A diverse spectrum of unique peptide-MHC class I complexes guides CD8 T cell responses toward viral or stress-induced Ags. Multiple components are required to process Ag and facilitate peptide loading in the endoplasmic reticulum. IFN-γ, a potent proinflammatory cytokine, markedly up-regulates transcription of genes involved in MHC class I assembly. Physiological mechanisms which counteract this response are poorly defined. We demonstrate that promoters of functionally linked genes on this pathway contain conserved regulatory elements that allow antagonistic regulation by IFN-γ and the transcription factor B lymphocyte-induced maturation protein-1 (also known as PR domain-containing 1, with ZNF domain (PRDM1)). Repression of ERAP1, TAPASIN, MECL1, and LMP7 by PRDM1 results in failure to up-regulate surface MHC class I in response to IFN-γ in human cell lines. Using the sea urchin prdm1 ortholog, we demonstrate that the capacity of PRDM1 to repress the IFN response of such promoters is evolutionarily ancient and that dependence on the precise IFN regulatory factor element sequence is highly conserved. This indicates that the functional interaction between PRDM1 and IFN-regulated pathways antedates the evolution of the adaptive immune system and the MHC, and identifies a unique role for PRDM1 as a key regulator of Ag presentation by MHC class I.
The MHC class I system is first evident early in the evolution of jawed vertebrates at the inception of the “adaptive” or “anticipatory” type immune response (1). The MHC class I Ag-processing and -presentation pathways co-opted and adapted existing cellular machinery to allow the sampling of proteins during normal and altered cellular conditions and subsequent presentation to CD8-restricted T cells (2). The immune system has evolved the ability to change the pattern of peptide presentation in response to inflammatory cues by adapting the cellular machinery engaged in the generation of the peptide repertoire. For example, in the context of viral infection or inflammation, IFN-γ modulates not only the transcription of the MHC class I genes, but also multiple components of the Ag-processing and -loading pathways. IFN-γ enhances the generation of peptides of an appropriate length for presentation by inducing the expression of alternate proteasome components, large multifunctional peptidase 2 (LMP2)3, LMP7, and multicatalytic endopeptidase complex subunit-1 (MECL1), to generate the immunoproteasome (3). It acts to facilitate the transport of such peptides into the endoplasmic reticulum (ER), where the MHC class I peptide complex is assembled, by inducing the expression of the peptide transporters TAP1 and TAP2 (4, 5). It enhances the expression of the TAP-associated chaperone protein TAPASIN which facilitates peptide loading into the MHC class I-binding groove (6, 7) and ER amino peptidase 1 (ERAP1) (8, 9), which is responsible for trimming peptides to fit the groove, a function vital to the generation of the mature peptide repertoire (10).
The cumulative effects of IFN-γ therefore provide the cell with both qualitatively different peptides and a quantitative increase in surface MHC expression with which to elicit T cell activation. Changes in the peptide repertoire can lead to inappropriate destructive immune responses and contribute to the initiation of autoimmune disease (11). The response to IFN-γ provides a paradigm for the transcriptional control of MHC class I-dependent Ag presentation (3). Although the mechanisms for activation have been extensively studied, the endogenous factors mediating opposing or repressive effects on IFN-γ-induced transcription to maintain normal levels of MHC class I expression and unaltered peptide repertoires are ill-defined. Nevertheless, the tight control inherent in other aspects of the immune response suggests that these are likely to exist.
B lymphocyte-induced maturation protein-1 (BLIMP-1), which is also known as PR domain-containing 1, with ZNF domain (PRDM1), is an evolutionarily conserved transcriptional repressor of the Krüppel family of zinc finger proteins (12, 13, 14, 15), which acts both through direct competition for promoter occupancy, and by recruiting epigenetic modifiers (16, 17, 18, 19, 20). It is best known as a regulator of terminal B cell differentiation (21, 22), but additional roles for PRDM1 have also been identified in the control of T cell (23, 24), macrophage (25), and sebaceous gland differentiation (26). The original identification of PRDM1 as a postinduction repressor of the IFNβ promoter during cellular viral infection suggested that PRDM1 is a key regulator of cellular responses to IFNs (12). The optimum PRDM1 DNA-binding sequence overlaps with the IFN regulatory factor-element (IRF-E) (27, 28), and we recently showed that PRDM1 has the ability to regulate an IFN-γ-responsive promoter by competing with IRFs for occupancy of an IRF-E (19). Given these characteristics, PRDM1 is a candidate to act as a transcriptional repressor controlling MHC class I-dependent Ag-presentation pathways.
Only a subset of IRF-E sequences, among all IFN-γ-responsive gene regulatory elements, are potential targets for PRDM1-dependent repression (19, 28). In this study, we demonstrate that the promoters of multiple linked genes on the MHC class I Ag-processing and -loading pathway contain conserved IRF-E sequences which allow competitive binding by PRDM1 and IRFs. PRDM1 represses the transcription of these genes and thus negatively controls MHC class I expression in response to IFN-γ. This reveals a unique function for PRDM1 as a repressor of MHC class I-dependent Ag presentation.
Materials and Methods
Cell lines, expression vectors, and transfections
U266, H929, and HeLa cell lines were maintained in RPMI 1640 containing 10% heat-inactivated FCS and COS cells were grown in DMEM containing 10% FCS. HeLa or COS cells were seeded to reach 60% confluency on the day of transfection. Derivatives of bicistronic expression vector pIRES2-EGFP (BD Clontech) encoding full-length (FL) PRDM1, the 528–789 C-terminal amino acids (Δ527) of PRDM1, IRF-1 and IRF-2, have been described previously (19). Transfection of HeLa and COS cells was performed with GeneJuice (Novagen) reagent according to the manufacturer’s instructions. For RNA-interference experiments, U266 cells were transfected with 400 pM short interfering RNA (siRNA) (200 pM of each PRDM1 oligo or 400 pM control oligo) by electroporation using a Bio-Rad Gene Pulser with settings of 250 V and 960 μF. Twenty-four hours posttransfection, cells were processed for protein and mRNA evaluation. The oligos used contained the following sequences: PRDM1 siRNA 1: sense sequence, 5′-GGAAAGGACCUCUACCGUU-3′; PRDM1 siRNA 2: sense sequence, 5′-GAUCUGACCCGAAUCAAUG-3′; control siRNA: sense sequence, 5′-CUACCUCUAGAACGGACGU-3′.
Luciferase vectors and assays
The following primers were used to amplify promoter sequences from human genomic DNA: ERAP1 forward (F) (5′-TATAAGATCTGGATCCGCGTTCAGAAAGG-3′), ERAP1 reverse (R) (5′-AATTAAGCTTCTCACCCTTGCGCCG-3′); TAPASIN F (5′-TCTAAAGCTTAAGGATGCGCTCTTTATTTC-3′), TAPASIN R (5′-AAAACCATGGCGCTGCGACCTC-3′); MECL1 F (5′-TATAAGATCTCCTGAACAAGTCCAGAA-3′), MECL1 R (5′-AATTAAGCTTAGGCAGAGGGGATTAGG-3′); LMP7 F (5′-TATAGGATCCGTACCTCTTACTGTAACC-3′), LMP7 R (5′-AATTAAGCTTATGACCGCCCAGCACCCA-3′).
The amplicons were then cloned into the vector pXPG (29) generating promoter constructs spanning −248 to +61 for ERAP1, −377 to +174 for TAPASIN, −203 to +93 for MECL1, and −239 to +45 for LMP7, relative to the transcriptional start sites of reference sequences indicated in Table I⇓. The TAPASIN promoter plasmid was cut with SmaI and religated, to contain the minimal IFN-responsive elements (30, 31
All wild-type and mutant constructs were sequence verified. For luciferase assays, three replicate transfections were performed for each condition. Experiments were done using the Promega luciferase assay system and analyzed on a Berthold Lumat LB Luminometer. Each experiment was performed in duplicate with similar results and confirmed with separate plasmid preparations. The data are from a representative experiment displayed as fold increase in light units relative to unstimulated cells cotransfected with the empty vector.
Rabbit antisera to PRDM1 have been described previously (19, 32). Rabbit polyclonal Abs to IRF-1 and IRF-2 were obtained from Santa Cruz Biotechnology. Nonimmune rabbit IgG was obtained from Upstate Biotechnology. Abs used for flow cytometry were mouse IgG2a conjugated to PE (BD Biosciences) and monoclonal anti-human HLA-ABC conjugated to PE (clone W6/32; DakoCytomation). Mouse mAb to TAPASIN, clone 16, was obtained from BD Biosciences, goat polyclonal Ab to human keratin (GTX28572) was obtained from GeneTex, and mouse mAb to β-actin, clone AC15, was obtained from Sigma-Aldrich. Secondary Abs for immunofluorescence were Alexa Fluor 488 donkey anti-mouse, Alexa Fluor 594 donkey anti-rabbit, and biotinylated donkey anti-goat coupled with AMCA-Avidin (Vector Laboratories) for blue fluorescence.
Nuclear extracts were prepared as previously described (33) from COS cells transfected with expression vectors for PRDM1, IRF-1, or IRF-2. For EMSA, the double-stranded probes used contained the following sequences: ERAP F (5′-AGGACCGAAAGTGAAAGTGGAGCCCGGGGA-3′), ERAP R (5′-TCCCCGGGCTCCACTTTCACTTTCGGTCCTG-3′); TAPASIN F (5′-TTTGGAGGAAAGTGAAAGTGAAAGGAGGAAG-3′), TAPASIN R (5′-CTTCCTCCTTTCACTTTCACTTTCCTCCAAA-3′); MECL1 F (5′-GAAGGGTAAAGGCGAAAGCGAAAGCAGGAAG-3′), MECL1 R (5′-CTTCCTGCTTTCGCTTTCGCCTTTACCCTTC-3′); LMP7 F (5′-CGGAGGAGGAAGTGAAAGCGAAAGCCACAGA-3′), LMP7 R (5′-TCTGTGGCTTTCGCTTTCACT TCCTCCTCCG-3′).
The underlined bases represent sites in which position 5 and 11 of the IRF-E were altered as described in Fig. 3⇓. DNA probes, [32P]-labeled with T4 polynucleotide kinase, were incubated with nuclear extract in the presence of poly(dI:dC) (Amersham) for 30 min at room temperature. Supershift was performed by the addition of antisera to the extract before mixing with radioactive probe and competition assays included the addition of unlabeled probe to the reaction mixture.
GAPDH and ACTIN. For quantitative RT-PCR, the following primers were used: GAPDH F (5′-AACAGCGACACCCACTCCTC-3′), GAPDH R (5′-CATACCAGGAAATGAGCTTGACAA-3′); ACTIN F (5′-CATCGAGCACGGCATCGTCA-3′), ACTIN R (5′-AGCACAGCCTGGATAGCAAC-3′); ERAP1 F (5′-TCACCAGCAAATCCGACATG-3′), ERAP1 R (5′-CCCACATTAAATTTGATCCATTCC-3′); TAPASIN F (5′-CAAGGATTCAAAGAAGAAAGCAGAGT-3′), TAPASIN R (5′-GGAGAGAGATTGGAGGGATTAGG-3′); MECL1 F (5′-TGTGGACGCATGTGTGATCA-3′), MECL1 R (5′-GGTTCCAGGCACAAAGTGGTA-3′); LMP7 F (5′-GGAGTGATTGCAGCAGTGGAT-3′), LMP7 R (5′-TGCCAAGCA GGTAAGGGTTAA-3′).
The primers used to PCR HLA-A, B, C have been published previously (34). The ACTIN primers were F (5′-AGAAAATCTGGCACCACACC-3′) and R (5′-CTCCTTAATGTCACGCACGA-3′).
Chromatin immunoprecipitation (ChIP)
Chromatin prepared from myeloma or transfected HeLa cells was precleared with BSA saturated protein A-Sepharose followed by precipitation with 2 μg of Ab to PRDM1, IRF-1, IRF-2, or control rabbit Ab and protein A-Sepharose. DNA was eluted and cross-links reversed by overnight incubation at 65°C. Input DNA was prepared from an equal volume of chromatin following RNase treatment and resuspended in the same final volume as the ChIP samples. A standard curve of input DNA was generated from each chromatin sample. Target sequences were analyzed by real-time PCR as described above. PCRs were monitored by melting curve analysis and representative products were verified by sequencing. The amount of precipitated material was calculated from the standard curve and normalized relative to control immunoprecipitates. The data are representative of a minimum of three independent chromatin preparations, and represent average and SD of at least two independent immunoprecipitations. The following primers were used, with positions shown relative to reference sequences indicated in Table I⇑: ERAP1–240 F (5′-GGATCCGCGTTCAGAAAGG-3′), ERAP1–137 R (5′-CCAGGAAGGGAATTGGTAAATG-3′); TAPASIN + 86 F (5′-CCAGGCACCTTCACCTAACC-3′), TAPASIN + 187 R (5′-CAGCCATGAAGCCTCCTCTT-3′); MECL1–135 F (5′-GGGCACAGCAAGGGACAT-3′), MECL1–46 R (5′-GTGGCGGTTTTCTGCATCTT-3′); LMP7 + 123 F (5′-GCTCGGACCCAGGACACTAC-3′), LMP7 + 202 R (5′-TACTGCCCCGACCTGCAT-3′); PA28α-6 F (5′-ACTACCCAGGAAGGCGGAG-3′), PA28α+77 R (5′-CGCACAAGGAGTGGAGTGG-3′); PA28β+202 F (5′-CGCCACTGAATACCCCCTTT-3′), PA28β+285 R (5′-GGCTTATAGCTAGGGCCAACTG-3′); TAP2–245 F (5′-CAGATAAAGTTGCCCTTGAGACAA-3′), TAP2–123 R (5′-CACTGTACAGGCCTGCAATGA-3′); KRT-10 F (5′-TGGACACACCCTCTCAGTATATAAAGG-3′), KRT-10 R (5′-AGAGTAGTGCTTGCTTGAGCTGTATC-3′).
Flow cytometry and immunofluorescence
Evaluation of MHC class I surface expression was performed on HeLa cells 48 h posttransfection. Cells were cultured in the presence or absence of IFN-γ for the final 24 h. HeLa cells were stained with PE-conjugated isotype control or PE-conjugated Ab to pan MHC class I. Samples were analyzed on a LSRII flow cytometer (BD Biosciences) with settings determined by nontransfected cells and isotype staining. Viable cells were distinguished based on forward scatter and side scatter characteristics and propidium iodide staining. Plots were generated using WinMDI software. The top 15% of enhanced GFP (EGFP) expressing transfected cells were sorted.
Immunofluorescent staining was performed on human tonsil sections after heat-mediated Ag retrieval. The staining protocol was performed essentially as previously described (35). Sections were viewed using a Zeiss AxioPlan2 imaging fluorescence microscope. Images were captured and processed using the ISIS3 image capture system (MetaSystems). The use of human tissue was approved by the local research ethics committee.
An overlapping PRDM1/IRF-E-binding site distinguishes the promoters of MHC class I Ag-processing and -presentation machinery genes
Our laboratory has previously reported that PRDM1 represses IFN-γ-mediated activation of the MHC CIITA (CIITA) promoter IV and competes with the IFN-γ-induced transcription factors IRF-1 and IRF-2 for occupancy of the IRF-E (19). The overlap between the optimum PRDM1 consensus and the IRF-E suggests that other target genes may be under similar antagonistic regulation (28). To establish whether IFN-γ-regulated components of the MHC class I pathway were likely to be PRDM1 targets, we examined the basal promoter sequences of these genes (4, 30, 36, 37, 38, 39, 40, 41) for evidence of an overlapping PRDM1/IRF-E consensus (28). Remarkably, we discovered that all eight promoters examined contain IRF-E sequences which match the PRDM1 consensus-binding site with at most a single base mismatch (Table I⇑). Moreover, only one of three mismatches lies within the extended core sequence of the PRDM1 site, G(T/C)GAAAG(T/C)(G/T), and in each case the presence of an overlapping PRDM1/IRF-E consensus in the basal promoter is evolutionarily conserved.
This degree of overlap between PRDM1 consensus and IRF-controlled sites is not a common feature of IFN-γ- or IRF-regulated genes. In a published comparison of IRF-responsive elements, other than the known PRDM1 site within the IFNβ promoter (12) only 1 of the other 32 sites listed displays this degree of overlap with the PRDM1 consensus (42). Particularly important examples in this context are the sequences present in the promoters of β2-microglobulin and the MHC class I genes themselves (Table I⇑). These differ from the PRDM1 consensus at two or more positions within the extended core, which include critical residues necessary for efficient PRDM1 binding (12, 19).
PRDM1 binds to the promoters of MHC class I-processing and -presentation machinery genes in vivo
The results above demonstrate the presence of a PRDM1/IRF-E consensus in the promoters of the genes encoding multiple components of the MHC class I-processing and -presentation machinery. To evaluate promoter occupancy in vivo, we used ChIP to examine the association between PRDM1 and these promoters in myeloma cell lines U266 and H929. Constitutive PRDM1 binding to the promoters of the identified target genes ERAP1, TAPASIN, MECL1, and LMP7 was observed (Fig. 1⇓A and data not shown). In contrast, PA28β was weakly bound, while the promoter sequences of PA28α and TAP2 were enriched <2-fold relative to control. Therefore, PRDM1 is probably not bound to all available consensus sequences.
We previously found that U266 and H929 myeloma cells also express IRF-1 and IRF-2 (19). To test for occupancy by these transcription factors, we performed additional ChIP experiments. A mutually exclusive pattern of promoter occupancy was previously described at CIITA-pIV and the bidirectional TAP1/LMP2 promoter (19). In contrast, we observed that ERAP1, TAPASIN, MECL1, and LMP7 promoters all showed substantial binding by IRF-1 and IRF-2, as well as PRDM1 (Fig. 1⇑, B and C). PA28α, PA28β, and TAP2 promoters were occupied by IRF-1 and IRF-2 suggesting that PRDM1 is unable to efficiently compete at these sites.
PRDM1 and IRF occupancy of the IRF-E sequences is mutually exclusive and dependent on a precise sequence
The occupancy of ERAP1, TAPASIN, MECL1, and LMP7 promoters by all three factors potentially reflects combinations of promoters bound by either PRDM1 or the IRFs, as expected in the context of dynamic and antagonistic transcriptional regulation. Alternatively, PRDM1, IRF-1, and IRF-2 might co-occupy the IRF-E target sequences. To distinguish between these possibilities, next we examined interactions in vitro by EMSA. Oligonucleotides representing the various IRF-E sequences were incubated with COS cell extracts containing PRDM1, IRF-1, or IRF-2 alone, or in combination. Binding to the IRF-E sequences was clearly observed with the individual factors. The sites were occupied by IRF proteins either in a single or multimeric fashion (Fig. 2⇓). When extracts containing PRDM1 and IRF-1 or IRF-2 were used in combination, no additional complexes were observed containing both PRDM1 and either of the IRFs (Fig. 2⇓). Thus, PRDM1, IRF-1, and IRF-2 occupy IRF-E sequences separately and do not co-occupy the same sites.
At the IRF-E of the CIITA promoter IV, positions 5 and 11, are critical in determining dual regulation by IFN-γ and PRDM1 (19). However, it is not known whether this is a common feature of other promoters containing similar sites. To investigate this possibility, we used oligonucleotides representative of the ERAP1, TAPASIN, MECL1, and LMP7 IRF-E altered at these positions in EMSAs. The relative efficiency of PRDM1, IRF-1, and IRF-2 binding to the divergent sites was assessed in cold competitor assays. The altered IRF-E sequences successfully competed with the wild-type IRF-E sequences for binding to IRF-1 and IRF-2. In contrast, competition for binding to PRDM1 was barely detectable (Fig. 3⇓). Thus, in the context of these IRF-Es, positions 5 and 11, determine the capacity to bind PRDM1. Our findings suggest that the presence of a G rather than a C at positions 5 and 11 reflect evolutionary selection for dual regulation of these promoters.
PRDM1 represses IFN-γ-dependent promoter activation of MHC class I Ag-processing and -loading genes in a fashion dependent on the IRF-E sequence
To demonstrate whether PRDM1 binding to the promoters of MHC class I Ag-processing pathway genes causes transcriptional repression, we used luciferase reporter assays. As shown in Fig. 4⇓, PRDM1 mediates substantial repression of both basal and IFN-γ-dependent promoter activation (Fig. 4⇓A). In the presence of PRDM1, promoter activity remains below basal levels even after IFN-γ stimulation. ERAP1, TAPASIN, and MECL1 gave similar results, while partial inhibition was observed for the LMP7 promoter. Furthermore, a vector encoding the PRDM1 DNA-binding zinc-finger domain was sufficient to repress IFN-γ-dependent promoter activation (data not shown).
In EMSA, the mutant oligonucleotides preserve IRF, but not PRDM1 binding. These same substitutions abolished PRDM1-dependent repression in response to IFN-γ at the ERAP1, TAPASIN, and LMP7 promoters and significantly reduced repression at the MECL1 promoter (Fig. 4⇑B). Together, these data demonstrate that PRDM1 represses both the basal and IFN-γ-dependent activities of these promoters via the IRF-E sequence.
Distant orthologs of PRDM1 repress human TAPASIN promoter activity with the same strict sequence dependence
PRDM1 orthologs act as important transcriptional regulators among both vertebrates and invertebrates (15, 43, 44, 45, 46), including sea urchins, which lack an anticipatory/adaptive immune system and MHC class I genes (47). Hence, PRDM1 function has been co-opted during evolution into regulation of the adaptive immune response. The degree of conservation among PRDM1 orthologs is greatest in the zinc-finger DNA-binding domains (15). The IRF-E/PRDM1 sites in the MHC class I Ag-processing pathway genes include perfect matches to the optimum human and mouse PRDM1 consensus binding site (27, 28), and thus provide ideal targets for testing conservation of PRDM1 function.
We therefore examined the ability of frog (48), zebrafish (43), and sea urchin (15) Prdm1 to substitute for human PRDM1 in mediating repression of the TAPASIN promoter in vitro. All three orthologs, including sea urchin Prdm1 (blimp1/krox), clearly have the ability to repress both basal and IFN-γ-dependent promoter activation at the TAPASIN promoter (Fig. 5⇓A). As for human PRDM1, the ability of the orthologs to repress the TAPASIN promoter was eliminated by mutation of the IRF-E at position 5 and 11 (Fig. 5⇓B). In contrast to frog and zebrafish (2), the sea urchin, Strongylocentrotus purpuratus, does not possess an anticipatory/adaptive immune system or MHC Ag-presentation pathways (47). Thus, the ability of PRDM1 to regulate IRF-E sites is ancient and must have existed in a common ancestor before the evolution of the adaptive immune system.
PRDM1 represses IFN-γ-dependent induction of surface MHC class I and expression of the endogenous target genes
Repression of multiple target genes would be expected to impact on induction of MHC class I surface expression in response to IFN-γ. To test this, HeLa cells were transfected with PRDM1 IRES2 EGFP or control IRES2 EGFP vector and treated with IFN-γ. MHC class I surface expression was then assessed. EGFPhigh and EGFP− untransfected populations were compared (Fig. 6⇓). In controls, we observed indistinguishable levels of either basal or IFN-γ-induced MHC class I expression. The EGFPhigh population in PRDM1-transfected cells showed only a slight decrease in basal MHC class I expression, but displayed a complete failure to induce surface MHC class I expression in response to IFN-γ. Thus, PRDM1 expression has a profound effect on the ability of cells to up-regulate MHC class I surface expression in response to IFN-γ.
To examine whether this inhibition of surface MHC class I expression might be due to repression of MHC class I mRNA induction, EGFPhigh populations were sorted and MHC class I mRNA induction was assessed by RT-PCR. Control and PRDM1-transfected EGFPhigh populations fail to display a significant difference in levels of MHC class I mRNA expression (Fig. 7⇓A). In contrast, IFN-γ-dependent induction of ERAP1, TAPASIN, MECL1, and LMP7 mRNA was markedly inhibited in EGFPhigh PRDM1-transfected populations relative to control at 6 h poststimulation (Fig. 7⇓B). Repression of ERAP1 and TAPASIN mRNA, but not MECL1 or LMP7, remained substantial at 24 h (Fig. 7⇓C). ChIP assays confirmed that repression of these promoters was accompanied by PRDM1 occupancy (Fig. 7⇓D). PRDM1 therefore blocks IFN-γ-dependent induction of MHC class I surface expression due to repression of these components of the Ag-processing and peptide-loading pathway, and not MHC class I structural genes.
We have previously shown that the myeloma cell line U266 expresses both IRF-1 and IRF-2 in the absence of IFN-γ stimulation. Furthermore, the presence of these factors is sufficient to mediate increased expression of an otherwise IFN-γ-dependent gene, when PRDM1 levels are reduced (19). To demonstrate that endogenous PRDM1 mediates repression of ERAP1, TAPASIN, MECL1, and LMP7, we examined the effect of PRDM1 knockdown in U266 cells on the expression of these genes. PRDM1 protein was substantially reduced following transfection with specific siRNA (Fig. 7⇑E). Loss of PRDM1 led to an increase in expression of all four target genes, ranging from 1.6- to 4.6-fold (Fig. 7⇑F). Collectively, these data establish that the expression of genes on the MHC class I Ag-processing and -loading pathway is directly responsive to the relative levels of PRDM1 and IRF-1 or IRF-2.
Mutually exclusive expression of PRDM1 and TAPASIN in tonsil crypt epithelium is consistent with transcriptional repression in vivo
The tonsil crypt epithelium generates a specialized niche for Ag sampling (49), and represents an important site for class I presentation of viral pathogens such as HIV and EBV (50, 51). The modified squamous epithelium of the crypt is heavily infiltrated by lymphocytes and potentially subject to chronic cytokine stimulation. This epithelium displays strong expression of PRDM1 in superficial epithelial cells (52). To evaluate whether PRDM1 expression correlates with repression of components of the MHC class I Ag-presentation pathway in this tissue, we focused on TAPASIN expression, because an available Ab to this protein worked effectively in multicolor immunofluorescence. TAPASIN was expressed strongly in the lymphoid compartments of the tonsil, but was also expressed in tonsil crypt epithelium identified by staining with a pan-keratin Ab (Fig. 8⇓). In crypt epithelium, TAPASIN and PRDM1 displayed inverse expression. TAPASIN was strongest in basal and intermediate, but absent in superficial layers. This contrasted with the restricted zone of PRDM1 staining in more superficial epithelium. This pattern of PRDM1 and TAPASIN expression in the tonsil epithelium strengthens the idea that PRDM1 functions to repress this pathway in vivo.
The absolute level of MHC class I surface expression and the repertoire of peptides presented by MHC class I control CD8 T cell activation and target cell recognition. IFN-γ induces expression of MHC class I genes (53) and key components of the machinery responsible for peptide generation, loading, and trimming. In this study, we have demonstrated that PRDM1 can exert a dominant opposing role and repress the MHC class I Ag-presentation pathway.
We have identified a panel of novel PRDM1 target genes on this pathway. Our data reveal that four promoters are substantially occupied by PRDM1 under physiological conditions in a myeloma cell line. When bound to the newly defined target promoters, PRDM1 is a potent repressor of basal activity and IFN-γ-dependent induction. TAP2, PA28α, and PA28β are also potential targets of PRDM1-mediated repression, based on the presence of an overlapping PRDM1/IRF-E consensus site, but gave little evidence of occupancy in this study. At the initial step of peptide generation, PRDM1 acts to repress the transcription of two, MECL1 and LMP7, of the three catalytic components that distinguish the immunoproteasome (3). Because these three components are assembled in a cooperative fashion, absence of one or more subunits is sufficient to compromise assembly of the immunoproteasome and the enhanced generation of antigenic peptides (54, 55). The next component of the pathway targeted by PRDM1, TAPASIN, acts as a chaperone protein tethering the empty MHC class I molecule to the TAP transporter and facilitating the loading of antigenic peptide (6, 7). TAPASIN-deficient cell lines and mice have profound defects in Ag presentation and the efficiency with which stable, optimized MHC class I peptide complexes are formed (6, 56, 57). Because MHC class I alleles display differences in their relative dependence on TAPASIN for optimal peptide loading (7), repression of TAPASIN by PRDM1 has the potential to alter both the overall efficiency with which MHC class I peptide complexes are formed and the relative level of surface MHC class I alleles expressed. Finally, PRDM1 represses ERAP1 which has recently been shown to be the primary peptidase responsible for trimming antigenic peptides to fit the MHC class I-binding groove (8, 9, 58). Because peptides generated by the immunoproteasome may be particularly dependent on N-terminal trimming to fit the MHC class I peptide-binding groove, repression of ERAP1 has the potential to profoundly alter the generation of MHC peptide complexes (10, 59, 60). The impact on IFN-γ-dependent induction of surface MHC class I expression therefore reflects the combined action of PRDM1 on multiple linked steps in this functional pathway. However, we cannot exclude that repression of one or more of these steps mediate a dominant effect or that selective effects may be evident, dependent on the relative levels of PRDM1 and IRF expression.
A striking feature at the level of promoter sequences is the selection for overlapping PRDM1/IRF-E-binding sites in this set of genes, which act at sequential steps of a single functional pathway. Because the degree of overlap seen at these promoters is not a common feature among IRF- or IFN-regulated promoters (42), this suggests a particular selection for dual regulation. The lack of such selection in the MHC class I promoters themselves (61) argues against a direct role for PRDM1 in repressing MHC class I gene transcription, and we demonstrate that the blocked induction of MHC class I surface expression by PRDM1 is not reflected in repression of MHC class I transcription. We have previously shown that PRDM1 represses the IFN-γ responsive promoter IV of CIITA (19). In addition to regulating MHC class II expression, CIITA has been shown to play a role in the IFN-γ-responsive transcription of MHC class I (62, 63). Repression of CIITA potentially provides yet another mechanism by which PRDM1 may target MHC class I expression. However, the data presented here demonstrate that repression of MHC class I surface expression by PRDM1 is not dependent on transcriptional inhibition of MHC class I structural genes.
The evolutionary conservation of the PRDM1 protein is striking and centered in the PR- and DNA-binding domains (15). We demonstrate that Prdm1 orthologs from both lower vertebrates and invertebrates have the ability to effectively substitute for human PRDM1 in mediating repression of IFN-γ-dependent promoter induction. Moreover, the strict dependence on the precise sequence of the IRF-E/PRDM1 sites at position 5 and 11 is conserved. The sea urchin and zebrafish Prdm1 orthologs are derived from species that diverged before, and soon after, the evolutionary acquisition of the adaptive immune system (1). Our results indicate that the PRDM1 DNA-binding specificity was fixed before acquisition of the adaptive immune system. The presence of an IRF-1/IRF-2 ortholog in the sea urchin immune system (47) suggests that an antagonistic module of PRDM1 and IRF transcription factors was pre-existent and co-opted as a functional unit into the adaptive/anticipatory immune system.
The wider role for PRDM1-dependent regulation of Ag presentation remains to be determined. However, we propose a model in which PRDM1 provides a mechanism for controlling Ag presentation during cellular stress responses. Under conditions of cellular stress, Ag presentation may be profoundly altered, and consequent T cell responses can contribute to the initiation of autoimmune disease (11). In addition to its induction during cellular viral infection (12), we have shown that PRDM1 expression is responsive to a range of stressors, in particular those targeting the unfolded protein response of the ER (32). Responses of this type are linked to MHC class I-dependent Ag presentation in a number of ways. First, the misfolding of the MHC class I allele HLA-B27, particularly in the context of IFN-γ stimulation, can initiate an ER stress response (64). Second, the generation of antigenic peptides is altered by the phosphorylation of the eIF2α subunit of the ribosome (65). The selection of cryptic ribosomal initiation sites generates alternate immunogenic peptide products, and a general role for defective ribosomal products as a source of antigenic peptides is well-established (66). Third, chemical modification of proteins and the generation of altered peptides can occur and contribute to autoimmune disease (67). We propose that PRDM1 induction contributes to the control of MHC class I Ag presentation during cellular stress to help prevent inappropriate and deleterious Ag presentation. It is notable that lineages which express PRDM1, such as macrophages, squamous epithelia and endothelium, are particularly prone to exposure to adverse environmental conditions. We conclude that PRDM1 is a unique regulator of Ag presentation.
We thank Peter Cockerill for pXPG vector, Philip Ingham for Danio rerio Prdm1/Blimp-1 expression vector and pCSII, Christoph Niehrs for Xenopus laevis Prdm1/Blimp-1 expression vector, and Eric Davidson for S. purpuratus Prdm1/blimp1/krox. We thank Liz Straszynski for cell sorting. We are indebted to Liz Bikoff for helpful advice on the manuscript.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was supported by a Medical Research Council Clinician Scientist Fellowship (to R.M.T).
↵2 Address correspondence and reprint requests to Dr. Reuben M. Tooze, Section of Experimental Haematology, Leeds Institute of Molecular Medicine, Wellcome Trust Brenner Building, St. James’s University Hospital, Beckett Street, Leeds LS9 7TF, U.K. E-mail address:
↵3 Abbreviations used in this paper: LMP, large multifunctional peptidase; MECL-1, multicatalytic endopeptidase complex subunit-1; ER, endoplasmic reticulum; ERAP1, ER amino peptidase-1; BLIMP-1, B lymphocyte-induced maturation protein-1; PRDM1, PR domain-containing 1, with ZNF domain; IRF, IFN regulatory factor; IRF-E, IRF element; ChIP, chromatin immunoprecipitation; EGFP, enhanced GFP; siRNA, short interfering RNA.
- Received February 16, 2007.
- Accepted September 11, 2007.
- Copyright © 2007 by The American Association of Immunologists