NCR1 Expression Identifies Canine Natural Killer Cell Subsets with Phenotypic Similarity to Human Natural Killer Cells

Canines spontaneously develop many cancers similar to humans – including osteosarcoma, leukemia, and lymphoma – offering the opportunity to study immune therapies in a genetically heterogeneous and immunocompetent environment. However, a lack of antibodies recognizing canine NK cell markers has resulted in suboptimal characterization and unknown purity of NK cell products, hindering the development of canine models of NK cell adoptive immunotherapy. To this end, we generated a novel antibody to canine NCR1 (NKp46), the putative species-wide marker of NK cells, enabling purification of NK cells for further characterization. We demonstrate that CD3−/NKp46+ cells in healthy and osteosarcoma-bearing canines have phenotypic similarity to human CD3−/NKp46+ NK cells, expressing mRNA for CD16 and the natural cytotoxicity receptors NKp30, NKp44, and NKp80. Functionally, we demonstrate with the calcein release assay that canine CD3−/NKp46+ cells kill canine tumor cell lines without prior sensitization and secrete IFN-γ, TNF-α, IL-8, IL-10, and granulocyte-macrophage colony-stimulating factor as measured by Luminex. Similar to human NK cells, CD3−/NKp46+ cells expand rapidly on feeder cells expressing 4-1BBL and membrane-bound IL-21 (median = 20,283-fold in 21 days). Furthermore, we identify a minor Null population (CD3−/CD21−/CD14−/NKp46−) with reduced cytotoxicity against osteosarcoma cells, but similar cytokine secretion as CD3−/NKp46+ cells. Null cells in canines and humans have reduced expression of NKG2D, NKp44, and CD16 compared to NKp46+ NK cells and can be induced to express NKp46 with further expansion on feeder cells. In conclusion, we have identified and characterized canine NK cells, including an NKp46− subset of canine and human NK cells, using a novel anti-canine NKp46 antibody, and report robust ex vivo expansion of canine NK cells sufficient for adoptive immunotherapy.

inTrODUcTiOn Canines are a large animal model with spontaneous development of many cancers, including osteosarcoma, leukemia, lymphoma, glioblastoma, prostate cancer, and mammary cancer. Canines provide an outbred, immune competent disease model with genetic heterogeneity and a shared environment with humans (1)(2)(3)(4). Testing of novel therapies in the canine model has educated the protocols for bone marrow transplants in humans and, more recently, has been used for the testing of immune therapies such as adoptive transfer of T-cells, HER2-Listeria vaccine, and Liposomal-muramyl tripeptide (L-MTP-PE; mifamurtide) (5)(6)(7)(8)(9)(10)(11)(12).
Despite the advantages of the canine model, NK cells are less well characterized in canines than mice and humans. The sequencing of the canine genome in the early 2000s revealed that like humans, canines have all of the natural cytotoxicity receptors along with NKp80 in their genome (13)(14)(15)(16)(17). The primary inhibitory receptors that mediate licensing of NK cells are the Ly49 and KIR families of receptors, both of which recognize self through binding to MHC Class I. Mice have 16 Ly49 genes but only 2 KIR, whereas humans have 16 KIR genes but only a pseudogene of the Ly49 family (18). The canine genome has no KIR and only one Ly49 gene, which has a predicted ITIM sequence suggesting that it functions as an inhibitory receptor (19).
The identification of NK cells in canines has been met with seemingly conflicting results with some studies reporting CD3 − cell populations with NK cell properties, while others report CD3 + cell populations with NK cell properties (20)(21)(22)(23). Recently, Grondahl-Rosado et al. provided more clarity on the phenotype of canine NK cells using a cross-reacting anti-bovine antibody to NCR1 (NKp46), the putative species-wide marker of NK cells in mammals (13)(14)(15)(16)(24)(25)(26)(27). Using this antibody, they identified a CD3 − /NKp46 + cell population in most canines that were also positive for Granzyme B. Furthermore, they confirmed that NKp46 is an activating receptor in canine. They also proposed that a CD3 − /NKp46 − /Granzyme B + cell subset may be a subset of canine NK cells (16,17). However, this anti-bovine NKp46 antibody is reported by the authors to not be suitable for sorting of CD3 − /NKp46 + cells, limiting the ability to further characterize the receptor expression and function of CD3 − /NKp46 + cells and this NKp46 − cell population (16,17). Additionally, expansion of canine NK-like cells, while more successful than ex vivo expansion of mouse NK cells, has been significantly less than reported in humans with expansions reported of up to 233-fold on average in 2-3 weeks (19-23, 28, 29).
We sought to further characterize canine NK cells for use in osteosarcoma, where survival for metastatic human OS patients has largely remained stagnant at only 30% 5-year survival rate for the last 30 years (30)(31)(32)(33). Canine OS is highly prevalent, with over 8,000 new diagnoses per year, and an average survival rate of only 1 year, allowing for the rapid testing of new therapeutics. While mouse models have provided important discoveries in OS pathogenesis and treatment, the spontaneous canine model of OS has been well characterized and is used as an additional important animal model of OS (1,2,34,35).
To this end, we defined canine NK cells by their expression of NKp46, using a novel anti-canine NKp46 antibody, and expanded canine NK cells on membrane-bound IL-21 expressing feeder cells. We report here the identification and characterization of NKp46 + and NKp46 − canine NK cells that have striking phenotypic and functional similarity to human NK cells. Canine NK cells from both healthy and OS-bearing canines expand 20,283fold in 3 weeks enabling their use in testing NK cell therapies in the spontaneous canine model of OS.

Peripheral Blood Mononuclear cell isolation
Animal research was conducted with approval from the Institutional Animal Care and Use Committee at MD Anderson Cancer Center (00001532-RN00). Healthy canine blood was obtained from established animal colonies at the following locations: Animal Blood Resources International, Hemopet, and Texas A&M University (IACUC Protocol: 2014-0294). Healthy canine blood from UC Davis was exempted from IACUC approval. All blood from client-owned animals was obtained with informed consent and was consistent with the established guidelines for safe canine blood draws. Blood from canine patients with suspected osteosarcoma who had not received chemotherapy within 1 month was obtained with informed consent from UC Davis under Protocol 18315. Osteosarcoma diagnosis was confirmed with radiographs or biopsies. Two out of the seven canine patients presented with potential chondroblastic osteosarcoma, with one of the two patients presenting with disease with alternative diagnosis of chondrosarcoma. Four out of the seven patients were females with ages ranging from 2 to 13 years old. Osteosarcoma tumor sites were frontal bone (1), metatarsal (1), tibia (1), femur (1), radius (2), and ileal body (1). Experiments using discarded buffy coats from normal human red blood cell (RBC) donations were conducted under MD Anderson Cancer Center IRB exemption PA13-0978.
Canine blood was drawn into lithium heparin tubes and was diluted upon receipt 1:5 in HBSS before Ficoll separation. Both canine and human blood were processed using Ficoll Plus (GE Healthcare; 17-1440-02), as described previously, and canine RBCs were lysed with RBC lysis buffer (Stem Cell Technologies, 07800) for 4 min on ice (36,37).

nK cell expansion
Five million canine PBMC were cocultured with 10 6 K562 Clone9.mbIL-21 in at day 0 (1:2 ratio), and additional K562s were added at a 1:1 ratio at days 7 and 14. Medium was supplemented with 6.1 ng/mL recombinant canine IL-2 (rcIL-2) (R&D systems, 1815-CL), and fresh medium was added every 2-3 days. For human IL-2 comparison expansions, human IL-2 was added at 100 IU/mL in place of canine IL-2. The concentration of human IL-2 was based on the dose of IL-2 used in previous publications on canine NK-like cells (21,22,39).
For CD3 depletion experiments, canine PBMC was stained with CD3-FITC antibody and sorted for all PBMC that were CD3-negative. Sodium azide in CD3 antibody was reduced using Amicon Ultra-0.5 mL Centrifugal Filter Units with Ultracel-50 Membrane (EMD Millipore, UFC505024). Expansion was done according to the previous protocol, comparing CD3-depleted PBMC to -undepleted PBMC. Human NK cells were isolated via RosetteSep Human NK Cell Enrichment Cocktail at day 0 (Stem Cell Technologies; 15065) and expanded with K562 mbIL-21 feeder cells with the same protocol using 50 IU/mL of human IL-2.
Fold expansion was calculated as the percentage of CD3 − / NKp46 + cells within the lymphocyte gate of total PBMC. Null cell calculations were determined by subtracting the percentage of CD21 and CD14 expressing cells in the lymphocyte gate from the CD3 − cell percentage.

cell sorting
Canine cells were sorted on a FACSAria Sorter at the flow cytometry core facility (UT MD Anderson Cancer Center) using fourway purity with post-sort purity verification for expression of CD3 and NKp46 to generate ≥99.5% pure populations of CD3 − / NKp46 + , CD3 − /NKp46 − , CD3 + , and CD3 − cells when indicated.
rT-Pcr/qPcr RNA was isolated from canine expanded cells using RNAeasy Kit, QiaShredder Columns, and RNAase-Free DNase Set (all Qiagen, 74104, 79654, 79254), and cDNA was synthesized using Omniscript RT Kit (Qiagen, 20511) with RNase inhibitor (New England Biolabs, M0307S) and Oligo(dt)20 (ThermoFisher Scientific, 18418-020). RT-PCR and qPCR reactions were performed on Roche 480 (Roche, USA) using Power Up Sybr Green Master Mix (ThermoFisher; A25742) or TaqMan Universal PCR Mastermix for qPCR for CD16 (ThermoFisher Scientific, 4304437) for 40 cycles. All qPCR was done in duplicate with at least three different donors. Semi-quantitative RT-PCR reactions were run on five different donors with 100 ng cDNA per sample and 6× DNA loading dye, and products were verified on a 1.5% agarose gel with TAE buffer and GelRed Nucleic Acid Stain (Phenix, RGB-4103) and 100 bp DNA Ladder (New England Biolabs, N3238S). Gels were imaged on BioRad ChemiDoc Touch Imaging System (BioRad, USA) for Faint Bands and analyzed using Image Lab 5.2.1. The following enhancements were uniformly applied to all gels (High: 65535, Low: 34, Gamma: 0.43).

statistical analysis
Statistical analysis was performed using GraphPad Prism 6.0. Luminex assays were analyzed with unpaired t-tests and Holm-Sidak method for multiple comparisons. Paired t-tests were run for the following: cytotoxicity assays between NK and T-cells or Null Cells, qPCR, and comparison of NKp46 + and NKp46 − human NK cell phenotype. Comparison between healthy and OS-bearing canines' percent NK cells and expansion used the Mann-Whitney test. All other statistical analyses were unpaired Student's t-tests (all two-tailed). Human versus canine IL-2 analysis used a one-tailed Wilcoxon-matched pairs test. Correlations were done using the Spearman r test. p Values less than 0.05 were considered significant. resUlTs construction of an anti-canine nKp46 antibody Hybridomas obtained by immunization with L-cells expressing the NKp46 fusion protein (NKp46:L-cells) were screened via ELISA and flow cytometry against NKp46:L-cells and Empty Vector:L-cells. Clone 48A had strong staining on NKp46:L-cells and was selected for all future experiments (Figure 1A). Clone 48A did not identify NKp46 by western blot or immunoprecipitation (data not shown).  Figure 1B). On average, 2.3% (median = 1.2%, IQR = 0.6, 3.7, n = 12) of lymphocytes from healthy canines were CD3 − /NKp46 + , consistent with an NK cell expression pattern (Figure 1B). NKp46 was also expressed on a subset of CD3 + /TCR + cells.

Function of cD3 − /nKp46 + cells
To determine whether CD3 − /NKp46 + cells are able to kill spontaneously, a hallmark of NK cell function expanded pure CD3 − / NKp46 + cells and CD3 + T-cells from the same donor were cultured with the canine osteosarcoma cell line, Gray. CD3 − /NKp46 + cells   Figure 4A). Next, we compared the cytotoxicity of CD3 − /NKp46 + cells against two additional canine osteosarcoma cell lines (OSCA78 and Abrams), along with MDCK as a normal control at a 10:1 E:T ratio (44). We found OSCA78, Abrams, and MDCK to be sensitive to NK cell killing by all three donors ( Figure 4B). CD3 − / NKp46 + cells displayed titratable, donor-dependent cytotoxicity against three different canine osteosarcoma cell lines, including metastatic (Gray) osteosarcoma. Furthermore, CD3 − /NKp46 + cells are highly cytotoxic against CTAC ( Figure 4C). Thus, CD3 − /NKp46 + cells appear to possess NK cell function and CD3 − /NKp46 + will be used interchangeably with NK cells for the rest of the manuscript.

nKp46 identifies Distinct subsets of nK cells
We observed that NKp46 was not expressed on all lineage negative (CD3 − /CD21 − /CD14 − ) cells in both PBMC ( Figure 5A) and expanded ( Figure 3A) cells, with donor-dependent variation in the percent of null cells (CD3 − /CD21 − /CD14 − /NKp46 − ) observed. Null cells made up 0-13.52% of lymphocytes at day 0 (median = 4.5%, IQR = 2.54, 6.40), and 0-18.4% of the final expanded cell product (median = 5.0%, IQR = 0, 11.8). A significant negative correlation between percent of CD3 − / NKp46 + cells at day 0 and CD3 − /NKp46 + fold expansion was observed (r = −0.81, p = 0.022, not shown), which led us to question whether Null cells were significantly contributing to the yield of NK cells at day 21. When null cells were included in the fold expansion calculations, the mean fold expansion was 50% less than when using NKp46 + cells only, representing at most a two-fold over-estimation of expansion rates if all of the Null cells represented an NK population ( Figure 5B).
Finally, we sought to determine if human NK cells have similar subsets of NKp46 expressing cells. Using mass cytometry data of human CD3 − /CD56 + primary NK cells and K562 Clone9.mbIL-21-expanded NK cells, we found several striking similarities between human and canine NK cells. NKp46 − / CD3 − /CD56 + make up 21% of primary NK cells and 21.7% of  mbIL-21 expanded NK cells (Figure 5G). NKp46 − NK cells are primarily found within the CD56 dim NK cell subset (data not shown). Next, we used Spade to determine if NKp46 − NK cells are a distinct population of human NK cells. We clustered for CD56, NKp46, NKp30, NKp44, NKG2A, NKG2C, and NKG2D on both primary and mbIL-21 expanded human NK cells and

nK cells in canine Osteosarcoma
We obtained PBMC from canines that were not currently undergoing chemotherapy and found no difference in the percent of NK cells in OS-bearing canines compared to healthy controls ( Figure 6A). Additionally, NK cells from OS patients had similar proliferation on K562 Clone 9.mbIL-21 feeder cells and cytotoxicity against three canine OS cell lines compared to healthy canines (Figures 6B,C).

DiscUssiOn
We sought to characterize canine NK cells and optimize their expansion for use as a comparative oncology model of NK immunotherapy of OS. To this end, we developed a monoclonal antibody specific to canine NKp46 and demonstrated that this antibody recognizes CD3 − /NKp46 + cells that have striking phenotypic and functional similarity to human NK cells -expressing all of the NCRs and secreting IFN-γ and TNF-α. These CD3 − / NKp46 + cells from both healthy and OS-bearing canines exhibit robust expansion with K562 Clone9.mbIL21 and canine IL-2 and are highly cytotoxic against OS. In addition, we identified a small population of NKp46 − NK cells that have reduced cytotoxicity but similar cytokine secretion and can be induced to express NKp46. This antibody also identifies a subset of T-cells that are NKp46 + . In humans, NKp46 + T-cells are rare, but NKp46 can be acquired upon activation in γδ T-cells (45,46). In bovines, NKp46 + T-cells have functional similarity to bovine NK cells as they were able to kill a tumor cell line spontaneously unlike T-cells that did not express NKp46 (47). We observed very little killing of the spontaneous OS lung metastasis cell line, Gray, by CD3 + canine T-cells, but it will be of interest to more fully characterize the phenotype and function of NKp46 + canine T-cells. Despite the low cytotoxicity of the T-cells against the OS cell line, it remains unknown whether the contaminating T-cells in K562 Clone9. mbIL-21 expanded cell cultures are capable of mediating graftversus-host disease. Substantial work in humans suggests that contaminating T-cells might mediate GvHD and be detrimental in an allogeneic transplant setting (48,49). Thus, to reduce this potential adverse effect by contaminating T-cells, we demonstrate that CD3-depleted PBMC can give rise to greater than 96% pure CD3 − /NKp46 + expanded NK cells.
Spontaneous animal models of cancer for the study of NK cell therapies in a syngeneic setting have to date remained elusive due to difficulty in expanding NK cells in other species outside of human and the high cost of primate research. Instead, preclinical models of adoptive NK cell therapy have relied largely on infusing human NK cells into xenogeneic mouse models which lack a complete tumor microenvironment and an intact immune system that may influence the effectiveness of immune-based therapies (36,50). We report here an improved expansion platform for canine NK cells on K562 Clone 9.mbIL-21 feeder cells obtaining a median fold expansion of over 20,000 in 3 weeks when supplemented with canine IL-2. To the best of our knowledge, this is the largest fold expansion of NK cells from any mammal outside of human and is significantly greater than previously reported expansions of NK-like cells in canine (20,22). Previous expansions of canine CD3 − NK-like cells have reported an average fold expansion of 140-fold in 3 weeks using cytokine alone (IL-2 and IL-15), 233fold expansion in 2 weeks with EL08-1D2 feeder cells following CD5 depletion, and 90-fold expansion with K562 feeder cells and soluble IL-2, IL-15, and IL-21 in 3 weeks (20,21). It will be of future interest to determine if the improved expansion of canine NK cells with mbIL-21 over mbIL-15 is also due to an increase in telomere length as we have previously demonstrated for human NK cells (36).
Surprisingly, we found that CD3 − /NKp46 + cells displayed cytotoxicity against the MDCK cells used as normal controls. MDCK can undergo tumorigenic transformation, which could induce the expression of activating ligands (44). In addition, the existence of only one Ly49 gene and no KIRs in the canine genome raises the question of whether the single Ly49 can bind to all dog leukocyte antigens (DLA) (51), which could increase NK cell killing in the allogeneic setting. It will be necessary to better define canine Ly49-DLA interactions in order to predict these responses.
Similar to Grondahl-Rosado et al., we found that NKp46 was not constitutively expressed on all CD3 − /Granzyme B + cells (17). In the present study, we added to this knowledge, by finding that these NKp46 − , Null cells could be induced to express NKp46 and have a distinct function -displaying reduced cytotoxicity compared to NKp46 + NK cells. In contrast to Grondahl-Rosado et al., we found here that NKp46 − cells produce similar amounts of IFN-γ among all other cytokines tested. These differences may be explained by different cell stimulation as Grondahl-Rosado et al. measured IFN-γ stimulation after culturing NKp46 + and NKp46 − cells together with human IL-2, IL-12, and IL-15 while, in this paper, we stimulated NKp46 + and NKp46 − cells after sorting with canine IL-2 and K562 expressing membrane-bound IL-21 ( Figure 5F) (17).
NKp46 − NK cells are not unique to canines but have been previously reported in both porcine and humans, where NKp46 expression also discriminates human NK cells with reduced cytotoxicity. In contrast to our data in canine, porcine NKp46 expression discriminate porcine NK cells with different IFN-γ secretion but not cytotoxicity (27,52). Similar to porcine NKp46 − NK cells, NKp46 could be induced in canine NKp46 − cells; however, NKp46 has not been inducible in human NKp46 − cells (27). This may be due to differences in the regulation of NKp46 expression across species, where it is inducible in some species' (porcine and canine) NK cells, but not in human, or NKp46 expression may require specific signals that have not been described yet (52).
We found striking phenotypic similarity between canine NKp46 − null cells and both primary and K562 Clone9. mbIL-21-expanded NKp46 − human NK cells (CD3 − /CD56 + ). In humans, NKp46 − NK cells are predominantly CD56 dim . Expanded NKp46 − cells in both species have reduced expression of NKp44, NKG2D, and CD16. Thus, we speculate that NKp46 − canine and human NK cells may represent similar subsets of NK cells in both species; however, additional functional and phenotypic characterizations of NKp46 − NK cells are necessary to elucidate this cell type to determine if they are homologous. Primary human NK cells also have reduced expression of NKp44 and CD16, suggesting that this population of NKp46 − NK cells is not an artifact of ex vivo expansion. There are several possible explanations for this cell type in humans that will be of interest to explore in future studies. CD56 dim /CD16 − /NKp46 − cells may represent an intermediate step in NK cell maturation between CD56 bright /CD16 low/− /NKp46 + and CD56 dim /CD16 + /NKp46 dim NK cells, or CD56 dim /NKp46 − /CD16 − cells may be an exhausted NK cell population that were formerly CD56 dim /CD16 + /NKp46 + cells and have downregulated CD16 and NKp46 because of senescence.
Contrary to human OS patients, we found no significant difference in circulating NK cells in canine OS compared to healthy canines (53). It will be of future interest to determine if these primary NK cells are functionally impaired, which would be facilitating tumor growth and has been found in several human cancers (54,55). However, our data demonstrate that expanded NK cells from canine OS patients are active against OS, both primary and metastatic cell lines. This parallels findings in human OS where cytokine-activated NK cells from human OS patients have similar cytotoxicity against both OS cell lines and autologous patient-derived biopsies compared to healthy controls (56). Based on these findings, autologous NK cells may be effective in canine OS.
In summary, our study furthers the understanding of the canine NK cell phenotype and function. Furthermore, we describe the largest expansion of NK cells from another mammal besides humans, using K562 Clone 9.mbIL-21 feeder cells and canine IL-2. These mbIL-21 expanded NK cells are highly cytotoxic against OS cell lines, suggesting potential for immunotherapy of dogs with OS and potential for therapeutic use in other canine cancers such as lymphoma, leukemia, melanoma, and glioblastoma. This approach will allow for the study of NK cell therapy in an immune-intact, outbred, and spontaneous animal model. Many of these cancers progress rapidly in canines, allowing for rapid testing of NK cell therapies, which is of particular interest in rare and pediatric cancers such as OS where trial accrual is slow. These results support the development of canine NK cell trials to help prioritize the most promising treatment regimens for human clinical trials. Of particular interest is a trial of intratumoral injection of canine NK-like cells expanded on K562 Clone9.mbIL-21 feeder cells after radiation therapy in canine OS that is currently underway (UC Davis, Michael Kent & Bob Canter). The effectiveness of canine NK cells in combination with other agents, such as cytokines, chemotherapy, and antibodies, will be of particular interest.