MICAgen Mice Recapitulate the Highly Restricted but Activation-Inducible Expression of the Paradigmatic Human NKG2D Ligand MICA

NKG2D is a potent activating immunoreceptor expressed on nearly all cytotoxic lymphocytes promoting their cytotoxicity against self-cells expressing NKG2D ligands (NKG2DLs). NKG2DLs are MHC class I-like glycoproteins that usually are not expressed on “healthy” cells. Rather, their surface expression is induced by various forms of cellular stress, viral infection, or malignant transformation. Hence, cell surface NKG2DLs mark “dangerous” cells for elimination by cytotoxic lymphocytes and therefore can be considered as “kill-me” signals. In addition, NKG2DLs are up-regulated on activated leukocytes, which facilitates containment of immune responses. While the NKG2D receptor is conserved among mammals, NKG2DL genes have rapidly diversified during mammalian speciation, likely due to strong selective pressures exerted by species-specific pathogens. Consequently, NKG2DL genes are not conserved in man and mice, although their NKG2D-binding domains maintained structural homology. Human NKG2DLs comprise two members of the MIC (MICA/MICB) and six members of the ULBP family of glycoproteins (ULBP1–6) with MICA representing the best-studied human NKG2DLs by far. Many of these studies implicate a role of MICA in various malignant, infectious, or autoimmune diseases. However, conclusions from these studies often were limited in default of supporting in vivo experiments. Here, we report a MICA transgenic (MICAgen) mouse model that replicates central features of human MICA expression and function and, therefore, constitutes a novel tool to critically assess and extend conclusions from previous in vitro studies on MICA. Similarly to humans, MICA transcripts are broadly present in organs of MICAgen mice, while MICA glycoproteins are barely detectable. Upon activation, hematopoietic cells up-regulate and proteolytically shed surface MICA. Shed soluble MICA (sMICA) is also present in plasma of MICAgen mice but affects neither surface NKG2D expression of circulating NK cells nor their functional recognition of MICA-expressing tumor cells. Accordingly, MICAgen mice also show a delayed growth of MICA-expressing B16F10 tumors, not accompanied by an emergence of MICA-specific antibodies. Such immunotolerance for the xenoantigen MICA ideally suits MICAgen mice for anti-MICA-based immunotherapies. Altogether, MICAgen mice represent a valuable model to study regulation, function, disease relevance, and therapeutic targeting of MICA in vivo.

NKG2D is a potent activating immunoreceptor expressed on nearly all cytotoxic lymphocytes promoting their cytotoxicity against self-cells expressing NKG2D ligands (NKG2DLs). NKG2DLs are MHC class I-like glycoproteins that usually are not expressed on "healthy" cells. Rather, their surface expression is induced by various forms of cellular stress, viral infection, or malignant transformation. Hence, cell surface NKG2DLs mark "dangerous" cells for elimination by cytotoxic lymphocytes and therefore can be considered as "kill-me" signals. In addition, NKG2DLs are up-regulated on activated leukocytes, which facilitates containment of immune responses. While the NKG2D receptor is conserved among mammals, NKG2DL genes have rapidly diversified during mammalian speciation, likely due to strong selective pressures exerted by species-specific pathogens. Consequently, NKG2DL genes are not conserved in man and mice, although their NKG2D-binding domains maintained structural homology. Human NKG2DLs comprise two members of the MIC (MICA/MICB) and six members of the ULBP family of glycoproteins (ULBP1-6) with MICA representing the beststudied human NKG2DLs by far. Many of these studies implicate a role of MICA in various malignant, infectious, or autoimmune diseases. However, conclusions from these studies often were limited in default of supporting in vivo experiments. Here, we report a MICA transgenic (MICAgen) mouse model that replicates central features of human MICA expression and function and, therefore, constitutes a novel tool to critically assess and extend conclusions from previous in vitro studies on MICA. Similarly to humans, MICA transcripts are broadly present in organs of MICAgen mice, while MICA glycoproteins are barely detectable. Upon activation, hematopoietic cells up-regulate and proteolytically shed surface MICA. Shed soluble MICA (sMICA) is also present in plasma of MICAgen mice but affects neither surface NKG2D expression of circulating NK cells nor their functional recognition of MICA-expressing tumor cells. Accordingly, MICAgen mice also show a delayed growth of MICA-expressing B16F10 tumors, not accompanied by an emergence

INTRODUCTION
Expression of the activating immunoreceptor NKG2D facilitates immunorecognition and elimination of "harmful, " activated, or stressed cells by cytotoxic lymphocytes (1)(2)(3). NKG2D ligates several MHC class I-related glycoproteins, which are usually not expressed on "healthy" cells but inducibly expressed upon viral infection, malignant transformation, or cellular activation (2,4). In humans, such inducible NKG2D ligands (NKG2DLs) include two members of the MIC (MICA and MICB) and six members of the ULBP family (ULBP1-6) of MHC class Irelated glycoproteins (1,4,5). Although all MIC and ULBP molecules share MHC class I-like α1α2 superdomains, only MICA/B molecules contain an additional immunoglobulin-like α3 ectodomain alike MHC class I molecules (4,6). Infection of human cells with various types of viruses, including herpesviruses and adenoviruses, strongly induces NKG2DL expression, which often is counteracted by NKG2DL-specific viral immunoevasins (7)(8)(9)(10). Also, activation of hematopoietic cells can induce a pronounced surface expression of NKG2DLs (11)(12)(13), which has been suggested to contribute to the modulation and control of immune responses (11,14). A functional relevance of NKG2DLs has also been implicated in many different autoimmune diseases (15). But foremost, the broad expression of MIC and ULBP molecules by tumor cell lines and human tumors in contrast to the corresponding non-malignant tissues has spawned most interest in harnessing the NKG2D/NKG2DL system for cancer immunotherapy (16)(17)(18). Specifically, MICA is expressed by many different tumors and therefore constitutes an attractive target for immunotherapy (16,19). However, expression of NKG2DLs by established human tumors also suggests that NKG2D-mediated cancer immunosurveillance is paralyzed in such tumors. Several immune escape mechanisms sabotaging NKG2D-mediated tumor surveillance have been reported, including down-regulation of NKG2D by cytokines such as TGFβ (18,20) or tumor-derived soluble NKG2DL (sNKG2DL) as well as of down-regulation of NKG2DL expression by proteolytic shedding, miRNA, or cytokines (20)(21)(22)(23). Hence, following the excitement created by early tumor rejection studies in mice (24,25), subsequent reports on various NKG2D immune escape mechanisms in human cancer dampened this enthusiasm to some extent. A major problem in the field of disease relevance of human NKG2DLs and their therapeutic targeting is that human NKG2DLs are not conserved in mice. Therefore, many hypotheses on the relevance and pathogenic role of human NKG2DLs in malignant and autoimmune diseases derived from in vitro studies or correlative studies cannot easily be verified or falsified in appropriate mouse models. This includes, for example, hypotheses on the functional relevance of sNKG2DL in malignant disease or on NKG2DL expression by the intestinal epithelium for gastrointestinal diseases. Here, we present a transgenic mouse model for the paradigmatic human NKG2DL MICA, which replicates central aspects of MICA expression reported for the human situation. We anticipate that this mouse model will allow insightful studies on the regulation of MICA expression and functional relevance of MICA in immune responses and disease settings in vivo. As a first example, we report that surface MICA expression can be induced by activation of splenocytes accompanied by shedding of soluble MICA (sMICA), which is also detectable in plasma of MICAgen mice and apparently does not systemically downregulate NKG2D expression.

Mice
MICA transgenic mice (MICAgen mice) harboring a 23.4 kb genomic fragment comprising the human MICA * 007 gene were generated as follows: one-cell embryos of (C57BL/6 × SJL) F1/J hybrid mice were microinjected with a 23.4 kb XbaI/SalI fragment isolated from a cosmid encompassing the entire human MICA gene (position 31,393,601 to 31,417,018 on Chr6p21 according to GRCh38.p12). The cosmid MB24/1 was isolated from a human cosmid library of human genomic DNA (26) by screening with a MICA cDNA probe. Eggs were transferred into the oviducts of B6 CBA mice, and offspring was tested for the presence of the MICA transgene using genomic DNA from tail snips. Resulting MICA transgenic mice were termed MICAgen mice and have been continuously backcrossed with C57BL/6J mice (Envigo, Horst, Netherlands) for more than 20 years at our local animal facilities before the beginning of the experiments. Non-transgenic littermates (nontgLM) of MICAgen mouse breedings were used for controls. The cosmid MB24/1 had also been transfected into mouse L-tk cells with MICA-expressing L-tk transfectants being used for cloning of the corresponding MICA cDNA by reverse transcription-PCR (RT-PCR). Transgenic mice expressing this human MICA cDNA (MICA * 00701 allele; GenBank accession number AY750850.1) under control of the H2-K b promoter have been termed H2-K b -MICA and been described elsewhere (27,28). Litters were tested transgenes by PCR using genomic DNA or for MICA surface expression on peripheral blood lymphocytes (PBL) by flow cytometry using anti-MICA mAb AMO1-AlexaFluor647. NKG2D-deficient (Klrk1 −/− ) mice (29) were kindly provided by Dr. Bojan Polic. Animals were maintained under specific pathogen-free conditions in the animal facilities of the University Hospital Frankfurt am Main, Germany. Animal experiments were approved by the local authorities (Regierungspräsidium Darmstadt, Germany, permit nos. FU/Anz. 1035, FU/Anz. 1037, and FU/1115) and performed in full compliance with the respective national guidelines.

Cells
B16F10-mock and B16F10-MICA * 001 transductants were previously described (27). B16F10 transductants were cultured in Dulbecco's modified Eagle medium (DMEM) (Sigma-Aldrich, Steinheim, Germany) supplemented with 10% fetal calf serum (FCS) (Biochrom, Berlin, Germany), 2 mM L-glutamine (Sigma-Aldrich), 100 U/ml penicillin, 100 µg/ml streptomycin (Sigma-Aldrich), 1 mM sodium pyruvate (Life Technologies, Darmstadt, Germany), and non-essential amino acids (Sigma-Aldrich). C1R cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 (Sigma-Aldrich) supplemented with 10% FCS, 2 mM L-glutamine, 100 U/ml penicillin, 100 µg/ml streptomycin, 1 mM sodium pyruvate, and 1.8 mg/ml G418 (Carl Roth, Karlsruhe, Germany). For isolation of blood lymphocytes, blood was collected from hearts of euthanized mice, and cells were washed with phosphate-buffered saline (PBS) and resuspended in 5 ml of BD pharm lyse buffer (BD Biosciences, Heidelberg, Germany) for 5 min at RT for lysis of erythrocytes. After washing with PBS, lysis of erythrocytes was repeated, and cells were washed and resuspended with fluorescence-activated cell sorting (FACS) buffer (PBS, 2% FCS, 2 mM EDTA, and 0.01% sodium azide). Freshly isolated spleens were minced and passed through a 40 µm cell strainer (Greiner Bio-One, Kremsmünster, Austria). Resulting single-cell suspensions were washed with PBS, centrifuged, and subjected to lysis of erythrocytes. After another passage through a 40 µm cell strainer, cells were washed again with PBS and resuspended in FACS buffer or medium. Livers, lymph nodes, and lungs of mice were minced and passed through a 40 µm cell strainer (Greiner Bio-One) to obtain a single-cell suspension. After washing with PBS, liver cells were resuspended with BD pharm lyse buffer (BD Biosciences) and incubated for 5 min at RT to eliminate erythrocytes. Subsequently, liver cells, as well as single-cell suspensions of lymph nodes and lungs, were washed with PBS and resuspended in FACS buffer for flow cytometry.

In vitro Assays With Splenocytes
For in vitro assays, freshly isolated mouse splenocytes (see above) were resuspended at 1 × 10 6 cells/ml in complete RPMI 1640 supplemented with 10% FCS, 2 mM L-glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, 100 µg/ml streptomycin, 50 µM β-mercaptoethanol, and non-essential amino acids. Induction of cell surface expression of MICA was assessed upon exposure to either 10 µg/ml lipopolysaccharide (LPS) or to a combination of 50 ng/ml phorbol myristate acetate (PMA) and 1 µM ionomycin (PMA/I) (all from Sigma-Aldrich). In some experiments, splenocytes were continuously exposed to either PMA/I or LPS for 8 to 72 h before analysis. In other experiments, splenocytes were short-term treated with PMA/I for either 0.5 h or 2 h. Afterwards, splenocytes were repeatedly washed with PBS and in vitro cultures continued in the absence of PMA/I for up to 96 h. To assess modulation NKG2D surface expression by membrane-bound MICA, fresh single-cell suspensions of spleens from nontgLM, MICAgen, and H2-K b -MICA mice were prepared in medium as described above. NK cells were purified from spleens of nontgLM using the mouse NK cell isolation kit II (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's protocol and labeled with carboxyfluorescein succinimidyl ester (CFSE) by incubation for 20 min with 0.5 µM CFSE (Thermo Fisher Scientific, Waltham, MA). After washing, CFSE-labeled NK cells were co-cultured with splenocyte cultures (at ∼1 × 10 6 cells/ml) for 24 h in a 24-well plate and subsequently analyzed for their NKG2D surface expression. To assess modulation of NKG2D surface expression by shed sMICA, splenocyte cultures (1 × 10 6 cells/ml) were seeded into wells of a 24-well plate and costar transwell permeable supports (24 well, 1 µm pore size) (Corning, Corning, NY) containing CFSE-labeled NK cells (5 × 10 6 cells/ml) placed atop. NKG2D surface expression of CFSE-labeled NK cells was determined by flow cytometry after 24 h in vitro culture at 37 • C.

Immunoblotting
For immunoblot analysis, ex vivo stimulated splenocytes were lysed using NP-40 lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, and 1% NP-40) containing the Complete Protease Inhibitor Cocktail (Roche). For deglycosylation of proteins, lysates, or culture supernatants were treated with PNGaseF (New England BioLabs, Frankfurt, Germany) according to the manufacturer's instructions for 2 h at 37 • C. Cellular lysates were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) in a Mini-PROTEAN Tetra Cell (Bio-Rad, Munich, Germany) under reducing conditions transferred to polyvinylidene difluoride (PVDF) membranes (GE Healthcare, Munich, Germany) by semi-dry blotting. Membranes were blocked by incubating with 5% milk powder dissolved in 0.1% TBST for 2 h at RT and MICA detected with either anti-MICA/B mAb BAMO1 followed by HRP-conjugated goat anti-mouse IgG antibodies or biotinylated BAMO1 followed by SA-HRP. BAMO1 was used at 10 µg/ml. Secondary reagents were from Jackson ImmunoResearch Laboratories and used at 1:10,000 dilution. Membranes were developed by HRPjuice PLUS (PJK, Kleinblittersdorf, Germany). Membranes were stripped with ReBlot Plus Mild Antibody Stripping Solution (Merck Millipore, Billerica, MA) and stained with HRPconjugated anti-β-actin (1:10,000 diluted, clone AC-15, Sigma-Aldrich) as loading control.

Cytotoxicity Assay
To activate NK cells, nontgLM, MICAgen, and Klrk1 −/− mice were injected with 15 µg poly(I:C)/g body weight [poly(I:C); InvivoGen, San Diego, CA]. Sixteen hours later, spleens were harvested and minced and passed through a 40 µm nylon mesh, and splenocytes were washed with ice-cold PBS. Erythrocytes were removed by gradient centrifugation with Ficoll-Paque Plus (GE Healthcare), and NK cells were isolated from splenocytes using the mouse NK cell isolation kit II (Miltenyi Biotec) according to the manufacturer's protocol. B16F10 transductants were labeled with 50 µCi of 51 Cr (PerkinElmer, Waltham, MA) for 2 h at 37 • C. After washing, 51 Cr-labeled cells were incubated for 4 h at 37 • C with purified NK cells at different effector (E) to target (T) ratios. Subsequently, supernatants were mixed with OptiPhase SuperMix scintillation mixture in an IsoPlate-96 and measured with a MicroBeta2 plate counter (all from PerkinElmer). Spontaneous chromium release of target cells was always less than 17% of the maximum release of target cells lysed in 1% Triton X-100. Percentage of lysis was calculated as follows: 100 × (experimental release -spontaneous release)/(maximum release -spontaneous release). Experiments were performed in triplicates.

Statistical Analysis
Statistical analyses as detailed in the figure legends were performed using Prism 7 GraphPad (GraphPad Software, San Diego, CA).

MICA Gene Expression by MICAgen Mice
To enable experimental in vivo studies on the regulation, function, and disease relevance of MICA, we sought to establish transgenic mice with the complete human MICA gene. To this aim, cosmid MB24/1 was isolated from a cosmid library of human genomic DNA (26) by screening with a human MICA cDNA. A 23.4 kb fragment of cosmid MB24/1 comprising the entire coding sequence of human MICA * 007 gene (six exons spread across 11.7 kb) plus 10 kb upstream of the translational start site and 1.7 kb downstream of the polyadenylation site (Figure 1) was introduced into the germline of (C57BL/6 × SJL) F1/J hybrid mice. These mice, termed MICAgen mice, were then repeatedly backcrossed to the C57BL/6 background for more than two decades. MICAgen mice present without any obvious abnormalities in phenotype, growth, fertility, and organs and are apparently healthy (not shown). In humans, ample MICA transcripts are present in a wide variety of tissues (32). Similarly, MICA transcripts were broadly detected in all examined organs/tissues of MICAgen mice (Figure 2A) revealing that the human MICA gene can be expressed in rodents, although they do not have MIC genes. MICA transcript levels varied not more than 10-fold between organs of MICAgen mice, with higher relative levels detected for both immune barrier organs (skin, lung, and intestines) and hematopoietic organs (bone marrow, lymph nodes, and spleen) as compared with nonbarrier, non-hematopoietic organs (brain, heart, kidney, and liver) (Figure 2A). An exception is the thymus where low MICA transcript levels probably are due to the low MICA expression by T cells (data not shown). In contrast, MICA transcript levels in organs of the previously reported H2-K b -MICA transgenic mice expressing the MICA * 007 cDNA under the control of the MHC class I H2-K b promoter (28) drastically varied (∼50-fold to 100-fold) between hematopoietic organs and non-hematopoietic organs (Figure 2B), which reflects the strongly enhanced activity of MHC class I promoters in hematopoietic cells vs. nonhematopoietic cells (33). In line with this, non-hematopoietic barrier organs harboring a high number of immune cells (e.g., the lung and intestines) exhibit intermediate MICA transcript levels ( Figure 2B). Importantly, MICA transcript levels in nonhematopoietic organs of MICAgen mice are similar to those of H2-K b -MICA mice ( Figure 2C) and, hence, in a physiologically relevant range.

Restricted Tissue Expression of MICA Glycoproteins in MICAgen Mice
In accord with the high abundance of MICA transcripts in hematopoietic organs of H2-K b -MICA mice and our previous reports on MICA expression by H2-K b -MICA hematopoietic cells (27,28), we readily detected MICA expression by leukocytes in tissue sections of lymph nodes, thymus, and the large and small intestines of H2-K b -MICA mice ( Figure 3A). In contrast, MICA molecules could be or not be unambiguously detected in sections of the corresponding organs of MICAgen mice, which presented alike the negative controls from nontransgenic littermates (nontgLM) (Figure 3A). Specifically, no MICA molecules were detectable in the thymus and intestinal epithelium of MICAgen mice, in contrast to reports of MICA expression in the corresponding human tissues (34). In skin, where MICAgen mice and H2-K b -MICA mice exhibit similar amounts of MICA transcripts (Figure 2), single MICA-positive cells are present in H2-K b -MICA, while such cells could not be unambiguously identified in MICAgen mice ( Figure 3B). Altogether, these data are in line with most studies not or only barely detecting MICA molecules in non-malignant, nondiseased human tissues (16,34) in spite of abundant MICA transcripts. These observations support the notion that the strict post-transcriptional control of MICA glycoprotein expression is maintained in MICAgen mice.

Restricted Expression of MICA Molecules by MICAgen Splenocytes
As previously reported, splenocytes of H2-K b -MICA mice show a strong MICA surface expression (28). In contrast, cell surface MICA was not readily detectable on total splenocytes of MICAgen mice by flow cytometry (Figure 4A). However, when gating on subsets of splenocytes ( Figure S1) and using a potent staining procedure, a differential MICA expression pattern was observed: while T cells of MICAgen mice completely are devoid of surface MICA, a marginal MICA surface expression was detected for both B cells and NK cells, which was validated by comparative stainings of non-transgenic splenocytes ( Figure 4B). Splenic myeloid cells (CD11b + Gr1 + ) showed a substantial MICA expression ( Figure 4B) in line with previous reports of MICA expression by human myeloid cells (12,35). Previous in vitro studies reported an induced MICA surface expression upon activation of human leukocytes (11)(12)(13)(14)35). Hence, we wondered whether such an activation-induced MICA expression is recapitulated by MICAgen splenocytes. When potently stimulating MICAgen splenocytes in vitro with PMA/ionomycin, surface MICA expression by the various lymphocyte subsets was induced within 24 h, most strongly on B cells, but also on NK cells and T cells (Figure 4C). For B cells, an increase of surface MICA expression was already detectable at 8 h of stimulation and peaked at ∼24 h ( Figure 4C). Similarly, MICA expression on NK and T cells reached a maximum at 24-48 h after stimulation, although at lower levels ( Figure 4C and data not shown). We then addressed induction of MICA expression by a physiologically relevant agent and therefore treated MICAgen splenocytes with the TLR4 ligand LPS. In line with TLR4 expression by mouse B cells, we detected a substantial induction of surface MICA on B cells 24 h after LPS exposure in vitro, but not on NK cells and T cells ( Figure 4D). Strong MICA up-regulation on activated B cells was validated by staining with both AMO1 and BAMO1, and by staining of equally treated B cells of nontgLM ( Figure S2). Next, we assessed permanency of MICA surface expression upon induction. To this aim, MICAgen splenocytes were exposed to a pulse of PMA/ionomycin treatment (0.5 or 2 h) and extensively washed, and cultures continued in the absence of PMA/ionomycin. Such a short pulse was sufficient to achieve similar MICA surface expression levels 24 h later such as in continuous presence of PMA/ionomycin. However, MICA surface expression steadily diminished over the following days and reached background levels 4 days after stimulation ( Figure 4E). Immunoblotting of in vitro-activated splenocytes demonstrated that induction surface MICA was due to an enhanced de novo synthesis of MICA glycoproteins upon stimulation and not due to surfacing of intracellularly retained MICA, since MICA molecules were undetectable in non-stimulated MICAgen splenocytes ( Figure 4F). Collectively, these data demonstrate that MICAgen mice are capable to express MICA glycoproteins at the cell surface. But expression of MICA glycoproteins apparently is subjected to a restrictive control, requires induction by activating stimuli such as TLR ligands, and reverts to background levels within a few days. Hence, MICA expression by MICAgen mice recapitulates both the tightly controlled and activation-inducible character of MICA surface expression previously reported for human cells (11)(12)(13)(14).

Soluble MICA in MICAgen Mice
MICA is proteolytically shed from the cell surface of tumor cells by ADAM proteases, and shed sMICA has been detected in sera of many cancer patients, but also in sera of individuals with acute bacterial infections or cholestasis (22,31,36,37). MICA shedding is considered as an immune escape mechanism of tumors evading NKG2D-mediated tumor immunosurveillance (3,38) and has recently been successfully targeted in an immunotherapeutic  approach using preclinical mouse models (39). Hence, we asked whether MICA shedding and its functional implications for cancer immunosurveillance can also be studied in MICAgen mice. To this aim, we first addressed whether MICA is shed by activated MICAgen splenocytes and therefore measured sMICA in culture supernatants of MICAgen splenocytes treated with either PMA/ionomycin or with LPS using a sMICAspecific sandwich ELISA. Low levels of sMICA were already detectable at 24 h of treatment with PMA/ionomycin, or after 48 h treatment with LPS, which strongly increased during the ensuing 24-48 h (Figure 5A). Immunoblotting confirmed the presence of proteolytically processed sMICA in the culture supernatants, which can be distinguished from cell-associated MICA by its reduced molecular mass ( Figure 5B). Next, we asked whether sMICA is also detectable in plasma of MICAgen mice. We consistently detected low but substantial levels of sMICA (∼0.4 ng/ml; range: 0.25-0.55 ng/ml) in plasma of unchallenged MICAgen mice by ELISA (Figure 5C). Such sMICA levels are ∼100-to 300-fold lower than in plasma of H2-K b -MICA mice (28), but roughly in the same range (0.1-2 ng/ml) as reported  for humans with malignant and non-malignant diseases (36,37). Since many studies have implicated elevated sera levels of sMICA in systemic NKG2D down-regulation of cancer patients, we analyzed NKG2D expression by blood NK cells of MICAgen mice. There was no systemic NKG2D down-regulation by sMICA in MICAgen mice, since NKG2D expression levels on peripheral blood NK cells of MICAgen mice were comparable with levels of nontgLM ( Figure 5F). NKG2D surface expression levels on NK cells in spleen, lymph nodes, liver, and lung of MICAgen mice were also not different to matched NK cells of nontgLM, indicating that there was no NKG2D downregulation neither by locally concentrated sMICA nor by any potential membrane-bound MICA (Figure 5F, Figure S3). This is in contrast to NK cells of H2-K b -MICA mice where a strong NKG2D down-regulation has previously been reported as a consequence of permanent engagement of ubiquitously expressed membrane-bound MICA (28). To corroborate our in vivo findings, we finally addressed the issue of MICAinduced NKG2D down-regulation in MICAgen mice in vitro, using MICAgen splenocytes in direct comparison with nontgLM splenocytes and H2-K b -MICA splenocytes. In co-cultures with H2-K b -MICA splenocytes, NKG2D on NK cells was drastically down-regulated as previously reported (28), while co-cultures with splenocytes from either nontgLM or MICAgen mice did not affect NKG2D surface levels ( Figure 5E). When NK cells were spatially separated from splenocytes using a transwell setting, NKG2D surface expression of NK cells was even not altered in cultures with H2-K b -MICA splenocytes shedding high amounts of sMICA, confirming previous results that NKG2D in H2-K b -MICA splenocytes is down-regulated by membrane-bound MICA and not by the high levels of sMICA ( Figure 5D) (28).

MICAgen Mice Are Immunotolerant for MICA and Functionally Recognize MICA + -Syngeneic Mouse Tumors in an NKG2D-Dependent Manner
Therapeutic strategies targeting human MICA molecules, for example, by employing MICA-specific antibodies in cancer immunotherapy, are increasingly getting into the focus of current translational research. In vivo evaluation of the therapeutic efficacy of such antibodies in syngeneic and immunocompetent mouse tumor models requires inoculation of mouse tumor cells ectopically expressing MICA. However, the xenoantigen MICA provokes strong humoral and cellular responses of the adaptive immune system confounding the data on therapeutic efficacy (39). Hence, in such a scenario, MICAgen mice may be of major advantage, as they expectedly exhibit immunotolerance for MICA. To assess MICAgen mice for both functional NKG2D-mediated immunorecognition of MICA + -tumors and for immunotolerance toward the xenoantigen MICA, we used B16F10 melanoma cells ectopically expressing human MICA * 001. B16F10-MICA cells brightly express MICA at the cell surface, which ligates both human and mouse NKG2D (Figures 6A,B). In contrast, control-transduced B16F10-mock cells do not bind mouse NKG2D, as B16F10 cell do not express mouse NKG2D ligands ( Figure 6B). MICA expression rendered B16F10-MICA cells strongly susceptible to cytolysis by purified splenic NK cells from both nontgLM and MICAgen mice, but not from NKG2D-deficient mice (Figure 6C), demonstrating unimpaired NKG2D-mediated immunorecognition of tumorassociated human MICA by MICAgen NK cells. A low number (1 × 10 4 ) of B16F10-MICA or B16F10-mock cells (in Matrigel) was subcutaneously inoculated into either MICAgen mice or nontgLM and tumor growth monitored. Growth of B16F10-MICA tumors as compared with B16F10-mock tumors was markedly delayed in both nontgLM and in MICAgen mice ( Figure 6D). Accordingly, MICA expression by B16F10 tumors significantly improved survival of both nontgLM and MICAgen mice (Figure 6E), indicating that MICAgen mice mount an effective immune response against MICA-expressing tumors. However, such an anti-MICA-directed immune response could be mediated not only via NKG2D but also by humoral and cellular immune responses against the xenoantigen MICA. Therefore, we asked whether MICA-expressing B16F10 tumors raise adaptive anti-MICA responses that may confound these results. In fact, all nontgLM bearing B16F10-MICA tumors raised a strong humoral anti-MICA response within 10-14 days post tumor inoculation, as their plasma contained high titers of MICA-specific antibodies up to 20 µg/ml (Figures 6F,G). In contrast, no anti-MICA antibodies were detectable in plasma of any of the B16F10-MICA tumor-bearing MICAgen mice testifying their immunotolerance for MICA (Figures 6F,G).
Hence, MICAgen mice are immunotolerant for MICA and respond to MICA-expressing tumors primarily in an NKG2Ddependent manner. These findings show that MICAgen mice are more suited for immunotherapeutic studies targeting MICA than conventional mice.

DISCUSSION
About 25 years after the description of the human MICA gene (40) and ∼20 years after the functional characterization of MICA as stress-inducible ligand of NKG2D (1), we recognize an as of yet unmet need to establish a bona fide mouse model for studies of MICA expression and function in vivo.
Although numerous in vitro studies or correlative analyses have investigated a role of MICA in the regulation of immune responses, in the recognition and elimination of tumor or virusinfected cells, and in the pathogenesis of various autoimmune disorders, conclusions from many of these studies are limited by the lack of corroborative in vivo analyses. We believe that the mouse model presented here, in spite of its obvious deficits, will be helpful to advance our understanding of the human NKG2D/NKG2DL system by enabling in vivo studies of MICA expression and function. For example, understanding regulation of MICA expression is key for its function as a "kill-me" signal for NKG2D-expressing cytotoxic lymphocytes. Our data presented here suggest that MICA expression in MICAgen mice is subjected to a similarly restrictive control as reported for humans: although we abundantly detected MICA transcripts in a wide variety of organs, we could not detect MICA molecules in situ paralleling most reports on a restricted MICA expression by healthy human tissues (16,34). However, we cannot exclude that our in situ detection methodology may have failed to detect very low levels of MICA expression. With the use of more sensitive flow cytometry, we detected some residual MICA expression on splenic subsets such as NK cells and myeloid cells ex vivo. Immunoblotting of freshly isolated splenocytes confirmed that expression of MICA glycoproteins is restricted in MICAgen mice and only becomes detectable upon activation of splenocytes. Specifically, we find that splenic B cells drastically up-regulate MICA upon activation, an observation that has not yet been reported for humans and needs to be addressed. In contrast, several studies had reported up-regulation of MICA by activated human myeloid cells, for example, upon treatment with TLR ligands, or by activated human T cells (11)(12)(13)(14). Hence, regulation and functional consequences of activation-induced up-regulation of MICA can be studied using MICAgen mice. MICAgen mice do not only provide unlimited material for in vitro studies of MICA induction but also allow to study MICA regulation and induction in vivo using genetically-altered mouse models, treatment with small molecules or carcinogens, or by inducing autoimmune diseases. In the context of cancer immunology, it will be of particular interest to address how MICA expression is induced and regulated during malignant transformation, how MICA expression tunes anti-tumor immune response, and how tumor mediated shedding of MICA, and sMICA itself impacts on the disease course. We also confirmed the immunological supposition that MICAgen mice are immunotolerant for MICA, which renders them a valuable tool for in vivo experiments with MICA-expressing tumor cells and/or MICA-targeting antibodies.
Notably, there are previous reports of anti-MICA antibodies generated in cancer patients in the course of disease and/or treatment. Melanoma and non-small cell lung carcinoma patients vaccinated with irradiated, autologous granulocyte-macrophage colony-stimulating factor (GM-CSF)-secreting tumor cells and partially treated with anti-CTLA-4 antibodies generated high levels of anti-MICA antibodies, which were ascribed antitumor effects, for example, by neutralizing sMICA, increasing NKG2D expression, and complement-mediated tumor lysis (41). In several patients with monoclonal gammopathy of undetermined significance (MGUS), characterized by a restricted clonal expansion of antibody-secreting plasma cells in the bone marrow, the occurrence of anti-MICA antibodies was reported paralleling a pronounced MICA expression on the expanding plasma cells (42). Of note, patients with multiple myeloma (MM), which occasionally evolves from MGUS, show lower MICA surface expression on MM cells and lower anti-MICA antibody levels but strongly increased sMICA levels, which may foster disease progression (42). It will be of considerable interest to cross MICAgen mice with mouse models developing MMlike diseases (43) to thoroughly investigate the mechanisms of MICA induction and shedding by malignant cells, the emergence of anti-MICA antibodies, and their therapeutic benefit for the course of disease. Our characterization of MICAgen mice also resulted in a couple of observations that may importantly contribute to our understanding of MICA biology and function. Since the first reports on MICA shedding and sMICA in sera of cancer patients (31,44), there have been dozens of studies with conflicting data on the effects of sMICA molecules on NKG2D surface expression (45)(46)(47). While, numerous studies claimed that sMICA molecules directly down-regulate NKG2D on NK cells and T cells, other studies found no evidence for an NKG2D down-regulation by sMICA (20,47). Here, we find that NKG2D is neither systemically down-regulated on peripheral blood NK cells of MICAgen mice in spite of substantial amounts of sMICA plasma levels comparable with levels detected in many cancer patients (36,37) nor on NK cells in various organs of MICAgen mice where sMICA levels may be locally concentrated (e.g., in spleen or lymph nodes). We further corroborate our previous findings (28) that membrane-bound MICA of H2-K b -MICA splenocytes downregulates NKG2D upon direct cellular interaction with NK cells, while there is no down-regulation in a transwell setting where only shed sMICA acts on NK cells. Collectively, all our data strongly argue against an NKG2D down-regulation by shed sMICA, while some previous observations on NKG2D downregulation in cancer patients may be attributable to MICA embedded in exosomes or other tumor-derived vesicles or by other tumor-derived soluble factors such as TGF-β (20,45,46). Further, our data suggest that sMICA in plasma of MICAgen mice is, at least in part, derived from hematopoietic cells that up-regulated and shed MICA upon activation (Figure 5). Such sMICA may differ in its effects on NKG2D modulation from tumor-released MICA, especially when embedded in exosomes (45,48,49). Of note, activation-induced surface MICA is vanishing over several days past activation, and it will be interesting to determine the mechanisms not only underlying MICA induction but also underlying the reversal of MICA surface expression. Also for this purpose, a robust mouse model such as MICAgen mice amenable to genetic experiments will be valuable.
However, since MICA is highly polymorphic with more than 150 alleles reported to date encoding for more than 90 different MICA proteins (50,51), certain results obtained from studies of this mouse model expressing MICA * 007 may not be generally applicable to all MICA variants: these limitations may specifically relate to functional in vivo studies with regard to MICA allelic variants substantially differing from MICA * 007 in the affinity for NKG2D (such as MICA * 004) (5,52,53) or in the membrane localization and mode of release (such as MICA * 008) (54,55).
Taken together, we here provide a thorough characterization of MICAgen mice, which apparently express the human MICA gene in a strictly controlled but inducible manner very much alike that of humans. Therefore, these mice will be helpful in the context of (i) translational and therapeutic research requiring MICA-immunotolerant mice, (ii) studies on the regulation and induction of MICA expression, and (iii) studies on the functional relevance of MICA in the context of immune responses to infections, tumors, and autoimmune diseases.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

ETHICS STATEMENT
The animal study was reviewed and approved by Regierungspräsidium Darmstadt.

AUTHOR CONTRIBUTIONS
YK and CB performed the experiments and analyzed the data. MB provided some reagents and critical input. AS conceptualized the study, designed the experiments, and wrote the manuscript with the support of YK and CB.