Extracellular vesicle microRNAs (EV-miRNAs) are among the range of EV cargo components that may act as an informative and accessible window to disease. As short, single-stranded RNA molecules comprised of approximately 22 nucleotides in length, EV-miRNAs have attracted attention for their biological significance and clinical potential. Among the EV researchers unravelling miRNA data is Michiel Pegtel, Principal Investigator at Amsterdam UMC and co-founder of EV-based diagnostics company Exbiome. After having the opportunity to speak to Michiel Pegtel, we share a snippet of his journey and share his thoughts on reproducibility, isolation, and the future of EV diagnostics.
Pegtel has a long history with EV research; while currently he is focused on translational studies, his fascination with the field began with questions about fundamental biology. “It all started when I was doing my PhD in the US,” recalls Pegtel. “There was a collaborator of my thesis advisor who was Dutch. He studied EBV (Epstein-Barr virus), and he said that he had seen some slides from cancer tissue, where he saw a protein of this tumour virus around – and potentially in – other cells that he knew for sure were not infected with the virus.”
That Dutch collaborator, Professor Jaap Middeldorp, suspected that the viral protein was secreted by infected cells. Paired with Pegtel’s interest in EBV miRNA, they had the beginnings of what would be a successful collaboration. Together, they rapidly obtained a grant to explore the functional transfer of viral miRNAs, and soon after published a paper in PNAS titled ‘Functional delivery of viral miRNAs via exosomes’.1 “We provided the first proof of functional transfer of microRNAs by EVs,” says Pegtel. While EBV only infects B cells and epithelial cells, Pegtel et al. (2010) detected significant numbers of EBV miRNAs in non-B cells (T cells), alongside further circumstantial evidence suggesting EV-mediated transfer; for instance, they showed that the EBV-miRNAs were released by B cells and present in EVs, and they demonstrated dose-dependent miRNA-mediated repression of confirmed EBV target genes when EVs were internalised by monocyte-derived dendritic cells.
“And not only did we show it in vitro – we showed that transfer happened in vivo. And I think that aspect was missed by a lot of people. We were very lucky to access samples from HIV patients for that study, which has now been cited >1000 times, so that was a good start,” says Pegtel. “But it was also difficult because at the beginning, people didn’t really believe exosome mediated cell-cell communication was a bonafide biological phenomenon.”
Exosomes are well known for their slow rise to fame and it took a long time for the scientific community to acknowledge their unique mode of biogenesis and roles beyond waste removal. This is partly due to the difficulty that lies in demonstrating a vesicle is generated from internal compartments.
To explore exosome release events, Pegtel and his colleagues completed an imaging experiment that he now looks back on as one of his proudest achievements.2 “We repurposed a tool that was generated in the neurophysiology field. It’s called pHluorin, which is a pH-sensitive form of green fluorescent protein. We tagged pHluorin to an external loop of CD63 and provided what I think is the only direct evidence that multivesicular bodies fuse with the plasma membrane to release exosomes,” says Pegtel. CD63-pHluorin remains quenched while inside the acidic lumen of multivesicular bodies (MVB), yet upon fusion of the MVB with the plasma membrane, a rapid increase in fluorescence is generated that can be detected by live microscopy. “All the prior evidence was that of static images which cannot discriminate between a small vesicle fusing with a membrane event or a membrane budding event releasing a vesicle. Also, because these are highly dynamic processes it’s very difficult to capture such events.”
Since those early days, Michiel Pegtel has been working towards several long-term goals, including the development of an EV-based diagnostic test while – importantly – producing reproducible data that will be of use to others.
Currently, there are many open questions surrounding the biology of EV-miRNA and how to utilise EV-miRNA data in a meaningful and reproducible way. For Pegtel, his key research questions stem from the challenge of separating EV-miRNA derived from living cells, against a background of other signals.
“First, exosomes carry RNA and some of that RNA really is in the lumen and protected from external RNases”, says Pegtel. “Or, it can be in the membrane, or on the outside and protected somehow, by being bound to a protein or in a complex. And a lot of people have shown that using enzymes – for example, if you use proteinase K – then all of a sudden this RNA becomes accessible to RNases and is degraded. It’s actually only a very small portion that is in the lumen or somehow buried in the membrane of these vesicles.”
The proportion of miRNAs within EVs may be small, but that does not make them less attractive biomarkers than non-EV circulating miRNAs. In a way, the protection of miRNA in EVs gives EV-miRNA advantages over other extracellular sources of miRNA. “The EV snapshot is more stable, I would say. The signals you get from the serum using total microRNA is more due to all kinds of processes you cannot control and is less likely to be informative on any disease state”, explains Pegtel. This is especially important in diagnostics, as a labile biomarker will never be a good biomarker. If there is a possibility that your biomarker could be degraded before you can measure it, then you risk making clinical mistakes based on unsound data.
“But you still have to separate these signals from non-vesicular sources, if you’re really interested in the communication aspect of it, or at least interested in identifying what is released from living cells – for example, rapidly dividing tumour cells.”
Having the ability to filter and interpret EV-miRNA data from rapidly dividing tumour cells, against a background of noise, would be a gamechanger for diagnostics. And this is precisely what Pegtel is working on now, using Hodgkin’s lymphoma as a model.
It is widely accepted that reproducible research and diagnostic tests require a foundation of reproducible EV isolation – and reproducibility is something Pegtel is clearly passionate about. “I think with a young field of science, sometimes it’s overly optimistic in saying ‘oh, look what we can do’… and then it can never be reproduced. And very often, it costs a lot to reproduce the data, so a lot of people don’t even do it. And if nobody can reproduce it… it's not useful, so how can you be proud of it? In fact, it only costs people money, because people are trying to reproduce it, and that’s a major blow.”
But it’s not just about the money and keeping research on the right track. In the diagnostics space, the importance of reproducibility cannot be overstated. “If I really want to make a test, would I be comfortable, for example, withholding a patient from a certain treatment because of my test? Can I say that to a patient? Is my test good enough? You know, that’s a different question than in research, where the bar of accuracy is so much lower although it should perhaps not be.”
“If the isolation procedure is not standardised, and it’s biased somehow, or one time it works and the other time it doesn’t then obviously, for diagnostics that’s not good. You need to have consistent results from the same plasma sample.”
Pegtel reflects on the early days of EV research, when ultracentrifugation (UC) was the standard: “We noticed early on that UC was very error prone. Then some people started using sucrose gradients, but that is not something that you can use for a sample of 1 mL, or even less for diagnostics. You need to use something that’s much more applicable and scalable than that.”
Over the years, Pegtel and his group have been using size exclusion chromatography columns, and more recently, qEV columns which have helped with scaling up the operation. “Now we purify the exosomes in a more systematic manner with automated SEC, and that seems to work very well for our purposes.”
“Hodgkin’s lymphoma has a relatively homogenous population of immune cells that are not supposed to be there, like T cell infiltrate, monocytes and neutrophils. And the simplistic idea is that if these immune cells are rapidly dividing, and they're spitting (more) exosomes, which they do in culture, then maybe we can see that back in the plasma – as long as we take an equal volume of plasma to run through the column. And then we look for the microRNAs that we discovered in vitro.”
Initially, Pegtel’s group used qRT-PCR to study EV-miRNAs. Despite finding it cumbersome, they made significant progress on developing a quantitative test that could distinguish PET-positive patients with Hodgkin’s lymphoma from those who were PET-negative. Pegtel notes that if they did everything carefully and correctly, they could identify elevated levels of specific EV-miRNAs in plasma prior to treatment and identify a drop in those levels with successful treatment. “You could see that these levels stayed down for years and went up again with relapse. So, I think that was the first demonstration that miRNA diagnostics at that level could work and tell you something about the tumour activity. And it corresponded really well with PET imaging,” says Pegtel.
“The only problem is, how do you do that in the clinic? And how do you make it into a real test? So we’ve been busy with that.”
Pegtel’s focus is now placed on interpreting EV-RNA sequencing data to a standard that would be acceptable for use in the clinic. This involves the detection of approximately 10,000 molecules representing 400 miRNAs and their variants, all from 1 mL of plasma. Making sense of such a large library of data is quite the challenge, and requires a systematic and computational approach. For this, Pegtel is taking inspiration from the Human Genome Project and working on developing a reference exosomal miRNA profile in humans.
“And if you have a deviation of that, then is that related to disease?”, asks Pegtel. “You can normalise all of the miRNAs towards each other and say well 10% of all the microRNAs are derived from, say, miRNA-21. And then if in one patient, one microRNA like let7 becomes more prominent, then that means the other microRNAs go down. And the moment you do that, it becomes quite sensitive.”
Pegtel continues, “Absolute quantitation is quite hard to do, but if you have relative quantitation that is accurate, then that’s when you can compare one patient more easily with other patients, and begin to think ok, where are these microRNAs coming from? What is their biological relevance? And how stable are these changes?”
Pegtel notes that it will take time to collate enough data that allows for the average EV-miRNA profile to be established in humans, and acknowledges some confounding factors that the group has identified in their studies. However, given the promising results seen so far from a huge number of data points, Pegtel is optimistic that with machine learning it will be possible. Importantly, Pegtel is aware of the need for precision and quality control, which will be key to seeing this project through to a clinically relevant tool.
“If we control everything – the sequencing, blood processing, ensuring there are no batch effects – if we can do that, we can diagnose this disease and identify different types of responders: very good responders, partial responders, non-responders, and patients that are really in clinical remission that have no active disease anymore. And based on the vesicle microRNA profiles, we can see differences there. So that’s exciting.”