Cholesin

You wouldn’t think that there was anything more to be said about cholesterol metabolism after decades of work by med school classmate Mike Brown and a host other researchers.  But there is.

The body can synthesize cholesterol starting from scratch and Mike found out how this is inhibited when cholesterol levels get too high.  Here is a brief summary of how this happens from a recent paper [ Cell vol. 187 pp. 1685 – 1700 ’24 ]

“Cholesterol biosynthesis and uptake are tightly regu-lated through a negative feedback mechanism that senses the cellular cholesterol levels. When cells are deficient in cholesterol, SREBP2, along with its escort protein SREBP cleavage-acti- vating protein (SCAP), is transported in coat protein complex II (COPII) vesicles from the endoplasmic reticulum (ER) to the Golgi apparatus. In the Golgi, SREBP2 is sequentially cleaved by site-1 and site-2 proteases. The N-terminal domain of SREBP2, released by this cleavage, travels to the nucleus, where it func- tions as a transcription factor to enhance the expression of genes involved in cholesterol synthesis and uptake. Conversely, when cellular cholesterol levels rise, cholesterol molecules bind to SCAP, triggering its interaction with insulin-induced gene (INSIG). This interaction retains SREBP in the ER and prevents the subsequent activation of SREBP and the expression of genes involved in cholesterol metabolism”.

 

Well now you can see why this took decades to figure out.

However a recently discovered protein cholesin cuts off cholesterol synthesis when you eat and absorb cholesterol, which is much more proactive as it doesn’t wait for cholesterol levels to increase.   Cholesin is secreted into the blood by the gut when cholesterol is absorbed (secretion into the blood is what makes it a hormone).   Human cholesin contains 195 amino acids and works its magic by binding to a G Protein Coupled Receptor (GPCR) called GPCR146 which shuts off signaling by protein kinase A (PKA). This prevents  SREPB2 from turn on cholesterol synthesis (primarily in the liver).

So obviously GPCR146 and cholesin do a biochemical dance together.  Amazingly, dance is more than a metaphor, and the two proteins are coded (and entwined) on opposite strands of the same genetic locus of chromosome #7 with the code for GPCR146 on one strand inside the code for cholesin on the other.

I find this both bizarre and fantastic.  The discoveries of molecular biology never cease to amaze (me at least, and you too if your molecular biological soul isn’t completely dead).

Another neuropharmacologic surprise.

Our genome contains 826 different genes for G Protein Coupled Receptors (GPCRs) which are targeted by at least 475 FDA approved drugs (Nature vol. 587 p. 553 ’20 ). Yet part of the fascination of reading the current literature is the surprises it brings.

Our basic understanding was that the GPCRs sit on the surface of the cell waiting for ligands outside the cell to bind to it, which produces a conformational change on the cytoplasmic side of the cell membrane, changing the way the GPCR binds to the G protein, triggering all sorts of effects inside the cell.

As far as I recall, we never thought that different GPCRs would bind to each other in the cell membrane, even though a single cell can express ‘up to’ 100 different GPCRs [ Mol. Pharm. vol. 88 pp. 181 – 187 ’15 ].  Neurons express GPCRs and some are thought to be involved in neuropathic pain

But that’s exactly what Proc. Natl. Acad. Sci. vol. 119 e2123511119  ’22  is saying.

First a few definitions, if you’re as rusty about them as I was

A cytokine is an extracellular protein or peptide  helping cells to communicate with each other.  A chemokine is an extracellular protein which attracts cells.

Our genome has over 50 chemokines.  Most are  proteins with about 70 amino acids. The are classified by where the cysteines lie in them.  We have 23 receptors for chemokines, 18 of which are GPCRs.   Binding is promiscuous — a given chemokine binds to multiple receptors, and a given receptor binds to multiple chemokines.

Clearly the chemokines and their receptors are intimately involved in inflammation which always involves cell migration.  Neurons express chemokine receptors GPCRs and some are thought to be involved in neuropathic pain.

We also know that the nervous system is involved in immune function, particularly inflammation.  One prominent neurotransmitter is norepinephrine, and a variety of receptors bind to it.  There are 3 alpha1 norepinephrine receptors (a, b and d), all of which are GPCRs.

What is so shocking is that alpha1 GPCRs bind to chemokine receptors (forming heteromers), and that this binding is required for chemokines to have any effect on cell migration.  Even more interesting is that binding of norepinephrine to the alpha1 component of the heteromer INHIBITs cell migration.

So how many of our 826 GPCRs bind to each other, and what effects do they have?

Reading the literature is like opening presents, you find new fascinating toys to play with, some of which may actually benefit humanity

 

Solid evidence for acupuncture at last

The early hype about acupuncture was so extreme (bathwater) that I stopped looking for the medical baby within.  Part of the hype was a reaction against all things western.

However when stimulation of a mouse at the knee point (ST36) decreases mortality due to exposure to lipopolysaccharide by 40%, it’s time to sit up and take notice [ Nature vol. 598 pp. 573 – 574, 641 – 645 ’21 ].

Not only that but the authors found the neurons responsible for the effect.  These neurons in the dorsal root ganglion express the G Protein Coupled Receptor (Prokr2) which is a  receptor for prokineticin, a secreted protein which increases gut motility.

Stimulation of these neurons (or the point behind the knee they innervate) produces anti-inflammatory effects.  Destruction of these neurons (by expressing diphtheria toxin in them) prevents low intensity stimulation of ST36 from dampening inflammation.

The paper even gives a possible explanation for some of the irreproducible results in the field.  High intensity of stimulation of ST36 activates the sympathetic system, while low intensity stimulation activates the parasympathetic nervous system.  The latter activates the vagus nerve which stimulates the adrenal medulla to produce catecholamines (which are anti-inflammatory).  So high intensity stimulation of the same site produces no useful therapeutic effect.

I never thought I’d see high quality work like this on acupuncture, but there it is.  More is sure to follow.

Moonlighting molecules

Just when you thought you knew what your protein did, it goes off and does something completely different (and unexpected). This is called moonlighting, and is yet another reason drug discovery is hard. You can never be sure that your target is doing only what you think it’s doing.

Today’s example is PACAP, a neuromodulator/neurotransmitter made by neurons. Who knew that PACAP can and does act as an antibiotic when the brain is infected. [ Proc. Natl. Acad. Sci. vol. 118 e1917623117 ’21 ] does (PNAS no longer pages its journals, as last year’s total was over 33,000 !).   PACAP is a member of the vasoactive intestinal polypeptide, secretin, glucagon family of neuropeptides (mammals have over 100 neuropeptides according to the paper).

PACAP stands for Pituitary Adenylate Cyclase Activating Polypeptide. It comes in two forms containing 27 or 38 amino acids, both cleaved from a 176 amino acid precursor. There are 3 receptors for PACAP, all G Protein Coupled Receptors (GPCRs). A zillion functions have been ascribed to it, setting the circadian clock, protecting granule cells of the cerebellum. Outside the nervous system it is produced by immune cells in response to inflammatory conditions and antigenic stimulation. It is one of the most conserved neuropeptides throughout the course of evolution. Now we probably know why.

Showing how hard protein chemistry really is, PACAP is structurally similar to cathelicidin LL-37 an antimicrobial peptide, despite having less than 5% amino acid sequences in common. PACAP is cationic. Different sides of the protein have different characteristics, with one side being highly positively charged, and the other being hydrophobic (e.g. the protein is amphipathic). This is typical of antimicrobial peptides, and perturbation of microbial membranes by inducing negative Gaussian curvature probably explains its antibacterial activity.

In mouse models of Staph Aureus or Candida infections, PACAP is induced ‘up to’ 50 fold in the brain (or spleen or kidney) where it kills the bugs. Yet another reason drug discovery is so hard. We are mucking about in a system we barely understand.

There are many other examples of moonlighting proteins. Probably the best known is cytochrome c which is is a heme protein localized in the compartment between the inner and outer mitochondrial membranes where it functions to transfer electrons between complex III and complex IV of the respiratory chain. Oxidation and reduction of the iron atom in the heme along with movement along the mitochondrial intermembrane space allows it to schlep electrons between complexes of the respiratory chain.

All well and good. But cytochrome c also can tell a cell to commit suicide (apoptosis) when mitochondria are sufficiently damaged that cytochrome c can escape the intermembrane space. Who’d a thunk it?

How many more players are there in the cell (whose function we think we know) that are sneaking around — doing more things in heaven and Earth, Horatio, than are dreamt of in your philosophy?

Do orphan G Protein Coupled Receptors self stimulate?

Self-stimulation is frowned on in the Bible — Genesis 38:8-10, but one important G Protein Coupled Receptor (GPCR) may actually do it.  At least 1/3 of the drugs in clinical use manipulate GPCRs, and we have lots of them (at least 826/20,000 protein coding genes according to PNAS 115 p. 12733 ’18).  However only 360 or so are not involved in smell, and in one third of them  we have no idea what the natural ligand for them actually is (Cell vol. 177 p. 1933 ’19).  These are the orphan GPCRs, and they make a juicy target for drug discovery (if only  we knew what they did)

One orphan GPCR goes by the name of GPR52. It is found on neurons carrying the D2 dopamine receptor.  GPR52 binds to G(s) family of G proteins stimulating the production of CAMP (which would antagonize dopamine signaling), enough to stimulate (if not self-stimulate) any neuropharmacologist.

Which brings us to the peculiar behavior of GPR52 as shown by Nature vol. 579 pp. 142 – 147 ’20.  The second extracellular loop (ECL2) folds into what would normally be the binding site for an exogenous ligand (the orthosteric site).  Well, it could be protecting the site from inappropriate ligands.  But it isn’t, as removing or mutating ECL2 decreases the activity of GPR52 (e.g. less CAMP is produced).  Pharmacologists have produced a synthetic GPR52 agonist (called c17).  However it binds to a side pocket, in the 7 transmembrane region of the GCPR.   This is interesting in itself, as no such site is known in any of the other GPCRs studied.

Most GPCRs have some basal (constitutive) activity where they spontaneously couple to their G proteins, but the constitutive activity of GPR52 is quite high, so c17 only slightly increases the rise in CAMP that GPR52 normally produces.

This might be an explanation for other orphan GPCRs — like a hermaphrodite they could be self-fertilizing.

How little we really understand about proteins

How little we really understand about proteins.  We ‘know’ that the 7 transmembrane alpha helices of G Protein Coupled Receptors (GPCRs) all contain hydrophobic amino acids, so they dissolve in the (hydrophobic) lipids of the membrane.  GPCRs have been intensively by chemists, molecular biologists, pharmacologists and drug chemists with the net result that as of last year “128 GPCRs are targets for drugs listed in the Food and Drug Administration Orange Book. We estimate that ∼700 approved drugs target GPCRs, implying that approximately 35% of approved drugs target GPCRs.” https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5820538/

So if you changed the hydrophobic amino acids found in the 7 transmembrane segments of GPCRs to hydrophilic ones — all hell should break loose.

Wrong says Proc. Natl. Acad. Sci. vol. 116 pp. 25668 – 25667 ’19 ].  The trick was to replace hydrophobic amino acids with hydrophilic ones with the same shape.

Thus leucine (L — single amino acid letter code) is replaced by glutamine (Q), Isoleucine (I) and Valine (V) is replaced by Threonine (T) and finally phenylalanine (F) is replaced by Tyrosine (Y).  They call this the QTY code.

Instead of destroying the structure of the GPCRs (CCR5 and CXCR4) they became water soluble, and bound their ligands CCL5 for CCR5  and CXCL12 for CXCR4 to the same extent.

Even more amazing, the QTYdesigned receptors exhibit remarkable thermostability in the presence of arginine and retained ligand-binding activity after heat treatment at 60 °C for 4 h and 24 h, and at 100 °C for 10 min.

I would never have expected this.  Would you?

Why did they even do it?  Because GPCR structures are hard to study. You either have to remove them en bloc from the membrane or dissolve them in other lipids so they don’t denature.  Why these two GPCR’s?    Because their ligands are proteins and can’t snuggle deep down inside the 7 alpha helices embedded in the membrane (they’re just too big), but bind to the outside surface.  CCL5 is an 8 kiloDalton protein (probably 80 amino acids, while CXCL12 has 93.  So just solublizing the GPCR without changing any of the amino acids external to the membrane, produces an object for study.

It would be amusing to do the same thing for a GPCR binding one of the monamines.  I doubt that they would bind, but I never would have believed this possible in the first place.

The neuropharmacological brilliance of the meningococcus

The meningococcus can kill you within 12 hours after the spots appear — https://en.wikipedia.org/wiki/Waterhouse–Friderichsen_syndrome.  Who would have thought that it would be teaching us neuropharmacology.   But it is —  showing us how to make a new class of drugs, that no one has ever thought of.

One of the most important ways that the outside of a cell tells the inside what’s going on and what to do is the GPCR (acronym for G Protein Coupled Receptor).  Our 20,000 protein coding genome contains 826 of them. 108 G-protein-coupled receptors (GPCRs) are the targets of 475 Food and Drug Administration (FDA)-approved drugs (slightly over 1/3).   GPCRs are embedded in the outer membrane of the cell, with the protein going back and forth through the membrane 7 times (transmembrane segment 1 to 7 (TM1 – TM7). As the GPCR sits there usually the 7 TMs cluster together, and signaling molecules such as norepinephrine, dopamine, serotonin etc. etc. bind to the center of the cluster.   This is where the 475 drugs try to modify things.

Not so the meningococcus. It binds to the beta2 adrenergic receptor on the surface of brain endothelial cells lining cerebral blood vessels, turning on a signaling cascade which eventually promotes opening junctions of the brain endothelial cells with each other, so the bug can get into the brain.  All sorts of drugs are used to affect beta2 adrenergic receptors, in particular drugs for asthma which activate the receptor causing lung smooth muscle to relax.  All of them are small molecules which bind within the 7 TM cluster.

According to Nature Commun. vol 10 pp. 4752 –> ’19, the little hairs (pili) on the outside of the organism bind to sugars attached to the extracellular surface of the receptor, pulling on it activating the receptor.

This a completely new mechanism to alter GPCR function (which, after all,  is what our drugs are trying to do).  This means that we potentially have a whole new class of drugs, and 826 juicy targets to explore them with.

Here is one clinical experience I had with the meningococcus.  A middle aged man presented with headache, stiff neck and fever.  Normally spinal fluid is as clear as water.  This man’s was cloudy, a very bad sign as it usually means pus (lots of white blood cells).  I started the standard antibiotic (at the time)  for bacterial meningitis — because you don’t wait for the culture to come back which back then took two days.  The lab report showed no white cells, which I thought was screwy, so I went down to the lab to look for myself — there weren’t any.  The cloudiness was due to a huge number of meningococcal bacteria.  I though he was a goner, but amazingly he survived and went home. Unfortunately his immune system was quite abnormal, and the meningitis was the initial presentation of multiple myeloma.

Proline rides again !

Proline is a kinky amino acid.  Kinky in the sense that it is only one of the twenty with a fixed configuration of its alpha carbon because of the ring (which may be why there is more of it in organisms living at high temperature) and kinky in the sense that when present in alpha helices it produces a kink.  The previous post shows how it is used to schlep the body weight’s worth of ATP we make each day out of our mitochondria — https://luysii.wordpress.com/2019/01/30/3939/.

Well here it is in one of the marijuana receptors (CB1).  Binding of delta9 THC in the 7 transmembrane alpha helix bundles of the G Protein Coupled Receptor (GPCR) causes an alteration in the kink allowing transmembrane helix 6 (TM6) to move outward toward the cytoplasm, creating a cavity on the intracellular side, where the G protein trimer can bind.

You can read much more about this in an exquisite paper [ Cell vol. 176 pp. 448 – 458 `19 ] describing the CB1 receptor bound to a synthetic ligand 20 times more potent that delta-9 tetrahydrocannabinol (delta9 THC).  It is a cryoEM study which used 177,000 projections to come up with a 3 Angstrom resolution structure of CB1 bound to MBDB-FUBINACA in complex with its G protein trimer.  They had to use a single chain variable fragment (scFv6) along with a positive allosteric modulator (PAM) called ZCZ-011 to stabilize the complex.

MBDB-FUBINACA is a story in itself.  It is presently the fentanyl of synthetic cannabinoids, which “has been linked to thousands of hospitalizations and numerous fatalities”  [ New England Journal of Medicine vol. 376 pp. 235 – 242 ’17 ].  I’m surprised I’ve never heard of it — have you? But then I’ve been retired from clinical practice for some time. Perhaps the mainstream press, pushing marihuana legalization as it has been, kept it quiet, or more likely there have been no further episodes of mass intoxication from the AMB-FUBINACA (aka the zombie drug) since 2017.

I’ve never knowingly used marihuana.  Frankly it scares me — for why please see — https://luysii.wordpress.com/2014/05/13/why-marihuana-scares-me/.

There are 4 molecular switches buried in GPCRs [ Current Med. Chem. vol. 19 pp. 1090 – 1109 ’12 ]

1. The ionic lock switch between the D/E R Y sequence at the cytoplasmic end of TM3 and E286 at the cytoplasmic end of TM6 (single letter amino acid code used) –http://130.88.97.239/bioactivity/aacodefrm.html

2. TM3 – TM7 lock switch.  In rhodopsin it is between the protonated Schiff base of lysine and a glutamic acid and it broken on light activation,.=

3. Toggle switch linked with the n P x x Y motif in TM7 (x stands for any amino acid) — much more about this later in the post.

4. Transmission switch — produced by agonist binding, the outward movement of TM6 to to ligand binding creating a hole fo the G protein to bind to the receptor on the cytoplasmic side.

So why did I call the Cell paper exquisite?  Because of the molecular detail it provides about just how MDMB FUBINACA activates CB1.  Here’s the structure of AB-FUBINACA — https://en.wikipedia.org/wiki/AB-FUBINACA.   Both look like drugs designed by a committee.  They both have a para-iodophenyl group, an amide, and a fused indole ring with an extra nitrogen (imidazole ring — I never could keep heterocyclic nomenclature straight).    MDMB has a methyl ester (in place of the amide) and a tertiary butyl group (in place of the isoPropyl group).

I don’t have time to look up how Pfizer came up with it.  The FUBINACAs do not resemble delta9 THC at all — https://en.wikipedia.org/wiki/Tetrahydrocannabinol.

The pictures in the paper show how the hydrophobic aromatic side chains of FIVE phenylalanines and 2 tryptophans create a nice oily space for delta9 THC and MBDB-FUBINACA to bind.

F200 (phenylAlanine 200) and W356 are the toggle twin switch which stabilize the inactive conformation of CB1.  The rotation of F200 to interact with the imidazole of FUBINACA, allows W356 to rotate outward, changing the kink produced the the proline #358  in TM6 allowing the helix to straighten and rotate outward toward the cytoplasm, creating a cavity for the G protein to bind to.

Definitely a tour de force for the blind watchman.

Time for drug chemists to go to the Multiplex

30 – 40% of all the drugs currently in clinical use are thought to target G Protein Coupled Receptors (GPCRs). Just how many GPCRs inhabit our genome isn’t clear. The latest estimate is 850 which is 4.2% of the 20,077 annotated protein genes we have. That being the case, it behooves drug chemists to know everything about them and how they work.

A recent paper [ Cell vol. 166 pp. 907 – 919 ’16 ] shows that a lot of the old thinking about GPCRs is wrong. Binding of a ligand to a GCPR results in a conformational change in its 7 transmembrane segments, so that the parts inside the cell bind to a heterotrimer of proteins which bind (and hydrolyze) GTP — this is the G protein. So far so good. The trimer splits up into its 3 constituents, unimaginatively called alpha, beta and gamma, each of which can act as a messenger that a ligand from outside the cell has landed on a GPCR, binding to other proteins causing all sorts of effects (e.g. can act as a second messenger)

All good things must end, and termination of GPCR signaling was thought to involve phosphorylation of the intracellular segment of the GPCR, binding of another protein (betaArrestin), removal from the cell membrane (so it can no longer bind its extracellular ligand) and then either destruction or recycling back to the cell membrane. So the old paradigm was betaArrestin binding equals the end of signaling.

It was thought that betaArrestin and the G protein competed for binding to the same intracellular amino acids of the GPCR. Not so says this paper. For some GPCRs both can bind, and signaling can continue, even though the complex of GPCR, G protein and betaArrestin is now inside the cell in an endosome. The complex is called the Multiplex. The examples given are GPCRs for parathyroid hormone (PTH) and Thyroid Stimulating Hormone (TSH). Blurry pictures are given of the complex. GPCRs have been divided into several classes and GPCRs for TSH and PTH are class B GPCRs — which contain a long phosphorylatable tail in the cytoplasm. The G protein binds to these GPCRs by its core region, while betaArrestin binds to the tail. Signaling continues apace.

A probably original (and probably wrong) idea

This post is not for the pharmacologically or chemically faint of heart.  Drug chemists should have no problem with it. You’ll need to bring your own background to the party, older posts on this blog don’t provide it.

It involves a particularly well studied G protein coupled receptor (GPCR), the beta2 adrenergic receptor.  Our genome codes for over 800 GPCRs, and at least 30% of all the drugs we currently use bind to them.

So here are my notes on  the paper that set me to thinking.  The idea follows the **** at the end.

       [ Science vol. 335 pp. 1055 – 1056, 1106 – 1110 ’12 ]  Conformation changes on ligand binding follow a common theme in the 3 activated GPCRs (including the beta2 adrenergic receptor) that have so far been crystallized.  Agonist binding results in an outward displacement of transmembrane segments 5 and 6 (TM5 and TM6) which opens a pocket on the intracellular surface of the receptor accepting the G protein.  Binding of beta-arrestin to GPCRs requires phosphorylation of sites on the cytoplasmic surface on TM7 toward the carboxy terminus of the GPCRs.

        This work attached 19F probes to cysteines at the intracellular ends of TM6 and TM7 of the beta2 adrenergic receptor for NMR spectroscopy.  They used 2,2,2 trifluoroethanethol to label 3 native cysteine residues (#265, #327, #341).  With no bound ligand two conformations in the ligand free state are in equilibrium with each other.  Then different ligands were added.  Carvedilol induced a shift to the TM7 active state, but had little effect on the conformational equilibrium at TM6.  Carvedilol is an antagonist which blocks G protein signaling and induces receptor desensitization and internalization.  

      Older work shows that the weakest agonists loosen interactions (the ionic lock and toggle), that juxtaposeTM3 and TM6 in the inactive state.  More potent agonists perturb the Asn Pro X X Y sequence of TM7 which permits phosphorylation and beta-arrestin binding.  Substituents on the hydroxylamine tail of the ligand are important — large nonpolar groups are particularly effective.  

       Agonist induced receptor bias for G protein vs. beta-arrestin signaling is one of many instances of the plastic and multistate behavior characteristic of GPCRs. 

*****

So what’s the idea?  Drug chemists spend a lot of their time designing and making slightly different GPCR ligands to have slightly different effects.  On p. 1109 of the paper, the structures of 8 are given.  Probably hundreds exist.  The work here describes two different conformations of the beta2 GPCR, each of which triggers a different cellular response,  because each conformation binds a different set of proteins.  The drugs (particularly carvedilol which is in clinical use) cause the GPCR to favor one conformation over another.

Based on what we’ve been taught about neurotransmitters and neurotransmission, this makes no evolutionary sense.  The ‘natureal’ ligand for beta2 adrenergic GPCR is norepinephrine.  What selective advantage can there be to having the GPCR flopping between 2 (or more) conformations?  Well there might be if other neurotransmitters bound to it producing different conformations.

Back in the day there was a lot of talk about false neurotransmitters.  These were small molecules which somehow got into the synaptic vesicle and were released into the synaptic cleft when the vesicle was exocytosed, but which didn’t bind to the receptor.  One such false neurotransmitter was octopamine (norepinephrine without the meta-hydroxyl group).  It was thought to be formed in excess in liver failure, enter into synaptic vesicles and do nothing useful when released, causing hepatic encephalopathy.

A less fanciful example is available [ Neuron vol. 46 pp. 1 – 2, 65 – 74 ’05 ] Serotonin, a true neurotransmitter we all know and love, can enter dopamine neurons, when reuptake is blocked (by a tricyclic antidepressant, or by a SSRI), where it accumulates in synaptic vesicles.

Back in the day when I was in medical school, there was something called the Dale Feldberg law, which said that an identical chemical transmitter is liberated at all terminals of a single neuron.  We now know that this is wrong — neurons release two classes of chemical transmitters (1)  biogenic amines such as norepinephrine, dopamine and serotonin and (2) neuropeptides, and many release one from each class.  Whether or not a single neuron releases more than one biogenic amine isn’t known (despite the example above).

So perhaps the body is a better drug chemist than we think, and the same GPCR can bind different neurotransmitters with different effects, explaining the multiple conformations they adopt.  If true, this means that there are probably other transmitters out there that we don’t know about.  In the case of the beta2 adrenergic receptor, they are probably small molecules, as the binding site is deep within the transmembrane region of the GPCR.

So is this sort of thing why GPCRs have multiple conformations?