One Virus Is Not Always Enough: Cell Study Finds Infection Threshold Varies by Cell Type

In HeLa cells, poliovirus needs about 14 particles to replicate reliably. In primary human macrophages, the same virus needs about 190. That gap sits at the center of a paper in Cell that questions the textbook idea of infection as one virus entering one cell and making copies.

The paper was led by first authors Kohiro Mukai and Dongya Xia, with Nihal Alton-Bonet as corresponding author, across NIH, Academia Sinica, the University of Maryland, and Middle East Technical University. It began with a narrower puzzle: why do viruses carried inside extracellular vesicles, such as enteroviruses, norovirus, polyomavirus and adenovirus, infect fewer cells but replicate more efficiently than free particles? The podcast This Week in Virology went through it with Columbia virologist Vincent Racaniello, immunologist Breanne Barker, and Texas A&M's Jolene Ramsey. They stressed that most experiments used plain free virions at varying doses, so the finding concerns how many genomes enter a cell at once.

At a low dose of 10^4 units, 97% of HeLa cells held at least one viral genome, but only 25% replicated it. At 100 times that dose, 95% replicated. In macrophages the figures were 59%, 94% and 100% as the dose rose. The authors modeled a per-cell threshold they call VITREP, the virion input threshold for replication. Poliovirus came in at 14 ± 1 in HeLa cells, 9 ± 1 in iPSC-derived neurons, and 190 ± 40 in macrophages. Coxsackievirus measured 37 in A549 cells and glia, and SARS-CoV-2 86 ± 9 in glia.

The proposed mechanism is counterintuitive: more virus produces a weaker innate immune response. CXCL10 and interferon-beta were higher at low input and lower at high input across macrophages, HeLa and A549 cells. Phosphorylated IRF3 appeared at low doses and faded as dose rose, with suppression visible within 30 to 60 minutes. Synthetic poly(I:C), a double-stranded RNA mimic with no viral protein, reproduced the effect, so the trigger appears to be the quantity of foreign nucleic acid. That also rules out defective genomes complementing each other.

The authors identified negative feedback regulators switched on by high RNA loads: two kinases that inhibit the sensors RIG-I and MDA5, and a PP2 regulatory subunit that binds the antiviral kinase PKR. Knocking them down restored strong interferon induction. In mice, inhibiting one regulator, DAPK6, lowered intestinal norovirus loads, though the hosts noted this says nothing about particle multiplicity. Barker called it a Goldilocks system: too little virus and nothing replicates, a middle dose triggers immunity, a high dose overcomes it. She suggested that a heavily infected cell will be cleared by CD8 T cells after presenting antigen on MHC, so damping interferon there may limit immunopathology. Racaniello doubted the authors' evolutionary rationale, noting a finding from the previous week that RIG-I binds host RNA without activating because host RNA lacks features such as a 5' triphosphate. He called the result a surprising non-monotonic phenomenon: more is less.

Neither host would call the threshold universal. Ramsey questioned whether imaging can prove a genome sits in the cytoplasm rather than a lysosome, and whether the 3.5-to-5.5 hour windows capture just one replication cycle. Barker praised the paper's careful use of the word guarantee: single particles can still infect, just not reliably. Racaniello said he doesn't think it is all or none. His sharpest objection concerned SARS-CoV-2, which in lung cells fuses at the plasma membrane and resists endosomal protease inhibitors like camostat, a poor fit for a vesicle-endocytosis model. Barker agreed the paper forgot about extracellular vesicles partway through.

The same episode covered other news. Moderna has an mRNA vaccine against Bundibugyo virus in a Phase 2/3 efficacy trial in the DRC, under six months after the outbreak drew global attention in May 2026, resting on preclinical and Phase 1 data and partnerships with CEPI and Gavi. The current response uses Ervebo, the Zaire Ebola vaccine, which is not a matched strain. The hosts weighed wider Ervebo access, a push associated with Médecins Sans Frontières, against a stepped rollout that preserves a comparison group. Separately, PAHO has begun reviewing whether US measles has circulated continuously for a year, the criterion for losing the elimination status won in 2000. Sequencing will now separate continuous spread from repeated introductions. The US has recorded over 5,000 cases and 10 deaths, all preventable by vaccination, and Racaniello noted the status can be regained by resuming vaccination.

The hosts want organoid or animal data before accepting the threshold as general. A companion lake study, resting on three lakes sampled once in summer 2023, carries similar limits.