Million-Cell Neutrophil Atlas Maps Immune States Across Disease

A million-cell atlas has revealed that human neutrophils form numerous transcriptional states that vary by tissue, infection, cancer, sex and metabolic health. The Single-Cell Atlas of Human Neutrophils, or SCAHN, integrates 1 million neutrophils from 781 samples and 43 datasets drawn from 40 studies, covering seven sequencing platforms, nine countries and more than 20 tissues and conditions. The researchers also analyzed 17,023 bulk blood transcriptome profiles from 40 datasets and two large electronic health record databases. This was a retrospective computational reanalysis of public transcriptomic data and de-identified clinical records rather than a new experiment, so its findings describe associations and biological patterns that still require prospective validation.

The atlas resolves broad developmental compartments before identifying finer states. Mature neutrophils expressed high levels of FCGR3B, CXCR1, CXCR2 and MME, whereas immature neutrophils expressed CEACAM8 and other granule genes. CEACAM8, widely used as a neutrophil marker, was detected only in immature cells in this analysis. Two immature groups were distinguished by granule programs: an AZU1 group expressing primary-granule genes including MPO and PRTN3, and an LTF group expressing secondary-granule genes such as LTF, LCN2, MMP8 and CAMP. AZU1 cells expressed proliferation genes including MKI67 and STMN1 and had the highest mitochondrial and ribosomal gene expression, consistent with a proliferating precursor state. A third granule-associated subset combined mature markers with the tertiary-granule gene MMP9.

Other clusters reflected specific immune programs. Three subsets expressed type I interferon genes, type II interferon genes or both, while six were named for prominent cytokine or inflammatory genes: CXCL, CCL3/4, NF-κB, IL1B, IL1RN and VEGFA. These cytokine and inflammation states were significantly more abundant in solid tissues than in whole blood or purified blood neutrophils. Additional clusters were characterized by MME, S100A4, IL1R2, AP-1, EGR1, PTGS2, G0S2, SLPI, heat-shock proteins, CD74 or TXNIP. Trajectory analysis placed immature cells upstream of mature blood neutrophils and tissue-associated inflammatory states. Immature subsets were also enriched in bone marrow, spleen and cord blood, consistent with these tissues as reservoirs of developing neutrophils, although granulopoiesis was not directly measured in every tissue.

To make the atlas applicable to conventional blood studies, the team derived five neutrophil-specific gene signatures covering immature, mature, degranulating, antiprotease and total neutrophil compartments. A separate signature summarized interferon-related genes. Comparison with 721,502 peripheral-blood immune cells from 200 samples showed that neutrophils contributed more than 78% of all reads for S100A8 and S100A9, more than 80% for several genes linked to immune suppression, metabolism or inflammation, and more than 50% for most interferon-related genes in the analysis. These scores provided a way to estimate neutrophil states from bulk blood transcriptomes, including datasets for which single-cell data were unavailable. The analysis also showed that neutrophils contribute substantially to expression of ARG1, OLR1 and CD274, genes associated in the paper with immune-suppression-related programs, although expression alone does not establish that a cell is functionally suppressive.

In infection, the atlas detected distinct patterns in blood neutrophils. Across seven infectious-disease single-cell datasets containing 167 samples, including 44 healthy or convalescent controls, 21 non-severe COVID-19 or influenza cases, 56 severe COVID-19 cases and 46 bacterial sepsis cases, the AZU1, LTF and MMP9 subsets were more abundant in severe COVID-19 and bacterial sepsis than in healthy or convalescent participants. Those increases were not seen in non-severe COVID-19 or influenza. Interferon-associated subsets increased in both non-severe and severe respiratory infections, while bacterial sepsis was associated with larger SLPI and IL1R2 populations. Expression of ARG1, OLR1, S100A8/S100A9 and CD274 was higher in severe COVID-19 and bacterial sepsis.

Independent bulk-transcriptome cohorts extended these findings. Across 3,381 patients from infection or critical-illness datasets, higher immature, degranulating and total neutrophil signature scores were associated with greater 28- or 30-day mortality risk after adjustment for age and sex. Some mature and antiprotease signatures were associated with severe disease but not with worse survival, illustrating why neutrophil heterogeneity may contain more information than a single abundance measure. Cancer analyses also connected transcriptional states with clinical measures. Among 131 samples from six cancer single-cell studies, four inflammatory subsets—VEGFA, CCL3/4, NF-κB and IL1RN—were more abundant in pretreatment tumors than in adjacent tissue. CD274 and several inflammatory markers were also elevated in cancer, whereas mature, degranulating and S100-family signatures were lower.

In 8,205 tumor biopsies from The Cancer Genome Atlas, higher expression of NF-κB, G0S2, VEGFA, SLPI, heat-shock and EGR1 signatures was associated with poorer overall survival or progression-free interval after adjustment for age, sex, stage and grade; the CD74 signature was associated with better survival. These are prognostic associations, not evidence that any signature caused the outcomes. In an analysis of the randomized SAVE-MORE trial, anakinra, an IL-1 blocker, was associated with lower 28- or 30-day mortality in the group with low baseline IL1B expression but not in the high-expression group. The finding was a retrospective, median-based subgroup analysis rather than a prospectively stratified treatment trial.

Cross-species mapping showed that several human neutrophil programs have counterparts in mice and nonhuman primates. Mouse immature, interferon-response and antigen-presentation hubs corresponded to the AZU1, interferon and CD74 human states, while mouse immunosuppression-related hubs mapped to several human inflammatory, immunosuppressive and antigen-presentation states. Mouse bone marrow was enriched for immature cells, and solid-tissue mouse neutrophils were enriched for inflammatory states, paralleling the human tissue pattern. Important differences remained: mouse cells more frequently expressed granule genes including LTF, LCN2 and MMP8/MMP9, while human cells more broadly expressed CXCL1 and S100A4. The comparison supports the use of mouse models for selected conserved programs while cautioning against assuming that detailed granule and cytokine responses are interchangeable across species.

Demographic and lifestyle factors also shaped the signatures. Across 33 healthy single-cell donors, females had a higher proportion of the combined type I and type II interferon subset, while males had higher CCL3/4 and CXCL subsets. Independent bulk studies of 977 healthy adults, including purified-neutrophil RNA sequencing, broadly reproduced higher immature signatures in men and higher mature and interferon signatures in women. In 822 Framingham Heart Study participants, adjustment for age, body mass index and smoking preserved those sex differences. Higher BMI was associated with higher immature, degranulating, antiprotease and total neutrophil scores, while smoking was associated with higher scores except the immature score. The observational design means these patterns cannot establish whether demographic or lifestyle factors cause the cellular changes.

Metabolic analyses linked neutrophil states to long-term risk. In Framingham data from 5,321 people, every neutrophil-specific score except the mature score correlated positively with glucose, glycated hemoglobin and triglycerides and inversely with high-density lipoprotein cholesterol; the relationships remained after people with diabetes were excluded. Scores were also higher in people already diagnosed with diabetes, and several declined over follow-up among participants who remained diabetes-free but increased or stayed unchanged among those who developed diabetes. After adjustment for age, sex, BMI, disease and smoking, immature, degranulating, antiprotease and total scores were associated with all-cause mortality over five, seven and ten years. The analysis also found associations with cancer diagnosis within seven years, but not with cardiovascular disease diagnosis.

Retrospective health-record analyses extended these findings from gene expression to routine blood counts. Among 205,817 Stanford patients with normal absolute neutrophil counts and no Charlson-index comorbidities, those in the highest quartile of counts were more likely to develop a chronic condition than those in the lowest quartile; the hazard ratio was 1.18, with a 95% confidence interval of 1.13 to 1.23. Intermediate counts carried intermediate risk. The highest-count group also had lower 10-year survival, with a hazard ratio of 1.98 and a 95% confidence interval of 1.62 to 2.42. Replication in 224,771 Arcadia patients produced a similar chronic-disease association, with a hazard ratio of 1.27 for the highest versus lowest count quartiles. These records cannot eliminate unmeasured differences between patients, including healthcare use or conditions outside the Charlson index. The atlas is therefore a reference for future work, not yet a validated clinical diagnostic or treatment-selection system; its strongest conclusion is that apparently similar blood cells can carry distinct programs with different relationships to health and disease.

Subject of Research: Single-cell mapping of human neutrophil states across health, infection, cancer and metabolic disease

Article Title: A single-cell atlas of human neutrophils reveals neutrophil heterogeneity across health and disease

Article References: Zheng, H., Maestri, E., Moore, A. R., Ganesan, A., Enriquez, J., Durán, V., Einav, S., van der Poll, T., Bouma, H. R., Atreya, M. R., Efron, P. A., Moldawer, L. L., Sevransky, J., Wright, D. W., Martin-Loeches, I., Karvunidis, T., Hinson, J., Hager, D. N., Rothman, R. E., … Khatri, P. (2026). A single-cell atlas of human neutrophils reveals neutrophil heterogeneity across health and disease. Nature Immunology. https://doi.org/10.1038/s41590-026-02671-8

Image Credits: AI Generated

DOI: 10.1038/s41590-026-02671-8

Keywords: neutrophils, single-cell RNA sequencing, immune heterogeneity, infectious disease, cancer immunology, metabolic health, immune suppression, mortality risk, transcriptomic signatures, human immunology, single-cell, atlas