Article Text

Download PDFPDF

1010 A single-cell transcriptomic atlas of human NK cells to guide cancer immunotherapy
  1. Ryan King1,
  2. Azim Amirabad2,3,
  3. Amir Bayegan1,
  4. Joachim Theilhaber1,
  5. Nicole Acuff1,
  6. Shannon McGrath1,4,
  7. Xiangming Li1,
  8. Franck Rapaport2,5,
  9. Jack Pollard1,6 and
  10. Donald Jackson1
  1. 1Sanofi, Cambridge, MA, United States
  2. 2Digital Data Science, Cambridge, MA, USA
  3. 3Current Affiliation: Johnson and Johnson, Cambridge, MA, USA
  4. 4Current Affiliation: Moderna Clinical Biomarker Lab, Cambridge, MA, USA
  5. 5Current Affiliation: Sanofi Precision Medicine and Computational Biology, Cambridge, MA, USA
  6. 6Current Affiliation: Moderna Oncology, Cambridge, MA, US


Background Natural killer (NK) cell can serve as an effective anti-cancer treatment,1,2 and enhancing NK function has been shown to enhance patient outcomes.1,2 Therefore, insight into NK cell states and subtypes may lead to new treatment options. Initial characterization of NK cells relied on flow cytometry with surface markers dividing NK cells into less mature CD56bright NKs and more mature CD56dim NKs.4 More recently, single-cell RNAseq profiling has enabled deeper characterization and revealed three novel subtypes: cytokine-induced memory-like (CIML), adaptive, low ribosomal, and type I IFN responding NKs, which enhanced the previously known CD56bright, CD56dim, and CD56dim CD57+ subtypes.5 Here, we extend this single cell subtyping by building a cancer-focused NK cell Atlas that integrates 25 public datasets from multiple types of cancers and delineates both the subtypes and states of the NK cells.

Methods Single cell sequencing datasets were downloaded from the studies listed in [table 1]. NK cells were identified using the HaiTam cell type prediction algorithm v1.6 Harmony v0.1.07 was used for batch correction and ACTIONet v3.0.08 was used for data integration and cell state identification. Cell state abundance comparisons were performed using 1-way ANOVA with Dunnett’s post hoc test.

Results We generated an Atlas of 89,704 NK cells from 281 donors in 21 cancer focused studies, 3 studies using healthy donors, and 1 study of ulcerative colitis [table 1]. We identified 12 unique NK cellular states in our Atlas [figure 1]. Some states recapitulated known NK subtypes such as CIMLs and type 1 IFN responders. By contrast, CD56dim NKs were represented by 2 cellular states and CD56bright NKs were represented by 3 states. We found that one of these CD56bright states is more abundant in blood from renal cell cancer patients than in blood from healthy donors (2.1-fold change; p < 0.02). Compared to other CD56bright states, this overrepresented state expresses higher levels of the cytotoxic gene GZMK and of migratory markers CD44, CXCR3 and SELL.

Conclusions Our Atlas describes 12 NK cell states reflecting maturation, activation, and exhaustion in cancer. These states provide a framework for assessing co-expression of targets for NK modulators and for understanding the effects of treatments on NK cells.


  1. Chu J, Gao F, Yan M, Zhao S, Yan Z, Shi B, Liu Y. Natural killer cells: a promising immunotherapy for cancer. eBioMedicine 2020;59:102975.

  2. Suen W, Lee W, Leung K, Pan X, Li G. Natural Killer Cell-Based Cancer Immunotherapy: A Review on 10 Years Completed Clinical Trials. 2018;36:431–457.

  3. Abou-El-Enein M, Bauer G, Medcalf N, Volk H, Reinke P. Putting a price tag on novel autologous cellular therapies. Cytotherapy 2016;18:1056–1061.

  4. Moretta L. Dissecting CD56dim human NK cells. Blood 2010; 116:3689–3691.

  5. Smith L, Kennedy P, Stacey K, Worboys J, Yarwood A, Seo A, Solloa W, Mistretta B, Chatterjee S, Gunaratne P, Allette K, Wang Y, Smith M, Sebra R, Mace E, Horowitz A, Thomson W, Martin P, Eyre S, Davis D. Diversity of peripheral blood human NK cells identified by single-cell RNA sequencing. Blood Adv 2020;4:1388–1406.

  6. Talk2data []

  7. Korsunsky I, Millard N, Fan J, Slowikowski K, Zhang F, Wei K, Baglaenko Y, Brenner M, Loh P, Raychaudhuri S. Fast, sensitive and accurate integration of single-cell data with Harmony. 2019;16:1289–1296.

  8. Mohammadi S, Velderrain J, Kellis M. A multiresolution framework to characterize single-cell state landscapes. Nature Communications 2020;11:5399.

Abstract 1010 Figure 1

UMAP of the cellular states in the NK atlasEach unique cellular state identified in this study has a unique color assigned to it and has been plotted in the UMAP

Abstract 1010 Figure 2

Public studies used to construct the NK atlasThe names of each dataset used are listed along with identifiers and disease focus

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.