Long COVID Immune Changes at 12 Months: Early Cell Data

Long COVID can persist long after the acute infection ends, and one small Canadian study offers a detailed look at what may remain altered in the blood: depleted innate-like immune cells, activated and exhausted-looking T-cell populations, inflammatory signals, and a possible pathway for cell death. The finding is scientifically interesting. It is also early evidence, not a settled explanation for Long COVID or a reason to experiment with immune checkpoint drugs.

The strongest new claim comes from a 2026 single-cell study by Shahbaz et al. of women with Long COVID meeting criteria for ME/CFS, including post-exertional malaise. At least 12 months after confirmed COVID-19, the group showed a distinct peripheral-blood immune pattern compared with women who had recovered from COVID-19. The study points to the Galectin-9-TIM-3 checkpoint pathway as one possible contributor to the loss of certain T cells. It does not prove that this pathway causes Long COVID in the body, nor does it identify a treatment that reverses it.

A cell-by-cell map of persistent immune remodeling

The study enrolled 10 women with Long COVID/ME/CFS and 10 recovered female controls in Edmonton, Canada. All Long COVID participants had PCR-confirmed infection, substantial symptoms on the DePaul Symptom Questionnaire, and documented post-exertional malaise; people admitted to intensive care during acute COVID-19 were excluded. Researchers performed flow cytometry and plasma-protein measurements in all 20 participants. Full single-cell RNA sequencing of peripheral blood mononuclear cells—the circulating immune-cell fraction—used a subset of five participants per group.

That distinction matters. Single-cell sequencing can identify which cell types carry a transcriptional signal rather than averaging RNA across the whole blood-cell sample. But five people per group is a very small discovery sample. Individual differences can look like group differences at that scale, especially in a condition as heterogeneous as Long COVID.

Within that sample, the Long COVID/ME/CFS group had fewer mucosal-associated invariant T (MAIT) cells, gamma delta T cells, natural killer cells, and regulatory T cells. MAIT and gamma delta T cells are unconventional T-cell populations that respond rapidly to signs of infection or cellular stress. Natural killer cells help destroy infected or abnormal cells. Regulatory T cells help restrain immune responses. A change in any one of these populations does not tell us whether immunity is generally "weak" or "overactive"; immune systems are less like a battery meter and more like a crowded control room.

At the same time, classical effector and terminal-effector T-cell subsets were expanded and co-expressed PD-1, TIM-3, and LAG-3. These receptors are commonly discussed as markers associated with T-cell exhaustion, a state in which chronically stimulated T cells can lose function. Their co-expression supports an exhaustion-like phenotype, but blood-marker expression alone does not establish what those cells can or cannot do in a person's tissues.

The researchers also reported transcriptional reprogramming in monocytes, cells that participate in inflammation and tissue repair. Plasma C-reactive protein and serum amyloid A were elevated, alongside markers linked to microbial translocation or intestinal-barrier injury: soluble CD14, intestinal fatty acid-binding protein, and lipopolysaccharide-binding protein. The results describe a persistent inflammatory immune signature at the 12-month mark in this cohort. They do not show whether it began before Long COVID symptoms, causes those symptoms, or follows some other unresolved process.

Galectin-9 is a plausible lead, not a proven cause

The study's most specific mechanistic observation involved Galectin-9, a protein that can bind TIM-3. Plasma Galectin-9 was elevated in the Long COVID/ME/CFS group. In an ex vivo experiment, the researchers exposed participants' peripheral blood mononuclear cells to recombinant human Galectin-9 at 0.5 micrograms per milliliter overnight. MAIT and gamma delta T-cell subsets showed more Annexin V-positive apoptosis after exposure.

This experiment connects three observations: higher circulating Galectin-9, TIM-3 expression on affected immune cells, and Galectin-9's capacity to induce cell death in a dish. That is a credible hypothesis for why these cell populations might be depleted. It remains a hypothesis because an overnight exposure outside the body cannot establish what drives depletion in living patients. Elevated Galectin-9 could be upstream, downstream, or part of a feedback loop.

The gut-related markers give one alternative—or additional—explanation worth taking seriously. Microbial products entering circulation can activate monocytes and promote systemic inflammation. In this study, elevated soluble CD14 and intestinal fatty acid-binding protein occurred alongside the Galectin-9 signal. Without analyses that separate these pathways, the study cannot tell whether gut-barrier disruption drives immune activation and Galectin-9, whether persistent antigen contributes directly, or whether both processes occur together.

TIM-3 and Galectin-9 are already of interest in cancer immunology because immune checkpoints can regulate T-cell function. A 2025 review of TIM-3 in cancer describes it as a potential immunotherapy target in that very different setting. That overlap does not make checkpoint blockade a Long COVID treatment. Checkpoint drugs can produce serious immune toxicities, and no controlled Long COVID trial has shown that blocking TIM-3 or Galectin-9 restores MAIT cells, gamma delta T cells, or symptoms.

A companion study adds context, not independent confirmation

A 2025 companion study from the same research group used bulk RNA sequencing, flow cytometry, targeted proteomics, and hormonal profiling in an overlapping Edmonton female Long COVID/ME/CFS cohort. It reported sex-specific immune and hormonal dysregulation and higher expression of RELN, FEZF2, and BRINP2, genes linked in the paper to neuroinflammation and hematopoietic regulation.

The work supports the idea that the blood abnormalities are not limited to one assay. Still, overlapping participants and the same laboratory mean this is not an independent replication. Bulk RNA sequencing also averages signals across circulating cells, so it cannot identify which cell type produced a given transcriptomic difference. Nor can blood RNA sequencing show whether a neuroinflammatory signal reflects pathology in the brain, signaling between the immune system and nervous system, or changes in immune-cell development.

Why the findings need replication

No direct contradictory study in this evidence set shows immune recovery at 12 months or refutes the specific Galectin-9 finding. Absence of a contradiction is not confirmation. The central studies come from one laboratory, include women only, and rely on a small cohort. The single-cell analysis included five Long COVID participants and five recovered controls. The available report also does not stratify participants by viral variant, vaccination status, or prior infections—factors that could affect immune trajectories.

Peripheral blood adds another limit. It cannot directly measure immune cells in the lung, gut lining, brain-associated tissues, or other compartments where disease processes may matter more. A blood signature may be a consequence of tissue disease rather than its cause.

Useful next steps are straightforward, if not easy: larger mixed-sex cohorts from multiple centers; longitudinal sampling; tissue-relevant studies; and experiments showing that modifying the Galectin-9-TIM-3 pathway rescues cell survival. A clinical trial would need to pair immune endpoints with patient outcomes such as post-exertional malaise, fatigue, function, and safety. Lowering a biomarker is not enough if people do not feel or function better.

Breathing and rehabilitation address a different question

The topic is sometimes paired with a breathing-exercise trial, but it answers a separate clinical question. In a 2025 randomized trial in Thailand, 60 older adults with Long COVID were assigned to slow deep breathing or usual activity. The intervention involved 30-minute sessions, five times weekly for eight weeks. The trial assessed cardiovascular measures, heart-rate variability, respiratory-muscle strength, pulmonary function, physical performance, biochemical and oxidative-stress markers, and self-reported stress. It reported improvements in cardiopulmonary function, physical performance, and stress-related measures in the exercise group.

The trial did not measure MAIT cells, gamma delta T cells, natural killer cells, TIM-3, or Galectin-9. It therefore cannot show immune reconstitution. It was also unblinded and conducted at one center in older Thai adults, so its results should not be generalized automatically to everyone with Long COVID or to people with Long COVID/ME/CFS and post-exertional malaise.

Rehabilitation still has value for the outcomes it actually measures. A 2025 systematic review and meta-analysis of randomized trials found pulmonary rehabilitation improved several Long COVID outcomes, including exercise capacity, lung function, health-related quality of life, fatigue, and dyspnea. A 2024 meta-analysis similarly reviewed seven studies with 188 randomized participants and reported improvements in exercise capacity, dyspnea, fatigue, and peripheral muscle strength. A 2025 systematic review of physical-exercise programs found functional-performance benefits across a limited set of trials. None of these reviews establishes that rehabilitation reverses the immune-cell pattern described in the single-cell study.

The verdict: This is compelling early immune biology, not a treatment roadmap. The Galectin-9-TIM-3 pathway deserves independent replication and careful testing. For now, breathing and rehabilitation may help selected functional and cardiopulmonary outcomes, while no intervention has shown that it restores the specific immune changes reported at 12 months.