The Interferon Axis: Type I vs Type II Signalling Explained

Both interferon families run through STAT1, and that single shared subunit is why they are so often conflated — and why conflating them is a mistake. Type I interferon assembles STAT1 into a three-part complex with STAT2 and IRF9 that reads one class of promoter element. Type II interferon assembles STAT1 into a homodimer that reads a different one. Same phospho-STAT1 band on a blot; two largely non-overlapping gene programmes; two different reasons to block them. This article walks both arms from induction to output, and ends with the part that matters commercially: the fact that the interferon a successful T-cell response produces is also the signal that installs the brake it will later run into.

One STAT1, two complexes — the distinction everything else follows from

Type I interferon — IFN-α and IFN-β — engages IFNAR1 and IFNAR2, which carry TYK2 and JAK1 respectively. The phosphorylated STATs then assemble with a third, non-STAT subunit: IRF9. That trimer — STAT1, STAT2 and IRF9, collectively ISGF3 — recognises interferon-stimulated response elements, ISREs, because IRF9 supplies the DNA-binding specificity.

Type II interferon — IFN-γ — engages IFN-γR1 (CD119) with JAK1 and JAK2, and the product is a STAT1 homodimer, classically called GAF, which reads gamma-activated sequences. No STAT2, no IRF9, a different promoter grammar, and a substantially different gene set.

The practical consequence is worth stating plainly, because it is a common experimental error: phospho-STAT1 does not identify which interferon a cell is seeing. If you need to know, read one ISRE-driven gene — MX1 or IFIT1 — against one GAS-driven gene, such as CXCL9. It also explains the clinical genetics: STAT1 deficiency compromises antiviral and antimycobacterial defence, because it disables both arms at the one point they share.

Induction: where this map meets the rest of the library

Type I interferon is not constitutive; something has to sense a nucleic acid in the wrong place. Cytosolic DNA runs through cGAS and STING; endosomal dsRNA runs through TLR3 and CpG DNA through TLR9. Both converge on the IRF transcription factors, and the split between them sets the kinetics.

IRF3 is constitutively expressed and gives the first wave, dominated by IFN-β. IRF7 is present at low level and is itself an interferon-stimulated gene — so the first wave induces IRF7, and IRF7 drives a much larger second wave of IFN-α subtypes. That is a positive feedback loop with a transcription factor in it, which is exactly the architecture that produces an all-or-nothing response, and exactly the architecture that becomes pathological when the brakes fail.

IFN-λ (type III) is worth a line because it is often left out. It has its own receptor but converges on the same ISGF3, so it produces the same antiviral programme — only its receptor distribution is largely restricted to epithelium. An antiviral state at the barrier without the systemic cost.

The type I output: seven ways to make a cell a bad host

The ISG programme is not one mechanism but a portfolio, and knowing which is which changes what you measure:

Note the last row. ISG15 stabilises USP18, which is the brake on the very pathway that induced it — the negative feedback is built into the effector list, not bolted on beside it.

The type II output: visibility and resistance from the same signal

IFN-γ works through IRF1, and the gene set splits into three groups that pull in different directions. First, visibility: MHC class I and class II go up, so a tumour cell or a graft becomes easier for CD8 and CD4 T cells to see. (Class II induction runs through CIITA, which is the one node on this map with no product available anywhere in the catalogue — worth saying rather than papering over.) Second, recruitment: CXCL9, CXCL10 and CXCL11 are induced, and they draw CXCR3⁺ effector T cells into the tissue. Third, effector chemistry: NOS2 / iNOS in activated macrophages.

And then the fourth group, which is the one worth building a strategy around. The same IFN-γ induces PD-L1 and IDO1. This is adaptive resistance: the tumour is not pre-emptively hiding, it is responding to being attacked, and the brake it installs is a direct consequence of the attack. Two implications follow. A PD-L1 stain is, in part, a local IFN-γ readout — which is why CXCL9 and CXCL10 often track response to checkpoint blockade at least as well as PD-L1 itself. And an interferon-high tumour is not unambiguously good news; it is a tumour where both the attack and the counter-measure are running.

The brakes, and why they are not interchangeable

USP18 is the one most often mis-filed as "another SOCS". It is not. USP18 binds IFNAR2 directly and displaces JAK1 from the receptor — a steric, receptor-level block, and one that is specific to the type I arm. That is the molecular basis of type I desensitisation: a cell that has recently seen IFN-β is genuinely less able to respond to the next dose, and it is USP18 that makes it so.

SOCS1 and SOCS3 act further down, on the kinases, and are induced by both arms — classic negative feedback rather than arm-specific desensitisation. The clinical relevance of getting this right is considerable: sustained type I signalling with inadequate braking is the "interferon signature" of lupus and the monogenic interferonopathies, while too little of the same signalling is a failure of antiviral defence. The therapeutic target is the same molecule read in opposite directions.

Blocking interferon in vivo

Both arms can be closed at either the ligand or the receptor, and the choice matters. On the type I side, anti-IFN-β or anti-IFN-α neutralises one family member; anti-IFNAR1 closes the receptor and therefore takes out every type I subtype at once, including the ones you were not thinking about — usually what you want, given how many IFN-α subtypes exist. On the type II side, anti-IFN-γ (clones XMG1.2 and H22) or anti-CD119 at the receptor.

Two experimental cautions. Endotoxin: LPS is a TLR4 agonist and TLR4 signalling induces type I interferon. A contaminated blocking antibody therefore induces the pathway you are trying to block — not background noise but an effect pointing the opposite way to your hypothesis. Low- and ultra-low-endotoxin functional-grade material is not optional in interferon work. Timing: because the IRF7 loop amplifies, blockade given after the second wave has started is working against a much larger ligand pool than blockade given before it. Reporting the dosing schedule relative to the stimulus is not a formality here.

Frequently asked questions

What is the actual difference between type I and type II interferon signalling?

The complex STAT1 ends up in. Type I builds ISGF3 — STAT1 with STAT2 and IRF9 — which reads ISRE elements and switches on the antiviral ISGs. Type II builds a STAT1 homodimer that reads GAS elements and works largely through IRF1 to raise MHC and induce the CXCR3 chemokines.

Why does IFN-γ induce PD-L1 if it is meant to help the immune response?

Because PD-L1 induction is a feedback control on inflammation, not a tumour invention. The tumour exploits a normal brake: the T-cell response produces IFN-γ, IFN-γ induces PD-L1 through IRF1, and PD-L1 dampens the T cells. This is adaptive resistance, and it is why checkpoint blockade often works best where interferon signalling is already active.

Should I block the ligand or the receptor?

The receptor, if you want completeness. Anti-IFNAR1 removes signalling from every type I subtype, whereas neutralising IFN-α or IFN-β leaves the others intact. Ligand neutralisation is the better choice when you specifically need to attribute an effect to one family member.

What is USP18 and why does it matter more than it appears to?

USP18 binds IFNAR2 and displaces JAK1 — a receptor-level brake that only affects the type I arm. It is why prior type I exposure blunts the response to the next dose, and why type I and type II desensitisation behave so differently. Unlike SOCS1 and SOCS3, it is not a general kinase brake.

Which read-outs best report interferon activity in vivo?

Pick one from each arm. For type I: MX1, IFIT1 or ISG15. For type II: CXCL9 or CXCL10, both of which are secreted and therefore measurable in serum, plasma and urine. A single phospho-STAT1 measurement cannot distinguish the arms.

Is a strong interferon signature good or bad?

It depends entirely on which arm and for how long. Acute, resolving type I signalling is antiviral defence. Sustained type I signalling with failing brakes is the interferon signature of lupus and the interferonopathies. Type II signalling in a tumour raises MHC and recruits T cells, but installs PD-L1 and IDO1 at the same time.

Featured products for this pathway

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