Perturb-PBMC: Functional Proteomics
of the PBMC Secretome

Figure 1: PBMC perturbation overview—stimulate PBMCs, perturb with cytokines across doses, quantify 191 secreted proteins with nELISA

nELISA to Map PBMC Cytokine Signalling

As part of our recent Nature Methods publication on nELISA, we mapped cytokine signaling in PBMCs. We’re now releasing this dataset, which we are calling Perturb-PBMC, for you to explore in the Portal.

Perturb-PBMC by the numbers:

• Samples: 7,392 PBMC supernatants
• Donors: 6 healthy individuals
• Stimulations: LPS, Poly(I:C), ConA, PMA-ionomycin (multi-dose)
• Perturbagens: 80 recombinant cytokines
• Readout: 191 secreted proteins (nELISA)
• Scale: ~1.4M protein measurements

Cytokine perturbagens x Cytokine read-out

Why run this screen?

Cytokines are pleiotropic: they act across pathways and produce complex, sometimes counterintuitive effects. They’re also both biomarkers (e.g., IFNγ-induced CXCL9/CXCL10) and drugs/targets (e.g., IL-2, GM-CSF).

Mapping their interactions therefore has direct translational value—for mechanism, biomarkers, and therapy design.

Global Structure of Responses

The interactive UMAP of all 7,392 samples shows rich structure: you can resolve donors, stimulation context, dose, and individual perturbagens. Zoom into subclusters, recolor by perturbagen, or facet by donor.

Figure 2: UMAP dimensionality reduction captures PBMC phenotypic diversity: Proteomic response profiles cluster by stimulus, donor, and dose, with several distinct clusterings driven predominantly by strong perturbagens.

Fold-Change Response Profiles Reveal Robust Cytokine Networks

Using response profile similarity analysis, we identified cytokines that produced consistent and biologically coherent response patterns across donors and perturbations.

Figure 3: Clustering of cytokine profiles highlights groups of perturbagens with similar effects. Heatmap dendrogram of perturbagen effects on cellular cytokine expression

After normalizing donor and stimulation effects, it’s clear that many perturbagens regulate IFNγ. Some act specifically (e.g., IL-21), while others are broad (e.g., IL-15). IFNβ (type I interferon) shows simultaneous pro- and anti-inflammatory effects. Notably, IL-4 (Th2) shares anti-inflammatory properties with IFNβ, clustering together despite their opposite effects on IFNγ.

This breadth is clinically relevant: interferon's pro-inflammatory properties are beneficial to treat some infections and tumors, whereas its anti-inflammatory components are beneficial in MS, but can undermine cancer immunotherapy.

Protein vs RNA: Complementary, Not Interchangeable

CytoSig is an RNA-based resource that infers cytokine activity from transcriptomic signatures, useful for mRNA-level effects. In contrast, nELISA directly measures secreted proteins, the functional read-out in question.

When looking at the same cytokine read-outs, we detected 447 cytokine interactions vs 137 in CytoSig, with 45 shared and ~87% directional agreement.

Figure 4: Proteomics reveals cytokine interaction insights transcriptomics missed. Correlation between cytokine interactions detected by nELISA and CytoSig, based on the fold change in expression of a protein in response to a perturbagen.

nELISA also revealed additional, nuanced effects absent in RNA (e.g. IL-1b suppression by Th1 cytokines, similarly to Th2 cytokines).

We also observed six interactions showing opposite responses at mRNA vs protein levels—likely signs of post-transcriptional regulation that remind us that cytokines often have unstable transcripts and regulated secretion.

Perturb-PBMC and Cytokine Repurposing Opportunities

Large-scale functional proteomics reveals unexpected relationships between cytokines. Here, IL-1 Receptor Antagonist (IL-1RA) and IFNβ both suppress innate immune responses, but only IFNβ induces pro-inflammatory mediators linked to flu-like side effects. Protein-level data suggest that IL-1RA, which IFNβ itself induces, mediates many of its therapeutic anti-inflammatory effects. The recombinant form, anakinra, already well tolerated in patients, could therefore offer a safer alternative for treating multiple sclerosis.

This illustrates how functional proteomics connects mechanism to therapy, enabling drug repurposing and explaining differential safety profiles that RNA alone can miss.

Figure 5: Novel putative responses to recombinant cytokines uncover actionable insights applicable to drug repurposing. Correlation plot of significant effects of IFNβ and IL-1 RA/RN on unstimulated PBMCs.