Skip to main content

National Institutes of Health

Eunice Kennedy Shriver National Institute of Child Health and Human Development

2023 Annual Report of the Division of Intramural Research

Membrane Interactions Underly Viral Spike Protein Porations, Microvesicle Assembly Sites on Umbilical Cord Endothelial Cells, and Glioblastoma Cell Migration.

Leonid Chernomordik
  • Joshua Zimmerberg, MD, PhD, Head, Section on Integrative Biophysics
  • Paul S. Blank, PhD, Staff Scientist
  • Svetlana Glushakova, MD, PhD, Staff Scientist
  • Matthias Garten, PhD, Visiting Fellow
  • Irene Jimenez Munguia, PhD, Visiting Fellow
  • Yuto Kegawa, PhD, Visiting Fellow
  • Avijit Sardar, PhD, Visiting Fellow
  • Adriana Golding, PhD, Postdoctoral Intramural Research Training Award Fellow
  • John E. Heuser, MD, Senior Biophysicist
  • Hang Waters, MS, Biologist
  • Jennifer Petersen, PhD, Electron Microscopist
  • Elena Mekhedov, MA, Contractor
  • Tatyana I. Tenkova-Heuser, PhD, Contractor
  • Glen Humphrey, PhD, Guest Researcher
  • Emily Feigen, BA, Postbaccalaureate Trainee
  • Garrett Tisdale, BA, Postbaccalaureate Trainee
  • Jaqulin Wallace, MS, Postbaccalaureate Trainee
  • Komala Shivanna, MS, Graduate Partnership Program

Eukaryotic life must create the many shapes and sizes of the system of internal membranes and organelles that inhabit the variety of cells in nature, i.e., membranes that must remodel for cells to repair damaged plasmalemma and deal with infectious agents such as viruses and parasites. Such basic membrane mechanisms must be highly regulated and highly organized in various hierarchies in space and time to allow the organism to thrive despite environmental challenges, genetic instability, unpredictable food supply, and physical trauma. We are using our expertise and the techniques we perfected over the years to address various biological problems that have in common the underlying regulation or disturbance of protein/lipid interactions. The overall goal of this project is to determine the physico-chemical mechanisms of membrane remodeling in cells.

Extracellular vesicles that secrete exosomes from internal multivesicular bodies

Signaling between cells, mediated by secreted membrane-enclosed organelles called extracellular vesicles (EVs), is a widespread form of intercellular communication, evolutionarily conserved from bacteria to plants and animals. Cells load EVs with a range of bioactive cargos, including lipids, membrane proteins, adhesion proteins, cytoskeletal elements, enzymes, signaling molecules, and nucleic acids. Once released into the extracellular milieu, EVs can signal locally or travel long distances in body fluids, such as blood, lymph, cerebrospinal fluid, amniotic fluid, to act on remote tissue targets. Upon reaching recipient cells, specific interactions between EVs and target cells promote binding and uptake by pinocytosis, phagocytosis, endocytosis, or direct fusion with the plasma membrane. EV–mediated intercellular signaling is a ubiquitous mechanism occurring under physiological and disease states. Plasma EVs are proving to be sensitive biomarkers of numerous disease states, such as cancer, and can be obtained by a liquid biopsy. Furthermore, EVs are being developed as vehicles for delivery of therapeutic agents. Despite these important physiological functions and medical utilities, much remains to be discovered about the biosynthesis of EVs.

It is generally viewed that EVs fall into two categories based on their site of biogenesis. Microvesicles arise at the plasma membrane, by outward budding and pinching off directly from the cell surface. In contrast, exosomes are of endosomal origin, produced when intraluminal vesicles (ILVs) contained in multivesicular bodies (MVBs) are exocytosed upon MVB fusion with the plasma membrane. Endothelial cells form the endothelium, a single cell–thick lining of the blood and lymphatic vessels, which controls the exchange of oxygen and nutrients between the vessel contents and underlying tissues. In their role on the 'front lines' of vessels, exposed to circulating body fluids, endothelial cells release a significant proportion of the EVs found in blood. EVs released by the endothelium contribute to its role in supporting vascular homeostasis, which includes maintenance of the anti-thrombogenic surface of the vessels (blood fluidity) and vasodilation, inhibition of inflammation, cell survival, and angiogenesis. To gain a better understanding of the mechanisms and structural aspects of EV release from endothelial cells under pro-angiogenic but non-inflammatory conditions, we used thin-section electron microscopy (EM) to examine HUVECs (human umbilical vein endothelial cells), and look for structural features consistent with microvesicle budding from the plasma membrane and exocytic release of exosomes from MVBs. Cells were preserved by ultra-fast freezing, which is optimal for capturing fast events such as exocytosis, and processed by a freeze-substitution protocol optimized for plasma membrane enhancement.

In thin sections, groups of protrusions were observed on the otherwise smooth HUVEC plasma membrane that were often branched and contained vesicular organelles, including MVBs with ILVs. Beyond cell peripheries, vesicles that contained MVB–like vesicles were observed, suggesting that they were microvesicles that had pinched off from the protrusions, diffused, and occasionally adhered to the coverslip. Serial sections through the presumptive microvesicles on the coverslip confirmed that they were not connected to cells by cellular extensions and that ILV–like vesicles were within the MVB–like vesicles. Further examination revealed omega figures, the structural hallmarks of exocytosis, occurring between MVB–like vesicles inside the microvesicles and their limiting membrane. On occasion, such omega figures contained small vesicles that were identical to ILVs. These observations support the notion that microvesicles containing many membrane compartments (referred to as multi-compartmented microvesicles, or MCMVs, a newly discovered class of microvesicles that bud from cellular protrusions clustered on the plasma membrane of HUVECs) pinch off from MVB–containing protrusions at specialized sites on the cell surface. MCMVs contain MVBs that apparently can release exosomes after transiting away from the parent cell.

Figure 1. Artistic representation based on data sets of a protrusion site on a cultured endothelial cell and omega figures on an MCMV a distance from a cell

Figure 1

Click image to view.

(a) Depiction of several protrusions clustered on a cultured endothelial cell surface, and a nearby pinched off MCMV (multi-compartmented microvesicles) floating in the extracellular milieu. A slice plane in the en face orientation relative to the coverslip is represented by the dashed line.

(b) A view at a slightly different angle shows that the protrusions are often branching.

(c) A slice through the protrusions and the MCMV shows internal round and tubular vesicular organelles, including MVBs containing ILVs, inside the protrusions and the MCMV. An omega figure on the MCMV–limiting membrane indicates exosome secretion from the MCMV.

The cellular membrane protrusions contain vesicular cargo that, when compared with cytoplasmic organelles, could be identified as MVBs with ILVs, endosomes (round, tubular, and clathrin-coated), ER, and mitochondria. Serial sections showed that the protrusions are a few hundred nanometers to 1 micron thick and in some cases extended up in the Z axis relative to the coverslips for more than approximately 1200 nm, beyond the scope of the serial sections analyzed. Protrusions were often branched and intermingled. At branch points and connections to the cell, the protrusions often became constricted to thin necks of 65–200 nm. Exploration of the coverslip surface between cells revealed MCMVs that were immobilized on the coverslip and contained MVB–like vesicles containing ILV–like vesicles, in addition to other vesicle types. The direct observation of omega figures joining membranes of internal vesicles of MCMVs with the peripheral membrane of MCMVs, and the many images of ILV–like vesicles in and immediately adjacent to the fusion pore of omega figures, and associated with the periphery of MCMVs, together suggest to us that ILVs can be released from MCMVs, a function akin to the exocytosis of ILVs from cells. Given that this could only be possible if MCMVs contain compartments akin to MVBs of cells, such a hypothesis would be unique to MCMVs. Preservation, by fast-freezing, of omega figures on the MCMV–limiting membrane showed a lack of any membrane coat. Thus, these omega figures are not the result of any coat-mediated endocytosis such as those mediated by clathrin or caveolin. It also seems unlikely, but not impossible, that ILV–sized vesicles are captured and internalized into MCMVs from the relatively vast volume of culture medium. Taken together, these observations suggest a novel pathway by which a subset of exosomes are released from a transiting MCMV after pinching off from a protrusion on the HUVEC surface.

A challenge for the study of EVs has been the isolation of EV subpopulations. Though possessing different sites of origin, microvesicles and exosomes share overlapping size ranges, molecular compositions, and densities, rendering biochemical enrichment and characterization of EV subsets a challenge. Our findings suggest that part of the difficulty may arise from EVs that consist of exosomes inside of microvesicles. Their presence could go unrecognized or be misinterpreted as apparent overlap in biophysical properties. Moreover, attempts to separate microvesicles from exosomes may prove futile when MCMVs are both. Of note, the in vitro cultivated HUVECs used in this study differ in some respects from tissue vascular endothelium, and future studies are needed to explore these findings in more a physiological system. The glycocalyx of HUVECs may differ from those in tissue, and an ILV could have more trouble traversing the glycocalyx of a tissue cell once released. Also, the vascular endothelium in tissue is not proliferating, whereas HUVECs in culture are exposed to several angiogenic growth factors in the medium and replicate many features of endothelial cells undergoing an angiogenic response. Angiogenic responses are known to alter the production and contents of endothelial-derived EVs, and EVs produced by endothelial cells can regulate angiogenic responses. The EV production we studied may be relevant to their role in angiogenesis. Furthermore, the HUVECs analyzed were sub-confluent to better observe potential sites of EV biogenesis on the cell peripheries; possibly, release of MCMVs occurs in response to a wound-like state to influence wound healing. Future studies are needed to determine whether MCMVs are released from confluent endothelial cells in vivo.

In summary, we have described a domain of protrusions extending from the plasma membrane of HUVECs that contain membrane-bound organelles including MVBs. MCMVs bud from the protrusions and contain vesicular compartments, including MVBs that can fuse with the MCMV–limiting membrane and release exosomes. This implies that the function of MCMVs is signaling rather than removal of cellular material, as has been proposed for exophers (membrane-bound EVs that are released by budding out of cells into the extracellular space) and migrasomes (EVs that are formed in migrating cells and mediate extracellular communication). To be functional as signaling entities, EVs must deliver messages, in the form of bioactive molecules, to recipient cells. Packaging cargos inside multiple layers of membrane, rather than a unilamellar carrier, could shield EV contents from degradation in the extracellular space, enabling them to voyage farther before being released from the MCMV or taken up into recipient cells. Multiple layers of membrane could also help vesicle contents avoid lysosomal degradation in the recipient cytoplasm and/or reach the nucleus. Additionally, grouping many vesicles of related signaling molecules into a single EV could deliver contents as a component kit, rather than relying on coincidental arrival of components in separate EVs, at the right place and in the right quantities, allowing for more efficient signaling. MCMVs can be evaluated as a new type of organelle-containing microvesicle, and a potential source of exosome release that occurs remotely from the parent cell, adding new considerations to when, where, and how EVs are assembled and released from the endothelium and potentially other cells and tissues.

Planar aggregation of the influenza viral fusion peptide alters membrane structure and hydration, promoting poration.

For all enveloped viruses, one or more glycoproteins on the surface of the viral membrane mediate the fusion of the envelope with the cell membrane for transport of the viral genome to the target cell cytoplasm, bringing about infection. In the first electron-microscopy visualizations of purified viral spike proteins from rabies, rubella, influenza, and other viruses, a striking similarity between spikes from different viruses was the assembly of these purified viral spike proteins into aggregates, termed rosettes, as their hydrophobic transmembrane domains (TMDs) aggregated. Most enveloped viruses enter their target cells via the endocytic pathway, where the viral envelope spike protein encounters an acidic pH. For the influenza virus spike protein hemagglutinin (HA), acidic pH activation of HA is necessary and sufficient for triggering fusion of the viral envelope with a variety of target membranes, including receptor-doped phospholipid bilayers. In the absence of their TMDs, activation of isolated soluble ectodomains of HA led anew to fresh rosettes of those trimers. The N-terminal domain of HA2 is responsible for this second aggregation of HA ectodomains; it is a short amphiphilic N-terminal sequence that became known as the fusion peptide (FP). As with the first rosette of HA, this second rosette formation is considered a consequence of the hydrophobic effect: the hydrophobic surface formed by one side of the FP would avoid water via association with the hydrophobic surface of another FP. The influenza FP comprises the N-terminal 21 amino acids of HA2, located within the HA ectodomain (at neutral pH), proximal to the HA trimer surface but near the TMD. At low pH, the FP is found in the target membrane, as evidenced by hydrophobic photolabeling. The FP is required for infection in vivo and membrane fusion in vitro and is featured in all hypotheses on HA–mediated fusion, although there is little agreement to date on structural, compositional, and mechanistic data on its exact role. Given that the FP is a highly conserved region of the influenza virus genome across many different subtypes of influenza virus, and a universal feature of enveloped viral fusion proteins, determining the FP's role in infectivity and membrane fusion is critical to finding variant-independent immunogens and pan-viral therapeutics to ameliorate morbidity and mortality.

In the first paper reporting this discovery, a study of the membrane mechanisms by which the influenza virus can disrupt a target membrane, we established that FPs underly this disruption: in target membranes, a reversible pore forms upon addition of FP in the absence of virus or even the rest of HA [Reference 2]. In molecular dynamics simulations crafted to understand the chemistry by which FPs act, a third kind of rosette emerged: the aggregation of FP via their lateral side chains (not their hydrophobic surfaces) into FP microdomains that displace lipids in the cis leaflet. This aggregated structure locally thinned the bilayer and significantly increased the probability of water entry. A new model is proposed to explain our data based on a tilting of FPs towards each other to further thin the remaining lipids immediately under even an FP dimer. For larger aggregates, this more hydrated, thinner membrane structure replaces the lipid bilayer in a small domain wherein a lipidic pore can form.

A second paper on the membrane mechanisms by which the influenza virus can disrupt a target membrane [Reference 3] described how the fusion peptide (FP) domain is necessary for the fusogenic activity of spike proteins in a variety of enveloped viruses, allowing the virus to infect the host cell; it is the only part of the protein that interacts directly with the target membrane lipid tails during fusion. There are consistent findings of poration by this domain in experimental model membrane systems, and, in certain conditions, the isolated FPs can generate pores. We used molecular dynamics simulations to investigate the specifics of how these FP–induced pores form in membranes with different compositions of lysolipids (derivatives of lipids resulting from hydrolytic removal of an acyl chain) and POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine). The simulations show that pores form spontaneously at high lysolipid concentrations via hybrid intermediates, where FP aggregates in the cis leaflet tilt to form a funnel-like structure that spans the leaflet and locally reduces the hydrophobic thickness, which must be traversed by water to form a pore. By restraining a single FP within an FP aggregate to this tilted conformation, pores can be formed in membranes of lower lysolipid content, including pure POPC, on the 100-ns timescale, much more rapidly than in unbiased simulations in bilayers with the same composition. The pore formation pathway is similar to the spontaneous formation at high lysolipid concentrations. Depending on the membrane composition, the pores can be metastable (as seen in POPC) or lead to membrane rupture.

Perivascular invasion of primary human glioblastoma cells in organotypic human brain slices: human cells migrating in human brain

Glioblastoma (GBM) is the most common and lethal adult primary brain tumor. Despite aggressive surgery, followed by standard-of-care chemo- and radiation therapy, median overall survival is only 14.6 months from the time of diagnosis. One significant feature of GBM is its highly invasive behavior, a feature that makes complete surgical resection virtually impossible, leading to inevitable recurrence and death. GBM cells undergo several biological modifications, including cellular volume changes, cytoskeletal modification, protein secretion, and development of functional multicellular network structures to actively invade the brain through the perivascular space, meninges, and white matter tracts. GBM invasion into brain tissue is influenced by many factors, including extra-cellular matrix (ECM) modification, the immune system, chemo-attractants, communication with other central nervous system (CNS) cells, and glucose and oxygen concentrations.

The brain-tumor microenvironment provides adequate extracellular matrix (ECM) components, cellular interactions, and mechanical properties to foment the invasion of tumor cells into healthy brain. Individual roles that microenvironment components play in the invasion characteristics of human GBM cells are studied in isolation, such as with the wound-healing assay, the microliter-scale migration assay, the spot assay, and the transwell migration assay. However, these systems do not attempt to model the totality of the brain microenvironment. Novel 3D-bioprinted models and microfluidics are designed to mimic the brain’s cytoarchitecture and to study cell-to-cell interactions, but they lack the specific ECM components and microenvironment of the human brain. While animal systems do utilize a living brain microenvironment in vivo, they do not represent the human brain microenvironment, and they are not always practical in terms of time, cost, and availability. Moreover, animal models do not accurately represent human cellular interactions, given the striking differences between human and rodent brain tissue architecture.

We evaluated the perivascular migration of human GBM cells in a human-derived ex vivo organotypic model, a model that allows one to see and study the movement of human cancer cells in human brain and their cellular interactions with the healthy brain cytoarchitecture. Brain viability after surgical excision and maintenance in the laboratory lasts for 14 days, ample time to study many aspects of GBM invasion. Live tissue fluorescence microscopy of brain slices inoculated with labeled GBM–derived cells directly demonstrated how GBM cells infiltrate the brain parenchyma, interacting with the surrounding microenvironment, as they move. After initiation of infiltration, inoculated GBM cells moved at a basal speed; their speed increased significantly when GBM cell processes contacted other structures such as blood vessels. Subsequently, GBM cells rapidly moved onto the surface of these structures, and then migrated along the structures at a speed slower than basal. The model can be used to experimentally manipulate and analyze GBM cell migration in human brain tissue ex vivo. It provides the closest representation of human brain cancer cell migration outside a living subject, and the list of potential applications is long, including studies of chemotherapeutics, mitogens, cell signaling pathways, and drug testing.

Methods to study human GBM cell invasion and migration

The strategies that have been traditionally used to study GBM cell invasion include 2D alternatives such as cell-culture surface manipulation, extracellular matrix substrates such as matrigel, and biomaterial scaffolds. Slightly more complex, quasi 3D models have been recently developed to study brain cancer migration; however, they all lack many aspects of the human brain parenchyma, as stated above. The use of human brain tissue sections offers the unique possibility of observing cell invasion in a more relevant microenvironment. Similar ex vivo approaches have been used to study glioma cell migration in rat brain pre-implanted with the glioma cell line C6 and then prepared as organotypic slices. Mammalian models in vivo include mice, rats, and pigs, but rely on non-human tissue. Studying human cell migration within non-human tissue is suboptimal for many reasons, including the need for immunocompromised hosts, the lack of human cell interactions, and the lack of human-specific ultrastructural differences. Organotypic tissue models are well established; they do require maintaining tissue ex vivo (outside of the organism’s body). Typically, this maintenance relies on the air-medium interface, whereby tissue explants are placed on a membrane at the interface between air and medium. Viability requires diffusion of substrates from the medium to the tissue explant. The model has primarily been described for rodent-based tissues; human-derived organotypic models are relatively uncommon. In previous work, our tumor organotypic model was maintained reproducibly for two weeks without significant change in tissue cyto-architecture, based on immunohistochemistry and electron microscopy. We have generalized this model to make use of the non-invaded adult human cortex to study human GBM cell migration and interactions ex vivo.

GBM invasion

Promptly upon implantation, GBM cells direct their migration towards blood vessels in a specifically oriented way, as determined by their persistence of directionality. Previous work demonstrated the attraction of glioma cells towards endothelial cells or endothelial-conditioned media. However, the identification of the extracellular signals that might regulate this initial attraction towards blood vessels is still a matter of investigation. The interplay between tumoral cells within the tumor microenvironment likely contributes significantly to tumor progression and resistance to therapy. Our model presents an opportunity to study these phenomena in human tissue. Similarly, it permits the study of molecules that allow GBM cell invasion through the brain parenchyma, such as matrix metalloproteases.

GBM invade human brain tissue sections using blood vessels as paths.

One reason for the preference of glioma cells for blood vessels could be a mechanistic one: they offer a path of least resistance for migration through an otherwise tight brain parenchyma. Another explanation may be that biological cues secreted or expressed by endothelial and pericyte cells could attract glioma cells. We observed that GBM–derived cells have a biased migration pattern towards blood vessels once placed directly on the surface of an organotypic slice. These cells increase their speed as they approach and make initial contact with a blood vessel and then slow down, once they fully attach to and follow its projection. These observations may indicate that human GBM cells have an intrinsic tropism towards blood vessels. They also suggest that when GBM cells lose their niche, migration increases until they find a new niche in the vicinity of the blood vessels. This property may be important because, following surgery, remaining GBM cells that are left in the resection cavity or surgical margin might lose their niche, causing them to accelerate their migration into less invaded brain parenchyma until a new niche is contacted. Preventing accelerated migration following surgery may require new therapeutic interventions.

In conclusion, we described an ex vivo organotypic explant model of human GBM invasion into human brain tissue. We demonstrated that GBM cells start migrating in the search of new perivascular niches and change their migratory dynamics once they contact blood vessels [Reference 4].


  1. Petersen JD, Mekhedov EM, Kaur S, Roberts DD, Zimmerberg J. Endothelial cells release microvesicles that harbour multivesicular bodies and secrete exosomes. J Extracell Biol 2023 2:e79.
  2. Rice A, Haldar S, Wang E, Blank PS, Akimov SA, Galimzyanov TR, Pastor RW, Zimmerberg J. Planar aggregation of the influenza viral fusion peptide alters membrane structure and hydration, promoting poration. Nat Commun 2022 13:7336.
  3. Rice A, Zimmerberg J, Pastor RW. Initiation and evolution of pores formed by influenza fusion peptides probed by lysolipid inclusion. Biophys J 2023 122:1018–1032.
  4. Ravin R, Suarez-Meade P, Busse B, Blank PS, Vivas-Buitrago T, Norton ES, Graepel S, Chaichana KL, Bezrukov L, Guerrero-Cazares H, Zimmerberg J, Quiñones-Hinojosa A. Perivascular invasion of primary human glioblastoma cells in organotypic human brain slices: human cells migrating in human brain. J Neurooncol 2023 164:43–54.
  5. Petersen JD, Lu J, Fitzgerald W, Zhou F, Blank PS, Matthies D, Zimmerberg J. Unique aggregation of retroviral particles pseudotyped with the delta variant SARS-CoV-2 spike protein. Viruses 2022 14:1024.
  6. Pfeiffer A, Petersen JD, Falduto GH, Anderson DE, Zimmerberg J, Metcalfe DD, Olivera A. Selective immunocapture reveals neoplastic human mast cells secrete distinct microvesicle- and exosome-like populations of KIT-containing extracellular vesicles. J Extracell Vesicles 2022 11:e12272.


  • Sergei Akimov, PhD, Frumkin Institute of Physical Chemistry and Electrochemistry, Russian Academy of Sciences, Moscow, Russia
  • Andrew Beaven, PhD, Unit on Membrane Chemical Physics, NICHD, Bethesda, MD
  • Wendy Fitzgerald, BA, Section on Intercellular Interactions, NICHD, Bethesda, MD
  • Vadim Frolov, PhD, Universidad del País Vasco, Bilbao, Spain
  • Timur Galimzyanov, PhD, Frumkin Institute of Physical Chemistry and Electrochemistry, Russian Academy of Sciences, Moscow, Russia
  • Samuel T. Hess, PhD, University of Maine, Orono, ME
  • Sukbir Kaur, PhD, Laboratory of Pathology, Center for Cancer Research, NCI, Bethesda, MD
  • Leonid Margolis, PhD, Section on Intercellular Interactions, NICHD, Bethesda, MD
  • Doreen Matthies, PhD, Unit on Structural Biology, NICHD, Bethesda, MD
  • Ana Olivera, PhD, Mast Cell Biology Section, NIAID, Bethesda, MD
  • Richard Pastor, PhD, Laboratory of Membrane Biophysics, NHLBI, Bethesda, MD
  • Annika Pfeiffer Daniels, PhD, Laboratory of Allergic Diseases, NIAID, Bethesda, MD
  • David D. Roberts, PhD, Laboratory of Pathology, Center for Cancer Research, NCI, Bethesda, MD
  • Alexander J. Sodt, PhD, Unit on Membrane Chemical Physics, NICHD, Bethesda, MD
  • Gary R. Whittaker, PhD, College of Veterinary Medicine, Cornell University, Ithaca, NY
  • Jack Yanovski, MD, PhD, Section on Growth and Obesity, NICHD, Bethesda, MD
  • Wei Zheng, PhD, Division of Preclinical Innovation, NCATS, Bethesda, MD
  • Fei Zhou, PhD, Unit on Structural Biology, NICHD, Bethesda, MD


For more information, email

Top of Page