Results and discussion

To initiate our study, we transduced a variety of cell types (EL4, 293, Daudi, P815) with N-terminal flag-tagged versions of Mill1 and Mill2. For Mill2, alternative splicing generates long (Mill2-L) and short (Mill2-S) isoforms that differ in size by 15 amino acids near the N terminus.³ Unless otherwise noted, Mill2 will refer to both isoforms in this report.

Kajikawa et al.¹⁸ recently reported that Mill family members associate with 2M, including the finding of an apparent absolute requirement of 2M for cell-surface expression of Mill1.Both findings were observed during the initial phase of our study, thus confirming these results (data not shown).

The specific properties of the association between Mill2-L and Mill2-S, and 2M remain unknown. Thus, a cell system initially devoid of 2M was designed to explore this issue. We chose Daudi cells as our reporter cell because they do not express 2M. Transduction efficiency in Daudi cells using lentivirus is poor (<5%). Therefore, we sorted Daudi cells transduced with an ecotropic receptor-eGFP fusion protein (DE cells), which are very sensitive to transduction with ecotropic retrovirus. Next, we generated a 2M-positive variant of DE cells (DEM cells) using a retrovirus encoding mouse 2M. DE or DEM cells were then transduced with Mill1 or Mill2. One week after transduction of DEM cells, approximately 50% of each population expressed cell-surface Mill1 or Mill2. As expected, DE cells supported expression of both Mill2 isoforms but not of Mill1 (Supplementary Figure S1). However, Mill1 was expressed intracellularly in DE cells (data not shown). We were surprised that neither Mill2-L nor Mill2-S required 2M for surface expression and speculated that bovine 2M in our in vitro culture system might be stabilizing Mill2. Accordingly, we positively sorted DE and DEM cells expressing Mill2-L or Mill2-S and removed cell-surface Mill with trypsin. These cells were then cultured for 72 h in serum-free medium and the recovery of Mill2 was assessed. Nearly 100% of Mill2-S transductants recovered full cell-surface expression irrespective of 2M expression. In contrast, only 20% of DE cells and 44% of DEM cells recovered cell-surface expression of Mill2-L, and the fluorescent intensity was more than 10-fold less compared to Mill2-S (Figure 1a). Thus, cell-surface expression of Mill2-L is partially dependent on 2M and a serum factor, whereas excision of exon 2 from the Mill2 transcript renders the surface expression of Mill2S completely independent of both 2M and a serum factor. Mill1 demonstrated similar dependence on a serum factor as Mill2-L but also required the expression of 2M (data not shown). Importantly, endogenous classical MHC-I reached normal levels irrespective of the transduced gene as long as 2M was co-expressed (data not shown). The presence of 1% fetal bovine serum or 1% mouse serum during the last 24 h of culture resulted in nearly full recovery of Mill2-L cell-surface expression, even when the serum was derived from 2M-deficient mice (Figure 1b). TAPI, a broad-spectrum metalloprotease (MMP) inhibitor, could not mimic the effects of serum when added to this system (data not shown). This suggested that a serum factor(s) other than 2M did not block cell-surface shedding of Mill2-L, but rather promoted the recovery or stabilized Mill2-L at the cell surface. The involvement of a serum factor in the cell-surface expression of Mill1 and Mill2-L was the first sign that these molecules play a role in nutrient metabolism.

Figure 1.

Mill1 and Mill2 are MHC-I homologues that more closely resemble HFE than MICA. (a) DE cells or DEM cells expressing Mill2L or Mill2S (as indicated) were treated with trypsin for 10 min at 37 °C to remove surface expression of the Flag-tagged Mill proteins. Cells were subsequently cultured for 72 h in SFM and stained with anti-Flag PE. (b) DE-Mill2L cells were trypsinized and cultured for 72 h without serum or 48 h without serum, followed by 24 h with 1% fetal bovine serum, 1% B6 serum or 1% B6 2M^-/- serum (as indicated). DE cells are shown in black. (c) EL4 cells stably expressing mouse IL4-R (control), Flag-tagged Mill1, Flag-tagged Mill2-L, Flag-tagged Mill2-S or ULBP1 were stained (as indicated) with mouse NKG2D-LZ and an anti-leucine zipper mAb. (d) B6 splenocytes were depleted of CD4+, CD8+ and IgM+ cells and cultured in IL-15 for 3 days. The resulting population, which was greater than 95% NK1.1+/TCR-, was stained (as indicated) with p3-Fc (negative control), Mill1-Fc, Mill2L-Fc, Mill2S-Fc or ULBP2-Fc (positive control). (e–i) Ribbon diagrams of predicted three-dimensional structure of Mill2 (e) and crystal structures of HFE (f) and MICA (g). 2M, when present, is coloured blue. (h–j) Space-filling models of grooves generated between the 1 and 2 domains for H2Kd (h), Mill1 (i) and Mill2S (j). Surface electrostatic charge is highlighted such that blue is positive and red is negative. HFE, hereditary haemochromatosis protein; mAb, monoclonal antibody; MHC-I, major histocompatibility complex class I-related proteins; SFM, serum-free medium; 2M, 2-microglobulin.

Full figure and legend (216K)

On the basis of the association of the Mill family of proteins with 2M, its possible interaction with a serum factor and the low amino-acid sequence homology with MICA (30%), we hypothesized that these molecules were unlikely candidates as ligands for NKG2D. Thus, we stained EL4 cells, retrovirally transduced with Mill1 or Mill2, with a soluble mouse NKG2D-leucine zipper fusion protein (mNKG2D-LZ). EL4 cells were chosen because they do not express NKG2D ligands endogenously.²² mNKG2D-LZ bound with EL4-ULBP1 (positive control²²), but not with EL4-Mill1 or EL4-Mill2 (Figure 1c). Soluble Mill1-hFc and Mill2-hFc fusion proteins bound neither mouse NK cells (Figure 1d) nor human NK cell-like YT cells transduced with mNKG2D (data not shown). Cytotoxicity assays using interleukin (IL)-2 or IL-15 activated C57BL/6 NK effector cells, and Mill-transduced EL4 target cells resulted in the same level of killing as that observed with EL4 cells transduced with IL-4R (control). As previously reported, EL4-ULBP1 were more than threefold more sensitive to lysis than the control (data not shown).²² Our data indicate that Mill family members are not ligands for NKG2D nor other NK cell receptors on mature NK cells. We do not exclude the possibility that they are involved in other immunological processes.

To identify an MHC-I family member that more closely resembled the characteristics ascribed to Mill1 and Mill2, the tertiary structures of Mill1, Mill2-L and Mill2-S were predicted using computer-based protein homology using HFE (PDB ID 1A6Z) as a template.²³ A complete model for Mill2-L could not be generated because the 15 N-terminal amino acids found in Mill2-L but not Mill2-S fell outside the alignment to the structure template. These 15 amino acids, nevertheless, are of obvious interest because in their absence, expression of Mill2-S is completely independent of 2M or a serum factor.

Our model revealed that Mill1 and Mill2-S contain 33 and 31 predicted contact sites with 2M, respectively. Alignment of Mill1 and Mill2 demonstrated that 20 2M contact sites were identical and 27 conserved in position. Alignment of Mill1 and Mill2 with H2-K^d established that although only nine 2M-binding residues were conserved; 26 of the 33 contact sites between H2-K^d and 2M were located in the same position with those found for Mill1 or Mill2 (Supplementary Figure S2). Thus, our model predicts that orientation rather than conservation of residues serves to explain the unexpected association between Mill and 2M. Although our models did predict greater surface contact between Mill1 and 2M versus Mill2 and 2M (40 Å²), they could not immediately explain the absolute dependence of Mill1 on 2M. Solving the crystal structure for these complexes will ultimately resolve this issue.

The MHC family is very closely related structurally. Figures 1e–g illustrate the similarity between the predicted structure for Mill2 and the crystal structures of HFE (1A6Z)²³ and MICA (1HYR).²⁴ In a previous report, MICA was predicted to be the closest relative of the Mill family.¹⁷ Using protein-threading analysis (in house method) against 6692 known structures (ASTRAL SCOP^{25, 26, 27, 28}), both Mill1 and Mill2 were most highly related to HFE. ZAG, an MHC class I homologue involved in lipid metabolism,^{15, 16} was fifth closest to both Mill1 and Mill2, whereas MICA was the 149th and the 13th thirteenth closest to Mill1 and Mill2, respectively. Analysis using other threading engines (3D-PSSM)²⁹ generated slightly different hierarchies but reiterated the conclusion that Mill family members most closely resemble HFE.

Hereditary haemochromatosis protein is an MHC-IB family member that associates with 2M^{14, 15, 16, 23, 30} and the transferrin receptor (CD71). The dimeric complex formed between HFE and CD71 binds to and internalizes transferrin. Binding of HFE to iron is pH-dependent and release occurs in the acidic environment of early endosomes.^{14, 31, 32} We therefore examined the electrostatics of the groove formed between 1 and 2 of Mill1 and Mill2. Unlike H2-K, both Mill1 and Mill2 were heavily charged. Mill1 was acidic, whereas Mill2 was basic (Figures 1h–j). Mill1 demonstrated similar dependence on a serum factor as Mill2-L but also required the expression of 2M (data not shown). The generation of soluble Fc-fusion proteins of Mill1 and Mill2 allowed us to search for cellular counter structures and to test the effect of pH on any observed binding. Supporting the specificity of binding with these fusion proteins, versions of Mill1 and Mill2 fused to mouse IgG1 Fc inhibited staining when added in excess (data not shown). Similar binding patterns were observed for both molecules, although the magnitude of binding was always greater with Mill2-Fc. A series of mouse cells and cell lines was stained with Mill1-Fc and Mill2-Fc. EL4 (thymoma) and 70Z/3 (pre-B lymphoma) were highly positive whereas P815 (mastocytoma) and WEHI (plasmacytoma) were negative. RMA (T-cell lymphoma) and RAW-264 (macrophage-like) were marginally positive (data not shown). In mouse bone marrow (BM), 10–15% and 40–50% of cells bound to Mill1-Fc and Mill2-Fc, respectively (Figure 2a). We identified the positive populations in the BM as B cells and erythrocyte precursors. In the B-cell lineage, staining with Mill-Fc decreased as the cells differentiated from early (CD43^-B220⁺IgM+IgD^-) to late pre-B cells (CD43^-B220⁺IgM/D^-) and finally to immature B cells (Figures 2b–c; Mill2-Fc shown); CD43⁺ pro-B cells did not stain (data not shown). An analogous pattern was observed in the erythroid lineage. Developing erythrocytes can be ordered from least (Figure 2d, gate 1) to most differentiated (Figure 2d, gate 4) via loss of CD71 expression on the Ter119⁺ population (Figure 2d). Figure 2e shows that staining with Mill2-Fc decreased as the erythrocyte population matured. Thus, for both lineages, soluble Mill proteins bound preferentially to less-differentiated/highly proliferative cells with greater nutrient demands. Similarly, for splenic T and B cells, we found that mitogen-activated and proliferating, but not resting, cells bound Mill2-Fc (Figures 2f–g; resting cells not shown). In this case, very little staining was observed with Mill1-Fc. For Flt3-L-mobilized plasmacytoid dendritic cells (PDC), we found that stimulation of PDC with CpG induced expression of a counter structure for Mill2-Fc. In contrast, stimulation of granulocyte-macrophage colony-stimulating factor-mobilized myeloid dendritic cells (MDC) resulted in a complete loss of Mill2-Fc staining (Figure 2g). This opposite result obtained with stimulated MDCs may be because, unlike T cells, B cells and PDC that proliferate upon stimulation, MDC do not. The highly charged grooves between 1 and 2 of Mill1 and Mill2 suggested that pH might change Mill-Fc staining patterns. Interestingly, a small decrease in pH from 7 to 6.8 resulted in a significant increase in fluorescence intensity for all cells that bound Mill-Fc (data not shown) including BM (Figure 2i). Our model predicted that a drop in pH from 7 to 6.8 would have no effect on groove charge for Mill1 or Mill2. This suggests that the pH-dependent increase in binding to Mill-Fc could be a function of the putative ligand or a minor structural change in the Mills. Thus, Mill1 and Mill2 bind to a ligand on proliferating cells in a pH-dependent manner, a finding similar to the binding properties of HFE for transferrin (that is, pH-dependent) and the expression pattern of CD71 on proliferating cells (that is, upregulated). Moreover, there was nearly a complete overlap of staining with an anti-CD71 monoclonal antibody (mAb) and Mill2-Fc on BM cells (data not shown). These data support the hypothesis that like HFE, Mill proteins may also be involved in nutrient metabolism.

Figure 2.

Characterization of cellular and soluble ligand(s) for Mill1 and Mill2. (a) BM cells from B6 mice were stained with either ULBP3-Fc (control), Mill1-Fc or Mill2-Fc (as indicated). (b) B6 BM gated on live (FCS/SSC) B220⁺ cells and subfractioned into (1) early pre-B cells, (2) late pre-B cells and (3) immature B cells. Staining of fractions 1–3 from (b) with Mill2-Fc (filled histograms) is depicted in (c). (d) B6 BM subfractioned into four subsets based on CD71 and Ter119 as follows: (1) pro-erythroblasts, (2) basophilic erythroblasts, (3) chromatophilic erythroblasts and (4) ortho-chromatophilic erythroblasts and erythrocytes. Staining of fractions 1–4 from (d) with Mill2-Fc (filled histograms) is shown in (e). (f) B6 splenocytes were activated with anti-CD3 for 48 h, gated on CD4⁺ and CD8⁺ cells (as indicated) and assessed for labelling with Mill1-Fc and Mill2-Fc (as marked). In the upper panel of (g), B6 splenocytes were activated for 72 h with LPS, gated on CD19⁺ cells and assessed for labelling with Mill2-Fc. In the bottom four panels, BM cells were cultured in GM-CSF (MDC) or Flt-3L (PDC) for 7 or 9 days, respectively and stained with Mill2-Fc prior to or after activation for 24 h (as indicated). For MDC, cells were gated as CD11b⁺/CD11c⁺. For PDC, cells were gated as CD11c⁺/CD11b^-. (h) B6 BM was stained with Mill2-Fc in the presence of the indicated molecules. Staining with ULBP3-Fc (control) is shown as the black histogram. (i) B6 BM was stained with Mill1-Fc or Mill2-Fc at pH 7 or pH 6.8 (as marked). For (c), (e–g) and (i), staining with ULBP3-Fc (control) is shown as open histograms. BM, bone marrow; GM-CSF, granulocyte-macrophage colony-stimulating factor; MDC, myeloid dendritic cells; PDC, plasmacytoid dendritic cells.

Full figure and legend (277K)

Repeated attempts to identify a cell surface ligand for Mill1 and Mill2 via immunoprecipitation and expression cloning failed. On the basis of our finding that a serum factor is involved in Mill expression, we hypothesized that a low molecular weight soluble serum factor (too small to be identified via gel electrophoresis) may bind non-covalently to proliferating cells. We therefore utilized an indirect approach to identify potential soluble ligands for Mill1 and Mill2 by screening an array of small charged molecules for their ability to block staining of BM with Mill1-Fc and Mill2-Fc. We observed that fetal bovine serum (Figure 2h) and MS (data not shown) completely blocked Mill-Fc staining (Mill2-Fc, shown). Our protein modelling showed that although the groove between 1 and 2 was too small to accommodate TAP-derived peptides,¹⁸ the charged grooves of Mill1 and Mill2 could accommodate small peptides of two or three amino acids, vitamins, sugars and nucleosides. The results of our screen indicated that the polyamines, spermidine and spermine (but not putrescine), highly charged oligomers of arginine, could completely inhibit binding of Mill-Fc to BM cells (Figure 2h; Mill2-Fc, shown). It was surprising that spermidine and spermine may bind both Mill family members as Mill1 and Mill2 possess oppositely charged grooves. On the basis of our models, however, these polyamines could bind in alternate modes to Mill1 and Mill2. Surface electrostatic calculations on these ligands while present in the binding groove indicate that the ligands possess a surface charge distribution that could allow their binding to these differentially charged grooves. Because highly charged molecules may associate non-specifically to Mill, we can only speculate that small charged molecules like polyamines resemble the true ligands. These likely bind within the grooves of Mill1 and Mill2, using different binding modes.

Using RT-PCR Kasahara³ reported that normal expression of Mill1 is scarce (mainly skin, thymus, ovary and kidney) whereas that of Mill2 is widespread. A similar pattern was observed using Taqman qRT-PCR (data not shown). Transcripts for Mill1 and Mill2 were nearly absent in whole fetus and embryonic stem cells, suggesting that their expression is limited to postnatal life. To examine Mill1 and Mill2 at the protein level, we generated mAbs specific for Mill1 (M575), Mill2-L (M247) and Mill2-S (M248; binds both Mill2-L and Mill2-S), which were used in flow cytometry and immunohistochemistry experiments. Immunohistochemical analysis of a panel of frozen tissue sections from BALB/c mice did not detect Mill1 protein expression on any of the tissues (data not shown). As detection of low-density surface antigens is often difficult using IHC but possible using flow cytometry, cell suspensions of thymocytes, 10–15% of which are dividing cells, were probed for expression of Mill1. Cycling but not resting thymocytes expressed low but appreciable levels of Mill1 (Figures 3a–b). Most resting thymocytes, however, did express intracellular Mill1 (Figure 3c), implying that Mill1 is transported to the cell surface when thymocytes undergo cell division. Within the proliferating thymocyte pool, Mill1 was only expressed on developing CD4⁺CD8⁺ and CD4^-CD8^- thymocytes (data not shown). Consistent with the expression of Mill1 on developing thymocytes, we also observed low but appreciable expression of Mill1 on EL4 thymoma cells (data not shown). The fibroblast cell line, 3T3, was identified as positive for Mill1 (Supplementary Figure S3c). For Mill2, every tumour cell line tested expressed low but significant levels of the protein. The highest levels were observed on 3T3 and the squamous cell carcinoma, SCCVII (data not shown). Standard IHC staining of normal BALB/c tissue detected Mill2 expression in the skin, bladder, uterus and colonic muscularis (Figures 3e–h; isotype controls are shown in Supplementary Figure S4). In these tissues, staining was confined to SMC and fibroblasts. Flow cytometric analysis of primary SMC from carotid artery explants confirmed our results for Mill2 while producing negative results for Mill1 and an isotype control (Supplementary Figure S4a). The expression of Mill2 on stromal cells serves to explain its widespread expression via RT-PCR and qRT-PCR. Interestingly, both 3T3 and SMC bound strongly to Mill1-Fc and Mill2-Fc, suggesting a possible role for autocrine-like activity in these cell types (Supplementary Figure S3b and d).

Figure 3.

Mill1 and Mill2 expression on normal and inflamed tissues (a–d). Balb/c thymocytes were gated on resting and dividing populations (based on SSC/FSC) and stained for cell surface (a and b) and intracellular (c and d) Mill1 using M575. Staining was visualized with an anti-rat PE antibody. (e–h) Immunohistochemical staining of a panel of cryopreserved tissues (as shown) from BALB/c mice stained with M248. The arrow in (g) indicates a fibroblast. (i and j) Cryopreserved sections of the colonic muscularis from non-colitic mdr1a^-/- mice (i) or mildly colitic mdr1a^-/- mice (j) were stained with M248.

Full figure and legend (229K)

Several aspects of our data suggested a role for Mill2 in nutrient metabolism during times of increased energy demands, such as cellular activation and proliferation, which occur during tissue injury and repair. To examine whether enhanced energy demands of SMC undergoing hyperplasia correlate with expression of Mill2, we stained cryopreserved colitic colon sections from two models of inflammatory bowel disease. Mdr1a^-/- mice suffer from chronic inflammation and hyperplasia of the Mill2⁺ muscularis externa,³³ whereas treatment of mice with dextran sulphate sodium results in acute destruction of Mill2^- epithelial cells and no change to the muscularis.³⁴ As anticipated, Mill2 was upregulated only in the hyperplastic muscularis of colitic mdr1a^-/- mice (Figures 3i–j) but not in dextran sulphate sodium-treated mice (data not shown).

To test whether the expression of Mill was merely a correlate to increased energy demands or was of functional consequence, we asked what effect mAbs specific for Mill1 and Mill2 would have on another form of tissue injury, dermal wound healing. This model was chosen because the skin co-expressed Mill1 and Mill2, allowing us to test the antibodies against both molecules in a well-established assay. A skin organ culture system was chosen to diminish the number of variables potentially affecting our results such as infiltrating lymphocytes. Excised wounds were allowed to heal on surgical gelfoam for 7 days with or without the addition of antibodies. Wound closure was calculated as described in Materials and Methods. We found that compared to skin organ culture performed in media alone or with the addition of an isotype control mAb, the addition of M247 (binds Mill2-L) resulted in wounds about 50% smaller on day 7 (Figures 4a–k; P<0.0001, Student's t-test). Addition of M575 (binds Mill1) and, to a greater extent, M248 (binds both Mill2-L and Mill2-S), appeared to augment wound closure to some degree but the values were not significant (P>0.01). Importantly, heat denaturation of the M247 abolished its ability to augment wound closure (Figure 4k). It is possible that the difference between the effects of M247 and M248 are related to epitope location as (as assessed by flow cytometry) SMC and 3T3 stained more weakly with M248 than with M247. Affinity differences were not formally assessed although both antibodies behaved similarly using transduced EL4 cells and flow cytometry. To address whether M247 augmented wound healing by promoting keratinocyte proliferation, we pulsed these organ cultures with BrdU to enumerate proliferating cells. Treatment with M247 did not increase the number of BrdU+ cells (data not shown). Mill2 expression during wound closure as assessed by IHC did not appear to change, but the fragility of the tissue after organ culture may have decreased the sensitivity of staining (data not shown). Nevertheless, we did observe specific staining for Mill1 in the hair follicles of wounded skin, an observation also made by Kajikawa et al.¹⁸ in the skin of 3-day-old mice but not older animals¹⁸ (Figure 4l).

Figure 4.

An antibody specific for Mill2-L augments dermal wound healing in SOC. (a–j) Two-millimeter wounds were introduced into dorsal skin sections of BALB/c mice and then cultured on surgical sponges for up to 7 days containing media alone, an isotype control mAb, M575, M247, M248 or heat-denatured M247 (as indicated). Antibody concentration was 10 g ml^-1 and media was changed on day 4. This experiment was repeated four times and the combined results are illustrated in (k). The y axis represents the percentage of the wound still open on day 7 compared to day 0. Each dot represents one piece of wounded skin. (l) Cryosections of wounded skin 5 days after wounding were stained with M575. mAb, monoclonal antibody; SOC, skin organ culture.

Full figure and legend (182K)

Although we do not yet understand the mechanism of action of M247, it is interesting to note that other than transferrin, the only known iron transport vehicles are the same polyamines that blocked Mill-Fc staining of BM, that is, spermine and spermidine but not putrescine.³⁵ Like HFE, the Mill family may also be involved in iron metabolism. It is worth noting that haemochromatosis is exacerbated in 2M^-/- mice compared to HFE^-/- mice, suggesting that a 2M-binding molecule other than HFE is involved in iron transport.^{36, 37} M247 augmented wound healing without inducing the proliferation of cells surrounding the wound. As Mill2 is expressed on contractile SMC, M247 may have functioned by inducing the skin to 'pinch in'. M247 may act directly by sending a contractile signal to the cells or indirectly by inducing internalization of Mill2L and associated iron. In support of this explanation, we found that crosslinking Mill2 on EL4 cells resulted in altered motility/contractility as measured by enhanced adherence and spreading of these to plastic (without the induction of a calcium flux; data not shown).

In conclusion, we have demonstrated that the Mill family of MHC-I homologues resemble the structure and function of HFE, rather than the previously presumed relationship with MICA. We provided evidence for this similarity not only by demonstrating that neither of the Mill isoforms binds to the receptor for MICA (NKG2D), but also by presenting data indicating that Mill1 and Mill2 are linked to nutrient metabolism and can influence wound healing. Our data provide a template for future experimental investigation.

Methods

Expression of Mill1 and Mill2

Mill1 and Mill2 were cloned from BALB/c uterus cDNA (Clontech, Mountain View, CA, USA) into pDONR222 (Invitrogen, Carlsbad, CA, USA). In some cases, the leader sequence was replaced with an IgK leader-Flag-M1 cassette. Mill1 and Mill2 were then recombined into either the transient expression vector pDC415G or the lentivector pLV410G (human EF1 promoter) or LZRS-G retroviral vector via LR reaction (Invitrogen). Lentiviral and retroviral packaging with the VSV.G or ecotropic envelope protein and transduction of EL4 or P815 cells with Mill1 or Mill2 were performed as previously described.^{38, 39} Daudi cells overexpressing the ecotropic receptor MCAT-1 fused to GFP (DE cells) were generated by transducing Daudi cells with pLV410-MCAT-1-eGFP (VSV.G) and sorting the eGFP+ cells (1–2%). These cells were then transduced with LZRS encoding mouse-2M (ecotropic), generating DE cells that express 2M and HLA-A/B/C (DEM cells).

Flow cytometry and Fc-fusion proteins

Fc fusion proteins of Mill1, Mill2-L and Mill2-S were generated as previously described.⁹ All primary and secondary antibodies for flow cytometry were purchased from Pharmingen (San Jose, CA, USA) and Jackson Labs (Bar Harbor, ME, USA), respectively, except PE-anti-Flag M1 (M1 conjugated with PE (Prozyme, San Leandro, CA, USA)).

Western blotting and immunoprecipitation

Western blotting and immunoprecipitation were performed using digitonin lysis buffer as previously described.³⁸

Computer-based protein modelling

The structures of H2-Kd, Mill1 and Mill2 were calculated using computer-based protein homology modelling based on the haemochromatosis protein HFE²³ (Protein Data Bank entry 1A6Z^{23, 40}). The protein structure models were built using the Homology Modelling method within the Molecular Operating Environment software package (Chemical Computing Group, Montreal, Canada). To verify the Mill fold family, threading analysis was performed on a template library built using protein-level representative structures using ASTRAL SCOP version 1.69^{25, 26, 27, 28} resulting in a fold library of 6692 structures. Threading is a method in which a protein sequence is matched onto a structure from a library of known protein structures and the resulting fold of the query sequence scored for suitability to the template fold.⁴¹

Monoclonal antibody generation and immunohistochemistry

Lewis rats were immunized with either Mill1-Fc or Mill2L-Fc protein as described.⁹ Hybridoma supernatants were screened via flow cytometry and immunohistochemistry on fibrin cell pellets using EL4 cells expressing Mill1 or Mill2. IHC studies were performed using standard techniques.

Skin organ culture and wound healing

Skin organ culture was performed as previously described⁴² using BALB/c mice between 8 and 10 weeks of age. Wound size was measured using Metavue Software (Metavue Corp., Madison, WI) on days 0, 4 and 7 of culture. Antibodies were added to the culture at 10 g ml^-1. Media was changed on day 4.

References

1. Wong P, Pamer EG. CD8T cell responses to infectious pathogens. Annu Rev Immunol 2003; 21: 29–70. | Article | PubMed | ISI | ChemPort |
2. Natarajan K, Li H, Mariuzza RA, Margulies DH. MHC class I molecules, structure and function. Rev Immunogenet 1999; 1: 32–46. | PubMed | ChemPort |
3. Kasahara M. The chromosomal duplication model of the major histocompatibility complex. Immunol Rev 1999; 167: 17–32. | Article | PubMed | ISI | ChemPort |
4. Gleimer M, Parham P. Stress management: MHC class I and class I-like molecules as reporters of cellular stress. Immunity 2003; 19: 469–477. | Article | PubMed | ISI | ChemPort |
5. Beckman EM, Porcelli SA, Morita CT, Behar SM, Furlong ST, Brenner MB. Recognition of a lipid antigen by CD1-restricted alpha beta+ T cells. Nature 1994; 372: 691–694. | Article | PubMed | ChemPort |
6. Braud VM, Allan DS, McMichael AJ. Functions of nonclassical MHC and non-MHC-encoded class I molecules. Curr Opin Immunol 1999; 11: 100–108. | Article | PubMed | ISI | ChemPort |
7. Cerwenka A, Bakker AB, McClanahan T, Wagner J, Wu J, Phillips JH et al. Retinoic acid early inducible genes define a ligand family for the activating NKG2D receptor in mice. Immunity 2000; 12: 721–727. | Article | PubMed | ISI | ChemPort |
8. Carayannopoulos LN, Naidenko OV, Fremont DH, Yokoyama WM. Cutting edge: murine UL16-binding protein-like transcript 1: a newly described transcript encoding a high-affinity ligand for murine NKG2D. J Immunol 2002; 169: 4079–4083. | PubMed | ISI | ChemPort |
9. Cosman D, Mullberg J, Sutherland CL, Chin W, Armitage R, Fanslow W et al. ULBPs, novel MHC class I-related molecules, bind to CMV glycoprotein UL16 and stimulate NK cytotoxicity through the NKG2D receptor. Immunity 2001; 14: 123–133. | Article | PubMed | ISI | ChemPort |
10. Bauer S, Groh V, Wu J, Steinle A, Phillips JH, Lanier LL et al. Activation of NK cells and T cells by NKG2D, a receptor for stress-inducible MICA. Science 1999; 285: 727–729. | Article | PubMed | ISI | ChemPort |
11. Diefenbach A, Jensen ER, Jamieson AM, Raulet DH. Rae1 and H60 ligands of the NKG2D receptor stimulate tumour immunity. Nature 2001; 413: 165–171. | Article | PubMed | ISI | ChemPort |
12. Groh V, Bruhl A, El-Gabalawy H, Nelson JL, Spies T. Stimulation of T cell autoreactivity by anomalous expression of NKG2D and its MIC ligands in rheumatoid arthritis. Proc Natl Acad Sci USA 2003; 100: 9452–9457. | Article | PubMed | ChemPort |
13. Cerwenka A, Baron JL, Lanier LL. Ectopic expression of retinoic acid early inducible-1 gene (RAE-1) permits natural killer cell-mediated rejection of a MHC class I-bearing tumor in vivo. Proc Natl Acad Sci USA 2001; 98: 11521–11526. | Article | PubMed | ChemPort |
14. Ehrlich R, Lemonnier FA. HFE: a novel nonclassical class I molecule that is involved in iron metabolism. Immunity 2000; 13: 585–588. | Article | PubMed | ChemPort |
15. Sanchez LM, Chirino AJ, Bjorkman P. Crystal structure of human ZAG, a fat-depleting factor related to MHC molecules. Science 1999; 283: 1914–1919. | Article | PubMed | ISI | ChemPort |
16. Todorov PT, McDevitt TM, Meyer DJ, Ueyama H, Ohkubo I, Tisdale MJ. Purification and characterization of a tumor lipid-mobilizing factor. Cancer Res 1998; 58: 2353–2358. | PubMed | ISI | ChemPort |
17. Kasahara M, Watanabe Y, Sumasu M, Nagata T. A family of MHC class I-like genes located in the vicinity of the mouse leukocyte receptor complex. Proc Natl Acad Sci USA 2002; 99: 13687–13692. | Article | PubMed | ChemPort |
18. Kajikawa M, Baba T, Tomaru U, Watanabe Y, Koganei S, Tsuji-Kawahara S et al. MHC Class I-Like MILL molecules are beta2-microglobulin-associated, GPI-anchored glycoproteins that do not require TAP for cell surface expression. J Immunol 2006; 177: 3108–3115. | PubMed | ChemPort |
19. Borrego F, Kabat J, Kim D-K, Lieto L, Maasho K, Pena J et al. Structure and function of major histocompatibility complex (MHC) class I specific receptors expressed on human natural killer (NK) cells. Mol Immunol 2002; 38: 637–660. | Article | PubMed | ISI | ChemPort |
20. Barten R, Torkar M, Haude A, Trowsdale J, Wilson MJ. Divergent and convergent evolution of NK-cell receptors. Trends Immunol 2001; 22: 52–57. | Article | PubMed | ISI | ChemPort |
21. Watanabe Y, Maruoka T, Walter L, Kasahara M. Comparative genomics of the Mill family: a rapidly evolving MHC class I gene family. Eur J Immunol 2004; 34: 1597–1607. | Article | PubMed | ISI | ChemPort |
22. Sutherland CL, Rabinovich B, Chalupny NJ, Brawand P, Miller R, Cosman D. ULBPs, human ligands of the NKG2D receptor, stimulate tumor immunity; enhancement by IL-15. Blood 2006; 108: 1313–1319. | Article | PubMed | ISI | ChemPort |
23. Lebron JA, Bennett MJ, Vaughn DE, Chirino AJ, Snow PM, Mintier GA et al. Crystal structure of the hemochromatosis protein HFE and characterization of its interaction with transferrin receptor. Cell 1998; 93: 111–123. | Article | PubMed | ISI | ChemPort |
24. Li P, Morris DL, Willcox BE, Steinle A, Spies T, Strong RK. Complex structure of the activating immunoreceptor NKG2D and its MHC class I-like ligand MICA. Nat Immunol 2001; 2: 443–451. | Article | PubMed | ISI | ChemPort |
25. Brenner SE, Koehl P, Levitt M. The ASTRAL compendium for protein structure and sequence analysis. Nucleic Acids Res 2000; 28: 254–256. | Article | PubMed | ISI | ChemPort |
26. Chandonia JM, Hon G, Walker NS, Lo Conte L, Koehl P, Levitt M et al. The ASTRAL Compendium in 2004. Nucleic Acids Res 2004; 32: D189–D192. | Article | PubMed | ChemPort |
27. Chandonia JM, Walker NS, Lo Conte L, Koehl P, Levitt M, Brenner SE. ASTRAL compendium enhancements. Nucleic Acids Res 2002; 30: 260–263. | Article | PubMed | ChemPort |
28. Murzin AG, Brenner SE, Hubbard T, Chothia C. SCOP: a structural classification of proteins database for the investigation of sequences and structures. J Mol Biol 1995; 247: 536–540. | Article | PubMed | ISI | ChemPort |
29. Ginalski K, Elofsson A, Fischer D, Rychlewski L. 3D-Jury: a simple approach to improve protein structure predictions. Bioinformatics 2003; 19: 1015–1018. | Article | PubMed | ISI | ChemPort |
30. Bennett MJ, Lebron JA, Bjorkman PJ. Crystal structure of the hereditary haemochromatosis protein HFE complexed with transferrin receptor. Nature 2000; 403: 46–53. | Article | PubMed | ISI | ChemPort |
31. Parkkila S, Waheed A, Britton RS, Bacon BR, Zhou XY, Tomatsu S et al. Association of the transferrin receptor in human placenta with HFE, the protein defective in hereditary hemochromatosis. Proc Natl Acad Sci USA 1997; 94: 13198–13202. | Article | PubMed | ChemPort |
32. Feder JN, Penny DM, Irrinki A, Lee VK, Lebron JA, Watson N et al. The hemochromatosis gene product complexes with the transferrin receptor and lowers its affinity for ligand binding. Proc Natl Acad Sci USA 1998; 95: 1472–1477. | Article | PubMed | ChemPort |
33. Wilk JN, Bilsborough J, Viney JL. The mdr1a-/- mouse model of spontaneous colitis: a relevant and appropriate animal model to study inflammatory bowel disease. Immunol Res 2005; 31: 151–159. | Article | PubMed | ChemPort |
34. Okayasu I, Hatakeyama S, Yamada M, Ohkusa T, Inagaki Y, Nakaya R. A novel method in the induction of reliable experimental acute and chronic ulcerative colitis in mice. Gastroenterology 1990; 98: 694–702. | PubMed | ISI | ChemPort |
35. Gaboriau F, Kreder A, Clavreul N, Moulinoux J-P, Delcros J-G, Lescoat G. Polyamine modulation of iron uptake in CHO cells. Biochem Pharmacol 2004; 67: 1629–1637. | Article | PubMed | ChemPort |
36. Bensaid M, Fruchon S, Mazeres C, Bahram S, Roth MP, Coppin H. Multigenic control of hepatic iron loading in a murine model of hemochromatosis. Gastroenterology 2004; 126: 1400–1408. | Article | PubMed | ISI | ChemPort |
37. Miranda CJ, Makui H, Andrews NC, Santos MM. Contributions of beta2-microglobulin-dependent molecules and lymphocytes to iron regulation: insights from HfeRag1-/- and beta2mRag1-/- double knock-out mice. Blood 2004; 103: 2847–2849. | Article | PubMed | ChemPort |
38. Rabinovich B, Li J, Wolfson M, Lawrence W, Beers C, Chalupny J et al. NKG2D splice variants: a reexamination of adaptor molecule associations. Immunogenetics 2006; 58: 81–88. | Article | PubMed | ChemPort |
39. Dunn C, Chalupny NJ, Sutherland CL, Dosch S, Sivakumar PV, Johnson DC et al. Human cytomegalovirus glycoprotein UL16 causes intracellular sequestration of NKG2D ligands, protecting against natural killer cell cytotoxicity. J Exp Med 2003; 197: 1427–1439. | Article | PubMed | ISI | ChemPort |
40. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H et al. The protein data bank. Nucleic Acids Res 2000; 28: 235–242. | Article | PubMed | ISI | ChemPort |
41. Bowie JU, Luthy R, Eisenberg D. A method to identify protein sequences that fold into a known three-dimensional structure. Science 1991; 253: 164–170. | Article | PubMed | ISI | ChemPort |
42. Jameson J, Ugarte K, Chen N, Yachi P, Fuchs E, Boismenu R et al. A role for skin gammadelta T cells in wound repair. Science 2002; 296: 747–749. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We thank Paul Kong for his excellent technical expertise in performing the immuno-histochemical analyses, Julie Jameson for her guidance on the skin organ culture system and Michael Bass for designing the software used for protein threading.

Supplementary Information accompanies the paper on Immunology and Cell Biology website (http://www.nature.com/icb)