Cellular Mechanisms Controlling Surfacing of AICL Glycoproteins, Cognate Ligands of the Activating NK Receptor NKp80

Natural killer cells are innate lymphocytes with cytotoxic capacity toward virus-infected, transformed, or stressed cells (1, 2). Hence, there is continued interest to harness NK cells for cancer immunotherapy (3–5). Furthermore, NK cells are important producers of cytokines, such as IFN-γ and TNF, thereby modulating immune responses and contributing to tissue homeostasis (6–8). More recently, the potential of NK cells to mediate certain memory responses and to develop into long-lived cells with enhanced effector functions has been appreciated (9–11). According to both the “missing-self” and the “induced-self” hypotheses, NK cells preferentially react against cells with poor MHC class I expression but induced expression of ligands of activating NKRs (2, 12–14). Such cell contact–dependent effector responses of NK cells are regulated by a variety of germline-encoded activating and inhibitory receptors (1, 3) that can be divided into Ig-like and C-type lectin-like receptors (CTLR). The latter are homodimeric or heterodimeric type II transmembrane glycoproteins with a single extracellular C-type lectin-like domain (CTLD) and are encoded in a certain genomic region, termed the NK gene complex (NKC) (15). In humans, NKC-encoded, activating CTLR include NKG2D, CD94/NKG2C, NKp80, and NKp65, whereas CD94/NKG2A and NKR-P1A (CD161) are ITIM-bearing inhibitory receptors (16). Among these, NKR-P1A, NKp80, and NKp65 are peculiar because they engage cognate ligands of the CLEC2 family that are encoded in the NKC in close proximity to their respective receptors (16, 17). Like their receptors, CLEC2 family members are dimeric type II glycoproteins with an amino-terminal cytosolic domain, a transmembrane domain, a short stalk region, and a single C-terminal CTLD, which is stabilized by two or three intramolecular disulfide bonds (15, 18).

NKR-P1A binds CLEC2D-encoded LLT1 glycoproteins expressed on activated B cells, thereby inhibiting cytotoxicity and IFN-γ secretion of NKR-P1A–expressing NK cells (19, 20). The activating receptor NKp65 on the human NK cell line NK-92 engages CLEC2A-encoded KACL molecules with high affinity–stimulating cytotoxicity and IFN-γ secretion (21–23). In contrast to NKp65, NKp80 is expressed on virtually all mature human NK cells (24, 25) but also on subsets of effector memory T cells (26) and binds to CLEC2B-encoded AICL glycoproteins which are inducibly expressed on activated myeloid cells and NK cells (25, 27). Engagement of NKp80 by AICL promotes an activating cross-talk between monocytes and NK cells (25) and between macrophages and effector T cells (26) in the presence of inflammatory cytokines, leading to the secretion of IFN-γ and TNF. Further, NKp80 ligation contributes to the cytolysis of malignant AICL-expressing myeloid cells, suggesting a role of NKp80–AICL interaction in NK-mediated surveillance of myeloid leukemia cells (25, 28).

Resting human NK cells were shown to contain intracellular stores of AICL molecules (27). Therefore, NK cells simultaneously express both NKp80 and its ligand AICL. Intracellular retention in resting lymphocytes has also been shown for the CD69 molecules (29, 30), which are encoded by the CLEC2C locus and are the only human CLEC2 family members without any known receptor acting in trans. However, for both CD69 and AICL, mechanism-governing intracellular retention is yet unknown. Exposure of NK cells to the inflammatory cytokines IL-12 and IL-18 results in surface expression of both AICL and CD69, whereas surface NKp80 becomes downregulated. Such cytokine-activated, AICL-expressing human NK cells are functionally recognized by autologous, resting NK cells in an NKp80-dependent manner, resulting in cytolysis of the stimulated NK cells and cytokine release by resting bystander NK cells (27).

The mechanism of intracellular AICL retention remained elusive but was attributed to the AICL CTLD (27). Hence, we aimed to dissect the molecular determinants in AICL molecules involved in regulation of AICL expression and cellular localization.

Human CD4⁺ T cell line CEM (31) and the calnexin-deficient derivative CEM-NKR (32) were obtained through the National Institutes of Health AIDS Reagent Program (https://www.aidsreagent.org/). Human promonocytic cell line U937 and embryonic kidney cell line 293 were purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (Braunschweig, Germany). CEM, NKR, and U937 cells were cultured in RPMI 1640 and 293 cells were cultured in DMEM. All media contained 10% FCS (Biochrom, Berlin, Germany), 2 mM l-glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 mg/ml streptomycin (all from Sigma-Aldrich, St. Louis, MO). AICL mutants were generated using site-directed mutagenesis (33) and cloned into the pIRES2 vector (Clontech Laboratories, Mountain View, CA) either with C-terminal FLAG and hexahistidine sequences (FH-tag) or without an FH-tag. The 293 cells were transfected using AppliFect (AppliChem, Darmstadt, Germany) and analyzed 3 d later by flow cytometry or immunoblotting. The 293 cells stably expressing AICL mutants were selected by addition of G418 (2 mg/ml), followed by FACSorting of enhanced GFP (eGFP)-positive cells. CEM and NKR cells were transduced with the retroviral vector pMXs-IP (kindly provided by T. Kitamura, University of Tokyo) that encodes for FH-tagged AICL or KACL glycoproteins and transductants selected by addition of puromycin (0.6 μg/ml). Inhibition of protein translation, of proteasomal degradation, and of lysosomal acidification was achieved by adding 30 μg/ml cycloheximide (CHX) (AppliChem), 10 μM MG-132, and 50 nM concanamycin A (CCMA) (both from Sigma-Aldrich), respectively.

Cells were washed with ice-cold buffer (PBS with 2% FCS, 2 mM EDTA, and 0.01% sodium azide) prior to staining with primary Abs for 30 min at 4°C. After washing, cells were stained with allophycocyanin-conjugated F(ab)₂ fragments of goat anti-mouse IgG (Jackson ImmunoResearch, Ely, U.K.) for 20 min at 4°C and washed again. Alternatively, cells were stained with previously described tetramers of NKp80 ectodomains (ED) (25) or NKp65-ED (21) for 30 min at 4°C. For intracellular staining of AICL proteins, cells were fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences, San Jose, CA) for 20 min on ice prior to Ab staining and washing with saponin buffer (PBS, 0.5% BSA, 0.1% saponin, and 0.01% sodium azide). Flow cytometry analysis was performed on a FACSCanto II (BD Biosciences), and data were analyzed using FlowJo 9.9.6 software (Tree Star, Ashland, OR).

Cells were washed twice with PBS and cell pellets lysed with ice-cold lysis buffer (1% Triton X-100, 50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, and 50 mM octyl glucopyranoside, Complete Protease Inhibitor [Roche, Mannheim, Germany]). Cell lysates containing 25–50 μg of total protein were then subjected to SDS-PAGE. For protein deglycosylation, cell lysates were treated with endoglycosidase H (Endo H) or peptide:N-glycosidase F (PNGase F) prior to SDS-PAGE (both from New England Biolabs, Ipswich, MA) according to the manufacturer’s protocol. After blotting onto a Hybond ECL nitrocellulose membrane (GE Healthcare Europe, Freiburg, Germany) or polyvinylidene difluoride membrane (Carl Roth, Karlsruhe, Germany), the membrane was either probed with anti-AICL mAb 7G4 (25) or anti-FLAG mAb (clone M2; Sigma-Aldrich) followed by detection of the unconjugated primary Ab with HRP-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA). For detection of calnexin, polyclonal rabbit anti-calnexin Abs (Cell Signaling Technology, Danvers, MA) and secondary HRP-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch) were used. Immunoblots were developed using enhanced ECL-applying West Pico (Thermo Fisher Scientific) or HRP-Juice PLUS (PJK, Kleinblittersdorf, Germany). For control, membranes were stripped with ReBlot Plus Mild Antibody Stripping Solution (Merck Millipore, Billerica, MA) and reprobed with HRP-conjugated anti–β-actin (clone AC-15; Sigma-Aldrich) or anti-GFP (clone B34; BioLegend) followed by detection with HRP-conjugated goat anti-mouse IgG (Jackson ImmunoResearch). Signal intensities were quantified with FusionCapt Advance software (Vilber Lourmat, Marne-la-Vallée, France) after visualization of chemiluminescence with a Fusion SL imaging system (Vilber Lourmat Sté).

U937 cells were washed twice with PBS and cultured in RPMI 1640 minimal medium without methionine or cysteine (Sigma-Aldrich) for 30 min at 37°C prior to labeling with 0.5 mCi/ml EasyTag EXPRESS [³⁵S] Protein Labeling Mix (PerkinElmer, Waltham, MA) containing [³⁵S]-l-methionine and [³⁵S]-l-cysteine. After incubation for 1 h (pulse), cells were washed thoroughly with PBS and resuspended for the indicated times in complete RPMI 1640 medium (chase). Cells were lysed, and supernatants were precleared four times by addition of Pierce Protein A/G UltraLink Resin (Thermo Fisher Scientific) loaded with isotype control Ab. AICL glycoproteins were immunoprecipitated by incubation with protein A/G beads linked to anti-AICL mAb 7F12 (25) for 3 h at 4°C. After extensive washing with radioimmunoprecipitation assay buffer containing 0.3 M NaCl, samples were boiled at 95°C and subjected to SDS-PAGE. Gels were dried and exposed to FujiFilm Super RX film for signal detection.

Peptide arrays (peptides of 18 aa with an offset by 1 aa) covering the entire AICL ED (Lys²⁶ to His¹⁴⁹, Fig. 1A) were synthesized by Fmoc chemistry at activated polyethylene glycol spacers on cellulose membranes by automated parallel peptide synthesis on a MultiPep RSi (INTAVIS Bioanalytical Instruments, Cologne, Germany) and used for binding experiments as previously described (34, 35). After saturation of unspecific binding sites, peptides were probed with AICL-specific mAb 7F12 or 7G4. Bound Abs were detected with a mouse Ig–specific, HRP-conjugated Ab (Jackson ImmunoResearch Laboratories) and visualized by ECL. Epitopes of 7F12 and 7G4 as defined by peptide binding (Fig. 1) were verified by mutational studies of AICL glycoproteins (data not shown).

The 293 cells stably expressing FH-tagged AICL or KACL glycoproteins were treated with 1.2% (v/v) paraformaldehyde (PFA) in PBS for 10 min at room temperature. Cross-linking was quenched by washing with 1.25 M glycine. The 293 cells were lysed with ice-cold lysis buffer, and FH-tagged glycoproteins were immunoprecipitated from lysates using anti-Flag mAb M2–loaded magnetic beads (Sigma-Aldrich) for 3 h at 4°C. After extensive washing with TBS, glycoproteins were deglycosylated with PNGase F, and samples were heated for 10 min at 65°C and subjected to SDS-PAGE on a gradient gel (Mini-PROTEAN TGX 4–20%; Bio-Rad Laboratories) under reducing conditions. After staining with InstantBlue (Expedeon, Cambridge, U.K.), lanes were cut into pieces prior to in-gel digestion of proteins performed as described previously (36). In summary, gel pieces were washed, destained, and dehydrated. Proteins were reduced with 10 mM DTT, alkylated with 55 mM iodoacetamide, and digested overnight with sequencing-grade trypsin (Promega, Madison, WI). Peptides were extracted using an increasing concentration of acetonitrile and were subsequently concentrated and desalted using the Stop and Go Extraction technique (37). Peptides were separated and eluted from a C18 column via a linear gradient of an increasing acetonitrile concentration on an EASY-nLC 1200 (Thermo Fisher Scientific) coupled via a nanoelectrospray ionization source to the quadrupole-based Q Exactive HF benchtop mass spectrometer (38). Mass spectrometry (MS) spectra were acquired using 3E6 as an automatic gain control target with a maximal injection time of 20 ms and a 60,000 resolution at 300 m/z. The mass spectrometer operated in a data-dependent, top 15 mode with subsequent acquisition of higher-energy collisional dissociation fragmentation MS/MS spectra of the top 15 most intense peaks. Resolution for MS/MS spectra was set to 30,000 at 200 m/z, automatic gain control target was set to 1E5, maximal injection time was set to 64 ms, and the isolation window was set to 1.6 Th. All acquired raw files were processed using MaxQuant (1.5.3.30) (39) and the implemented Andromeda search engine (40). For protein assignment, electrospray ionization-tandem MS fragmentation spectra were correlated with the UniProt human database (version 2017), including a list of common contaminants. Relative label-free quantification of proteins was done using the MaxLFQ algorithm integrated into MaxQuant (41). Perseus software (1.5.3.0) (42) was used for further analysis. Data were filtered for contaminants, reverse entries, and proteins that were only identified by site.

Statistical significance was calculated by paired t test or two-way ANOVA, followed by Bonferroni posttest, as detailed in the respective figure legends using the GraphPad Prism 7 software (GraphPad Software, La Jolla, CA).

AICL is a type II transmembrane glycoprotein consisting of a very short amino-terminal cytoplasmic part, a transmembrane domain, a short extracellular stalk region, and a single C-terminal CTLD (Fig. 1A). The latter contains all six highly conserved cysteines typically found in CTLD of NKC-encoded CTLR and is known to stabilize the compact structure of the CTLD via three intramolecular disulfide bonds (15, 18). Additional intermolecular disulfide bonds by cysteines located in the stalk region result in disulfide-linked homodimerization of many NKC-encoded CTLR (15, 17). We had previously demonstrated that AICL also forms disulfide-linked homodimers (21, 25). However, AICL does not contain cysteines in the stalk region in contrast to other disulfide-linked CTLR (Fig. 1A). However, inspection of the sequence of the AICL CTLD revealed an additional cysteine (Cys⁸⁷) not found in other related CTLR. Hence, in absence of a crystal structure of AICL, we hypothesized that Cys⁸⁷ may link AICL homodimers through disulfide bonding. Indeed, accurate positioning of Cys⁸⁷ into the available crystal structures of the closely related CTLR CD69 (43, 44) and LLT1 (45) (Fig. 1B) strongly suggested that Cys⁸⁷ residues are located at the C terminus of the α2 helix of AICL, where the adjacent polypeptide chains of the AICL monomers are located in close proximity to each other and thus may readily enable the formation of intermolecular disulfide bonds between the AICL monomers (Fig. 1B). To directly test the presumption that AICL dimerization is mediated by disulfide linkage via Cys⁸⁷, we substituted cysteine at position 87 of AICL for serine (AICL-C87S). Immunoblot analyses of 293 cells transfected with AICL-C87S revealed that such AICL mutants are monomeric under both reducing and nonreducing conditions, whereas wild-type (wt) AICL forms stable dimers under nonreducing conditions (Fig. 1C). These data demonstrate that disulfide linkage of AICL homodimers is mediated exclusively through Cys⁸⁷ at the membrane-distal surface of the AICL CTLD. A longer exposure of the AICL immunoblots additionally revealed a lower abundance of higher molecular mass glycoisoforms among AICL-C87S mutants as compared with AICL-wt (Fig. 1D). Higher molecular mass glycoisoforms are usually due to complex glycosylation processes in the Golgi complex and are indicative of glycoproteins that had passed the Golgi complex and reached the cell surface. To further corroborate such a maturation defect of nondisulfide-linked AICL-C87S glycoproteins, AICL-transfected 293 lysates were treated with Endo H, which selectively deglycosylates only such glycoproteins that have not yet passed the Golgi complex, whereas cell surface glycoproteins are resistant to Endo H digestion (46). A substantially higher portion of AICL-wt glycoproteins, as compared with AICL-C87S glycoproteins, was resistant to Endo H digestion (Fig. 1E), indicating an impaired maturation of AICL-C87S molecules. The impaired cell surface expression of nondisulfide-linked AICL-C87S molecules was further confirmed by flow cytometry, showing that surface expression of AICL-C87S as compared with AICL-wt is reduced by ∼50% (Fig. 1F–H). Reduced surface expression was evident using the two AICL-specific mAb 7F12 and 7G4, which bind two different linear epitopes of the AICL CTLD (Fig. 1I) as determined by epitope mapping using arrays of overlapping peptides. AICL mAb 7F12 binds to a linear epitope formed by a β strand at the outer rim of the AICL CTLD distantly located from position 87 (Fig. 1I) and is unlikely to be affected by the C87S mutation. Unexpectedly, when addressing NKp80 binding by AICL-C87S molecules, we did not detect any binding of NKp80 tetramers to the cell surface of AICL-C87S transfectants (Fig. 1J). Altogether, these data demonstrate that AICL homodimers are linked in a unique manner by a disulfide bridge at the membrane-distal surface of the CTLD and that the absence of such a disulfide linkage results in an impaired AICL maturation, cell surface expression, and a fully abrogated NKp80 binding.

A previous study demonstrated that Kaposi sarcoma–associated herpesvirus (KSHV)–encoded ubiquitin E3 ligase activity targets AICL glycoproteins for lysosomal degradation via their cytoplasmic lysine residues (47). We wondered whether these cytoplasmic lysine residues (Fig. 1A) might also regulate turnover of AICL glycoproteins independently of KSHV infection. To this aim, 293 cells ectopically expressing either AICL-wt or mutant AICL-K6A_K7A molecules, in which Lys⁶ and Lys⁷ had been substituted by alanines, were cultured in the presence of the inhibitors CCMA or MG-132, respectively. Whereas CCMA blocks acidification of lysosomes and consequently inhibits lysosomal protein degradation, MG-132 inhibits protein degradation by the proteasome. Flow cytometry analysis revealed a substantial and significant increase in AICL-wt levels at the cell surface (∼3-fold) upon treatment with MG-132, whereas surface levels of AICL-K6A_K7A molecules apparently remained unchanged (Fig. 2A–C). In contrast, treatment with CCMA had no significant effect on AICL cell surface expression levels. Similarly, immunoblotting of cellular lysates of 293 transfectants treated with MG-132 revealed markedly increased cellular levels of AICL-wt molecules (∼3-fold) but not of AICL-K6A_K7A molecules (Fig. 2D). Treatment with CCMA led only to slightly increased levels of both AICL-wt and AICL-K6A_K7A molecules (∼1.4-fold). Hence, cytoplasmic lysine residues facilitate tuning of cellular levels of AICL glycoproteins via proteasomal degradation in contrast to the lysosomal degradation promoted by the KSHV-encoded K5 ubiquitin E3 ligase activity (47).

Immunoblotting had demonstrated that cellular AICL molecules occur as several glycoisoforms of different molecular masses irrespective of the cellular context [i.e., in primary NK cells as well as in immortalized cell lines (e.g., NK-92; U937) as well as in transfected cell lines (e.g., P815-AICL, 293-AICL) (21, 25, 27)]. Usually, immunoblotting reveals two characteristic Endo H–susceptible AICL glycoisoforms of distinct molecular masses (∼30 and ∼33 kDa), hereafter called “Endo H–susceptible doublet,” as well as diverse Endo H–resistant glycoisoforms of higher molecular mass (range ∼45–55 kDa) (Fig. 3A). The abundance of the latter correlates well with the levels of AICL surface expression as determined by flow cytometry (Fig. 1D, 1F) (27). To address the nature and the sequence of biosynthesis of the AICL glycoisoforms, we investigated AICL maturation by metabolic-labeling studies. U937 cells that endogenously express large amounts of AICL glycoproteins were metabolically labeled, and AICL glycoproteins precipitated from lysates were analyzed by autoradiography. Several AICL species in the range from 22 to 33 kDa were detected that were absent in the control precipitation (Fig. 3B). These included the Endo H–susceptible doublet also detectable in the AICL immunoblot of U937 (Fig. 3A) but also several less-prominent glycoisoforms of lower molecular mass that likely correspond to early endoplasmic reticulum (ER)–resident precursors with various modifications of the N-linked carbohydrate branch (48, 49). In addition, there were several Endo H–resistant AICL glycoisoforms of higher molecular mass (>45 kDa), which accordingly, had already passed the Golgi complex (Fig. 3B). Subsequently, cellular maturation of AICL glycoproteins was analyzed in pulse–chase experiments with metabolically labeled U937 cells. Above-mentioned early ER-resident AICL precursors became readily undetectable after a few hours of chase (Fig. 3C). In contrast, the Endo H–susceptible doublet successively became more prominent with the lower molecular mass species appearing at first and the higher molecular mass species with delay of several hours (Fig. 3C). Endo H–resistant higher molecular mass AICL species were not clearly attributable to a specific band in the immunoblot due to an increased nonspecific background. Next, cellular maturation of AICL was assessed by treating U937 cells with the translation inhibitor CHX (Fig. 3D). The lower molecular mass species of the Endo H–susceptible doublet faded shortly after start of CHX treatment and almost completely disappeared 24–48 h later, whereas the higher molecular mass species remained apparently unaltered, similar to the high molecular Endo H–resistant glycoisoforms. Accordingly, Endo H treatment of these lysates revealed a decline of the Endo H–sensitive AICL molecules within 24 h of CHX treatment, whereas the amounts of Endo H–resistant AICL glycoproteins remained rather constant (Fig. 3D). Altogether, these data show that the two major Endo H–susceptible AICL glycoisoforms (∼30 and ∼33 kDa) clearly differ with regard to synthesis and stability. Although the synthesis of the ∼33 kDa species is delayed as compared with the ∼30 kDa species, the latter is rather short-lived as compared with the ∼33 kDa species, which exhibits an unusually long dwell time.

For unknown reasons, a major portion of AICL glycoproteins, unlike the closely related KACL glycoproteins, is intracellularly retained in the ER and/or the Golgi complex of cells both endogenously or ectopically expressing AICL (27). Our data indicate that achievement of full N-linked glycosylation is a rate-limiting step in the cellular maturation of AICL glycoproteins (Fig. 3). The CTLD of AICL contains three putative N-glycosylation sites (Asn⁵⁷, Asn⁶², and Asn¹⁰⁰; consensus sequence NxS/T) and, in addition, one so-called atypical N-glycosylation site at position 44 (Asn⁴⁴; consensus sequence NxC) (Figs. 1I, 4A). This atypical N-glycosylation site is located at the base of the CTLD (Fig. 1I) and includes Cys⁴⁶ that was predicted to form an intramolecular disulfide bond with Cys³⁵ (Fig. 1A). Such a rare atypical N-glycosylation is also present in the closely related CD69 molecules (50). To test whether these four putative N-glycosylation sites affect the cellular maturation of AICL molecules, we generated a series of AICL mutants in which one or several of these asparagines were replaced by aspartate (Fig. 4B).

The N-glycosylation of these mutants was analyzed by immunoblotting of cell lysates of the respective 293 transfectants using anti-AICL mAb 7G4 (Fig. 4C). In 293 transfectants, the Endo H–susceptible doublet of AICL-wt glycoproteins (∼30 kDa; ∼33 kDa) was less distinct than in U937 cells endogenously expressing AICL (Fig. 3A), presumably because of ectopic overexpression (Fig. 4C). As expected, the mutant D1/2/3/4, which lacks all four putative N-glycosylation sites, was detected with an apparent molecular mass of ∼17 kDa, corresponding to the calculated molecular mass of the unglycosylated AICL polypeptide (17.3 kDa) and to the observed molecular mass of deglycosylated AICL molecules (Figs. 1E, 3A). AICL mutant D1/2/3 with an intact N-glycosylation site at Asn¹⁰⁰ homogenously showed the expected increased molecular mass of ∼20 kDa, demonstrating that Asn¹⁰⁰ is a functional and complete N-glycosylation site. However, additional presence of the atypical N-glycosylation site Asn⁴⁴ in AICL mutants D2/3/4 (as compared with D1/2/3/4) and D2/3 (as compared with D1/2/3), respectively, resulted in the appearance of an additional minor fraction of AICL glycoproteins with an extra N-glycosylation, indicating that N-glycosylation at Asn⁴⁴ is partial and inefficient (Fig. 4C). Accordingly, when only the atypical N-glycosylation site at Asn⁴⁴ was altered (mutant D1), the upper molecular mass AICL species (∼33 kDa) of the Endo H–susceptible doublet disappeared and only the lower molecular mass species (∼30 kDa) remained. These data strongly suggest that partial glycosylation at Asn⁴⁴ gives rise to the two major Endo H–susceptible AICL glycoisoforms detected as Endo H–susceptible doublet in many cells and cell lines endogenously expressing AICL (21, 25, 27). Differential glycosylation at Asn⁴⁴ also likely accounts for the predominant AICL doublets detected in lysates of mutants D2/4 and D2, D3, and D4. However, the mutant D1/4 also gave rise to an AICL doublet, indicating that mutation of Asn¹⁰⁰ impairs glycosylation at Asn⁵⁷ and/or Asn⁶². Comparative analysis of mutants D2/4 and D3/4 revealed additional occurrence of nonglycosylated AICL proteins (∼17 kDa) in D3/4 lysates, but not in D2/4 lysates, demonstrating that the glycosylation at Asn⁵⁷, but not at Asn⁶², is impaired when glycosylation at Asn¹⁰⁰ is missing.

Next, we investigated the impact of AICL N-glycosylation on intracellular retention and surface expression of AICL molecules. Lysates of the various mutants (Fig. 4B) were treated with Endo H prior to immunoblotting to address the cellular distribution of these mutants by assessing the fraction of Endo H–resistant AICL species (>18 kDa; mostly at cell surface) versus the fraction of Endo H–susceptible deglycosylated AICL species (∼17 kDa; intracellular) (Fig. 4D). These data were then assessed together with the detection of cell surface AICL expression on the same 293 transfectants by flow cytometry (Fig. 4E–J). Mutation of the atypical glycosylation site Asn⁴⁴ increased AICL surface expression in the context of almost all mutants. A most drastic increase in AICL cell surface expression was observed when Asn⁴⁴ was mutated on the background of the D2/3 mutant (mutant D1/2/3 versus D2/3) (Fig. 4D, 4G, 4J). Although mutant D2/3 was almost completely intracellularly retained and hardly detectable at the cell surface, additional mutation of Asn⁴⁴ (mutant D1/2/3) almost fully restored AICL surface expression to AICL-wt levels (Fig. 4D, 4G, 4J). There was also a marked increase in AICL surface expression when mutant D1 was compared with AICL-wt (Fig. 4E, 4J) and Asn⁴⁴ was mutated on the background of the mutant D4 (mutant D1/4 versus D4) (Fig. 4F), although the latter was not significantly different (Fig. 4J). In contrast, mutating any of the three conventional glycosylation sites (mutants D2, D3, and D4) reduced AICL cell surface expression (Fig. 4D–F). Mutation of two conventional glycosylation sites in parallel further reduced surface expression of AICL (Fig. 4I), and mutation of all three conventional glycosylation sites (D2/3/4) fully abrogated surface expression of AICL (Fig. 4D, 4H). These data demonstrate that glycosylation of at least one of these conventional N-glycosylation sites is essentially required for AICL maturation and cell surface export and that all three conventional N-glycosylation sites contribute to AICL maturation in an additive manner. Of note, mutation of Asn⁴⁴ did not improve AICL maturation when all conventional glycosylation sites were mutated (mutant D1/2/3/4 versus D2/3/4) (Fig. 4D, 4G, 4H), further corroborating that glycosylation of at least one conventional N-glycosylation site is essentially required for AICL maturation. In addition, the impact of the individual conventional glycosylation sites on AICL cell surface expression appears to be different, with Asn⁶² having the strongest impact, followed by Asn⁵⁷ and Asn¹⁰⁰ having the least impact (Fig. 4D–F, 4I). When analyzing AICL mutants for binding of NKp80 tetramers, mutation of the glycosylation site at Asn⁶², but not at other glycosylation sites, appeared to slightly impair NKp80 binding in a direct manner beyond the reduction of NKp80 binding because of the diminished AICL surface expression (Fig. 4K).

In summary, these data demonstrate that N-glycosylation of AICL proteins is a prerequisite for cell surface expression. Individual glycosylation at the three conventional N-glycosylation sites Asn⁵⁷, Asn⁶³, and Asn¹⁰⁰ cumulatively promotes AICL surface expression, with glycosylation at Asn¹⁰⁰ also improving the efficacy of glycosylation at Asn⁵⁷. In contrast, presence of the atypical glycosylation site Asn⁴⁴ impairs cellular maturation and surface expression of AICL glycoproteins.

To identify AICL-interacting proteins involved in the intracellular retention and maturation of AICL glycoproteins, MS of AICL coimmunoprecipitates was performed. To this aim, 293 cells stably expressing FH-tagged AICL glycoproteins were treated with PFA to cross-link AICL-binding proteins. For control, 293 cells expressing FH-tagged KACL, which is not intracellular retained, and mock-transfected 293 cells were processed in parallel. Following PFA treatment, cellular lysates were incubated with anti–Flag-tag mAb-loaded beads, and resulting immunoprecipitates deglycosylated prior to gel electrophoresis and subsequent MS. Evaluation of MS data revealed a significant enrichment of calnexin in AICL immunoprecipitates as compared with immunoprecipitates of KACL- or mock-transfected cells. Calnexin is an ER-resident transmembrane protein that acts as a chaperone for folding and maturation of glycoproteins (51). Preferential association of calnexin with AICL was verified by coimmunoprecipitation of calnexin with AICL, but not with KACL, from lysates of PFA–cross-linked 293 transfectants (Fig. 5A) despite significantly stronger ectopic expression of KACL as compared with AICL. To directly assess the functional impact of calnexin on AICL maturation, the expression pattern of AICL by the calnexin-deficient cell line NKR (32) was compared with the parental calnexin-sufficient CD4⁺ T cell line CEM (31). CEM and NKR cells were transduced to stably express either FH-tagged AICL or KACL glycoproteins, and cellular lysates were analyzed by immunoblotting after treatment with Endo H or PNGase F (Fig. 5B). In agreement with previous data, the vast majority of AICL glycoproteins was Endo H–susceptible and thus intracellularly retained in both cell lines, whereas the vast majority of KACL glycoproteins exhibited Endo H resistance indicative of cell surface expression and the lack of retention. No difference in abundance of KACL molecules was observed for CEM versus NKR cells, whereas abundance of AICL molecules was significantly reduced in NKR cells as compared with CEM cells (Fig. 5C). A similar reduction in AICL expression by NKR cells as compared with CEM cells was also observed by flow cytometry of both intact and permeabilized cells (Fig. 5D, 5E), whereas transcript levels of AICL or KACL in CEM and NKR cells were comparable (data not shown). Accordingly, also binding of NKp80 tetramers to AICL-transduced NKR cells was reduced (data not shown).

Altogether, these data demonstrate that AICL glycoproteins, but not KACL molecules, preferentially interact with calnexin, and that absence of calnexin markedly impairs AICL expression.

Activation of myeloid cells by TLR ligands or exposure of NK cells to proinflammatory cytokines induces AICL surface expression, enabling recognition of such activated innate immune cells by mature NK cells via NKp80 (25–27). Engagement of NKp80 by AICL promotes cytokine release and cytotoxicity of NK cells, thereby potentially modulating nascent immune responses (25, 27). Of note, resting NK cells contain intracellular stores of AICL glycoproteins that appear to be specifically retained in the Golgi complex (27). In this study, we investigated the regulation of cellular expression and maturation of AICL molecules by posttranslational modifications and aimed to address mechanisms underlying intracellular retention of AICL.

Like many other NKC-encoded CTLR, functional AICL is a disulfide-linked homodimer (21, 52–54). We demonstrate in this study that AICL homodimers are linked by an intermolecular disulfide bond through juxtaposed cysteines at position 87, located at the membrane-distal surface of the AICL CTLD. Such positioning of the intermolecular disulfide bond is peculiar among NKC-encoded CTLR because disulfide linkage of these dimeric CTLR commonly occurs through cysteines of the membrane-proximal stalk region as suggested for the AICL relatives LLT1 (45, 55), CD69 (44), and BACL (56) but also for many other NKC-encoded CTLR (16). In contrast, the fourth human CLEC2 family member KACL is a nondisulfide-linked homodimer (21) lacking cysteines for intermolecular disulfide bonding. Apart from promoting homodimerization, our results also demonstrate that Cys⁸⁷ affects maturation and cell surface expression of AICL glycoproteins because AICL-C87S mutants were impaired in that regard. Hence, we conclude that loss of the disulfide bond at position 87 not only disfavors AICL homodimerization but also negatively impacts the stability and/or surfacing of AICL glycoproteins. Possibly, loss of this disulfide bond also alters the conformation of the AICL CTLD, thereby interfering with cellular quality control mechanisms supervising export of AICL glycoproteins to the cell surface. This idea is further supported by our finding that AICL-C87S mutants cannot be engaged by NKp80 at the cell surface.

The rather short cytoplasmic stretch of AICL glycoproteins is basic with three lysines and one histidine and may be involved in regulating AICL expression. In fact, a previous study has shown that these lysine residues promote lysosomal degradation of AICL glycoproteins by a KSHV-encoded ubiquitin E3 ligase, supporting the idea that AICL induction is involved in the cellular immunosurveillance of KSHV and possibly other viruses (47). We wondered whether cytoplasmic Lys⁶ and Lys⁷ regulate cellular AICL expression also in the absence of viral infection. Our results with AICL mutants show that cytoplasmic Lys⁶ and Lys⁷ render AICL molecules susceptible for degradation by the proteasome but not for lysosomal degradation. Hence, proteasomal degradation appears to represent another regulatory mechanism contributing to a strictly regulated AICL expression and opens the possibility to tune AICL expression via differentially expressed E3 ubiquitin ligases. Regulation of protein expression levels upon NK cell stimulation by altered expression of ubiquitin E3 ligases has previously been shown (57).

We extensively characterized the contribution of N-linked glycosylation to the maturation of AICL glycoproteins based on the observation that AICL gives rise to two predominant Endo H–susceptible glycoisoforms that are intracellularly retained through an unknown mechanism. In addition to three conventional N-glycosylation sites (NxS/T), the AICL CTLD also contains an atypical N-glycosylation site (NKC) at the membrane-proximal surface, which is only partially and inefficiently glycosylated, and thus gives rise to the Endo H–susceptible doublet observed in immunoblotting. A similar glycosylation pattern was described for CD69, which also harbors an atypical N-glycosylation site in addition to a conventional N-glycosylation site and which also intracellularly accumulates (50). The inefficient glycosylation at these rare atypical N-glycosylation sites may depend on reduced kinetics of the glycosylation machinery in case of a cysteine-comprising glycosylation site (58). According to the hydrogen bond acceptor theory, the glycosylation at such an atypical site can only occur before the cysteine engages in disulfide bonding. In addition, the sulfhydryl group of the cysteine leads to a reduced glycosylation efficiency compared with the hydroxyl groups in serine or threonine present in conventional N-glycosylation sites (58). Therefore, the atypical glycosylation site is often nonglycosylated in case of high translation rates which limit the time for glycosylation prior to the formation of disulfide bonds (59).

Accordingly, we observed in pulse–chase experiments that the ∼30 kDa AICL species of the Endo H–susceptible doublet appears immediately, whereas generation of the ∼33 kDa AICL species, which requires glycosylation at the atypical N-glycosylation site, is time delayed. However, the ∼30 kDa AICL species appeared much less stable than the ∼33 kDa AICL species upon CHX treatment, indicating that it is further processed either by glycosylation or by degradation. As the ∼33 kDa AICL Endo H–susceptible species are unusually stable over time as revealed in CHX-treated U937 cells, but also in activated PBMC (27), we hypothesize that these fully glycosylated AICL molecules are specifically retained in the ER or in the Golgi complex. Apart from human CLEC2 family members CD69 and AICL, atypical N-glycosylation sites are also present in mouse CLEC2 family members, including CD69 and most Clr molecules. Of note, evidence for intracellular retention has also been reported for some of these molecules such as CD69 and Clr-a that are predominantly localized intracellularly governed by unknown mechanisms (29, 35, 60). The extracellular domain of AICL has been shown to mediate intracellular retention (27), whereas the stalk and parts of the cytoplasmic domain were shown to contribute to the impaired cell surface expression of Clr-a (35).

Analysis of AICL glycomutants also demonstrated that conventional N-glycosylation of AICL is essentially required to enable surface expression and that the extent of surface expression correlates with the number of functional conventional N-glycosylation sites. In contrast, CD69 surface expression was not markedly affected by either treatment with a glycosylation inhibitor or mutation of the conventional, the atypical, or both N-glycosylation sites, demonstrating that N-glycosylation is not required for the maturation and surfacing of CD69 molecules (61, 62). The recently described brain-associated non-NKC–encoded CLEC2 family member BACL is efficiently expressed at the cell surface in absence of glycosylation (56). Thus, N-glycosylation is not a common prerequisite for efficient surface expression of CLEC2 family members such as those reported in this study for AICL. Intriguingly, presence of the atypical glycosylation site rather impaired the surface expression of AICL molecules contrary to the conventional glycosylation sites. Furthermore, our studies with the glycosylation mutants indicate an important role of glycosylation at Asn¹⁰⁰ for an efficient glycosylation of AICL at position Asn⁵⁷ and an impact of the glycosylation at Asn⁶² on the NKp80 binding site. In summary, N-glycosylation of the AICL CTLD critically regulates AICL maturation and surfacing. N-glycosylation at conventional sites (Asn⁵⁷, Asn⁶², and Asn¹⁰⁰) promotes AICL maturation and surface expression in an additive manner, whereas presence of the atypical N-glycosylation site (Asn⁴⁴) tends to negatively impact AICL surfacing.

A MS screen for proteins interacting with AICL identified the ER-resident chaperone calnexin. Calnexin binds to glycoproteins shortly after their translation into the ER and supports their correct folding or oligomerization and prevents aggregation and premature transport into the Golgi complex (63–65). A sustained interaction of calnexin with AICL molecules was corroborated by coimmunoprecipitation and functionally addressed using the calnexin-deficient cell line NKR. As in all previously tested cell lines, we detected a strong intracellular retention of AICL glycoproteins in both CEM and NKR cells, whereas the AICL-relative KACL was expressed at the cell surface. However, the expression of AICL was strongly reduced in NKR cells, indicating that proper folding, glycosylation, and possibly dimerization of newly synthesized AICL glycoproteins is dependent on the chaperoning function of calnexin. In contrast, KACL expression was not affected by the absence of calnexin as it was already shown for several other proteins in calnexin-deficient cells (51, 66, 67). Notably, an interaction with the calnexin-relative calreticulin, another ER-resident chaperone involved in glycoprotein folding, has been reported for CD69 (68).

Taken together, in this study, we report several posttranslational molecular mechanisms governing maturation and cell surface expression of AICL glycoproteins either positively or negatively. Our data also further support the hypothesis that AICL molecules are intracellularly retained by interacting with yet unidentified molecules via their CTLD. Further investigations aiming at the identification of such putative AICL interactors and at the elucidation of the mechanisms underlying intracellular retention of AICL are needed to obtain a better understanding of the immunological function of AICL and its receptor NKp80.

We thank Dr. J. P. Jacobs and Dr. P. Cresswell for the cell lines CEM and CEM-NKR and are particularly grateful to Praveen Mathoor for assistance in cell sorting.

A.S. and the Eberhard Karls University of Tübingen have a patent application entitled “Methods for blocking the interaction between nkp80 and its ligand aicl” filed through the Eberhard Karls University of Tübingen. The other authors have no financial conflicts of interest.