Fresh from the lab: our latest work on microglia and adult neurogenesis in schizophrenia

You can find out latest preprint, “Haploinsufficiency of the schizophrenia risk gene Cyfip1 causes abnormal postnatal hippocampal neurogenesis through a novel microglia dependent mechanism” here. This work, which I did most, but certainly not all the work for, links three very different areas of research together, areas which have not been linked before. These are adult neurogenesis, genetic risk for schizophrenia (and other psychiatric diseases), and microglia.

Most of the brain can’t make any new cells in adulthood. However, the hippocampus is an exception to this rule. Here, in a structure called the subgranular zone of the dentate gyrus sit neural stem cells, which can make new neurons throughout life in a process called adult hippocampal neurogenesis. These adult born neurons are specifically used in several functions of the hippocampus, mostly to do with memory. Interestingly, there is some evidence hippocampal neurogenesis is disrupted is psychiatric patients, including in schizophrenia, and might be involved in some of the symptoms of these diseases.

Making new neurons in the adult brain broadly requires four phases. The first is cell division. This is when the stem cells that sit in the subgranular zone divide to generate new cells that will eventually become neurons. The second phase is differentiation. This is when a cells slowly change from a stem cell to something that will become a neuron. Call it neuronal puberty. In parallel with this, the third phase starts, which is cell migration. The developing neurons move from their birthplace into the adjacent granular zone to their eventual final location. This is where the last phase takes place, which is maturation. This is when the new neuron wires into the exiting circuitry and starts functioning. In rodents, this process takes up to eight weeks.

As a lab, we use a lot of genetic models of schizophrenia. Now, I’ve written about schizophrenia models before, so I’m not going to go on about their validity. What we used here was a mouse model for a human genetic risk factor called the 15q11.2(BP1-BP2) CNV. This means carriers have a small deletion in one copy of chromosome 15. Carrying this deletion doubles your risk for schizophrenia, as well as increasing risk for autism, epilepsy, and developmental delay amongst other things. One of the genes in this deletion is called Cyfip1. This gene is interesting because it has previously been shown to affect neuronal connectivity. For this and other reasons, we used a mouse model, which lacks one of the two copies of Cyfip1. In this mouse, we looked at adult hippocampal neurogenesis.

We looked at all phases of neurogenesis in these mice. We showed that there is no effect of the mutation on the rate of cell division, or the birth of the new neurons, either in the brain or when we grow the stem cells in a dish. However, what we do see in both conditions are larger numbers of immature neurons, which are in the process of differentiating, again both in the brain and in the dish. This difference stays when you look later on when the cells have migrated to their final positions and have matured.

Neurogenesis in a dish. Stem cells isolated from mouse brain generating new neurons while we watch. This is a great model system for studying how neurogenesis is regulated

If more neurons aren’t being made, but you’re still seeing more neurons at the end, there is one obvious way to do this, that is to have more of them survive. In normal neurogenesis, well over half of all potential neurons die during the maturation process. If fewer die, you would end up with more neurons without having to make more at the start. Turns out this is exactly what is happening in our mice missing a copy of Cyfip1. Both in a dish and in the brain, there are fewer immature neurons dying.

Now, it was known that this cell death happens, but what causes it wasn’t well known yet. For this we looked at micorglia, the immune cells in the brain. These cells turn out to regulate many things in the brain, as I’ve talked about before here and here. We wondered if they could be the cells regulating the death of these new born neurons. First we looked at whether this was the case. Turns out that if you take out microglia from growing stem cells isolated from mouse brain, neurons survive better, and vice versa if you add more microglia, new neurons die more. So yes, microglia do regulated the survival of neurons at this stage.

So is this the mechanism through which loss of one copy of Cyfip1 affects neuronal survival? To start to work this out, we looked at the effects of the mutation on the functioning of microglia. Although microglia do make Cyfip1 (which nobody knew to start with), and their activation may be somewhat affected, overall the mutated microglia were pretty normal.

So are microglia not to blame? No, they are! The last set of experiments I did shows this clearly. If you grow cells in a dish, you do this in a nutrient solution called medium. While the cells are growing in this, the secret all sorts of factors into the media. Turns out that if you put medium in which microglia have grown on stem cell cultures, this can still cause the death of the new neurons (this is called a conditioned medium experiment). However, what happens when you put on medium in which Cyfip1 mutant microglia have grown on the stem cells? Nothing. Nada. Absolutely zilch.

So what do these results tell us? It tells us that in the normal situation, microglia regulate the numbers of adult born neurons by inducing cell death in some of them, which we didn’t know before. In the Cyfip1 mutant animals, microglia have somehow lost this ability, which means we get more surviving neurons than normal. (If you’re with me so far and have thought this through there could be another possibly, Cyfip1 mutant neurons can’t react to the signals from microglia. I tested this, and they can do this just fine. So it’s definitely the microglia)

Of course, there are still plenty of remaining questions. What do these excess neurons actually do in the brain? How exactly do microglia cause cell death in immature neurons? What does Cyfip1 actually do in microglia? We’re working hard on answering those questions right now, so all I can say have patience and you’ll hear about it here first.

---

Stay tuned for more posts here on Neuroscience Ramblings, and in the mean time, follow me on Twitter: @DrNielsHaan

PS: There seems to be something going on with cell migration in these animals as well, but you’ll have to wait for the next preprint to hear about that….

Anti-microglial treatments in schizophrenia – why aren’t they going anywhere?

Why do we need new treatments for schizophrenia?

A valid question. After all, we’ve had anti-psychotic drugs for decades, haloperidol (Haldol) has been around since the 60s, and second generation atypical anti-psychotics such as olanzapine (Zyprexa) and risperidone (Risperdal) came around in the 70s and 80s. Since then we’ve just been making variations on a theme, with no new classes of drug being developed so far.

The current drugs are pretty effective in treating the so-called positive symptoms (a spectacularly inappropriate term that somehow stuck) of hallucinations and delusions. They do this by blocking signalling of a neurotransmitter called dopamine, which is increased in schizophrenia.

However, they don’t do anything about the so-called negative symptoms (far more appropriate term), which include things like apathy, emotional blunting, social problems and memory issues. These things aren’t caused by dopamine. What are they caused by? Good question. We know there are differences in brain structure in schizophrenia patients compared to neurotypical people. Some people look towards the immune system to explain this.

What happening with the immune system in schizophrenia?

Let’s start at the beginning; why did people think targeting microglia, the immune cells of the brain, was a good idea? There is a long history of research into the role of the immune system in schizophrenia. We know a significant risk factor for in individual to develop schizophrenia is for their mother to contract an infection during pregnancy. The thinking is that the immune response by the mother affects brain development of the foetus.

Other risk factors include severe and chronic stress during childhood, for instance being exposed to abuse, famine or warfare. This chronic stress also leads to immune system activation. Lastly, we’ve now characterised well over 150 genetic risk factors which increase the risk for schizophrenia. Many of these are also involved in the immune response.

This isn’t just theorising; a common finding in the blood or cerebrospinal fluid of schizophrenia sufferers is elevated levels of immune factors, suggesting their immune system is somehow over-active.

Over-active microglia in schizophrenia?

Ok, so looks like there is involvement of the immune system in schizophrenia. As I described here, the immune system in the brain is different from the rest of the body, the brain has its own special immune cells, the microglia. Obviously, the next logical step would be to look at if microglia in schizophrenia patients are more active. People have done this using brain scans with a (mildy) radioactive probe for microglia and found that, indeed, this is the case (see for instance here).

Overzealous microglia are bad news, as I described in the second section here. It is a reasonable assumption that, if there is excess microglia activity, this leads to loss of neuronal connections and altered neuronal functioning, and that this could underlie some of the symptoms. So, the obvious next step is to calm down the microglia.

How can we hit microglia?

Ah, at least here there is a simple answer: minocycline. This was developed as an antibiotic, but along the line someone clever noticed it also inhibits the activity of microglia.

Minocycline, not just an antibiotic

Actually, minocyline is quite a ‘dirty’ drug, it is not very specific. It does other things in the brain as well, but for the purposes of the things we’re looking at today, the microglia effects are the most important.

Well? Does it work?

Between 2010 and 2015 the initial studies on minocyline in schizophrenia were reported. The tally? Four showing beneficial effects (1, 2, 3 and 4), two showing no change (1 and 2) in schizophrenia symptoms.

A recent meta-analysis (collating several of the earlier small studies) found small positive effects that were borderline significant. Promising, but nothing to write home about. These early trials were followed by larger trials trying to confirm these effects. These are now reporting their findings.

The results of the latest mincocycline trial (BeneMin, one of the largest yet) were presented by Bill Deakin at last month’s British Association for Psychopharmacology Summer Meeting, see my tweet with some of his results here. Long story short, absolutely nothing happened. No effects whatsoever.

This was in direct contradiction to one of the earlier positive studies, led by Bill Deakin as well. So, this leaves us with a net score of 3 vs 3, with the three “winners” generally being close calls.

So why isn’t minocycline the silver bullet for schizophrenia?

Pick your poison:

• We’re not treating long enough
• We’re not treating early enough
• We’re not using the right drug
• We aren’t treating the right patients
• All microglia are not the same
• Microglia really don’t have anything to do with the disease
• All of the above (ie we have no idea what we’re doing)

The first two options are the most charitable interpretation of recent data. The reshaping of neuronal circuits and the other jobs they do can be slow processes, and it may take a long time for the drug to have enough of an effect in the brain to see a difference in symptoms.We would just need to treat for longer. It may also be that by the time symptoms are severe enough to need treatment, there are so many changes in the brain that just inhibiting microglia isn’t enough. You could think of treating people at high risk of developing the disease.

Are we using the right drug? Difficult question. Minocycline is just one option. People have tried other, more general, anti-inflammatory drugs such as NSAIDS, or immune suppressors like methotrexate, These drugs don’t target the microglia themselves, but the inflammatory process that activates them. So far, these trials have shown at most modest effects or less.

It may be that there are subtypes of schizophrenia and treating everyone in this way isn’t helpful. This concept is already being used in depression trials, where a distinction is being made between “inflammatory” and “non-inflammatory” patients. It might very well be the case that stratifying patients and targeting treatment that way could work in schizophrenia as well. I think this is likely (and also shows we still really don’t understand what is going on).

Maybe not all microglia are created equal. It’s known that there is a lot of diversity of microglia, in different brain regions, even between cells in the same region. Some may be more activated, and some may even be beneficial. Just shutting all of them down may be like taking a sledgehammer to a mosquito, and we should be targeting specific populations of cells. How that could be done is left as an exercise to the reader….

It could also be we’re barking up completely the wrong tree. Microglia are a logical and tempting target, given available data. But what about that data? For instance, the brain scans showing over-activation of microglia? Turns out the radioactive tracer used isn’t actually specific for microglia at all, and measures who knows what else as well. So are microglia actually more activated in schizophrenia sufferers? We don’t know for sure. The possibility still exists that microglia play no role at all, though I doubt that.

To be honest, the last option (“we have no idea what we’re doing”), is probably closest to the truth at the moment….

Where to go from here?

Well, walk before we run, I’d say. Scientists are people too, and we get excited as well. I think that when the data on raised inflammatory factors, and later the scans supposedly showing microglial over-activation, came through, people made the logical mental leap to start hitting microglia, without knowing what is actually going on (surprisingly common in trials, actually).

I think the (circumstantial) evidence that microglia are involved in schizophrenia is there, but we need to be cleverer in targeting them. Just hitting all of them clearly isn’t the answer.  What do we need to hit and how to do it? More research required! (Get back to me in five years time…..)

---

Stay tuned for more posts here on Neuroscience Ramblings, and in the mean time, follow me on Twitter: @DrNielsHaan

“You can’t ask a mouse if he hears squeaks” – on modelling schizophrenia

When people ask me what I work on, I frequently say “schizophrenia”. That’s a lie, really. I might be more accurate and say “I work on immune effects in animal models of schizophrenia”. Still a lie. Not the “immune effects” or “work” parts – though some colleagues might disagree with the latter – but the lie is the “schizophrenia models” part. That is because we actually don’t really have a model of schizophrenia.

Schizophrenia is very complex, both in how it develops and in its symptoms. Things that put people at risk for schizophrenia include genetic factors, environmental influences during gestation, childhood, and adulthood, as well as things such as drug usage. All these things do is increase your risk for developing the disease, not causing it.

Symptoms vary from patient to patient, but generally include ‘positive symptoms’, such as hallucinations and delusions, and ‘negative symptoms’, problems with memory and personal interactions. Although the positive symptoms can often be treated fairly well with drugs, the negative symptoms are very hard to treat.

There are two big categories of “schizophrenia models” – those that model genetics and those that model environmental factors. Genetic models are easiest, in many ways. We now know of almost 150 genes that can increase your risk for schizophrenia. For many of these, mouse or rat models exist. However, these genes are risk factors. It does not mean that having these gene variants always means you develop the disease. So what we’re doing in rodent models of these risk genes is modelling processes that might contribute to risk of schizophrenia. This is still very informative, as it tells us what these genes do, and what the precise role of them may be in disease development. This is one of the things I do.

The other main group of models is environmental. Like risk genes, there are also risk factors in the environment. These are things like malnutrition or infections of the mother during pregnancy, childhood abuse and neglect, and use of some recreational drugs. Again, we can model these in rodents. In our group, we use a model of pubertal stress in rats, which produces negative symptom-like behaviour once the rats have grown up.

Of course, there’s nothing stopping you from modelling multiple risk factors, such as juvenile stress or drug challenges in genetic models, and we are doing these sorts of experiments as well.

Do rodents actually get schizophrenia? Well, there’s the biggest problem. How would you tell? As the title says, you can’t ask a mouse if he is hearing squeaks ie is hallucinating. In other words, you can’t test if they have many classical schizophrenia symptoms. You can test many other symptoms though. Memory testing is easy in rodents, as is testing of social behaviour. We can also look at aspects such as brain structure and gene expression, and see how these match up with schizophrenia patients.

So why do I say we don’t have a schizophrenia model? Because a) none of the models reproduce all symptoms of the disease and b) we can’t even test if they do. Does this make our models useless? Not in the slightest! No rodent model ever reproduces the human condition perfectly, but an imperfect model is still very much better than no model at all. Our models can teach us a great deal.

We just need to be careful with the questions we ask. For instance, in a genetic model, the question isn’t “How does this gene contribute to schizophrenia?”, as we can’t test that.  What we’re asking is “What does this gene do in brain functioning, and what can that tell us about what it might do in schizophrenia?”.

Of course, the million dollar question is why you still read about schizophrenia models? The answers are simple, money and attention. “Schizophrenia model” sounds far more interesting and straightforward than “Model of schizophrenia risk factors”.  Which gets more attention, “Alterations in X, Y, and Z in offspring of virally infected mothers” or “Alterations in X, Y, and Z in a maternal infection model of schizophrenia”? Science (or indeed me) is not immune to the wonders of spin!

Stay tuned for more posts here on Neuroscience Ramblings, and in the mean time, follow me on Twitter: @DrNielsHaan