgesia, lose the rewarding effect, and lose withdrawal signs after chronic dosing; the benefits and the harms disappear together ^[7]. That is why a plain mu agonist cannot be tuned into a safe drug by adjusting the dose. Mu is also the most complicated of the four at the gene level: OPRM1 is heavily alternatively spliced, and the variants differ in signalling and in which agonists they answer to ^[8].

Delta is the mood-leaning member. Delta agonists produce antidepressant-like and anxiolytic-like effects in animals along with modest analgesia; the catch is that several are convulsant, and whether the seizure risk separates from the benefit is still open ^[9]. Delta receptors are also more likely than mu to sit inside the cell until stimulation brings them to the surface, so their pharmacology is state dependent.

Kappa is the cautionary mirror image of mu. Kappa agonists are genuine analgesics and can suppress itch, but strong central kappa activation is reliably dysphoric and can be psychotomimetic. Its endogenous ligand, dynorphin, is released by stress, and the dynorphin/kappa arm lowers the value of reward and drives aversion; that is why the same receptor is now pursued from both directions, with agonists as non-addictive analgesics and antagonists as candidate treatments for anhedonia ^[10].

NOP is the newest member and the odd one out. It was found by homology to the other three and is structurally an opioid receptor, but it binds classic opioid ligands poorly and naloxone does not block it. Its own peptide, nociceptin, can oppose or assist mu depending on the circuit: given in the brain it reduces opioid reward and can produce hyperalgesia, while given in the spinal cord it is analgesic ^[6]. That bidirectionality is why NOP has taken far longer to turn into drugs.

Saying a receptor is Gi/o-coupled is shorthand for two effects that arrive together. The alpha subunit inhibits adenylyl cyclase, lowering cyclic AMP and everything downstream of protein kinase A. The beta-gamma dimer does the faster work: it opens inwardly rectifying potassium channels and closes voltage-gated calcium channels of the N and P/Q types ^[1]. Where the receptor sits decides which effect matters. On a presynaptic terminal the calcium arm dominates and less transmitter is released; on a postsynaptic cell body the potassium arm dominates and the cell becomes harder to excite. One receptor, one G protein, two different outcomes.

This also answers a question that puzzles most people the first time they meet it: how can a purely inhibitory receptor produce euphoria? Through disinhibition. In the ventral tegmental area, mu receptors sit densely on the GABA interneurons that hold dopamine neurons in check. Quieting the inhibitor releases the brake, dopamine neurons fire faster, and dopamine rises in the nucleus accumbens. Nothing was switched on; something that was switching things off was switched off. See the dopaminergic system.

The same logic makes overdose lethal. Mu receptors are expressed on brainstem neurons that generate respiratory rhythm and on circuits that sense rising carbon dioxide, so quieting them slows breathing and blunts the reflex that would force harder breathing ^[1]. That is a location fact rather than a dose fact, which is why tolerance to the pleasant effects does not protect against it. These are real objects rather than cartoons: the mu receptor was crystallised in 2012 bound to a morphinan antagonist, showing a wide, solvent-exposed pocket that explains why chemically unrelated scaffolds all fit ^[11].

Finally, the signal has to stop. Agonist binding recruits G protein-coupled receptor kinases that phosphorylate the receptor's tail; phosphorylation recruits beta-arrestin, which uncouples the receptor from its G protein and can pull it into the cell to be recycled or degraded. That machinery is the physical basis of both signal termination and tolerance ^[12].

A GPCR is not a simple switch. When an agonist binds mu, the receptor can hand its signal to two partners: G proteins and beta-arrestin. The influential hypothesis came from knockout mice, which lacked beta-arrestin-2 and showed enhanced, prolonged morphine analgesia, with follow-up work reporting less constipation and less respiratory depression in the same animals ^[14]. That suggested G-protein signalling carries the good part and arrestin much of the bad part, so a drug biased toward G protein should relieve pain with a wider safety margin.

The hypothesis produced real drugs. Oliceridine reached the market as the first mu agonist sold on a biased-ligand rationale for acute pain, with trial data showing analgesia at doses producing less respiratory and gastrointestinal effect than morphine ^[18]. It also drove interest in Mitragynine, a partial mu agonist that recruits little arrestin and antagonises kappa and delta; that is an unusual profile for a natural product, and its 7-hydroxy metabolite is a far more potent mu agonist than the parent ^[19].

The honest update is that the story is contested, and the contest is instructive. A careful comparison of the new agonists concluded that their cleaner profiles track low intrinsic efficacy, meaning they are weak partial agonists, rather than any true separation of pathways; matched for efficacy, the bias largely disappears ^[16]. Independent work then found that morphine still depresses respiration in mice lacking beta-arrestin-2 altogether, which the original hypothesis does not predict ^[17].

The sharpest result is a knock-in mouse whose mu receptor cannot be phosphorylated and therefore cannot recruit arrestin properly. If the hypothesis were right, that animal should be the ideal case. It showed better analgesia and less tolerance, as predicted, but worse respiratory depression and constipation, the opposite of the prediction ^[15]. When an experiment designed to confirm an idea reverses half of it, the idea needs rebuilding.

None of this makes biased agonism worthless. A drug with a ceiling on its maximum effect is genuinely safer than one without. What changed is the explanation: the safety margin looks like a property of how strongly a ligand activates the receptor rather than of which pathway it chooses.

Once the four subtypes are in place, almost any opioid can be described with three questions, asked in order. Which receptors does it touch? How much efficacy does it have at each? Where in the body can it get? Answer those and most of a drug's clinical personality follows without memorising anything.

Efficacy is the axis people skip, and it decides the most. A full agonist produces the maximum effect the receptor can deliver, so its dose response has no ceiling until the person stops breathing. A partial agonist cannot reach that maximum however much you give, which caps both the benefit and the harm, and which also means it can displace a full agonist and precipitate withdrawal if it binds tightly enough. An antagonist produces nothing itself and simply occupies the site.

Access is the axis people forget. Two drugs with identical receptor profiles behave completely differently if one crosses the blood brain barrier and the other does not. The antidiarrhoeal loperamide is a good mu agonist pumped straight back out of the brain by P-glycoprotein, so at ordinary doses it acts only on the gut; at very high doses it stops being peripheral and becomes dangerous. See pharmacokinetics for why transport and metabolism decide as much as binding.

One caution about names. Not everything filed near this family is an opioid. Dextromethorphan is a morphinan by chemistry and an NMDA antagonist plus sigma ligand by pharmacology, which is why it belongs with the dissociatives. The skeleton says where a molecule came from; only the receptor profile says what it does.

Tolerance to opioids is not one process, and treating it as one is where most confusion starts. At the cellular level, the oldest and clearest observation is that chronic exposure makes adenylyl cyclase rebound: cells exposed to morphine long enough upregulate the very pathway the drug inhibits, so cAMP looks normal while the drug is present and overshoots when it is removed ^[13]. That overshoot is much of what withdrawal physically is.

At the receptor level, sustained agonist exposure drives phosphorylation, arrestin recruitment, uncoupling and internalisation, and whether the internalised receptor is recycled or degraded depends on the agonist ^[12]. Different agonists drive those steps to very different extents, so tolerance to one opioid carries over only partly to another. That incomplete cross-tolerance is the basis of opioid rotation, and why a dose that felt routine on one drug can be an overdose on the next.

The part that matters most for safety is that tolerance does not develop to every effect at the same rate. Analgesia and euphoria fade relatively quickly. Constipation and pupil constriction barely fade at all. Respiratory depression develops tolerance more slowly and less completely than analgesia ^[23]. So a person escalating to keep the pain relief walks toward the respiratory ceiling rather than away from it. Tolerance is also lost far faster than it is gained, which is why overdose deaths spike immediately after release from custody or discharge from detox.

Opioid-induced hyperalgesia is the strangest member of this family and is frequently mistaken for tolerance. It is not that the drug works less well; the person has become more sensitive to pain overall, sometimes in areas unrelated to the original problem. The two look identical from outside and demand opposite responses, since more drug helps tolerance and worsens hyperalgesia ^[23].

Finally, dependence and addiction are separate things, and mixing them up causes real harm. Physical dependence is an expected adaptation that appears in anyone dosed long enough, including people who never experience craving. Addiction is compulsive use despite harm. See tolerance and dependence.

The whole promise and peril of opioids lives in this subtype map. Mu delivers the analgesia everyone wants, and the same receptor in reward and brainstem circuits delivers euphoria, dependence and lethal respiratory depression; the knockout evidence says those are one package rather than three ^[7]. Hence the constant attempts to hit mu differently, or to recruit other subtypes to help.

Kappa is where the map earns its keep in the other direction. The plant diterpene salvinorin A from Salvia divinorum is a potent, selective kappa agonist and a vivid demonstration of what strong kappa activation feels like; repeated exposure shifts reward sensitivity in animals rather than simply blunting pain ^[21]. Because the dynorphin/kappa arm is engaged by stress and lowers the value of reward, the opposite approach is now being tested: Aticaprant and related kappa antagonists have been studied for anhedonia, and a randomised proof-of-mechanism trial found that kappa antagonism moved the intended brain readout, ventral striatal activation during reward anticipation. That justifies further trials rather than proving a treatment ^[22].

The atypical agents are the best argument for learning this map. Tianeptine looks like a tricyclic antidepressant and was sold as one for decades, but it is a full mu agonist with weak delta activity and none at kappa, which explains its mood effects and its dependence liability at the doses people escalate to ^[20]; that several antidepressants touch the opioid system is now an active research thread ^[24]. Mitragynine from Kratom is a partial, arrestin-sparing mu agonist that also blocks kappa and delta, which helps explain why kratom feels stimulating at low doses and opioid-like at high ones ^[19].

For the broader signalling picture see neurotransmitters 101; for what happens when a receptor system adapts to constant stimulation see tolerance and dependence.

Whether signalling bias exists in a way that helps patients. Low intrinsic efficacy explains the current biased agonists at least as well as pathway selectivity does, and the phosphorylation-deficient mouse reversed the prediction on half its measures ^[15][16]. A genuinely biased ligand may be achievable; no marketed drug is established as one.

Whether delta agonists can be antidepressants without being convulsants. The mood effects are reproducible in animals and the seizure risk is agonist dependent rather than universal, so separation looks possible, but no delta agonist has completed a convincing controlled trial in people ^[9].

What NOP does in humans. The receptor opposes mu reward and modulates pain in both directions depending on where it is engaged, and mixed mu and NOP ligands are in development, but the human picture is thin next to the rodent one ^[6].

What the mu splice variants do. OPRM1 produces a large family of variants whose signalling differs, and that is the leading candidate explanation for incomplete cross-tolerance and for why individuals respond so differently to different opioids; the link from a variant to a clinical response has not been made ^[8]. The same applies one level up, where more than twenty peptides come from four precursors with heavily overlapping preferences and nobody knows what the redundancy is for ^[5].

How much of everyday experience runs on this system. Endogenous opioids are implicated in placebo analgesia, exercise-related mood change and social attachment, and naloxone-reversal studies support a role in each; the size of that role is not quantified. Saying so plainly is more useful than the confident version written elsewhere.