Sonia, I, and our collaborators at Massachusetts General Hospital (MGH) are proud to announce that a paper describing the first results from our clinical research study of biomarkers in people at risk for genetic prion disease is now available online: Cerebrospinal fluid and plasma biomarkers in individuals at risk for genetic prion disease. In this post, I will introduce the motivation and goals of the study, what we did, and what we learned.
Sonia and I would like to prevent her disease before it ever starts. That means we need a paradigm in which to test a potential drug in pre-symptomatic people like Sonia: people who harbor a prion protein gene (PRNP) mutation that puts them at high risk for genetic prion disease. Of course, we would love to see the same drug prove effective both at preventing prion disease, and at treating it after symptoms strike. But we can’t rely on clinical trials in symptomatic patients as our only hope for getting a drug approved to prevent prion disease. One reason is that plenty of interventions that effectively prevent a disease only work preventively (imagine taking cholesterol-lowering drugs at the moment of a heart attack). Another reason is that many, perhaps all, drugs that have proven effective in mouse models of prion disease work better the earlier they are given. In other words, there is a risk that a drug that could save Sonia’s life, and the lives of many others in her position, might not prove effective in symptomatic prion disease patients. If we count on clinical trials in sick patients as our only drug development path, we risk missing out on a really effective preventive drug.
If we want to do a clinical trial in pre-symptomatic people at risk for genetic prion disease, we will ultimately need to show that the drug delays their age of onset, which is our goal. But, we’ve found that the unpredictable age of onset and small patient population in prion disease combine to make it numerically infeasible to demonstrate such a delay in onset in the type of clinical trial usually designed to support drug approval: a randomized trial lasting just a few years [Minikel 2019]. Instead, demonstrating that a drug delays the disease will take many years and will have to be done as a “post-marketing” study after a drug is already approved. In order to gain approval of that drug, then, we will need to choose something else as our primary endpoint — the main thing we will measure that will tell us whether the drug worked. It appears that our only option will be to rely on a biomarker: a molecular readout of whether the drug is doing its job.
One can imagine two distinct paradigms for how a biomarker could allow us to conduct meaningful clinical trials of a drug in pre-symptomatic people at risk for genetic prion disease, which we’ll call primary prevention and secondary prevention.
primary prevention biomarkers
In a primary prevention trial, one aims to intervene before any sign that anything has gone wrong. People would be recruited for the clinical trial based solely on their high genetic risk of eventually developing disease. Not only are these people asymptomatic, but there is no expectation that the disease process has necessarily even begun on the molecular level. Therefore, one must aim to measure, as your primary endpoint, a biomarker that is present in healthy people, in everyone. And yet, that biomarker must be so central to the nature of the disease that, if it changes, you can be confident enough that the drug is working that you would then want to take that drug for years in the interest of preventing the disease. What could such a biomarker be? We are lucky that prion disease boils down to a very simple molecular blueprint: everything centers on prion protein, PrP. PrP is not just a function of the disease state, but rather, is present in everyone. And yet, we know with certainty that PrP is the causal protein in prion disease, and we are very confident that if we can lower PrP, this will delay the onset of disease.
Recognizing this, a few years ago we began studying PrP in cerebrospinal fluid to evaluate whether it could serve as a biomarker in primary prevention trials, beginning with leftover CSF samples from other studies that other researchers generously shared with us [Vallabh 2019]. Fortunately, we found evidence that PrP in CSF was coming from the brain, as opposed to blood, suggesting it could tell us whether a drug was working in the brain. But there were two potential catches, one technical and one biological.
The technical problem was that PrP was exquisitely sensitive to plastic exposure. Every time you move CSF from one plastic tube to another — which happens a lot in the process of acquiring and analyzing samples — you might lose half of the PrP that was there. We suspected that this was a major contributor to the large variation we and others had observed in CSF PrP levels: different people could have CSF PrP concentrations that apparently differed by 100-fold. If CSF PrP differed by this much from day to day within one person, then we’d be sunk: there would be no way to measure CSF PrP before and after a drug and show that it went down. We found that handling samples very meticulously, and adding a small amount of detergent to CSF could help solve the plastic problem. And, in a small set of leftover CSF samples from an Alzheimer’s trial that had been handled very carefully, we found that each person’s CSF PrP level was stable over at least a couple of months. Still, until we collected our own samples and handled them our way from the very beginning, we couldn’t be certain that we’d identified all the variables that were affecting our ability to measure PrP.
The biological problem was that PrP concentration in CSF goes down in people who are at a symptomatic stage of prion disease. Several other researchers had found this previously, and we confirmed it in our lab as well [Vallabh 2019]. We wondered if perhaps it was just a problem with the method we used to measure CSF PrP, which relies on antibodies that, who knows, might not be able to measure PrP that has misfolded or been cut up by enzymes, both of which are things that happen in prion disease. So we developed a completely different method using mass spectrometry, only to find the exact same result [Minikel & Kuhn 2019]. The fact that CSF PrP drops during disease is a bit non-intuitive, because PrP actually builds up in the brain over the course of disease. But it may simply be that misfolded PrP in the disease state is too caught up in aggregates within brain tissue to make it out into the CSF. In any case, this raised a concern: what if CSF PrP concentration was already beginning to drop in people who were still pre-symptomatic? If PrP was already going down anyway, then it would be really hard to measure CSF PrP in a clinical trial and show that it went down as a result of a drug. While it was good that we’d been able to show CSF PrP was stable over a couple of months in the Alzheimer’s patients we had been able to measure, we really needed to test for the same stability in people at high risk of developing prion disease.
With all this in mind, we designed our study with the primary goal of assessing the stability of PrP levels in people at risk for genetic prion disease. We would bring people in for an initial visit, a short-term retest visit 2-4 months later and, funding allowing (which thankfully it did), a longer-term visit a year or more later. Across all these visits, we would collect CSF, handle it very carefully to minimize any variables related to plastic or other technical issues, and we could ask whether the CSF PrP concentration in one person was stable over time. If it was, then we could hope to one day measure a drug-dependent decrease in CSF PrP in this population.
secondary prevention biomarkers
In a secondary prevention trial, one aims to find people who are still asymptomatic, but in whom one can find evidence of a disease process already underway. People would be recruited for a clinical trial based on a molecular measurement, a brain imaging profile, or some other type of readout indicating their brains are already abnormal in some way. This abnormality would need to be present before the onset of obvious symptoms, and would ideally have a trajectory of getting more and more severe leading up to symptoms. In a trial, you would select such an abnormal marker as the primary endpoint, and measure how it responded to drug treatment. If the biomarker went back to normal, or at least stopped getting worse, or at very least got worse more slowly than it had been, you’d have a signal the drug was working.
In recent years, the secondary prevention concept has gained a lot of currency in the Alzheimer’s field. Researchers have figured out that there exists a decades-long prodrome — a period of pre-symptomatic pathological changes — in people at risk of developing Alzheimer’s. Decades before dementia strikes, one can see changes in concentrations of particular proteins in CSF, accumulation of protein plaques in the brain, shrinkage of overall brain volume, and so on, in a fairly predictable cascade leading to symptomatic disease [Bateman 2012]. Similar changes have been documented decades before the onset of Huntington’s disease, where measuring markers of damaged or dying neurons in blood looks like a particularly promising way to track people’s trajectories toward disease [Byrne 2017]. In both of those diseases, people are now designing trials to try to measure whether a drug changes the slope of that prodromal trajectory before people get sick.
A big difference is that Alzheimer’s and Huntington’s are slowly progressive dementias, while most forms of prion disease are quite rapid. Decades of imaging studies in prion disease have not managed to find much evidence of a prodrome — researchers have been able to find subtle changes only about 1 year before disease onset, changes that were not necessarily obvious at the time, but that in hindsight could be interpreted to presage disease. But outside of a couple of case reports, not much was known about how potential biomarkers in CSF or blood behave before symptom onset in prion disease.
As we were gearing up to bring at-risk people in to donate CSF for measuring PrP, we realized we also had an opportunity to look at potential prodromal markers for a secondary prevention model. The CSF and blood samples could be stored in the freezer for future analyses of a wide variety of potential markers, but for the immediate term, we decided to focus in on the few markers that seemed to have the best odds of being informative on a prodromal state, based everything we knew about fluid biomarkers up to that point. Total tau (T-tau) and neurofilament light chain (NfL), two markers of dying or damaged neurons, are well-studied proteins, with good tools available to measure them, that are very highly elevated in the spinal fluid and blood of symptomatic prion disease patients, and are also known to rise before symptom onset in more slowly progressive dementias, so both seemed like obvious candidates. Real-time quaking-induced conversion (RT-QuIC), an in vitro fibrillization assay for misfolded PrP in CSF, is the front line diagnostic test in prion disease, with excellent sensitivity and specificity at the symptomatic stage, so that seemed like another excellent candidate. We reasoned that if any disease process was occurring before symptom onset in people with PRNP mutations, it would be most likely to be reflected in one or more of these biomarkers.
With these considerations in mind, as we designed our study, while evaluating CSF PrP was the primary goal, we also planned to analyze these other fluid biomarkers. We knew that it was, fortunately, unlikely that we would see anybody become symptomatic during the first year or two of the study, so we might not be able to directly observe the dynamics of conversion to the disease state anytime soon. But even with shorter-term data, we could do a cross-sectional analysis to evaluate whether mutation carriers differed from controls on these biomarkers, and whether there was any obviously progressive change underway.
what we did
With the above goals in mind, we set out three years ago to launch a clinical research study that would bring PRNP mutation carriers and controls to Boston to donate CSF and blood for our research. We were lucky to connect with Dr. Steven Arnold, a neurologist who had just been recruited from Penn, was still building up his research program at MGH and had the bandwidth to think about taking on an ambitious new project. He was bright and compassionate, scientifically curious, and we sensed he’d be a great click with our patient community. It was a match: we and Dr. Arnold wrote a study protocol together, Sonia and I scraped together every last dollar of Prion Alliance money that we could, and in July 2017 — just under the wire, four days before our daughter was born — we launched the study. The response from the patient community was tremendous: within the first 48 hours, over 20 people called to sign up, exceeding the initially scoped study size of 10 mutation carriers and 10 controls. Since then, with additional funding coming in from Prion Alliance, from CJD Foundation, and just recently from NIH, we have been able to continue to expand the study. We have expanded both by bringing in more people, and by bringing people in for repeat visits as time goes on. The paper we just released describes initial findings from 43 participants over the first couple of years, but the study is very much ongoing (details here if you want to volunteer) and we hope it will continue to generate important insights.
Although prion disease is not subtle and we don’t expect that we could have failed to notice someone being symptomatic, participants underwent a battery of 20 different cognitive and motor tests, functional and psychiatric inventory questionnaires, and so on, to allow us to really rigorously assess and quantify whether they were indeed “asymptomatic”. We also sequenced DNA, for research purposes only — we were not set up to return results to participants. The study recruited people who knew they had a mutation, knew they didn’t have a mutation, or were at risk but didn’t know their status. On the research side we only see de-identified codes, we don’t know who is who, but we need to match each sample to its exact mutation. Meanwhile we looked at the fluid biomarkers of interest. CSF samples came to our lab at the Broad Institute, where Sonia measured PrP, T-tau, NfL, and RT-QuIC “seeding activity”. Blood plasma samples were shipped out to our collaborator Henrik Zetterberg in Sweden, who measured T-tau and NfL. We analyzed the biomarker data in a few different ways. In cross-sectional analyses we compared mutation carriers versus controls. We looked at test-retest reliabilty — stability between visits 2-4 months apart. And, while it’s early days, we did a pilot longitudinal analysis looking for change, or lack thereof, over the full duraton of the study so far, up to 20 months.
what we learned
After testing DNA, we found that this dataset of 43 individuals includes 27 mutation carriers and 16 controls. (Note that because many participants came in already knowing they had a mutation, the overall 27:16 ratio does not reflect the odds that a person of unknown status turned out to be positive.) The three PRNP mutations that account for the majority of genetic prion disease cases worldwide — P102L, D178N, and E200K — are all represented, which suggests that our study should be a reasonable sampling of the population of interest. We also saw four less common mutations (grouped under “other”), and having this diversity of different mutations in the cohort is a strength as well. The demographics of the groups are well-matched: in terms of age, sex, and how many study visits people made, mutation carriers and controls are about the same. This means that we can make meaningful group-wise comparisons between the people who turned out to have a PRNP mutation and those who did not. And, across all 20 measures of brain function that we assessed, mutation carriers look totally normal — we saw no evidence of a difference between groups. The pre-symptomatic mutation carriers truly are, as far as we can tell, without any symptoms.
The most important finding of the study is that CSF PrP concentration is stable over time. Any way we looked at it — in mutation carriers or controls, over the short term of a few months or the medium term of a year-plus, each person’s CSF PrP concentration was remarkably steady (with a coefficient of variation or CV of just 7%). That is really, really good news, because it means that if we can one day lower PrP by, say, 40% with a drug, it should be easy to measure the effect of that drug in a small number of people — the signal should be much larger than the noise. The fact that we were able to measure such rock solid stability of CSF PrP over time tells us that the protocol we implemented, where samples are handled very carefully and uniformly from the moment of collection, is working. In the leftover CSF samples we had access to before we launched this study, where CSF PrP had varied wildly both between and within individuals, we now think that much of the variability we saw must have simply been due to plastic exposure or other such “pre-analytical” variables, rather than genuine biological variation. Equally important, the stable baseline of CSF PrP in our population suggests that the decline in CSF PrP concentration observed in symptomatic disease is not happening yet in these pre-symptomatic individuals. That means we don’t need to worry about such a prodromal decline as a confounder in a trial in pre-symptomatic patients. We did notice that CSF PrP was a bit lower in people with the D178N mutation, in line with one previous study [Villar-Pique & Schmitz 2019]. But we found that this lower baseline was nonetheless stable over time: we didn’t see a progressive decline in CSF PrP in individuals with that mutation. Several studies have found that PrP with the D178N mutation is apparently less stable, with more of it being degraded before reaching the cell surface — so our interpretation is that the lower CSF PrP level in people with this mutation is probably a lifelong state, rather than part of a prodromal disease course.
The other key finding of the study is that, by and large, we do not see evidence of a prodrome in mutation carriers. The markers of neuronal damage that we looked at, T-tau and NfL, were normal in our mutation carriers, at low levels indistinguishable from controls, in both CSF and blood. As far as we can see, our participants’ neurons are not yet sick. This is a sharp contrast to studies of pre-symptomatic people with mutations that cause early-onset Alzheimer’s or Huntington’s disease, where neuronal damage can be detected a decade or more before the onset of clear symptoms. And this is all the more remarkable because, at the symptomatic stage, these markers are much more elevated in prion disease than in those slower dementias, presumably because neurons are dying so much more rapidly. Meanwhile, we generally did not see RT-QuIC “seeding activity” in the CSF of mutation carriers either, or in other words, we found no evidence that prions were yet replicating in people’s brains. Here, there was one exception: just one mutation carrier, out of 23, was positive by RT-QuIC. This was true at both study visits, despite steady levels of CSF PrP and low levels of T-tau and NfL still well within the normal range, suggesting no ongoing neuronal damage — and the individual remained asymptomatic a year later. So the prognostic or predictive value of RT-QuIC in a pre-symptomatic individual remains unclear, and cross-sectionally, we can say that most mutation carriers are not positive for this marker.
We launched this study to help us design clinical trials in pre-symptomatic individuals at risk for genetic prion disease, by assessing potential biomarkers that we would need for primary and secondary prevention trial paradigms. And while the clinical study is ongoing and we will continue to learn more, we released the new paper because we believe we have already learned some really important lessons to help us move forward.
First, CSF PrP should work as a biomarker for primary prevention trials. Our handling protocol for CSF appears to have reined in the variables that affected previous measurements of PrP. We now think that each person’s CSF PrP concentration is remarkably stable and reliable, so if we measure it today, a few months from now, and a year after that, we will get virtually the exact same answer. Therefore, we can test a PrP-lowering drug in pre-symptomatic people and measure how much that drug did its job in the brain, by measuring PrP in CSF.
Second, a secondary prevention model may be challenging in prion disease. We did not see any evidence of disease process underway in any mutation carrier in terms of neuronal damage, nor in 22/23 mutation carriers in terms of prion “seeding activity.” That means that a “secondary prevention” model, in which one would recruit people in a prodromal disease state and then try to track whether a drug lessened the progressive worsening of a prodromal biomarker, is unlikely to be realistic in prion disease.
Together, these findings give us a crucial insight to help us design drug trials in pre-symptomatic individuals: we should be focusing largely on primary prevention, designing trials to recruit people based solely on genetic risk, and to measure a drug-dependent change in PrP level as the primary endpoint of the trial.
Beyond the practical implications for who we recruit and what we measure in clinical trials, we believe that the findings of the study also carry a personal meaning. As far as we can tell, if you’re a mutation carrier, your brain is not already deteriorating out from underneath you years before you ever realize it. Your brain is as healthy today as it was the day you were born. And that means we have an opportunity — a mandate, even — to keep it that way.
Of course, an important limitation is that our study is young. With less than two years of follow-up data on people and (thankfully) no one having become symptomatic during the study, we certainly can’t rule out a prodromal state that might exist for some period of time in some people before disease onset. Two other studies — Michael Geschwind’s study at UCSF and Simon Mead’s study at UCL — have followed PRNP mutation carriers for far longer than we have, and will likely provide important insights on this question. Indeed, earlier this year at Prion2019, Andrew Thompson from UCL presented some valuable data suggesting that there is some rise in plasma NfL levels, indicative of neuronal damage, in the year or two before onset in some individuals. Those are important findings that motivate continuing to study potential prodromal markers. Based on the currently available data, such a marker would be more likely to provide supportive data as a secondary or exploratory endpoint in a trial, than to be the primary endpoint of a trial. Because the vast majority of carriers, at any given time, do not seem to have any evidence of a progressive disease process underway, it might be hard to structure a whole trial around a prodromal marker as a primary endpoint, and given our small patient population, restricting to only the very few people who are prodromal at any given time might not leave enough people to enroll a trial. Still, even if a trial enrolled 50 people and a prodromal marker was observed in just 2 or 3, and that marker stopped progressing or even improved after drug treatment, those data could potentially help to prove that the drug worked. For this reason, even though our data so far argue that any prodrome in genetic prion disease is likely to be brief, uncommon, or both, this question remains important, and we and others will continue to track NfL and other markers over time in the research volunteers who we are following.
We owe a huge debt of gratitude to so many people for making this study possible. Dr. Arnold and the whole clinical team at MGH have been spectacular, and our friends in the patient community who tell us they’ve participated in the study often mention what a pleasure everyone there was to interact with. We’re grateful to our collaborators in Sweden who analyzed the blood samples, and in Australia who provided the positive control CSF samples we used for comparison. But more than anything, we owe this all to the patient community. It was your sweat and tears (figuratively) and blood and spinal fluid (literally) that gave us these insights, and it is a tremendous gift that we cherish. Participating in the study wasn’t easy: many people drove or flew in from hours away, using their limited vacation days off of work to come out and participate repeatedly. And it was donations from patients and families, to Prion Alliance as well as CJD Foundation, that funded this study. It’s a debt we can never really repay, but we promise this: we will continue to give our lives to this work, striving every day to find a way to not only treat this disease, but prevent it from ever starting in the first place. From the bottom of our hearts, thank you for your help.