Mass Spectrometry Defines Lysophospholipids as Ligands for Chicken MHCY Class I Molecules

Overall, birds and mammals generate innate and adaptive immune responses in similar ways. Although not entirely identical, related molecular and cellular features are frequently observed. Both mammals and birds have, for example, MHC gene regions. Surprisingly, in chickens and some other birds, there is a second MHC-like region containing polymorphic MHC class I– and class II–like genes for which there is no clear equivalent in mammals (1–3). This second MHC-like gene region in chickens, called MHCY (also known as Rfp-Y and MHC-Y), contains genes belonging to five gene families including specialized, polymorphic MHC class I genes, several MHCY class IIβ loci, multiple C-type lectin-like genes, LENG9 loci, and various zinc finger protein genes (1, 3–8). Members of these gene families are distributed in sequence blocks amid repeat elements and in haplotypes that vary widely in size and in the number of genes present (8).

Associations between MHCY and immune responses are emerging in a variety of experiments. MHCY class I genes are widely expressed (9, 10). Changes in MHCY class I gene expression have been noted in several unbiased assays examining changes in gene expression during immune responses (11–14). MHCY genotypes and Ab titers are associated in chicken lines developed under long-term bidirectional selection for high and low Ab responses to an experimental Ag, sheep RBCs (15). MHCY haplotypes also appear likely to influence the colonization of chickens by Campylobacter jejuni (16), a bacterium with an unusually high lysophospholipid content (17). In cell-mediated cytotoxicity assays using MHC class I expressed from transgenes, there is some evidence that MHCY class I molecules have a role in guiding the natural killing activity of embryonic CD8αα⁺, CD3⁻ splenocytes that have been stimulated in vitro with chicken rIL-2 (18).

Chicken MHCY class I molecules have features that set them apart from classical polymorphic MHC class I molecules (9, 19). Crystallographic analysis of YF1*7.1/β₂-microglobulin (β₂m) heterodimers revealed a binding groove that is hydrophobic and too narrow to accommodate peptide Ag, but not the identity of bound ligand (19, 20). The display of polymorphic residues on MHCY class I molecules is different from that of classical peptide-binding MHC class I molecules (8, 9, 19). There is no evidence in dN/dS analyses that MHCY sequences are under the overdominant selection associated with the polymorphism observed among classical MHC class I loci (8).

MHCY was not among the alloantigen systems found during early investigations of blood groups in chickens (21), but alloimmune responses can be raised against MHCY class I molecules as demonstrated by the isoform-specific alloantiserum produced against YF1*7.1 (10). This highly specific alloantiserum revealed that MHCY class I is present on the surfaces of erythrocytes, lymphocytes, granulocytes, monocytes, and thrombocytes within the spleen of developing and newly hatched chicks (10).

This study was undertaken to define ligands bound by MHCY class I molecules using mass spectrometry as the primary analytical tool. Two MHCY isoforms were examined. MHCY heterodimers were reconstituted by dilution from recombinant MHCY class I and β₂m proteins produced in Escherichia coli. Heterodimers were reconstituted in the presence and absence of lipids added to the refolding buffer. Formation of YF1/β₂m heterodimers were verified by SDS-PAGE. Ligands present in heterodimers were identified by mass spectrometry with excess β₂m in separate fractions in the same preparations serving as controls.

Cloned DNA for the YF1*7.1 (AF218783) H chain (α1–α3 domains) and for chicken β₂m (M84767.1) were transferred from previously described pMAL-p4x clones (20) into the pET-17b expression vector. DNA for the α1–α3 domains of YF1*RJF34 (MW423628) were cloned into pET-17b from a YF1*RJF34 cDNA clone in pBluescript. After the transfers were verified to be correct, the pET-17b clones were expressed in BL21(DE3)-Gold E. coli cells grown in 500-ml cultures in terrific broth plus 3-(N-morpholino)propanesulfonic acid and 100 μg/ml carbenicillin medium. Protein expression was induced with 1 mM isopropyl β-d-thiogalactopyranoside (20, 22). Cells were harvested by centrifugation and pellets were frozen. The pelleted cells were thawed at 4°C in a solution containing 25% sucrose, 50 mM Tris-Cl (pH 8), 10 mM DTT, 0.1% NaN₃, and Roche complete protein inhibitors and then disrupted by two passages through a French press. After disruption, the broken cells were contained in a volume of ∼60 ml. This slurry was brought to 50 mM NaCl and 1 mM MgCl₂. Then, 10 mg of lysozyme and 250 μg of DNase I were added. Triton X-100 and Nonidet P-40 detergents were then added to final concentrations of 1% and the slurry was stirred for 30–60 min at room temperature. EDTA was added to a final concentration of 7 mM. The slurry was snap-frozen in liquid nitrogen. Following thawing at 37°C for 30 min, MgCl₂ was added to 8 mM. Incubation was continued for 30–60 min, during which the viscosity of the slurry decreased. An additional aliquot of EDTA was added, the slurry centrifuged, the supernatant discarded, and the pellet containing inclusion bodies was retained. The inclusion bodies were washed twice with sonication in 100 mM NaCl, 50 mM Tris-Cl (pH 8.0), 1 mM EDTA, 1 mM DTT, and 0.1% NaN₃, first with and then without 0.5% Triton X-100. The inclusion bodies were pelleted and solubilized in buffered 8 M urea. The solubilized YF1*7.1, YF1*RJF34, and β₂m proteins were purified by fast protein liquid chromatography (FPLC) on a HiLoad 16/600 Superdex 75 prep grade (pg) column (MilliporeSigma) and concentrated with spin columns.

YF1*7.1 H chain was mixed at an approximate 1:1 molar ratio with β₂m and renatured by dilution. Initially, standard folding buffer (100 mM Tris-HCl [pH 8], 0.4 M l-arginine, 5.0 mM reduced glutathione, 0.5 mM oxidized glutathione, 2 mM EDTA) was used (22). In later experiments, the folding buffer was modified to include 10% RPMI 1640 (without phenol red). RPMI 1640, which contains no proteins or lipids, was added when it was observed in additional experiments not included here to reduce precipitation during refolding and to increase yields. The conditions for renaturing of YF1*7.1/β₂m were as follows: 1) without added candidate ligands, 2) in the presence of synthetic lysophosphatidylethanolamine (lyso-PE) 17:1 (product no. 856707 from Avanti Polar Lipids, Alabaster, AL, also known as LPE 17:1, and 1-(10Z-heptadecenoyl)-sn-glycero-3-phosphoethanolamine), and 3) in the presence of E. coli total lipid extract (product no. 100500 from Avanti Polar Lipids). The lipids, supplied in chloroform, were dried under argon and redissolved in DMSO before adding ∼1.7 mg to 200 ml of refolding buffer. The second isoform, YF1*RJF34, was similarly prepared, mixed with β₂m, and refolded by dilution in the absence of added ligand and in the presence of the E. coli total lipid extract. After refolding, samples were concentrated using a 10 kDa cutoff Amicon filter and purified by FPLC on a size-exclusion HiLoad 16/600 Superdex column (MilliporeSigma). Column fractions containing refolded heterodimers (MHCY class I H chain and β₂m proteins in the same fractions; sample) and β₂m alone in later fractions (control) were identified in Coomassie Blue–stained SDS polyacrylamide gels. Sample column fractions and control column fractions were pooled separately, concentrated, and passed through 0.22-μm filters prior to mass spectrometry.

Concentrated sample and control fractions were separately injected directly onto a Phenomenex Kinetex C18 column (100 × 2.1 mm, 1.7 μm particle size, 100 Å pore size) in a Dionex UltiMate 3000 UHPLC (Ultra HPLC) system (Thermo Fisher Scientific, San Jose, CA). Separation was achieved by reversed-phase chromatography. Two elution systems were used in this study. All but two experiments were performed with elution system 1. For elution system 1 the solvents were acetonitrile containing 0.1% formic acid (solution B) and water containing 0.1% formic acid (solution A). A flow rate of 200 μl/min was used, and the column was maintained at 45°C. The gradient steps were as follows: 1) 3% solution B for 2 min, 2) an increase to 30% solution B during 4 min, 3) an increase to 98% solution B during 2 min, and 4) maintenance at 98% solution B for 6 min before returning to initial conditions during 2 min to re-equilibrate the column at 3% solution B for 4 min. Total time was 20 min. The analyses to generate the E. coli total lipid extract profile presented (Supplemental Table I) and for the YF1*7.1/β₂m heterodimer formed in the presence of E. coli total lipid extract (see Table II) were performed with elution system 2, which allowed more hydrophobic lipids from the E. coli total lipid extract to elute from the column. For elution system 2 the solvents were as follows: solution A, 60% water with 40% acetonitrile and 10 mM ammonium formate; and solution B, 90% isopropanol, 10% acetonitrile, with 10 mM ammonium formate. A flow rate of 300 μl/min was used, and the column was maintained at 55°C. The gradient steps were as follows: 1) starting and maintaining the column in 60% solution A and 40% solution B for 1 min, 2) an increase to 100% solution B during 15 min, 3) maintaining solution B for 4 min, 4) moving back to solution A during 2 min, and 5) finally re-equilibrating the column for 5 min in 60% solution A and 40% solution B. Total time was 27 min.

Eluents from the chromatography were ionized and analyzed in a Thermo Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, San Jose, CA) operating in negative ion mode. For MS¹, spectra were acquired at 120,000 resolution in m/z range 200–1,400, maximum injection time 100 ms, automatic gain control target 200,000, and S lens 50. For MS², ions were isolated with the quadrupole at an isolation width of 1.3 Da and fragmented by higher energy collisional dissociation 25% collisional energy and then acquired in the Orbitrap at 15,000 resolution for as many ions as possible. An exclusion list, generated from a blank run, was used to avoid triggering MS² spectra from background ions. MS² spectra were analyzed with LipidSearch software 4.1 (Thermo Fisher Scientific). Lipids were identified from the MS² spectra by in silico database searches. Only grade A and B identifications are included in the results (23). Lipids identified in FPLC fractions of the sample were compared with the lipids identified in the β₂m control using the same software. Only lipids with a fold difference >2 and p value <0.05 between sample and control were considered as potential ligands.

The structure of YF1*7.1/β₂m was downloaded from the Protein Data Bank (PDB; ID: 3P73) (19). The protein structure was prepared using the Protein Preparation Wizard in Maestro (Schrödinger, New York, NY) (24). The lipid structures were built using the Maestro graphical user interface, and MacroModel was used to generate multiple energetically feasible conformations using ligand torsional sampling (Schrödinger release 2021-4: MacroModel). These lipid conformations were docked to the binding site in YF1*7.1 using standard precision Glide (25). The first 16 carbon atoms of lyso-PE 18:1 were positioned according to the lipid coordinates in the crystal structure of YF1*7.1 (PDB ID: 3P73), as modeled based on the electron density reported earlier (19). For this, the ligand similarity-based restraint in Glide was used. The head group atoms were freely sampled. Additionally, the van der Waals radii of the ligand heavy atoms were scaled by 0.5 to allow more diversity in the docked poses. The resulting docked poses were clustered by ligand root-mean-square deviation, and one representative from each cluster was retained. The final docked poses were selected based on optimal packing of the head groups within the binding cavity and the highest number of polar contacts. The binding cavities of the selected poses were minimized using Prime (26). The lyso-PE 18:1 poses within the MHCY class I binding groove were visualized using PyMOL (PyMOL Molecular Graphics System, version 1.2r3pre, Schrödinger).

The initial aim of this study was to identify the ligand reported to be held within the MHCY class I heterodimer YF1*7.1/β₂m for which a structural determination (PDB ID: 3P73) has been made. Hee et al. (19) reported the presence of a ligand with an extended hydrophobic chain of 17 atoms and a tetragonal head group. The ligand was not identified but was considered to possibly be cetrimonium, a quaternary ammonium surfactant commonly used in laboratory settings that might have been present within the protein preparations in this earlier study. Initially, we renatured YF1*7.1/β₂m heterodimers in the presence and absence of added cetrimonium, as we sought to confirm the presence of cetrimonium in the binding groove using mass spectrometry. We did confirm the presence of cetrimonium in the samples renatured in its presence, but, more importantly, we found that YF1*7.1/β₂m heterodimers in the control preparations refolded in the absence of cetrimonium. When it became apparent that refolding occurred in the absence of added ligand, as is readily evident when fractions from column purification were analyzed by SDS-PAGE (Fig. 1, left), we focused on what might be in the binding groove in the absence of added ligand. Mass spectrometry revealed the presence of lysophospholipids (Table I, top). The lysophospholipids are presumed to be carried over from the E. coli cells in which the recombinant proteins were produced. Lyso-PE 17:1 was found in greatest abundance.

To verify that lyso-PE 17:1 was correctly identified, we renatured YF1*7.1 and β₂m in the presence of added synthetic lyso-PE 17:1. The added lyso-PE 17:1 significantly increased the yield of YF1*7.1/β₂m refolded protein (Fig. 1, right) compared with YF1*7.1 and β₂m renatured without added lyso-PE 17:1 (Fig. 1, left). In the mass spectrometry results (Table I, bottom), the yield of the lyso-PE 17:1 in the heterodimer fractions (sample) doubled whereas the β₂m alone fractions (control) remained unchanged. Other lysophospholipids, that is, lyso-PE 16:0 and lyso-PE 16:1, were found to be less commonly present. These findings provide evidence that lysophospholipids are ligands for YF1*7.1/β₂m.

To begin to test the specificity of the YF1*7.1/β₂m heterodimers for binding of lyso-PE 17:1, we renatured YF1*7.1 and β₂m in the presence of the complex mixture of lipids contained in a total lipid extract from E. coli. This extract contains hundreds of diverse lipids (Supplemental Table I). In the presence of the total lipid extract, lysophospholipids were again the ligands preferentially bound in the YF1*7.1/β₂m heterodimer (Table II, bottom). Lyso-PE 18:1 was found to be present in highest abundance with lyso-PE 17:1 also present, but was less frequently found. The sample/control ratio value of 5980 for lyso-PE 18:1 indicates high specificity binding. Also identified, but less common, were two lysophosphatidylglycerols (lyso-PGs), lyso-PG 17:1 and lyso-PG 16:0. When YF1*7.1/β₂m was renatured in buffer only, lyso-PE 17:1 was again the dominant ligand as found earlier (Table II, top). Also identified in the absence of added lipid extract were lyso-PE 19:1 and three diacyl lipids (PE 17:1/16:0, PE 16.0/18.1, and PE 16:0/16:1). The low sample/control ratios for the diacyl lipids indicate weak specificity in binding.

To gain insight into molecular interactions underlying the observed selection of lysophospholipids for binding in the YF1*7.1 binding groove, we modeled lyso-PE 18:1 in the YF1*7.1 groove starting with the hydrocarbon chain coordinates reported earlier (19). In the full view of the model in (Fig. 2A, the overarching portion of the binding groove has been made semitransparent so that residues within the groove can be viewed. Thirteen hydrophobic residues within the YF1*7.1 binding groove, named and identified with side chains in green, are predicted to interact with the lyso-PE 18:1 hydrocarbon chain. The lyso-PE headgroup is located to the right in a pocket outside the hydrophobic region. In this portion of the groove, there are 12 residues within 5 Å of the lyso-PE 18:1 headgroup (Fig. 2B). Thick sidechain sticks identify five of these predicted to be involved in polar contacts with the headgroup atoms. Salt bridges are predicted between Arg-82, Arg-142, and Glu-146 and the headgroup. Residues Asn-75 and Trp-143 are predicted to form hydrogen bonds with the headgroup. Seven additional YF1*7.1 residues identified with thin sidechain sticks are predicted to be within 5 Å of the headgroup but to have little or no interaction with it.

To begin to examine how sequence differences among MHCY class I isoforms might affect ligand binding, we expressed and analyzed a second MHCY isoform from another cDNA clone, YF1*RJF34. YF1*RJF34 differs from YF1*7.1 at 35 positions in the α1 and α2 domains that form the MHCY ligand binding region. Mass spectrometry of YF1*RJF34 revealed lyso-PE 18:1 to again be the dominant ligand when refolding was carried out in the presence of total lipid extract from E. coli (Table III, bottom). Similar to earlier findings for YF1*7.1, lyso-PE 17:1 and lyso-PE 19:1 were found in YF1*RJF34/β₂m renatured in the absence of added lipid extract (Table III, top).

To better understand how lyso-PE 18:1 could be the dominant ligand bound in both YF1*7.1 and YF1*RJF34, we aligned the sequences of their ligand binding domains (Fig. 3A) and mapped the positions with polymorphism in a molecular model (Fig. 3B). Most of the residues predicted to interact lyso-PE 18:1 are identical or conserved substitutions (noted by o and ϕ in (Fig. 3A). Among the 35 positions where the two sequences differ, marked with orange in (Fig. 3A and 3B, only 7 are predicted to interact with the lyso-PE 18:1 ligand (marked by ϕ, Δ, and *). Among these, conservative hydrophobic residue substitutions, unlikely to significantly change binding, occur at two of the positions (Leu>Met at 9 and Iso>Leu at 32, marked by ϕ in (Fig. 3A and labeled in (Fig. 3B). Residue differences at the remaining five positions are also unlikely to substantially affect the binding with the lyso-PE 18:1. This includes where hydrophobic residues in YF1*7.1 are replaced with polar residues in YF1*RJF34 (Met>Lys at 94 and 112 Tyr>Asp at 112, marked by * in (Fig. 3A and labeled in (Fig. 3B). The substituted polar residues at 94 and 112 are of opposite charge and closely positioned in the folded molecule. It is anticipated that they will produce a charge-neutral surface available for interaction with lyso-PE 18:1. Substitutions at three remaining positions (Trp>Glu at 74, Asn>Gly at 75, and Trp>Arg at 43, marked by 8 in (Fig. 3A and labeled in (Fig. 3B) are also not likely to preclude interaction with lyso-PE 18:1. These substitutions provide alternate polar residues for interactions with the lysolipid headgroup. The replacement of Trp with Glu at 74 provides the possibility of an additional salt bridge with the lyso-PE headgroup. Replacement of Asn with Gly at 75 might allow hydrogen bonding if the YF1*RJF34 α1 backbone is rearranged in the vicinity. Glycine residues are known to induce backbone rearrangement by disrupting α helices (27). Replacement of Trp with Arg at 143 would likely provide another stable salt bridge with the phosphate of the headgroup. The remaining positions where amino acid differences occur between the two isoforms are located away from the region involved with ligand binding. In summary, modeling of the binding of lyso-PE 18:1 in YF1*7.1 and YF1*RJF34 supports the mass spectrometry findings.

In this study, we report findings from mass spectrometry providing evidence that lysophospholipids are bound within the MHCY1*7.1 and MHCY1*RJF34 class I molecules when they are refolded with β₂m in vitro in the presence of a complex mixture of lipids from E. coli. The findings are consistent with the earlier observation in a crystallographic study of an unidentified molecule with an extended hydrocarbon chain and a tetragonal headgroup present within the MHCY1*7.1 binding groove (19).

The two MHCY class I isoforms in this study provide insight into the specificity of lipid binding in MHCY class I molecules. The results define single-chain hydrocarbon lysolipids as ligands preferentially bound by MHC class I molecules refolded in vitro in the presence of a complex mixture of lipids (Supplemental Table I). In folding of both isoforms, lyso-PE 18:1 was found to the dominant ligand. This suggests that lyso-PE 18:1 may be the best fitting ligand. However, it also happens that lyso-PE 18:1 is the lysophospholipid present at the highest concentration in the lipid mixture in the E. coli extract used in this test (Supplemental Table I). It is not yet clear how much the dominant binding of lyso-PE 18:1 is affected by its abundance in the folding buffer. The binding of other lysolipids do provide some evidence for specificity in the binding of lysolipids. Lipids lyso-PG 17:1 and lyso-PG 18:1 are seventh and sixth in order of concentration among the lysolipids present in the E. coli lipid extract (Supplemental Table I), but they are second in abundance in the binding grooves of MHCY1*7.1 and MHCY1*RJF34, respectively (see bottom portions of Tables II and III). More work is needed to define selectivity in the binding of lysolipids within the MHCY class I binding groove.

At this time, nothing is known about the binding of ligands by MHCY class I molecules expressed in vivo. It could be that lyso-PE and lyso-PG will be the principal MHCY class I ligands revealed in molecules folded in vivo. Although lyso-PE and lyso-PG are commonly found in bacteria, these lysolipids are typically minor components in the array of lipids present in eukaryotic cells. Lyso-PE in eukaryotic cells is typically membrane bound, contributing to membrane fluidity, membrane curvature, and tubule formation (28). Because of low availability, it might be that endogenous lyso-PE typically fills only a portion of the MHCY class I molecules present in vivo and that a variety of other lipids are held within the hydrophobic groove in the absence of available lyso-PE and lyso-PG. It is also possible that MHCY class I molecules may frequently be empty in vivo or are filled with loosely bound low-specificity lipids in the absence of infection. When infection occurs, structurally preferred lysolipid ligands might become more available and bind preferentially in the MHCY binding groove. Binding of lysolipids originating from infectious organisms might lead to increased stability of MHCY class I molecules on the cell surface. Greater recognition of MHCY class I might then occur, allowing increased immune recognition and the promotion of immune responses. Lyso-PE and lyso-PG are fairly abundant in Gram-negative bacteria. Lyso-PE and lyso-PG have been found to be required for motility of C. jejuni under low oxygen tension (17), and short-chain lyso-PE is a C. jejuni virulence factor for permeabilizing host cell membranes (29). It will be interesting to learn whether lysolipids play a role in the recently reported link between MHCY and immune responses to C. jejuni (16). In the human thymus, lyso-PE derivatives are known to guide development and maturation of semi-invariant NKT cells (30). It might be that lyso-PE binding in MHCY class I has a similar function in guiding the development of specialized immune cells in chickens. MHCY class I molecules with bound ligands of microbial origin might directly stimulate natural killing responses or stimulate a series of responses via the secretion of cytokines that lead to advanced immune responses. It will be interesting to learn more about the specificity of responses guided by the different MHCY class I isoforms. More work is needed including experiments to determine whether other isoforms share the same ligand specificity revealed in this study for YF1*7.1 and YF1*RJF34.

For understanding the function of MHCY class I molecules, it is important to identify ligands bound by MHCY class I molecules in vivo. Defining ligands bound in vivo will require purification of native MHCY class I molecules or the products of transgenes expressed in chicken cells. A direct approach for isolating MHCY by immunoprecipitation from chicken tissues and then analyzing the product by mass spectrometry would be ideal, but mAbs specific for MHCY class I molecules are lacking. A mAb to β₂m could possibly be used, but copurification of MHCB class I molecules might complicate ligand analysis. Expression of FLAG epitope-tagged MHCY class I molecules in chicken cells followed by immunoaffinity purification based on the FLAG epitope provides an alternative approach. This method provided good yields of MHCB class I molecules from transgenes expressed in chicken cells that allowed identification of peptide ligands by mass spectrometry (31). Parallel experiments for MHCY class I molecules also provided good yields and mass spectrometry data indicating the absence of peptide ligands in MHCY class I (data not shown).

What role amino acid variability among MHCY class I isoforms has in signaling is not yet clear. Current findings for two isoforms indicate that there is a conservation of amino acids within the binding groove that is apparently of importance in stabilizing ligands. The role of variability elsewhere in the MHCY class I molecules remains to be defined. The nonbinding groove variability suggests that these residues may contribute to specificity in the interactions of MHCY class I molecules with receptors. Experiments are needed to define cells bearing receptors that recognize MHCY class I and to identify these receptors. Interactions could be similar to those known to occur with other specialized MHC class I–like molecules and their cognate receptors, which are typically specialized TCRs. Human CD1c (32), CD1d (33), and MR1 (34) all bind specialized ligands and are known to guide early immune responses through specific recognition interactions with specialized T cells. The details of cellular interactions and the outcome of MHCY class I presentation are especially important to understand because, in contrast to these other better known MHC class I–like molecules, MHCY class I molecules possess polymorphism that may confer an added level of specificity in their contributions to immunity. It will be interesting to see whether MHCY class I genetic variability underlies heritable differences in the responses of chickens to the challenge of infectious disease.

We are grateful to Pamela J. Bjorkman and Beth Stadtmueller, Caltech, and Fiyaz Mohammed and Benjamin Willcox, University of Birmingham, for valuable guidance in the production and refolding of recombinant MHCY class I molecules. We thank Timothy O’Conner, City of Hope, for unlimited use of the AKTA Pharmacia FPLC. We thank Mary E. Delany, UC Davis, for providing us with RNAzol preserved UCD line 001 (RJF) tissue for MHCY class I cDNA cloning.

The authors have no financial conflicts of interest.