Category Archives: Chemistry (relatively pure)

The world’s longest allosteric effect

I think there is some very interesting protein physical chemistry to be discovered/worked out based on a recent report [ Nature vol. 537 pp. 107 – 111 ’16 ]. It involves a long (2,200 Angstrom) coiled coil protein called EEA1 (Early Endosome Antigen 1). It contains 1,400 amino acids 1,275 of which form a coiled coil.

If you are conversant with the alpha helix and how two of them form a coiled coil, jump to ****. Otherwise here is some background and links to pictures which should help.

The alpha helix is a type of protein secondary structure in which the protein backbone assumes the shape of a coiled spring. There are 3.64 amino acids per turn. A single turn is 5.4 Angstroms high and 11 Angstroms wide. The alpha helix is right handed. That is to say, that if you orient the chain so that your thumb points from the N terminal to C terminal amino acid, the chain will twist in the direction of the fingers of the right hand as it rises. For some reason I can’t provide a link to a very large number of images for you hit. However, when I go to Google and type images of alpha helices you see them immediately — you’ll have to do the same to get there.

Coiled coils have two alpha helices winding around each other. This means that for secure interactions, the same types of amino acids must repeat again and again. A 7 residue periodicity (abcdefg)n in the distribution of nonpolar and charged amino acid residues is a feature characteristic of proteins which form alpha helices coiled about each other (coiled coil molecules). The 7 amino acids are lettered a – g from amino to carboxy. Positions a and d are usually hydrophobic amino acids (Leu, Ile, Val, Ala), positions e and g are usually polar or charged. The nonpolar a and d side chains associate by means of complementary knobs into holes packing. Each individual alpha helix is right handed, but the two helices wind around each other with a left handed turn. There are 3.64 amino acids per turn of an alpha helix, so for a regular repeating structure an amino acid should appear at the same position in space on the alpha helix (which forms a rigid rod). To see all the pictures you want — go to Google and type “Images of the Alpha Helix”.

To get the number of amino acids down so there are 3.5 per/turn (so the structure can repeat exactly every 7 amino acids –e.g. after 2 alpha helical turns) left handed supercoiling of each helix occurs (it’s a chicken and the egg situation). The helices are at an angle of 18 degrees to each other, and every 3.5 amino acids still form a 5.4 Angstrom (when one helix is viewed in isolation), but due to the tilt, they take up 5.1 Angstroms. This means that the same type of amino acid is found at positions 1, 8, 15, 22 etc. All intermediate filament proteins (keratin, neurofilaments, vimentin, etc.) contain a coiled coil structure. So to see all the pictures you could want — go to Google and type “Images of coiled coil proteins”

So the 1,275 amino acids of EEA1 divided by 3.5 and multiplied by 5.1 give you a coiled coil of fairly enormous length for a protein (1,858 Angstroms) — average protein diameter (if there is such a thing) is under 50 Angstroms

Functionally, EEA1 seems to be used as a tether with one end free and the other end hooked to a target membrane which wants to ‘catch’ the early endosome. The target membrane isn’t specified in the paper. Apparently EEA1 when not binding the endosome, is in a fully extended state, at around 2,000 Angstroms.

A protein called Rab5 is found on the early endosome membrane, and when EEA1 contacts it, the long coiled coil helix collapses, dragging the endosome toward the target membrane.  This is entropy in action, there being far more configurations of a collapsed protein than a rigidly extended one. To feel entropy for yourself, just pull on a rubber band, entropic effects just like this one are what you feel pulling back.

The collapse of EEA1  is an allosteric effect and a very long one, although the authors note long range allosteric effects are “not uncommon among coiled coil proteins”.

EEA1 is more complicated than initialy described. It contains amino acids which disrupt the 7 amino acid periodicity of the coiled coil (making it a jointed structure). The authors then made an EEA1 protein without the joints (so it was a perfect very long coiled coil). Binding of this protein to Rab5 on an endosome doesn’t result in collapse. So clearly normal EEA1 collapses at the ‘joints’.

The authors talk about some hypotheses as to how this happens in the Supplementary material (but I was unable to find).

So here’s a good research proejct for an enterprising grad student: either find out why and how a protein with multiple joints should exist in a fully extended configuration, or figure out how binding of Rab5 at one end of EEA1 produces such profound allosteric changes through this long linear protein. Happy hunting and thinking.

I must say it’s a pleasure to get back to chemistry after writing about the neurologic and medical issues of the presidential candidates.

Addendum 29 September — I wrote one the following to one of the authors (Dr. Grill) sending him the post above

Dr. Grill

Greatly enjoyed the paper.  I could never find the discussion of possible mechanism in the supplementary material.  You might enjoy the following post written about the paper

He replied as follows:

“Dear Luysii thank you very much for the kind words, and I really like your title!

With the supplementary discussion, besides the method part there is an additional supplement file on the Nature website that is easy to miss…I attach it here for you. We discuss this a bit more, but I must admit that this is not very satisfactory at the moment. We just don’t know how this works, and much of our efforts at the moment are dedicated to understand”
So for other readers of the original paper who also can’t find the supplement with the authors’ speculations as to what is going on– here  is what he sent.

” A key question is how Rab5 can induce such a long-range global molecular transition in flexibility of EEA1. Indeed, long-range allosteric effects have been observed for other coiled-coil proteins. In the case of myosin, the presence of discontinuities in the coiled-coil heptads drive structural changes to flexibility. Other tethering factors may bend through large breaks in coiled-coil structure acting as joints, although it remains to be shown whether and how conformational changes are triggered by Rab binding, as shown for EEA1.

Furthermore, a dynamically flexible coiled-coil is mostly extended, provided its ends are free60. However, when the ends of this coiled coil are tethered, bent, or when torsion is locally applied, compensatory structural changes are propagated and even amplified through the length of the structure. Our results suggest that a change in intrinsic static curvature may contribute but is not the major cause for the reduction in end-to-end distance. However, a more rigorous assessment would require visualizing the thermal fluctuations of the bound and unbound EEA1 very rapidly and in three dimensions.

Force generation due to entropic effects plays a key role in many processes in biology ranging from DNA cytoskeletal filaments to motor proteins. Switching a molecule from stiff to flexible could be an effective and general mechanism of many coiled-coil proteins for generating an attractive force, thereby pulling two objects together or allowing reactions otherwise hindered by polymer rigidity. Future experiments will test to what extent the entropic collapse is a general mechanism used not only by membrane tethers but also in other biological processes.”


Baudelaire comes to Chemistry

Could an evil molecule be beautiful? In Les Fleurs du Mal, a collection of poems, Baudelaire argued that there was a certain beauty in evil. Well, if there ever was an evil molecule, it’s the Abeta42 peptide, the main component of the senile plaque of Alzheimer’s disease, a molecule whose effects I spent my entire professional career as a neurologist ineffectually fighting. And yet, in a recent paper on the way it forms the fibrils constituting the plaque I found the structure compellingly beautiful.

The papers are Proc. Natl. Acad. Sci. vol. 113 pp. 9398 – 9400, E4976 – E4984 ’16. People have been working on the structure of the amyloid fibril of Alzheimer’s for decades, consistently stymied by its insolubility. The authors solved it not by Xray crystallography, not by cryoEM, but by solid state NMR. They basically looked at the distance constraints between pairs of isotopically labeled atoms, and built their model that way. Actually they built a bouquet of models using computer aided energy minimization of the peptide backbone. Another independent study produced nearly the same set.

The root mean square deviation of backbone atoms of the 10 lowest energy models of the bouquets in the two studies was small (.89 and .71 Angstroms). Even better the model bouquets of the two papers resemble each other.

There are two chains of Abeta42, EACH shaped like a double horseshoe (similar to the letter S). The two S’s meet around a twofold axis. The interface between the two S’s is form by two noncontiguous areas on each monomer (#15 – #17) and (#34 – #37).

The hydrophilic amino terminal residues (#1 – #14) are poorly ordered, but amino acids #15 – #42 are arranged into 4 short beta strands (I only see 3 obvious ones) that stack up and down the fibril into parallel in register beta-sheets. Each stack of double horseshoes forms a thread and the two threads twist around each other to form a two stranded protofilament.

Glycines allow sharp turns at the corners of the horseshoes. Hydrogen bonds between amides link the two layers of the fibrils. Asparagine side chains form ladders of hydrogen bonds up and down the fibrils. Water isn’t present between the layers because the beta sheets are so close together (counterintuitively this decreases the entropy, because water molecules don’t have to align themselves just so to solvate the side chains).

Each of the horseshoes is stabilized by hydrophobic interactions among the hydrophobic side chains buried in the core. Charged residues are solvent exposed. The interface between the two horsehoes is a hydrophobic interface.

Many of the famlial mutations are on the outer edges of double S structure — they are K16N, A21G, D23N, E22A, E22K, E22G, E22Q.

The surface hydrophobic patch formed by V40 and A42 may explain the greater rate of secondary nucleation by Abeta42 vs. Abeta40.

The cryoEM structures we have of Abeta42 are different showing the phenomenon of amyloid polymorphism.

The PNAS paper used reombinant Abeta and prepared homogenous fibrils by repeated seeding of dissolved Abeta42 with preformed fibrils. The other study used chemically synthesized Abeta and got fibrils without seeding. Details of pH, peptide concentration, salt concentration differed, and yet the results are the same, making both structures more secure.

The new structure doesn’t immediately suggest the toxic mechanism of Abeta.

To indulge in a bit of teleology — the structure is so beautiful and so intricately designed, that the aBeta42 peptide has probably been evolutionarily optimized to perform an (as yet unknown) function in our bodies. Animals lacking Abeta42’s parent (the amyloid precursor protein) don’t form neuromuscular synapses correctly, but they are viable.

What is a hormone? What is an endocrine organ?

We all knew what hormones were back in the day. They were chemicals released by an endocrine gland into the blood where they went everywhere and affected distant organs. The classic example were the sex hormones (estrogen, progesterone, testosterone) eleased by the gonads affecting the reproductive organs, and not least the brain.

Things have changed mightily, and just about every tissue in the body does this now. There are at least 20 adipokines released by fat — examples are adiponectin, adipsin, and of course leptin. Muscle may be also getting into the act with irisin (although that is controversial). Other muscle produced hormones (myokines)  include atrial natriuretic peptide released by the heart and skeletal muscle releases at least 8 more.

There is even more stuff released into local tissue fluids which don’t get into the blood so they aren’t hormones. You can regard all neurotransmission this way. Paracrines are compounds which act only on cells close to them (because they don’t get into the blood). Examples include the huge class of prostaglandins and polypeptide growth factors such as the 22 member fibroblast growth factor family.

What to make of [ Cell vol. 166 pp. 424 – 435 ’16 ] which describes PM20D1 (Peptidase M20 Domain containing 1) which is secreted by fat cells. It’s an enzyme which builds compounds from substances already in the blood. The chemistry is simplicity itself — it takes a long chain fatty acid and an amino acid and forms the fatty acid amide — or an N-acyl amino acid.

What does the product do? It causes uncoupling of oxidative phosphorylation by mitochondria, so it just produces heat (something useful to an animal in the cold). Administration of N-acyl amino acids to mice increases energy expenditure and improves glucose metabolism. It’s possible that they could be used therapeutically.

Another example of how little we knew about what is going on inside us.

You are alive because the lipid bilayer of your plasma membrane is asymmetric

You are an organism with trillions of cells. A mosquito bit you depositing millions of viruses in your tissues. The virus can reproduce only within one of your cells and it has exploited all sorts of protein protein chemistry to get in. Antibodies (if you are fortunate enough to have them) can get rid of the extracellular critters. However, 500,000 have made into the same number of your cells, and are merrily trying to reproduce.

How does the asymmetry of the lipid bilayer of your plasma membrane help you survive. If each virus infected cell killed itself before the virus reproduced, you’d survive. Although 500,000 is a large number is is less than 1 millionth of your cell total.

Well you do have intracellular defenses against viruses, called the innate immune system. One of them is a protein with the ugly name of gasdermin D. The activated innate immune system (in the form of inflammatory caspases) cleaves gasdermin. This breaks up the inhibition of the amino terminal part of gasdermin by the carboxy terminal part giving a fragment which binds to one particular membrane component (phosphatidyl serine) which makes up 20% of the inner leaflet of the cell membrane. It then forms a large diameter (to a cell 140 Angstroms is quite large) pore in the cell membrane. No cell can survive this, so it dies, releasing cellular contents (probably some viral components but not fully formed one). For details see [ Nature vol. 535 pp 111 – 116, 153 – 158 ’16 ]

Wait a minute. The toxic gasdermin fragment is also released. So how come it doesn’t kill everything in sight? Because our cellular membranes keep phosphatidyl serine confined to the inner membrane, normal cells don’t show it on their exterior, so they can be bathed in gasdermin with no ill effect. What is responsible for this asymmetry — believe it or not an ATP consuming enzyme called flippase (about this more later) which takes any phosphatidyl serine it finds on the outer leaflet and schleps it back inside the cell.

There is all sorts of elegant chemistry which explains just how gasdermin binds to phosphatidyl serine and none of the many other phospholipids found on the inner leaflet. There is more elegant chemistry explaining how flippase works (see later).

What chemistry cannot explain, is why organisms would ‘want’ an asymmetric membrane. As soon as you get into the function of a particular compound in an organism, chemistry is powerless to tell you why. Nothing else can explain how a given molecule does what it does on the molecular level but that is not enough for a satisfying explanation.

One further explanation before some hard core cellular biochemistry follows (after ***). Our cells are dying all the time. The lining of your gut is replaced every 5 days. Even the longest lasting element of your blood is gone after half a year, and most other elements are turned over at least once a month. When these cells die, they must be cleaned up, without undue fuss (such as inflammation). The cleaners are cells called macrophages. A dying cell releases chemical signals, actually called ‘eat me’, one of which is phosphatidyl serine found on the membrane fragments of a dead cell. The fact that flippases keep it on the inner leaflet means that macrophages won’t attack a normal cell.

Slick isn’t it?


Flippase is a MgATPdependent aminophospholipid translocase. It localizes phosphatidylserine and phosphatidylethanolamine to the inner membrane leaflet by rapidly translocating them from the outer to the inner leaflet against an electrochemical gradient. The stoichiometry between amino phospholipid translocation and ATP hydrolysis is close to one (how will the cell have enough ATP to do anything else?). The flippase is inhibited by high calcium, and by pseudosubstrates such as vanadate, acetylphosphate and para-nitrophenyl phosphate, and by SH reactive reagents such as N-ethylmaleimide and pyridyldithioethylamine (PDA) a specific inhibitor of phospholipid translocation

[ Proc. Natl. Acad. Sci. vol. 109 pp. 1449 – 1454 ’12 ] P4-ATPases are a subfamily of P-type ATPases. They transport aminophospholipids from the exoplasmic to the cytoplasmic leaflet (and are known as flippases). Man has 14 P4-ATPases, expressed in various cell types. They are thought to be similar to the catalytic subunits of the Ca++ ATPase, and the Na, K ATPase, consisting of cytoplasmic, N, P and A domains and a membrane domain made of 10 transmembrane helices (M1 – M10).

[ Proc. Natl. Acad. Sci. vol. 111 pp. E1334 – E1343 ’14 ] The P4-ATPases are thought to resemble the classic P-type ATPase cation pumps — a transmembrane domain of 10 helices and 3 cytoplasmic domains (P for phosphorylation, N for nucleotide binding and A for actuator). ATP8A2 forms an intermediate phosphorylated on aspartic acid (E2P)and undergoes a catalytic cycle similar to the sodium pump (Na+, K+ ATPase). Dephosphorylation of E2P is activated by the transported substrates phosphatidyl serine (PS) and phosphatidyl ethanolamine (PE), similar to the K+ activation of dephosphorylation in the sodium pump.

PE and PS are 10x as large as the cations transported by the sodium pump. This is known as the giant substrate problem. This work shows that isoleucine #364 (mutated in — patients with the ataxia, retardation and dysequilibrium syndrome Eur. J. Hum. Genet. vol. 21 pp. 281 – 285 ’13 aka CAMRQ syndrome ) forms a hydrophobic gate separating the entry and exit sites of PS. I364 likely directs the sequential formation and annihilation of water filled cavities (as shown by molecular dynamics simulations) allowing transport of the hydrophilic phospholipid head group, in a groove outlined by TMs 1, 2, 4 and 6, with the hydrocarbon chains following passively, still in the membrane lipid phase (and presumably outside the channel) — this must disrupt the hell out of the protein as it passes. They call this the credit card model — only the interaction with part of the molecule is important — just as the magnetic stripe is the only important thing about the credit card.

Another fail safe mechanism used by the cell — readthrough

Nothing is perfect in this world, not even the translation of mRNA into protein. The error rate is one amino acid misincorporated into a protein for every 10,000 or so done correctly — but these results are for one celled organisms (E. Coli, yeast). I can’t find a number for mammals, primates etc. etc.

This means that occasionally one of the 3 codons which tell the ribosome to quit (stop codons), will be misread as an amino acid. This is called readthrough, and means that the ribosome will merrily march on producing a much larger protein than coded for by the mRNA until one of two things happens. l. the ribosome reaches the end of the mRNA and stops. 2. the mRNA contains another stop codon (there are 3). The probability of this is 3/64 per codon. If stop codons are randomly distributed (which they are most certainly not in the protein coding segment of an mRNA) the chances of 100 codons in a row not containing a stop codon is under 1% (.822 % to be exact). So any protein containing more than 100 amino acids is a statistical freak in this sense. Since the 3′ untranslated region (3’UTR) of mRNA doesn’t code for protein, they should have stop codons randomly distributed (there being no selective pressure to keep them away).

Enter Nature vol. 534 pp. 719 – 723 ’16 — if you attach a 3′ UTR section of an mRNA to a normal protein sequence (mimicking readthrough) you get much less protein. The authors think the 3’UTRs code for peptide sequences destabilizing the attached protein. They don’t know what this might be, so it’s terra incognita for researchers, and a worthwhile PhD project to figure it out. Another example of ‘coding’ by a presumably nonCoding sequence in the genome. It may also tell us something about protein structure.

You gotta love this structure

Science vol. 352 pp. 1555 – 1559 has a structure you have to love. It is a molecular knot containing a mere 30 pyridines and 10 benzenes all tied together in a knot which looks like a five pointed star. The tying was done by metathesis of benzenes with CH2 CH2 CH = CH2 dangling from them. To think of what needed to be tied to what was extremely clever.

Surprisingly with all this going on the knot coordinates just a single halogen atom. This shows why you must build a model of a complicated organic compound to see what it really looks like, something I learned with adamantane years ago — you can draw all the chairs you want, making it look rather spiky, but the damn thing is actually spherical. Well, no model was built, but the structure was determined using Xray crystallography (figure 3) Anyone playing with tinkertoys back in the day (or Legos now) and loving it will have a natural affinity for organic chemistry

Why you do and don’t need chemistry to understand why we have big brains

You need some serious molecular biological chops to understand why primates such as ourselves have large brains. For this you need organic chemistry. Or do you? Yes and no. Yes to understand how the players are built and how they interact. No because it can be explained without any chemistry at all. In fact, the mechanism is even clearer that way.

It’s an exercise in pure logic. David Hilbert, one of the major mathematicians at the dawn of the 20th century famously said about geometry — “One must be able to say at all times–instead of points, straight lines, and planes–tables, chairs, and beer mugs”. The relationships between the objects of geometry were far more crucial to him than the objects themselves. We’ll take the same tack here.

So instead of the nucleotides Uridine (U), Adenine (A), Guanine (G), Cytosine (C), we’re going to talk about lock and key and hook and eye.

We’re going to talk about long chains of these four items. The order is crucial Two long chains of them can pair up only only if there are segments on each where the locks on one pair with the keys on the other and the hooks with the eyes. How many possible combinations of the four are there on a chain of 20 — just 4^20 or 2^40 = 1,099,511,621,776. So to get two randomly chosen chains to pair up exactly is pretty unlikely, unless in some way you or the blind Watchmaker chose them to do so.

Now you need a Turing machine to take a long string of these 4 items and turn it into a protein. In the case of the crucial Notch protein the string of locks, keys, hooks and eyes contains at least 5,000 of them, and their order is important, just as the order of letters in a word is crucial for its meaning (consider united and untied).

The cell has tons of such Turing machines (called ribosomes) and lots of copies of strings coding for Notch (called Notch mRNAs).

The more Notch protein around in the developing brain, the more the proliferating precursors to neurons proliferate before differentiating into neurons, resulting in a bigger brain.

The Notch string doesn’t all code for protein, at one end is a stretch of locks, keys, hooks and eyes which bind other strings, which when bound cause the Notch string to be degraded, mean less Notch and a smaller brain. The other strings are about 20 long and are called microRNAs.

So to get more Notch and a bigger brain, you need to decrease the number of microRNAs specifically binding to the Notch string. One particular microRNA (called miR-143-3p) has it in for the Notch string. So how did primates get rid of miR-143-3p they have an insert (unique to them) in another string which contains 16 binding sites for miR-143-3p. So this string called lincND essentially acts as a sponge for miR-143-3p meaning it can’t get to the Notch string, meaning that neuronal precursor cells proliferate more, and primate brains get bigger.

So can you forget organic chemistry if you want to understand why we have big brains? In the above sense you can. Your understanding won’t be particularly rich, but it will be at a level where chemical explanation is powerless.

No amount of understanding of polyribonucleotide double helices will tell you why a particular choice out of the 1,099,511,621,776 possible strings of 20 will be important. Literally we have moved from physicality to the realm of pure ideas, crossing the Cartesian dichotomy in the process.

Here’s a copy of the original post with lots of chemistry in it and all the references you need to get the molecular biological chops you’ll need.

Why our brains are large: the elegance of its molecular biology

Primates have much larger brains in proportion to their body size than other mammals. Here’s why. The mechanism is incredibly elegant. Unfortunately, you must put a sizable chunk of recent molecular biology under your belt before you can comprehend it. Anyone can listen to Mozart without knowing how to read or write music. Not so here.

I doubt that anyone can start from ground zero and climb all the way up, but here is all the background you need to comprehend what follows. Start here —
and follow the links (there are 5 more articles).

Also you should be conversant with competitive endogenous RNA (ceRNA) — here’s a link —

Also you should understand what microRNAs are — we’re still discovering all the things they do — here’s the background you need —

Still game?

Now we must delve into the embryology of the brain, something few chemists or nonbiological type scientists have dealt with.

You’ve probably heard of the term ‘water on the brain’. This refers to enlargement of the ventricular system, a series of cavities in all our brains. In the fetus, all nearly all our neurons are formed from cells called neuronal precursor cells (NPCs) lining the fetal ventricle. Once formed they migrate to their final positions.

Each NPC has two choices — Choice #1 –divide into two NPCs, or Choice #2 — divide into an NPC and a daughter cell which will divide no further, but which will mature, migrate and become an adult neuron. So to get a big brain make NPCs adopt choice #1.

This is essentially a choice between proliferation and maturation. It doesn’t take many doublings of a NPC to eventually make a lot of neurons. Naturally cancer biologists are very interested in the mechanism of this choice.

Well to make a long story short, there is a protein called NOTCH — vitally important in embryology and in cancer biology which, when present, causes NPCs to make choice #1. So to make a big brain keep Notch around.

Well we know that some microRNAs bind to the mRNA for NOTCH which helps speed its degradation, meaning less NOTCH protein. One such microRNA is called miR-143-3p.

We also know that the brain contains a lncRNA called lncND (ND for Neural Development). The incredible elegance is that there is a primate specific insert in lncND which contains 16 (yes 16) binding sites for miR-143-3p. So lncND acts as a sponge for miR-143-3p meaning it can’t bind to the mRNA for NOTCH, meaning that there is more NOTCH around. Is this elegant or what. Let’s hear it for the Blind Watchmaker, assuming you have the faith to believe in such things.

Fortunately lncND is confined to the brain, otherwise we’d all be dead of cancer.

Should you want to read about this, here’s the reference [ Neuron vol. 90 pp. 1141 – 1143, 1255 – 1262 ’16 ] where there’s a lot more.

Historically, this was one of the criticisms of the Star Wars Missile Defense — the Russians wouldn’t send over a few missles, they’d send hundreds which would act as sponges to our defense. Whether or not attempting to put Star Wars in place led to Russia’s demise is debatable, but a society where it was a crime to own a copying machine, could never compete technically to produce such a thing.

Mind the gap (junction that is)

Gap junctions don’t get much play in pharmacology, or even in neurology, where they are widespread in the central nervous system, linking neurons to neurons, astrocytes to astrocytes. They may get quite a bit more if blocking them is a way of treating metastatic disease (see later).

A bit of background if you’re unfamiliar with them. This is from my notes Molecular Biology of the Cell 4th Edition p. 1074

The gap junction is a cylindrical oligomer composed of 6 identical rod shaped subunits (called connexins). They have 4 transmembrane segments and two extracellular loops which contain a beta-strand structure (and which are an essential structural basis for the docking of the two connexons). Multiple connexons in a membrane tend to form hexagonal arrays.

The gap junction spans the lipid bilayer creating a channel along the central axis. The pore is made of two such protein hexamers one from each cell (called a hemichannel or a connexon) arranged end to end. Different tissues have different specific gap junction proteins (connexins). Man has 14 distinct connexins each encoded by a separate gene (20 homologous proteins in man PNAS 103 pp. 5213 – 5218 ’06). Most cell types express more than one. Connexins are capable of assembling into a heteromeric connexon Adjacent cells expressing different connexins can form intercellular channels in which the two aligned dihalf-channels are different. Each gap junction can contain a cluster of a few to MANY THOUSANDS of CONNEXONs.

Neuroscientists should be interested in them as they form a functional ‘synapse’ between cells, e.g. a way of transferring information between them. For the afficienado there will be much more at the end. To flog a nearly dead horse, this is yet another way a wiring diagram of the brain won’t help you understand it — gap junctions don’t show up when you’re looking at classic synapses. For details see

A recent paper in Nature implied that cancer cells can form gap junctions with astrocytes (a glial cell of the brain). Usually we think of gap junctions being of the same cell type, but not here apparently.

Then they describe a mechanism for the cancer cell tweak the astrocyte so it produces something enabling the cancer cell to survive. Here’s whqt they claim

[ Nature vol. 533 pp. 493 – 498 ’16 ] Human and mouse breast and lung cancer cells express protocadherin7 (PCDH7) whicboth promotes (how?) the assembly of carcinoma – astrocyte gap junctions made of connexin43. PCDH7 normally is only expressed in brain. It joints the stialyl transferase ST6GALNAC5 and neuroserpin as brain restricted proteins which metastastic cells from breast and lung cancer use to colonize the brain.

Metastastic cells then uswe the channels to transfer cGAMP to astrocytes activating the STING pathway, which results in InterferonAlpha (IFNalpha) and Tumor Necrosis Factor (TNF), paracrine signals. These activate STAT1 and NFkappaB in the metastatic cells, supporting tumor growth and chemoresistance.

Meclofenamate and tonabersat are ‘modulators’ of gap junctions, breaking the loop between metastatic cancer cell and the astrocyte. Adding them to the tissue culture studied in the paper, inhibited tumor growth. So here might be a way treat metastatic cancer — particularly since meclofenamate is an FDA approved generic drug available without a prescription.

I think the mechanism described above is incomplete — why should a tumor cell transfer something to another cell to have it secrete something which makes the original cell use something it already had.

Now for a few of the things gap junctions are doing in the brain.

[ Neuron vol. 90 pp. 810 – 823 ’16 ] ManhyGABAeric interneurons (are there other kinds?) IN VITRO are coupled by gap junctions. This work used dual patch clamp recordings of interneurons IN VIVO. They studied coupled cerebellar Golgi cells, and showed that, in the presence of spontaneous background synaptic activity, electrically coupled cerebellar Golgi cells showed robust milliSecond precision correlated activity. This was further enhanced by sensory stimulation.

The electrical coupling equlized membrane potential fluctuations, so that coupled neurons approach action potential threshold together. They say that something called spike triggered spikelets transmitted through gap junctions conditionally triggered postJunctional spikes, if both neurons were close to threshold.

Spikelets are brief low amplitude potentials which look like action potentials but which are much smaller. A spike cannot be generated without a much larger potential change than provided by a spikelet, because the spikelet voltage is too small to activate the ion channels of electrically excitable membranes.

So gap junctions controls the temporal precision and degree of both spontaneous and sensory evoked correlated activity betwen interneurons, by the cooperative effects of shared synaptic depolarization and spikelet transmission.

[ Neuron vol. 90 pp. 912 – 913, 1043 – 1056 ’16 ] It has been found that the strength of electrical coupling between neurons in a network is highly variable (even in the same neuron, so it could be coupled at different strengths with each of its partners). Site specific modulation of electrical coupling quickly reconfigures networks of electrically coupled neurons in the retina. Phosphorylation of connexin36 alters its conductivity.

The number of gap junctions determines the strength of ele tical coupling between cerebellar Golgi cells. Ultrastructural analysis shows that gap junctions vary widely in size, which also influences coupling strength (according to a computer simulation). These are dendro-dendritic electrical synapses (widespread in the brain between inhibitory interneurons).

Only 18% or so of the channels present at the gap junctions account for the boserved strength of electrical transmission between cerebellar golgi cells.

Somato-somatic junctions occur in the mammalian trigeminal mesencephalic nucleus. Could the excess junctions be acting as adhesion molecules.

In one system, the turnover of gap junction channel proteins is rapid and comparable with that of glutamic acid receptors.

Gap junctions are ‘low pass filters’ (they pass slow fluctuations of membrane potential better than they pass rapid fluctuations). This is why the electrical synapses are inhibitory — each action potential from a Golgi cell consists of a rapid (but brief) depolarizing spike followed by a relatively deep and protracted afterhyperpolarization — which is 200 times longer than the spike — and transmitted much more effectively.

Inhibition by sparse excitatory input breaks up Golgi network synchronization, because the coupling to adjacent cells is different for each one, causing dispersion of the spikes.

In quietly attentive animals cerebellar Golgi cells generate rhythmic synchronous activity at 8 Hertz. The same behavior is seen in cerebellar slices. The hyperpolarizing electrical post-synaptic potentials (PSPs) are the only synchronizing force. This is the default state, but it can be disrupted by a variety of sensory stimuli (or by movements) which reduce spiking frequency and rhythmicity.

Golgi cells can inhibit thousands of granule cells, and every granule cell gets inhibitory input from 4 – 8 Golgi cells. The transient nature of network desynchronization ‘could’ allow the cerebellar input layer to act as a timing device over the 10 milliSecond to 1 second timescale.

In a gambling mood?

If a pair of posters to be presented Monday 6 June at the 2016 American Society of Clinical Oncology Annual Meeting in Chicago, Illinois, contains the results of a phase III clinical trial of rigosertib, and if the results are as good as a paper discussed below the stock Onconova Therapeutics (ONTX) will jump by a factor of 10 to 100.

Full disclosure: I own some. The posters may just describe the clinical trial rather than report the results in which case all bets are off. In that case, I’ll just hold the stock until the results are in. This isn’t the ‘pump and dump’ beloved of boiler room operators everywhere. The rationale for the drug and my take on the original paper (3 May ’16) are reproduced below.

Has the great white whale of oncology finally been harpooned?

The ras oncogene is the great white whale of oncology. Mutations in 20 – 40% of cancer turn its activity on so that nothing can turn it off, resulting in cellular proliferation. People have been trying to turn mutated ras off for years with no success.

A current paper [ Cell vol. 165 pp. 643 – 655 ’16 ] describes a new and different way to attack it. Once ras is turned on (either naturally or by mutation) many other proteins must bind to it, to produce their effects — they are called RAS effectors, among which are the uneuphoniously named RAF, RalGDS and PI3K. They bind to activated ras by the cleverly named Ras Binding Domain (RBD) which has 78 amino acids.

The paper describes rigosertib, a not that complicated molecule to the chemist, which inhibits the binding (by resembling the site on ras that the RBD binds to). It is a styryl benzyl sulfone and you can see the structure here —

What’s good about it? Well it is in phase III trials for a fairly uncommon form of cancer (myelodysplastic syndrome). That means it isn’t horribly toxic or it wouldn’t have made it out of phase I.

Given the mechanism described, it is possible that Rigosertib will be useful in 20 – 40% of all cancer. Can you say blockbuster drug?

Do you have a speculative bent? Buy the company testing the drug and owning the patent — Onconova Therapeutics. It’s quite cheap — trading at $.40 (yes 40 cents !). It once traded as high as $30.00 — symbol ONTX. I don’t own any (yet), but for the price of a movie with a beer and some wings afterwards you could be the proud owner of 100 shares. If Rigosertib works, the stock will certainly increase more than a hundredfold.

Enough kidding around. This is serious business. In what follows you will find some hardcore molecular biology and cellular physiology showing just what we’re up against. Some of the following is quite old, and probably out of date (like yours truly), but it does give you the broad outlines of what is involved.

The pathway from Ras to the nucleus

The components of the pathway had been found in isolation (primarily because mutations in them were associated with malignancy). Ras was discovered as an oncogene in various sarcoma viruses. Mutations in ras found in tumors left it in a ‘turned on’ state, but just how ras (and everything else) fit into the chain of binding of a growth factor (such as platelet derived growth factor, epidermal growth factor, insulin, etc. etc.) to its receptor on the cell surface to alterations in gene expression wasn’t clear. It is certain to become more complicated, because anything as important as cellular proliferation is very likely to have a wide variety of control mechanisms superimposed on it. Although all sorts of protein kinases are involved in the pathway it is important to remember that ras is NOT a protein kinase.

l. The first step is binding of a growth factor to its receptor on the cell surface. The receptor is usually a tyrosine kinase. Binding of the factor to the receptor causes ‘activation’ of the receptor. Activation usually means increasing the enzymatic activity of the receptor in the tyrosine kinase reaction (most growth factor receptors are tyrosine kinases). The increase in activity is usually brought about by dimerization of the receptor (so it phosphorylates itself on tyrosine).

2. Most activated growth factor receptors phosphorylate themselves (as well as other proteins) on tyrosine. A variety of other proteins have domains known as SH2 (for src homology 2) which bind to phosphorylated tyrosine.

3. A protein called grb2 binds via its SH2 domain to a phosphorylated tyrosine on the receptor. Grb2 binds to the polyproline domain of another protein called sos1 via its SH3 domain. At this point, the unintiated must find the proceedings pretty hokey, but the pathway is so general (and fundamental) that proteins from yeast may be substituted into the human pathway and still have it work.

4. At last we get to ras. This protein is ‘active’ when it binds GTP, and inactive when it binds GDP. Ras is a GTPase (it can hydrolyze GTP to GDP). Most mutations which make ras an oncogene decrease the GTPase activity of RAS leaving it in a permanently ‘turned on’ state. It is important for the neurologist to know that the defective gene in type I neurofibromatosis activates the GTPase activity of ras, turning ras off. Deficiencies (in ras inactivation) lead to a variety of unusual tumors familiar to neurologists.

Once RAS has hydrolyzed GTP to GDP, the GDP remains bound to RAS inactivating it. This is the function of sos1. It catalyzes the exchange of GDP for GTP on ras, thus activating ras.

5. What does activated ras do? It activates Raf-1 silly. Raf-1 is another oncogene. How does activated ras activate Raf-1 ? Ras appears to activate raf by causing raf to bind to the cell membrane (this doesn’t happen in vitro as there is no membrane). Once ras has done its job of localizing raf to the plasma membrane, it is no longer required. How membrane localization activates raf is less than crystal clear. [ Proc. Natl. Acad. Sci. vol. 93 pp. 6924 – 6928 ’96 ] There is increasing evidence that Ras may mediate its actions by stimulating multiple downstream targets of which Raf-1 is only one.

6. Raf-1 is a protein kinase. Protein kinases work by adding phosphate groups to serine, threonine or tyrosine. In general protein kinases fall into two classes those phosphorylating on serine or threonine and those phosphorylating on tyrosine. Biochemistry has a well documented series of examples of enzymes being activated (or inhibited) by phosphorylation. The best worked out is the pathway from the binding of epinephrine to its cell surface receptor to glycogen breakdown. There is a whole sequence of one enzyme phosphorylating another which then phosphorylates a third. Something similar goes on between Raf-1 and a collection of protein kinases called MAPKs (mitogen activated protein kinases). These were discovered as kinases activated when mitogens bound to their extracellular receptors.There may be a kinase lurking about which activates Raf (it isn’t Ras which has no kinase activity). Removal of phosphate from Raf (by phosphatases) inactivates it.

7. Raf-1 activates members of the MAPK family by phosphorylating them. There may be several kinases in a row phosphorylating each other. [ Science vol. 262 pp. 1065 – 1067 ’93 ] There are at least three kinase reactions at present at this point. It isn’t known if some can be sidestepped. Raf-1 activates mitogen activated protein kinase kinase (MAPK-K) by phosphorylation (it is called MEK in the ras pathway). MAPK-K activates mitogen activation protein kinase (MAPK) by phosphorylation. Thus Raf-1 is actually mitogen activated protein kinase kinase kinase (sort of like the character in Catch-22 named Junior Junior Junior). (1/06 — I think that Raf-1 is now called BRAF)

8. The final step in the pathway is activation of transcription factors (which turn genes off or on) by MAP kinases by (what else) phosphorylation. Thus the pathway from cell surface is complete.

Internal Energy, Enthalpy, Helmholtz free energy and Gibbs free energy are all Legebdre transformations of each other

Sometimes it pays to be persistent in thinking about things you don’t understand (if you have the time as I do). The chemical potential is of enormous interest to chemists, and yet is defined thermodynamically in 5 distinct ways. This made me wonder if the definitions were actually describing essentially the same thing (not to worry they are).

First, a few thousand definitions

Chemical potential of species i — mu(i)
Internal energy — U
Entropy — S
Enthalpy — H
Helmholtz free energy — F or A (but, maddeningly, never H)
Gibbs free energy — G
Ni — number of elements of chemical species i
Pressure — p
Volume — V
Temperature — T

Just 5 more
mu(i) == ∂H/∂Ni constant S, p
mu(i) == ∂S/∂Ni constant U, V
mu(i) == ∂U/∂Ni constant S, V
mu(i) == ∂F/∂Ni constant T, V
mu(i) == ∂G/∂Ni constant T, p

Clearly, at a given constellation of S, U, F, G the mu(i)’s won’t all be the same number, but they’re essentially the same thing. Here’s why.

Start with a simple mathematical problem. Assume you have a simple function (f) of two variables (x,y), and that f is continuous in x and y and that its partial derivatives u = ∂f/∂x and w = ∂f/∂y are continuous as well so you have

df = u dx + w dy

u and dx are conjugate variables, as are w and dy

Suppose you want to change df = u dx + w dy to

another function g such that

dg = u dx – y dw

which is basically flipping a pair of conjugate variables around

Patience, the reason for wanting to do this will become apparent in a moment.

The answer is to use what is called the Legendre transform of f which is simply

g = f – y w

dg = df – y dw – w dy

plug in df

dg = u dx + w dw – y dw – w dy == df – y dw – w dy Done.

Where does the thermodynamics come in?

Well, you have to start somewhere, so why not with the fundamental thermodynamic equation for internal energy U

dU = ∂U/∂S dS + ∂U/∂V dV + ∑ ∂U/∂Ni dNi

We already know that ∂U/Ni = mu(i)

Because of the partial derivative notation (∂) it is assumed that all the other variables say in the expression for dU e.g. V and Ni are held constant in ∂U/∂S. This will reduce the clutter in notation which is already cluttered enough.

We already know that ∂U/∂Ni is mu(i). One definition of temperature T, is as ∂U/∂S, and another for p is -∂U/∂V (which makes sense if you think about it — decreasing volume relative to U should increase pressure).

Suddenly dU looks like what we were talking about with the Legendre transformation.

dU = T dS – p dV + ∑ mu(i) dNi

Apply the Legendre transformation to U to switch conjugate variables p and V

H = U + pV ; looks suspiciously like enthalpy (H) because it is

dH = dU + p dV + V dp + ∑ mu(i) dNi

= T dS – p dV + ∑ mu(i) dNi + p dV + V dp

= T dS + V dp + ∑ mu(i) dNi

Notice how mu(i) here comes out to ∂H/dNi at constant S and P

Start with the fundamental thermodynamic equation for internal energy

dU = T dS – p dV + ∑ mu(i) dNi

Now apply the Legendre transformation to T and S and you get
F = U – TS ; looks like the Helmholtz free energy (sometimes written A, but never as H) because it is.

You get

dF = – S dT – p dV + ∑ mu(i) dNi

Who cares? Chemists do because, although it is difficult to hold U constant or S constant (and it is impossible to measure them directly) it is very easy to keep temperature and volume constant in a reaction, meaning that changes in Helmholtz free energy under those conditions is just
∑ mu(i) dNi. So here mu(i) = ∂F/∂Ni at constant T and p

If you start with enthalpy

dH = T dS + V dp + ∑ mu(i) dNi

and do the Legendre transform you get the Gibbs free energy G = H – TS

I won’t bore you with it but this gives you the chemical potential mu(i) at constant T and p, conditions chemists easily arrange all the time.

To summarize

Enthalpy (H) is one Legendre transform of internal energy (U)
Helmholtz free energy (F) is another Legendre transform of U
Gibbs free energy (G) is the Legendre transform of Enthalpy (H)

It should be clear that Legendre transforms are all reversible

For example if H = U + PV then U = H – PV

If you think a bit about the 5 definitions of chemical potential, you’ll see that it can depend on 5 things (U, S, p, V and T). Ultimately all thermodynamic variables (U, S, H, G, F, p, V, T, mu(i) ) often have relations to each other.

Examples include H = U + pV, F = U – TS, G = H -TS

Helping keep things clear are equations of state from the things you can easily measure (p,V, T). The most famous is the ideal gas law p V = nRT.