Roughly 48 percent of participants (550 of 1,148) had at least one chronic virus detectable in the first 40 days after admission. Most of those had only one (372 of 550). EBV transcripts were present in 24 percent of participants near the time of admission. HSV1 turned up in 43 percent of endotracheal aspirates sampled around three weeks. The two families that dominated were Herpesviridae and Anelloviridae.

Viruses reactivate in tissue, not in blood. HSV1 lives in sensory ganglia, CMV in myeloid cells, EBV in lymphoid tissue. Sampling blood and airway almost certainly undercounts what was actually happening.

EBV peaked at admission and declined from there. Anelloviridae held steady until about day 20 and then tapered. CMV and HSV1 arrived late, peaking near three weeks and mostly in respiratory samples. The team replicated these curves in an independent whole-blood cohort.

Elevated EBV IgG was already present at admission, which places EBV reactivation before the hospitalization rather than after it. The wider consequence is that sampling time determines what a study finds. A single draw captures whichever virus is replicating that day and misses the others entirely, which is a reasonable explanation for why the prior literature disagrees with itself.

Detection of EBV, CMV, HSV1 and Anelloviridae all associated with COVID severity across the five trajectory groups, and the associations survived adjustment for circulating immune cell frequencies. Among the critically ill, detectable CMV, EBV or HSV1 in a respiratory compartment associated with death within a year. Only 17.4 percent of the Anelloviridae positive patients were on immunosuppressive medication, and the herpesviruses showed no association with immunosuppression at all.

This is association, and the authors put that in the abstract rather than the discussion, which is more than the press coverage managed. Severe illness plausibly causes reactivation rather than the other way around. The paper cannot separate the two and does not claim to.

Controlling for severity, age and sex, each virus carried its own cytokine signature. EBV in blood tracked with IL-6, IL-10, CCL2 and CXCL10. CMV and HSV1 shared a different set including IL-18, CXCL9, CXCL11 and TNF. The IL-10 finding is the one with a mechanism attached, since both EBV and CMV encode their own IL-10 mimics.

Severity adjusted is not severity free. Trajectory group is a coarse instrument, and residual confounding by "this patient was sicker" is the obvious alternative reading.

Reactivation during the acute illness did not predict which long COVID group a patient landed in. But Anelloviridae transcripts found in convalescent blood samples were significantly more common in the group reporting physical deficits, holding age, sex, immunosuppression and acute severity constant. The associated gene expression signature was the same in convalescence as it had been during the acute illness.

The patient reported outcomes were designed in 2020, before long COVID had an agreed definition, and may not map onto how we phenotype it now. Dropout was heaviest among exactly the patients who had reactivated, which cuts power where it was most needed. Anelloviridae were pooled at the family level because individual species were too sparse to model. And notably, EBV did not show the acute association that earlier antibody work had implied.