Written in the Bone: How COVID Rewires Your Immune System at the Source

Three antiviral trials for established Long COVID. Three failures. Paxlovid, monoclonal antibodies, nirmatrelvir/ritonavir — none of them worked. The question isn't why they failed. The question is what they reveal about where Long COVID actually lives.

If Long COVID were driven by lingering virus, antivirals should help. They don't. PAX LC (100 patients, 15-day Paxlovid, Lancet Infectious Diseases April 2026): negative. AER002 (36 patients, SARS-CoV-2 monoclonal antibody): failed primary endpoint. STOP-PASC (nirmatrelvir/ritonavir): transient immune shifts that normalized by week 10.

The pattern is now incontrovertible. By the time Long COVID is established, the problem has moved. It's no longer in the virus. It's in the cells your body makes. Specifically, it's in the bone marrow — where your immune system's stem cells have been reprogrammed to continuously produce a new kind of inflammatory cell.

This post synthesizes six papers across three years to describe a vicious cycle that may be the central engine of Long COVID. It connects IFN-γ signaling, peroxisome destruction, microclots, heme metabolism, and mast cell neuropathy into one mechanism. The cell at the center of it all is a monocyte — and the instructions that made it wrong are written in the bone.

The Paradox: Monocytes Die in Days. Long COVID Lasts Years.

A circulating monocyte lives roughly one to three days in the bloodstream before migrating into tissue and differentiating into a macrophage. If inflammatory monocytes were the problem, the problem should burn itself out in a week. It doesn't.

In 2023, a team at Columbia used a technique called PBMC-PIE to enrich circulating hematopoietic stem and progenitor cells (HSPCs) from COVID patients' blood. What they found changed the framing of post-viral immune dysfunction entirely.

Severe COVID leaves lasting epigenetic changes in hematopoietic stem cells — changes that persist for months to at least one year after infection, and that are inherited by every monocyte those stem cells produce.

— Cheong et al., Cell, 2023

The mechanism: IL-6, surging during acute infection, reprograms HSPCs at the epigenetic level — histone modifications and transcription factor binding (IRF, AP-1, CTCF) that skew the differentiation program. The stem cells begin overproducing granulocyte-monocyte precursors (GMPs). Every monocyte born from these altered stem cells carries the inflammatory signature. The memory isn't in the monocytes — it's upstream, in the factory that makes them.

Critically, anti-IL-6 receptor antibody blocked this reprogramming in both human cells and mice. The stem cells could be protected — if you intervened early enough.

LC-Mo: The Cell That Emerged

In January 2026, Kumar et al. published the cellular identity of what those reprogrammed stem cells produce. They called it LC-Mo — a distinct monocyte transcriptional state found in Long COVID patients who had mild-to-moderate acute infections.

LC-Mo correlates with fatigue severity and respiratory symptoms. It is not a bystander. It is the effector — the cell doing the damage.

Source: Kumar et al., Nature Immunology, January 2026.

Mild COVID, Severe Disease — A Different Immunological Entity

A companion paper by Antar et al. in the same issue of Nature Immunology resolved a paradox that has confused clinicians since 2020: how can a mild acute infection cause disabling Long COVID?

The answer: it's a different disease. Mild acute COVID that progresses to Long COVID follows the TGF-β/WNT-β-catenin (LC-Mo) pathway. Severe acute COVID that progresses to Long COVID may follow entirely different immune mechanisms. They share the label "Long COVID" but they are immunologically distinct.

This matters therapeutically. A drug targeting the LC-Mo pathway would help the mild-acute cohort — potentially the majority of Long COVID patients — but might do nothing for the severe-acute cohort. It also explains why clinical trials enrolling mixed populations see diluted effects.

Source: Antar et al., Nature Immunology, January 2026.

The Fuel That Keeps the Fire Burning

If IL-6 programs the stem cells during acute infection, what maintains the program months and years later, long after the cytokine storm has subsided?

Heme.

In 2021, Jentho et al. demonstrated that free heme — released during tissue damage and hemolysis — induces trained immunity in both monocytes and HSPCs. The pathway is specific: Syk/JNK signaling (not mTOR, which means rapamycin wouldn't block it), producing H3K27ac histone acetylation changes that last far longer than the heme exposure itself.

This is where my earlier reporting converges. In Post #15, I covered the IMPACC study — a multiomics analysis of 500+ hospitalized COVID patients that found elevated heme metabolism as a leading signal predicting Long COVID. At the time, I noted heme was a biomarker. It's more than that. Heme is a damage-associated molecular pattern (DAMP) that actively reprograms bone marrow stem cells to produce more inflammatory monocytes. It's not a readout — it's a driver.

Source: Jentho et al., PNAS, 2021.

And in March 2025, Rehill et al. added the second output arm: trained monocytes don't just make cytokines — they make clots. Heme-trained macrophages show enhanced procoagulant activity through acid sphingomyelinase-mediated tissue factor decryption. Bone marrow HSPCs from trained mice also produce prothrombotic monocytes for more than four weeks — longer than a monocyte's lifespan, confirming the memory is central, not peripheral.

This directly links trained immunity to the microclot findings in Post #14. The 19.7-fold microclot elevation in Long COVID patients isn't just from NET-stabilized clots. The monocytes themselves are prothrombotic factories.

Source: Rehill et al., Science Advances, March 2025.

The Vicious Cycle

Here is the mechanism I've been building toward across six posts. Each element has been published independently. No single paper has assembled them into one cycle. This synthesis is mine.

The cycle works like this. During acute COVID, an IL-6 surge reprograms bone marrow stem cells. Those stem cells produce LC-Mo monocytes carrying profibrotic and prothrombotic programs. The profibrotic arm destroys peroxisomes in lung macrophages, triggering inflammasome activation and fibrosis. The prothrombotic arm activates tissue factor, generating microclots. Both arms produce tissue damage and hemolysis. The released heme acts as a DAMP, entering the bone marrow via the Syk/JNK pathway and reprogramming more HSPCs. The cycle is self-sustaining.

Additional inputs amplify the cycle: mast cells (activated via TLR4) pump IL-1β, IL-6, and TNF-α into the environment, pushing more monocytes into the LC-Mo state. DHEA-S depletion — documented in both Long COVID and ME/CFS — removes a key anti-inflammatory brake. And persistent IFN-γ in the blood, which I covered in Post #10, reflects an alarm the LC-Mo monocytes can no longer respond to: trained exhaustion.

Why Antivirals Can't Fix This

This framework explains the three antiviral failures cleanly. Paxlovid can inhibit viral replication. Monoclonal antibodies can neutralize circulating virus. Neither can erase H3K27ac marks on bone marrow stem cells. By the time Long COVID is established — months after acute infection — the epigenetic program is autonomous. The virus lit the match. The fire is now self-sustaining through a heme-driven trained immunity loop that doesn't require viral persistence to continue.

This doesn't mean viral persistence is irrelevant. SARS-CoV-2 RNA has been found in tissue reservoirs months after infection, and it may contribute to ongoing immune activation in some patients. But the antiviral trial data now clearly shows that eliminating residual virus is not sufficient to resolve Long COVID. The cellular machinery has been permanently altered.

The STOP-PASC systems immunology study makes this concrete. Maestri, Bonilla et al. profiled 152 participants with the Olink 5400-protein panel and found that nirmatrelvir/ritonavir caused transient immune changes at day 15 — myeloid and complement shifts — that completely normalized by week 10. A meta-analysis across 9 independent cohorts identified a conserved 60-protein Long COVID signature (including IL6, SIGLEC1/10, AZU1, GZMH), but antivirals didn't durably shift it.

What Can Be Targeted

The vicious cycle has multiple entry points. Several trials are — unknowingly or not — targeting different parts of it.

Target	Cycle Component	Trial / Drug	Status
IL-6 / JAK-STAT	HSPC reprogramming maintenance	REVERSE-LC (baricitinib)	550 pts, 17 sites
JAK + TGF-β	Upstream cytokines + fibrotic output	LC-REVITALIZE (upadacitinib + pirfenidone)	348 pts, 6 countries
NF-κB / TLR4	LC-Mo epigenetic driver	ADDRESS-LC (bezisterim)	200 pts, results H1 2026
IL-6 receptor	HSPC reprogramming initiation	Tocilizumab / sarilumab	Emulated trial: 58% lower LC risk
TGF-β alone	Fibrotic output	FIBRO-COVID (pirfenidone)	Failed — pirfenidone alone insufficient
Syk / JNK	Heme-driven perpetuation	Fostamatinib	No LC trial exists
Epigenetic marks	H3K27ac / trained memory	HDAC inhibitors / methyltransferase inhibitors	Preclinical only

Two things stand out. First, LC-REVITALIZE is the most mechanistically aligned trial currently running — it combines a JAK inhibitor (blocking upstream cytokine signaling) with pirfenidone (blocking downstream fibrotic output). The FIBRO-COVID trial showed pirfenidone alone wasn't enough; you likely need to shut off the upstream factory while also treating the downstream damage.

Second, there is a critical gap.

The Drug Nobody Has Tried

Fostamatinib is an FDA-approved Syk inhibitor (approved for immune thrombocytopenia). In a Phase 3 trial for acute COVID (280 patients, published in Science Advances), it reduced oxygen dependence from 7.6 to 4.8 days and — crucially — restored monocyte homeostasis. It increased nonclassical monocytes, increased HLA-DR expression on classical monocytes, and restored interferon responses. It normalized the very monocyte dysfunction that defines the LC-Mo state.

Fostamatinib would target the heme-driven perpetuation arm of the vicious cycle — the arm that no current trial addresses. Heme activates Syk. Syk drives the JNK pathway that reprograms HSPCs. Block Syk, and you may break the cycle at the point where tissue damage feeds back into bone marrow reprogramming.

No one has proposed this trial. No one has tested fostamatinib for established Long COVID. It is a novel therapeutic hypothesis — one that emerges only when you map the full cycle and notice which segment has no drug aimed at it.

What This Means

The LC-Mo framework doesn't replace what we knew. It organizes it. IFN-γ, peroxisome destruction, microclots, heme metabolism, mast cell activation — these are not competing hypotheses for Long COVID. They are outputs of the same machine: an epigenetically reprogrammed immune system that keeps producing broken cells from altered stem cells, powered by a heme-driven feedback loop.

This is why Long COVID persists. It's not that the virus won't leave. It's that the immune system was rewritten — and the instructions are in the bone marrow, where they are read out into every new monocyte, every day, for as long as the cycle runs.

Five days from now, on March 28, the RECOVER-AUTONOMIC trial reports results at the American College of Cardiology — the first randomized data on ivabradine for Long COVID POTS. In weeks, ADDRESS-LC bezisterim results are expected. Later this year, REVERSE-LC neurocognitive data arrives. Each of these trials targets a piece of the cycle. Whether they succeed or fail will tell us which parts of this machine are most vulnerable to intervention.

The answer to Long COVID will not be a single drug. It will be the right drugs hitting the right parts of the cycle. This framework — written across six papers, three years, and the bone marrow of millions of people — shows where to aim.