Category Archives: Neurology & Psychiatry

Spot the flaw

Mathematical talent varies widely. It was a humbling thing a few years ago to sit in an upper level college math class on Abstract Algebra with a 16 year old high school student taking the course, listening with one ear while he did his German homework. He was later a double summa in both math and physics at Yale. So do mathematicians think differently? A fascinating paper implies that they use different parts of their brain doing math than when not doing it. The paper has one methodological flaw — see if you can find it.

[ Proc. Natl. Acad. Sci. vol. 113 pp. 4887 – 4889, 4909 – 4917 ’16 ] 15 expert mathematicians and 15 nonMathematicians with comparable academic qualifications were studied (3 literature, 3 history 1 philosophy 2 linguistics, 1 antiquity, 3 graphic arts and theater 1 communcation, 1 heritage conservation — fortunately no feminist studies). They had to listen to mathematical and nonMathematical statements and decide true false or meaningless. The nonMathematical statements referred to general knowledge of nature and history. All this while they were embedded in a Magnetic Resonance Imager, so that functional MRI (fMRI) could be performed.

In mathematicians there was no overlap of the math responsive network (e.g. the areas of the brain activated by doing math) with the areas activated by sentence comprehension and general semantic knowledge.

The same brain networks were activated by all types of mathematical statement (geometry, analysis, algebra and topology) as opposed to nonMathematical statement. The areas activated were the dorsal parietal, ventrolateral temporal and bilateral frontal. This was only present in the expert mathematicians (and only to mathematical statements) These areas are outside those associated with language (inferior frontal gyrus of the left hemisphere). The activated areas are also involved in visual processing of arabic numbers and simple calculation. The activated areas in mathematicians were NOT those related to language or general knowledge.

So what’s wrong with the conclusion? The editorialist (pp. 4887 – 4889) pointed this out but I thought of it independently.

All you can say is that experts working in their field of expertise use different parts of their brain than they use for general knowledge. The nonMathematicians should have been tested in their field of expertise. Shouldn’t be hard to do.

NonAlgorithmic Intelligence

Penrose was right. Human intelligence is nonAlgorithmic. But that doesn’t mean that our physical brains produce consciousness and intelligence using quantum mechanics (although all matter is what it is because of quantum mechanics). The parts (even small ones like neurotubules) contain so much mass that their associated wavefunction is too small to exhibit quantum mechanical effects. Here Penrose got roped in by Kauffman thinking that neurotubules were the carriers of the quantum mechanical indeterminacy. They aren’t, they are just too big. The dimer of alpha and beta tubulin contains 900 amino acids — a mass of around 90,000 Daltons (or 90,000 hydrogen atoms — which are small enough to show quantum mechanical effects).

So why was Penrose right? Because neural nets which are inherently nonAlgorithmic are showing intelligent behavior. AlphaGo which beat the world champion is the most recent example, but others include facial recognition and image classification [ Nature vol. 529 pp. 484 – 489 ’16 ].

Nets are trained on real world images and told whether they are right or wrong. I suppose this is programming of a sort, but it is certainly nonAlgorithmic. As the net learns from experience it adjusts the strength of the connections between its neurons (synapses if you will).

So it should be a simple matter to find out just how AlphaGo did it — just get a list of the neurons it contains, and the number and strengths of the synapses between them. I can’t find out just how many neurons and connections there are, but I do know that thousands of CPUs and graphics processors were used. I doubt that there were 80 billion neurons or a trillion connections between them (which is what our brains are currently thought to have).

Just print out the above list (assuming you have enough paper) and look at it. Will you understand how AlphaGo won? I seriously doubt it. You will understand it less well than looking at a list of the positions and momenta of 80 billion gas molecules will tell you its pressure and temperature. Why? Because in statistical mechanics you assume that the particles making up an ideal gas are featureless, identical and do not interact with each other. This isn’t true for neural nets.

It also isn’t true for the brain. Efforts are underway to find a wiring diagram of a small area of the cerebral cortex. The following will get you started — https://www.quantamagazine.org/20160406-brain-maps-micron-program-iarpa/

Here’s a quote from the article to whet your appetite.

“By the end of the five-year IARPA project, dubbed Machine Intelligence from Cortical Networks (Microns), researchers aim to map a cubic millimeter of cortex. That tiny portion houses about 100,000 neurons, 3 to 15 million neuronal connections, or synapses, and enough neural wiring to span the width of Manhattan, were it all untangled and laid end-to-end.”

I don’t think this will help us understand how the brain works any more than the above list of neurons and connections from AlphaGo. There are even more problems with such a list. Connections (synapses) between neurons come and go (and they increase and decrease in strength as in the neural net). Some connections turn on the receiving neuron, some turn it off. I don’t think there is a good way to tell what a given connection is doing just by looking a a slice of it under the electron microscope. Lastly, some of our most human attributes (emotion) are due not to connections between neurons but due to release of neurotransmitters generally into the brain, not at the very localized synapse, so it won’t show up on a wiring diagram. This is called volume neurotransmission, and the transmitters are serotonin, norepinephrine and dopamine. Not convinced? Among agents modifying volume neurotransmission are cocaine, amphetamine, antidepressants, antipsychotics. Fairly important.

So I don’t think we’ll ever truly understand how the neural net inside our head does what it does.

Is a rational treatment for Multiple Sclerosis in our future?

Two very recent papers taken together point the way to a rational treatment of multiple sclerosis (and probably all autoimmune disease). The short story:
Paper #1 found a way to find the antigen or antigens patients with MS are reacting to
Paper #2 found a way to selectively impair the response to an inciting antigen without clobbering the whole immune system

Some history: Some evening in 1966 or 1967 a fellow neurology resident and I were sitting on the ward having dealt with the complications of high doses corticosteroids for a case of optic neuritis (often the first sign of MS). I said, some day they’ll look at what we’re doing the way we look at docs of 200 years ago using leeches (and bloodletting). As a kid, I remember my parents driving into Philly. Shortly after getting over the Ben Franklin bridge we’d pass a pharmacy offering leeches on its sign.

It was obvious even back then that MS in some way was an attack by the immune system on the brain. Finding the particular antigen the system was reacting to would lead us to the cause and hopefully less simplistic treatment than clobbering the immune system. We didn’t know all the proteins we had or even how many, so people would look for antibodies to a variety causes (which they’d arrived at by reasoning, not data). Increased antibody titers to a variety of viruses were found, but that led nowhere. No one ever isolated a virus from MS brain, although sightings on electron microscopy were eagerly reported. Eventually it became obvious that the immune system was on high alert with increased antibodies to lots of things.

This leads to paper #1 [ Proc. Natl. Acad. Sci. vol. 113 pp. 2188 – 2193 ’16 ] To make a long story short they used something called the Human Protein Atlas Program to find what proteins the antibodies in MS patients were reacting to. So rather than having a theory about what MS patients might be reacting to and testing it, they looked at all proteins and watched. It’s the difference between being a Greek philosopher reasoning things out from first principles and collecting data. Only when the technology is available can you stop a priori theorizing and just look. Don’t be too hard on the earlier researchers, they didn’t have the tools.

The found that MS patients were reacting to a protein called anoctamin2, which actually showed increased expression near and inside the demyelinating plaques of MS.

For the gory details keep reading, otherwise skip to **** where I’ll discuss paper #2

Gory details — The Human Protein Atlas produces human protein fragments, selected on the basis of their low similarity to other proteins in the proteome. [ Science vol. 347 1260419 (23 Jan) ’15 ] The atlas hopes to find out where and how much of each our proteins is at the tissue and cellular level. It is based on antibody based profiling on tissue microarrays (of proteins?). This based on transcript expression (RNA-Seq), and immunohistochemistry (24,028 antibodies coresponding to 16,975 protein coding genes). 44 tissues were studied. The antibodies produced more than 13 million tissue based mmunohistochemistry images. They also report subproteomes (secreted proteins n = 3,171, and membrane bound proteins n = 5,570). Interstingly there was an overall concurrence between mRNA and protein levels for a given gene product across various tissues.

The PNAS paper profiled 2,169 plasma samples from MS cases and population based controls (with neurologic disease) using bead arrays built with 384 human protein fragments seleted from an initial screening with 11,520 antigens. There was increased reactivity to anoctamin2 (aka TMEM16B) in MS vs. controls (by how much?). This was corroborated in independent assay with alternative protein constructs and by epitope mapping with peptides covering the identified region of anoctamin 2.

ImmunoFLuorescence in human MS brain tissue showed increased anoctamin2 expression as small cellular aggregates near and inside MS lesions. The controls had other neurologic disease. There was a 5.3 fold change in fluorescence intensity in the MS group. The antibodies are directed against the amino terminal region.

*****

*****

Paper #2 — [ Nature vol. 530 pp. 422 – 423, 434 – 440 ’16 ] basically found a way to knock out the immune system’s response to a single antigen — not all of them. The point is that just an antigen by itself isn’t enough to turn the immune system on. A costimulatory molecule must also be present on the antigen presenting cell. If it isn’t there the immune system is actually turned off by forming regulatory T cells (which even though they are part of the immune system they actually turn it off).

One can form models of human autoimmune disease in mice. Two such are EAE (Experimental Allergic Encephalomyelitis) formed by giving the animal myelin basic protein (a constituent of myelin which is attacked in MS), and rheumatoid arthritis (formed by giving collagen to the animals). What is so great about this paper is that MHC II carrying peptides from collagen suppress disease in a mouse model of rheumatoid arthritis, but NOT in mice with EAE. MHC-II carrying CNS antigen peptides control EAE but not collagen induced arthritis.. In addition neither treatment impaired the immune response to infection — something that almost always happens when you clobber the immune system.

Well it’s a long way from the lab to the bedside, but imagine finding what the immune system is reacting to and stopping it (without stopping the immune system). That’s what these two papers portend. Exciting times.

Nicastrin the gatekeeper of gamma secretase

Once a year some hapless trucker from out of town gets stuck trying to drive under a nearby railroad bridge with a low clearance. This is exactly the function of nicastrin in the gamma secretase complex which produces the main component of the senile plaque, the aBeta peptide.

Gamma secretase is a 4 protein complex which functions as an enzyme which can cut the transmembrane segment of proteins embedded in the cell membrane. This was not understood for years, as cutting a protein here means hydrolyzing the amide bond of the protein, (e.g. adding water) and there is precious little water in the cell membrane which is nearly all lipid.

Big pharma has been attacking gamma secretase for years, as inhibiting it should stop production of the Abeta peptide and (hopefully) help Alzheimer’s disease. However the paper to be discussed [ Proc. Natl. Acad. Sci. vol. 113 p.n E509 – E518 ’16 ] notes that gamma secretase processes ‘scores’ of cell membrane proteins, so blanket inhibition might be dangerous.

The idea that Nicastrin is the gatekeeper for gamma secretase is at least a decade old [ Cell vol. 122 pp. 318 – 320 ’05 ], but back then people were looking for specific binding of nicastrin to gamma secretase targets.

The new paper provides a much simpler explanation. It won’t let any transmembrane segment of a protein near the active site of gamma secretase unless the extracellular part is lopped off. The answer is simple mechanics. Nicastrin is large (709 amino acids) but with just one transmembrane domain. Most of it is extracellular forming a blob extending out 25 Angstroms from the membrane, directly over the substrate binding pocket of gamma secretase. Only substrates with small portions outside the membrane (ectodomains) can pass through it. It’s the railroad bridge mentioned above. Take a look at the picture — https://en.wikipedia.org/wiki/Nicastrin

This is why a preliminary cleavage of the Amyloid Precursor Peptide (APP) is required for gamma secretase to work.

So all you had to do was write down the wavefunction for Nicastrin (all 709 amino acids) and solve it (assuming you even write it down) and you’d have the same answer — NOT. Only the totally macroscopic world explanation (railroad bridge) is of any use. What keeps proteins from moving through each other? Van der Waals forces. What help explain them. The Pauli exclusion principle, as pure quantum mechanics as it gets.

Freud was right, there is an unconscious mind and it’s pretty smart.

Freud has fallen out of favor, with his analogies of the workings of the mind to a steam engine (drives, pressures, releases, displacements), the dominant technology of his day–as the computer is to ours. However, the following paper [ Proc. Natl. Acad. Sci. vol. 113 pp. E616 – E6125 ’16 ] shows that we have an unconscious mind, and that it is mathematically sophisticated (although I don’t think the authors made this point).

[ Proc. Natl. Acad. Sci. vol. 113 pp. E616 – E625 ’16 ] The work used magnetoencephalography (MEG) https://en.wikipedia.org/wiki/Magnetoencephalography, to record brain activity in response to a series of tone pips. MEG is conceptually quite similar to the electrocardiogram (EKG) or the electroencephalogram (EEG), both measuring voltage differences between two electrodes over time. Well, a voltage difference causes an electric current to flow through a conductor, and the nice wet brain is nothing if not that. Anyone who has studied how an electric motor works, knows that an electric current produces a magnetic field, which is what the MEG measures. The great advantage of MEG is that it is temporally precise, and changes can be measured in milliSeconds.

So what did they do? They presented tone pips to an unspecified number of subjects. The relation of one pip to another could either be completely random (RAND) or part of a repeating pattern — say pip pip pip silence silence pip pip pip silence silence (PAT). In one series of experiments, subjects were asked to press a button as soon as the pip sequence went from random to patterned (RAND –> PAT), all this while the MEG was being recorded. In another, they were asked to do this for PAT –> RAND. The subjects were as good as something called the the mathematical ideal observer of the variable order Markov model. It only took one or two random pips after a patterned sequence to notice it and press the button. They could also pick up that a pattern was formed midway through the second repetition of a pattern.

The MEG showed abrupt changes at either transition (RAND –> PAT or PAT –> RAND). The work didn’t stop with just sequences of just one tone. They could use an ‘alphabet of tones’. The subjects could pick up when the number of tones in the alphabet changed, again with MEG values to match.  So they had an independent signal from the MEG show that the brain picked up the transition without requiring any cooperation from the subjects.  All very nice, but anyone who likes music can do this.

Then the subjects were then asked to perform the n-back task, in which the subject is presented with a sequence of stimuli;  and the task consists of indicating when the current stimulus matches the one from n steps earlier in the sequence. Tricky isn’t it? Certainly, something that requires concentration. The load factor n can be adjusted to make the task more or less difficult. You’ve got to hold the sequence just presented in your head so the n-back task is a test of working memory.

Drum roll —

If the tone pips were presented while the subjects were doing the n-back task, the MEG still picked up RAND –> PAT and PAT –> RAND transitions, something the subjects weren’t consciously trying to do.

We know the brain does all sorts of things unconsciously — e.g. breathing. But they are pretty simple. The tests here are conceptually subtle. Your unconscious brain picks up statistical regularities and irregularities without your consciously trying. Who knows what else it does — maybe Freud was right.

Why should this be useful? Well, you’d want to know if a predator is sneaking up on you. The same work should be done with animals performing a task they’ve been trained to do.

Why some of us gamble

If you are one of the hapless schlubs who bought a Powerball ticket or two and didn’t win (like me), modern neuroscience can tell you why (but not without a bit of pompous terminology). They call a small chance of winning large amounts — e.g. Powerball along with a large chance of losing a little a positively skewed gamble. Impressed? I’m not.

Nonetheless [ Neuron vol. 89 pp. 63 – 69 ’16 ] is a very interesting paper. Functional Magnetic Resonance Imaging (fMRI – https://en.wikipedia.org/wiki/Functional_magnetic_resonance_imaging) has shown that increased blood flow in one area of the brain (the nucleus accumbent sept) predicts risk seeking choices, while increased blood flow in another (the anterior insula) predicts aversion to risk. The unproven assumption behind fMRI is that increased blood flow is due to increased neural activity.

The neurochemistry of the two regions is quite different. The accumbens gets dopamine input while the insula gets norepinephrine projections.

BTW the insula is unique to man. Our cortex has grown so much (particularly in the frontal region) that it folds over on itself burying the insular cortex — https://en.wikipedia.org/wiki/Insular_cortex.

We are now actually able to measure axon bundles (white matter fiber tracts) in the living brain, using something called diffusion weighted MRI. By and large, fiber tracts tend to have the fibers parallel and running in the same direction. It is far easier for water to flow along parallel fibers than across them, and this is what the technique measures. For the mathematically inclined, what is actually measured is a tensor field, because at any given point in the brain it varies in direction, unlike a vector field which points just one way at a given point (the mathematically inclined know that this is a simplification because vectors are actually a type of tensor).

At any rate, the work used diffusion wieghted MRI to study the white matter tracts connecting the two areas. The larger the tract between insula and accumbens was the more risk averse an individual was. The implication being that a larger tract is a better connection between the two. So your nucleus is your impulsive child and the anterior insula is your mother telling you not to do anything crazy.

Fascinating, but like all this stuff it needs to be replicated, as it probably confirms the original idea for the study.

The most interesting thing to an evolutionist is not that APOE4 increases the risk of Alzheimer’s disease

Neurologists were immensely excited by the discovery 25 years ago that the APOE4 variant of APOlipoprotein E increases the risk of Late Onset Alzheimer’s Disease (LOAD). 24,000 papers later (Google Scholar) we still don’t know how it does it. Should all this work have been done ? Of course ! !  Once we know the mechanism(s) by which APOE4 increases Alzheimer’s risk we’ll have new ideas to help us attack.

The APOE gene has 3 variants (alleles) APOE2, 3 and 4. The protein is average sized (299 amino acids). The 3 alleles differ at two positions (amino acids #112 and #158) where either cysteine or arginine can be found. The frequency of APOE4 is 14% in the adult white population, that of E3 is 78% and that of E2 is 8%.

Fascinating as this all is, it’s not what’s interesting from an evolutionary point of view.

[ Proc. Natl. Acad. Sci. vol. 113 pp. 17 – 18, 74 – 79 ’16 ] Postmenpausal longevity in females is not limited to humans. Humans, orcas and pilot whales are the only vertebrate species known to have prolonged postreproductive lifespans. Our fertility ends at about the same age that fertility ends in other female hominids (the great apes). However, apes rarely live into their 40s (even in captivity).

Unlike APOE4, APOE2 and APOE3 protect against late onset Alzheimer’s.

The fascinating point is that APOE2 and APOE3 aren’t found in the great apes. They are a human invention. Now LOAD occurs well past reproduction, so there should be no reason in terms of reproductive success for them to arise and be more common in human populations than the original APOE4.

Even more interesting is some work on another protein CD33, found on immune cells and glia in the brain [ Neuron vol. 78 pp. 575 – 577, 631 – 643 ’13 ] A minor allele (21% frequency in human populations) of CD33 (SNP rs 3865444) protects against Alzheimer’s. The allele is associated with reductions in CD33 expression in microglia, and also with reduction in levels of insoluble Abeta42 in (Alzheimer’s) brain. The numbers of CD33+ microglia correlate with insoluble Abeta42 levels and amyloid plaque burden. So decreasing (or inhibiting) CD33 function might help Alzheimer patients.

Again the protective allele is only found in man. The great apes don’t have it just the major (nonprotective) allele.

Again, there is no way that having the allele directly improves your reproductive success. By the time it is protecting you, you’re infertile.

What in the world is going on? Why did alleles protective against Alzheimer’s arise in two very different proteins in the course of human evolution?

“There is something fascinating about science. One gets such wholesale returns of conjecture out of such a trifling investment of fact.” — Mark Twain.

The reason these alleles probably arose gets us in to an ancient battle in evolutionary theory — what is the actual unit of selection? It may be the group rather than the individual. Face it, human infants and children are helpless for longer than other primates, and need others to care for them, for at least 5 years. Who better than grandma and grandpa? So the fact that with granny around more children survive to reproduce constitutes group selection (I think).

As Theodosius Dobzhansky said “Nothing in Biology Makes Sense Except in the Light of Evolution”

It ain’t the bricks it’s the plan — take II

A recent review in Neuron (vol. 88 pp. 681 – 677 ’15) gives a possible new explanation of how our brains came to be so different from apes (if not our behavior of late).

You’ve all heard that our proteins are only 2% different than the chimp, so we are 98% chimpanzee. The facts are correct, the interpretation wrong. We are far more than the protein ‘bricks’ that make us up, and two current papers in Cell [ vol. 163 pp. 24 – 26, 66 – 83 ’15 ] essentially prove this.

This is like saying Monticello and Independence Hall are just the same because they’re both made out of bricks. One could chemically identify Monticello bricks as coming from the Virginia piedmont, and Independence Hall bricks coming from the red clay of New Jersey, but the real difference between the buildings is the plan.

It’s not the proteins, but where and when and how much of them are made. The control for this (plan if you will) lies outside the genes for the proteins themselves, in the rest of the genome (remember only 2% of the genome codes for the amino acids making up our 20,000 or so protein genes). The control elements have as much right to be called genes, as the parts of the genome coding for amino acids. Granted, it’s easier to study genes coding for proteins, because we’ve identified them and know so much about them. It’s like the drunk looking for his keys under the lamppost because that’s where the light is.

We are far more than the protein ‘bricks’ that make us up, and two current papers in Cell [ vol. 163 pp. 24 – 26, 66 – 83 ’15 ] essentially prove this.

All the molecular biology you need to understand what follows is in the following post — https://luysii.wordpress.com/2010/07/07/molecular-biology-survival-guide-for-chemists-i-dna-and-protein-coding-gene-structure.

The neuron paper is detailed and fascinating to a neurologist, but toward the end it begins to fry far bigger fish.

Until about 10 years ago, molecular biology was incredibly protein-centric. Consider the following terms — nonsense codon, noncoding DNA, junk DNA. All are pejorative and arose from the view that all the genome does is code for protein. Nonsense codon means one of the 3 termination codons, which tells the ribosome to stop making protein. Noncoding DNA means not coding for protein (with the implication that DNA not coding for protein isn’t coding for anything).

Well all that has changed. The ENCODE Consortium showed that well over half (and probably all) our DNA is transcribed into RNA — for details see https://en.wikipedia.org/wiki/ENCODE. This takes energy, and it is doubtful (to me at least) that organisms would waste this much energy if the products were not doing something useful.

I’ve discussed microRNAs elsewhere — for details please see — https://luysii.wordpress.com/2010/07/14/junk-dna-that-isnt-and-why-chemistry-isnt-enough/. They don’t code for protein either, but control how much of a given protein is made.

The Neuron paper concerns lncRNAs (long nonCoding RNAs). They don’t code for protein either and contain over 200 nucleotides. There are a lot of them (10,000 – 50,000 are known to be expressed in man. Amazingly 40% of them are expressed in the brain, and not just in adult life, but during embryonic development. Expression of some of them is restricted to specific brain areas. It is easier for an embryologist to tell what type a cell is during brain cortical development by looking at the lncRNAs expressed than by the proteins a given cell is making. The paper contains multiple examples of the lncRNAs controlling when and where a protein is made in the brain.

lncRNAs can contain multiple domains, each of which has a different affinity for a particular RNA (such as the mRNA for a protein), or DNA, or protein. In the nucleus they influence the DNA binding sites of transcription factors, RNA polymerase II, the polycomb repressor complex. The review goes on with many specific examples of lncRNA function — synaptic plasticity, neurotic extension.

Getting back to proteins, the vast majority are nearly the same in all mammals (this is where the 2% Chimpanzee argument comes from). Here is where it gets interesting. Roughly 1/3 of lncRNAs found in man are primate specific. This includes hundreds of lncRNAs found only in man. The paper gives evidence that hundreds of them have shown evidence of positive selection in humans.

So the paper provides yet another mechanism (with far more detail than I’ve been able to provide here) for why our brains are so much larger, and different in many ways than our nearest evolutionary ancestor, the chimpanzee. This is the largest molecular biological difference found so far for the human brain as opposed to every other brain. Fascinating stuff. Stay tuned. I think this is a watershed paper.

A new form of matter?

Has cellular biology and biochemistry shown us a new form of matter? It’s certainly something I never studied in PChem back in the day. It goes by multiple names, and may be more than one thing.

Start with the nucleolus — it’s been known for years, a visible agglomeration of proteins and RNA in the nucleus, not bound by a membrane. Then there is the processing body (aka stress granule), also made of proteins and RNA (but different ones — transcription factors and mRNAs). Then there is the nuclear pore, made of ‘low complexity sequence tails of proteins surrounding the pore (mostly phenylalanine glycine repeats — aka FG repeats) thought to form a barrier to protein movement through the pore. Then there are RNA granules – said to occur by a phase transition to a hydrogel-like phase (whatever that is). Neurologists have long been interested in FUS/TLS a protein which is mutated in some forms of Amyotrophic Lateral Sclerosis and dementia.

I do think that we’re at the blind men and the elephant stage trying to sort all this out (which, of course, makes it fascinating and a fit subject for scientific work — apologies Wittgenstein — “What we cannot speak about we must pass over in silence”

So in what follows you’ll find a lot of information about these matters, which does not have a neat and tidy explanation. This is what science looks like when it’s being done.

[ Cell vol. 149 pp. 753 – 767, 768 -779 ’12 ] RNA granules don’t just occur in dendrites — they are found in (1) germ cell P granules of C. elegans embryos (2) polar granules of Drosophila embryos (3) stress grnules appearing in cultured yeast and mammalian cells on nutrient deprivation or other forms of metabolic stress (4) neuronal granules transporting mRNAs to dendrites.

Unsurprisingly, the granule contains RNA binding proteins (with KH or RNA Recognition motif (RRM) domains). These domains allow proteins containing them to recognize 3’ untranslated regions of target mRNA in a sequence specific manner (really?).

This work shows that structures resembling RNA granules can be reversibly aggregated and disaggregated in a soluble cellfree system in response to a small molecule (a biotinylated isoxazole ) The proteins in the granules contain low complexity sequences (LC sequences). which show little diversity in their amino acid composition (which is usually repetitive). One example is the leucine rich domain. LC sequences are all you need for aggregation by the isoxazole. The domains undergo a concentration dependent phase transition to a hydrogel-like state with no chemical present?? The hydrogels are made of uniformly polymerized amyloidlike fibers. The fibers form and dissolve and don’t cause trouble (unlike classic amyloid).

LC sequences are particularly enriched in RNA and DNA binding proteins. FUS (FUsed in Sarcoma) is an RNA binding protein containing an LC domain (Gly/Ser Tyr Gly/Ser repeats). Hydrogel droplets formed from the LC sequence of FUS can retain proteins containing either the FUS LC sequence or other LC sequences.

This work finds a potential use for LC sequences — they allow the movement of regulatory proteins into and out of organized subcellular domains, via reversible polymerization into dynamic amyloidlike fibers. It’s possible that something similar occurs in Cajal bodies, nuclear speckles and nuclear factories involved in RNA splicing.

[ Proc. Natl. Acad. Sci. vol. 99 pp. 13583 – 13588 ’02 ] They range in size from 2000 Angstroms to several microns. None of them are bounded by a membrane. It is thought that the same processes leading to the formation of nuclear bodies (e.g. a phase transition) is responsible for similar bodies occuring in the cytoplasm) — e.g. P bodies (Processing bodies), stress granules.

Each type is identified immunologically by antibodies against its components (either signature proteins or ribonucleoproteins or even small nuclear RNAs. They include
l. The Cajal body (the coiled body)
2. The promyelocytic body (PML body, POD)
3. Splicing related bodies
a. SC35 speckles (interchromatin granule cluster)
4. The GEM body
5. The matrix associated deacetylase body
6. HAP body
7. nucleoli associated paraspeckles
8. Nucleoli themselves.

The integrity of a nuclear body can be disrupted after depletion of its normal components — PODs are disrupted in acute PML.

The Cajal body and GEM are colocalized, but otherwise there doesn’t seem to be much association among the different nuclear bodies.

[ Cell vol. 162 pp. 1066 – 1077 ’15 ] FUS forms liquid compartments at sites of DNA damage in the nucleus and in the cytoplasm on stress. With time liquid droplets of FUS convert with time to an aggregated state, a conversion accelerated by mutations (in the prionlike domain) derived from patients.

Why is the compartment called liquidlike? FUS molecules rapidly rearrange within the compartment. The comaprtments formed by FUS are spherical. Two FUS compartments can fuse and relax into one sphere.

FUS compartments belong to a set of RNA protein compartments (P granules, nucleoli) which ‘probably’ form by liquid liquid demixing (phase separation) from cytoplasm.

The conversion between a liquid to a solidlike state is concentration dependent, and mutations blocking nuclear localization sequence (NLS) functgion produce increased concentrations in the cytoplasm with aggregation.

The prionlike domain of FUS is intrinsically disordered.

[ Neuron vol. 88 pp. 678 – 690 ’15 ] Mutations in a bunch of RNA binding proteins (TDP43, FUS, ataxin2, hnRNPA1, hnRNPB2) are associated with ALS/FTD (Amyotrophic Lateral Sclerosis/FrontoTemporal Dementia). Poorly soluble assemblies of the mutant RNA binding protein are found in the nucleus and cytoplasm in the patients.

The assemblies differ from amyloids in the following ways
l. They are soluble in urea
2. They have low beta sheet content
3. They have a mixed granular/fibrillar appearance on EM
4. They don’t bind dyes diagnostic for amyloid (e.g. thioflavin T)
5. When fluorescently labeled, they don’t show the reductions in in vivo fluorescent lifetimes typical of conventional amyloid.

This work shows that the LC domain (Low Complexity domain) of normal FUS undergoes phase transitions, reversibly shifting between dispersed liquid droplets and hydrogel-like phases (defined how). FUS mutants limit the ability to shift between phases, instead increasing the propensity of FUS to condense into poorly soluble stable (e.g. irreversible) fibrillar hydrogel-like assemblies (e.g. a new type of phase. Spontaneous occurrence of this might explain sporadic ALS/FTD with FUS pathology even when no mutations are present. These assemblies selectively entrap other ribonucleoproteins, impair local RNP granule function and decrease new protein synthesis in axon terminals of cultured neurons. The work was done in C. elegans.

“The biophysics of conversion from liquid droplet to reversible hydrogel is not yet clear”. Thw two differ only slightly in viscosity.

The FG repeats (phenylalanine, glycine repeats) of nucleoporins show structural characteristics typical of natively unfolded proteins (e.g. highly flexible proteins lacking ordered secondary structure). They can be quite long (200 – 700 amino acids in yeast). Protease sensitivity shows that most FG repeat containing nucleoporins are disordered in situ within the nuclear pore complexes of purified yeast nuclei. This makes it likely that they form a meshwork of random coils at the pore through which nuclear transport proceeds. Natively unfolded proteins show the following biochemical features

l. multiple domains allowing simultaneous interactions with multiple binding partners
2. nonrigid binding domains that can accomodate a variety of interacting partners
3. fast molecular association and dissociation rates.

Another model has FG domains interacting with each other in the pore to form a protein meshwork which acts as a separate hydrophobic phase. Transport complexes can partititon into this phase because they can bind to the GF repeats. Proteins unable to bind to the FG repeats are excluded from the hydrophobic phase. Molecules below 30 – 40 kiloDaltons get through the water filled holes in the gel.

To get through the pore a midsize protein must recruit a large receptor to pass through a narrow channel. The receptors replace the FG – FG binding of the nups with each other by binding to themselves — they essentially dissolve into the gel.

An alternate view holds that FG repeats form a network of unlinked polymers whose thermally activated undulations create a zone of ‘entropic exclusion’. The entropic penalty in collapsing the chains allows a barrier to form. However by binding to the repeats, carriers can circumvent the exclusion — replacing one type of bond with another.

There are several models for the FG repeats in the nuclear pore. The most convincing (to me) is the ‘selective phase’ model — a sievelike meshwork is formed within the NPC via interactions between FG repeats. The size of the FG mesh determines the upper limits of the diffusion gate (e.g. — the molecules getting through without help — in this case under 30 kiloDaltons). The binding of nuclear transport receptors (NTRs) to the FG repeats is proposed to locally dissolve the FG-FG network, allowing passage of whatever is bound to the NTRs.

‘Sufficiently concentrated’ solutions of cohesive FG domains spontaneously form FG hydrogels (which excludes inert molecules over 50 Angstroms in diameter ). Cargo NTR complexes migrate into such hydrogels ‘up to’ 20,000 times faster than the respective cargoes alone. The intragel diffusion rate of a typical importinBeta:cargo complex predicts a similar NPC passage time (10 milliSeconds) as was actually ssen in living NPCs.

The FG repeat domain of the yeast nucleoporin Nsp1 forms a hydrogel-like structure in vitro which requires hydrophobic interactions between the aromatic rings of the phenylalanines. This work assembled FG hydrogels in vitro, and studied protein entry into them and diffusion through them usingfluorescence microscopy. The influx of various nuclear transport receptors of the importin beta family into the Nsp1 FG hydrogel was 1000 times faster than the entry of a control protein. Access of a model cargo bound to importin beta was accelerated by over 20,000 fold (compared to free cargo). However, not every FG hydrogel shows selectivity. To achieve selective permeability the total FG concentration within the gel had to be raised above 50 milliMolar. This has led the authors to introduce the concept of the saturated hydrogel, in which all the FG repeats must extend completely and undergo a maximum number of interactions. It seems likely (to the authors of the editorial not the authors of the paper) that newly made FG proteins would immediately curl up and form intramolecular FG bridges (rather than intermolecular ones) In vitro gel formation can only be induced from lyophilized proteins under extreme pH and salt. The authors suggest that nuclear transport receptors act as chaperones preventing intramolecuular FG interactions after synthesis. Under more physiologic conditions, the FG domain of Nsp1 formed neither homo nor heterotypic interactions with other FG nucleoporins.

FG repeat domains (they contain a hydrophobic patch, usually FG, FxFG, or GLFG, surrounded by more hydroplic spacers) account for 12 – 20% of the mass of a nuclear pore complex. Up to 50 FG repeat domains may occur in a single protein. FG repeats occur in various flavors — examples are FxFG repeats

So there you have it — quite a mess. Figure it out and get on the boat to Sweden

Man’s best friend

I usually pay little attention to animal models of neurologic disease. After all, our brain is what separates us from animals (recent human behavior excepted). Neuromuscular disease is different because our peripheral nerves and muscles work the same way as animals. An astounding paper from Harvard and Brazil, gives us an entirely new angle to treat muscular dystrophy, particularly the Duchenne form. I ran a muscular dystrophy clinic for 15 years in the 70s and 80s and haplessly watched young boys deteriorate and die from Duchenne. The major therapeutic advance during that time was — hold your breath — lighter weight braces, allowing the boys to stay out of wheelchairs a bit longer.

Some background for those who don’t know, the molecular defect in Duchenne was found in ’87. Interestingly Kunkel, one of the authors on the original paper [ Cell vol. 51 pp.; 919 – 928 ’87 ] is an author on the present one [ Cell vol. 163 pp. 1204 – 1213 ’15 ]. Duchenne dystrophy affects only males, as the gene for the protein (dystrophin) is found on the X chromosome, so women with a normal X and a mutant X escape. To show how pathetic things were back then, we tried to find out if a sister of a patient was a carrier. How did we do it. By measuring an enzyme released by damaged muscle (CPK) on several occasion. Carriers often showed an elevation.

The mutated protein is called dystrophin. It hooks the contractile apparatus of a muscle cell to the membrane. Failure of this makes muscle cells more fragile when they contract resulting in eventual loss. From a molecular biological point of view the protein is fascinating. The gene is one of largest known, stretching over 2,220,233 positions (nucleotides) on the X chromosome and containing 79 exons. Figuring a transcription rate of 100 nucleotides a second, it takes 6 hours to make the messenger RNA (mRNA) for it. The protein has 3,685 amino acids and figuring a translation rate of 3 – 6 amino acids/second it takes 10 minutes for the ribosome to make it. Given that it takes only 3 nucleotides to code for an amino acid, the protein coding part of the gene takes up only .5% of the gene. Correctly splicing out the introns is a huge task, which we all perform well. This size and complexity of the gene explains why mutations are so common, making it the most common form of hereditary muscular dystrophy (most are).

There are currently all sorts of efforts underway to correct the mutation, particularly in a milder form called Becker dystrophy. Derek has covered them and they constitute a logical direct attack on the pathology.

What is so remarkable about the current Cell paper is that it gives us an entirely new and different way to attack Duchenne (and possible all forms of muscular dystrophy). It involves a colony of dogs in Brazil. They have GRMD (Golden Retriever Muscular Dystrophy) with a mutation in one of the many splice sites in dystrophin (it has 79 exons in man) leading to a premature stop codon and no functional dystrophin in the dogs’ muscles. The animals weaken and become non ambulatory with a shortened lifespan. However, a few of the dogs in the colony seemed pretty normal. So they went to work. The obvious reason was that gene was in some way repaired so the animals had normal amounts of dystrophin. Not so, even though ambulatory, the animals’ muscles had no dystrophin. So the whole genome was sequenced. What they found was that a mutation at an upstream site of a protein called Jagged1 lead to increased transcription of the gene and increased levels of the protein.

Jagged1 is a protein ligand for the Notch system of receptors. The Notch system is important in muscle regeneration. The myoblasts of the animals had more proliferative capacity. The Notch system is far too complicated to go into here — https://en.wikipedia.org/wiki/Notch_signaling_pathway, but expect to see a lot more research money pumped into it.

What I find so fabulous about this paper, is that it gives us an entirely new way of thinking about Duchenne, totally unrelated to the genetic defect, which had been our focus up to now. It also rubs our noses in how little we understand about our molecular biology and cell physiology. If we really understood things, we’d have been focused on Notch years ago. Yet another reason drug discovery is so hard. We are trying to alter a system we only dimly understand.

Follow

Get every new post delivered to your Inbox.

Join 92 other followers