The study, led by Michael J. Peluso and Dylan Ryder of the University of California, San Francisco, together with a broad consortium of collaborators at the University of Pennsylvania, Harvard Medical School, Aerium Therapeutics, and the Patient-Led Research Collaborative, was designed as a mechanistic trial rather than a conventional efficacy study. Thirty-six participants were enrolled and randomized in a 2:1 ratio to receive either a single infusion of AER002 or a placebo, with both participants and investigators blinded to allocation. After an intensive baseline characterization that included patient-reported outcomes, objective measures of physical and neurocognitive function, and a battery of blood-, imaging-, and tissue-based biomarkers, participants were followed for a full 360 days. The primary endpoint was the PROMIS-29 Physical Health Summary Score, a validated patient-reported measure of physical health, assessed at 90 days after infusion. The trial was registered as NCT05877508 and was funded in part by the Patient-Led Research Fund, with additional support from the PolyBio Research Foundation and the National Institute of Neurological Disorders and Stroke.

The choice of AER002 reflected the specific immunological reasoning behind the trial. The antibody is a monoclonal neutralizing antibody directed against SARS-CoV-2, engineered to bind the spike protein of the virus. In theory, if residual virus or viral antigen in reservoir sites such as the gut or other tissues was driving ongoing symptoms, a potent neutralizing antibody circulating at high concentrations could bind to and help clear that antigen, interrupting the pathological feedback loop. Monoclonal antibodies of this class proved their worth during the acute phase of the pandemic, reducing viral load and hospitalization when given early in infection. The Long COVID setting, however, poses a fundamentally different problem: the target is not a rapidly replicating population of virus in the airway but a poorly characterized, possibly intracellular or sequestered reservoir, and the therapeutic window is measured in months or years rather than days.

When the results were analyzed, the trial returned a clear and largely negative verdict on those hopes. There were no significant differences between the AER002 arm and the placebo arm in physical health as measured by the primary endpoint, in quality of life, in objective measures of physical function or cognition, or in blood-based biomarkers of immune activation and viral persistence. The consistency of the null result across both subjective and objective measures is notable, because mechanistic trials of this kind often show discordance between what patients report and what laboratory assays detect. Here, neither signal moved. The investigators emphasize that the study was exploratory and small, with only 36 participants, which limits its statistical power and means that a modest treatment effect in a molecularly defined subgroup cannot be definitively excluded. Still, as a proof of concept for the idea that a single dose of a spike-specific monoclonal antibody can ameliorate Long COVID symptoms, the trial did not deliver support.

The safety data, at least, were unambiguous. AER002 was well tolerated across the follow-up period, with no significant safety concerns identified between the treatment and control arms. That finding matters for the field because it establishes that potent anti-SARS-CoV-2 antibodies can be administered to people with chronic post-acute illness without provoking unexpected adverse events, including immune-complex-mediated reactions that some theorists had worried about when antibody binds to circulating viral antigen. For a condition in which patients have been desperate for any intervention at all, and in which unproven remedies circulate widely, the demonstration that a rigorously tested biological agent can be given safely is a small but genuine contribution, even when the efficacy signal is absent.

Beneath the headline null result, however, the trial produced one of the more intriguing post-hoc observations in the Long COVID literature to date. In an analysis not specified in advance, the researchers found that participants who entered the study with lower baseline levels of SARS-CoV-2 antibodies, specifically antibodies against the spike protein, the S1 subunit, and the receptor-binding domain, and who also achieved higher drug exposure after infusion, were more likely to perceive a treatment benefit on the Patient Global Impression of Change scale, with the associations reaching statistical significance at p less than 0.05 for all three antibody measures. The pattern is biologically coherent: patients with low endogenous anti-spike titers might be those whose immune systems have been least effective at controlling a persistent viral reservoir, and who would therefore stand to gain the most from an exogenous dose of neutralizing antibody. Higher drug exposure, meanwhile, would ensure that the antibody actually reached relevant tissue compartments in sufficient concentration.

The post-hoc finding must be interpreted with considerable caution. Post-hoc analyses in small trials are exploratory by definition, and the Patient Global Impression of Change is a subjective measure that can be influenced by expectation and the natural fluctuation of symptoms over time. With 36 participants divided across two arms, the subgroup analysis rests on a handful of individuals, and the possibility of chance findings cannot be dismissed. Nevertheless, the investigators argue that the signal is valuable precisely because it is hypothesis-generating: it suggests that future trials of antiviral or antibody-based approaches in Long COVID should stratify or enrich enrollment based on baseline humoral immunity, rather than treating the syndrome as a single homogeneous entity. If viral persistence drives disease in only a subset of patients, trials that enroll unselected populations will dilute any true treatment effect into statistical invisibility, a problem that has plagued the field broadly.

The trial’s mechanistic ambition also sets a template for how Long COVID studies should be conducted. Rather than relying solely on symptom questionnaires, the investigators collected objective measures of physical function, including cardiopulmonary exercise testing, neurocognitive testing batteries, imaging procedures, and optional gut biopsies, alongside an extensive panel of blood-based biomarkers. This integrated design means that even a negative efficacy result yields a rich dataset on the natural history of the syndrome and the relationship between molecular markers and clinical status over a full year of follow-up. The involvement of the Patient-Led Research Collaborative in the design and implementation of the study, with advice from patient researchers Lisa McCorkell and Hannah Davis, also reflects a growing recognition that Long COVID research benefits from the lived expertise of the affected community, both in choosing meaningful endpoints and in sustaining recruitment.

For the viral persistence hypothesis itself, the trial is neither a confirmation nor a refutation. A single monoclonal antibody dose is a blunt instrument against a reservoir whose size, location, and biology remain unknown. If persistent virus resides intracellularly or within immune-privileged sites that antibodies penetrate poorly, or if the relevant antigen is present at vanishingly low levels, then even a highly potent neutralizing antibody may fail to change the trajectory of disease. Alternatively, if viral persistence is only one of several converging mechanisms, alongside autoimmunity, microbiome disruption, latent virus reactivation, or vascular injury, then neutralizing the virus in a mixed population would predictably produce the kind of null result observed here. The authors themselves frame the trial as a foundation rather than a conclusion, stating that although AER002 was not efficacious in this proof-of-concept study, the findings could inform future trials using monoclonal antibodies to target viral persistence in Long COVID.

What comes next will likely be shaped by the trial’s two principal lessons. First, patient selection must become molecular: enrichment strategies based on baseline anti-spike antibody levels, viral antigen detection in plasma or tissue, or other biomarkers of active persistence could dramatically increase the signal-to-noise ratio of future antiviral trials. Second, dosing and duration may need rethinking, since a single infusion may be insufficient against a reservoir that has had years to establish itself, and combination approaches, for example pairing antibodies with antiviral drugs that act intracellularly, may be required. The AER002 trial, conceived and executed with unusual rigor and community partnership, closes one door while pointing clearly toward the next: the search for the right drug, given to the right patients, at the right biological moment. For the millions living with Long COVID, that search continues, now with better maps than before.

Subject of Research: A randomized phase 2a mechanistic trial testing the SARS-CoV-2 monoclonal antibody AER002 as a treatment for Long COVID

Article Title: SARS-CoV-2-specific monoclonal antibody AER002 in Long COVID: an exploratory randomized phase 2a mechanistic trial

Article References: Peluso, M. J., Ryder, D., Dalhuisen, T., Chu, D. H. T., Williams, M. C., Rodriguez, A. E., LaFranchi, B. H., Vinden, J., Fehrman, E. A., Huang, B., Hoh, R., Asare, K. A., Bellon Pizarro, K., Rahman, M. S., de Narvaez, E., Painter, M. M., Wherry, E. J., Swank, Z. N., Hansen, L. L., … Deeks, S. G. (2026). SARS-CoV-2-specific monoclonal antibody AER002 in Long COVID: an exploratory randomized phase 2a mechanistic trial. Nature Communications. https://doi.org/10.1038/s41467-026-77925-y

Image Credits: AI Generated

DOI: 10.1038/s41467-026-77925-y

Keywords: Long COVID, SARS-CoV-2, AER002, monoclonal antibody, viral persistence, phase 2a trial, PROMIS-29, Patient Global Impression of Change, anti-spike antibodies, mechanistic trial, Nature Communications, placebo-controlled