A Soluble Fragment of the Tumor Antigen BCL2-associated Athanogene 6 (BAG-6) Is Essential and Sufficient for Inhibition of NKp30 Receptor-dependent Cytotoxicity of Natural Killer Cells*

Background: Cellular ligands of the activating natural killer (NK) cell receptor NKp30 are poorly characterized. Results: The identified domain of the cellular ligand BCL2-associated athanogene 6 (BAG-6) is essential and sufficient to bind NKp30 and inhibits NK cell function. Conclusion: The BAG-6 domain from amino acid 686 to 936 is an important element of BAG-6-dependent tumor immune escape. Significance: This study gives the first molecular insights into BAG-6-mediated inhibition of NKp30-dependent NK cell cytotoxicity. Immunosurveillance of tumor cells depends on NKp30, a major activating receptor of human natural killer (NK) cells. The human BCL2-associated athanogene 6 (BAG-6, also known as BAT3; 1126 amino acids) is a cellular ligand of NKp30. To date, little is known about the molecular details of this receptor ligand system. Within the current study, we have located the binding site of NKp30 to a sequence stretch of 250 amino acids in the C-terminal region of BAG-6 (BAG-6686–936). BAG-6686–936 forms a noncovalent dimer of 57–59 kDa, which is sufficient for high affinity interaction with NKp30 (KD < 100 nm). As our most important finding, BAG-6686–936 inhibits NKp30-dependent signaling, interferon-γ release, and degranulation of NK cells in the presence of malignantly transformed target cells. Based on these data, we show for the first time that BAG-6686–936 comprises a subdomain of BAG-6, which is sufficient for receptor docking and inhibition of NKp30-dependent NK cell cytotoxicity as part of a tumor immune escape mechanism. These molecular insights provide an access point to restore tumor immunosurveillance by NK cells and to increase the efficacy of cellular therapies.


Immunosurveillance of tumor cells depends on NKp30, a major activating receptor of human natural killer (NK) cells.
The human BCL2-associated athanogene 6 (BAG-6, also known as BAT3; 1126 amino acids) is a cellular ligand of NKp30. To date, little is known about the molecular details of this receptor ligand system. Within the current study, we have located the binding site of NKp30 to a sequence stretch of 250 amino acids in the C-terminal region of BAG-6 (BAG-6 686 -936 ). BAG-6 686 -936 forms a noncovalent dimer of 57-59 kDa, which is sufficient for high affinity interaction with NKp30 (K D < 100 nM). As our most important finding, BAG-6 686 -936 inhibits NKp30-dependent signaling, interferon-␥ release, and degranulation of NK cells in the presence of malignantly transformed target cells. Based on these data, we show for the first time that BAG-6 686 -936 comprises a subdomain of BAG-6, which is sufficient for receptor docking and inhibition of NKp30-dependent NK cell cytotoxicity as part of a tumor immune escape mechanism. These molecular insights provide an access point to restore tumor immunosurveillance by NK cells and to increase the efficacy of cellular therapies. NK 2 cells eliminate tumor cells by the polarized release of cytotoxic granules containing perforin and granzymes or by binding of the tumor necrosis factor family members FAS/ CD95, TRAIL receptor (TRAILR), and tumor necrosis factor receptor (TNFR) on tumor cells to their cognate ligands FAS-L, TRAIL, and TNF, respectively, on NK cells (1,2). The mechanisms leading to NK cell activation are described by the principles of "missing self" and "induced self," which imply that target cells with low or absent expression of major histocompatibility complex (MHC) class I or other non-MHC class I inhibitory ligands (missing self) and/or stress-induced expression of ligands for activating NK receptors (induced self) are preferentially eliminated by NK cells (3). Thus, a balance of activating and inhibitory signals determines whether NK responses are initiated or not. The major activating receptors involved in recognition and killing of malignantly transformed cells include the natural cytotoxicity receptors (NCRs) NKp30, NKp44, and NKp46, whose few ligands identified so far remain poorly characterized on the molecular level, and NKG2D, recognizing several related ligands such as the MHC class I polypeptide-related sequence A and B (MICA, MICB) and the UL16-binding proteins (ULBPs) (4 -7). The importance of the NCRs for NK cell activation is underscored by the fact that expression of an insufficient amount of NCRs results in resistance of leukemia cells to NK cell cytotoxicity in patients with acute myeloid leukemia (8 -10). Moreover, blocking of these receptors resulted in significantly decreased killing of virus-infected and malignantly transformed cells (6). Recently, the BCL2-associated athanogene 6 (BAG-6, also known as BAT3) was identified as a cellular ligand of NKp30 (11,12). The domain organization and threedimensional structure of BAG-6 are mostly unknown except for an N-terminal ubiquitin-like domain (Protein Data Bank (PDB) IDs: 4EEW and 4DWF). Additionally, a domain of unknown function (DUF3538) and a C-terminal BAG domain were identified (see Fig. 1A). BAG-6 is present in various tissues, on the plasma membrane of immune cells and tumor cells as well as on exosomes. In addition, BAG-6 is expressed as a soluble protein upon cellular stress (11)(12)(13)(14). Therefore, BAG-6 represents an important tumor antigen and might be part of a tumor immune escape strategy reminiscent of the NKG2D ligands ULBP2 and ULBP3 (14,15).
Recent data suggest that BAG-6 on the plasma membrane of immature DCs triggers NK cell killing, whereas mature DCs escape from killing by up-regulation of MHC class I expression (12). By killing of immature DCs (cells that are associated with the induction of tolerogenic responses), activated NK cells might select a more immunogenic subset of DCs during a protective immune response (16). However, it remains puzzling whether up-regulation of BAG-6 at the plasma membrane of immature dendritic cells is a consequence of NK cell cytotoxicity via a different NK cell receptor-ligand system or a prerequisite for NKp30-BAG-6-dependent killing of immature dendritic cells.
The molecular details of the interaction between NKp30 and BAG-6 are largely unknown. However, detailed knowledge about this receptor-ligand system is required to understand NK cell function and for the development of NK cell-based therapies. Therefore, within the current study, we investigated the interaction of NKp30 and BAG-6 in immunosurveillance of tumor cells.
Flow Cytometry of Cells-NK-92 (0.5-1 ϫ 10 6 cells) were blocked with 2% FCS (v/v) and 5% BSA (w/v) in PBS prior to incubation with specific antibodies or recombinant protein for 1 h at 4°C. Following detection with secondary fluorophoreconjugated antibodies for 1 h at 4°C, a minimum of 20,000 cells were analyzed on a FACSCanto II instrument (BD Biosciences).
Protein Production and Purification-NKp30-IgG1-Fc fusion proteins were produced as described previously (17). BAG-6 686 -936 (with a C-terminal hexahistidine tag) was heterologously expressed in Escherichia coli BL21 cells as a soluble cytosolic protein and in High Five insect cells as a secreted protein (with a C-terminal Strep-tag II).
Fragments of the BAG-6 gene were amplified from human cDNA clone IRAUp969B1019D (isoform 2 (P46379-2); Source BioScience) by PCR using gene-specific primers and were cloned into the pET21a expression vector. Transformed E. coli BL21 cells were grown to an A 600 of 0.6, and protein expression was induced with 0.3 mM isopropyl-␤-D-thiogalactoside for 4 h. Bacteria were harvested by centrifugation and disrupted by sonication. Cell debris was removed by centrifugation following purification of BAG-6 protein from the supernatant by immobilized metal ion affinity chromatography (IMAC) on nickelnitrilotriacetic acid-agarose beads (Qiagen) and subsequent size exclusion chromatography (SEC) in 100 mM HEPES buffer supplemented with 25 mM NaCl (pH 7.5) on a Superdex 200 column (GE Healthcare). To produce BAG-6 variants in insect cells, BAG-6 genes were cloned into transfer vector pFastBac1 with an N-terminal secretion sequence from gp67 and a C-terminal Strep-tag II. Transformed E. coli DH10Bac YFP (kind provided by Imre Berger, Grenoble, France) integrate target genes from the transfer vector into bacmid DNA by Tn7 transposition. Sf9 cells were transfected with bacmid DNA to produce recombinant baculovirus within 72 h at 27°C. Initial viral supernatant (V 0 ) was used to amplify viral particles in Sf9 cells in shaking flasks for a further 72 h at 27°C (V 1 ). Afterward, BAG-6 variants were produced in High Five cells (7 ϫ 10 5 cells/ ml) using V 1 (1:100 (v/v)). After removal of viral particles by ultracentrifugation (2 h, 100,000 ϫ g), harvested supernatants were sterile filtered and purified using StrepTactin Sepharose (IBA) and subsequent SEC in 100 mM Tris/HCl (pH 8) supplemented with 150 mM NaCl and 1 mM EDTA.
Size Exclusion Chromatography-Multiangle Laser Light Scattering (SEC-MALS)-Size exclusion chromatography coupled with multiangle light scattering (SEC-MALS) was performed using a TSK-GEL G3000SWXL column (15 ml, Tosoh Bioscience), a light-scattering detector (miniDAWN TREOS) and a refractometer (Optilab rEX) from Wyatt Technology, and a UV detector, an HPLC, pump, and degasser from Jasco. The system was equilibrated with 2 column volumes of SEC buffer (25 mM sodium P i , 25 mM NaCl (pH 7.0), filtered through 0.1-m pore size VVLP filters (Millipore)) following a recirculation through the system for at least 1 day at 0.1 ml/min to improve the base line. Per measurement, 150 -300 g of protein in 100 l were injected and analyzed at a flow rate of 0.5 ml/min at 4°C. The obtained signals were processed with the ASTRA software package version 5.3.4.13 (Wyatt Technology) to calculate the molecular mass using the refractive index (RI) signal for concentration determination taking a differential index of refraction of 0.185 ml/g.
Analytical Ultracentrifugation-Sedimentation velocity experiments were conducted with 300 l of protein samples at a protein concentration of 0.3-0.6 mg/ml in 100 mM Tris/HCl (pH 8) supplemented with 150 mM NaCl and 1 mM EDTA or 25 mM sodium phosphate buffer (pH 8) with an Optima XL-A centrifuge (Beckman Coulter Instruments). Absorbance data were acquired at a wavelength of 280 nm, at rotor speeds of 40,000 rpm, using a 0.003-cm step scan with three averages, and at a temperature of 20°C. Data were analyzed with the sedfit program (18). The protein partial-specific volume, the buffer density, and viscosity were calculated using the SEDNTERP program, kindly provided by Dr. J. Philo.
Peptide Arrays-Peptide arrays of human BAG-6 (P46379-1; peptides of 18 amino acids offset by one amino acid) were synthesized by Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry at activated PEG spacers on cellulose membranes by automated parallel peptide synthesis on a MultiPep RS instrument (Intavis) and used for binding experiments as described previously (19,20). After saturation of unspecific binding sites, peptide membranes were probed with soluble human NKp30-IgG1-Fc fusion proteins (17). Bound IgG1-Fc fusion proteins were detected with a human IgG1-Fc-specific HRP-conjugated antibody (Sigma-Aldrich) and visualized by chemiluminescence imaging.
CD Spectroscopy-The CD spectra of BAG-6 (7 l; 2 mg/ml) in 25 mM phosphate buffer (pH 8.0) were recorded by using a CaF 2 cuvette with a path length of 50 m. The spectra were collected on a J-720 CD spectrometer (Jasco) with a resolution of 0.1 nm; 10 scans per spectra were taken. The sample holder was temperature-controlled by an external water bath.
Immunoprecipitation-Tagged BAG-6 686 -936 proteins were mixed with IgG1-Fc fusion proteins for 1 h and incubated with 20 l of magnetic beads covalently coated with polyhistidine tag-specific mouse monoclonal antibodies (GenScript) or StrepTactin-coated magnetic beads (IBA) for 1 h at 4°C. As negative control, an IgG1-Fc fusion protein of the human interferon receptor subunit 2 (IFNAR2-IgG1-Fc) was used (17). Bead-associated proteins were eluted with SDS sample buffer by boiling for 5 min (His-tagged BAG-6 686 -936 ) or with 2 mM biotin (Strep-tag II-tagged BAG-6 686 -936 ) and analyzed by SDS-PAGE and Western blotting.

NKp30
Interacts with the C-terminal Part of BAG-6-To determine the binding interface of NKp30 and BAG-6, we probed overlapping peptide arrays (peptides of 18 amino acids in length, offset by one amino acid) covering the entire amino acid sequence of human BAG-6 ( Fig. 1A) with soluble NKp30-IgG1-Fc fusion proteins. After detection with a human Fc-specific antibody, we identified 10 clustered interaction hot spots (I-X) for NKp30 within the C-terminal half of BAG-6 ( Fig. 1B and supplemental S1 and supplemental Table S1). The reactivity of the hot spots ranged from weak (III, IV, VIII, IX) over moderate (VI, VII) to strong (I, II, V, X). For reference, only one spot within region X was recognized on the peptide arrays by the human Fc-specific antibody alone (supplemental Fig. S1), demonstrating high specificity of the binding assay for the other regions. The strongest specific interaction hot spots (I, II, V) were localized in the sequence stretch from amino acids 686 to 936 (BAG-6 686 -936 , nomenclature according to isoform 2 of BAG-6; spots 692-925, compare supplemental Table S1).
Expression and Purification of BAG-6 686 -936 -The BAG-6 fragment, which contained the strongest specific interaction hot spots (I, II, V) for NKp30 (BAG-6 686 -936 ), was cloned with a C-terminal hexahistidine tag or Strep-tag II. These constructs were expressed as soluble cytosolic proteins in E. coli BL21 bacteria (hexahistidine tag) and High Five insect cells (Strep-tag II) after secretion into the culture supernatant. Both proteins were purified to homogeneity by IMAC or by StrepTactin-Sepharose as demonstrated by SDS-PAGE and Western blotting (Fig. 2, A  and B). Per liter of culture, roughly 1.5 and 3 mg of pure protein were obtained from E. coli and insect cells, respectively.
The BAG-6 686 -936 protein appeared as a monomer of roughly 26 kDa (as predicted from its primary sequence, BAG-6 686 -936 (E. coli) 28.6 kDa; BAG-6 686 -936 (High Five cells) 29.3 kDa) in SDS-PAGE and corresponding Western blot analyses. Reducing and nonreducing SDS-PAGE gels demonstrate the absence of covalently linked oligomers (Fig. 2, C and D). Notably, BAG-6 686 -936 from insect cells appears as a double band in SDS-PAGE and corresponding Western blot analyses. The reason for these two species is unknown. However, phosphorylation as well as N-and O-linked glycosylation were excluded because they were not detected by phosphotyrosine-specific antibodies and their electrophoretic mobility remained unchanged in enzymatic deglycosylation assays. Interestingly, BAG-6 686 -936 appeared as a monodisperse peak, which corresponds to a dimer of 57-58 kDa in size exclusion chromatography on a Superdex 200 column (Fig. 2, E and F). The peak annotation and protein purity were validated by Coomassie Blue-stained SDS-PAGE and Western blotting.

BAG-6 686 -936 Forms a Noncovalent Dimer with a Predominantly ␣-Helical
Conformation-To determine the oligomeric state of BAG-6 686 -936 , we performed SEC-MALS measurements. In these experiments, the predominant protein species had a molecular mass of 57 kDa corresponding to a dimer of BAG-6 686 -936 (Fig. 3, A and B). To analyze the oligomerization state of BAG-6 686 -936 in solution, we performed sedimentation velocity measurements. We found that the BAG-6 686 -936 proteins migrate at a sedimentation coefficient of 3.5 s as a single peak corresponding to more than 80% of the material loaded (Fig. 3C), confirming BAG-6 686 -936 dimerization. Based on the SEC-MALS and sedimentation velocity measurements, the molecular mass of the BAG-6 686 -936 dimer is 57-59 kDa, which is in agreement with its theoretical mass (BAG-6 686 -936 (E. coli) 57.2 kDa; BAG-6 686 -936 (High Five cells) 58.6 kDa) calculated from primary sequences. To assess the folding and secondary structure of the BAG-6 686 -936 dimers, we employed CD spectroscopy. The far UV-CD spectrum between 178 and 260 nm (at 22°C) indicates a predominantly ␣-helical conformation (roughly 80%, Fig. 3D). These findings are in accordance with secondary structure prediction performed with the Phyre2 web interface (21). Here, the ␣-helix content of BAG-6 686 -936 is 70% with a high confidence score, whereas annotations for the rest of BAG-6 (except for the ubiquitin-like domain) are poor.
NKp30 Interacts Specifically with BAG-6 686 -936 -To analyze the interaction of BAG-6 686 -936 with NKp30 on a molecular level, we initially performed co-immunoprecipitation experiments. Therefore, recombinant BAG-6 686 -936 from E. coli or High Five insect cells was mixed with a 15-fold molar excess of recombinant NKp30-IgG1-Fc fusion proteins. Protein complexes were recovered on magnetic beads coupled to polyhistidine tag-specific antibodies or StrepTactin and analyzed by SDS-PAGE and Western blotting (Fig. 4, A and B). Based on our results, NKp30 binds specifically to both BAG-6 686 -936 proteins, demonstrating that the BAG-6 686 -936 domain is essential and sufficient for binding to NKp30. In contrast, IgG1-Fc fusion proteins of the interferon receptor subunit IFNAR2 (IFNAR2-IgG1-Fc), used as a negative control, showed no binding to BAG-6 686 -936 , confirming specificity of the assay.
For binding affinity measurements of NKp30 and BAG-6 686 -936 , we developed an ELISA-based assay. In brief, recombinant BAG-6 686 -936 proteins were immobilized on an ELISA plate and probed with graded amounts of NKp30-IgG1-Fc or IFNAR2-IgG1-Fc fusion proteins (negative control). NKp30 bound specifically to BAG-6 in a concentration-dependent manner, whereas IFNAR2 binding was close to the detection limit of the assay (Fig. 4, C and D). After fitting of the binding curves to a 1:1 Langmuir binding model, equilibrium binding constants (K D ) of 32 Ϯ 4 and 64 Ϯ 3 nM were calculated for NKp30 binding to BAG-6 686 -936 produced in E. coli and insect cells, respectively.
Following detailed biochemical characterization of the interaction between NKp30 and BAG-6 686 -936 , we investigated whether the BAG-6 686 -936 fragment binds to NKp30 in the plasma membrane of living NK cells. Therefore, NK-92 cells were incubated with purified BAG-6 686 -936 and analyzed by flow cytometry after detection with tag-specific antibodies. In accordance with the results from ELISA measurements, BAG-6 686 -936 bound specifically to NKp30 in the plasma membrane of NK cells (Fig. 4, E and F).
In summary, BAG-6 686 -936 comprises a subdomain of BAG-6, which is essential and sufficient for receptor docking and inhibition of NKp30-dependent NK cell cytotoxicity. Interestingly, recombinant BAG-6 686 -936 derived from E. coli and High Five insect cells displayed equal biochemical and biophysical properties with respect to its binding affinity to NKp30, folding, and oligomerization. However, although BAG-6 686 -936 derived from E. coli has the ability to suppress degranulation and IFN-␥ production of NK cells, it failed to inhibit NKp30-CD3 chaindependent signaling. In contrast, High Five insect cell-derived BAG-6 686 -936 is a strong inhibitor of NKp30. The precise reason for this phenomenon is unknown. Possible reasons might be subtle changes in protein folding, post-translational modifications other than glycosylation or phosphorylation, or a certain degree of disturbing LPS contamination in E. coli-derived BAG-6 686 -936 (22).

DISCUSSION
The human BAG-6 (also known as BAT3; 1126 amino acids) is a multifunctional protein as indicated by its diverse cellular localization. As a nuclear protein, BAG-6 is involved in p53-dependent DNA repair (23) and in modulation of histone methylation, thus regulating gene expression (24). Further, BAG-6 is cleaved by caspase-3 at position 1001 (DEQD), and the resulting C-terminal fragment of 131 amino acids mediates ricininduced apoptotic morphological changes (25). BAG-6 also promotes stress-induced apoptosis by stabilizing the apoptosisinducing factor AiF (26). In addition, BAG-6 binds to heat shock protein 70 (Hsp70) and acts as co-chaperone (27). BAG-6 links the targeting and ubiquitination pathway as it plays a role in translocation of tail-anchored proteins into the endoplasmic reticulum membrane (28,29). Full-length BAG-6, but not a C-terminally truncated form of BAG-6, binds to TGF-␤ receptors, modulates TGF-␤ signaling, and enhances TGF-␤1-induced type I collagen expression in mouse mesangial cells (30). The fact that homozygous BAG-6 knock-out in mice led to embryonic or perinatal lethality, associated with pronounced developmental defects in several tissues, indicates that BAG-6 is critical for normal mammalian development (31). BAG-6 is up-regulated in colorectal tumor tissue (13). In addition, BAG-6 is released by heat-shocked macrophages, and treatment of macrophages with purified BAG-6 led to reduced nitric oxide after INF-␥ stimulation as well as IL-1␤ and IL-12p70 production upon LPS treatment.
Only very recently, BAG-6 on tumor cells and immature DCs was proposed to be a cellular ligand of the major activating NK cell receptor NKp30 (11,12). However, the molecular details of the NKp30/BAG-6 receptor-ligand system are poorly characterized.
Within the current study, we have localized the binding site of NKp30 on BAG-6 to a sequence stretch of 250 amino acids in the C-terminal half of BAG-6 (BAG-6 686 -936 ). These data are in accordance with a previous study, which identified a C-terminal fragment of BAG-6 (BAG-6-CT, amino acids 555-(⌬ 1055-1111 )-1032) by a yeast two-hybrid screen that interacts with the extracellular domain of NKp30 (11). The identified cDNA fragment of BAG-6 was characterized by a deletion of the conserved BAG domain (amino acids 1055-1111), which is responsible for binding of BAG-6 to Hsp70 (27,32). Both variants, either including or lacking the BAG domain, are expressed in tumor tissues, cell lines, and monocyte-derived dendritic cells (11,13,33).
BAG-6 contains a lengthy proline-rich stretch (amino acids 195-681), making it an attractive molecular scaffold that could mediate multiple protein-protein interactions in several cellular compartments. By contrast to the previous study (11), the identified NKp30 binding fragment of BAG-6 (BAG-6 686 -936 ) does not contain sequences from this proline-rich stretch.
Soluble recombinant BAG-6 686 -936 formed stable noncovalent dimers and bound with nanomolar affinity to NKp30, suggesting that BAG-6 686 -936 forms a separate domain fold within the BAG-6 protein. Therefore, the current study provides the first information on the structural organization of the sequence stretch of BAG-6 in between the N-terminal ubiquitin-like domain (amino acids 1-87) and a C-terminal BAG domain (amino acids 1055-1111). Notably, ab initio modeling of BAG-6 did not reveal significant results.
Strikingly, BAG-6 686 -936 inhibited NKp30-dependent signaling in a reporter cell assay and in NK cells (inhibition of IFN-␥ release, degranulation, and cytotoxicity). Based on these data, we show for the first time that soluble BAG-6 is an inhibitory ligand of NKp30 and might thus be part of a potent tumor immune escape mechanism. Further, BAG-6 is the first inhibitory ligand of NKp30 that acts via binding to the ectodomain of NKp30, mechanistically different from cytomegalovirus pp65, which acts via dissociation of the NKp30-CD3 chain complex (34).
Taken together, these data suggest that soluble, and most likely also exosomal BAG-6 acts as a potent tumor immunoevasin reminiscent of exosomal ULBP3 and soluble ULBP2, which lead to down-modulation of NKG2D (15).