Bench philosophy: Brain organoids

Souls in Suspension
by Steven Buckingham, Labtimes 04/2017

Organoids are produced when cells in 3D culture organise themselves into structures similar to real organs. ­Several organs have been mimicked in this way. However, when it was announced that you could pull the same trick to make not just any organ but real brains, it was not surprising that the popular press became interested. So, just how convincing are these “mini-brains” – and what can we do with them?

Doing experiments directly on organs may be the most realistic way to answer biological and physiological questions but it can be a real pain, especially when those organs are still situated where they belong: inside an animal. We all agree that experiments on animals, even on extracted tissue, are messy and full of variability, and each experiment often costs the life of an animal. No wonder the idea of culturing tissues in a dish is so attractive. Keep the tissues fed with nutrient, keep them in an incubator and any time you want to do an experiment, you just take them out, put them into the experimental chamber and off you go. In my years as a patch-clamp electrophysiologist, I have found myself all too willing to trade in the realism of working with real brains for the convenience of using a cell line.

But if the convenience of culturing individual cells in a dish is an almost irresistible pull, how much more is the attraction of growing whole tissues? And if that were not enough, how about the idea of growing whole organs in culture?

Whole organs in a dish? Really? Is that possible? Well, yes – sort of. We have known for some years now that if you start off with the right sort of cell, e.g. induced pluripotent stem cells, you can coax the cells to divide, multiply and differentiate into something remarkably like a real organ. Hans Clevers at the Hubrecht Institute, Utrecht (Netherlands) has, since 2009, been making organoids from both human and mouse tissue. In fact, to date Clevers' organ factory has churned out dishes of small intestines, colon, stomach and liver.

A team centred around Madeline Lancaster and Jürgen Knoblich recently constructed microfilament-engineered cerebral organoids, displaying “improved cortical development”. Photo: IMBA

Organ factory

Making these organoids is surprisingly easy. To take one example, let's look at Clevers' way of making a prostate gland organoid from a mouse. You take one mouse or human prostate gland and digest it with collagenase and then trypsin, to separate the cells. Seed these cells into Matrigel and overlay with a medium. Now, this medium has to be just right. In the present case, it has to have three growth factors and a pinch of testosterone. With hardly any change to the protocol, you can do exactly the same thing using tissue from human prostate. And what do you have to do to get the organoids to form? Well – not much, really. That's the point: like well-trained soldiers setting up a camp, the cells somehow just organise themselves.

Impressive. Okay, so how about making, say, a stomach? Pretty much the same. Clevers' group followed an almost identical procedure using stem cells they found near the gastric glands in the mouse stomach (Cell Stem Cell 6, 25-36). There is a catch, though – you have to get the culture medium just right. The culture media reported in Clevers' papers each contain their own cocktail of growth factors and inhibitors, and each organ requires a slightly different recipe. Who would have guessed, for instance, that to make a stomach you need to add “10 μM ROCK inhibitor Y-27632 ... to avoid anoikis”? (anoikis is a special kind of apoptosis that is induced by inadequate or inappropriate cell-matrix interactions).

If there are any cynics out there, let's be clear about one thing. What you get when you follow these protocols is not just a lump of cells that have been labelled somewhat over-generously as an organised structure. When you look at the organoids closely, they really do bear an amazing resemblance to real organs. They have the right cell types in roughly the right places.

When Clevers cultivated organoids from human small intestine, for instance, they found that all the important cell types were in there (Gastroenterology 141). Alkaline phosphatase staining showed the presence of mature enterocytes, periodic acid-Schiff staining revealed the presence of goblet cells and synaptophysin stained up enteroendocrine cells. This isn't the exception: several papers reporting the development of organoids also confirm the presence of key genetic markers that are specific to the tissues and organs they are culturing.

All key genetic markers are on board

The presence of such a high level of structure in organoids is important for many reasons, not least because it implies that as the organoid grows, it is performing the complex developmental dance that has to be played out just right for the highest level of organisation to emerge. And, if the development is normal (or at least normal enough to end up with something convincingly like the real thing), it suggests the function is probably normal too. If that is the case, it means we can use organoids as a proxy for the real organ in experiments designed to unravel the organ's physiology in health and disease.

That last point is an important one. Several labs have shown that you can take cells from sick people, turn them into pluripotent stem cells and grow organoids from them. The resulting organoids have the same genetic constitution as the sick person and you can then develop routes to therapy in extremely realistic conditions, only in a dish. Who knows, perhaps we will one day have banks of organoids ready for therapeutic implantation.

So organoids from human and animal tissues recapitulate the development of high-order structure seen in real organs. This opens up the possibility of in vitro experiments aimed at just the kind of questions that address themselves to this high level of structure. And surely, the organ where this level of structure is the most critical has to be the brain.

How does organoid-making work with something as complex as brains? Actually, the story goes back quite a few years, ever since lazy patch clampers like me opted for cell culture. For many neuronal types, a simple route is to dissociate them from real nervous tissue, settle them onto a culture dish and let them regrow. They will sprout, send out axons and behave generally as individual neurons do. But that is as far is it usually goes. The cultured neurones won't form synapses, for one thing, at least not unless you grow them along with other cell types. And even then, they don't make anything like the structures you see in the brain.

One of the earliest steps towards proper 3D culture came when Reynolds and Weiss at the Department of Pathology, University of Calgary Faculty of Medicine, Alberta, Canada showed that you could take cells from adult mouse striatum and induce proliferation with growth factors (Science 255, 1992). The resulting “neurospheres” didn't exactly look like a brain, or even a striatum come to that, but they did contain both neurons and glia, and expressed more than one type of neurotransmitter. Tantalisingly close to a genuine recapitulation of development.

And then we discovered stem cells. First, there were embryonic stem cells and the first attempts to exploit this for neuroscience showed that you could direct their development in vitro into a neuronal phenotype. Growing these neurons in a dish resulted in “neuronal rosettes”, which bear an uncanny resemblance to a sagittal view of a developing neural tube. In fact, the resemblance is more than superficial, in that the pattern of gene expression, and even to some extent the pattern of cell migration, looks remarkably like a real developing neural tube.

But the big breakthrough was yet to come: the discovery of induced pluripotency, less than ten years ago. By knocking down some key genes, it was found that you could take cells from an adult animal (or person) and push them back to a ready-to-go stem-cell state. A couple of years ago, Sergiu Pasca at Stanford University reported that you could take human induced pluripotent stem cells growing as colonies in feeders, push them towards neuronal differentiation by inhibiting the BMP and TGF-signalling pathways, and grow them in serum-free medium (Nat Methods 12, 671-78). The result was astonishing: spheroids of neurons, in which you could make out the cell types and the layering you see in real cortical tissue. The neurons were also interspersed with non-spiking astrocytes. No wonder they called them “cortical spheroids”.

Structure-forming neurons

But a cortex is not a whole brain. That is where Madeline A. Lancaster comes into the story. A few years ago, then with Jürgen Knoblich’s lab at the Austrian Academy of Sciences’ Institute of Molecular Biotechnology in Vienna, Austria, Lancaster noticed that when she was culturing neurons on a plate, some would float off and form little structures. As she played around with these odd structures, she found that they looked quite organised. In other words, she had discovered neurospheres for herself.

She reasoned that if neurons can sort themselves out like that, one thing that would get in their way would be dealing with the problems in the middle of the lump. The poor cells in the middle would have problems getting oxygen and nutrients. To see what would happen if the vascularisation hurdle were overcome, she cultured the neurons in Matrigel, overlaid with medium and simulated vascularisation by constantly moving the medium around. After a bit more tweaking of the parameters, Lancaster found that, whereas other 3D iPSC brain-growers were making parts of brains (striatum, cortex etc), her growth conditions resulted in much more complex assemblies of brain structures. “Brains in a dish,” the press cried.

There are two major routes by which this discovery can be exploited. The first is in medical research. Lancaster, now running her own lab at the MRC Laboratory of Molecular Biology, Cambridge, UK, has worked in collaboration with several labs to show that if you take cells from people with a developmental neuronal disorder, the resulting brain organoids recapitulate that disorder. The organoid has the same genetic background as the donor, so results of attempts to rescue the phenotype are much more likely to translate into a treatment that works for that person. Lancaster is now working on adapting the same approach to less tractable neuronal disorders, such as depression and schizophrenia. Schizophrenia is a good choice, since there is a well-supported view that it has a large developmental component.

Does the system actually work?

The second avenue is in basic research and the most promising application of brain organoids in this area is in developmental biology. It is difficult to determine whether the convincing architecture seen in the mini-brains is matched by the presence of equally realistic information-processing circuits. Although the building blocks are there both at the cytoarchitectural and histological levels, and although it may look like the right cells are expressing the right proteins, that does not guarantee the system actually “works”. I have a laptop at home that has all the right parts but when you press the “on” button, nothing happens. It is, however, clear that something like real neurodevelopment is actually taking place and we can investigate that with new tools such as CRISPR/Cas.

Growing brains, or any other organoid come to that, is like growing tomatoes. Get the conditions right and the brain (organoid, tomato...) will grow, we know not how. But success comes down to knowing what those conditions are. Read the original papers, and if they do happen to tell you how they hit on the right conditions (many don't), it will be perfectly clear that it is a matter of intelligent guesswork and try-it-and-see. But even when you get conditions right, organoid research is plagued with variability. Reproducibility in research is a must, it is never an option, and organoid research must overcome this bottleneck. Just like growing tomatoes, this will happen incrementally as individual problems get solved and general patterns of good practice emerge.

Last Changed: 28.08.2017