The Molecular Kidnapping: How Reoviruses are Taking Advantage of Our ATXN2L protein to Produce Their Babies

 Mammalian reovirus is a diverse family of virus that has segmented double-stranded RNA genome contained in two protein shells, which distinguishes them from many other viral families. Despite causing disease in wide range of mammalian hosts, reovirus is not typically considered as a human pathogen due to the absence of obvious symptoms or known clinical impact. Because of this, it becomes a perfect 'model organism' for studying cellular processes that can be safely handled in the lab, including translation, which is the main subject of this paper. It is the last step in converting DNA into proteins that are the actual workers of our cells.

In our cells, translation is made possible by the presence of a hat (5'cap) and a tail (poly-A tail) appended to mRNAs (the temporary carrier of genetic information), which helps them to exit the nucleus, protect them from enzymes that can degrade them, and help initiate translation (circularize the mRNA and reload translation machinery). Then mRNA will be converted into protein in the cytoplasm. In comparison, reovirus RNA genome gets translated within viral factories, which are little houses with no walls (often in membrane-less compartments called liquid-liquid phase separation) containing necessary tools for baby viruses to assemble and mature. Reovirus µNS protein forms the scaffold of these houses and interact with the core, where viral mRNAs are transcribed with the hat but without the tail. Surprisingly, these viral mRNAs can be readily translated in infected cells without the assistance and protection of the poly-A tail! This paper seek to find out the host cell factor that reoviruses has 'kidnapped' to produce viral proteins for themselves.

Since µNS protein is the important building block of viral factories where viral translation takes place, the 'manipulated' cellular protein probably interact with µNS protein. Based on this logic, the researchers used a technique called proximity-dependent Biotin identification, which is basically a spray-paint process. µNS protein is tagged with the spray---biotin protein ligase, which will enable the biotinylation of nearby proteins---paint. The 50 painted proteins labeled with biotin are identified and underwent a protein-protein interaction cluster analysis. There's one cluster that matches well with the expectations: proteins associated with SG (stress granule) and translation initiation. Followed by a long process of screening that identified proteins that would lead to failure in viral replication if deleted and if temporarily suppressed through CRISPR and siRNA whole genome screening respectively, a single common candidate that pop out in all filtered results is identified: ATXN2L.

Then the hypothesis that ATXN2L is the essential mediator of viral translation without poly A tail is tested stepwise. If ATXN2L indeed facilitates the translation of viral protein, it should be indispensable for viral replication. The researchers hence used CRISPR/Cas9 gene editing technique, which basically messes up the nucleotides at the specific desired site (gene ATXN2L) and render it nonfunctional, and it resulted in greatly impaired viral replication compared to the control group with undisrupted ATXN2L gene (WT) and with ATXN2L gene re-introduced (KO+). The green fluorescence is the virus-specific the antibody staining, indicating the infectivity. 

Additionally, if ATXN2L plays an essential role in viral translation, it probably interact with viral mRNA in some ways. Therefore, the researchers then tried to confirm this hypothesis and identify the specific interacting region. By creating ATXN2L with different combinations of deletions of regions that potentially participates in RNA binding (LSm and LSmAD) or contacts with poly-A-tail-binding protein (PAM2 motif) (shown in the image below, Δ meaning deletion), they found out that removal of either LSm and LSmAD alone or together significantly decreases ATXN2L's interaction with µNS and viral yield, with LSm being the primary one. This result suggests that RNA binding domains of ATXN2L are required for viral replication and it's highly possible that ATXN2L physically contact with mRNA at these specific sites. 

Then the researchers used Biotin-Streptavidin pull-down to find out if they two indeed interact with each other and where the region of contact is. Biotin is the small molecule that is attached to the viral mRNA and Streptavidin is a protein that acts like a high-powered magnet that can pull out Biotin and the bound mRNA along with proteins associates with it. Researchers tested three versions of viral s4 mRNA (as surrogate or the "representative" one among the ten segments): the normal version (s4 3'UTR), a version with the last ten nucleotides truncated (s4 3'UTR Δ10) and one with the last ten nucleotides scrambled. They chose the last ten nucleotides because they contain a five-nucleotide highly conserved sequence (meaning that it's found among almost all viral mRNAs), which can have evolutionary significance. They found that for both the truncated and scrambled condition, Streptavidin was not able to pull biotinylated mRNA bound by ATXN2L out from the cell lysates whereas with the control condition, they detected ATXN2L binding to the biotinylated mRNA. This suggests that mRNA and ATXN2La indeed interact with each other, and the conserved UCAUC sequence is the critical "landing pad" for ATXN2L, likely mediated by the its LSm and LSmAD domains.

Then researchers also discovered other evidences that indicated ATXN2L's participation in viral protein production, including: the absence of ATXN2L blocked the enlargement of viral factories ( appears to be punctate), which occurs when more viral proteins are contained. In the image below, KO refers to the condition where ATXN2L is knocked out. ATXN2L is also found to associate with the translation complex (polysome), including 80S ribosome as well as translation initiation factors elF4G1 and elf4G3.

Finally and most importantly, the researchers asked if ATXN2L indeed facilitates translation of mRNA WITHOUT poly-A tail. They created two mRNA that can glow when translated (NanoLuciferase translation reporter construct), one with the poly-A tail and the other without, so that they can compare the translation level through detecting luminescence exhibited. They found that the translation level of mRNA with poly-A tail is maintained regardless of the presence of ATXN2L but the translation of mRNA without the tail is impaired significantly by the absence of ATXN2L. Eventually, the researchers arrived at the conclusion that ATXN2L is the protein that's being 'kidnapped' by reovirus to complete their translation as it's doing its normal work in managing cell stress as a SG-associated gene. 

"What is the point of walking me through this extremely long research journey?" you might think. First of all, how RNA viruses translate their mRNA without a poly-A tail has long been an unresolved puzzle. This research helps us to find the molecular bridge that connects the viral factories and mRNAs as well as the probable contact sites between mRNA and ATXN2L. The researchers presented to us a very rigorous and comprehensive way of identifying the protein that's playing the major role in viral translation, from initial screening based on certain criteria and confirming its interacting partner proteins as well as its "jobs" in viral replication from different angles. Additionally, now that the important role of ATXN2L in viral translation is established, it can be used as the potential target of antivirals. If we can develop a drug that blocks either the interaction of ATXN2L with µNS or viral mRNA, the reoviruses will be deprived of their ability to produce viral proteins in our cells. Taking even another step further, since reoviruses are known for their oncolytic potential (preferentially replicate in malignant tumor cells), engineering the interactions among ATXN2L, viral factories and mRNA can also increase the efficiency of delivering therapeutic agents into the cancer cells.  Conceivable next steps of this study can be testing the role of ATXN2L in viral yield in different cell types and families of viruses as well as investigating the specific mechanisms of how ATXN2L facilitate translation, such as circularization of mRNA and reloading the ribosome.

Bibliography:

Primary research article:

Somoulay, Xayathed, et al. “Ataxin-2-like Promotes Translation of Nonpolyadenylated Reovirus MRNA.” Nature Communications, vol. 17, no. 1, 17 Dec. 2025, www.nature.com/articles/s41467-025-67547-1, https://doi.org/10.1038/s41467-025-67547-1. Accessed 21 Apr. 2026.

Other Citations:

DouvilleRenée N., et al. “Reovirus Serotypes Elicit Distinctive Patterns of Recall

Immunity in Humans.” Journal of Virology, vol. 82, no. 15, Aug. 2008, pp. 7515–7523,

https://doi.org/10.1128/jvi.00464-08. Accessed 4 May 2022.

Lemay, Guy, and Simon Boudreault. “The Reovirus μ2 Protein, an Enigmatic

Multifunctional Protein with Numerous Secrets yet to Be Uncovered.” Virology, vol.

601, Jan. 2025, p. 110275, https://doi.org/10.1016/j.virol.2024.110275. Accessed 10

June 2025.

Lakshmipriya, Thangavel, et al. “Biotin-Streptavidin Competition Mediates Sensitive

Detection of Biomolecules in Enzyme Linked Immunosorbent Assay.” PLoS ONE, vol.

11, no. 3, 8 Mar. 2016, www.ncbi.nlm.nih.gov/pmc/articles/PMC4783082/,

https://doi.org/10.1371/journal.pone.0151153.

Sharifi, Negar, et al. “Reovirus Oncolysis and the next Frontiers for This Unique

Oncoviral Immunotherapy.” Seminars in Immunology, vol. 80, 30 Sept. 2025, p.

101995, www.sciencedirect.com/science/article/abs/pii/S1044532325000673,

https://doi.org/10.1016/j.smim.2025.101995.

Images:

https://www.creative-biolabs.com/proximity-dependent-biotin-identification-bioid-service.html

https://www.nature.com/articles/s41580-021-00417-y

https://www.preprints.org/manuscript/202506.0175/v1

Im

‌

‌

‌

‌

‌

 

 

Molecular analysis of human norovirus VP2 elucidates critical molecular interactions for virus assembly

 The article can be read here, and more about the primary investigator can be found here.  

 ***

            Norovirus is one of the leading causes of gastrointestinal illness around the world. Outbreaks are common in schools, nursing homes, and other crowded places and typically stem from contaminated food. As this paper highlights, there are estimated to be 680 million cases per year. Despite the high burden of disease, there is not currently a vaccine or a specific antiviral treatment for norovirus.

One of the main knowledge gaps holding back research into treatments and vaccines is a lack of understanding of the norovirus lifecycle, including how the viral proteins come together to make a new infectious particle. In this study, the authors identify and investigate a crucial component of this process.

 

The building blocks of the norovirus capsid

            The norovirus capsid, or protein shell surrounding the genome, is made of two proteins. The bulk of the capsid is formed from major capsid protein VP1. Compared with the rest of the virus, VP1 is well understood. Its structure can be studied through x-ray crystallography and cryo-electron microscopy, allowing researchers to visualize the protein. In contrast, the minor capsid protein VP2 has not yet been resolved through these methods. While prior research has shown that VP2 and VP1 interact with each other, the structure of VP2 can only be visualized through predictive software like AlphaFold and its function remains largely unknown. This study attempts to answer some of these questions surrounding VP2.

            A third component of this system is VPg, a protein that attaches to the norovirus genome. VPg is crucial for norovirus success, and without it the virus can neither infect cells nor make more virus once inside. VPg is also thought to be part of encapsidation, the process of packaging the viral genome into the capsid after the capsid has been built. To escort the genome inside the capsid, VPg might interact with either VP1 or VP2. As part of their attempt to understand VP2, the authors also determine the nature of its interactions with VPg.

 

Figure 2B: A model generated using the protein folding program AlphaFold shows the predicted shape and structure of norovirus protein VP2.

            The first thing the authors do is figure out if VP2 is different between genogroups of norovirus, and what they find is that VP2 is highly diverse. Most of the protein looks considerably different between norovirus genogroups, but at both ends of the protein there are regions that stay largely the same. This suggests that these are probably key components of the protein, since the lack of variation indicates that any changes to those areas would prevent the virus from reproducing or infecting new cells. With this in mind, the authors focus on these two regions as they search for the function of VP2.

 

Connecting the proteins

            The authors looked for an interaction between the major capsid protein VP1 and VP2 at the N-terminal end of VP2 (the region shown on the left of Figure 2B). To start, they looked to see if connections between the proteins stayed if VP2 was shortened. The authors found that if VP2 was shortened by more than 50 amino acids, the interaction with VP1 was lost. This told them that VP2 connected with VP1 at a location between the end of the protein and 50 amino acids in. 

 

Figure 3B: The N-terminal end of VP2 is the location of interactions with VP1. In this immunoprecipitation experiment, VP1 and VP2 are added to a mixture together and allowed to interact before VP1 is removed from the mix with anything that is attached to it. This figure shows that VP2 is found attached to VP1 when the N-terminal end of the protein is intact (represented by the green bars in the second row down) but when everything before the 50^th amino acid is lost VP2 no longer interacts with VP1. This shows that the site where VP2 and VP1 interact is before the 50^th amino acid in VP2.

            After determining the general area where the VP1-VP2 interaction happens, the authors then focus on isolating the specific amino acids on VP2 that connect with VP1. They find that changing amino acids 40-43 causes the mutated VP2 to lose its interaction with VP1, suggesting that these amino acids are where the interaction occurs. Further investigation suggests that all amino acids in this region contribute to the interaction and all of them need to be mutated for the interaction between VP1 and VP2 to be completely lost. 

 

Figure 3C: Mutations to amino acids 40-43 in VP2 cause the protein to lose its interaction with VP1. The same experimental procedure is followed as in 3B. Again, VP1-Vp2 interaction is represented by the presence of green bars in the second row down. These bars disappear when amino acids 40-43 are changed, but not when amino acids 39-42 are.

            Next, the authors investigated how the interactions between VP1, VP2, and VPg that form the norovirus capsid occur. The authors first confirm that the same region responsible for VP1-VP2 interaction, amino acids 40-43 on VP2, is not also the place where VPg interacts with VP2. They also show that VP1 and VPg cannot interact on their own. In figure 5C, the authors localize the interaction between Vp2 and VPg to the C-terminal end of VP2. This suggests that VP2 might act as a bridge between the two other capsid proteins, with interactions at one end connecting VP1 and capturing VPg with the other end. 

 

Figure 5C: VP2 interacts with VPg at the C-terminal end. Using a similar experiment to figure 3B, the authors see if VPg (represented by the blue band in the second row down) can interact with either the first or the second half of VP2. They find that when only the first half of VP2 is present VPg cannot bind, but VPg can attach when only the second half of VP2 is present.

            The final question the authors look into is whether VP2 can interact with both VP1 and VPg at the same time. Using the same type of experiment as before, they find that when VP2 is intact and present, pulling VP1 out of a sample also pulls out VP2 and VPg despite the established lack of direct interaction between VP1 and VPg. This confirms the authors’ hypothesis that VP2 bridges the gap between the other two capsid proteins, allowing the norovirus capsid to form.

 

But how does this keep me from getting food poisoning?

            For all the discomfort norovirus causes around the world, we know surprisingly little about it. The more we know about this virus, the more chances we will have to develop an effective vaccine to prevent disease or treatments for after people get sick. Understanding the norovirus capsid is particularly important for this. Viral capsids are often what gets noticed by the immune system (and what we can engineer an immune response against for a vaccine). Additionally, without an intact viral capsid the virus cannot spread to a new cell. Early in the paper, the authors mention that without intact VPg norovirus is not infectious. If we can figure out a way to inactivate VPg with a medication, it could help us limit and treat severe norovirus cases. This work is an important step forward in understanding the function of the proteins in the norovirus capsid, and lays the groundwork for future discoveries towards vaccine or antiviral design. While there is much still to be learned, understanding the basics of how the norovirus capsid is built is a crucial development in the field.

Cell Type Differences in HSV-1 Infection and HSV-1's Dependency on ICP27

You can find the article here and more information about the Dembowski Lab and the awesome work they do here! 

Introduction

Herpes simplex virus type 1 may affect up to 90% of the adult population and is known to infect 64% of the human population globally.1 Viral research into the infection, efficiency of infection, and outcome of infection of herpes simplex virus type-1 has allowed for treatments to be derived for extreme outbreaks which specialize in managing the infection and preventing further spread of the virus. Yet, these viral studies utilized different permissive cell types which best benefit the needs of their study, whether alleviating cost or cell type availability or using cells which are best for the specific aspect of HSV-1 they intend to study. Numerous cell types are used in HSV-1 studies, including African green monkey kidney epithelial cells (Vero)2, human lung fibroblasts (MRC-5)3, human foreskin fibroblasts (HFF)4, human diploid keratinocytes (N/TERT-2G)5, and human cervical cancer epithelial cells (HeLa)6. This lack of standardization can affect the outcomes of these viral studies. Here, Sabrina L. Rutan and Jill A. Dembowski of the Dembowski Lab at Duquesne University explored how different cell types can vary the infection, efficiency of infection, and outcome of infection in the aforementioned cells in their paper titled Cell type-specific differences in herpes simplex virus type 1 infection and dependency on ICP27, published in the Journal of Virology by the American Society for Microbiology. Furthermore, Sabrina and Jill aimed to investigate how reliant different cell types are on the infected cell protein (ICP27), a cell-type specific essential viral immediate early gene product which promotes efficient processing and transport of viral mRNAs. ICP27 helps to regulate viral gene expression through mRNA export and 3’ end processing.7 

Initial Findings

Initially, the kinetics of viral DNA replication between cell types was explored by infection of Vero, MRC-5, N/TERT-2G, HFF, and HeLa cells with HSV-1 strain KOS at a multiplicity of infection (MOI) of 10 plaque-forming units (PFU)/cell. Infected cells were harvested before the qPCR was analyzed, allowing for calculation of the number of viral genomes per cell. Through this process, HeLa cells were noted as having the number of infecting viral genomes decrease shortly after infection, with a delay in viral DNA replication. Replication kinetics for Vero, MRC-5, and HFF cells were similar, whereas N/TERT-2G cells were noted for having viral replication begin earlier and at a faster rate. All cell types had a similar maximum viral capacity. (Figure 1: seen below). Furthermore, HeLa cells were found to produce less infectious particles, despite producing similar quantities of viral genomes per cell as 24 hours past infection. On the other hand, HFF cells produced a notable larger quantity of infectious particles at 24 hpi, and Vero, MRC-5, N/TERT-2G, and HFF cells produces 1,000 PFU/cell by 96 hpi, whereas HeLa cells only produced 4 PFU/cell. 

Fig 1: Characterization of HSV-1 DNA replication and infectious virus output in several cell types. A) Comparative chart describing the quantities of cellular and viral genomes calculated using qPCR for each cell type. B) Comparison of viral yield between cell types after high MOI infection (10 PFU/cell). C) Comparison of viral yield after low MOI infection (0.01 PFU/cell). D) Coomparison of levels of deffective virus particles between cell types after high MOI infection after 24 hours. All values were obtain via a one-way ANOVA with a Tukey’s multiple comparisons test. 

HeLa cells produced more defective viral particles

    These findings were further explored, as HeLa cells were noted for producing more defective viral particles than all other cell types. By subjecting an aliquot of supernatant media from infected cells to proteinase K digestion, followed by DNA isolation and viral genome quantification through qPCR, the number of defective particles per PFU could be measured by cell type. Here, Sabrina and Jill found that Vero cells produced, on average, 60 defective particles per PFU (dp/PFU), HFF cells produced 27 dp/PFU, MRC-5 and N/TERT-G2 cells produced 82 dp/PFU, and HeLa cells produced an astonishing 170 dp/PFU. (Figure 2)

Fig 2) Comparison of HSV-1 KOS protein expression between cell types after infection at an MOI of 10 PFU/cell. Total protein was collected 2, 4, 6, and 8 hpi.

Viral Protein Expression Timing

Next, the researchers observed the timing of viral protein expression through observation of whole cell lysates at 2, 4, 6, and 8 hpi via western blotting. This method allowed for the observation of immediate early (IE), early (E), leaky late (LL), and late (L) viral proteins. ICP27 and ICP8 were found to have highly variable timing of expression between cell types. LL viral proteins were similar among all cell types except for HeLa cells, which saw a delay in LL protein expression. Lastly, L proteins were observed at 6 hpi for HFF and N/TERT-G2 cells and consistently at 8 hpi for all cell types. From these findings, the researchers concluded that despite relative temporal expression of viral protein consistency across cell types, the LL and L gene products expressed after the onset of viral DNA replication were unique enough that they may contribute to differences in observed viral DNA replication and viral particle production. 
Viral gene expression was then investigated through the quantification of expression of representative viral transcripts for each gene class. Total RNA was isolated at 2, 4, 6, and 8 hpi and after DNAse treatment, reverse transcription, and qPCR, mRNA copies of each gene per ug total RNA were documented. Viral mRNA levels were observed to be the highest in N/TERT-2G cells compared to other cell types at early times after infection. By 8 hpi, mRNA levels were similar for HFF, MRC-5, and N/TERT-2G cells for all HSV-1 gene classes. For N/TERT-2G cells, IE and E genes were observed until 4 hpi, whereas in HFF, MRC-5, and HeLa cells IE and E genes plateaued at 4 hpi. Also, in N/TERT-2G cells, LL gene expression peaked at 4 hpi and decreased onwards, whereas in all other cell types, LL genes steadily increased throughout the experiment. Consistently, HeLa cells displayed fewer E, LL, and L viral transcripts than the other cell types. (Figure 3) 

Fig 3) Viral gene expression kinetics varying between cell types during strain KOS infection. A) Immediate Early (IE) genes. B) Early (E) genes. C) Leaky Late (LL) genes. D) Late (L) genes. E) 18S rRNA and GAPDH controls for comparison.

Differences in the absence of ICP27

With differences in gene and protein expression noted for each cell type and different times past infection, the researchers then moved to explore the cell type-specific differences in infection in the absence of ICP27, focusing first on viral DNA replication. When ICP27 is absent, viral replication compartments still form in MRC-5 cells, where in Vero cells incoming viral genomes form small punctate structures which never grow to full compartments, indicating that without ICP27, viral replication cannot occur in Vero cells, which was proven through Edc-labeled DNA. Replication was also found to vary dependent on cell type, showing a cell type dependence on ICP27 for DNA replication. Viral gene and protein expression were roughly consistent across cell types during infection with ICP27 absent. 

Fig 5) Viral Gene expression kinetics during 5dl1.2 (ICP27 absent) infection. mRNA levels obtained at 2-8hpi for A) IE, B) E, C) LL, D), L genes with E) 18S rRNA as a control. All mRNA levels were quantified through qPCR and differences were noted through unpaired two-tailed Student’s t-test.

Conclusion

    Overall, Sabrina and Jill found that HeLa cells had a lower number of infecting viral genomes and that genome amplification was delayed. Although viral genomes produced across cell types were similar at 24 hpi, HeLa cells produced the least infectious particles, whereas N/TERT-2G cells produced the most infectious particles per PFU. N/TERT-2G cells were found to have viral DNA replication and gene expression begin earlier, but plateau earlier as well. These findings indicate that there are distinct differences between cell types for the kinetics of viral DNA replication. In addition, cell cycle-specific differences were observed for the infectious cycle of HSV-1 mutants who lacked ICP27, where HFF cells see nearly normal DNA replication, MRC-5 and N/TERT-2G see reduced DNA replication, and HeLa and Vero cells see no DNA replication. Their results highlight the importance of consistency and standardization in viral studies, as selection of cell type should be considerate of differences in viral infection kinetics. Despite these findings, no cell type is noted as the particular best for investigation of HSV-1 infection study. Instead, the authors go out of their way to highlight that there are advantages and disadvantages to each cell type, and that these must be considered when selecting cell type and comparing studies. Thus, viral infection kinetics vary depending on cell type, and viral DNA replication is cell type-dependent during HSV-1 infection in the absence of ICP27.

References:
1) Organization WH. 2023. Implementing the global health sector strategies on HIV, viral hepatitis and sexually transmitted infections, 2022–2030: report on progress and gaps 2024, Vol. 2, p 1–70.

2) Huleihel M, Shufan E, Zeiri L, Salman A. Detection of Vero Cells Infected with Herpes Simplex Types 1 and 2 and Varicella Zoster Viruses Using Raman Spectroscopy and Advanced Statistical Methods. PLoS One. 2016 Apr 14;11(4):e0153599. doi: 10.1371/journal.pone.0153599. PMID: 27078266; PMCID: PMC4831712.

3) Gleaves CA, Wilson DJ, Wold AD, Smith TF. Detection and serotyping of herpes simplex virus in MRC-5 cells by use of centrifugation and monoclonal antibodies 16 h postinoculation. J Clin Microbiol. 1985 Jan;21(1):29-32. doi: 10.1128/jcm.21.1.29-32.1985. PMID: 2981901; PMCID: PMC271574.

4) Pan X, Xie J, Zhang Z, Guo X, Li J, Lin D, Qian Y, Xu J, Hu Y, Shi J. Serotype-specific host proteome remodeling in human foreskin fibroblasts during lytic HSV-1 and HSV-2 infection. Virol J. 2025 Jul 14;22(1):239. doi: 10.1186/s12985-025-02803-w. PMID: 40660299; PMCID: PMC12257817.

5) Kite J, Russell T, Jones J, Elliott G. 2021. Cell-to-cell transmission of HSV1 in human keratinocytes in the absence of the major entry receptor, nectin1. PLoS Pathog 17:e1009631.

6)Wang, X., Diao, C., Yang, X. et al. ICP4-induced miR-101 attenuates HSV-1 replication. Sci Rep 6, 23205 (2016). https://doi.org/10.1038/srep23205

7) Smith RWP, Malik P, Clements JB. 2005. The herpes simplex virus ICP27 protein: a multifunctional post-transcriptional regulator of gene expression. Biochem Soc Trans 33:499–501.

SARS-CoV-2 Entry: The Potential EV Hitchhiker

You can find the original article here, and the Ya-Wen Chen Lab (Icahn Medical School at Mount Sinai) here.

Introduction

The COVID-19 pandemic, caused by SARS-CoV-2, is an event whose modern-day impact precludes introduction. In December 2019, an outbreak of SARS-CoV-2 emerged from a cluster of patients in Wuhan, situated in China’s Hubei province (1). By November 2025, upwards of 7.1 million deaths worldwide had been attributed to the pandemic, especially among adults above the age of 65 (2). SARS-CoV-2 infection is usually characterized, within the first 0-7 days post-infection, by its constitutional symptoms (the generic, body-wide symptoms associated with “normal” disease progression). In the vast majority of infections, this is followed by a gradual return to health following bodily elimination of the virus. For certain groups of individuals (especially older adults, immunocompromised individuals, and those with underlying pulmonary or respiratory illness), however, internal dysregulation or severe COVID-19 disease may manifest instead. This is usually due to the emergence of a “cytokine storm” (essentially an overproduction of molecules that cause inflammation), and SARS-CoV-2 infection of cell types outside of its normal hosts. There is also “Long COVID”, a poorly understood condition wherein patients experience internal dysregulation months to years after virus elimination (2). Overall, SARS-CoV-2 creates a very diverse set of illnesses and conditions, and the reasons for this remain poorly understood. Given these discrepancies, there has been a major focus in the scientific enterprise over the last several years to fully understand and characterize SARS-CoV-2 infection mechanisms, with the hope of producing more effective and specific therapies for affected individuals, especially given the large number of coronaviruses circulating in mammal populations. This paper, published from the Ya-Wen Chen Lab at the Icahn Medical School at Mount Sinai, explores a poorly understood but intriguing aspect of SARS-CoV-2 infection: how it can infect cells that do not possess its attachment receptors (molecules on cells that the virus uses to gain entry).

An Upheaval in SARS-CoV-2 Entry

Viruses build their outer proteins into super-specific shapes, and in the world of proteins, shape determines what you can do and what you can interact with. By doing so, viruses, under evolutionary pressure, design their outer proteins to specifically and strongly interact with host cells' outer proteins. This allows them to gain entry to these cells and determines which cells they can interact with. In the case of SARS-CoV-2, established models affirmed that the virus interacts with Angiotensin Converting Enzyme 2, or ACE2, on pulmonary epithelial cells (cells that comprise the surfaces of much of the lungs. This places the virus in an endosome (an internal cellular compartment) that then showers the virus with acids that allow it to release its genetic material. An alternate entry mechanism, however, involves TMPRSS2, a protein that changes SARS-CoV-2’s outer protein shapes, which in turn allows the virus’s outer membrane to fuse with the membrane of the host cell. 

The latter explanation generally assumes that these proteins are on the surface of cells, since the virus needs to access them first in order to gain entry. However, imaging of the cells reveals that TMPRSS2 is almost exclusively found on the inside of cells, near but not breaching the surface:

Fig. 1A, C: These results show that TMPRSS2 exists inside the cells. To the left, Figure 1A shows images of cells with dyes that stain TMPRSS2 and EPCAM, a surface protein. As can be seen, TMPRSS2 exists in an inner band within the EPCAM signals, indicating that it exists inside the cells, rather than on the surface. On the right, we can see the results of flow cytometry, where the further along an axis a signal is, the greater the amount of protein is in that sample. In the surface samples, we can see high levels of EPCAM-APC but low levels of TMPRSS2. In the inner samples, we observe high levels of both signals, indicating that TMPRSS2 is exclusively intracellular.

Fig. 2A and Fig. S5A: These results show that TMPRSS2 exists in extracellular vesicles or EVs (which are small compartments that carry stuff out of cells). To the left, there are more images showing TMPRSS2 in green and common “markers” of EVs in red. As can be seen, there is significant overlap between these signals. To the right, we can see that a great proportion of these EV markers are in the same places as TMPRSS2 (and vice versa), indicating that TMPRSS2 may exist inside of EVs.

Fig 3A-B: These results show that TMPRSS2 is expressed on the surface of pulmonary epithelial cells in diseased lungs. In A, we see that healthy lung cells have low levels of TMPRSS2 compared to EPCAM, indicating little to no TMPRSS2 is on their surfaces. In B, we see that amongst patients with idiopathic pulmonary fibrosis (IPF), fibrosis, or interstitial lung disease (ILD), there are small but nonzero levels of TMPRSS2 on cell surfaces

From the above results, the authors begin to challenge the latter SARS-CoV-2 entry mechanism. Only in diseased lungs is TMPRSS2 on the surface, since in healthy lung cells, TMPRSS2 remains in extracellular vesicles. EVs are notable in that they tend to be absorbed by other cells, especially phagocytic cells (cells that engulf and consume given materials). Based on these findings, then, the authors began to look for cells that contain TMPRSS2 but aren’t capable of producing the molecule, as they may be absorbing it through these EVs.

A Striking Discovery: SARS-CoV-2 can enter Cell Types that do not Produce TMPRSS2

In their search, the authors find that alveolar macrophages, or AMs, and endothelial cells (ECs) expressed TMPRSS2 on their surfaces despite not creating the genetic material needed to produce TMPRSS2. Alveolar macrophages are phagocytic cells that patrol the alveoli for incoming disease threats. Endothelial cells are cells that comprise the lining of blood vessels, and do not normally express TMPRSS2 at all. However, the authors find that some of these cells nevertheless have TMPRSS2 on their surfaces:

Fig 5A-B: These results show that, in some diseased lungs, endothelial cells and alveolar macrophages possess TMPRSS2 on their surfaces, despite these cell types being incapable of doing so. We can see in A and in B that TMPRSS2 levels are high in both cell populations. 

Fig 6C, E: These results show that, within purified alveolar macrophages and endothelial cells, adding the aforementioned EVs caused the sudden appearance of TMPRSS2 on their surfaces. In C and E, we can see that adding EVs caused a massive increase in the TMPRSS2 signal. This perhaps explains why, in certain cases, cells have TMPRSS2 on their surfaces even if they are incapable of producing it, and a potential way in which they then acquire TMPRSS2. 

Fig 7B-C: These results show that adding EVs to stem cells increases their susceptibility to SARS-CoV-2 infection. We can see in both panels that viral titers (concentrations) are massively increased when EVs are added. Stem cells are cells that haven’t developed into any given cell type, so they shouldn’t normally produce TMPRSS2 since they haven’t developed in pulmonary epithelial cells. However, by adding EVs, SARS-CoV-2 was able to infect them nonetheless, suggesting that the EVs were installing TMPRSS2 onto their surfaces, and thus allowing the virus to gain entry.

The aggregate of these experiments reveals a striking new mechanism through which SARS-CoV-2 gains access to cells. Extracellular vesicles can carry TMPRSS2 and install it onto the surface of phagocytic cells, expanding the range of cells SARS-CoV-2 can infect. This newfound host diversity may begin to explain some of the greater discrepancies regarding COVID-19 disease, especially within severe COVID-19. As aforementioned, this installation of TMPRSS2 onto cell surfaces occurs more frequently in diseased lungs, providing potential directions regarding COVID-19’s abnormal pathologies (the manifestation of disease) in such cases. One major question that remains is whether this release of EVs is a deliberate mechanism of SARS-CoV-2, meant to prolong or intensify infection, or if this release is a hijacked or deliberate mechanism of the body. The authors posit that these EVs may serve as “decoys”, meant to absorb viruses so that they cannot enter actual cells. It’s also possible, as the authors also consider, that SARS-CoV-2 deliberately “hitchhikes” within the EVs and waits for them to be absorbed by cells, which would allow the viruses to enter undetected. More experiments in these regards will be needed to meaningfully explore these avenues, but this paper nonetheless explores a fascinating aspect of SARS-CoV-2 infection. In exploring such questions, it may be possible to create better treatments against coronaviruses, especially for vulnerable individuals, and to better understand SARS-CoV-2 infection. Coronaviruses are diverse and circulate in animal populations, so there is an omnipresent threat that a new virus may emerge in humans, as has occurred a few times this century (such as SARS-CoV in 2006, MERS-CoV in 2011, and of course, SARS-CoV-2 in 2019). By studying SARS-CoV-2 in its entirety, we can better safeguard ourselves against such new diseases, improving outcomes for untold numbers of people as we continue in our ever-advancing day.

References

N.A., (2024, July 4).  CDC Museum COVID-19 Timeline. Center for Disease Control,

    https://www.cdc.gov/museum/timeline/covid19.html#print.

Zhu, Y., Sharma, L., & Chang, D. (2023). Pathophysiology and clinical management of 

coronavirus disease (COVID-19): a mini-review.Frontiers in immunology, 14, https://doi.org/10.3389/fimmu.2023.1116131. 

Rea-Moreno, M., Tian, L., Tavakol, T.N. et al. (2026). Unveiling alternate pathways for 

SARS-CoV-2 infection via extracellular vesicle-mediated transfer of ACE2 and TMPRSS2. Nature Communications, https://doi.org/10.1038/s41467-026-71680-w.