The PRISM project: pointing the way to a more precise psychiatry
The PRISM project aims to develop a quantitative biological approach to the understanding of neuropsychiatric diseases. The project focuses on schizophrenia (SZ), Alzheimer’s disease (AD), and major depression (MD), disorders that share in part common symptomatologies, including social withdrawal and certain cognitive deficits, such as attention, working memory and sensory processing.
After the success of the first phase of the project, PRISM1, the Innovative Medicine Initiative 2 (a joint undertaking between the EU and the European Federation of Pharmaceutical Industries and Associations [EFPIA]) has backed the programme with a second round of funding to explore the underlying biology of Alzheimer’s disease (AD), schizophrenia (SZ) and major depressive disorder (MDD) associated with social dysfunction. The PRISM2 (Psychiatric Ratings using Intermediate Stratified Markers 2) project aims to identify quantitative biological features common across the diseases, opening the possibility of developing targeted treatments irrespective of traditional diagnosis. ECNP also supports the project.
Dr Hugh Marston, head of Global CNS Diseases at Boehringer Ingelheim (BI), is the industry lead on the PRISM project. He discusses the industry perspective with ECNP Press Officer, Tom Parkhill.
TP: PRISM has been underway for several years, and you are now in the second period of funding, PRISM2 – tell me about it, how’s it going?
HM: We finished PRISM1, and it took a while to complete the initial analyses of the excellent, but complex, data set. We had two attempts to get additional IMI funding, which turned out to come with a little bit of a silver lining. As time went by we were able to process ever more information coming out of PRISM1, so were able to develop a progressively more robust working hypothesis which really has now become the core of PRISM2. We now feel that we can measure social dysfunction more accurately and slightly differently than has historically been the case. The robustness of this finding in PRISM1 along with the ability to generalise beyond SZ and AD to MDD are key components of the PRISM2 plan. In addition to “how” we measure social dysfunction, PRISM1 also suggests “what” may be the biological cause. Converging, and independent, lines of evidence coming from the imaging studies, the electrophysiology studies, the genetics studies, and the behavioural data all focus our attention on one brain network. The convergence of these multiple lines of evidence give me confidence that there really is something for us to drill down into in PRISM2.
Often in complex projects like PRISM you may find one or two things quickly which are interesting. You therefore go after them, but then as more analyses emerge from the complexity of the data set these then disqualify the initial findings and muddies the waters as you decide what to proceed with. In this case, all four lines of evidence converged on one system, namely the default mode network or DMN. Therefore, the core proposal we went back to the IMI with was to demonstrate the robustness of the findings of PRISM1 with respect to social dysfunction, schizophrenia and AD and the role of the DMN. We also want to further develop the digital tools that we are using to assess social dysfunction. The hope is that we can collect enough evidence to be able to continue work with the EMA, taking the necessary steps to gaining formal approval of the digital platform as a qualified biomarker – a proper clinical tool for looking at social dysfunction. So now we have a clear scientific hypothesis which allows us to probe further to see if it really is the default mode network that underlies social dysfunction, as well as a programme in which we can test if the evidence is robust and reproducible. Along with this we have a platform for which we are making good progress towards gaining formal approval as a new clinical biomarker.
The core idea behind PRISM was to look, in an unbiased way, at the transdiagnostic concepts. In other words, is there a common biology of, in this case, social dysfunction that does not align with “traditional” diagnoses, such as schizophrenia, but rather ranges across diagnoses (transdiagnostic) such as schizophrenia and AD? We could only manage AD and schizophrenia in the first iteration, but in the second iteration we now can add depression and major depression to PRISM’s scope. This will allow us to see if we can extend the findings and determine whether the role of the DMN is not just a robust finding, but whether the transdiagnostic concept can be expanded to include social dysfunction across three clinical classes rather than two.
So the output is diagnostic?
There are a number of outputs. First, to confirm whether the way we were stratifying the population based on social dysfunction is robust and reproducible. Secondly, to see if we can get enough data to really demonstrate that the passive monitoring smartphone app developed Martien Kas – the PRISM project co-ordinator – and his colleagues, the BeHapp (see footnote below), offers a step change in the ability to measure and monitor social dysfunction. If so this would enable Martien and the team to progress the process of gaining qualified biomarker status by the EMA. If this can be achieved, this will deliver a powerful new tool for both the preclinical and clinical communities to use. Third, we aim to make significant progress re-defining the phenotype, or phenotypes, of social dysfunction. There do appear to be different types of social dysfunction and quantifying and understanding these differences biologically would be a major step forward. Taken together – and this is the reason that industry is involved – this gives us the patients, the tools and a circuit and system to focus on. In other words, this gives us something biologically tangible and quantifiable to target as we develop new therapeutic concepts. Indeed, there is another important workstream in PRISM2, which is taking the human findings in PRISM1 and now back-translating these into animal studies using novel technologies that we developed during PRISM1.
When we started PRISM we didn’t, for instance, have the technology needed to look at electrophysiological changes in the default mode network in groups of behaving animals. We developed and validated these through PRISM1, so we are now able to use this new technology to see if what we found in humans can now be reproduced in socially behaving animals.
So this preclinical research is being driven by industry?
It’s a joint effort between industry and academe. Martien Kas’s lab in Groningen is working closely with our labs in Biberach on these technologies, an effective handshake between industry and academe.
Which companies are involved?
Several of the original companies have dropped out. Indeed, during PRISM1 I was project leader but representing Lilly. Boehringer Ingelheim was heavily involved in PRISM 1. Bernd Sommer, my predecessor here at BI, was both co-creator and co-project leader. For PRISM2 BI is now the lead company, and indeed we have put significantly more resources into PRISM2 than we did onto PRISM1. A number of major pharma companies expressed interest, but unfortunately none could raise sufficient resources to allow them to continue to participate in the consortium. On the EFPIA side an American company, Psychogenics, has joined and now – we are very pleased – has a stake. The big change though is Cohen Veterans Bioscience, a US-based non-profit research organisation, joining as our major new partner. CVB bring great expertise in bioanalysis, databasing and multimodal data processing. This is vital as we want to make sure that when PRISM1 and PRISM2 are finished, there remains a sustainable and useful legacy.
Unfortunately, when many public-private partnerships come to an end with no further funding, vast amounts of data and the knowledge gained just get lost or buried. What’s great about CVB is that they have a data platform called Brain Commons, and we are engineering all the PRISM1 and PRISM2 data so that it can be made available on Brain Commons. CVB will be able curate the resource, we hope, indefinitely. Thus, if PRISM stops at the end of PRISM2, a legacy will remain of resources living on and accessible to the research community. In addition, we have continuing great support from P1Vital, BioTrial and SBGNeuro, all SMEs expert in the areas of experimental CNS medicine studies, clinical electrophysiology and complex CNS data analysis.
Which universities and academic institutions are involved?
We are really pleased that the great majority of academic teams that formed the clinical and pre-clinical networks in PRISM1 have all been able to re-join for PRISM2. As a consequence, we have been able, for instance, to reactivate the clinical centres in Spain and the Netherlands quickly and effectively. It is also great that ECNP have been able to continue as part of the consortium.
At the moment you are concentrating on PRISM2. Are you in a position to say whether there might be a PRISM3, or if the project should be funded in another way?
I think the second part of your question is the more relevant. I don’t think there should necessarily be a PRISM3. I think that what we should be doing is using PRISM1 and PRISM2 as a springboard for other things. Certainly, when we realised that we wouldn’t get the same level of funding for PRISM2 as we were hoping, we started to consider how we can use PRISM2 as a co-ordinating point. Coming to the ECNP Congress in Vienna and talking to colleagues about this, there are already several studies and initiatives that we are beginning to consider that we can arrange around PRISM2 that we hope will then go far beyond PRISM2.
Looking at things from the outside, pharmaceutical companies tend to be competitive, how does that work?
I’d actually challenge you on that, Tom. I think drug companies are fundamentally co-operative at the pre-competitive level. You can often find two or three companies interested in a particular area that are keen to co-operate. Each will have its own agenda but all realise that co-operating in the pre-competitive space makes sense. Co-operating allows us to tackle bigger problems than any one company can take on alone. For example, if social dysfunction is going to become a recognised disorder, no single company, however large, could easily shift the regulatory authorities, the clinical mindset, etc., on their own. We, both industry and academe, have to do this in a collaborative way. We all realise that pooling both funding and resources and doing it together (a) reduces the risk, (b) spreads the costs, but, and most importantly, (c) has greater impact. So, at that level I think we’re very collegial, much more so than in academe on occasion. Indeed, look at the IMI: European pharma came up with 182 co-operative projects budgeted at € 5.3 billion, generating amongst other things more than 3,800 publications. Clearly, once you have a concept, and we’ve developed a drug or something similar, then we are fiercely competitive, so on that I accept your point. It’s where the IP (intellectual property) sits. There is no IP in defining social dysfunction in a new way (that’s just good for everybody), but the way you treat social dysfunction may well have IP – and yes, we are competitive on that.
Of course, this brings out roles for trans-organisational bodies such as EFPIA, like IMI. PRISM has developed out of the RDoC concept, largely developed by Tom Insel and Bruce Cuthbert at the National Institute of Mental Health. Do these industry co-operative bodies exist in the USA?
Much less so. We are very pleased that Bruce Cuthbert is still on the scientific advisory panel for PRISM. We also still keep in touch with the RDoC office at the NIMH, though I accept we still need to think about the best ways to communicate the PRISM results in the US. But, in a follow-up commentary to their original RDoC paper, which came out in Science just as PRISM1 was starting to recruit, Tom Insel and Bruce were kind enough to observe that, though the RDoC concept was the subject of much discussion in the US, Europe was ahead in getting on with the actual research. In that sense, things in Europe are well ahead of the US. The concept of getting pre-competitive, government-funding for work in the industry-academe sector is far less common in the US. By contrast, getting pre-competitive funding for studies within and between academic institutions in the US is often far better funded than in Europe. This allows far bigger academic studies to be performed. But getting industry, government and academe to come together is not so easy. That’s why Europe has been able perhaps to take a lead in some areas. But there are still challenges: in Europe, bureaucracy is a challenge while politics such as Brexit often complicates matters. Losing Britain largely from the European research ecosystem has had, and is still causing, problems. Obtaining the maximum benefit from European cross-border scientific co-operation on the big challenges such as the unmet needs of those suffering from psychiatric disorders is clearly more difficult now.
PRISM began by looking at fairly tightly defined areas of mental health, but if PRISM works, then I’d imagine we can see the methodology rolled out more generally. How would you approach this?
I need to think back to the first discussions my predecessor at Boehringer Ingelheim, Bernd Summer, and I had. What Bernd and I wanted to do was to demonstrate that we could come up with ways to drive novel treatment development in a different way. In a sense, whether we found a profound new scientific finding with PRISM or not was secondary, as long as we could demonstrate that the thought experiment could be turned into a real experiment. The thought experiment was: if we discard the current diagnostic labels and look, using quantitative biological techniques, at a group of patients with similar symptoms in a particular domain, do we confirm the existing diagnostic label or discover a new transdiagnostic concept.
If you look at much of the excellent work presented at the ECNP Congress in Vienna, it has been based on “modelling” psychiatric behaviour in animals. But to “model” something accurately you need to have a good understanding of what you are modelling. Really what the PRISM consortium is trying to illustrate is that as rats and mice don’t suffer from psychiatric disorders, humans can now be used as the legitimate biological starting point. Without studies such as PRISM we will never really know the biological detail of what is going wrong in a particular patient. Therefore, we can’t really expect to accurately back-translate and model aspects psychiatric disorders in animals. Indeed, we’ve been trying to do this for 50 years and really we have not gone very far. However, our ability to probe, with an array of innovative technologies, what is biologically happening in the human disorder is quickly shifting the balance. So with the success of PRISM, and now a few other studies, we are now maybe – and it is a ‘maybe’ – moving in the right direction.
Starting with the patient, and trying to understand what’s really happening biologically with that patient, gives us new, testable hypotheses which we can back-translate. I am afraid that for the foreseeable future we will need to continue to probe the fine detail and pharmacology in animal systems. The aim though is that the starting point can now be with studies such as PRISM telling us that, say, 75% of patients with social dysfunction have a particular gene deletion or specific neurocircuit deficit. From the human biology we thus back-translate to develop a new therapeutic.
If successful, this then brings the second main advantage. When we now want to forward-translate and go back to the patient, we will have a pre-defined “what” to measure and know in “which” patient. This is the key to what we need to do differently. At the moment we are often trying to forward-translate from a face valid animal model to a clinical world defined by the DSM-5. This DSM framework originates in spirit from the elegant descriptive clinical findings of the 1870s and 1880s. It is not though deeply seated in a clear biological understanding of the perturbations that are causal to the patient’s problems.
Completely descriptive. But we also need to remember that, to a great extent, the effective psychotherapeutics we have today have come from the principally serendipitous drug discoveries of the 1960s and 70s. Nobody really knew what an anti-depressant was pharmacologically when they first became available. Indeed, the chemistry of the first anti-depressants and anti-psychotics was developed from that of the anti-histamines. If you look at the chemical structures of chlorphenamine (antihistamine), chlorpromazine (anti-psychotic) and fluoxetine (SSRI), all have bicyclic structures with a linker and a central tail, fundamentally the same chemistry. You shift a couple of groups in the structure and you change it from being an anti-histamine to a dopamine antagonist or a serotonin reuptake inhibitor. In many cases when these molecules first went into man, we didn’t know what the drugs actually did pharmacologically. It was through careful clinical observation that certain structures were shown to be “stimulant – antidepressant” or “calming – neuroleptics” and hence of very significant clinical value. How did you determine whether you are going to give someone one drug or another if you had no biological test to guide you? A framework based on the symptoms you could observe that correlates with clinical benefit works reasonably well. If a patient fits one classification you give them an SSRI, another you give them an anti-psychotic. Being slightly provocative, has DSM evolved, at least in part, to fit the available pharmacology? If it has, that makes the life of a drug discovery scientist even more difficult, because by definition everyone sitting in an SSRI-defined bucket should respond to an SSRI. Developing a better SSRI is therefore fairly straightforward but many patients don’t respond to an SSRI. It’s these ones who don’t fit into a bucket that we are trying to treat, but we know little about them other than they are “treatment-resistant”.
As an aside, we have a BI programme where we have targeted a specific brain circuit which pretty much everybody agrees is dysfunctional in certain patients. As it becomes unbalanced, it flips patients’ emotions at an inappropriate moment. However, those with this pattern of circuit dysfunction range across a number of DSM diagnoses, but it does not occur in all patients within a diagnosis. As a consequence, we have a strong biological hypothesis but one that does not align with the current diagnostic framework. However, patients need to be recruited within the current descriptive framework. Therefore, we are having to run three phase 2 programmes in parallel in patients with similar symptoms but with different DSM diagnoses; in other words, trying to work transdiagnostically.
The framework of clinical practice is currently defined but also, I would suggest, limited by the complete reliance on these diagnoses. That’s why we have to come up with new concepts that may be “trans-diagnostic” approaches. You can see this in PRISM, because there we’re getting something which is biologically similar in both AD, schizophrenia and we hope in MDD – so that’s transdiagnostic, it doesn’t fit the “DSM” type of mind set.
This is really what we want PRISM to start doing, to say “think about your patient in a different way, think about the causal biology”. You may then be able to apply the current therapies in a different but hopefully better way. But, importantly, it also gives us new opportunities for research starting points and new avenues for treatment. So it’s changing that mindset, that’s really what we are trying to do.
The BeHapp is a passive monitoring smartphone application developed by Martien Kas, first at the University of Utrecht and now at the University of Groningen in the Netherlands. If a subject authorises the upload of the app it operates “invisibly,” logging and transmitting in an encrypted form data from the normal function of the smartphone. Over 50 parameters, including geolocation, movement, incoming and outgoing communications, attachment to networks, etc., are recorded and uploaded to a secure server. On the server the data is held in an anonymised and encrypted form such that a researcher cannot identify a specific subjects identity.