A team led by the Dell Medical School at the University of Texas at Austin, drawing on the multi-centre IMPACC cohort, shows that severe COVID-19 frequently reactivates chronic viruses that we all carry in a dormant state. Among 1,154 hospitalized and unvaccinated patients, 47.9% present at least one viral reactivation during the acute phase; the intensity of reactivation tracks disease severity, systemic inflammation and prognosis. The most novel result concerns the Anelloviridae, a near-ubiquitous viral family reputed to be harmless: their detection during convalescence is associated with the form of long COVID dominated by physical deficits (fatigue, disability). The authors insist: this is an association, not proof of cause.

Source: nature.com

In plain terms

We permanently host a small menagerie of "dormant" viruses — the mononucleosis virus (Epstein-Barr), cytomegalovirus, herpes, and a very common, benign family, the anelloviruses. Under normal circumstances our immune system keeps them in check. This study, covering more than a thousand patients hospitalized for severe COVID, shows that the illness can lower that guard: in nearly one patient out of two, at least one of these viruses "wakes up" and starts producing detectable traces again. The more severe the COVID, the more frequent the awakenings. And in people who retain physical sequelae long afterwards ("long COVID"), the anellovirus signature is found more often. This does not prove that these viruses cause long COVID — they may be no more than a witness to an overwhelmed immune system — but it offers a measurable lead for spotting, and perhaps one day treating, these lasting forms.

Fact sheet — Discovery

Technical explanation

1. The actual mechanism: what is a "reactivation"? A latent chronic virus maintains its genome inside our cells without producing new particles (silent phase). Reactivation is the resumption of viral transcription: the virus starts making its messenger RNAs again, and sometimes virions after that. The study does not detect the "sleeping" virus but precisely these reappeared transcripts (RNA), through RNA sequencing in three compartments: peripheral blood mononuclear cells (PBMC), nasal mucosa, and tracheal aspirate in ventilated patients. Detecting transcripts = catching the virus in activity, not merely its dormant presence.

2. A timeline specific to each virus. The study establishes that the awakenings are not synchronous. EBV (Epstein-Barr) reactivates early: 24% of patients have detectable transcripts as soon as days 1-8 (260/1,080), after which the signal declines. The Anelloviridae remain stable until day 20 before slowly decreasing. CMV and HSV1 arrive later (peak around 22 days) and above all in the respiratory compartments — HSV1 in 43% of tracheal aspirates (13/30). This distinct temporal signature is a strong argument: a technical artefact would strike all viruses in the same way; dynamics specific to each family betray a real biology.

3. What each method proves. Multi-tissue RNA-seq measures viral transcriptional activity in situ — hence reactivation, not mere seropositivity. EBV/CMV serology provides the control evidence: patients with CMV transcripts already had higher CMV seropositivity at enrolment (97.2% vs 76.1%, P = 0.0018), which confirms that the transcriptional signal does trace a reactivation in carriers rather than a primary infection. Cytokines (PEA) and the metabolome link each reactivation to the concomitant inflammatory state. Mixed modelling (random effects, with Benjamini-Hochberg correction for multiple testing) controls for confounders — age, sex, immunosuppression — and makes it possible to state that anellovirus remains associated with severity independently of these variables (P = 8.6 × 10⁻⁵).

4. The conceptual reversal. The dogma held that chronic viral reactivation was above all a consequence of immunosuppression (transplant patients, chemotherapy). Here, patients immunosuppressed by medication account for only 17.4% of the carriers of reactivated anelloviruses: the awakenings occur massively in immunocompetent subjects in a state of severe illness, correlated with heightened systemic inflammation. Reactivation is therefore not only a sign of collapsed defences — it accompanies, and could amplify, the inflammatory storm.

5. The link with long COVID — and its limit. In convalescence (≥ 2 months post-admission), anelloviruses and enteroviruses are the most frequently detected viruses. Anelloviridae transcripts are significantly more present in patients of the "physical deficits" cluster (high scores on the PROMIS physical function scale), even after controlling for sex, age, immunosuppression and acute severity (adjusted P = 0.012). In the acute phase, by contrast, no link with the sequelae clusters appears. The cautious reading: anellovirus is a candidate marker of a physically dominant long COVID, not a demonstrated cause.

Why it worked

The strength of the study is its scale and its integration: 1,154 patients, 20 centres, up to ten samples per patient over one year, and above all the superposition of six layers of data (host and viral transcriptome, cytokines, cytometry, metabolome, proteome) on the same individuals. An external validation cohort reproduces "strikingly similar temporal dynamics" for EBV, CMV and anelloviruses over the first 40 days — an internal replication test that rules out sampling chance.

The gap between the media announcement and the evidence must be held. The easy headline would be "dormant viruses cause long COVID"; that is not what the study shows, and the authors repeat it: "our results do not establish causality between viral reactivation and clinical outcomes". Two structural limits are acknowledged. First, systemic detection (PBMC) underestimates local reactivations in tissues: the absence of a blood signal is not the absence of reactivation. Second, the long-COVID analysis is weakened by attrition: chronic viruses are associated with mortality, so the most severely affected patients leave the convalescent cohort, reducing statistical power. The anellovirus–physical disability association (P = 0.012) is robust to covariate control, but it calls for replication before any clinical use.

Causal chain

We all carry 8 to 12 latent chronic viruses, kept in check by immunity → a severe illness (here COVID-19) triggers massive systemic inflammation → this inflammation, and not immunosuppression alone, lifts the control over dormant viruses → nearly one hospitalized patient in two reactivates at least one virus, with a timeline specific to each family → the intensity of reactivation tracks severity and partly predicts mortality → some viruses (anelloviruses) persist into convalescence → their presence is associated with the physical-deficit form of long COVID → a lead towards a prognostic biomarker and a potential therapeutic target (to be validated) → beyond COVID, an open question: is viral reactivation a general mechanism of post-infectious sequelae?

Anecdote

Anelloviruses are the paradox of this story. Their flagship member was discovered in 1997 by Nishizawa and colleagues in a Japanese patient with post-transfusion hepatitis of unknown origin; it owes its "TT" abbreviation to that first patient's initials (the Latin name Torque teno virus came afterwards). They are so widespread that they are regarded as permanent and benign passengers of human blood — to the point of serving as a marker to gauge the state of the immune system in transplantation: the lower the immunity, the higher their load climbs. That this discreet virometer should re-emerge here as a possible signature of long COVID illustrates a reversal: the virus once used to measure immunity could also bear witness to its lasting dysregulations.

Legacy and current data

The WHO has recorded more than 774 million COVID-19 cases and nearly 7 million deaths; long COVID affects a substantial fraction of survivors, with no biomarker validated to date for distinguishing its subtypes. That is precisely the gap this study begins to fill: not a treatment, but immune, transcriptomic and metabolic signatures associated with reactivation, which could one day serve to prognosticate the lasting forms. Since the IMPACC cohort enrolled unvaccinated patients in 2020-2021, transposability to the post-vaccine era and to recent variants remains to be established.

The researcher's view — open questions

(Interpretation, not results of the study.) Three experiments would settle the status of anellovirus. The fine-grained timing test: does the anellovirus load rise before the physical deficits set in, or only at the same time? Temporal precedence (a tight longitudinal design) is the first step out of mere correlation towards causality. The intervention test: in patients with a high persistent anellovirus load, does immune or antiviral modulation change the trajectory of symptoms? That is the trial that would distinguish a passive witness from an actor. The generality test: is the anellovirus–physical sequelae signature also found after other severe infections (influenza, sepsis)? If so, what we hold is not a peculiarity of SARS-CoV-2 but a general mechanism of degraded convalescence — a hypothesis with far heavier consequences.

Sources

References verified during the fact-checking audit (12 August 2026). Full text (abstract, body, numerical values and timelines) consulted on nature.com — open access article.

• Maguire C., Chen J., Rouphael N., Morse B. A., Langelier C. R., Melamed E. et al. (IMPACC Network). Virus reactivation in acute and long COVID-19. Nature, online 5 August 2026, open access. DOI: 10.1038/s41586-026-10740-z — peer-reviewed.
• Dataset: Maguire C., Morse B. A. & Melamed E., Zenodo, 10.5281/zenodo.19657241 (2026).
• Editorial context: Nature News, "COVID can wake up a slew of dormant viruses inside you", d41586-026-02443-2.

Background references

(Historical context, distinct from the study's primary sources.)

• Nishizawa T. et al. A novel DNA virus (TTV) associated with elevated transaminase levels in posttransfusion hepatitis of unknown etiology. Biochem. Biophys. Res. Commun., 1997, 241(1), 92 — discovery of the TTV anellovirus, named after the first patient's initials.

Confidence statement

• Solidly established (measured): 47.9% acute reactivation; distinct timelines by virus; severity/mortality associations at the adjusted P values cited; anellovirus–physical deficits association (P = 0.012) after covariate control; replication in an external cohort. Open access data, Zenodo deposit.
• Not established (and flagged as such by the authors): any causal relationship between reactivation and clinical outcome. The regime is strictly observational → wording capped at "associated with".
• Acknowledged limits: systemic detection (PBMC) underestimating tissue reactivations; reduced power in convalescence due to mortality-related attrition; unvaccinated 2020-2021 cohort, post-vaccine transposability to be established.
• Technical checks: DOI and Zenodo deposit resolve; numerical values extracted from the full text; dates consistent (publication 5 August 2026).
• Uniqueness verdict: UNIQUE. Key doi:10.1038/s41586-026-10740-z absent from the registry; no prior COVID/virology/immunology article on the site.
• What the article adds: informs the reader of a major result on the candidate mechanisms of long COVID, while firmly holding the association/causation distinction that media coverage tends to erase.