Immunodeficient Mouse Strains Display Marked Variability in Growth of Human Melanoma Lung Metastases

Mouse models of human cancer serve as essential experimental systems, and genetically defined immunodeficient mouse strains constitute a valuable tool for studying tumorigenesis. Athymic (nude) mice have been the standard for establishing in vivo models of human malignancies (1). However, the presence of residual adaptive and innate immunity can interfere with the establishment of tumor xenografts. Athymic mice develop small numbers of mature αβTCR lymphocytes with age; in addition, robust natural killer (NK) cell activity is present and increases with age (2, 3). Numerous reports confirm variable rates of human tumor growth in athymic animals (4–6); for example, of 200 human breast cancer samples tested in nude mice, just 25 (12.5%) grew as xenografts at the site of s.c. implantation (7). Severe combined immunodeficiency (scid) mice have relative B- and T-cell deficiencies and are often used as recipients of human xenografts. Improved tumor engraftment rates have been reported in the nonobese diabetic (NOD)-scid strain, where introduction of the scid mutation into the NOD background results in reduced macrophage and NK function, as well as an absence of complement-dependent hemolytic activity (8, 9). Recently, two additional immunodeficient strains have been described as follows: NOD-scid β2m^null and NOD-scid IL2Rγ^null (10). The NOD-scid β2m^null strain was developed by backcrossing the β2m^null mutation to the NOD-scid strain resulting in mice deficient in MHC class I expression (NOD-scid β2m^null). Accumulating data suggest that NK cells that develop in a MHC class I–deficient background are unlicensed and, hence, unable to kill susceptible targets upon activation (11). The NOD-scid IL2Rγ ^null strain was developed by introduction of the IL2Rγ^null mutation into the NOD-scid strain (12). Absence of IL2Rγ, the common cytokine- receptor γ-chain shared by the interleukin (IL)-2, IL-4, IL-7, IL-9, IL-15, and IL-21 receptors leads to impaired NK cell development due to absence of IL-15 signaling (13). Improved engraftment of human cord blood CD34+ cells has been reported in NOD-scid β2m^null and NOD-scid IL2Rγ^null highlighting the potential use of these strains as recipients of tumor xenografts (14).

Animal models of human melanoma are limited. Clark et al. (15) generated highly metastatic human melanoma (A375) cells through in vivo selection of lung metastasis in nude mice. A genomic analysis of metastatic A375 variants showed that RhoC, an RAS-related guanosine triphosphatase, is an important determinant of tumor cell invasion. Further studies have shown that RhoC plays a role in cytoskeleton organization and is essential for tumor metastasis (16). In a recent series of studies, Beige-scid mice were used to evaluate therapeutics for human melanoma A375-RhoC pulmonary metastasis (17, 18). However, NK cell function was not studied and because no other mouse strains were evaluated, the relative effect of residual innate immunity remains undefined in the A375-RhoC model.

In the present study, we evaluate human melanoma (A375-RhoC) pulmonary metastases in four immunodeficient mouse strains: scid, NOD-scid, NOD-scid β2m^null, and NOD-scid IL2Rγ^null strains. Tumor growth was monitored noninvasively by imaging using bioluminescence (BLI) and magnetic resonance imaging (MRI) combined with histopathologic assessment. BLI allowed for accurate, real-time serial in vivo quantitation of tumor burden, whereas MRI permitted three-dimensional structural imaging of tumor. Our results show that NOD-scid IL2Rγ^null mice are highly permissive for engraftment of human melanoma metastases, whereas endogenous NK activity in the other three strains delayed or, in certain instances, completely prevented the formation of pulmonary metastasis. Importantly, we show that expression of human NKG2D ligands MHC class I chain–related A (MICA) and B (MICB) molecules by melanoma confers susceptibility to murine NK-mediated cytotoxicity.

Mouse strains. CB17-Prkdc^scid/J (scid), NOD.CB17-Prkdc^scid/J (NOD-scid), NOD.Cg-Prkdc^scid B2m^tm1Unc/J (NOD-scid β2m^null), and NOD.Cg-Prkdc^scid IL2rγ^tm1Wjll/SzJ (NOD-scid IL2Rγ^null) mice were obtained from The Jackson Laboratory and bred and housed according to the guidelines of Washington University, Division of Comparative Medicine. Strain background is BALB/c (H-2^d) for scid and NOD (H-2^g7) for all other strains. The animal ethics committee approved all experiments. All mice used were between ages 7 and 14 wk.

Melanoma cell lines. The human melanoma cell line A375P-RhoC-GFP (15) was transduced with a retrovirus expressing a click beetle red luciferase (cbr-luc)/enhanced yellow fluorescence protein (eYFP) fusion gene. A stable cell line was selected by flow cytometry on a MoFlo sorting for cells expressing high GFP (for RhoC expression) and eYFG (for cbr-luc expression) levels. Stable A375 RhoC-luciferase–expressing cell line is called A375RC-Luc. Human melanoma lines DM6 and Lox (American Type Culture Collection) are previously described (19). CG mel is a primary human melanoma (S100+ HMB45+) generated in our laboratory from a resected lymph node metastases. DM6, Lox, and CG lines stably expressing high GFP (for RhoC expression) and eYFG (for cbr-luc expression) levels were generated as described above for A375. M14 melanoma cells (20) were transfected with MICA and MICB cDNA (21) using lipofectamine as directed by manufacturer's instructions (Invitrogen). Clones were selected using G418 at 1 mg/mL and stable expression of MICA and MICB confirmed by flow cytometry using anti-MICA/B monoclonal antibody (mAb) 6D4 (Biolegend; ref. 22). Expression of ULBP1-3 by melanoma cell lines was determined by flow cytometry using antibodies (23) obtained from Axxora.

In vivo imaging. For serial analysis of tumor growth, mice were injected i.v. with 2.5 × 10⁶ RhoC/luciferase-expressing tumor cells in 200 μL PBS. After tumor inoculation (2-3 h), mice were injected i.p. with 150 mg/kg D-luciferin (Biosynth)/PBS and imaged 10 min later. Imaging was done using a charge-coupled device camera (IVIS 50; Caliper Corporation; exposure time, 1-30 s; binning, 8; field of view, 12; f/stop 1, open filter) at the Molecular Imaging Center (Washington University, St. Louis) as described previously (24). Mice were anesthetized using isoflourane (2.5% vaporized in O₂). For analysis, total photon flux (photons per second) was measured from a fixed region of interest over the thorax/liver area using Living Image 2.50 and IgorPro software (Wavemetrics; ref. 24). Animals were monitored biweekly for health and signs of tachypnea and weight loss. MRI experiments were done in the Biomedical MR Laboratory (Washington University, St. Louis) as previously described (25, 26). Briefly, respiratory gated, spin-echo MR images were collected in a 4.7 T Oxford magnet interfaced with a Varian NMR Systems INOVA console. Animals were anesthetized with isoflourane (1% v/v in O₂) and animal core body temperature was maintained at 37°C ± 1°C by circulation of warm air through the bore of the magnet. Synchronization of MR data collection with animal respiration was achieved with a home-built respiratory-gating unit (27) and all images were collected during postexpiratory periods. Imaging parameters were as follows: TR, 3 s; TE, 20 ms; field of view, 2.5 cm; slice thickness, 0.5 mm. BLI was done weekly; MRI experiments were done at selected time points as described.

Histopathology. After MRI on day 28 (for NOD-scid IL2Rγ^null) or day 35 (all other strains), animals were sacrificed, lungs were excised, and fixed in 10% formalin overnight. Paraffin-embedded tissues were sectioned and stained by H&E. Tissues were examined in a blinded manner; at least five sections of tumor were examined by a pathologist. Two animals per strain were evaluated. Representative slides from each animal were scanned with modified 35-mm slide scanner, areas of tumor were quantified and expressed as a percentage of total lung tissue using NIH Image Software (NIH).

Immunologic assays. Spleen cell suspensions were prepared and stained using anti-CD3, anti-CD45.1, anti-CD49b-PE (DX5), and/or anti–NKG2D-biotin (28). Antibodies were obtained from eBioscience or Biolegend. Cells were incubated with anti-CD32/CD16 for 5 min at 4°C, followed by specific antibodies, washed, and, in the case of anti-NKG2D biotin, followed by SA-APC. Staining with murine (mu)NKG2D and control tetramer was done as described (29). Multiparameter flow cytometry analysis was done using a BD Biosciences FACScan flow cytometer modified with a second, 633 nm 25 mW laser, and 2 additional detectors with bandwidths at 660 and 760 nm. For assessment of NK function, spleen cell suspensions were depleted of RBC and cultured at 2 × 10⁶ cells/mL in 24-well tissue culture plates for 48 h in RPMI/10% FCS/2 mmol/L glutamine, HEPES, Pen-Strep, murine IL-15 (100 ng/mL; Peprotec) and murine IL-18 (100 ng/mL; R&D Systems). For ⁵¹Cr-release assays, target cells (10⁶ cells/mL) were labeled with 1 μmol/L ⁵¹Cr for 1 h, washed and added to spleen cultures at various effector/target (E/T) ratios, and incubated in triplicate for an additional 4 h. ⁵¹Cr-release was quantitated in a Tri-Lux Wallac as described (30).

Statistical analysis. Unpaired t test was done to evaluate tumor growth kinetics among strains, and the differences were significant if P value was <0.05. The Interquartile range were determined to evaluate tumor growth variability among strains, and Kaplan-Meier product-limit method was used to calculate survival rates; differences between groups were determined using log-rank analysis; P values of <0.05 indicate significance in survival rates. Repeated measure one-way ANOVA was done to evaluate lysis of melanoma by NK cells; for pairwise comparison, a Tukey's test was done post-ANOVA analysis. All analysis was done using Prism version 5 (GraphPad Software, Inc.).

Human melanoma pulmonary metastases show different rates of engraftment in immunodeficient mouse strains. Human melanoma cells expressing RhoC (A375-RhoC) exhibit lung tropism after i.v. transplantation in athymic mice (15). To evaluate the effect of residual murine innate immunity on the development of human melanoma pulmonary metastases, we assessed A375-RhoC growth in scid, NOD-scid, NOD-scid β2m^null, and NOD-scid IL2Rγ^null. BLI and MRI imaging modalities were used to monitor rates of tumor engraftment (25, 31). A375-RhoC cells stably expressing a luciferase reporter (A375RC-Luc, 2.5 × 10⁶) were administered i.v. on day 0 and animals were examined using BLI 2 to 3 hours after tumor inoculation and weekly thereafter to assess the kinetics of tumor growth (31). Serial BLI images of one representative mouse per strain are shown in Fig. 1. Tumor growth as assessed by BLI signal intensity is most rapid in NOD-scid IL2Rγ^null animals followed by the NOD-scid β2m^null strain (Fig. 2A). The growth of melanoma lung metastasis was delayed in scid and NOD-scid mice when compared with NOD-scid IL2Rγ^null and NOD-scid β2m^null (Fig. 2A). The mean time to reach a photon flux of 1 × 10⁹ photon/second was as follows: scid, 31d; NOD-scid, 28d; NOD-scid β2m^null, 14d; and NOD-scid IL2Rγ^null, 8d (Fig. 2A). As tissue attenuation of BLI signal is fixed at any given depth, different times to reach equivalent BLI signals are likely to reflect distinct tumor growth kinetics among strains. Tumor growth was confined to the lung in all strains as no BLI signal was detected outside this organ (Fig. 1). Imaging results were verified by necropsy on day 28 to 35 (data not shown).

To further characterize tumor growth and correlate BLI with anatomy, MRI was done on two animals per strain. Coronal, respiratory-gated spin-echo images of a representative mouse per strain (day 28 for NOD-scid IL2Rγ^null; day 35 all other strains) are shown together with histopathology of lung tissue (H&E-stained sections; Fig. 2B). Two views, ×10 and ×100, of lung tissue sections are shown for comparison. Based on total lung parenchyma, quantitative tumor burden for each strain (n = 2 mice per strain) was assessed microscopically in a blinded manner: scid, 7%; NOD-scid, 35%; NOD-scid β2m^null, 62%; and NOD-scid IL2Rγ^null, 95%. MRI detection of pulmonary metastases is less sensitive than BLI; in scid mice, BLI shows clear signal on day 35, although no solid tumor is detected by MRI. In contrast, innumerable pulmonary metastases are visualized in both NOD-scid β2m^null and NOD-scid IL2Rγ^null strains (Fig. 2B, right column). Although several lung nodules are evident in the NOD-scid animals, the tumor burden is substantially less and is consistent with the BLI results.

Human melanoma shows variable engraftment in NOD-scid β2m^null. We did additional experiments to further characterize the kinetics of tumor growth in the three most permissive strains: NOD-scid, NOD-scidβ2m^null, and NOD-scid IL2Rγ^null. Figure 3A to C shows a summary of tumor growth rates as determined by BLI for individual animals at weekly time points. BLI signal is reproducibly detected in all animals within 2 to 3 hours after i.v. injection of A375RC-luc cells. In spite of brisk tumor growth in both NOD-scid β2m^null and NOD-scid IL2γ^null, tumor engraftment is highly variable in NOD-scid β2m^null as shown by the larger interquartile range among mice in the NOD-scid β2m^null strain (Fig. 3B). Between day 28 and 35, 60% of NOD-scid IL2Rγ^null mice died due to tumor progression. In contrast, no deaths were seen before day 35 in any tumor-bearing NOD-scid and NOD-scid β2m^null animal; NOD-scid mice show significantly delayed tumor growth compared with NOD-scid β2m^null and NOD-scid IL2Rγ^null mice. Survival curves of NOD-scid β2m^null and NOD-scid IL2Rγ^null in a representative experiment (n = 10 mice per cohort) are shown in Fig. 3D. The median survival was 42 days for NOD-scid β2m^null and 32 days for NOD-scid IL2Rγ^null; in contrast, all NOD-scid mice survived past 48 days (data not shown).

Rapid engraftment of multiple human melanoma lines in NOD-scid IL2R^γnullmice. Due to the consistent and reproducible growth of A375RC-luc observed in NOD-scid IL2Rγ^null, we selected this mouse strain to evaluate the growth of three additional Rho-C/luciferase expressing human melanoma cell lines. DM6 (19) and Lox are well-characterized cell lines, whereas CG mel is a primary (low passage) melanoma line isolated in our laboratory. NOD-scid mice (strain background NOD: H-2^g7) were used as the comparator strain with NOD-scid IL2Rγ^null mice as these strains share identical genetic background (10). Average BLI values for mice (n = 3) at each time point is shown in Fig. 4A to D. Each melanoma studied exhibited a distinct rate and level of engraftment in NOD-scid and NOD-scid IL2Rγ^null mice as assessed by BLI. Rates and levels of engraftment of A375, Lox, and CG were higher in NOD-scid IL2Rγ^null relative to NOD-scid mice. Interestingly, despite similar BLI signal at day 0 in DM6-bearing NOD-scid and NOD-scid IL2Rγ^null, no BLI signal was detected after day 7 in NOD-scid mice, suggesting complete rejection of DM6 in this strain. In agreement with this finding, no evidence of tumor was observed upon autopsy of DM6-bearing NOD-scid lungs (day 50 after tumor inoculation; data not shown). Altogether, these results show that NOD-scid IL2Rγ^null mice are a permissive host for engraftment of human melanoma pulmonary metastasis.

NK characterization and function in immunodeficient strains. The superior engraftment of human melanoma in NOD-scid IL2Rγ^null, a strain devoid of NK cells, implicated murine NK cells as the effector population responsible for tumor rejection. Flow cytometry analysis confirms the presence of NK cells in scid (48.9%), NOD-scid (27.3%), and NOD-scid β2m^null (30.5%) spleen CD45+ populations as determined by the coexpression of CD49b (DX5) and NKG2D. No reactivity of NOD-scid IL2Rγ^null spleen cells was observed with anti-CD49b or anti-NKG2D, indicating that NK cells were absent in this mouse strain as previously noted (Fig. 5A and B; ref. 13).

To assess NK effector function, spleen cells from the various strains were activated in vitro with IL-15 + IL-18 and, 48 hours later, assessed for their capacity to recognize human melanoma cells as targets in a ⁵¹Cr-release assay. NOD-scid IL2Rγ^null spleen cells cultured in IL-15 + IL-18 yielded no appreciable numbers of NK cells and displayed no killing against human melanoma or YAC cells, confirming the complete absence of NK cells in this strain (data not shown). IL-15 + IL-18–activated scid, NOD-scid, and NOD-scid β2m^null spleen cells were able to recognize and kill YAC cells (data not shown), as well as multiple human melanoma lines (Fig. 5C). These findings indicate that murine NK cells from selected immunodeficient mice have the ability to recognize human melanoma.

Expression of NKG2D ligands MICA, MICB, and UL-16 binding proteins 1 to 3 by human melanoma. MICA and MICB and UL-16 binding proteins (ULBP) 1 to 3 are the human ligands for NKG2D, a NK activating receptor (32). Several reports provide evidence that muNKG2D can bind human NKG2D ligands such as MICB, ULBP-1, and ULBP-2 (33, 34). To further investigate the mechanism of murine NK recognition of human melanoma, we determined the expression of NKG2D ligands on selected melanoma cell lines, including M17 a melanoma cell line obtained from the European Searchable Tumor Cell Line and Data Bank and reported as MICA/B negative (35). As shown in Fig. 6A, ULBP-1, ULBP-2, and ULBP-3 expression among melanoma cell lines was heterogeneous. Most lines were negative for ULBP-1, and expressed variable levels of ULBP-2 and ULBP-3. Only one cell line, DM6, expressed all three ULBP molecules. No correlation was observed between ULBP expression and the ability of murine NK cells to recognize human melanoma. In contrast, seven of nine melanoma cell lines expressed MICA/B as detected by the 6D4 mAb (22). Interestingly, low levels of MICA/B expression in M14 and M17 cells correlated with low susceptibility to murine NK cell cytotoxicity (Figs. 6A-B and 7A), suggesting a potential role for MICA/B in murine NK cytotoxicity against human melanoma cells.

MICA and B expression by human melanoma confers susceptibility to muNKG2D-mediated cytotoxicity. To investigate MICA/B-muNKG2D interactions and its functional consequences, M14 cells were transfected with MICA or MICB and evaluated for binding of muNKG2D tetramer (29). As shown in Fig. 6B, M14 cells show no reactivity with MICA/B mAb and exhibit low levels of muNKG2D tetramer binding. This finding is consistent with low level expression of ULBP-2 by M14 because muNKG2D tetramer can bind ULBP-2 as shown by Sutherland et al. (34). M14 cells were transfected with cDNA encoding MICA or MICB and expression was confirmed using the 6D4 mAb. Moreover, expression of MICA and MICB in M14 cells results in increased binding of muNKG2D tetramer as shown in Fig. 6B. In cytotoxicity assays, neither M17 nor M14 were susceptible to lysis by activated NK cells (Fig. 7A), whereas A375 and DM6 were killed in a dose-dependent manner at the indicated E/T ratios. Expression of MICA and MICB rendered M14 cells sensitive to murine NK lysis as shown at multiple E/T ratios (Fig. 7B) compared with control M14 cells. In multiple independent experiments (n = 6), the expression of either MICA or MICB was sufficient to render M14 melanoma sensitive to NOD-scid NK cell lysis (Fig. 7C). The involvement of muNKG2D receptor in melanoma recognition was examined using the antagonistic anti-muNKG2D mAb C7 (28). Pretreatment of activated NK cells with anti-NKG2D mAb resulted in a significant decrease in MICA and MICB recognition with no statistically significant effect on control M14 cells (Fig. 7D). Altogether, these data support the finding that MICA and MICB are capable of binding to and activating murine NK cells through interaction with their NKG2D receptor.

In the present study, we compared scid, NOD-scid, NOD-scid-β2m^null, and NOD-scid IL2Rγ^null mice as recipients for human melanoma. Engraftment of human melanoma pulmonary metastasis in NOD-scid IL2Rγ^null mice is brisk and reproducible relative to the other strains examined. Despite multiple attempts to isolate and grow NK cells from NOD-scid IL2Rγ^null mice, we were unsuccessful and, thus, confirm the original observation that ablation of the common cytokine-receptor γ-chain (IL2Rγ) confers an absolute NK cell deficiency (13). Our results also suggest that murine NK cells use NKG2D to recognize MICA/B, and we propose that this mechanism accounts for the increased xenograft rejection observed in scid, NOD-scid, and NOD-scid-β2m^null mice.

We initially observed dramatic differences in human melanoma engraftment among NOD-scid β2m^null littermates and were perplexed by this inconsistency. Although NK cells from NOD-scid β2m^null mice are present in similar percentages as scid and NOD-scid animals, it seems that NK cells that develop in a MHC class I–deficient environment are functionally impaired (11, 36). However, our findings suggest the presence of residual/partial NK cell activity in NOD-scid β2m^null mice because some animals clearly have delayed tumor engraftment (Fig. 3B). The variable engraftment rates of human CD34+ stem cells in NOD-scid β2m^null animals supports this conclusion (37). In support of NK cells retarding the growth of xenografts, administration of anti-CD122 antibody to deplete endogenous NK cells in NOD-scid mice has been shown to improve engraftment of human hematopoietic stem cells (38) as well as solid tumor stem cells (39). The inconsistent tumor growth seen is likely to reflect variable levels of NK function among mouse strains and not heterogeneity in tumor sample as similar results were obtained using tumor lines or clones obtained by limiting dilution (data not shown). The inherent advantage of using the NOD-scid IL2Rγ^null strain is that gene ablation of the IL2Rγ chain leads to absolute NK deficiency (12, 13) and results in remarkable consistency of human melanoma engraftment (this study). Additionally, NOD-scid IL2Rγ^null do not develop thymic lymphoma with age, thus allowing for long-term studies in these mice. In contrast, NOD-scid and NOD-scid β2m^null animals have a high incidence of thymic lymphomas that are fatal (10).

NK function is regulated by inhibitory and activating cell surface receptors (40). NKG2D, a C-type lectin-like molecule, is a major NK activating receptor that interacts with a diverse array of ligands (41, 42). NKG2D ligands include retinoic acid early transcript 1 proteins and minor histocompatibility protein H60 and MULTI in mice and MICA, MICB, and ULBPs in human (32). Recent data in both NKG2D-deficient mice and gain-of-function studies using a novel chimeric NKG2D construct support the critical role of NKG2D in tumor surveillance (43, 44). Human melanomas and various carcinomas have been reported to express MICA/B and/or ULBPs and both ligands have been shown to be involved in human NK recognition of tumors (20, 45). Interestingly, NKG2D from one species can bind ligands from another; for example muNKG2D has been previously shown to bind human ligands MICB, ULBP-1, and ULBP-2 (33, 34). Several previous studies have examined NK cell function in NOD mice and shown some degree of impairment in target cell recognition (compared with the C57BL/6 strain; refs. 46, 47); however, little direct information regarding NOD-scid NK cell function in the context of xenograft recognition is available. Our results show that MICA/B-muNKG2D interaction leads to recognition of human melanoma, which may, in turn, result in decreased tumor engraftment in certain immunodeficient strains. The beige-scid mouse has also been reported to be permissive to human melanoma xenografts and as shown by Elsner and colleagues (48), MICA/B seems to be the tumor associated rejection ligand recognized by muNKG2D. Interestingly, beige-scid mice have large numbers of DX5+ NK cells that express low levels of NKG2D (data not shown). Thus, our results are consistent with the observation that murine NKG2D recognition of MICA/B and ULBPs may prevent/delay engraftment of many human solid tumors in immunodeficient mouse strains.

Small animal imaging provides significant advantages to monitor tumor growth in experimental models (49). Whole body BLI with luciferase reporters is remarkably sensitive, facile to execute, serially, and allows detection of early metastases in various tissues. MRI is less sensitive but has inherent advantages including superior spatial resolution and no requirement for reporter constructs. A recent report evaluating positron-emission tomography, X-ray computed tomography, and BLI modalities documents the ability for detecting human melanoma lung metastases (A375-M) in scid mice at day 45 (50). In our model, ¹⁸F-FDG-PET imaging was done on several animals, which could, in fact, detect A375-RhoC lung metastases as early as day 30 (data not shown).

In summary, our study highlights muNKG2D recognition of MICA/B as an important receptor-ligand interaction used by NK cells in immunodeficient strains to limit engraftment of human tumors. Because NOD-scid IL2Rγ^null mice have an absolute NK cell deficiency as well as additional innate and adaptive defects, this strain seems to be the best recipient for human melanoma.

No potential conflicts of interest were disclosed.

We thank Richard Hynes, John DiPersio, Timothy Graubert, Marco Colonna, and Thomas Spies for reagents; Julie Ritchey and Matthew Holt for excellent technical assistance; Siteman Cancer Center High Speed Cell Sorter Core and Small Animal Cancer Imaging Core facilities for expert assistance; and Wayne Yokoyama and Todd Fehniger for advice.