Human cell models of Alzheimer’s disease – moving beyond the familial

Simon Lovestone discusses the failures of animal models in Alzheimer's research, and how new work on induced pluripotent stem cells, published in Alzheimer's Research & Therapy, could point the way to solving these.

In what seems like a whole era ago, there was a running joke that the Alzheimer’s disease research field was a battle between two quasi-religious groups – the Baptists and the Tauists. It wasn’t a very funny joke but it did capture a truth: that there were those who thought beta-amyloid (the Bap-tists) to be the essential component of the Alzheimer’s disease (AD) process. Meanwhile, there were others, the tau-ists, who thought that dementia essentially resulted from tangle pathology and that the overwhelming concentration of work on amyloid was misplaced.

These are of course caricatures and most recognized that both pathologies are important and part of a process. Indeed, this process had a name – the amyloid cascade hypothesis – and some of us argued that thinking hard about the cascade itself would be the way to find therapeutic targets. How exactly does the generation of beta-amyloid peptide induce tau phosphorylation and tangle formation?

A decade or so later it seems we are not so much closer to understanding the cascade element of Alzheimer’s although the number of actors in this particular drama have certainly increased. No longer ‘just’ amyloid and tau; we now have inflammation of a variety of flavors, oxidative stress, metabolic dysfunction, cholesterol biosynthesis and any number of other components, all with their advocates and many citing considerable genetic and observational data to support a role in AD.

Yet, despite the explosion of other mechanisms that might be involved, an understanding of the core or canonical process of AD seems to evade us. Why is this?

A lack of explanatory models

At least one reason, I would argue, is the lack of really explanatory models. Mice don’t seem to be of much help. In fact the most consistent observation from the very considerable number of mouse models with amyloid related pathology is that they do not show convincing evidence of an amyloid cascade. No amount of Abeta generation in rodents translates to a tau pathology resulting in neurodegeneration.

It is high time that there was a diversification both of experimental approaches and of pathways under the metaphorical, and literal, microscope.

This is remarkable in itself. Removed from the brain, rodent neurons seems to be a little more accommodating and Abeta does induce some tau phosphorylation and loss of synapses, and it certainly kills the cells. But it takes large amounts of peptide to do so and any lab that has used Abeta peptide knows just how variable this experimental paradigm can be.

With neither an animal nor a cellular model of the amyloid cascade it is no wonder that researchers have begun to turn elsewhere. Much of this is not only to be expected but a good thing too. It is high time that there was a diversification both of experimental approaches and of pathways under the metaphorical, and literal, microscope. And yet, arguably, it isn’t our understanding of mechanism but our models that are lacking.

Focusing on human neurons for better models

The paper by Ochalek and colleagues might be a sign of how this just might be about to change. There has been considerable hope that human neurons, generated from induced pluripotent stem cells (iPSCs), might be a better model of AD. If Alzheimer’s is a near-exclusively human disease as we have recently argued, then it would make sense to focus our efforts on human neurons as well as trying to understand why rodents and other animals seem so enviably resistant to AD related insults.

Recent studies showing that human neurons generate markers of pathology including both amyloid and tau related are encouraging. 3D cultures of human neurons have even been reported to show structures similar to plaques and tangles and careful generation of different neuronal types has been used to explore the intriguing conundrum of selective vulnerability in different neuronal populations.

It has to be admitted that such stem cell derived models are far from perfect. Whilst the terminally differentiated cells have a neuronal-like phenotype they represent, at best, immature neurons. Moreover, most of the studies reported to date have been on neurons carrying mutations in APP or PS-1, the genetic changes of early-onset familial AD (fAD). This extremely rare autosomal dominant condition might not be quite the same as the common late onset form of the disease that causes so much worry to politicians and the public alike.

Which is why Ochalek et al.’s paper might be so important. Here they show iPSCs derived from people with late-onset, or sporadic, disease generate increased Abeta peptides, tau phosphorylation and have increased activity of the tau kinase GSK3beta. This is remarkable. This data would suggest first that fAD is indeed a variant, albeit an aggressive variant, of ‘ordinary’ AD.

Secondly, as the driver of GSK3 activity and tau phosphorylation in the fAD cells is something to do with the over production of Abeta and as the sAD cells also overproduce Abeta and have the same GSK3 activity and tau kinase, then it follows that there is indeed an amyloid cascade and it is active in both cell types.

Thirdly, and most importantly, if the above is true then it follows that – for the first time – we now have a model in which we can dissect, and of course seek strategies to disrupt, this cascade.

Further research

This work now needs to be urgently replicated and extended. Cells from only ten individuals were studied; are these representative? Do all sAD derived neurons differ from controls and from fAD neurons? Does the protocol for iPSC generation and terminal differentiation make a difference? The protocol used here ages cells to 70 days – are there methods that speed the process and enable scaling up of experiments whilst also preserving the phenotypes observed here? Will other labs reproduce the findings with these same cell lines and with others?

As always, as many questions as answers. But nonetheless, this work is highly encouraging. Quite apart from the critical issues of replication, it would be an enormously good idea to start banking cells for iPSC generation from well characterized research participants. We recently reported on the very deep and very frequent phenotyping of a group of more than twenty individuals as part of a pilot for a much larger study, itself a part of the MRC Dementias Platform UK. Reading the work of Ochalek et al. I’m delighted that we have iPSC derived neurons slowly maturing in our lab and I’m looking forward to seeing if we can replicate some of their observations in these cells. I’m pretty sure others will be doing the same.

View the latest posts on the On Health homepage