Most cells in your adult body are “terminally differentiated” – meaning that they have developed from stem cells into the final liver, or heart, or muscle or endothelial cell that they were meant to be.  From that point onward, cells are able to “remember” to stay in this final state – in part – via stable patterns of DNA methylation that reinforce the regulation of “the end state” of gene expression for that cell.  As evidence for this role of DNA methylation, it has been observed that levels of DNA methyl transferase (DNMT) decline when cells are fully differentiated and thus, cannot modify or disrupt their patterns of methylation.

NOT the case in the brain! Even though neurons in the adult brain are fully differentiated, levels of methyl transferases – DO NOT decline.  Why not? Afterall, we wouldn’t want our neurons to turn into liver cells, or big toe cells, would we?

One hypothesis, suggested by David Sweatt and colleagues is that neurons have more important things to “remember”.   They suggest in their fee and open research article, “Evidence That DNA (Cytosine-5) Methyltransferase Regulates Synaptic Plasticity in the Hippocampus” [doi: 10.1074/jbc.M511767200] that:

DNA methylation could have lasting effects on neuronal gene expression and overall functional state. We hypothesize that direct modification of DNA, in the form of DNA (cytosine-5) methylation, is another epigenetic mechanism for long term information storage in the nervous system.

By measuring methylated vs. unmethylated DNA in the promoter of the reelin and BDNF genes and relating this to electrophysiological measures of synaptic plasticity, the research team finds correlations between methylation status and synaptic plasticity.  More specifically, they find that zebularine (an inhibitor of DNMT) CAN block long-term potentiation (LTP), but NOT block baseline synaptic transmission nor the ability of synapses to fire in a theta-burst pattern (needed to induce LTP).

This suggests that the epigenetic machinery used for DNA methylation may have a role in the formation of cellular memory – but not in the same sense as in other cells in the body – where cells remember to remain in a terminally differentiated state.

In the brain, this epigenetic machinery may help cells remember stuff that’s more germane to brain function … you know … our memories and stuff.

just a pointer to: Genetic Future’s pointer to the recent article, “Family become first to have DNA sequenced for non-medical reasons“.    The father suggests, “it will be ethically improper if you don’t have your children sequenced“.

Early days.

It sounds simple on paper – just collect lots of folks with disorder X and look at their genomes in reference to a demographically matched healthy control population.  Voila! whatever is different is a candidate for genetic risk.  Apparently, not so.

The missing heritability problem that clouds the birth of the personal genomes era refers to the baffling inability to find enough common genetic variants that can account for the genetic risk of an illness or disorder.

There are any number of reasons for this … (i) even as any given MZ and DZ twin pair shares genetic variants that predispose them toward the similar brains and mental states, it may be the case that different MZ and DZ pairs have different types of rare genetic variation thus diluting out any similar patterns of variation when large pools of cases and controls are compared …  (ii) also, the way that the environment interacts with common risk-promoting genetic variation may be quite different from person to person – making it hard to find variation that is similarly risk-promoting in large pools of cases and controls … and many others I’m sure.

One research group recently asked whether the type of common genetic variation(SNP vs. CNV) might inform the search for the missing heritability.  The authors of the recent paper, “Genome-wide association study of CNVs in 16,000 cases of eight common diseases and 3,000 shared controls” [doi:10.1038/nature08979] looked at an alternative to the usual SNP markers – so called common copy number variants (CNVs) – and asked if these markers might provide a stronger accounting for genetic risk.  While a number of previous papers in the mental health field have indeed shown associations with CNVs, this massive study (some 3,432 CNV probes in 2000 or so cases and 3000 controls) did not reveal an association with bipolar disorder.  Furthermore, the team reports that common CNV variants are already in fairly strong linkage disequilibrium with common SNPs and so perhaps may not have reached any farther into the abyss of rare genetic variation than previous GWAS studies.

Disappointing perhaps, but a big step forward nonetheless!  What will the personal genomes era look like if we all have different forms of rare genetic variation?

If you’ve started to notice the arrival of spring blossoms, you may have wondered, “how do the blossoms know when its spring?“  Well, it turns out that its not the temperature, but rather, that plants sense the length of the day-light cycle in order to synchronize their  own life cycles with the seasons.  According to the photoperiodism entry for wikipedia, “Many flowering plants use a photoreceptor protein, such as phytochrome or cryptochrome, to sense seasonal changes in night length, or photoperiod, which they take as signals to flower.”

It turns out that humans are much the same. Say wha?!

Yep, as the long ago descendants of single cells who had to eek out a living during day (when the sun emits mutagenic UV radiation) and night cycles, our very own basic molecular machinery that regulates the transcription, translation, replication and a host of other cellular functions is remarkably sensitive – entrained – in a clock-like fashion to the rising and setting sun.  This is because, in our retinas, there are light-sensing cells that send signals to the suprachiasmatic nucleus (SCN) which then – via the pineal gland – secretes systemic hormones such as melatonin that help synchronize cells and organs in your brain and body.  When this process is disrupted, folks can feel downright lousy, as seen in seasonal affective disorder (SAD), delayed sleep phase syndrome (DSPS) and other circadian rhythm disorders.

If you’re skeptical, consider the effects of genetic variation in genes that regulate our circadian rhythms, often called “clock” genes – very ancient genes that keep our cellular clocks synchronized with each other and the outside environment.  Soria et al., have a great paper entitled, “Differential Association of Circadian Genes with Mood Disorders: CRY1 and NPAS2 are Associated with Unipolar Major Depression and CLOCK and VIP with Bipolar Disorder” [doi: 10.1038/npp.2009.230] wherein they reveal that normal variation in these clock genes is associated with mood regulation.

A few of the highlights reported are rs2287161 in the CRY1 gene,  rs11123857 in the NPAS2 gene, and rs885861 in the VIPR2 gene – where the C-allele, G-allele and C-allele, respectively, were associated with mood disorders.

I’m not sure how one would best interpret genetic variation of such circadian rhythm genes.  Perhaps they index how much a person’s mood could be influenced by changes or disruptions to the normal rhythm??  Not sure.  My 23andMe data shows the non-risk AA genotype for rs11123857 (the others are not covered by 23andMe).

Obviously, Wolfe was a gifted artist, despite whatever psychiatric diagnosis was suggested at the time.  Nevertheless, clinical psychiatrists might be quick to point out that such work reflects the presence of an underlying thought disorder (loss of abstraction ability, tangentiality, loose associations, derailment, thought blocking, overinclusive thinking, etc., etc.) – despite the undeniable aesthetic beauty in the work.  As an ardent fan of such art,  it made me wonder just how “well ordered” my own thoughts might be.  Given to being rather forgetful and distractable, I suspect my thinking process is just sufficiently well ordered to perform the routine tasks of day-to-day living, but perhaps not a whole lot more so.  Is this bad or good?  Who knows.

However, Krug et al., in their recent paper, “The effect of Neuregulin 1 on neural correlates of episodic memory encoding and retrieval” [doi:10.1016/j.neuroimage.2009.12.062] do note that the brains of unaffected relatives of persons with mental illness show subtle differences in various patterns of activation.  It seems that when individuals are using their brains to encode information for memory storage, unaffected relatives show greater activation in areas of the frontal cortex compared to unrelated subjects.  This so-called encoding process during episodic memory is very important for a healthy memory system and its dysfunction is correlated with thought disorders and other aspects of cognitive dysfunction.  Krug et al., proceed to explore this encoding process further and ask if a well-known schizophrenia risk variant (rs35753505 C vs. T) in the neuregulin-1 gene might underlie this phenomenon.  To do this, they asked 34 TT, 32 TC and 28 CC individuals to perform a memory (of faces) game whilst laying in an MRI scanner.

The team reports that there were indeed differences in brain activity during both the encoding (storage) and retrieval (recall) portions of the task – that were both correlated with genotype – and also in which the CC risk genotype was correlated with more (hyper-) activation.  Some of the brain areas that were hyperactivated during encoding and associated with CC genotype were the left middle frontal gyrus (BA 9), the bilateral fusiform gyrus and the left middle occipital gyrus (BA 19).  The left middle occipital gyrus showed gene associated-hyperactivation during recall.  So it seems, that healthy individuals can carry risk for mental illness and that their brains may actually function slightly differently.

As an ardent fan of Art Brut, I confess I hoped I would carry the CC genotype, but alas, my 23andme profile shows a boring TT genotype.  No wonder my artwork sucks.  More on NRG1 here.

Here’s a great paper, “Association between ADORA2A and DRD2 Polymorphisms and Caffeine-Induced Anxiety” [doi: 10.1038/npp.2008.17] wherein polymorphisms in the adenosine A2A receptor (ADORA2A encodes the protein that caffeine binds to and antagonizes) – as well as the dopamine D2 receptor (DRD2 encodes a protein whose downstream signals are normally counteracted by A2A receptors) — show associations with anxiety after the consumption of 150mg of caffeine (about an average cup of coffee – much less than the super-size, super-rich cups that Starbucks sells).  The variants, rs5751876 (T-allele), rs2298383 (T-allele) and rs4822492 (G-allele) from the ADORA2A gene as well as rs1110976 (-/G genotype) from the DRD2 gene showed significant increases in anxiety in a test population of 102 otherwise-healthy light-moderate regular coffee drinkers.

My own 23andMe data only provides a drop of information suggesting I’m protected from the anxiety-promoting effects.  Nevertheless, I’ll avoid the super-sizes.
rs5751876 (T-allele)  C/C – less anxiety
rs2298383 (T-allele) – not covered
rs4822492 (G-allele) – not covered
rs1110976 (-/G genotype) – not covered

DON’T tell the grant funding agencies

Let’s say we ask a population of folks to perform a task – perhaps a word memory task – and then we use neuroimaging to identify the areas of the brain that (i) were associated with performance of the task, and (ii) were not only associated with performance, but were also associated with genetic variation in the population.  Indeed, there are already examples of just this type of “imaging-genetic” study in the literature.  Such studies form a crucial translational link in understanding how genes (whose biochemical functions are most often studied in animal models) relate to human brain function (usually studied with cognitive psychology). However, do these genes relate to just this task? What if subjects were recalling objects? or feelings?  What if subjects were recalling objects / experiences / feelings / etc. from their childhoods?  Of course, there are thousands of common cognitive operations one’s brain routinely performs, and, hence, thousands of experimental paradigms that could be used in such “imaging-genetic” gene association studies.  At more than $500/hour (some paradigms last up to 2 hours) in imaging costs, the translational genes-to-cognition endeavor could get expensive!

DO tell the grant funding agencies that this may not be a problem any longer.

The recent paper by Liu and colleagues “Prefrontal-Related Functional Connectivities within the Default Network Are Modulated by COMT val158met in Healthy Young Adults” [doi: 10.1523/jneurosci.3941-09.2010] suggests an approach that may simplify matters.  Their approach still involves genotyping (in this case for rs4680) and neuroimaging.  However, instead of performing a specific cognitive task, the team asks subjects to lay in the scanner – and do nothing.  That’s right – nothing – just lay still with eyes closed and just let the mind wander and not to think about anything in particular – for a mere 10 minutes.  Hunh?  What the heck can you learn from that?

It turns out that one can learn a lot.  This is because the neural pathways that the brain uses when you are actively doing something (a word recall task) are largely intact even when you are doing nothing.  Your brain does not “turn off” when you are laying still with your eyes closed and drifting in thought.  Rather, your brain slips into a kind of default pattern, described in studies of  “default networks” or “resting-state networks” where wide-ranging brain circuits remain dynamically coupled and actively exchange neural information.  One really great paper that describes these networks is a free-and-open article by Hagmann et al., “Mapping the Structural Core of Human Cerebral Cortex” [doi: 10.1371/journal.pbio.0060159] from which I’ve lifted their Figure 1 above.  The work by Hagmann et al., and others show that the brain has a sort of “connectome” where there are thousands of “connector hubs” or nodes that remain actively coupled (meaning that if one node fires, the other node will fire in a synchronized way) when the brain is at rest and when the brain is actively performing cognitive operations.  In a few studies, it seems that the strength of functional coupling in certain brain areas at rest is correlated (positively and negatively) with the activation of these areas when subjects are performing a specific task.

In the genetic study reported by Liu and colleagues, they found that genotype (N=57) at the dopaminergic COMT gene correlated with differences in the functional connectivity (synchronization of firing) of nodes in the prefrontal cortex.  This result is eerily similar to results found for a number of specific tasks (N-back, Wisconsin Card Sorting, Gambling, etc.) where COMT genotype was correlated with the differential activation of the frontal cortex during the task.  So it seems that one imaging paradigm (lay still and rest for 10 minutes) provided comparable insights to several lengthy (and diverse) activation tasks.  Perhaps this is the case. If so, might it provide a more direct route to linking genetic variation with cognitive function?

Liu and colleagues do not comment on this proposition directly nor do they seem to be over-interpreting their results in they way I have editorialized things here.  They very thoughtfully point out the ways in which the networks they’ve identified and similar and different to the published findings of others.  Certainly, this study and the other one like it are the first in what might be a promising new direction!

I’m wondering how much physical and/or social stress is enough to cause changes in the epigenome?  Does the concern about epigenetics only apply to exposure to severe stress?  or run of the mill forms of stress?  How much do we know about this?

This year, I hope to explore this line of inquiry further.  For starters, I came across a fantastic paper by Fraga et al., entitled, “Epigenetic differences arise during the lifetime of monozygotic twins” [doi:10.1073/pnas.0500398102].   The group carries out a remarkably straightforward and time honored approach – a twin study – to ask how much identical twins differ at the epigenetic level.  Since identical twins have the same genome sequence, any differences in their physiology, behavior etc. are, strictly speaking, due to the way in which the environment (from the uterus to adulthood) shapes their development.  Hence, the team of Fraga et al., can compare the amount and location of methyl (CH3) and acetyl (OCCH3) groups to see whether the environment has differentially shaped the epigenome.

An analysis of some 40 identical twin pairs from ages 3-74 years old showed that – YES – the environment, over time, does seem to shape the epigenome (in this case of lymphocytes).  The most compelling evidence for me was seen in Figure 4 where the team used a method known as Restriction Landmark Genomic Scanning (RLGS) to compare patterns of methylation in a genome-wide manner.  Using this analysis, the team found that older twin pairs had about 2.5 times as many differences as did the epigenomes of the youngest twin pairs.  These methylation differences also correlated with gene expression differences (older pairs also had more gene expression differences) and they found that the individual who showed the lowest levels of methylation also had the highest levels of gene expression.  Furthermore, the team finds that twin pairs who lived apart and had more differences in life history were more likely to have epigenetic differences.  Finally, measures of histone acetylation seemed consistent with the gradient of epigenetic change over time and life-history distance.

Thus it seems that, as everyday life progresses, the epigenome changes too.  So, perhaps, one does not need extreme forms of stress to leave long-lasting epigenetic marks on the genome?  Is this true during early life (where the team did not see many differences between pairs)?  and in the brain (the team focused mainly on lymphocytes)?  Are the differences between twins due to the creation of new environmentally-mediated marks or the faulty passage of existing marks from dividing cell-to-cell over time?  Will be fun to seek out information on this.

In their forecast “The World in 2010” special issue, the Economist points to “The looming crisis in human genetics” wherein scientists will reluctantly acknowledge that, even with super-cheap genome sequencing tools, we may not soon understand how genetic variation contributes to complex illness.  The argument is a valid one to be sure, but only time will tell.

A paper I read recently, reminded me of the long hard slog ahead in the area of genomics and psychiatric illness.  The authors in “Association of the Glutamate Transporter Gene SLC1A1 With Atypical Antipsychotics–Induced Obsessive-compulsive Symptoms” [Kwon et al., (2009) Arch Gen Psychiatry 66(11)] are trying to do something very important.  They would like to understand why certain (most) psychiatric medications have adverse side-effects and how to steer patients clear of adverse side-effects.  This is because, nowadays, a patient learns via a drawn-out trial-and-error ordeal about which medications he/she can manage the benefits/costs.

Specifically, the authors focused their efforts on so-called obsessive-compulsive symptoms that can arise from treatment with atypical antipsychotic medications.  Working from 3 major medical centers (Samsung Medical Center, Seoul National University Hospital and Asan Medical Center) Kwon et al., were able to cobble together a mere 40 patients who display these particular adverse side-effects and matched them with 54 patients based on several demographic and medication-based criteria.  Keep in mind that most genetic studies use upwards of 1,000 samples and still – hardly – are able to obtain significant effects.

Nevertheless, the authors note that the glutamate transporter gene (SLC1A1 or EAAC1) is a most logical candidate gene, being a located in a region mapped for obsessive-compulsive disorder risk and also a gene that appears to be down-regulated in response to atypical anti-psychotic treatment (particularly clozapine).  A series of statistical association tests for 10 SNPs in this gene reveal that two SNPs (rs2228622 and rs3780412) and a 3-SNP haplotype (the A/C/G haplotype at rs2228622-rs3780413-rs3780412) showed modestly significant association (about 4-fold higher risk) with the adverse symptoms.

To me, this is a very noteworthy finding.  A lot of work went into a very important problem – perhaps THE most pressing problem for patients on anti-psychotic medications today – and the results, while only of modest significance, are probably biologically valid.  The authors point out that rs2228622 and rs3780412 have previously been associated with OCD in other studies.

But when you compare these modest results (that these authors fought hard to obtain) with the big promises of the genomic era (as noted in the Economist article), well then, the results seem rather diminutive.  Will all patients who carry the risk haplotype be steered away from atypical antipsychotics?  Will big pharma (the authors of this paper disclose a great many ties to big pharma) support the fragmentation of their blockbuster drug markets into a hundred sub-populations?  I doubt it.  But some doctors and patients will experiment and continue to explore this avenue of inquiry – and it will take a long time to work out.  Better check back in 2020.

What the heck is an “in”sensitive period then?   Well, whereas visual input is clearly a “good thing” for the sensitive period of visual development, perhaps some inputs are “bad” and it may be useful to shield or protect the brain from exposure.  Maybe some environmental inputs are “bad” and one would not want the developing brain to be exposed to them and say, “OK, this (bad stuff) is normal“.  As a parent, I am constantly telling my children that the traffic-filled street is a “bad place” and, like all parents, I would not want my children to think that it was OK to wander into the street.  Clearly, I want my child to recognize the car-filled street as a “bad thing”.

In the developing brain, it turns out that there are some “bad things” that one would NOT like (the brain) to get accustomed to.  Long-term exposure to glucocorticoids is one example – well-known to cause a type of neuronal remodelling in the hippocampus, that is associated with poor cognitive performance (visit Bruce McEwen’s lab at Rockefeller University to learn more about this).  Perhaps an “in”sensitive period – where the brain is insensitive to glucocorticoids – is one way to teach the brain that glucocorticoids are “bad” and DO NOT get too familiar with them (such a period does actually occur during early post-natal mammalian development).  Of course, we do need our brains to mount an acute stress response, if and when, we are being threatened, but it is also very important that the brain learn to TURN-OFF the acute stress response when the threat has passed – an extensive literature on the deleterious effects of chronic exposure to stress bears this out.  Hence, the brain needs to learn to recognize the flow of glucocorticoids as something that needs to be shut down.

OK, so our developing brain needs to learn what/who is “good vs. bad”.  Perhaps sensitive and insensitive periods help to reinforce this learning – and also – to cement learning into the system in a sort of permanent way (I’m really not sure if this is the consensus view, but I’ll try and podcast interview some of the experts here asap).  In any case, in the case of the visual system, it is clear that the lack of visual input during the sensitive period has long lasting consequences.  In the case of the stress response, it is also clear that if there is untoward stress early in development, one can be (somewhat) destined to endure a lifetime of emotional difficulty.  Previous posts here, here, here cover research on behavioral/genomic correlates of early life stress.

Genes meet environment in the epigenome during sensitive and insensitive periods?

As stated at the outset – genes are not destiny.  The DNA cannot encode a system that knows who/what is good vs. bad, but rather can only encode a system of molecular parts that can assemble to learn these contingencies on the fly.  During sensitive periods in the visual system, cells in the visual system are more active and fire more profusely during the sensitive period. This extra firing leads to changes in gene expression in ways that (somewhat) permanently set the connectivity, strength and sensitivity of visual synapses.  The expression of neuroligins, neurexins, integrins and all manner of extracellular proteins that stabilize synaptic connections are well-known tagets of activity-induced gene expression.  Hence the environment “interacts” with the genome via neuronal firing which induces gene expression which – in turn – feeds back and modulates neuronal firing.  Environment –> neuronal firing –> gene expression –> modified neuronal firing.  OK.

Similarly, in the stress response system, the environment induces changes in the firing of cells in the hypothalamus which leads (through a series of intermediates) to the release of glucocorticoids.  Genes induced during the firing of hypothalamic cells and by the release of glucocorticoid can modify the organism’s subsequent response to stressful events.  Environment –> neuronal firing –> gene expression –> modified neuronal firing.  OK.

Digging deeper into the mechanism by which neuronal firing induces gene expression, we find an interesting twist.   Certainly there is a well-studied mechanism wherein neuronal firing causes Ca++ release which activates gene expression of neuroligins, neurexins, integrins and all manner of extracellular proteins that stabilize synaptic connections – for many decades.  There is another mechanism that can permanently mark certain genes and alter their levels of expression – in a long-lasting manner.  These are so-called epigenetic mechanisms such as DNA methylation and acetylation.  As covered here and here, for instance, Michael Meaney’s lab has shown that DNA CpG methylation of various genes can vary in response to early-life stress and/or maternal care. In some cases, females who were poorly cared for, may, in turn, be rather lousy mothers themselves as a consequence of these epigenetic markings.

A new research article, “Dynamic DNA methylation programs persistent adverse effects of early-life stress” by Chris Murgatroyd and colleagues [doi:10.1038/nn.2436] explores these mechanisms in great detail.  The team explored the expression of the arginine vasopressin (AVP) peptide – a gene which is important for healthy social interaction and social-stress responsivity.  Among many other interesting results, the team reports that early life stress (using a mouse model) leads to lower levels of methylation in the 3rd CpG island which is located downstream in a distal gene-expression-enhancer region.  In short, more early-life stress was correlated with less methylation, more AVP expression which is known to potentiate the release of glucocorticoids (a bad thing).   The team reports that the methyl binding MeCP2 protein, encoded by the gene that underlies Rett syndrome, acts as a repressor of AVP expression – which would normally be a good thing since it would keep AVP levels (and hence glucocorticoid levels) down.  But unfortunately, early-life stress removes the very methyl groups to which MeCP2 binds and also the team reports that parvocelluar neuronal depolarization leads to phosphorylation (on serine residue #438) of MeCP2 – a form of MeCP2 that is less accessible to its targets.  So, in  a manner similar to other examples, early life stress can have long-lasting effects on gene expression via an epigenetic mechanism – and disables an otherwise protective mechanism that would shield the organism from the effects of stress.  Much like in the case of Rett syndrome (as covered here) it seems that when MeCP2 is bound – then it silences gene expression – which would seem to be a good thing when it comes to the case of AVP.

So who puts these epigenetic marks on chromosomes and why?

I’ll try and explore this further in the weeks ahead.  One intriguing idea about why methylation has been co-opted among mammals, has to do with the idea of parent-offspring conflict.  According to David Haig, one of the experts on this topic, males have various incentives to cause their offspring to be large and fast growing, while females have incentive to combat the genomic tricks that males use, and to keep their offspring smaller and more manageable in size.  The literature clearly show that genes that are marked or methylated by fathers (paternally imprinted genes) tend to be growth promoting genes and that maternally imprinted genes tend to be growth inhibitors.  One might imagine that maternally methylated genes might have an impact on maternal care as well.

Lastly, the growth promoting/inhibiting effects of paternal/maternal genes and gene markings is now starting to be discussed somewhat in the context of autism/schizophrenia which have have been associated with synaptic under-/over-growth, respectively.

Building a brain is already tough enough – but to have to do it amidst an eons-old battle between maternal and paternal genomes.  Sheesh!  More on this to come.