In 2007, Johan Skog, a new postdoc in Xandra Breakefield’s laboratory at Harvard Medical School, tried to culture fresh human glioblastoma tissue from biopsied material given to him by a neurosurgeon. When he put the cells into culture and looked at them under a light microscope, “they were the weirdest-looking cells we ever saw,” Breakefield recalls. “They were covered with bumps. And we were thinking, ‘What is this?’”
“This,” as it turns out, was vesicles. Lots and lots of vesicles, some as large as half a micron in diameter. Under an electron microscope, the cells were actually pumping out much smaller particles, too—as many as 10,000 a day, Breakefield says.
Extracellular vesicles (EVs), membrane-encapsulated packages secreted by cells into the circulatory system and found in all bodily fluids, can be as large as 2 microns and as small as about 50 nm in diameter; exosomes, one particularly well-studied subtype of EV, range from 50 to 150 nm. Researchers have been aware of them for years. But until the mid-2000s, they were largely dismissed as being cellular debris or perhaps carriers of interesting protein signals. As Skog recalls, that struck him as odd. “I was looking at these vesicles and thinking, ‘It would be strange if they could not contain RNA.’”
As it turns out, they did: The vesicles were chock-full of RNA that reflected the mutational status of the original tumor. Equally significant, those RNAs could serve as intracellular messengers, inducing recipient cells to change their behavior. Add purified glioblastoma vesicles to endothelial cells in culture, for instance, and they initiate blood vessel formation, or angiogenesis, says Breakefield. “They just about form tubules in front of your eyes.”
“Message in a bottle”
Today, researchers like Breakefield and Skog (who after her postdoc went on to found a company dedicated to this new science, called Exosome Diagnostics), are working hard to tease apart the biology of EVs and to translate that information into clinical action. They have made some exciting observations. For instance, EVs’ contents do not necessarily match that of the cells from which they arise.
“They’re like a message in a bottle,” says Andrew Hill of the Department of Biochemistry and Genetics at La Trobe University in Melbourne, Australia, the president-elect of the International Society for Extracellular Vesicles (ISEV).
The research community is finally taking note. Nearly 4,600 publications in PubMed include the keyword “exosomes,” 4,200 of them published since 2006. In 2013, the National Institutes of Health launched its Common Fund-backed Extracellular RNA Communication Consortium (ERCC). According to Julie Saugstad, an associate professor in the Department of Anesthesiology and Preoperative Medicine at Oregon Health and Science University, who is funded under the ERCC, the ISEV annual meeting has grown from hundreds of attendees at the inaugural 2011 meeting, to thousands. “It’s like everyone just woke up one day and said, ‘Oh my, these are very cool.’ And they are,” she agrees. But that doesn’t mean they’re easy to study.
One method does not fit all
Esther Nolte-’t Hoen, assistant professor of Biochemistry and Cell Biology at the University of Utrecht, has been investigating EVs since she was a graduate student. Back then, EVs were isolated (or really, enriched) from biofluids via differential centrifugation or density-based fractionation. Those techniques are still widely used today, she notes, but they are not terribly practical for nonexperts. Moreover, they cannot be used to purify vesicle subpopulations that differ in molecular composition and function.
More recently, size-exclusion chromatography has been added to the toolbox, as have various commercial options, including the Total Exosome Isolation reagents from Thermo Fisher Scientific, which recover exosomes via precipitation, and the exoRNeasy Serum/Plasma Kits from Exosome Diagnostics (distributed by Qiagen), which rely on spin columns.
The limitation of all these methods, says Alexander “Sasha” Vlassov, Senior Manager for R&D at Thermo Fisher Scientific, is that they aren’t specific for any one class of vesicles. Cells secrete “at least 10 different types of nanovesicles, but they are very difficult or impossible to differentiate due to similar size, density, and surface markers.”
At least some functions attributed to vesicles may in fact be carried out by free ribonucleoprotein complexes, which also tend to copurify with vesicles. And all these different entities, whether membrane-enclosed are not, are likely formed via different pathways, carry different cargoes, and perform different functions. Exosomes, for instance, are produced via the endosomal pathway—they are made inside the cell. Some larger vesicles, in contrast, bleb from the cell surface like viruses.
Naturally, researchers are working to identify protein markers that can aid in distinguishing these different vesicles. Clotilde Théry, a principal investigator at the French National Institute of Health and Medical Research (INSERM) and the Institut Curie in Paris, France, for instance, recently used liquid chromatography-mass spectrometry to probe the protein content of different vesicle fractions. Her analysis identified at least six distinct vesicle classes: large, medium, and small extracellular vesicles (which can be distinguished by centrifugation), with the latter category further subdivided into four subclasses based on protein signatures, including the general vesicle markers CD9, CD63, and CD81.
Cancer cells, of course, produce unique constellations of proteins, and researchers are particularly interested in identifying proteins that mark cancer-derived vesicles. In one recent example of such research, Raghu Kalluri, professor and chairman of the Department of Cancer Biology at the University of Texas MD Anderson Cancer Center in Houston, and colleagues identified glypican-1 (GPC1) as a pancreatic cancer-specific vesicle marker. (Kalluri cofounded and holds equity in Codiak Biosciences, a company that exploits exosomes for the diagnosis and treatment of various diseases.) GPC1-positive vesicles seem doubly informative, Kalluri says: GPC1-positive vesicle abundance correlated with disease severity, while genetic analysis of vesicular RNA using quantitative polymerase chain reaction (qPCR) revealed a tumor’s mutational status.
Similarly, researchers at Exosome Sciences demonstrated recently in a study of 78 former professional football players that extracellular vesicles enriched in the neurological protein tau—which the company calls “TauSomes”—are elevated in athletes with chronic traumatic encephalopathy, a neurological condition that currently cannot be definitively diagnosed antemortem.
Go it alone
Though many researchers study EV populations en masse, John Nolan, a professor at the Scintillon Institute in San Diego, California, prefers a different approach.
Just as cells differ in their protein and gene-expression properties, so too do their secreted vesicles. The only way to get at those differences, Nolan says, is to analyze those particles one by one. His method of choice: flow cytometry.
Adapting flow cytometry to nanometer-sized vesicles isn’t easy, Nolan notes. A 10-micron cell might have 100,000 copies of an abundant protein on its surface, and even low-abundance proteins are present at a few thousand copies. But an EV measuring 100 nm in diameter is 100 times smaller than that, with 10,000 times less surface area and 1 million times less volume, and thus contains far fewer proteins for antibodies to latch onto.
Nolan and his team have built custom instrumentation designed to maximize light generation and detection, and coupled it with exceptionally bright fluorophores, protocols, and calibration standards. “We make a bunch of changes on the light source, the fluidics, the detectors, and the light collection, all of which improves performance by 50 percent to 200 percent,” he explains.
Nolan used that system to quantify vesicles studded with specific markers—CD61 and annexin V—in rat plasma. He could also distinguish particles based on size, because mixing EVs with di-8-ANEPPS, a membrane-binding dye, produces a signal proportional to the surface area of the membranes.
According to Nolan, commercial systems like the Beckman Coulter CytoFLEX are now coming online that also have the fluorescence sensitivity required for vesicle analysis. That should make the technology more accessible to the wider research community, he says. But he notes that there’s still a challenge to understanding vesicles as biomarkers of disease: Nobody knows what a normal vesicle distribution looks like.
Still, researchers are forging ahead. For instance, Nolte-’t Hoen and her coworkers have used a modified BD Biosciences Influx system to sort vesicles derived from mast cells. Using antibodies to either CD9 or CD63, they demonstrated that some vesicles contain one protein and some contain the other. “We think it may have to do with their route of biogenesis, that they come from different parts of the cell,” she says.
It’s also unclear whether the two vesicle types perform different functions. And that may not be easy to determine, says Nolte-’t Hoen, “because of course now you have an antibody attached to your vesicle, which may influence the functionality.” To circumvent that problem, she is now investigating “negative sorting” strategies, in which vesicles are enriched based on the proteins they do not contain.
Exosomes are produced via
the endosomal pathway—
they are made inside the cell.
Despite the yawning knowledge gap in basic vesicle biology, many researchers’ eyes are fixed elsewhere, specifically on EVs’ clinical potential.
Many on the diagnostics front, for instance, are pursuing so-called “liquid biopsies.” Rather than diagnosing, staging, and monitoring disease (especially cancer) via a solid tumor biopsy or noninvasive imaging, clinicians can theoretically extract similar information from blood, urine, and other biofluids such as circulating tumor cells, circulating tumor DNA—and exosomes.
Exosome Diagnostics’ ExoDx Lung(ALK) assay, for instance, uses quantitative real-time PCR to profile the EML4-ALK gene fusion from blood plasma, a genetic marker of susceptibility to the kinase inhibitor crizotinib. A recent study in JAMA Oncology suggests that the company’s prostate cancer assay, which is not yet commercially available, can likewise stratify patients into low- and high-risk categories based on the expression of three genes.
At Caris Life Sciences, a “molecular-profiling” company that focuses on EVs, researchers use a subtractive-binding technology called “ADAPT” (Adaptive Dynamic Artificial Poly-ligand Targeting) to identify protein signatures of disease, says David Spetzler, the company’s chief scientific officer. Recently the company used a library of 2,000 peptides on 500 patient samples to develop a signature that was better able to detect cancer in dense breast tissue than was mammography.
Similarly, Saugstad has studied the RNA content of cerebrospinal fluid (CSF) to identify a potential signature of Alzheimer’s disease. Starting from a set of 756 known human microRNAs (miRNAs), her team identified 36 whose abundance in CSF appears to correlate with the disease. Given that miRNAs are regulatory noncoding molecules, that information could identify novel proteins involved in pathophysiology, she says.
In Boston, Hakho Lee, director of the Biomedical Engineering Program at the Center for Systems Biology at Massachusetts General Hospital, is taking yet another approach to exosome diagnostics: microfluidics.
Lee has developed liquid biopsy analytical tools based on multiple principles over the years, including electrochemical detection, magnetics, acoustics, and more. His current state-of-the-art technology, he says, exploits surface plasmon resonance (SPR).
SPR is a mature technology that has been commercialized by companies such as Biacore (now part of GE Healthcare) to quantify protein-protein and protein-ligand interactions. To generate SPR, antibodies are affixed to a thin sheet of gold atop a prism. Light passing through the prism bounces off the bottom of the gold strip at a defined angle. As molecules bind to the opposite face of the gold sensor, that angle changes in proportion to the degree of binding, providing a real-time readout of molecular interaction.
In Lee’s version of the technology, antibodies are spotted on tiny gold sensors in a “periodic nanohole array,” which is arranged on a microfluidic chip. As vesicles bind to these sensors, their spectral responses change proportionally to the degree of binding. Best of all, the measured vesicles can subsequently be purified for downstream genetic or protein analysis.
According to Lee, the system is highly scalable. In a proof of principle study, for instance, his lab developed a “nano-plasmonic exosome” (nPLEX) sensor with 1,089 detection sites. From a pool of 71 proteins expressed on ovarian cancer cell lines, they identified a two-protein exosomal signature that they subsequently applied to 20 cancer and 10 control subjects. That signature tracked treatment response in the nPLEX assay, Lee notes, with the marker expression dropping in responding patients but increasing in nonresponders.
Researchers are also investigating exosomes as vehicles for delivering therapeutics. EVs, says Joshua Leonard, associate professor of Chemical and Biological Engineering at Northwestern University, seem to exhibit some of the properties—especially low toxicity—that researchers have been struggling to achieve with synthetic vesicles.
For instance, researchers can load EVs with specific cargo using electroporation, or by expressing nucleic acids in EV-producing cells. In 2011, Matthew Wood and colleagues at the University of Oxford used both approaches to show that they could use exosomes to downregulate neuronal gene expression in the mouse brain by loading the exosomes with a neuron-targeting peptide and specific short interfering RNAs. That result, Leonard says, suggests EVs can overcome at least three significant hurdles: crossing the blood-brain barrier, getting taken up specifically by neurons, and successfully delivering content inside the cells.
More recently, Leonard’s team, led by graduate student Michelle Hung, has begun teasing apart the rules governing RNA-loading in EVs. The team fused an exosomal protein to a bacteriophage protein normally involved in loading nucleic acids into viruses, coupled that protein’s signal sequence hairpin to RNAs of different lengths, and monitored what RNAs ended up in the resulting particles. All RNAs could be loaded, they found, though longer sequences and messenger RNAs tended to load less efficiently. “We tried to come up with a first pass at a quantitative map of the rules for loading EVs,” he explains.
It will take considerable effort to convert such observations into clinical realities, of course. But given the engagement of the research community, expect those advances sooner rather than later. “There’s probably a language here, and we’re [only] at the level of knowing something about the alphabet,” concedes Breakefield. “We don’t know the grammar, we don’t know who’s talking to whom, or when, or why. But we’re figuring it out.”
Learn more about new methods for detecting EVs in our recent webinar