The murine CD94/NKG2 ligand, Qa-1b, is a high-affinity, functional ligand for the CD8αα homodimer

The immune co-receptor CD8 molecule (CD8) has two subunits, CD8α and CD8β, which can assemble into homo or heterodimers. Nonclassical (class-Ib) major histocompatibility complex (MHC) molecules (MHC-Ibs) have recently been identified as ligands for the CD8αα homodimer. This was demonstrated by the observation that histocompatibility 2, Q region locus 10 (H2-Q10) is a high-affinity ligand for CD8αα which also binds the MHC-Ib molecule H2-TL. This suggests that MHC-Ib proteins may be an extended source of CD8αα ligands. Expression of H2-T3/TL and H2-Q10 is restricted to the small intestine and liver, respectively, yet CD8αα-containing lymphocytes are present more broadly. Therefore, here we sought to determine whether murine CD8αα binds only to tissue-specific MHC-Ib molecules or also to ubiquitously expressed MHC-Ib molecules. Using recombinant proteins and surface plasmon resonance–based binding assays, we show that the MHC-Ib family furnishes multiple binding partners for murine CD8αα, including H2-T22 and the CD94/NKG2-A/B-activating NK receptor (NKG2) ligand Qa-1b. We also demonstrate a hierarchy among MHC-Ib proteins with respect to CD8αα binding, in which Qa-1b > H2-Q10 > TL. Finally, we provide evidence that Qa-1b is a functional ligand for CD8αα, distinguishing it from its human homologue MHC class I antigen E (HLA-E). These findings provide additional clues as to how CD8αα-expressing cells are controlled in different tissues. They also highlight an unexpected immunological divergence of Qa-1b/HLA-E function, indicating the need for more robust studies of murine MHC-Ib proteins as models for human disease.

MHC-Ib are less polymorphic than class Ia and can demonstrate cellular and tissue specificity (13,14). This allows them to serve more diverse and specialized functions than their class Ia counterparts, whose primary role is the presentation of intracellular peptides to CD8 T cells. For example, H2-T3/TL is restricted to epithelial cells of the intestine (15,16), H2-Q10 is a soluble MHC (17,18) that is overexpressed in the liver (19), and H2-T13 is only expressed in the blastocyst and placenta (20). In contrast, H2-M3 is widely expressed (13) but is restricted to the presentation of N-formylated peptides, making it a recognition element for CD8 T cells directed at intracellular bacteria (21). Similarly, Qa-1 b is ubiquitously expressed (13) and acts to present the leader sequence from class Ia MHC to subsets of natural killer (NK) and T cells expressing CD94/NKG2 heterodimers (22). In this regard, Qa-1 b appears to be the sole homologue of a human MHC-Ib, HLA-E, with which it shares its function (22,23). As such, studies using Qa-1 b have been heavily utilized as preclinical models for HLA-E responses to a variety of infections and cancers (24, 25).
The observation that H2-T3/TL and H2-Q10 demonstrate tissue-restricted expression whereas CD8␣␣ expressing cells are distributed throughout nonlymphoid organs suggested that other MHC-Ib could act as ligands for CD8␣␣. Our results support this hypothesis and demonstrate that the interaction between MHC-Ib and CD8␣␣ occurs within a hierarchical framework. Surface plasmon resonance demonstrated that Qa-1 b had the highest affinity for CD8␣␣ and was also capable of activating CD8␣␣ expressing T cells. These interactions indicate that CD8␣␣ ligands are more widespread than previously anticipated. Importantly, they also demonstrate a func-tional divergence between Qa-1 b and HLA-E and highlight the need for more robust studies of murine MHC-Ib as models for human disease.

Qa-1 b binds CD8␣␣ with a higher affinity than H2-T3/TL and H2-Q10
We have previously demonstrated that the interaction between H2-Q10 and CD8␣␣ is of a higher affinity (ϳ300 nM) than that between H2-T3/TL and CD8␣␣ (ϳ800 nM) (12). Given these observations, we sought to determine the affinity between Qa-1 b and CD8␣␣ as a means to determine a binding hierarchy among H2-T3/TL, H2-Q10, and Qa-1 b . When using Qa-1 b as the ligand (Fig. 2a), the K D calculated was ϳ300 nM (averaged over several independent experiments, Fig. 2c), whereas using Qa-1 b in the analyte phase (Fig. 2b), the K D calculated was ϳ200 nM (averaged over several independent The open histogram is the unstained control whereas the light-shaded histogram is CD8 staining on RMA-s parental cells. The filled histogram is CD8 staining on RMA-s-CD8␣␣ cells. Results are representative of at least three independent experiments. All histograms have been offset to stack vertically above one another and scaled to maximum count for clarity. b, staining of RMA-s-CD8␣␣ cells with class Ia and MHC-Ib tetramers demonstrates that H2-T22, TL, H2-Q10, and Qa-1 b bind CD8␣␣. The open histograms are unstained controls, the gray histograms are the indicated tetramers on RMA-s cells and the filled histograms are tetramer staining on RMA-s-CD8␣␣ cells. Results are representative of at least three independent experiments. All histograms have been offset to stack vertically above one another and scaled to maximum count for clarity. c, median fluorescent intensity (MFI) of H2-T3/TL, H2-Q10 and Qa-1 b staining on RMA-s-CD8␣␣ cells. The MFI was pooled from five independent experiments using equivalent tetramer concentrations and identical laser voltages. *, p ϭ 0.0476 and **, p ϭ 0.0079.

Qa-1b as a ligand for CD8␣␣
experiments, Fig. 2c). Therefore, the SPR in the two different orientations yielded broadly consistent results and reinforced our tetramer data. Collectively, the data from Figs. 1 and 2 demonstrate that MHC-Ib represents an extended family of ligands for CD8␣␣ and that a hierarchy exists among these molecules with Qa-1 b currently at the apex.

HLA-E and Qa-1 b differ in their ability to bind to CD8␣␣
Finally, we sought to determine whether the interaction between Qa-1 b and CD8␣␣ was shared with their human counterparts. To do this, we generated Jurkat cells expressing homodimers of human CD8␣␣ and stained them with tetramers of HLA-E, Qa-1 b , and HLA-G (Fig. 4). In line with a previous study (26), we observed minimal binding of HLA-E and Qa-1 b to human CD8␣␣. In contrast, we did observe binding of the known human CD8␣␣ ligand HLA-G (Fig. 4a). Examination of the reverse interaction indicated that HLA-E and HLA-G had minimal binding to mouse CD8␣␣ (Fig. 4b). These data dem-onstrate that the interaction between Qa-1 b and CD8␣␣ is highly specific and suggest a significant divergence between Qa-1 b and HLA-E with respect to CD8␣␣ binding.

Discussion
Our data provide new evidence that the mouse MHC-Ib provides multiple ligands for CD8␣␣, expanding on previous observations that identified the prototypical ligand, H2-TL (10,27,28) and H2-Q10 (24). The expression pattern of H2-TL is restricted to epithelial cells of the small intestine (15,16), where it plays a central role in the activation of CD8␣␣ ϩ cells (29), whereas H2-Q10 is restricted to the liver (19), where it regulates the development of CD8␣␣ ϩ ␥␦T cells. However, subsets of CD8␣␣ ϩ cells exist in the lung and kidney (5-10), but none of these organs expresses H2-TL. Our observation that H2-T22 and Qa-1 b are ligands for CD8␣␣ now provides a rationale for the presence of CD8␣␣ expressing cells outside of the small intestine and liver.
The signals controlling the development and activation of CD8␣␣ expressing cells are not completely understood. These cells can be found in mice lacking all class I (30) as well as those lacking just class Ia MHC or H2-TL (29 -32). This suggests that the role of MHC-Ib, and in particular Qa-1 b , is more likely to be associated with controlling the responses of CD8␣␣ expressing cells. Interestingly, the immune systems of Qa-1 b -deficient mice develop normally but they do exhibit defects in secondary immune responses (33). The answer to this may lie within the capacity of activated CD8 T cells to up-regulate the CD8␣␣ homodimer (10). Indeed, the expression of CD8␣␣ promotes the survival and differentiation of memory T cell precursors (10). Although an up-regulation of H2-TL is observed on activated antigen-presenting cells (10), our new data suggest that the capacity of H2-TL to bind CD8␣␣ on activated CD8 T cells would be outcompeted by Qa-1 b . Qa-1 b can also be up-regulated on antigen-presenting cells by the actions of TLR ligands and the presence of IFN-␥ (34, 35), making it a likely target for activated T cells expressing CD8␣␣. Intriguingly, activated dendritic cells can stimulate the activation of Qa-1 b restricted TCR␥␦ cells expressing CD8␣␣ (36). These cells can then target and eliminate self-reactive CD4 T cell clones and attenuate the severity of experimental autoimmune encephalomyelitis (35). Although our experiments did not completely discount the coexpression of additional receptors that might also engage Qa-1 b on CD8␣␣ϩ TCR␥␦ϩ cells, our data convincingly showed that Qa-1 b is a bona fide ligand for CD8␣␣ itself. Hence, it will be interesting to determine whether CD8␣␣ can act as a co-stimulator or co-repressor of the activation of Qa-1 b restricted CD8␣␣ϩ TCR␥␦ cells. Given that CD8␣␣ does not act as a co-receptor for CD8-dependent TCRs (37,38) and that H2-TL is thought to act as a co-repressor (3), it is likely that the presence of CD8␣␣ on these Qa-1 b -restricted T cells is limiting their functionality. In addition, it will be important to determine whether binding can occur between Qa-1 b and the CD8␣␤ heterodimer. It would be expected that if Qa-1 b can interact with CD8␣␤, it would be at a significantly lower affinity than to CD8␣␣, to prevent CD8␣␤ϩ T cells from becoming activated by the presence of Qa-1 b .

Qa-1b as a ligand for CD8␣␣
Complicating the study of Qa-1 b and CD8␣␣ in vivo are the observations that Qa-1 b also binds CD94/NKG2 complexes (22,39) and the TCR (37). Importantly, HLA-E also binds CD94/NKG2 (23) and the TCR (40) but not CD8␣␣ (Fig. 4). What our data suggest is that Qa-1 b will preferentially bind CD8␣␣ over CD94/NKG2A as the affinity between Qa-1 b and CD8␣␣ is two orders of magnitude higher than that between Qa-1 b and CD94/NKG2A (ϳ17 M for CD94/NKG2A) (41). The affinity of Qa-1 b -restricted TCRs is yet to be determined but recognition of HLA-E by CMV-specific TCRs ranges from 3 to 37 M (38, 40). Should these affinities be reflective of those between murine TCRs and Qa-1 b , this suggests that a TCRindependent interaction with CD8␣␣ will predominate. However, further exploration of the responses controlled by Qa-1 b

Qa-1b as a ligand for CD8␣␣
will require a better understanding of the biochemical partners involved in binding. The observation that CD8␣␣ binds to the ␣3 region of the MHC (28), whereas CD94/NKG2 and the TCR bind to the ␣1-2 regions (41, 42) provides a basis for future studies in this area.
The generation of mice in which the Qa-1 b CD8␣␣ interaction is abolished will be an important approach given the observation that Qa-1 b is a high-affinity ligand for CD8␣␣, whereas HLA-E is not. This is especially relevant given the interest in ␥␦T cells during cytomegalovirus (CMV) infection. Recent evidence has demonstrated that ␥␦T cells confer protection against murine cytomegalovirus infection (MCMV) (43) and the expansion of ␥␦T cell subsets is also associated with the resolution of human cytomegalovirus infection (HCMV) (44). On the flip side, HCMV infection is known to cause a specific expansion of NK cells expressing NKG2C, the ligand for HLA-E (45), whereas this is not seen during MCMV infection (46). Similarly, a large proportion of murine ␥␦T cells express CD8␣␣ (6), although this population is not as pronounced in humans (47, 48). Thus, there is the potential for Qa-1 b to play a major role in the ␥␦T cell response to MCMV, whereas its homologue during HCMV infection involves an NK cell response, mediated by the interaction between HLA-E and NKG2C. Collectively, this indicates that there are alternate strategies for immunity involving analogous molecules in different species. Given the interest in Qa-1 b /HLA-E in models of transplantation (49) and cancer (25) our results provide the basis for the generation of preclinical mouse models that are a more faithful representation of the human condition.
Given that CD8␣␣ recognition occurs within a hierarchy, further investigation into other members of the mouse MHC-Ib is required. We currently understand the basic biochemistry and immunology of only 6 of the 30 MHC-Ib, hence the capacity for future research into the function of this family is warranted. Indeed, it remains possible that many other members of the MHC-Ib family bind CD8␣␣, and even that another MHC-Ib may eclipse Qa-1 b as the apex partner.

Experimental procedures
Mice C57BL/6 mice were from Alfred Medical Research and Education Precinct (AMREP) Animal services. All mice were used at between 6 and 8 weeks of age. All experiments were in accordance with the animal ethics guidelines of the National Health and Medical Research Council of Australia and were approved by the AMREP Animal Ethics Committee.

Cell culture
RMA-s and Jurkat cells were sourced from the Peter MacCallum Cancer Centre Tumor cell bank. Cells were cultured in RPMI supplemented with 10% FCS, L-glutamine, penicillin and streptomycin. 293T and Phoenix E cells were cultured in DMEM supplemented with 10% FCS, L-glutamine, penicillin and streptomycin.

Cloning and expression of recombinant CD8
A codon-optimized gene encoding the full-length mouse CD8␣ (NM_001081110.2) was purchased from GenScript (Pis-cataway, NJ) and ligated into MSCV vectors. Orientation and correct sequence was confirmed by DNA sequencing using T7 forward and reverse primers.

Retroviral transduction
Retroviral transduction was performed on RMA-s and Jurkat cells; retrovirus-containing supernatant was produced by transfecting packaging cells with murine stem cell virusinternal ribosome entry site-GFP (MSCV-I-GFP) or MSCV-I-mCherry using standard calcium phosphate transfection methods. Viral supernatant was used to transduce RMA-s and Jurkat cells on retronectin (TaKaRa Bio, Shiga, Japan) precoated plates (BD Biosciences). After 5-7 days, GFP only (CD8␣␣) events were subjected to two rounds of cell sorting (FACSAria, BD Biosciences) to produce stable cell lines. Expression of CD8␣␣ was confirmed by flow cytometry.

Generation of recombinant CD8␣␣
The sequence of soluble mouse CD8␣␣, allele CD8A*02, was designed based on a previously published sequence to fold soluble mouse CD8␣/CD8␤ following Escherichia coli expression (50), GenBank accession number GQ247790.1, except for the following modifications whereby residue numbering is based on Kern et al. (50). (i) Cys-36 was mutated to Ser as based on crystal structures of CD8␣␣, Cys-36 is not involved in disulfide bonds (50,51), and we suspected that mutating this residue might improve folding efficiency; (ii) the gene was shorted to cover residues Gly-5 to Lys-128, as in previously determined crystal structures that included mouse CD8␣␣ or CD8␣␤ there were no data beyond this region; (iii) the gene was codon-optimized for E. coli expression by GenScript (Piscataway, NJ). The gene was purchased from GenScript (Piscataway, NJ) and subcloned into a pET30 expression vector, expressed in BL21 E. coli competent cells and purified from inclusion bodies. The inclusion bodies were solubilized in 6 M guanidine-HCl and 120 mg/liter of total protein (split over three injections) was rapidly diluted in a buffer containing 0.4 M arginine hydrocholoride, 100 mM Tris-HCl (pH 8), 2 mM EDTA, 3 mM reduced GSH, and 0.3 mM oxidized GSH and allowed to sit for 2-3 days at 4°C. The refold was then dialyzed against 25 mM HEPES (pH 7.4) overnight, followed by filtration through a 0.45 M filter. The CD8␣␣ was purified by cation exchange using an SP-Sepharose column (Amersham Biosciences) and eluted using 25 mM HEPES containing a gradient of NaCl. The major peak was pooled, concentrated, and further purified by size exclusion chromatography using Superdex 75 column with a buffer containing 25 mM HEPES (pH 7.4) with 150 mM NaCl.
Activation and analysis of sorted ␥␦T cells-Sorted cells were cultured in RPMI supplemented with 10% FCS, L-glutamine, penicillin, and streptomycin. 10 5 CD8␣␣ ϩ or CD8␣␣ Ϫ subsets in 100 l were added to U-bottom 96-well plates in which 5 g/ml Qa-1 b or HLA-E had been crosslinked overnight at 4°C. Supernatants were harvested at 6 and 18 h post stimulation and frozen at Ϫ20°C. The CBA Flex System Kit (BD Biosciences) was then used to measure IFN-␥ production according to the manufacturer's protocols.
Flow cytometric analysis of RMA-s and Jurkat-Cells were cultured in tissue culture flasks for 2 days prior to removal with Tryple (Invitrogen). Cells were washed two times in PBS and nonspecific binding blocked with 2.4G2. Cells were then stained with mAb to CD8␣ (mouse 53-6.72, human SK1; Bio-Legend, San Diego, CA), and tetramers prior to fixation in 2% paraformaldehyde. Cells were then washed twice in PBS, resuspended in FACS buffer and acquired on an LSR-II or X-20 Fortessa flow cytometer (BD Biosciences). Doublets were excluded using a FSC-A versus FSC-H acquisition profile followed a FSC-A versus SSC-A profile to exclude debris. At least 10,000 gated events were collected for analysis.

Surface plasmon resonance
Surface plasmon resonance (SPR) was performed essentially as described (52) in two different orientations, one where Qa-1 b was on the chip (ligand) and the other where Qa-1 b was in solution (analyte). For Qa-1 b in the ligand phase, biotinylated MHC was captured on the surface of a ProteOn NLC neutravidin chip (Bio-Rad, ϳ150 RU of each). An empty flow cell with neutravidin alone served as a control. Serially diluted CD8␣␣, produced as described above, was injected simultaneously over the control and test surfaces. To verify results, SPR was also performed in the reverse orientation. In these experiments, an antibody to CD8␣␣ (53-6.72) was amine coupled to two flow cells of a ProteOn GLC chip (ϳ500 RU) and recombinant CD8␣␣ was injected over the chip at 30 l/min, and ϳ500 RU was captured by the antibody. The running and sample buffer was 10 mM HEPES (pH 7.4), 150 mM NaCl, and 0.05% Tween-20. The other flow cell containing antibody alone served as a control cell. Qa-1 b was then injected over both flow cells. After subtraction of data from the control flow cells, K D was calculated by kinetic analysis using the ProteOn Manager software (Bio-Rad). At least two independent SPR experiments were performed in both orientations.

Statistical analysis
The nonparametric, two-tailed Mann-Whitney U test was used to determine the statistical significance of data sets; p values of less than 0.05 were considered significant.