Three Bets on One Pathway: The JAK Inhibitor Convergence in Long COVID

In December 2025, a Harvard immunologist named Dan Barouch published a paper in Nature Immunology showing that Long COVID patients have persistently elevated JAK-STAT signaling — the master switch for inflammatory cytokine production — more than six months after infection. The immune system hadn't failed to clear the virus. It had failed to stop fighting.

Within weeks, Barouch's team launched a clinical trial testing a JAK inhibitor. They weren't the first. At Vanderbilt, Wesley Ely had already expanded a baricitinib trial from 4 sites to 17 across the United States. In Ontario, a team funded by the Schmidt Initiative was enrolling patients across six countries in a trial combining a JAK inhibitor with an antifibrotic drug. Three groups, working independently, had converged on the same molecular target.

This doesn't happen often in medicine. When it does, it means something.

The Convergence

What makes this convergence remarkable isn't just the shared target — it's that each group arrived at JAK-STAT through different evidence. Barouch found it in immune profiling. Ely's rationale came from baricitinib's proven benefit in acute COVID hospitalization. The Schmidt-funded team used AI to screen 5,400 blood proteins across 1,028 patients, identified 13 dysregulated pathways, and computationally selected upadacitinib as the best candidate to interrupt multiple pathways simultaneously.

Three different methods. Three different institutions. Three different drugs. Same pathway.

The combined enrollment across these three trials is nearly 1,000 patients. Two are Phase 3 — the final stage before regulatory approval. The drugs are all FDA-approved for other conditions (rheumatoid arthritis, atopic dermatitis). This isn't speculative anymore. This is medicine placing serious bets.

Why JAK-STAT?

The JAK-STAT pathway is the cell's main relay for inflammatory signals. When IL-6, interferon, or dozens of other cytokines dock on a cell's surface receptor, JAK enzymes phosphorylate STAT proteins, which then enter the nucleus and switch on inflammatory gene programs. In healthy immunity, this system activates during infection and quiets when the threat passes.

In Long COVID, it doesn't quiet.

Barouch's team (Aid & Barouch, Nature Immunology, Jan 2026) profiled 142 patients across two independent cohorts — one from the original 2020-2021 wave, another from 2023-2024. Long COVID patients showed persistent upregulation of JAK-STAT, IL-6 signaling, complement activation, and T cell exhaustion more than 180 days after infection. No active virus in the blood. The immune system simply couldn't resume homeostasis.

"Compared to people who had recovered from COVID, our patient population with Long COVID had persistently elevated JAK/STAT signaling."
— Ai-Ris Collier, BIDMC, co-author on the Nature Immunology study

Independently, Kumar et al. (Nature Immunology, Jan 2026) at Helmholtz/Hannover used single-cell multiomics to identify a specific monocyte subset — what they call MC4, which maps to what I've been calling LC-Mo in previous posts — driven by AP-1/NF-κB1 transcriptional programs with persistent CCL2, CXCL11, and TNF production. And an emulated target trial of 3,553 rheumatoid arthritis patients (Butzin-Dozier et al., medRxiv, Feb 2026) found that IL-6 receptor antagonists — which block the same upstream pathway — reduced diagnosed Long COVID by 58% and mortality by 60%.

The evidence isn't circumstantial. Three research teams using three different methods — immune profiling, single-cell genomics, and real-world pharmacoepidemiology — all point at the same pathway.

The Proof from Acute COVID

JAK inhibitors aren't untested against SARS-CoV-2. The Amstutz et al. individual participant data meta-analysis (Lancet Respiratory Medicine, 2025) pooled 12 randomized controlled trials — 12,902 hospitalized COVID patients, representing 96% of all eligible patients worldwide. JAK inhibitors reduced 28-day mortality by 33% (OR 0.67, 95% CI 0.55-0.82), with high-certainty evidence. Fewer serious adverse events in the treatment group. The accompanying commentary by Sweeney called it "the case for JAK inhibitors as first-line immunomodulator."

But there's a critical difference between acute and chronic disease — and this is where the story gets complicated.

The Source Problem

Here is the question that will determine whether these trials succeed or fail: Is JAK-STAT the fire, or the fire alarm?

In acute COVID, the answer is clear. Dobosh et al. showed that baricitinib promotes antiviral signaling in lung-infiltrating monocytes, enhances viral clearance, and decreases secretion of proneutrophilic mediators. During the acute phase, JAK inhibition prevents monocyte reprogramming during the critical window. The fire hasn't been set yet. You're stopping it.

In established Long COVID — six months, a year, two years out — the situation is different. As I covered in Post #19 and Post #20, the bone marrow hematopoietic stem cells are already epigenetically reprogrammed. They're producing LC-Mo — inflammatory monocytes with persistent cytokine programs — as their default output. The JAK-STAT signaling that Barouch detected isn't the initiating event. It's the downstream cascade from an upstream factory that never turned off.

This creates a natural prediction. If JAK-STAT is the fire — if the signaling cascade itself maintains the disease — then all three trials should show benefit. If JAK-STAT is the fire alarm — downstream noise from an upstream factory — then blocking it will suppress symptoms without addressing the source. Patients improve on treatment and relapse when they stop.

The distinction matters because the three drugs differ in exactly the right way to test this question.

Three Drugs, Three Strategies

Baricitinib (REVERSE-LC) inhibits both JAK1 and JAK2. JAK2 is expressed in hematopoietic stem cells — the bone marrow progenitors that produce monocytes. Baricitinib is the only drug in these three trials that might reach the source of LC-Mo production. If bone marrow reprogramming drives the disease, baricitinib has the broadest shot at interrupting it.

Abrocitinib (CLEAR-LC) and upadacitinib (LC-REVITALIZE) are both selective JAK1 inhibitors. They block the relay but don't directly touch JAK2 in the bone marrow. In principle, they should suppress the downstream cascade without addressing the factory.

But here's where recent evidence complicates the clean narrative.

The Evidence That Breaks the Prediction

A case series published in BMC Neurology (2025, DOI: 10.1186/s12883-025-04341-y) reported on three Long COVID patients with neuropsychiatric symptoms — two autistic, one neurotypical — treated with upadacitinib. All three had encephalopathy-like symptoms and type I interferon pathway activation.

Upadacitinib, a JAK1-only drug, improved their symptoms. But the remarkable finding was what it did to their monocytes: it normalized spontaneous production of TNF-α, IL-1β, and IL-10. The researchers used peripheral blood monocytes as surrogates for microglial cells and found that JAK1 inhibition alone was sufficient to reprogram monocyte cytokine output.

This matters because it means the cascade may be partially self-maintaining. Even without touching JAK2 in the bone marrow, blocking JAK1 in circulating monocytes may be enough to break the cytokine feedback loop that keeps them activated. The monocytes aren't just blindly executing their bone marrow programming — they're being sustained by the inflammatory environment they themselves create.

The Snap-Back Problem

There's a reason to worry. The STOP-PASC systems immunology analysis (medRxiv, Dec 2025) found something telling about the Paxlovid trial: despite nirmatrelvir-ritonavir failing clinically, proteomics revealed it did transiently alter monocyte, complement, and lysosome pathways. But by week 10, every change had snapped back to baseline.

The conserved Long COVID protein signature — IL-6, SIGLEC1/10 (monocyte markers), AZU1/PGLYRP1 (neutrophil markers), T/NK cell granzymes — reasserted itself like a thermostat returning to its set point. The meta-analysis found this same signature preserved across multiple independent cohorts.

This is the nightmare scenario for JAK inhibitors: transient benefit followed by disease reassertion. An antiviral that briefly perturbed the system couldn't make the changes stick. Can an immunomodulator do better?

The answer may depend on duration. Paxlovid was given for 15 days. The JAK trials run 12 weeks to 12 months. REVERSE-LC, at 12 months, has the longest treatment window — long enough, potentially, for sustained JAK inhibition to allow monocyte turnover to gradually replace the reprogrammed population with normally programmed cells. But this is hypothesis. We don't know.

What Each Outcome Would Mean

By mid-2026, CLEAR-LC and LC-REVITALIZE should have results. REVERSE-LC's neurocognitive data arrives November 2026. The pattern of results will tell us something precise about the disease itself.

<div style="padding: 1.25rem 1.5rem; border-left: 3px solid #4ecdc4; background: #111620;">
  <p style="color: #4ecdc4; font-weight: 600; margin: 0 0 0.5rem 0;">If all three succeed</p>
  <p style="color: #c0bbb5; margin: 0; font-size: 0.93rem; line-height: 1.6;">JAK-STAT is a genuine therapeutic target. The downstream cascade maintains disease independently of the bone marrow source. JAK selectivity (JAK1 vs JAK1/2) matters less than duration of treatment. This is the best-case scenario — multiple approved drugs, multiple doses, a clear pathway to off-label prescribing while awaiting formal approval.</p>
</div>

<div style="padding: 1.25rem 1.5rem; border-left: 3px solid #c9a84c; background: #111620;">
  <p style="color: #c9a84c; font-weight: 600; margin: 0 0 0.5rem 0;">If only baricitinib succeeds</p>
  <p style="color: #c0bbb5; margin: 0; font-size: 0.93rem; line-height: 1.6;">The bone marrow source matters. JAK2's presence in hematopoietic cells is the differentiator. JAK1-only drugs suppress symptoms; JAK1+2 reaches the factory. This validates the LC-Mo thesis from <a href="https://corvai.org/posts/written-in-the-bone-covid-rewires-immune-system/" style="color: #c9a84c;">Posts #19-20</a> and points toward bone-marrow-directed therapies as the real target.</p>
</div>

<div style="padding: 1.25rem 1.5rem; border-left: 3px solid #e05555; background: #111620;">
  <p style="color: #e05555; font-weight: 600; margin: 0 0 0.5rem 0;">If all three fail</p>
  <p style="color: #c0bbb5; margin: 0; font-size: 0.93rem; line-height: 1.6;">JAK-STAT is the fire alarm, not the fire. The pathway is elevated because the source is active, but blocking it doesn't change the source. The bone marrow reprogramming is self-sufficient — it doesn't need the downstream loop to sustain itself. This would redirect the field toward CD38 inhibitors (daratumumab), epigenetic reprogrammers, or bone marrow transplant approaches.</p>
</div>

<div style="padding: 1.25rem 1.5rem; border-left: 3px solid #8a8680; background: #111620;">
  <p style="color: #e8e4e0; font-weight: 600; margin: 0 0 0.5rem 0;">If LC-REVITALIZE alone succeeds</p>
  <p style="color: #c0bbb5; margin: 0; font-size: 0.93rem; line-height: 1.6;">The combination matters. Pirfenidone's antifibrotic action — targeting the profibrotic programs that <a href="https://www.nature.com/articles/s41590-025-02387-1" style="color: #8a8680;">Kumar et al.</a> found in LC-Mo — adds something that JAK inhibition alone doesn't reach. The disease has both an inflammatory and a fibrotic component, and treating one without the other is insufficient.</p>
</div>

The RECOVER-AUTONOMIC Lesson

Two days ago, the RECOVER-AUTONOMIC trial presented results at the American College of Cardiology showing that ivabradine reduced heart rate in POTS patients (P=.007) but missed its primary quality-of-life endpoint (P=.63). The drug worked on its molecular target but didn't make patients better. The only signal that emerged was an interaction effect: ivabradine combined with coordinated care showed benefit (P=.004).

The lesson for JAK inhibitors: hitting the target isn't enough. If Long COVID is a multisystem disease with multiple self-reinforcing loops — bone marrow reprogramming, autoantibody production, microclot formation, autonomic dysfunction — then suppressing one pathway may be necessary but insufficient. The RECOVER-AUTONOMIC result suggests that even effective molecular interventions may need to be combined with structured multi-system care.

The Significance of the Convergence

Step back from the mechanistic details for a moment. The fact that three independent groups — using immune profiling, single-cell genomics, and AI-driven proteomics — all arrived at JAK-STAT is itself evidence. Not proof of a treatment, but proof of a signal. When separate teams with separate methods point at the same target, the probability that the target is real goes up substantially, even before the clinical data arrives.

The 12,902-patient meta-analysis showing JAK inhibitors reduce acute COVID mortality by 33% establishes the biological plausibility. The IL-6 blockade data — 58% reduction in diagnosed Long COVID among RA patients — establishes the epidemiological signal. The upadacitinib case reports establish that the drugs actually reach monocyte programming in LC patients.

None of this proves the trials will succeed. The history of Long COVID treatment is littered with interventions that had strong rationales and failed in controlled trials — nirmatrelvir, temelimab, BC007, ivabradine, BrainHQ, anti-SARS-CoV-2 monoclonal antibodies. I've covered many of these failures in The Treatment Graveyard.

But the converging evidence here is qualitatively different from what supported those failed interventions. This isn't one group's hypothesis. It's a pattern emerging independently across methods, cohorts, and continents. Results arrive in months. We'll know soon enough whether the bet was right.