After Tocilizumab, IL-6 Goes Up. That Is the Drug Working.

Tocilizumab reverses cytokine release syndrome within hours, and it never touches the cell that started it. That sentence contains the whole mechanism. The trigger is a T cell — engaged by a CAR, a bispecific, or an infection. The syndrome is built by monocytes and macrophages downstream. And it funnels through one cytokine tightly enough that blocking a single receptor collapses the entire thing while leaving the therapy's anti-tumour activity intact. IL-6 is that cytokine, and the reason receptor blockade works so completely is that IL-6 signals two different ways through two different populations of cells. Click any protein for the matching Assay Genie ELISA kit or In Vivo antibody.

CRS end to end: the engaged T cell, the myeloid amplifier that does the actual damage, the IL-6 core with its classic and trans arms, the hepatocyte acute-phase readout, endothelial failure, and the measurement panel — including the free-drug assays for the three antibodies that treat it. Open the interactive version to click any protein for its role and the matching validated reagent.

The T cell pulls the trigger; the monocyte builds the storm

This is the single most useful thing to know about CRS, and it is why the standard treatment does not destroy the therapy. An engaged T cell — through CD3ε, against CD19 or CD20 — releases IFN-γ, IL-2 and GM-CSF within hours. Those act on monocytes and macrophages, which produce the bulk of the IL-6, IL-1β and TNF that cause the fever, the hypotension and the capillary leak.

Inside the monocyte the route is the inflammasome: NLRP3 assembles, caspase-1 cleaves pro-IL-1β and pro-IL-18, and the IL-1 arm fires. IL-10 rises alongside all of it — counter-regulatory, but here it is a marker of how big the response is rather than a sign that it is resolving, which is a distinction worth holding onto when reading a multiplex panel.

The evidence is not circumstantial. Serum IL-6 in severe CRS is overwhelmingly myeloid in origin, not T-cell. Deplete monocytes in a mouse CAR-T model and the syndrome largely disappears while the CAR-T cells still kill the tumour. That experiment is what settled the field, and it is the reason an anti-IL-6R antibody is a supportive-care drug rather than an antagonist of the treatment.

Classic versus trans-signalling — and why the target matters more than the ligand

Membrane IL-6R is restricted: hepatocytes, neutrophils, monocytes, some lymphocytes. That is classic signalling, and if it were the whole story IL-6 would be a fairly contained cytokine. ADAM17 changes that by shedding IL-6R into a soluble form. Soluble IL-6R plus IL-6 can then engage gp130, which is on every cell in the body — endothelium very much included. That is trans-signalling, and it is the arm that drives the vascular pathology.

Tocilizumab binds the receptor and blocks both arms. Siltuximab binds the ligand. Blocking at the receptor is the more complete intervention here, which is the same lesson the chemokine and VEGF maps teach from other directions. Downstream is short and drugged: JAK1 to STAT3, restrained by SOCS3, which is why JAK inhibitors are in trials for this indication. SOCS3 induction is also why IL-6 signalling is normally self-limiting — and part of why CRS is not.

The assay trap that catches everyone

After tocilizumab, serum IL-6 rises sharply — often several-fold. This is not treatment failure and it is not a paradox. Membrane IL-6R is a major route of IL-6 clearance; block the receptor and the ligand accumulates in the circulation because nothing is taking it out. The patient improves while the number you are watching goes up.

Two related timing errors are worth naming. CRP lags IL-6 by roughly a day, so a falling IL-6 with a still-climbing CRP is expected. And IL-2 peaks early, so a normal IL-2 on day three says nothing about day one. Serum amyloid A rises faster and further than CRP and is underused as a result.

When it stops being CRS

The acute-phase programme is written by STAT3 in the hepatocyte: CRP, SAA, fibrinogen, hepcidin and ferritin, with albumin suppressed. Hepcidin is the direct link from IL-6 to functional iron deficiency, which explains the anaemia of inflammation in a single step.

Fibrinogen is the one to watch, because it moves in both directions. It rises as an acute-phase protein and then falls if consumptive coagulopathy or secondary HLH supervenes. A dropping fibrinogen in a patient who looks septic is a warning sign, not reassurance. Ferritin above roughly 10,000 µg/L together with a high soluble CD25 and a very high IL-18 should prompt the question of whether this is CRS or HLH — because the management diverges at that point.

Then the endothelium fails. Angiopoietin-2, von Willebrand factor, ICAM-1 and VCAM-1 mark the transition from a febrile patient to a hypotensive one, and ANG-2 and vWF come from the same Weibel-Palade granules so they rise together. A high pre-treatment ANG-2:ANG-1 ratio identifies the patients who will get severe CRS before they get it, which is the kind of measurement that changes management. CCL2 and IL-8 track alongside and are among the analytes most consistently associated with neurotoxicity.

Reading the panel

The In Vivo angle — and what is not in it

Ten of the 37 nodes carry a functional-grade In Vivo antibody, and it is worth stating the gap plainly rather than talking around it: IL-6 and IL-6R are not among them. The central node of this map is measurable, not blockable, in the functional-grade range. Anyone modelling tocilizumab in a mouse needs to know that before they plan the study.

What is blockable is the layer either side. Upstream: CD3, CD19, CD20, IL-2 and IFN-γ — enough to build or interrupt the trigger. Downstream in the amplifier: IL-1β, TNF, GM-CSF, ICAM-1, and most usefully IL-1R1 (CD121a). That last one is anakinra's target and the second-line CRS drug, and it is genuinely blockable in vivo. Given that IL-1β rises before IL-6 in CRS models — which is why early anakinra can prevent the syndrome rather than merely treat it — the IL-1 arm may be the more interesting experiment anyway. Read IL-1Ra alongside IL-1β: the ratio sets the signal, not the ligand alone.

And unusually, the drugs themselves are measurable. Free-drug ELISAs exist for tocilizumab, siltuximab and canakinumab. That closes a loop most pathways leave open: before concluding that a patient or a model failed to respond, you can confirm the drug was actually there at the concentration you assumed. Under-exposure and true non-response look identical on a cytokine panel and completely different on a drug-level assay.

The one-line version

T cell triggers, monocyte amplifies, IL-6 carries it, and soluble IL-6R lets it reach every cell in the body. Block the receptor rather than the ligand, expect the IL-6 number to rise afterwards, and watch fibrinogen fall if it is turning into something else.

For the therapy that causes it, see CAR-T cell signalling and T-cell engagers and bispecific antibodies; for the TNF arm in detail, see the TNF receptor superfamily; and for the signalling module underneath, see cytokine JAK–STAT signalling.

For research use only. Not for use in diagnostic or therapeutic procedures.