The Inhibitory NKR-P1B:Clr-b Recognition Axis Facilitates Detection of Oncogenic Transformation and Cancer Immunosurveillance

Natural killer (NK) cells represent a subset of innate lymphoid cells (ILC) with the capacity to recognize and eliminate a variety of pathologic target cells, including transformed, infected, transplanted, antibody-coated, and stressed cells. NK cells discriminate healthy ‘self’ cells from malignant altered-self or non-self targets through interactions between a variety of germline-encoded NK-cell receptors and their cognate ligands on target cells, which are modulated during cellular pathologies (1). The missing-self hypothesis was initially formulated on the basis of the observation that class I MHC (MHC-I)-deficient tumor cells are more efficiently recognized and eliminated by NK cells (2). More recently, this paradigm has been expanded to include both induced-self (3) and MHC-independent (4) modes of NK-cell recognition.

One MHC-independent recognition system is the inhibitory NKR-P1B:Clr-b interaction (5). Both the inhibitory NKR-P1B receptor and its cognate Clr-b ligand are type-II transmembrane C-type lectin-related proteins encoded within and genetically linked to one another in the NK gene complex (NKC; refs. 6, 7). Clr-b is broadly expressed on most nucleated hematopoietic cells, similar to MHC-I molecules (7, 8). Moreover, the loss of Clr-b has been shown to be involved in missing-self recognition of tumor cells (6, 7), cells infected with cytomegaloviruses (RCMV, MCMV) or poxviruses (Vaccinia, ectromelia; refs. 9–13), as well as cells undergoing genotoxic or cellular stress (14). More recently, the generation of Clr-b^−/− and NKR-P1B^−/− mice have revealed a role for this system in hematopoietic transplants, where NKR-P1B:Clr-b interactions function to inhibit NK-cell–mediated rejection of healthy syngeneic and allogeneic cells (8, 15). On the other hand, cytomegaloviruses have been shown to encode novel decoy or Clr-like surrogate ligands used to evade innate immune detection (9, 11–13). Moreover, enforced maintenance of Clr-b ligand levels during tumor development has been postulated to play a role in immunoediting and immune escape of malignant cells in the Eμ-cMyc spontaneous lymphoma model, whereby this selective pressure is mitigated in Nkrp1b^−/− receptor–deficient mice (15). This prompted a study of Clr-b modulation during oncogenic transformation and immune surveillance in Clr-b^−/− ligand–deficient mice.

The Ras family of proto-oncogenes was first discovered as transforming elements of the Harvey and Kirsten strains of murine sarcoma viruses in mice and rats (16). Subsequently, mutated alleles of RAS genes were identified as dominant oncogenes in various types of human tumors, with approximately 30% of human malignancies bearing activating Ras mutations (16). The Ras proteins are small GTPases that act as molecular switches by cycling between a GTP-bound active state and a GDP-bound inactive state. Three mammalian ras genes (H-Ras, N-Ras, and K-Ras) functionally encode the founding members of a larger family of at least 35 related isoforms (16). These common isoforms are highly homologous except for the C-terminal 24–25 amino acids, collectively known as the hypervariable region (HVR). It is generally accepted that functional differences among these Ras isoforms are attributed to the HVR (16). They exhibit distinct posttranslational modifications, trafficking routes, and localization in the plasma membrane, and mutations in each isoform are associated with specific types of tumors (16).

NK cells are known to detect and eliminate transformed and tumor cells through the integration of signals delivered by various stimulatory (e.g., DNAM-1, NCR, NKG2D) and inhibitory receptors (e.g., Ly49/KIR, CD94/NKG2A, NKR-P1B; refs. 17–19). Thus, a “missing-self” lack of inhibitory ligands, such as MHC-I molecules, or an “induced-self” upregulation of stimulatory proteins, such as NKG2D ligands (NKG2D-L; ref. 20) will alter the balance of signals in favor of NK-cell–mediated target cytotoxicity. Oncogenic transformation has been demonstrated to induce innate immune recognition in several models, including the Eμ-cMyc spontaneous lymphoma model, in which NKG2D-L were shown to be upregulated upon lymphoma development, in turn increasing susceptibility to NK cytotoxicity (21, 22). NKG2D-L were also found to be induced via the DNA damage response pathway on primary tumors derived from oncogene-transformed cells in vitro (23). Recently, it was shown that the constitutive-active H-Ras^G12V oncogene upregulates NKG2D-L, including Rae1α/β (on mouse fibroblasts), and MICA/B and ULBP1-3 (on human cell lines), via the Raf–MEK–ERK and PI3K pathways (24). In NIH3T3 fibroblasts, Ras-mediated oncogenic transformation was also shown to promote a loss of cell surface MHC-I molecules (25), which serve as ligands for inhibitory Ly49 receptors.

How transformation or primary tumorigenesis influence Clr-b expression are currently unknown. Here, we show that overexpression of wild-type (WT) or constitutive-active (CA), but not dominant-negative (DN), Ras isoforms downregulate surface Clr-b levels on NIH3T3 fibroblasts and C1498 leukemia cells. The loss of surface Clr-b coincides with a decrease in nascent Clr-b (Clec2d) transcripts and promoter activity. Similar to NKG2D-L induction, Ras^G12V-mediated Clr-b downregulation occurs via the Raf–MEK–ERK and PI3K pathways. Specifically, shRNA-mediated MEK1/2 double-knockdown or ERK2 knockdown reversed Ras^G12V-mediated Clr-b loss. Overexpression of various ERK2 mutants show that both kinase-dependent and -independent functions of ERK2 are involved. Functionally, loss of Clr-b surface protein also diminishes recognition of Ras-transformed cells via the NKR-P1B inhibitory receptor, in turn enabling “missing-self” cytotoxicity of leukemia target cells by NK cells in vitro and enhanced rejection by WT mice in vivo. Finally, genetic crosses involving Clr-b^−/− ligand–deficient mice with Eμ-cMyc transgenic mice demonstrate enhanced survival upon ablation of either one or two Clr-b alleles, whereas selective rejection of Clr-b^−/− hematopoietic cells was previously observed following transplantation into naïve WT mice in vivo (8, 15). Taken together, these findings demonstrate an important role for the inhibitory NKR-P1B:Clr-b recognition system in the detection of oncogenic transformation, innate immunosurveillance, and the immune escape of primary lymphoma cells in vivo. These results suggest that the corresponding human NKR-P1A:LLT1 interaction may be a valid target for combinatorial immune checkpoint therapy.

C57BL/6 (B6) and B6.CD45.1 mice were purchased from Jackson Laboratories. FVB/N mice were purchased from Taconic. Eμ-cMyc mice originally purchased from Jackson Laboratories were transferred from Terrence Donnelly Centre for Cellular & Biomolecular Research (Toronto, Ontario, Canada). B6.Clrb^−/− (Clr-b^−/−; refs. 8, 26) and Eμ-cMyc mice (27) have previously been described. Eμ-cMyc mice were crossed to Clr-b^−/− mice to generate Eμ-cMyc-transgenic Clr-b^+/+, Clr-b^+/−, and Clr-b^−/− cohorts for the primary lymphoma progression model. Experiments were performed and mice maintained under protocols approved by Animal Care Committee at Sunnybrook Research Institute (Toronto, Ontario, Canada) in compliance with guidelines from the Canadian Council on Animal Care.

NIH3T3 cells and C1498 cells were purchased from ATCC. HEK 293T and BWZ.36 cells were obtained from Drs. David Raulet and Nilabh Shastri (UC Berkeley, Berkeley, CA; ref. 28), and authenticated by flow cytometric analysis and β-galactosidase reporter cell activity, respectively. BWZ.CD3ζ/NKR-P1B cells were generated previously (7). All cells were cultured in complete DMEM-HG, supplemented with 2 mmol/L Glutamax (Life Technologies), 100 U/mL penicillin, 100 μg/mL streptomycin, 50 μg/mL gentamicin, 110 μg/mL sodium pyruvate, 50 μmol/L 2-mercaptoethanol, 10 mmol/L HEPES, and 10% FBS. Ex vivo cells were cultured in supplemented RPMI. All cell lines were tested for Mycoplasma (MycoAlert; Lonza) at the outset of the studies.

SB203580, U0126, FR180204, SP600125 and PD98059 were purchased from Sigma-Aldrich, BIX02189 from Cedarlane Laboratories, Ly294002 from Calbiochem, and CI-1040 and FTI-277 from Tocris Bioscience. Chemicals were dissolved according to manufacturer's instructions in DMSO, water or PBS. Twenty-four hours after transfection, media were replaced and the cells were rested for 4 hours before dosing with inhibitors or vehicle alone for 24 hours before flow cytometry.

Selected CA, dominant-negative DN, and WT pathway vector sets for Ras and Raf were purchased from Clontech. Cotransfection of these vectors was performed with internal control pCMV-GFP-NLS at a 10:1 ratio. The bicistronic pBudCE4.1 (Thermo Fisher Scientific) modified to harbor eGFP downstream of the CMV promoter was a gift from Dr. Eleanor Fish (University of Toronto, Toronto, Ontario, Canada), which was then modified to express turboRFP (tRFP) and used to express ERK2 mutants under the EF-1α promoter. Ras and ERK isoforms were cloned using gene-specific primers described in Supplementary Table S1, employing Gene SOEing. All cloning was performed using the Expand^PLUS High Fidelity PCR System (Roche), and all products were confirmed by sequencing.

shRNA for MEK1, MEK2, ERK1, ERK2, and ERK5 in the pLKO.1 lentiviral vector were obtained from The RNAi Consortium (University of Toronto), with the following targeting sequences: MEK1 (Map2k1) GCC ATC CAA CAT TCT AGT GAA; MEK2 (Map2k2) CTT CCA GGA GTT TGT GAA TAA; ERK2 (Mapk1) GAC ATG GAG TTG GAC GAC TTA; ERK1 (Mapk3) CAA CAC CAC CTG CGA CCT TAA; ERK5 (Mapk7) AGA CCC ACC TTT CAG CCT TAA. Proviral pLKO.1 vectors were cotransfected with the packaging vectors. Supernatants were collected after 48 hours, then NIH3T3 fibroblasts were transduced for 24 hours. Stable transductants were analyzed after 5 days of selection with 2.5 μg/mL puromycin.

NIH3T3 fibroblasts were plated in 6-well plates 4 hours prior to transfection. H-Ras WT, CA, DN, or double-mutants in pBudCE4.1-eGFP, and ERK2 WT and mutant isoforms in pBudCE4.1-tRFP, were either single- or double-transfected at a 1:1 ratio using PolyJet (SignaGen Laboratories). Surface Clr-b expression was analyzed 48 hours after transfection.

Cell suspensions were stained in FACS buffer for 25 minutes on ice using biotinylated mouse Clr-b mAb (4A6; ref. 7) then stained for 20 minutes with secondary SA-PE or SA-APC (Thermo Fisher Scientific). Additional mAbs were purchased from Thermo Fisher Scientific, BioLegend, or BD Biosciences, as follows: IgM (II/41), CD45R/B220 (RA3-6B2), H-2D^b (KH95), H-2K^b (AF6-88.5), panRae-1 (186107, plus anti-rat IgG-APC), CD155/PVR (TX56), CD112 (MAB3869), CD48 (HM48-1), Qa-1 (6A8.6F10.1A6), CD54/ICAM-1 (YN1/1.7.4), CD3ϵ (145-2C11), NKp46 (29A1.4), NK1.1 (PK136), CD45.2 (104), CD16/CD32 (2.4G2). NKR-P1B^B6 mAb (2D12) was a gift from Drs. K. Iizuka and W. Yokoyama (Washington University, St. Louis, MO; ref. 6). Cells were analyzed using an LSR II flow cytometer (BD Biosciences), and FlowJo software (FlowJo, LLC). All plots show gated viable cells, as determined by forward scatter, side scatter, and lack of DAPI uptake.

Cells were lysed in RIPA buffer and protein concentrations were determined using a protein assay kit (Bio-Rad Laboratories). Normalized protein amounts were mixed with Laemmli buffer in reducing conditions and boiled. Lysates were loaded on 10% Mini-Protean TGX Pre-Cast gels (Bio-Rad), transferred to Immobilon-P^SQ membranes (Millipore), and then immunoblotted. Blots were developed using antibodies specific for MEK1 (61B12), MEK2 (13E3), ERK1, ERK2, ERK1/2 (137F5), ERK5, c-Myc, or β-tubulin (9F3), secondary anti-rat-IgG-HRP, and secondary anti-rabbit-IgG-HRP (Cell Signaling Technology), or Ras (259; Santa Cruz Biotechnology), and visualized with Immobilon Western HRP Chemiluminescent reagent (Millipore) on a ChemiGenius2 (Syngene) or MicroChemi4.2.

PB-tetO2-mcMyc-IRES-βgeo, PB-CBA-rtTA (Adv-rbgpA), and pCMV-piggyBac transposase vector were provided by Dr. A. Nagy (University of Toronto, Toronto, Ontario, Canada). The mouse cMyc insert was replaced with mouse H-Ras^WT, H-Ras^G12V, or H-Ras^S17N. In some experiments, the βGeo selection marker was replaced with the puromycin resistance gene. NIH3T3 fibroblasts or C1498 mouse leukemia cells were cotransfected with the three piggyBac vectors using the Nucleofection kit R (Lonza), Nucleofection kit V, or Lipofectamine 3000 (Thermo Fisher Scientific). On day 1, cells were rested in complete media for 6 hours, then 1.5 μg/mL (NIH3T3 cells) or 0.5 μg/mL (C1498 cells) doxycycline (Sigma) was added, followed by selection in 0.8–3.2 mg/mL G418 (Bioshop) or 2.5 μg/mL puromycin (Gibco) for 6 days. Doxycycline and G418 were replenished in culture every 48 hours.

BWZ.CD3ζ/NKR-P1B and control BWZ.36 reporter assays were described previously (7). Reporter cells were mixed overnight with stable doxycycline-inducible NIH3T3 or C1498 transductants that were treated for 24 hours or 48 hours with 1.5 μg/mL doxycycline. Cells were then lysed using CPRG lysis buffer and the assays were developed over various times. Plates' readings were obtained using a Varioskan plate reader, by measuring OD_595–655, as described previously (13).

NK-LAK effector cells were prepared from mouse FVB/N-strain splenocytes grown in complete RPMI containing 2,500 U/mL of recombinant human IL2 (Novartis) and sorted for NKp46⁺CD3⁻ NK cells, subsetted as NKR-P1B^+/− using the PK136 mAb (BD Biosciences) on day 4, then cultured for 2–3 more days. Doxycycline-inducible Ras and cMyc C1498 cells were induced for 24 hours and 48 hours, respectively. Lysis assays were performed as described previously (13). Supernatants (100 μL) were transferred to scintillation plates (PerkinElmer), dried, and counted using a Top Count NXT Microplate Scintillation Counter (Packard Instrument Company).

Eμ-cmyc⁺ tumor lines or inducible C1498 transfectants (±doxycycline) were differentially labeled with Maroon CellVue and PKH67, injected intraperitoneally into B6.CD45.1 or FVB/N mice and recovered after 16 hours. CD45.2⁺ donor cells were gated and recovery ratios were calculated using frequencies of residual labeled cells normalized relative to the input ratios.

To gain insight into the underlying basis for Clr-b loss during cellular transformation and on tumor cell lines (7), we screened a panel of WT, DN, and CA proto-oncogene and tumor suppressor vectors (Supplementary Table S2), and assessed their effects on Clr-b levels upon transfection into mouse NIH3T3 fibroblasts. The vectors were cotransfected with a vector encoding GFP to compare Clr-b levels between GFP⁺ (transfected) and GFP⁻ (nontransfected) populations. Notably, overexpression of WT H-Ras was found to promote a significant downregulation of surface Clr-b, with even more striking Clr-b loss using the CA-mutant, H-Ras^G12V, while modest Clr-b upregulation was observed for the DN-mutant, H-Ras^S17N (Fig. 1A and B). Clr-b downregulation was also observed upon overexpression of c-Raf, an immediate downstream effector of Ras, but this effect was only significant using a CA-mutant (Raf^CAAX; Fig. 1A and B).

Different Ras isoforms possess distinct posttranslational modifications, trafficking routes, plasma membrane localizations, and signaling outputs. To determine the effect of different Ras isoforms on Clr-b levels, NIH3T3 fibroblasts were transfected with WT or CA isoforms of H-Ras, N-Ras, or K-Ras using a bicistronic GFP reporter vector, which provided tighter monitoring of Ras and GFP coexpression. Here, WT Ras isoforms promoted significant Clr-b downregulation, while the CA-mutant Ras isoforms showed greater Clr-b loss (Supplementary Fig. S1A and S1B). No differences were observed between the distinct Ras isoforms. Taken together, these findings identify the Ras signaling pathway as a major mediator of Clr-b loss on normal fibroblasts.

Ras-mediated signaling and subsequent oncogenic transformation require tethering to the plasma membrane mediated by posttranslational modifications. To determine whether Ras-mediated Clr-b loss was linked to the immediate effects of the Ras signaling pathway, we assessed the effects of pharmacologic inhibition in NIH3T3 cells transfected with Ras^G12V in a bicistronic GFP reporter vector. FTI-277 is a farnesyltransferase inhibitor that blocks Ras isoprenylation, maturation, and subsequent activation of the Raf–MEK pathway (16). Following treatment with FTI-277, Ras^G12V-induced Clr-b downregulation was reversed in a dose-dependent manner (Fig. 1C and D), confirming the involvement of the downstream Ras signaling pathway in Clr-b modulation.

While different Ras isoforms vary in their relative abilities to activate downstream signaling cascades (16), they share common effector pathways. To further dissect the mechanism of Ras-mediated Clr-b downregulation, we investigated the involvement of three main Ras effectors, c-Raf, PI3K, and RalGDS. For this purpose, we tested the effects of Ras double-mutants on surface Clr-b levels. Single amino acid mutations in the Ras effector region, in combination with the activating G12V substitution, are known to induce partial loss-of-function effects, such that interactions promoting certain Ras effector pathways are maintained, while others are lost (29). Ras^G12V/D38E (which interacts only with c-Raf) largely maintains a capacity to downregulate Clr-b levels, while Ras^G12V/E37G (which loses interactions with c-Raf and PI3Kα) does not promote significant Clr-b loss (Fig. 2A and B). This further validates the importance of the c-Raf effector pathway in promoting Ras-mediated Clr-b loss. In addition, Ras^G12V/Y40C (which interacts only with PI3Kα/γ) exhibits an intermediate effect, suggesting a partial involvement of PI3K effectors in Ras-mediated Clr-b loss. Thus, the c-Raf and to a certain extent PI3Kα/γ effector pathways both appear to be involved in Ras-mediated Clr-b loss.

Ras^G12V signaling via Raf and PI3K can influence a number of downstream kinases, including MEK, JNK, and p38 MAPK. Therefore, the involvement of selected downstream signaling pathways on Ras-mediated Clr-b loss were tested via pharmacologic inhibition, including NFκB, JNK, p38 MAPK, MEK1/2/5, ERK1/2/5, and PI3K (Fig. 2C and D; refs. 30, 31). Inhibition of p38 MAPK or JNK in isolation using SB203580 and SP600125, respectively, had little effect on Ras^G12V-mediated Clr-b loss, as did inhibition of PI3K using LY294002, or NFκB using parthenolide. However, inhibitors of MEK1/2/5, U0126 and CI-1040 (PD184352), were capable of blocking Ras^G12V-mediated Clr-b downregulation. Notably, CI-1040 had a partial effect at 1 μmol/L, which is near the IC₅₀ for the MEK1/2 proteins, but exhibited a complete block at 20 μmol/L (32). This may be suggestive of redundancy in distinct MEK pathways. The ERK1/2 inhibitor, FR180204 (33), had only a partial effect on Ras^G12V-mediated Clr-b loss, suggesting possible involvement of the MEK5–ERK5 pathway in the observed effect of U0126 and CI-1040. However, a MEK5 inhibitor, BIX02189, that prevents ERK5 phosphorylation (34), did not block Ras-induced Clr-b downregulation. This supports the involvement of MEK1/2-ERK1/2 pathway.

A caveat to pharmacologic inhibitors is their pleiotropic, off-target, or undocumented effects, for example CI-1040 and U0126 can cross-inhibit MEK1/2/5 signaling. To circumvent this, we used lentiviral shRNA vectors to knockdown MEK1, MEK2, ERK1, ERK2, and ERK5 in NIH3T3 cells. Knockdown was confirmed by Western blotting, which demonstrated a reproducible efficiency of 80%–90% for MEK1, 90%–100% for MEK2, 80%–90% for ERK1/2, and 50%–70% for ERK5 (Fig. 3A–D). Importantly, shRNA knockdown of MEK1 did not affect MEK2 levels, and vice versa; this was also the case for ERK1 and ERK2. Flow cytometric analysis showed that MEK1/2 double-knockdown, but not MEK1 or MEK2 single-knockdown, significantly inhibited Ras^G12V-mediated Clr-b loss compared with a control LacZ shRNA (Fig. 3B and C). This is supportive of redundancy between MEK1 and MEK2 in Ras^G12V-mediated Clr-b downregulation; however, residual MEK1 protein was still evident in MEK1/2 double-knockdown cells.

With respect to the ERK isoforms, ERK2 knockdown, but not ERK1 or ERK5 knockdown, caused a significant block in Ras^G12V-mediated Clr-b loss, relative to control LacZ shRNA (Fig. 3E and F). This finding was further confirmed using an ERK2-deficient NIH3T3 cell line engineered using CRISPR/Cas9 gene-editing technology, which displayed reduced downregulation of Clr-b upon Ras^G12V transfection, an effect reversed by complementation with constitutive-active ERK2 (Fig. 3G). This implicates ERK2 as a central effector in Ras^G12V-mediated Clr-b downregulation, with little redundancy downstream of MEK1/2.

As ERK2 is necessary for promoting Ras^G12V-mediated Clr-b loss, we next tested whether ERK2 kinase function was sufficient. Thus, WT ERK2 and several mutants (Supplementary Table S2) were overexpressed, either alone or with Ras^G12V cotransfection (using bicistronic Ras^G12V-GFP and ERK2-tRFP vectors). These included a DN kinase-dead ERK2^K52R mutant (35), a CA-kinase ERK2^L73P/S151D mutant (36), a nuclear import-deficient ERK2^S321-7A mutant (37), and an ERK2^T183A/Y185F mutant lacking the TEY motif required for MEK-mediated activation and homodimerization (38). In addition, ERK2 may have kinase-independent functions (39), including a reported role in transcriptional repression (ERK2^K257A/R259A mutant; ref. 40).

Notably, none of the ERK2 constructs alone was sufficient to promote a loss of Clr-b surface expression, in the absence of Ras^G12V cotransfection. The inability of the CA-kinase ERK2^L73P/S151D to promote Clr-b loss without Ras^G12V coexpression suggests that upstream activation of MEK1/MEK2 may be required for full ERK2-dependent function (41). However, Ras^G12V cotransfection experiments (Fig. 3H) show that WT ERK2, CA-kinase ERK2^L73P/S151D, and nuclear import-deficient ERK2^S321-7A isoforms all enhanced Ras^G12V-mediated Clr-b downregulation; in contrast, the DN kinase-dead ERK2^K52R, TEY-mutant ERK2^T183A/Y185F, and purported repressor-deficient ERK2^K257A/R259A mutants had no effect on Ras^G12V-mediated Clr-b loss. Collectively, this suggests that ERK2 kinase activity alone is not sufficient to promote Clr-b loss, even though ERK2 function is required for Ras^G12V-mediated Clr-b loss. In turn, this suggests that upstream activation of kinase-dependent and kinase-independent functions of ERK2 by Raf/PI3K and/or MEK1/2 are involved in synergizing with Ras signaling to promote Clr-b loss.

It was recently shown that Clr-b (Clec2d) promoter activity, nascent transcripts, and surface protein expression are downregulated in response to viral infection (9, 11–13), and that Clr-b transcripts are lost in response to genotoxic/cellular stress (14). To investigate Clec2d transcripts in the context of oncogenic Ras signaling, stable NIH3T3 transfectants were generated to inducibly express Ras^G12V or Ras^S17N upon the addition of doxycycline, using a modified piggyBac transposon vector system, and inducible Ras expression and Clr-b loss were confirmed by Western blot analysis and flow cytometry (Fig. 4A and B). Ras^G12V induction using doxycycline promoted a real-time loss of surface Clr-b on stable transfectants compared with Ras^S17N or empty vector alone (Fig. 4A).

Subsequent analysis of Clec2d nascent transcripts was performed using intronic qRT-PCR, which measures the amount of Clr-b unspliced pre-mRNA across introns 1–4. Notably, doxycycline induction of Ras^G12V promoted a significant decrease in Clec2d nascent transcripts, relative to Ras^S17N or empty vector alone (Fig. 4C). Hence, oncogenic Ras signaling downregulates Clr-b transcripts in a manner consistent with that previously observed following viral infection or genotoxic stress (9, 11–14), perhaps indicative of an oncogenic stress pathway.

To assess this effect at the Clec2d promoter level, a dual-luciferase reporter assay (DLRA) was performed using the pGL3 vector harboring incremental fragments of the Clec2d promoter (100–500 bp upstream of transcriptional start site, TSS), and luciferase activity was measured relative to control Renilla luciferase activity in the presence of Ras^G12V or Ras^S17N cotransfection into NIH3T3 fibroblasts (12). Here, the activity of the 200 bp Clec2d fragment was significantly decreased by CA Ras^G12V cotransfection relative to DN Ras^S17N (Fig. 4D). This suggests that potential regulatory elements may exist within 200 bp of the Clec2d TSS in response to CA Ras^G12V signaling, perhaps downstream of ERK2 (40).

The loss of Clr-b ligand has been shown to enable “missing-self” recognition by NK cells expressing its cognate receptor, NKR-P1B, under conditions of genotoxic stress (14), virus infection (9–13), and hematopoietic transplantation using Clr-b^−/− bone marrow cells (8, 15). To test the effects of oncogenic signaling on NKR-P1B–mediated recognition of Clr-b during cellular transformation, NIH3T3 cells expressing doxycycline-inducible Ras^S17N and Ras^G12V were used as stimulators in BWZ.CD3ζ/NKR-P1B (BWZ.P1B) reporter cell assays. DN Ras^S17N transfectants, which possess slightly elevated Clr-b levels compared with control cells, show significantly increased BWZ.P1B reporter cell stimulation, whereas Clr-b ligand function is significantly diminished using CA Ras^G12V transfectants (Fig. 4E). This confirms that modulation of Clr-b ligand levels mediated via the oncogenic Ras signaling pathway directly affects NKR-P1B receptor–dependent recognition of target cells.

To determine whether oncogene-mediated Clr-b loss promotes missing-self recognition by NK cells, we employed doxycycline-inducible expression of the various Ras isoforms (WT Ras, DN Ras^S17N, CA Ras^G12V) in the B6-strain C1498 leukemia cell line (which expresses endogenous Clr-b), for use in functional assays. As seen using inducible NIH3T3 cells, Clr-b was significantly downregulated on C1498 leukemia cells within 24 hours of doxycycline induction for WT Ras and CA Ras^G12V transfectants, while Clr-b was upregulated by DN Ras^S17N induction (Fig. 5A). Notably, Ras isoform induction was confirmed by Western blot analysis (Fig. 5B).

Next, the inducible C1498 transfectants were used as stimulators in BWZ.P1B¹²⁹ reporter cell assays to confirm Clr-b ligand function. Here, we observed significantly diminished BWZ.P1B¹²⁹ reporter activity using doxycycline-induced WT Ras and CA Ras^G12V stimulators, compared with uninduced cells, while reporter activity was augmented using doxycycline-induced DN Ras^S17N cells (Fig. 5C). This confirms decreased Clr-b ligand function upon expression of WT and CA Ras isoforms, yet increased Clr-b ligand function upon expression of the DN Ras isoform.

Furthermore, the inducible C1498 transfectants were used as target cells in ⁵¹Cr-release cytotoxicity assays for sorted allogeneic FVB-strain NKp46⁺ lymphokine-activated killer (NK-LAK) effector cells, subsetted according to NKR-P1B expression (using the NK1.1 mAb, PK136). Importantly, cytotoxicity of doxycycline-induced WT Ras and CA Ras^G12V targets was significantly augmented using NKR-P1B⁺ NK-LAK effectors, in comparison with noninduced target cells, while cytotoxicity of doxycycline-induced DN Ras^S17N cells was diminished using NKR-P1B⁺ NK-LAK effectors (Fig. 5D). This modulation in NK cytotoxicity was NKR-P1B–specific, as similar effects were not observed using sorted NKR-P1B⁻ NK-LAK effectors (Fig. 5E), in turn highlighting that the observed changes in cytotoxicity are at least due in part to changes in Clr-b ligand function.

To confirm this, we generated doxycycline-inducible Ras^G12V C1498 cells lacking Clr-b (ΔClr-b) by CRISPR/Cas9-mediated genome editing and repeated BWZ.P1B reporter cell and ⁵¹Cr-release cytotoxicity assays (Supplementary Fig. S2A–S2E). Here, increased killing of doxycycline-induced Ras^G12V C1498.ΔClr-b cells was observed equally for NKR-P1B⁻ and NKR-P1B⁺ NK-LAK effectors (Supplementary Fig. S2D and S2E), suggesting the contribution of additional NK receptor–ligand axes. Screening of a number of NK ligands revealed that only CD155/PVR showed significant upregulation on C1498.ΔClr-b cells, an effect that was also observed using parental WT C1498 cells (Supplementary Fig. S2F; Supplementary Fig. S3A). Notably, however, cytotoxicity of the WT C1498 cells was significantly increased using NKR-P1B⁺ relative to NKR-P1B⁻ NK-LAK effectors, an effect that was not observed for C1498.ΔClr-b target cells (Supplementary Fig. S2E); this confirms an independent role for Clr-b. Together, these data demonstrate that Ras oncogene-mediated Clr-b modulation promotes missing-self NK recognition via the NKR-P1B:Clr-b axis; however, additional pathways likely contribute to augmented NK cytotoxicity. Notably, there are three known cognate receptors for CD155/PVR, including DNAM-1, TIGIT, and CD96.

To further test the effect of short-term oncogene induction on recognition of tumor cells in vivo, we compared the ability of WT B6.CD45.1 hosts to selectively reject labeled Ras-induced C1498 leukemia cells versus control noninduced cells. While control empty vector transfectants were not selectively rejected regardless of doxycycline induction, WT and CA Ras-induced C1498 cells were selectively rejected compared with noninduced control cells (Fig. 5F). Similar selective rejection of CA Ras-induced C1498 cells was also observed using allogeneic FVB recipient hosts (Fig. 5F).

To further corroborate the above findings using the Ras oncogene, we also investigated inducible expression of c-Myc, another well-characterized proto-oncogene and a downstream target of the Ras signaling pathway. Importantly, inducible overexpression of the c-Myc oncogene also promoted Clr-b loss on C1498 leukemia cells (Fig. 6A and B), commensurate with diminished Clr-b ligand function observed using BWZ.P1B reporter cells (Fig. 6C), and modestly augmented NK cytotoxicity mediated by NKR-P1B⁺, but not NKR-P1B⁻ NK-LAK effector cells (Fig. 6D and E). Notably, few changes were observed for other NK ligands upon c-Myc induction, with the exception of modest CD155/PVR upregulation, akin to that observed for Ras induction, albeit to a lesser extent (Supplementary Fig. S3B). Similar to Ras-induced C1498 leukemia cells, labeled cMyc-induced C1498 leukemia cells were selectively rejected in WT B6.CD45.1 hosts compared with control noninduced cells (Fig. 6F).

Collectively, these findings further reinforce the finding that oncogenic transformation promotes Clr-b downregulation and enhances NKR-P1B-dependent “missing-self” NK recognition, but additional effects such as CD155/PVR induction may also be involved and require further investigation.

To evaluate the role of Clr-b in tumor immune surveillance in vivo, we utilized the Eμ-cMyc spontaneous lymphoma model. The Eμ-cMyc transgene was backcrossed and intercrossed onto a Clr-b^−/− background by two different breeding strategies (Fig. 7A and B). In both cases, heterozygous Clr-b^+/− and homozygous Clr-b^−/− Eμ-cMyc⁺ littermate cohorts displayed significantly enhanced survival relative to WT Clr-b^+/+ mice [either WT non-littermate transgenic founder mice (Fig. 7A), or transgenic WT littermates (Fig. 7B)]. These results are consistent with the following model: Clr-b^+/+ mice may succumb faster due to the capacity of their lymphomas to inhibit NKR-P1B⁺ NK cells by maintaining high Clr-b ligand levels. Conversely, Clr-b^−/− lymphomas lack the ability to utilize this immune evasion mechanism, and hence are more readily controlled by NK cells (Fig. 7A and B). On the other hand, Clr-b^+/− mice may be protected compared with Clr-b^+/+ mice by more sensitive or finely tuned missing-self NK recognition combined with a decreased capacity of their lymphomas to maintain high Clr-b levels to evade NK cells (Fig. 7). Notably, a previous report examining Eμ-cMyc⁺ mice with or without the NKR-P1B receptor revealed that lymphomas from WT Nkrp1b^+/+ transgenic mice displayed higher Clr-b levels relative to lymphomas from Nkrp1b^−/− transgenic mice, highlighting that malignant lymphomas are capable of immunoediting of Clr-b levels (15). Hence, both Clr-b^−/− (Fig. 7) and Nkrp1b^−/− (15) transgenic mice survive longer than WT Eμ-cMyc⁺ mice.

To further address this, we examined the spontaneous lymphomas from moribund Eμ-cMyc⁺ mice, which possessed B lineage lymphomas (Fig. 7C; Supplementary Table S3). Next, clones or pooled clones of primary lymphoma lines were labeled, and the secondary rejection rates of Clr-b^−/− lymphomas were compared against Clr-b^+/+ or Clr-b^+/− lymphomas injected intraperitoneally in vivo. While the majority of individual mice showed selective rejection of Clr-b^−/− relative to Clr-b^+/+ or Clr-b^+/− lymphoma lines (Fig. 7D), the results were variable and not statistically significant, perhaps due to the clonal nature of the tumors. As NK cells have been previously shown to be involved in the control of lymphomas in the Eμ-cMyc model (42), we examined the expression of other NK ligands on the primary lymphomas. Notably, variable levels of distinct ligands were expressed (Fig. 7E), likely leading to variable rejection rates among clones (Fig. 7D). Interestingly, apart from Clr-b levels, the CD48 ligand for the 2B4/CD244 receptor was significantly lower on Clr-b^+/− and Clr-b^−/− lymphoma lines compared with WT Clr-b^+/+ lines (Fig. 7E); however, the reason for this is unclear at present. Collectively, these results suggest that the inhibitory NKR-P1B:Clr-b axis plays a role in both the elimination and escape (18) of malignant lymphoma cells in vivo.

Oncogenic transformation is an early event in tumorigenesis. The early detection of cells that escape the checkpoints preceding pathologic alterations is important in host immunosurveillance. NK cells have been demonstrated to be important in the recognition of cells expressing oncogenes, such as c-Myc (21, 22) and Ras (24), which are known to induce NKG2D-L and/or promote loss of surface MHC-I molecules (25). On the basis of the rationale that the NKR-P1B-ligand, Clr-b, is frequently lost on tumor cell lines in vitro (7) and malignant lymphomas in vivo (in Nkrp1b^−/− mice) (15), and that oncogenic signaling and transformation are early events in the immortalization of tumor cells, we sought to investigate the role of the Ras and c-Myc proto-oncogenes in Clr-b modulation.

Here, we demonstrate that Clr-b is rapidly downregulated during cellular transformation induced by oncogenic Ras and c-Myc signaling. While different Ras isoforms have unique properties, and their mutations are associated with specific types of tumors (16), all isoforms tested were found to promote Clr-b downregulation to a similar degree. We also demonstrate that the downstream Ras effectors, Raf, PI3K, MEK1/2, and ERK2 are required for the observed Clr-b downregulation. Moreover, the demonstration that c-Myc overexpression mediates similar effects suggests that oncogenic transformation in general promotes Clr-b loss to enhance NK-cell–mediated “missing-self” recognition.

While the MEK1/2–ERK2 pathway is necessary for Ras-mediated Clr-b loss, overexpression of WT or CA ERK2 alone were not sufficient to mimic Ras signaling, and it was difficult to identify ERK2 DN mutants capable of blocking the Ras-driven signal. Therefore, a lack of ERK2 sufficiency may be due to other kinase-dependent and -independent functions activated upstream of ERK2, including MEK1/2. Alternatively, Ras-mediated Clr-b loss was abrogated in ERK2^−/− cells, yet restored by complementation with exogenous CA ERK2. Importantly, there was demonstrable synergy of ERK2 overexpression with the Ras^G12V signal to promote enhanced Clr-b downregulation, suggesting a kinase dosage effect downstream of Ras. Taken together, these data implicate the Ras-Raf/PI3K–MEK1/2–ERK2 pathways in regulating Clr-b levels, in turn enabling NK-cell recognition via NKR-P1B during oncogenic transformation.

Previous work has demonstrated that Clr-b transcripts and cell surface protein are both lost upon viral infection and genotoxic stress, and that a short Clr-b surface half-life allows rapid communication of cellular fitness to facilitate “missing-self” recognition by NK cells (9, 11–14). In this study, we further highlight that this mechanism operates during Ras/Myc–induced oncogenic transformation. Reciprocally, the DN Ras mutant promoted an increase in Clr-b surface expression and ligand function, suggesting that cellular quiescence may be linked to high, “healthy-self” Clr-b levels. Mechanistically, oncogene signaling during cancer development may also involve other regulatory mechanisms shown to influence Clr-b levels, including epigenetic regulation (43), posttranscriptional regulation of mRNA stability (44), and posttranslational regulation of surface Clr-b levels via the ubiquitin–proteasome and endo-lysosomal degradation pathways (14). In other studies, noncoding miRNA have been linked to cancer (45), Ras activation (46), and the silencing of NKG2D-L induction (47). Interestingly, the regulatory mechanisms underlying NKG2D-L “induced-self” recognition share numerous similarities with those controlling Clr-b “missing-self” recognition, suggesting a convergent evolution of synergistic pathways controlling the susceptibility of target cells to NK-cell–mediated self-nonself discrimination. For example, NKG2D-L are induced by virus infection and genotoxic stress (23, 48, 49), and Rae1 and ULBP were recently shown to be upregulated by Ras signaling via the Raf and PI3K effectors, culminating in eIF4E-dependent posttranscriptional regulation (24). Our data also implicate oncogenes in the induction of CD155, ligand for the DNAM-1, TIGIT, and CD96 receptors. Further insight may elucidate additional factors that control Clr-b downregulation in response to oncogenic signaling and transformation.

Functionally, Ras/Myc–mediated oncogenic transformation and Clr-b downregulation resulted in diminished Clr-b ligand function in reporter cell assays (for both NIH3T3 fibroblasts and C1498 leukemia cells), and increased susceptibility of C1498 target cells to NK cytotoxicity in vitro, specifically by the NKR-P1B⁺ subset of NK cells. This shows that oncogene-mediated Clr-b loss directly affects NK-cell recognition of transformed target cells via NKR-P1B. Comparison of Clr-b-deficient to WT C1498 targets confirmed this was due in part to Clr-b recognition, but also highlighted a Clr-b–independent role for other NK ligands, including induction of CD155.

While many spontaneous tumor cell lines lack surface Clr-b expression altogether (7), and oncogene-induced Clr-b loss enhances NK cytotoxicity in vitro, an open question remained as to the effects of Clr-b deficiency during primary tumor progression in vivo. Here, oncogene induction selectively augmented rejection of C1498 leukemia cells in vivo in WT mice. In addition, we show that the loss of a single Clr-b allele (heterozygous-sufficient) or both alleles (homozygous Clr-b^−/−) protects Eμ-cMyc-transgenic mice from primary lymphoma progression in vivo. This could be due to a gene dosage effect resulting from more efficient NKR-P1B–mediated missing-self NK recognition of monoallelic Clr-b levels. Alternatively, as hematopoietic cells from homozygous Clr-b^−/− mice have been shown to spontaneously upregulate an activating NKR-P1F ligand (8), heterozygous Clr-b^+/− mice may possess a more balanced or finely tuned “induced-self” recognition of lymphoma cells via stimulatory NKR-P1F:Clr-c/d/g.

In any case, despite a documented beneficial role for the NKR-P1B:Clr-b axis in the detection of oncogenic transformation via innate immune surveillance, genetic ablation of either the NKR-P1B receptor (15) or Clr-b ligand (this study) both appear to prolong disease-free survival, suggesting a pathologic role for the NKR-P1B:Clr-b axis in primary tumor development in vivo. This can be explained by the previous finding that Eμ-cMyc⁺ lymphomas selectively lose Clr-b in vivo in Nkrp1b^−/− mice (as do oncogene-transformed cells in vitro), while malignant Eμ-cMyc⁺ lymphomas in WT mice were found to selectively maintain high Clr-b levels (15), indicative of immunoediting and the development of resistance under pressure from immune selection (17–19). Thus, as the NKR-P1B:Clr-b axis is inhibitory in nature, it may ultimately represent a tractable model for immune checkpoint blockade in vivo during cancer progression.

By extension, these results suggest that the homologous human inhibitory NKR-P1A:LLT1 axis may also play a role in cancer immunosurveillance and immune escape of malignant lymphoma cells in patients with cancer. In turn, these studies implicate both the human NKR-P1A receptor and its cognate LLT1 ligand as novel targets for therapeutic immune checkpoint blockade.

Collectively, these findings contribute to our current knowledge of innate recognition of cancer cells, and more specifically, NK-cell recognition of healthy versus pathologic target cells.

No potential conflicts of interest were disclosed.

Conception and design: M. Tanaka, J.H. Fine, C.L. Kirkham, A. Martin, D.S.J. Allan, J.R. Carlyle

Development of methodology: M. Tanaka, J.H. Fine, C.L. Kirkham, O.A. Aguilar, A. Belcheva, T. Ketela, J. Moffat, J.R. Carlyle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Tanaka, J.H. Fine, C.L. Kirkham, T. Ketela, J. Moffat

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Tanaka, J.H. Fine, C.L. Kirkham, O.A. Aguilar, J. Moffat, D.S.J. Allan, J.R. Carlyle

Writing, review, and/or revision of the manuscript: M. Tanaka, J.H. Fine, C.L. Kirkham, O.A. Aguilar, J.R. Carlyle

Study supervision: D.S.J. Allan, J.R. Carlyle

We thank Dr. Geneve Awong, Courtney McIntosh, and Vincent Cheng for cell sorting, and Dr. A. Nagy for the cMyc-inducible piggyBac vector set. This work was funded by scholarships from the University of Toronto (to M. Tanaka), Natural Sciences and Engineering Research Council of Canada (to J.H. Fine, O.A. Aguilar, C.L. Kirkham), Ontario Graduate Scholarships (to C.L. Kirkham), a Burroughs Wellcome Fund Investigator in the Pathogenesis of Infectious Disease Award, a Canadian Institutes of Health Research (CIHR) New Investigator Award, and CIHR Operating Grants (106491 to J.R. Carlyle).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.