Transcriptomic analysis of functional versus desensitized natural killer cells

Natural Killer (NK) cell desensitization induced by persistent stimulation limits durable antitumor immunity, yet the molecular mechanisms governing this dysfunctional state remain poorly defined. To identify conserved regulators of NK cell desensitization, we performed comparative transcriptomic analyses across multiple murine models of persistent activation and dysfunction. We set up multiple robust in vivo and in vitro conditions to compare desensitized NK cells with their functional counterparts. First, we compared functional vs. desensitized NK cells that naturally arise through NK cell education. Normal lymphoid cells lacking MHC I are killed by NK cells, indicating that healthy cells provide stimulatory signals that are normally counterbalanced by inhibitory receptors recognizing self MHC I molecules, such as Kb and Db in C57BL6/J (B6) mice. In B6 mice, these receptors include Ly49C and Ly49I (bind to Kb), and CD94/NKG2A (binds to a Db-derived peptide presented by the Qa-1 nonclassical MHC I molecule). Individual NK cells can express any combination of Ly49C, Ly49I or CD94/NKG2A (abbreviated C, I and N, respectively), or none of them. In B6 mice, NK cells lacking all 3 receptors (C-I-N- NK cells; triple negative, TN) are inhibited the least and therefore persistently stimulated the most by neighboring cells in steady state conditions, resulting in a high degree of desensitization. NK cells expressing only one of the receptors, such as Ly49I+ single positive (SP) cells, are only partially inhibited and therefore less desensitized, and triple positive (TP) NK cells are inhibited the most and therefore least desensitized. Consequently, in response to activating receptor engagement ex vivo, the degranulation and cytokine production responses follow the pattern TP>SP>TN. The same pattern held when we gated on the most mature, CD27-CD11b+ NK cell subset. In MHC I-deficient B2m–/– mice, all NK cells are persistently stimulated and therefore strongly desensitized, similar to TN cells in B6 mice. We next established additional modes of desensitization, including 22-hour stimulation with ionomycin, a calcium ionophore that mimics receptor-induced Ca2+ influx and is known to induce a state of dysfunction in both T cells and NK cells. We confirmed that a 22-hour treatment with ionomycin desensitized murine NK cells. As yet another mode of desensitization, we reproduced persistent ‘induced self’ signals by simultaneously stimulating normal NK cells in vitro via engagement of multiple activating receptors for an extended period. NK cells were desensitized by stimulation for 48-hours with plate-bound antibodies targeting NKp46 in combination with NKG2D or DNAM-1, combinations that optimally activate NK cells. We performed bulk RNA-seq to compare desensitized NK cells and their functional counterparts in the aforementioned conditions: 1) SP vs. TP, TN vs. TP, or TN vs. SP NK cells in B2m+/+ mice (labeled "CIN"), 2) TP NK cells from B2m–/– vs. B2m+/+ mice (B2m), 3) NK cells treated with ionomycin vs. vehicle (iono), and 4) NK cells stimulated for 48 hours with combinations of activating receptor antibodies (Ab combo). Furthermore, in MHC I-deficient tumors, NK cells initially recognize the loss of MHC I and attack tumor cells, such that small tumor loads can be eradicated, but with larger tumor loads NK cells eventually become desensitized and lose the capacity to control the tumors. We further provide gene expression profiles of NK cells isolated from MC38-B2m–/– colorectal tumors as a representation of such case.