MHC Class I and II Peptide Homology Regulates the Cellular Immune Response

Abstract

Mammalian immune responses are initiated by “danger” signals - immutable molecular structures known as PAMPs. When detected by fixed, germline encoded receptors, PAMPs subsequently inform the polarization of downstream adaptive responses depending upon identity and localization of the PAMP. Here we report the existence of a completely novel “PAMP” that is not a molecular structure but an antigenic pattern. This pattern - the incidence of peptide epitopes with stretches of 100% sequence identity bound to both dendritic cell MHC class I and MHC class II - strongly induces T_H1 immune polarization and activation of the cellular immune response. Inherent in the existence of this PAMP is the concomitant existence of a molecular sensor complex with the ability to scan and compare amino acid sequence identities of bound class I and II peptides. We provide substantial evidence implicating the multi-enzyme aminoacyl-tRNA synthetase complex and its AIMp1 structural component as the key constituents of this complex. The results demonstrate a wholly novel mechanism by which T_H polarization is governed and provide critical information for the design of vaccination strategies intended to provoke cell-mediated immunity.

Introduction

Mammalian immune responses are initiated by canonical danger signals known as pathogen-associated molecular patterns or PAMPs. PAMPs are conserved molecular structures so fundamental to the virulence, survival, or reproduction of a pathogen as to render impractical or impossible the evolution of escape mutants unable to be recognized by germline-encoded receptors called pattern recognition receptors or PRRs. Examples of such fixed molecular patterns include fundamental subunit components of lipopolysaccharide, peptidoglycan, flagellin, unmethylated CpG dinucleotides, and certain nucleic acid variants associated with viral genomes or their replication intermediates (1). Viruses are an interesting class of pathogen in that they are not obviously amenable to innate PRR detection given their direct derivation from the mammalian host and possession of few immutable non-mammalian patterns. Even viral nucleic acid products discernable from mammalian counterparts by structural differences are detected as much upon the basis of their compartmental localization within the cell than upon unique molecular patterning (2–7). The existence of at least nine different innate nucleic acid sensors (TLR3, TLR7, TLR8, TLR9, RIG-I, MDA5, LGP2, STING, cGAS) participating in the recognition of pathogenic nucleic acids (4, 8) testifies to the inherent difficulty in detecting numerous different categories of viral infection. Further, the diversity and redundancy of nucleic acid-detecting PRRs is also a testament to the multiplicity of virulence factors that pathogens continually evolve to subvert innate PRR detection (4, 9, 10). Yet despite any perceived challenges to immune discernment of viral infection, such infections are generally recognized and cleared efficiently and without development of concomitant autoimmunity, raising the possibility that other signals or complimentary mechanisms of detection could exist in parallel with known PRR.

T_H polarization of downstream adaptive immune responses is significantly influenced both by the specific identity of detected PAMPs as well as the cellular compartment in which detection occurs. When pathologic nucleic acid variants are detected by intracellular PRR, T_H1 polarization and adaptive cellular immunity are induced (11–16). Similarly, fixed pattern ligands associated with extracellular pathogens and detected by extracellular TLRs induce T_H2 polarization which includes humoral immunity, the production of IgE, and the activation of phagocytes (17–22). In this regard, both the specificity and the localization of any given PAMP have been shown to significantly influence the downstream T_H polarization of subsequent adaptive responses.

The existence of a novel, noncanonical pattern that induces cell-autonomous T_H1 polarizing characteristics in dendritic cells was first described a decade ago. This “PAMP” is not a fixed molecular pattern but rather a unique environmental byproduct that physiologically occurs almost exclusively within the context of fulminant viral infection (23). The fundamental character of a viral infection obliges the intracellular and extracellular antigenic milieu of virally-infected cells to be substantially similar given that the viral proteins bathing the cellular exterior (24–26) are also simultaneously being synthesized on the inside of infected cells. Professional antigen presenting cells (APC) of the adaptive immune system efficiently sample both the intracellular and extracellular antigenic environments, binding peptide antigens of each environment to presentation complexes termed major histocompatibility (MHC) class I and MHC class II respectively. At a certain threshold of identity, the presence of an equivalent intra- and extracellular antigenic environment become associated with increasingly high likelihood of intracellular infection, specifically when detected in conjunction with canonical intracellular nucleic acid danger PAMPs or other inflammatory signals like interferons, TNF-α, IL-6, and/or IL-1β (23). According to this hypothesis, the presence of substantial sequence homology between MHC class I and II peptide antigens in conjunction with pathological nucleic acid patterns or inflammatory cytokines is perceived as viral infection, necessitating activation of T_H1 adaptive immunity to mediate clearance. Indeed, in a variety of different studies, the provision of dendritic cells with obligately similar internal and external antigenic environments in the form of transfected nucleic acids and extracellular proteins corresponding to the products of those transfected nucleic acids resulted in cell autonomous upregulation of DC costimulatory marker and IL-12 secretion, reorganization of the DC transcriptome, and preferential generation of IFN-γ-secreting CD8⁺ T-cells with enhanced cytolytic and memory functions (23, 27–29). However, many different potential T_H polarizing cues are associated with nucleic acid transfections and protein or lysate production that might also modulate T_H polarization (30–34), and the possibility remains that previous experiments, no matter how well-controlled, did not or could not adequately account for all potential associated or contaminating pattern-based signals. Therefore, to definitively validate the hypothesis that high intra- and extracellular antigenic homology serve as a T_H1-polarizing signal, it was necessary to both a) develop defined antigenic model systems free of TLR and cytokine confounders and b) identify experimental output signal(s) specific to homologous antigenic environments that are also unresponsive to the presence of pattern-based ligands. With these goals in mind, we sought to expand upon previous studies (23, 27–29) to characterize the specificity of class I and II peptide antigenic homology required to induce T_H1 polarizing immune responses as well as the molecular mechanisms that underpin this phenomenon.

Methods

Reagents Used

Antibodies: αHuman/mouse CTLA-4 (WB) (Abcam; Cambridge, MA); αHuman/mouse AIMp1 (Lifespan Biosciences Inc, Seattle, WA); αHuman AIMp2 (Abcam); αMouse CD8 (flow cytometry), αMouse CD25, αMouse CD3, and αMouse CD4 (BD Biosciences); αHuman/mouse β-actin was purchased from Santa Cruz Biotechnologies (Santa Cruz, CA). αHLA-A/B/C was purchased from BioLegend, San Diego, CA. HLA typing antibodies: αHLA-A2-FITC (BD Biosciences), αHLA-B8-biotin (Abcam), and unconjugated αHLA-DR3/DR6 (Lifespan Biosciences). αLysyl-tRNA synthetase (GeneTex, Inc, Irvine, CA); αTyrosyl-tRNA synthetase (Abcam); αHistidyl-tRNA synthetase (Acris subsidiary of OriGene, Rockville, MD); αAlanyl-tRNA synthetase (GeneTex); αMethionyl-tRNA synthetase (Pierce subsidiary of Thermo Fisher Scientific, Waltham, MA); αGlycyl-tRNA synthetase (Acris). Confocal microscopy antibodies and reagents: αCTLA-4-biotin clone BNI3 (Cat No. 555852; BD Pharmingen, San Jose, CA); Streptavidin-APC (Cat No. 554067; BD Pharmingen); αRab5 (Cat No. 108011; Mouse-Monoclonal Synaptic Systems, Goettingen, Germany); αRab11 (Cat No. 610656; BD Biosciences, San Jose, CA); Giantin (Courtesy of Dr. Richard Sifers, Baylor College of Medicine, Houston, TX); Alexa-fluor Ms546 (Courtesy of Dr. Anna Sokac, Baylor College of Medicine); Alexa-fluor Rb546 (Courtesy of Dr. Anna Sokac); CD3-FITC (Cat No. 555332; BD Pharmingen); CD11c clone 3.9-Alexa-fluor 488: (Cat No. 301618; Biolegend); and DAPI: Slow Fade Gold Antifade Mountant (Cat No. S36938; Molecular Probes, Grand Island, NY). TLR agonists: TLR-3 agonist poly(I:C)-rhodamine, TLR-9 agonist CpG ODN-FITC, and TLR-5 agonist flagellin were obtained from InvivoGen (San Diego, CA). TLR-4 agonist LPS was obtained from Sigma-Aldrich (St. Louis, MO). DC uptake of poly(I:C)-rhodamine and CpG-FITC was confirmed and quantitated by fluorescent microscopy and flow cytometry using an LSR II flow cytometer (BD Biosciences) and analyzed with Flow Jo version 10.0.00003 for the MacIntosh (Tree Star Inc, Ashland, OR). All TLR agonists were used at a concentration of 1 μg/ml. Peptides: LCMV glycoprotein peptides gp33–41 (KAVYNFATC), gp31–45 (GIKAVNFATCGIFA), and gp66–80 (DIYKGVYQFKSVEFD) were synthesized by United Bio Systems (Herndon, VA). Influenza A New Caledonia hemagglutinin peptides WLTGKNGL, RNLLWLTGKNGLYPN, VLLENERTL, and ELLVLLENERTLDFH (described previously(23)) as well as methionine-for-glycine substituted derivatives WLTMKNML and RNLLWLTMKNMLYPN were synthesized by United BioSystems. Ovalbumin H-2K^b immunodominant peptide SIINFEKL was synthesized by Anaspec (Freemont, CA). H-2D^b CLIP-overlapping MRMATPLLM was synthesized by United Biosystems. Recombinant endotoxin-free ovalbumin protein was purchased from In vivo Gen. Other: AIMp1 (SCYE1, mouse and human), β2-microglobulin (mouse and human), HLA-DM (human), and TAP1 (mouse) siGenome SMART Pools and non-targeting siRNA pools were purchased from Fisher Thermo Scientific (Wilmington DE). The H-2D^b KAVYNFATC LCMV gp tetramer was provided by the Baylor College of Medicine MHC Tetramer Production core.

Mice

Eight-to-twelve-week-old C57BL/6 and Balb/c mice were obtained from Harlan Laboratories (Indianapolis, IN) or from the Jackson Laboratory (Barr Harbor, ME). H2-DM knockout mice in the C57BL/6 background were a kind gift from Dr. Jenny Ting at the University of North Carolina, Chapel Hill. All mice were maintained in accordance with the specific IACUC requirements of Baylor College of Medicine and animal protocol AN-1478.

Preparation of Vaccine Materials, DC Characterization and Loading, and In Vitro CoCulture.

Peptides were reconstituted in an 80:20 dH₂O:DMSO solution at 10 mg/ml and stored at −80° C. To generate MHC class I (mRNA) or II (lysate) determinants (Fig 2A only), tissue fractions were first disrupted using a Polytron PT1200E tissue homogenizer (Kinematica, Inc, Bohemia, NY) after which cell lysates and total mRNA were prepared as described previously (27). Peptide loading of DC was performed as described previously (29). Briefly, immature DC were resuspended in Viaspan (Barr Laboratories, Pomona, NY) at 40 × 10⁶ cells/ml and incubated with 10 μg/ml class I peptide on ice for 10 min. DC were then electroporated with exponential decay pulse (250 V, 125 μF, Ω = ∞) using Gene Pulser-X cell electroporator (Bio-Rad, Hercules, CA) in Gene Pulser cuvettes with a 4 mm gap width (Bio-Rad). Following electroporation, cells were diluted in AIM-V medium (Invitrogen, Carlsbad, CA) supplemented with 5% FBS and incubated with class I and class II peptide, both at a final concentration of 10 μg/ml, in a six-well plate. After 3 hours of peptide incubation, DC were diluted to a concentration of 1 × 10⁶/ml in AIM-V medium (Invitrogen, Carlsbad, CA) supplemented with 10% FBS and maturation cytokines were added as described previously (27). Use of siRNA was performed according to the manufacturer’s instructions (FisherThermo Scientific - Dharmacon). Dendritic cells were generated and in vitro co-cultures performed as described previously (23, 27). Human monocyte-derived DC and murine BMDC were characterized by flow cytometry as described previously (23, 27–29, 35–37) with additional lineage characterizations indicated. Maturation cocktail of H2-DM^−/− DC was additionally supplemented with 1 μg/ml CpG-ODN (InvivoGen).

Figure 2. Dendritic cell expression and secretion of CTLA-4⁺ microvesicles is antagonized by AIMp1.

A. RT-PCR was performed on WT or AIMp1^−/− DC using primers specific for CTLA-4, indicating a 30–50 fold upregulation of the CTLA-4 mRNA transcript in the absence of AIMp1. One of n = 4 representative experiments shown. B. Loading of WT DC class I and II compartments with identical antigens resulted in upregulated secretion of AIMp1 and downregulated CTLA-4⁺ microvesicle secretion; however loading of WT DC class I and II compartments with non-identical antigens resulted in the opposite: downregulation of AIMp1 secretion and upregulation of CTLA-4⁺ microvesicle secretion. Results assayed by western blot analysis and image J quantification. One of five representative experiments shown. C. Loading of WT DC class I and II compartments with identical antigens +/− 200 nM AIMp1 siRNA (or non-targeting siRNA) demonstrated dependence of this secretion pattern on AIMp1. Results assayed by western blot analysis and image J quantification. One of n = 5 representative experiments shown. D. AIMp1 and CTLA-4 western blots derived from stripped and re-probed membranes of wild type (left) and H2DM^−/− (right) murine DC culture supernatants following single, homologous, or heterologous loading. Membranes were stripped and re-probed a third time with anti-β-actin to control for both load and cell death. Representative experiment shown. E/F. Densitometry quantitation of AIMp1/CTLA-4 ratios of the data depicted in (D) over three independent experiments normalized to background AIMp1/CTLA-4 expression levels in unloaded DC (shown in representative experiment 2D, left panel). E. wild type DC. F. H2-DM^−/− DC. G. Loading of H2-DM^−/− DC with a class I H-2D^b peptide possessing amino acid sequence homology to Ii-CLIP recapitulates WT regulation of AIMp1 and CTLA-4. Representative experiment shown. Also shown: sequence overlap of class I H-2D^b CLIP with class II Ii-CLIP. H. High AIMp1/CTLA-4 ratios in response to loading of H2-DM^−/− DC with class I H-2D^b CLIP but not H-2K^b SIINFEKL mediated typical downstream augmentation of activated CD8⁺ T-cell expansion in vitro following coculture of H-2D^b CLIP-loaded H2-DM^/- DC with autologous splenocytes. Triplicate experiment shown. For all experiments shown, error bars = +/− SD. *p < 0.05, ^**p < 0.01 by one-way ANOVA with Bonferroni post-hoc.

Vaccination

Mice vaccinated with SIINFEKL/Ova loaded DC were vaccinated once in the footpad with 2.5 × 10⁵ matured DC suspended in 50 μl PBS. Seven days post-vaccination, circulating PBMC were collected by retro-orbital bleed and analyzed by flow cytometry. Mice vaccinated with LCMV peptides received footpad injections of 50 μg in 50 μl vehicle if administered a single peptide or 25 μg (each) in 50 μl vehicle if administered two peptides on days 0, 7, and 14. Concurrent with peptide administration, animals also received an i.p. injection of 500 μg imiquimod (Sigma-Aldrich, St. Louis, MO) in DMSO. Circulating PBMC were analyzed on day 24.

Immunofluorescence and confocal microscopy

As described previously (36), DC were cultured and matured in a six-well plate and subsequently collected onto poly-l-lysine coated coverslips (Corning, Inc., Corning, NY) in a 24-well plate by centrifugation. Cells were fixed in 4% formaldehyde and permeabilized by incubating the coverslips in 0.5% Triton-X-100 (Thermo Fisher Scientific, Waltham, MA). Blocking was performed with TBS-T/1% BSA, and the cells were incubated in the primary antibody overnight at 4° C. The following day cells were washed and incubated with appropriate secondary antibody. The cells were then counterstained with DAPI (Molecular Probes division of Life Technologies, Grand Island, NY). Cover slips were mounted using Prolong Gold antifade reagent (Molecular Probes), and image acquisition was performed on a Zeiss LSM 710 confocal microscope with a 60·/0.95 numerical aperture oil immersion objective (Carl Zeiss, Inc., Peabody, MA). Images were collected at a zoom factor of two with a resolution of 104nm per pixel. All images shown are representative of at least three independent experiments.

RT-PCR

Total RNA was extracted from DC by means of the Trizol method and cDNA synthesis was performed by reverse transcriptase-PCR using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies) according to the manufacturer’s instructions. Real time PCR was performed with the 7500 Real-time PCR system (Applied Biosystems, Foster City, CA) using the Taqman Realtime PCR assay (Thermo Fisher Scientific) according to the manufacturer’s instructions. The primers used were IL-12A (Hs01073447_m1, FAM), IL-12B (Hs00233688_m1, FAM), CADM1 (Hs00942509_m1, FAM), SIRPA (Hs00388953_g1, FAM), and 18s rRNA (4319413E, VIC).

Quantitation of Western Blot Images

Western blot chemiluminescent signal was detected using a ChemiDoc XRS digital imaging system running Image Lab software Version 2.0.1 (Bio-Rad Laboratories, Hercules, CA). All Western blots were quantitated by densitometry of Ponceau S (Sigma-Aldrich) stained membranes. Contamination of supernatants with residual cell lysate or debris from cell death was controlled for by immunostaining with anti-β-actin (Santa Cruz) and additional densitometry. Densitometry was performed using ImageJ software (NIH; Bethesda, MD). For detection of both CTLA-4 and AIMp1 on a single membrane, the membrane was typically probed first with anti-CTLA-4 after which it was stripped in Western Blot Restore buffer (Pierce, Rockford, IL) according to the manufacturer’s instructions and re-probed with anti-AIMp1.

Co-Immunoprecipitation Assay

DC were loaded as indicated and matured for two days before lysis with 1% NP-40 buffer + protease cocktail inhibitor (both from Sigma-Aldrich). Debris was pelleted at 14,000 rpm for 20 minutes at 4° C in a tabletop microfuge, and subsequent cell lysates were precleared with Protein G plus-Agarose Bead suspension IP04 (EMD Millipore; Darmstadt, Germany) for 1 hour at 4° C. Lysates were then rotated overnight at 4° C with Protein G plus antibody-coated beads. Beads were then washed three times in 1% NP-40 buffer, twice in PBS, and immunoprecipitate was collected by boiling in 2% SDS (Sigma-Aldrich) denaturing buffer prior to analysis by PAGE.

MHC Surface Stabilization Assay

H-2^b haplotype B16-F10 cells or C57BL/6 BMDC were electroporated in the presence of 200 nM TAP1 siRNA or non-targeting siRNA under conditions as described above (BMDC) or previously (B16-F10). (38) Following verification of MHC cell surface reduction, cells were incubated with 0.1 μg/ml, 1 μg/ml, or 10 μg/ml H-2 D^b CLIP or positive control LCMV peptide to induce TAP-independent MHC stabilization as previously described in TAP-deficient RMA/S cells. (39) Following overnight peptide incubation, surface MHC was quantitated by flow cytometry and percent MHC stabilization at each peptide concentration was determined by the formula: (MHC MFI of siTAP1 electroporated cells + peptide - MHC MFI of siTAP1 electroporated cells w/o peptide)/(MHC MFI of siNT electroporated cells) for each given peptide concentration.

Statistical Analysis

Statistical significance was defined as p < 0.05 and was determined by Student’s unpaired or paired t-test with one or two tails as statistically appropriate. Statistical differences between multiple groups were validated by one-way or two-way ANOVA. Pearson’s chi-squared test and Tukey HSD post hoc were also performed where appropriate. Statistical tests were performed with Microsoft Excel 2008 for the Macintosh Version 12.0. All normalized quantitation graphs were derived from three independent experiments unless stated otherwise and with error bars = +/− SD.

Results

TLR agonism does not modulate dendritic cell AIMp1 expression.

AIMp1 is a secreted inflammatory cytokine previously reported to be released from a variety of cell types in response to cell stress and other stimuli (29, 40–43). Recent work has substantially clarified the function of AIMp1, the dendritic cell expression of which was demonstrated to be essential for T_H1 polarization (29). Earlier studies demonstrated that genetic ablation of AIMp1 significantly enhanced T_H2-polarized airway hyperreactivity in a model of allergic inflammation (44), while in vitro studies showed that recombinant AIMp1 protein induced upregulated IL-12 secretion from bone marrow-derived DC and enhanced the generation of IFN-γ-secreting CD4⁺ T-cells (43, 45). Recombinant AIMp1 also induced B-cell activation, proliferation, and class switch recombination toward the T_H1-specific IgG₂ isotype with increased antigen-specific antibody production (46). More recent work demonstrated that dendritic cell-expressed AIMp1 is a critical component of the T_H1 regulatory cascade and absolutely required for IL-12 production, downstream T_H1 polarization, and functional antiviral and antitumor immunity in C57BL/6 and 129Sv mice (29).

In previously published work, upregulated AIMp1 expression and secretion in DC were shown to be induced by the presence of equivalent internal and external antigenic environments, i.e. the loading of DC class I and II compartments with highly similar or identical antigens (28, 29). The model system utilized in these experiments involved electroporation of DC with the immunodominant H-2K^b OVA peptide SIINFEKL coupled to incubation with whole OVA protein to provide the corresponding amino acid sequence in the context of MHC class II (29). To determine if upregulated AIMp1 expression was specific to the presence of homologous antigenic environments in DC or if AIMp1 upregulation might also be induced by other means, H-2K^b murine DC were stimulated with a variety of TLR ligands as well as type I interferon and assayed for upregulation of AIMp1 expression. As shown in the representative Fig S1 western blot and Fig 1A quantitation of five independent experiments, only DC loaded with both intracellular SIINFEKL and extracellular OVA together were able to alter AIMp1 expression from mock-treated DC. Neither loading with SIINFEKL or OVA alone nor activation with TLR ligands nor type I interferon was able to influence AIMp1 expression. This experiment suggested that upregulated AIMp1 expression might serve as a specific readout for DC detection of homologous antigenic environments.

Figure 1. Upregulation of AIMp1 expression is independent of TLR agonism and associated with T_H1 immune responses.

A. Only treatment of DC with SIINFEKL MHC class I H-2K^d binding epitope and whole OVA protein was sufficient to modulate AIMp1 expression. AIMp1 expression was not modulated from baseline by treatment with SIINFEKL or OVA alone nor DC activation with flagellin, poly(I:C), CpG, LPS, or type I interferon. Compilation of n = 5 independent experiments. B - E. Cohorts of mice administered DC loaded with SIINFEKL + OVA generated significantly increased B. CD3⁺CD8⁺, and C. CD3⁺CD8⁺CD25⁺ T-cell proliferation in vivo in comparison to DC loaded in any other fashion. This effect was abrogated by treatment of SIINFEKL + OVA-loaded DC with AIMp1 siRNA. Irrelevant heterologous class II antigen was a lysate derived from mouse seminal vesicle. Repetition of this experiment using SIINFEKL + OVA-loaded H2-DM^−/− DC further demonstrated no significant increase of D. CD3⁺CD8⁺, and E. CD3⁺CD8⁺CD25⁺ T-cell populations in vivo when MHC class II could not bind exogenous antigen. n = 3 mice per cohort. F/G. Results generated with ex vivo-derived and loaded DC were recapitulated in vivo using only defined overlapping and non-overlapping MHC binding peptides. F. Representative flow cytometry analysis. G. Quantitation of n = 5 mice per cohort. For all experiments MC = maturation cocktail. S+OVA = SIINFEKL + OVA. siNT = non-targeting siRNA. siAIMp1 = AIMp1 siRNA. Error bars = +/− SD. ^**p < 0.01, ^***p < 0.005, ^****p<0.001 by one-way ANOVA with Bonferroni post-hoc.

To validate that SIINFEKL/OVA loaded DC could also mediate enhanced adaptive T_H1 responses, cohorts of mice were vaccinated with SIINFEKL/OVA loaded DC or one of five other important controls including a) unloaded WT DC, b) WT DC loaded with SIINFEKL only, c) WT DC loaded with OVA protein only, d) WT DC loaded with SIINFEKL and an irrelevant heterologous cell lysate, and e) WT DC loaded with SIINFEKL and OVA and treated with AIMp1 siRNA. All other WT DC populations were treated with non-targeting (NT) siRNA. Seven days following footpad vaccination with 250,000 DC, circulating PBMC were analyzed by flow cytometry. As indicated in Fig 1B–C, only mice vaccinated with WT SIINFEKL + OVA loaded DC showed within the lymphocyte gate a) expansion of the CD3⁺CD8⁺ cell population (118% above unloaded WT DC, ^**p<0.01) and b) expansion of the activated CD3⁺CD8⁺CD25⁺ cell population (127% above unloaded WT DC, ^****p<0.001). T-cell populations derived from mice vaccinated with SIINFEKL + OVA-loaded differed substantially from every other cohort, indicating upregulated T_H1 responses could be induced by homologous loading of class I and II antigens and that such T_H1 responses proceeded in an AIMp1-dependent fashion.

We next repeated this experiment using DC populations derived from H2-DM^−/− mice. The H2-DM molecular chaperone is responsible for removing the CLIP peptide of the invariant chain from the MHC class II binding pocket. In the absence of H2-DM, CLIP is bound almost irreversibly to the class II binding pocket in the I-A^b haplotype (C57BL/6 background), thereby abrogating the ability to load exogenous antigen onto MHC class II among mice of the H-2^b haplotype (47, 48). The inability to load exogenous antigen is the only molecular deficit of H2-DM^−/− DC. H2-DM^−/− DC have no known defects in TLRs, NLRs, or other PRRs and have previously been shown to respond appropriately to TLR agonism (49). The two H2-DM^−/− control DC populations used here consisted of one loaded with SIINFEKL alone and another loaded with both SIINFEKL and OVA. As with the first set of experiments, WT DC loaded with SIINFEKL+OVA mediated upregulation of CD3⁺CD8⁺ (^**p<0.01) and CD3⁺CD8⁺CD25⁺ (^****p<0.001) cell populations in vivo whereas H2-DM^−/− DC were unable to do so in response to loading with SIINFEKL + OVA (Fig 1D–E). Taken together, these results indicated that DC AIMp1 expression is not influenced by common TLR ligands including LPS and may serve as a reliable readout for DC recognition of homologous class I and II determinants.

The preceding experiments were performed with bone marrow-derived DC (BMDC) because mouse BMDC do not cross-present exogenous antigen on MHC class I at physiologic concentrations unless stimulated through TLR-3 or TLR-9 (50). Given that no nucleic acid ligands were used to mature the BMDC (other than those tested individually on AIMp1 expression in Fig 1A) used in these experiments, we were able to avoid this potentially confounding caveat. However, it has also been reported that mouse BMDC may be a mixture of cDC, moDC, and macrophage lineages, the precise composition of which can vary from laboratory to laboratory or even batch to batch and which might also not correspond well to any in vivo-derived DC lineage (51). Therefore to validate that this phenomenon was not restricted to in vitro-derived BMDC (and with discussion on the caveats of cross-presentation provided subsequently), we performed a wholly in vivo experiment using only well-characterized LCMV peptide epitopes known to bind the MHC H-2^b haplotype (class I gp33–41; overlapping class II gp31–45; non-overlapping class II gp66–80, all shown in Fig S2) (52) and for which the class I gp33–41 tetramer is available for analysis of CD8⁺ TCR cognate antigen specificity (53). To perform this experiment, cohorts of mice were footpad injected with saline only, class I gp33–45 only, class II gp31–45 only, class I gp33–41 together with non-overlapping class II gp66–80, and class I gp33–41 together with overlapping class II gp31–45. All mice in all cohorts also received an i.p. injection of the TLR7 agonist imiquimod. Ten days after the final footpad injection, mice were bled retro-orbitally, and PBMC were analyzed for the generation of CD8⁺ LCMV tetramer gp33–41⁺ cells. As shown in the representative flow plots in Fig 1F and the compiled data in Fig 1G, only co-injection of overlapping class I and II epitopes resulted in significant production of antigen-specific CD8⁺ cells (up to 5.5% of circulating CD8⁺ T-cells, ^****p<0.001) whereas single peptides or non-overlapping combinations produced negligible increases in CD8⁺ tetramer⁺ cells. Additionally, mice injected with gp33–41 in one footpad and overlapping gp31–45 in the contralateral foot pad also generated no significant antigen-specific CD8⁺ responses (not shown). These data indicate that stimulation of T_H1 immunity in response to provision of overlapping class I and II antigens may happen in vivo is not an artifact limited to in vitro-derived BMDC.

Dendritic cell expression and secretion of CTLA-4⁺ microvesicles is antagonized by AIMp1.

In direct contrast to the function of AIMp1, dendritic cell-expressed and secreted CTLA-4⁺ microvesicular bodies [characterized extensively in (36)] powerfully inhibit CD8⁺ T-cell priming and are associated with T_H2 polarization. While T-cell expressed CTLA-4 has long been known as an important effector molecule of regulatory T-cell function as well as a self-limiting brake that counteracts expansion of activated conventional T-cells (54–56), dendritic cell expressed and secreted CTLA-4 acts as a constitutive homeostatic signal that prevents CD8⁺ T-cell priming by mature DC in the absence of T_H1 polarizing cues that curtail its secretion (36). Previous work demonstrated that CTLA-4 mRNA expression is substantially downregulated when dendritic cells are loaded with obligately identical internal and external antigenic environments (23), and further, that CTLA-4⁺ microvesicle secretion is significantly inhibited by the provision of such equivalent antigenic environments (i.e. loading with similar/identical class I and II antigens) (28). To determine if CTLA-4⁺ microvesicle secretion might also be used as a specific readout signal of homologous antigenic environments in murine DC, we first validated upregulation of the CTLA-4 mRNA transcript in DC derived from AIMp1 knockout animals by RT-PCR (29). As anticipated, genetic ablation of AIMp1 resulted in a 30–50 fold upregulation of DC CTLA-4 transcription under all in vitro conditions tested (representative experiment shown, Fig 2A), suggesting that AIMp1 negatively regulates CTLA-4 expression. In a manner analogous to the previous SIINFEKL/OVA experiments, we next loaded WT murine DC with identical or with disparate class I and II antigens and performed western blot analysis on supernatants after 48 hours of maturation. As shown in Fig 2B, differentially loaded DC secreted high levels of AIMp1 and low levels of CTLA4⁺ microvesicles into the supernatant when loaded with similar class I and II antigens; however, the opposite pattern of secretion, low AIMp1 and high CTLA-4, was observed when the loaded class I and II antigens were not the same. To demonstrate dependence of CTLA-4 secretion levels on AIMp1 expression, DC loaded with identical class I and II antigens were treated with either non-targeting (NT) or AIMp1 siRNA and the supernatants analyzed after 48 hours. As shown in Fig 2C, DC treated with NT siRNA exhibited the high AIMp1/low CTLA-4 supernatant secretion pattern of the previous experiment whereas DC treated with AIMp1 siRNA exhibited substantially elevated levels of CTLA4⁺ microvesicle secretion. Given that AIMp1 and CTLA-4 secretion move in opposite directions in response to DC loading with homologous class I and II antigens, this ratio of AIMp1 to CTLA-4 secretion could therefore serve as a powerful and specific signal by which to objectively assess TLR-independent T_H1 polarization of DC in response to loading with homologous class I and II antigens.

Dendritic cell recognition of homologous antigenic environments requires antigen loading onto MHC.

In previous experiments we noted that H2-DM^−/− DC could not mediate upregulated T_H1 responses in vivo after loading with SIINFEKL and OVA, indicating that inhibition of peptide binding to MHC class II was functionally relevant. Interestingly, H2-DM^−/− DC have been reported to display a T_H2 polarized phenotype that is physically dependent upon the presence of bound CLIP peptide (57). To determine the ability of H2DM^−/− DC to modulate secretion of AIMp1 and CTLA-4⁺ microvesicles in response to class I and II antigenic homology, we loaded both WT and H2-DM^−/− DC with class I determinant only, class II determinant only, homologous class I and II determinants, or heterologous class I and II determinants and assayed supernatant by western blot for AIMp1 then stripped and re-probed to assay for CTLA-4 secretion. To control for both cell death and the amount of supernatant loaded, membranes were stripped and re-probed a third time with anti-β-actin to detect trace actin in supernatant cell debris. As indicated in Fig 2D and four independent experiments quantitated in Fig 2E–F, H2-DM^−/− DC constitutively secreted low levels of AIMp1 and high levels of CTLA-4+ microvesicles, displaying no differential regulation in response to the homology of antigenic load. In contrast, WT DC secreted high levels of AIMp1 and low levels of CTLA-4⁺ microvesicles in response to loading with homologous class I and II antigens, indicating that regulation of these depends upon peptide binding to class II MHC and not to pattern recognition receptors.

Because CLIP is tightly bound within the MHC class II binding groove in the absence of the H2-DM chaperone in H-2^b haplotype mice (47, 48), one theoretical manner by which to stimulate antigen-mediated T_H1 polarization in H2-DM^−/− DC would be via loading of a class I binding peptide with significant homology to CLIP, i.e. with amino acid sequence overlapping the MHC class II bound-CLIP sequence LPKSAKPVSQMRMAT-PLLMRPMSM (58). To test this hypothesis, we designed a class I peptide predicted to bind H-2D^b (MRMATPLLM) that possessed full sequence overlap with CLIP. We then loaded H2-DM^−/− DC with either this CLIP-derived H-2D^b class I peptide or the well-established H-2K^b class I SIINFEKL peptide as a control. Upregulated AIMp1 expression and diminution of CLTA-4 secretion was subsequently observed in a dose-dependent fashion only from the cells loaded with H-2D^b CLIP but not SIINFEKL (Fig 2G). Accordingly, coculture of loaded H2-DM^−/− DC with wild type, syngeneic splenocytes in vitro led to an increase in CD8⁺CD25⁺ T cells only when H2-DM^−/− DC were loaded with H-2D^b CLIP but not SIINFEKL (Fig 2F). Enhancement of AIMp1/diminution of CTLA-4 secretion was well-correlated with the ability of H-2D^b CLIP to stabilize surface MHC in H-2^b TAP1deficient cells at the indicated peptide concentrations (Fig S3A/B), suggesting that this process was dependent upon binding of the H-2D^b CLIP peptide to MHC. (39) In these experiments H-2D^b CLIP stabilized surface MHC on B16 cells and C57BL/6 BMDC (both of haplotype H-2^b) in a fashion statistically identical to that of the control LCMV gp33–41 H-2 D^b binding peptide, indicating strong MHC binding. These data suggested high-level specificity of an intrinsic DC mechanistic process that compares the amino acid sequences of peptides concurrently bound to MHC class I and II.

To characterize the degree of peptide specificity required to alter the regulation of AIMp1 and CTLA-4, we moved from a mouse to a human system to take advantage of a validated and published system of human HLA-binding influenza HA (A/New Caledonia/20/99) peptide epitopes (Fig 3A) previously shown to generate T_H1 polarized immune responses when loaded on DC in homologous fashion (23, 37, 59). As with the murine SIINFEKL/OVA system, we validated that the AIMp1/CTLA-4 ratio was impacted only when DC were loaded with homologous peptide pairs and not heterologous pairs or TLR agonists. As indicated (Fig 3B/C), only loading of DC with homologous class I and II peptides resulted in significant alteration of the AIMp1/CTLA-4 ratio. In addition to western blot, this phenomenon could also be shown by intracellular flow cytometry (Fig S4) at 48 hours post-maturation. At this time point, CTLA-4⁺AIMp1⁺ intracellular positivity was primarily driven by retention of CTLA-4 that occurred concomitantly with decreased CTLA-4 secretion. A time course experiment confirmed that upregulation of AIMp1 was an early event, seen as early as three hours after peptide loading [and as reported previously (29)], whereas downregulated secretion/increased retention of CTLA-4 was a later event not observed until 24–48 hours after peptide loading (Fig S5). To mechanistically explore the parallel phenomena of decreased CTLA-4 secretion/increased CTLA-4 retention in more depth, DC were loaded with either homologous (overlapping) or heterologous (nonoverlapping) class I and II peptides and characterized by confocal microscopy at 24 and 48 hours post maturation. By 48 hours post-maturation, DC loaded with homologous peptides had greatly reduced their secretory capacity as indicated by significant downregulation of Rab5 [a marker of secretory vesicle exocytosis (60, 61)] in comparison to unloaded DC or DC loaded with heterologous peptides. Correspondingly, DC loaded with homologous peptides accumulated much more CTLA-4 internally than counterparts loaded with heterologous peptides and unloaded controls (Fig 4A–C). Confocal microscopy also allowed visualization and quantitation of CTLA-4⁺Rab5⁺ intracellular and budding vesicle export complexes. As shown (Fig 4D–E), at both 24 and 48 hours post-maturation these were still commonly visualized/enumerated among unloaded controls and DC loaded with heterologous peptides but became increasingly scarce among DC loaded with homologous peptides (^****p < 0.0001).

Figure 3. Antigenic homology but not TLR agonism modulates human DC CTLA-4 and AIMp1 secretion.

A. Schema depicting the sequence overlap of class I B8–166 influenza HA peptide with its homologous (DR3–162) and heterologous (DR3–440) counterparts. B. CTLA-4 and AIMp1 western blots of human DC culture supernatants following loading with homologous or heterologous class I and II peptide pairs and in the presence of various TLR agonistic stimuli. Representative experiment of three shown. C. Densitometry quantitation of AIMp1/CTLA-4 ratios of the data depicted in (B) over three independent experiments. As indicated, TLR agonism alone or in conjunction with heterologous class I and II peptide loading did not modulate the AIMp1/CTLA-4 ratio in the same fashion as loading with homologous class I and II peptides. Note background AIMp1/CTLA-4 expression ratio of unloaded DC on far left. Error bars = +/− SD. ^**p < 0.01 by one-way ANOVA with Bonferroni post-hoc.

Figure 4. DC loading with homologous class I and II peptide antigens downregulates CTLA-4⁺ vesicle exocytosis.

DC were loaded with homologous or heterologous peptide pairs and matured. Forty-eight hours later, cells were stained for CTLA-4 (green) and Rab5 (red), a marker of secretory vesicle exocytosis, and analyzed by confocal microscopy. A. As seen, DC loaded with homologous peptides exhibited significantly reduced secretory capacity as indicated by significant downregulation of Rab5 in comparison to unloaded DC or DC loaded with heterologous peptides. Correspondingly, DC loaded with homologous peptides accumulated much more CTLA-4 internally than counterparts loaded with B. heterologous peptides or C. unloaded controls. Confocal microscopy also allowed visualization and quantitation of CTLA-4⁺Rab5⁺ intracellular and budding vesicle export complexes. As shown in the representative field D. these were still commonly visualized/enumerated among unloaded controls and DC loaded with heterologous peptides but became increasingly scarce among DC loaded with homologous peptides E. Quantitation of this phenomenon at both 24 and 48 hours post-maturation among cells present in 20 randomly selected fields. ^****p < 0.001 by Pearson’s chi-squared.

Recognition of class I and II peptide homology requires a contiguous stretch of amino acid sequence identity.

To determine the degree of amino acid homology required to modulate the AIMp1/CTLA-4 ratio, we generated an additional series of class I and II binding peptides based upon the characterized influenza HA class I and class II peptides (i.e. Fig 3A) used in the previous experiments. By mutating the two non-anchor glycine (GLY) residues in the class I/II sequence overlap region to methionine (MET) residues, each mutated peptide maintained sequence homology at 6 of 8 residues in the overlap region with the original parent peptide yet no longer possessed a contiguous stretch of more than three amino acid residues when loaded in tandem with the corresponding glycine-containing non-cognate peptide. Full sequence overlap of the methionine mutant class I and II peptides could be restored if loaded together in cognate fashion rather than with the reciprocal but non-cognate glycine-containing counterpart. The diagram in Fig 5A illustrates the four ways that these class I and II peptides were combined with each other in subsequent DC loading experiments. As indicated by the representative experiment shown in Fig 5B and the compilation of 16 independent experiments shown in Fig 5C, loading of DC with the fully homologous class I and II peptides was sufficient to induce upregulation of AIMp1 expression and downregulation of CTLA-4 secretion; however, loading of a MET-substituted class II with a GLY-possessing class I peptide or loading of a MET-substituted class with a GLY-possessing class II peptide was sufficient to completely abrogate the AIMp1/CTLA-4 secretion pattern associated with T_H1 polarization. To validate that AIMp1 upregulation and CTLA-4 downregulation in this system could be well-correlated with traditional T_H1 output parameters, DC loaded with each peptide combination were also analyzed by RT-PCR for il12a and il12b transcript expression. As shown in Fig 5D and E, transcription of both il12a (^***p<0.005) and il12b (^****p<0.001) was substantially upregulated only when class I and II peptide homology was completely maintained (as reported previously(28)). Loading of DC with non-homologous peptides resulted in either no significant transcriptional change (of IL-12 p35 encoded by il12a) or transcriptional downregulation (of IL-12 p40 encoded by il12b, *p<0.05). Peptide loaded DC were also co-cultured with autologous PBMC in vitro. As shown in Fig S6, significant expansion of activated, IFN-γ-secreting T-cells was observed only when class I and II peptide homology was completely maintained (as reported previously in other antigenic systems (23, 27, 28, 35) whereas loading of DC with non-homologous peptides resulted in generation of very few activated IFN-γ-secreting T-cells.

Figure 5. Detection of class I and II peptide homology requires a contiguous stretch of amino acid sequence identity.

A. Schematic representation of the original and amino acid-substituted peptide pairs used in these experiments. DC loading with fully homologous class I and II peptides induced characteristic modulation of AIMp1 and CTLA-4; however, loading of a MET-substituted class II with a GLY-possessing class I or loading of a MET-substituted class I with a GLY-possessing class II completely abrogated these characteristic alterations. B. Representative experiment. C. Quantitation of 16 independent experiments. To validate the correlation of AIMp1/CTLA-4 secretion with traditional T_H1 output parameters, DC loaded with each peptide combination were also analyzed by RTPCR for D. il12a and E. il12b transcript expression. Transcription of both il12a ^(***p<0.005) and il12b (^****p<0.001) was substantially upregulated only when class I and II peptide homology was completely maintained. Loading of DC with non-homologous peptides resulted in either no significant transcriptional change (of IL-12 p35 encoded by il12a) or transcriptional downregulation (of IL-12 p40 encoded by il12b, *p<0.05). Normalized quantitation of three experiments shown. F/G. Treatment of DC with either β2microglobulin or HLA-DM siRNA was sufficient to abolish upregulation of the AIMp1/CTLA-4 ratio observed following loading with homologous peptide pairs, rendering it statistically identical to that observed following loading with heterologous peptides. F. Representative experiment. Note background AIMp1/CTLA-4 expression levels of unloaded DC on far right of each panel. G. Compilation of four independent experiments. Error bars = +/− SD. *p < 0.05 by Student’s two-tailed t-test. H. Transcription of the T_H1 lineage marker Cadm1 was significantly upregulated (^****p<0.001) only when only when DC-loaded class I and II peptide homology was completely maintained whereas transcription of the T_H2 lineage marker Sirpa was substantially upregulated (^****p<0.001) when class I and II peptide homology was not maintained. Normalized quantitation of four experiments shown. All comparisons by Student’s two-tailed t-test.

To verify that this phenomenon was dependent upon peptide binding to HLA, these experiments were repeated in conjunction with siRNA targeting of either β2-microglobulin to prevent MHC class I peptide loading in the ER or HLA-DM to prevent removal of Ii-CLIP and therefore loading of MHC class II with exogenous peptide. As indicated by the representative experiment shown in Fig 5E and the compilation of four independent experiments shown in Fig 5F/G, treatment of DC with either β2-microglobulin or HLA-DM siRNA was sufficient to abolish upregulation of the AIMp1/CTLA-4 ratio, rendering it statistically identical to that observed following loading with heterologous peptides. Taken together, these results suggest a physiologic mechanism that can identify and respond to stretches of amino acid identity no less than four residues in length that are shared between HLA class I and II-bound peptides.

Lineage characterization.

Recent elegant work has identified highly specific and phylogenetically conserved phenotypic signatures that define cDC1 (i.e. T_H1 polarizing) and cDC2 (i.e. T_H2 polarizing) functional dendritic cell subsets. Of Lin^neg HLA-DR⁺ CD11c⁺ cells, DC1 is defined as CADM1^high SIRPα^low whereas DC2 is defined as CADM1^low SIRPα^high (62). Interestingly, previously-published work reported significant downregulation of the SIRPα (also known as CD172a, SHPS1, and PTPNS1) mRNA transcript only among HLA-DR⁺CD11c⁺ human monocyte-derived DC that had been loaded with identical MHC class I and class II antigenic determinants (23). To validate that peptide sequence homology of class I and II antigens was sufficient to regulate cDC1/2 lineage marker expression, HLA-DR⁺CD11c⁺ monocyte-derived DC loaded with the homologous and nonhomologous peptide combinations outlined in Fig 5A were analyzed in quadruplicate by RT-PCR to characterize transcriptional expression levels of CADM1 and SIRPα. As indicated in Fig 5H, loading of DC with the homologous Gly/Gly or Met/Met class I/II peptide pairs resulted in significant upregulation of CADM1 and significant downregulation of SIRPα transcription, indicating the CADM1^high SIRPα^low cDC1 phenotype. In contrast, loading of DC with the non-homologous Gly/Met or Met/Gly class I/II peptide pairs resulted in significant downregulation of CADM1 and significant upregulation of SIRPα transcription, indicating the CADM1^low SIRPα^high cDC2 phenotype (^****p<0.0001 for both). These results indicated that loading of monocyte-derived DC with homologous or heterologous antigens was sufficient to impart the well-characterized lineage phenotypes of, respectively, physiologic cDC1 or cDC2 gene signatures.

Characteristics and protein-protein interactions of the multi-enzyme aminoacyl-tRNA synthetase (mARS) complex during DC homologous peptide loading.

Aminoacyl-tRNA synthetases possess an alluring molecular quality required by any putative sensor system capable of discerning sequence similarities between class I and II peptide epitopes: each of the 19 aminoacyl tRNA-synthetase isoforms can recognize and bind its cognate amino acid with exquisite selectivity and specificity (63). Interrelated with its role as a signaling mediator of T_H1 immune processes, AIMp1 also serves as a subunit component of a very large protein complex termed the multi-enzyme aminoacyl-tRNA synthetase (mARS) complex which, despite incorporation of many aminoacyl-tRNA synthetase isoforms, does not play a significant role in protein translation (64, 65). This molecular complex of over 2,000 kDa is comprised of multiple aminoacyl-tRNA synthetases arrayed in dimeric fashion and bound together by three core structural proteins termed AIMp1, 2, and 3. AIMp1 is released from this mARS complex and secreted in cytokine-fashion under conditions of cell stress (40, 66, 67). Recent elegant work demonstrated that multiple mARS subunit components are differentially phosphorylated and dissociate from the complex in response to VSV and influenzavirus infection. Once dissociated, these subunits bind and activate essential components of innate antiviral immunity, thereby implicating the mARS as both a sensor and regulator of the immune response to viral infection (68, 69). Canonically, the mARS complex is reported to possess only 8 of the 19 known mammalian aminoacyl-tRNA synthetase isoforms (glutamyl-prolyl, isoleucyl, leucyl, glutaminyl, methionyl, lysyl, arginyl, and aspartyl) (40, 64, 65, 68) which, if true under all homeostatic conditions, could limit its utility as a sensor capable of discerning amino acid homology between epitopes not possessing one or more of the cognate amino acid residues. To determine if the mARS complex might be able to acquire aminoacyl-tRNA isoforms other than the canonical eight, we immunoprecipitated the mARS complex using an AIMp2 monoclonal antibody in both immature and mature dendritic cells and performed mass spectroscopy on each immunoprecipitate. In immature DC, five complex-associated aminoacyl-tRNA synthetase isoforms were identified, and all five of these (glutamyl-prolyl, isoleucyl, aspartyl, glutaminyl, and lysyl) were one of the canonical eight reported previously. In mature DC however, the composition of the mARS complex was substantially different (*p<0.05). Of nine aminoacyl-tRNA synthetases identified, five were identical to those seen in immature DC; however, the other four identified (asparaginyl, alanyl, histidyl, and the most abundant tryptophanyl) had not previously been reported to associate with the complex (Fig 6A), indicating that complex-associated aminoacyl-tRNA isoform components can fluctuate under inflammatory conditions in DC. The ability to bind and/or colocalize with HLA would also be an essential characteristic of a sensor complex with the ability to scan HLA-bound peptide epitopes. As anticipated by this hypothesis, HLA was one of the most abundant proteins that co-precipitated with the mARS complex in both immature and mature DC. In immature DC (Fig 6B), only MHC class I HLA-B and HLAC proteins immunoprecipitated with the complex; however, in mature DC all three MHC class I proteins as well as MHC class II co-precipitated with the complex (^****p<0.0001). Interestingly associated MHC class II comprised both the α and β chains of HLA-DR; however, no HLA-DP or HLA-DQ chains were observed. These data indicated that the mARS complex under inflammatory conditions in DC binds MHC class I and MHC class II simultaneously as would be required of a hypothesized sensor complex.

Figure 6. Characteristics and protein-protein interactions of the multi-enzyme aminoacyl-tRNA synthetase (mARS) complex during DC homologous peptide loading.

A. Immunoprecipitation of mARS complex structural subunit AIMp2 and analysis by mass spectrometry indicated binding of only canonical aminoacyl-tRNA synthetase isoforms in immature DC; however, bound noncanonical isoforms were identified in matured DC. Blue - total number of unique tRNA synthetase isoforms detected. Red - total number of unique tRNA-synthetase peptide identifiers detected. B. Similarly, while the complex appeared to bind only MHC class I HLA-B and HLA-C in immature DC, it was shown to bind all three HLA proteins as well as MHC class II (HLA-DR) in matured DC. C. AIMp1 IP followed by IB with anti-HLA-A/B/C and anti-HLA-DR indicated mARS binding to MHC class I and II under conditions of heterologous peptide loading that was substantially abolished when DC were loaded with homologous peptides, suggesting release of AIMp1 from HLA-bound complex under such conditions. Representative experiment shown. D. Loading of DC with the same class I and II peptide combinations depicted in Fig 5A followed by mARS complex IP with anti-AIMp2 demonstrated preferential accumulation of aminoacyl-tRNA synthetase isoforms recognizing cognate amino acid residues located within those peptides only when loaded class I and II peptides were homologous. E. A similar experiment performed with a second set of peptides (schema and colorcoded amino acids shown) yielded similar results. Representative of each shown. *p < 0.05, ^****p < 0.001 by Pearson’s chi-squared.

Knowing that the mARS complex binds HLA under inflammatory conditions and knowing that mARS structural component AIMp1 is released from the complex in response to a variety of stimuli (40, 66, 67) and is also indispensable for promulgation of T_H1 immunity (29), we next asked the question as to whether or not AIMp1 could be released from the mARS complex in response to DC loading with homologous class I and II peptides. To perform this experiment, DC were loaded with homologous or heterologous combinations of the same influenza HA binding peptides first described in Fig 5A. As previously, these pairs of class I and II peptides were identical to each other with the exception that two nonanchor glycine residues in the homology overlap region were mutated to methionine residues. As shown in Fig 6C, immunoprecipitation of the complex with an AIMp1 antibody co-precipitated both MHC class I and MHC class II under all conditions of heterologous peptide loading; however, when DC were loaded with homologous peptides, almost no MHC co-precipitated with AIMp1, indicating release of AIMp1 from HLA-bound mARS complexes under these conditions.

Given that the content and identity of aminoacyl-tRNA isoforms in the mARS complex appears malleable in DC, we sought to characterize any correlations between aminoacyl-tRNA synthetase isoform identity in the complex and identity of amino acid residues in the homology overlap region of HLA-bound class I and class II peptide epitopes. Continuing with the peptide system outlined in Fig 5A, DC were loaded with each of the four indicated class I and II peptide combinations, and the mARS complex was subsequently immunoprecipitated with an antibody against AIMp2, a core mARS complex structural component that does not dissociate. We then immunoblotted with anti-glycyl- and anti-methionyl-tRNA synthetase antibodies given that the glycine-substituted methionine residues were all that differentiated the peptides from each other. As shown in Fig 6D, only when both the class I and the class II peptide possessed either the glycine or the methionine residues did the corresponding aminoacyl-tRNA synthetase isoform accumulate within the mARS complex. Substantial accumulation of glycyl-tRNA synthetase within the complex was only seen when glycine residues were present in the homology overlap region of both the class I and class II peptides but not when present on only one or the other. Similarly, substantial accumulation of methionyl-tRNA synthetase within the complex was only observed when methionine residues were present in the homology overlap region of both the class I and II peptides but not when present on only one or the other. Because this experiment is highly critical to the understanding of mechanism, additional experimental repetitions are shown in Fig S7A/B with the uncut blots shown in Fig S8A–E. To explore this phenomenon further, we next loaded DC with either the homologous or the heterologous peptide pairs introduced in Fig 3A (also reproduced in Fig 6E) and again immunoprecipitated the complex with anti-AIMp2. Once precipitated, the immunoprecipitate was analyzed for content of lysyl- and glycyl-tRNA synthetases, two aminoacyl-tRNA synthetase isoforms that corresponded to cognate amino acid residues within the class I and II homology overlap region. The precipitate was also analyzed for content of the bispecific glutamyl/prolyl-tRNA synthetase isoform, corresponding to cognate glutamic acid and proline residues located outside the class I and II homology overlap region. As shown in Fig 6E, lysyl- and glycyl-tRNA synthetases accumulated preferentially within the complex only when DC were loaded with homologous class I and II peptides. Preferential accumulation of these isoforms did not occur following loading with a heterologous class II peptide, despite the presence of the cognate amino acid residues on the class I peptide. Further, preferential accumulation of glutamyl/prolyl-tRNA synthetase did not occur following loading of either peptide combination despite the presence of a proline residue located outside the homology overlap region on the homologous class II peptide as well as three glutamic acid residues in the heterologous class II peptide. These results strongly suggested that the modulation of aminoacyl-tRNA synthetase isoforms within the mARS complex is related to the identity of cognate amino acid residues located within homology overlap regions of MHC-bound class I and class II peptide epitopes.

While the specific compartment in which class I and class II antigenic comparison might occur is not known, there exist dendritic cell intracellular compartments in which loaded class I and class II colocalize, most prominently the late endosome (70–72). Because late endosomes give rise to the exosomal bodies that are secreted from dendritic cells (72–74), the contents of the late endosome may be indirectly sampled by exosome analysis. To determine if the mARS complex is present in the late endosome, exosomal bodies were isolated and analyzed for mARS structural subunits AIMp1 and AIMp3. As indicated in the analysis of four different independent preparations (Fig S9A), the mARS structural subunits were abundantly present in DC exosomes. To determine if any tRNA synthetases found in exosomes were bound in a very large complex, we performed serial UV crosslinking with subsequent analysis by western blot. Because the very large (several million kDa) mARS complex is too large to enter a standard 12% SDS-PAGE acrylamide gel, tRNA-synthetases bound in a large complex should disappear in a dose-dependent fashion following UV cross-linking, the result that was observed (Fig S9B). The data suggest an opportunity for the mARS complex to surveil loaded MHC class I and II in the late endosomal compartment.

Discussion

TLR and other PRR ligands (PAMPs) have long been understood to provide key environmental cues that influence the direction of T_H polarization while simultaneously signaling the presence of danger (6, 8, 11–20, 32, 33). Previously-characterized PAMP ligands are conserved structures so fundamental to the virulence, survival, or reproduction of a pathogen as to render impractical or impossible the evolution of variants unable to be recognized by fixed, genome-encoded receptors. Here we describe the existence of a different type of PAMP that is not a fixed molecular structure but rather an antigenic pattern consisting of bound MHC class I and II epitopes that overlap in amino acid sequence specificity. Detection of this antigenic pattern in isolation does not serve as a danger signal in and of itself, yet when detected in conjunction with other inflammatory cues induces a robust T_H1 polarization. The physiologic significance of this antigenic pattern may be hypothesized both by the types of immune responses it induces (23, 29, 36) as well as the homeostatic circumstances under which it naturally occurs in vivo (75, 76), with both these lines of evidence suggesting that surveillance for this pattern provides an additional level of immune governance over the antiviral immune response (Fig 7 schematic). It was important that these confirmatory experiments were carried out in the absence of viruses or viral vectors given the many potential TLR, NOD, and RLH pattern ligands that are experimentally introduced by the use of these.

Figure 7. Schematic representation of regulation by MHC peptide epitope homology.

In the presence of fulminant viral infection, bound MHC class I and II peptide epitopes among infected dendritic cells become increasingly likely to overlap in amino acid sequence identity. Identification of MHC class I and II bound peptide sequence overlap by the mARS complex results in the release of AIMp1, a direct mediator of enhanced downstream T_H1 polarizing responses.

While previous work indicated that the physiologic loading of DC with homologous class I and II antigenic determinants could induce T_H1 polarization and an augmentation of downstream CD8⁺ effector responses in vitro and in vivo (23, 27–29), the present work definitively identified amino acid sequence homology of bound MHC class I and II peptide epitopes as the signal responsible for this phenomenon in experimental systems devoid of potential TLR ligands or other immunomodulators. It was further demonstrated that class I and II epitope homology drives upregulated expression and secretion of AIMp1, a gene product known to be absolutely required for T_H1 polarizing function and secretion of IL12 (29), as well as downregulated secretion of DC-expressed CTLA-4⁺ microvesicles previously shown to inhibit CD8⁺ T-cell priming (36). TLR agonism was unable to alter AIMp1 and CTLA-4 expression/secretion, nor was loading of DC with identically prepared heterologous peptide pairs able to modulate AIMp1 and CTLA-4. siRNA knockdown of β2-microglubulin or HLA-DM eliminated the ability of DC to respond to homologous class I and II binding peptides. Further, murine DC lacking H2-DM and thus unable to load exogenous antigen were also unable to mediate T_H1 polarization when loaded with homologous peptide pairs though MHC was intact and pattern recognition receptors remained functional (49). However, loading of H2-DM^−/− DC with a synthetic H-2D^b binding peptide that overlapped the amino acid sequence of class II-bound Ii CLIP permitted homologous loading of DC that restored modulation of AIMp1 and CTLA-4 and produced activated CD8⁺ T-cells in a dose responsive fashion. To further demonstrate T_H1 polarization dependent upon antigenic homology rather than innate PRR, we substituted two non-anchor amino acid residues in a previously characterized (23) class II binding peptide so that contiguous class I and II sequence homology of more than three amino acids was interrupted. This minor disruption of homology was sufficient to abrogate polarizing AIMp1/CTLA-4 ratios. Applying the complimentary amino acid substitutions to the class I peptide - thereby restoring complete sequence homology - was sufficient to recapitulate the high AIMp1/CTLA-4 ratio associated with T_H1 polarization. Taken together, the data suggest a novel T_H1 polarizing checkpoint in DC dependent upon a high degree of sequence homology between MHC class I and II-bound peptides.

AIMp1 is a structural component of the mARS, a large protein complex comprised of many different aminoacyl-tRNA synthetase isoforms arrayed in dimeric fashion and previously implicated in the transduction of antiviral immunity (40, 64–68). Given that this complex possesses alluring molecular qualities required of a sensor able to discern amino acid sequence similarities between peptide epitopes, we characterized protein-protein interactions and aminoacyl-tRNA synthetase isoform composition of the complex following DC loading with either homologous or heterologous class I and II peptide pairs. While these characterizations did not a priori establish the mARS complex as a sensor capable of discerning amino acid sequence similarities between HLA-bound class I and II peptide epitopes, the experiments performed provided good evidence in favor of this hypothesis by demonstrating a) that the complex binds to both class I and II HLA under inflammatory conditions, b) that the complex releases the critical T_H1 regulator AIMp1 in response to DC loading with homologous but not heterologous class I and II peptide pairs, and c) that the aminoacyl-tRNA synthetase composition of the mARS complex in DC are directly influenced by the identity of cognate amino acids on loaded peptide epitopes but only when those amino acid residues are located within a region of sequence homology present on both the class I and the class II HLA epitope. While the length of sequence homology required for recognition of homology was not directly determined, the present work established that three consecutive amino acid residues were not sufficient whereas previous work (23) indicated that five consecutive amino acid resides are sufficient.

The phenomenon of DC cross-presentation might be perceived as a potential confounder of the work presented here. While a free flow of antigens between class I and class II compartments among all DC subsets would indeed hamper the ability to interpret our experimental results, cross-presentation does not proceed in such an undefined manner. Cross-presentation is most efficient and best characterized among cDC1 subsets defined as CD8α⁺ or CD103⁺ in mice and CD141⁺Clec9A⁺ in humans; however, cross-presentation among cDC2, pDC, monocyte-derived DC, or among the in vitro-derived BMDC used in this study is more controversial, with a plurality of sources suggesting that if these subsets cross-present, they do so only under defined circumstances [e.g. provision of BMDC with TLR-3 or TLR-9 ligands (50)] or in experimental systems that rely upon supraphysiologic concentrations of antigen or following provision of apoptotic signals (77–79). Moreover, the intracellular compartments from which antigens are cross-presented are typically characterized as early endosomes whereas antigen targeted to late endosomes is not cross-presented (70). Lastly, in the process of antigenic comparison, there appears to be a clear threshold effect (i.e. Fig 2G) required for significant release of AIMp1 to occur, and we speculate that the level of antigen transferred during cross-presentation is typically insufficient to meet this threshold. Hence, despite efficient cross-presentation of antigen by certain DC subsets under defined conditions, the division of labor that exists between DC subsets and among the intracellular compartments of cross-presenting subsets permits the identities of class I and class II epitopes to provide meaningful information regarding the content of internal and external antigenic environments under many circumstances, including the experimental circumstances by which the present work was performed. Nonetheless, the manner by which the processes of antigenic comparison and cross-presentation are integrated within the broader context of DC immunobiology is a clear area of focus for which future studies will be highly instructive.

Related to the potentially-confounding issue of cross-presentation, the cellular compartment in which comparison of class I and class II epitopes occurs has yet to be identified. The canonical representation of loaded MHC trafficking to the cell surface suggests MHC class I and MHC class II traffic separately in endosomal bodies derived from the Golgi (class I) and the phagolysosome (class II). Nevertheless, a variety of different studies have shown that MHC class I and class II extensively colocalize in internal cellular compartments. Most prominently, Kleijmeer et al convincingly demonstrated abundant colocalization of MHC class I with MHC class II by double immuno-gold labeling in both early endosomes and multi-vesicular late endosomes, with approximately 10% of total cellular MHC class I localized in MHC class II compartments (72). At least some of this colocalization has been shown to result from the fusion of endosomal and lysosomal compartments, providing ample opportunity for loaded class I and class II to comingle (71). We speculate that the late endosome is the more likely compartment in which antigenic comparison might occur given that it is not thought to be a significant cross-presentation compartment (70). Therefore loaded MHC class I and II in the late endosome may more faithfully represent the actual internal and external antigenic environments. Accordingly, sampling of the late endosomal contents through analysis of extracellular exosomes indicated an abundant presence of mARS structural subunit and tRNA-synthetase components. Nonetheless, additional studies will be required to definitively demonstrate that antigenic comparison occurs in this compartment.

In summary, here we have demonstrated the existence of a variable antigenic PAMP, the detection of which induces T_H1 polarization. The use of experimental systems devoid of traditional TLR ligand PAMPs as well as novel readouts not responsive to the presence of TLR ligands or other pattern-based PAMPs were essential in establishing the validity of this wholly novel concept. The data indicate that provision of DC with homologous class I and II peptide epitopes, corresponding to equivalent internal and external antigenic environments, drives T_H1 polarization when provided in conjunction with neutral inflammatory cues. The precise manner by which this novel signal is detected was not fully established; however, substantial evidence was provided that the multienzyme aminoacyl-tRNA synthetase (mARS) complex plays a central role in detection. Future studies will be aimed at the full elucidation of this remarkable process.

Non-Standard Abbreviations

AIMp

Aminoacyl-tRNA synthetase complex interacting multifunctional protein

ANOVA

Analysis of variance

APC

Antigen presenting cell

BMDC

Bone marrow dendritic cells

BSA

Bovine serum albumin

CADM1

Cell adhesion molecule 1

CD

Cluster of differentiation

cDC

Conventional dendritic cell

cGAS

Cyclic GMP-AMP synthase

Clec9A

C-type lectin domain containing 9A

CLIP

Class II-associated invariant chain peptide

CpG

Cytosine-phosphate-guanine

CTLA-4

Cytotoxic T-lymphocyte-associated protein 4

DC

Dendritic cell

DMSO

Dimethyl sulfoxide

ER

Endoplasmic reticulum

FAM

6-carboxyfluorescein

FITC

Fluorescein isothiocyanate

gp

Glycoprotein

H-2

Histocompatibility 2

HA

Hemagglutinin

HLA

Human leukocyte antigen

HSD

Honestly significant difference

IFN

Interferon

Ig

Immunoglobulin

Ii

Invariant chain

IL

Interleukin

LCMV

Lymphocytic choriomeningitis virus

LGP2

Laboratory of genetics and physiology 2

mARS

Multi-enzyme aminoacyl-tRNA synthetase

MDA5

Melanoma differentiation-associated protein 5

MHC

Major histocompatibility complex

moDC

Monocyte-derived dendritic cell

NLR

Nod-like receptor

NT

Non-targeting

ODN

Oligodinucleotide

OVA

Ovalbumin

PAGE

Polyacrylamide gel electrophoresis

PAMP

Pathogen-associated molecular pattern

PBMC

Peripheral blood mononuclear cells

PBS

Phosphate buffered saline

PCR

Polymerase chain reaction

poly(I:C)

Polyinosinic:polycytidylic acid

PRR

Pattern recognition receptor

Rab5

Ras-related protein 5

RIG-I

Retinoic acid-inducible gene I

RLH

RIG-like helicase receptor

rRNA

Ribosomal RNA

RT-PCR

Reverse transcription polymerase chain reaction

SD

Standard deviation

SDS

Sodium dodecyl sulfate

siRNA

Small interfering ribonucleic acid

SIRPα

Signal regulatory protein alpha

STING

Simulator of interferon genes

TBS-T

Tris buffered saline-tween 20

T_H

T-helper

TLR

Toll-like receptor

TNF

Tumor necrosis factor

tRNA

Transfer ribonucleic acid

VIC

Victoria

VSV

Vesicular stomatitis virus

WB

Western blot

WT

Wildtype