Little cell-type specificity in the data from Cui et al. (2024) . The conclusions that IL1β induces distinct gene programs in almost every cell type (Cui et al., 2024) conflicted markedly with present observations (and with conclusions from our previous work [Baysoy et al., 2023]) that responses overlapped highly between cell types. This discordance was surprising because the experimental strategy of injecting cytokines in vivo was similar between the two studies, with minor technical differences (i.v. versus s.c. injection, 2- versus 4-h harvest, spleen versus LN) and mostly differing by the type of RNAseq, sorted cell types versus pseudobulking of single-cell RNAseq of cell types defined by clustering). We thus reanalyzed the original data from Cui et al. (2024), both primary data obtained from the authors and the genes selected as cell-type specific in Fig. 2 B and Table S4, which were used to support the claims of cell-type specificity. (A) Comparison of population RNAseq (this paper) and single-cell RNAseq results (Cui et al., 2024), effects in naive CD4+ T cells or in monocytes. Note the generally good overlap between the effects of IL1β revealed by both studies, albeit with some dropout of transcripts in the single-cell data. Thus, the discordant conclusions did not stem from the data in themselves. (B) Closer analysis of the IL1β response in CD4+ T cells. The primary data obtained from the authors showed a marked response, with 406 transcripts significant at adjusted P value <0.05 (green highlights, top panel). This response matched well with genes identified as IL1β-responsive in the present study (red and blue highlights, middle panel), although some IL1β-responsive transcripts were missed in the single-cell data, mostly those with signal intensity <0.1 cp10K), as expected from the poorer sensitivity of single-cell data at low expression values. A more important issue, however, was that the genes listed in Table S4 of Cui et al. (2024) represented only a small proportion of all induced genes that passed Padj <0.05 in the single-cell data. This selection resulted from downsampling introduced in the analysis of Cui et al. (2024) to compare matched numbers of cells; unfortunately, it resulted in a strong loss of power to compare differentially expressed genes (DEGs) in different cells. (C) FoldChange/FoldChange comparison of the response to IL1β in CD4+ and CD8+ T cells (data from Cui et al. [2024]), showing a quasicomplete concordance between the two cell types, contrary to the representation in Table S4, which listed 40 genes uniquely responsive in CD4+ or CD8+ T cells, versus 25 shared. Most problematically, transcripts listed as specifically induced in either cell (red and green highlights) were all clearly induced in both. (D) As in C, comparing responses to IL1β in plasmacytoid dendritic cells (pDC) and migratory dendritic cells (MigDC). Here again, the vast majority of transcripts represented to be specifically induced in one cell or the other were clearly induced in both (especially for genes identified as MigDC specific), albeit with quantitative differences. (E) Same comparative analysis for the response to IL2 in natural killer (NK) cells. Every single transcript claimed to be NK specific in Table S4 was found in the upper right quadrant, denoting induction in CD8+ T cells, with less than a twofold difference in induction (dashed lines), and sometimes even stronger in CD8+ T cells (the same applies to genes listed as IL2-responsive specifically in CD8+ T cells, although in this instance ∼15% do appear specific). (F) Simple recomputation of cell-type specificity in data from Cui et al. (2024). For each cell type, the number of genes induced by IL1β was determined (FoldChange >2 and adjusted P value <0.05). Among those, genes induced by >1.3 in any one of the other cell types were counted (note that this conservative estimation under-represents sharing in the low response range). On average, 20–30% of transcripts were cell-type specific, with the exception of responses in neutrophils and mast cells. In summary, this reanalysis shows that cell specificity displayed in Fig. 2 B and Table S4 of Cui et al. (2024) is misleading and is weakened or eliminated once actual inductions and repression values are considered, especially when comparing cell types within a branch. The misrepresentation stems in part from the inherently lower sensitivity of single-cell RNAseq to detect changes (B), but mostly from deducing cell specificity non-quantitatively by intersecting lists of DEGs derived at arbitrary statistical cutoffs (the common danger of Venn diagrams), which was compounded here by the downsampling procedure that had to be applied for cell–cell comparisons but further affected discriminating power. While there is certainly some cell-type specificity to cytokine responses, and some of this specificity can have important functional consequences, the data of Cui et al. (2024) equally support our claim of extensive redundancy between responses.