Human cytomegalovirus microrna mir-us4-1 inhibits CD8 + T cell responses by targeting the aminopeptidase ERAP1

Transcription

1 Human cytomegalovirus microrna inhibits CD8 + T cell responses by targeting the aminopeptidase ERAP1 211 Nature America, Inc. All rights reserved. Sungchul Kim 1, Sanghyun Lee 1, Jinwook Shin 2, Youngkyun Kim 1, Irini Evnouchidou 3, Donghyun Kim 1, Young-Kook Kim 4, Young-Eui Kim 5, Jin-Hyun Ahn 5, Stanley R Riddell 6, Efstratios Stratikos 3, V Narry Kim 4 & Kwangseog Ahn 1 Major histocompatibility complex (MHC) class I molecules present peptides on the cell surface to CD8 + T cells, which is critical for the killing of virus-infected or transformed cells. s of MHC class I presented peptides are trimmed to mature epitopes by the aminopeptidase ERAP1. The US2 US11 genomic region of human cytomegalovirus (HCMV) is dispensable for viral replication and encodes three micrornas (mirnas). We show here that HCMV specifically downregulated ERAP1 expression during viral infection. Accordingly, the trimming of HCMV-derived peptides was inhibited, which led to less susceptibility of infected cells to HCMV-specific cytotoxic T lymphocytes (CTLs). Our findings identify a previously unknown viral mirna based CTL-evasion mechanism that targets a key step in the MHC class I antigen-processing pathway. Major histocompatibility complex (MHC) class I molecules bind peptides and present them on the cell surface for recognition by CD8 + T cells. Cytosolic peptides generated by the proteasomes and prematurely translated peptides are transported to the endoplasmic reticulum by means of the transporter associated with antigen processing complex 1,2. Some of these peptides are further processed by aminopeptidases that reside in the endoplasmic reticulum, such as ERAP1 (also called A-LAP, ARTS-1 or PILS-AP), and peptide trimming by ERAP1 in the endoplasmic reticulum is a crucial step for determining the quality and quantity of optimal production of antigenic peptides and the stability of the heterotrimer of MHC class I β 2 -microglobulin peptide 3 6. ERAP1 trims relatively long peptides efficiently in a sequence-specific manner, which results in the accumulation of optimal peptides eight to nine amino acids in length 5,7,8. ERAP1 therefore acts as a molecular ruler for the production of antigenic peptides 9. In addition, genome-wide association studies have associated nonsynonymous single-nucleotide polymorphisms in ERAP1 with ankylosing spondylitis 4. Additionally, ERAP1 has non peptide-processing functions through its role in the shedding of cytokine receptors 4. MicroRNAs (mirnas) are small RNAs nucleotides in length that regulate gene expression by complete or partial base pairing with the 3 untranslated region (UTR) of their target mrna, which leads to the cleavage, destabilization or translational repression of mrna 1. Since the first report in 24 that viruses express mirnas 11, many viral mirnas have been discovered, mainly related to viral proliferation and survival-related evasion of the immune response (immunoevasion), although this is based on a limited number of studies 12. The β-herpesvirus human cytomegalovirus (HCMV) expresses at least 14 mirnas during productive infection 13,14. One HCMV-encoded mirna, mir-ul112-1, targets three HCMV genes involved in viral replication 15, and another HCMV-encoded mirna, mir-us25-1, can downregulate many host genes associated with cellcycle control and tumor progression by interacting with the 5 UTR of the target mrna 16. In addition, mir-ul112-1 targets one cellular gene, encoding MHC class I related chain B, to enable escape from the natural killer cell mediated immune response 17, and mir-ul112-1 can also repress the expression of UL114, a uracil DNA glycosylase that can influence viral replication 18 and whose gene is antisense to mir-ul However, the cellular or host targets of many viral mirnas remain to be elucidated. Many viruses have evolved strategies that target crucial stages of the MHC class I antigen-presentation pathway to prevent viral peptides from being presented to CD8 + T cells 1. The 9-kilobase US2 US11 region in the HCMV genome encodes at least five glycoproteins (US2, US3, US6, US1 and US11) specifically dedicated to interfering with the presentation of antigenic peptides to CD8 + T cells 1, Because deletion of the US2 US11 genomic segment has no influence on viral replication 24, the US2 US11 region is considered a reservoir of viral genes that encode molecules whose functions are to escape the presentation of viral antigens by MHC class I molecules. 1 National Creative Research Initiative Center for Antigen Presentation, Department of Biological Sciences, Seoul National University, Seoul, Republic of Korea. 2 Department of Molecular Genetics and Microbiology, Duke University Medical Center, Durham, North Carolina, USA. 3 Institute of Radioisotopes and Radiodiagnostic Products, National Centre for Scientific Research Demokritos, Athens, Greece. 4 Department of Biological Sciences, Seoul National University, Seoul, Republic of Korea. 5 Department of Molecular Cell Biology, Samsung Biomedical Research Institute, Sungkyunkwan University School of Medicine, Suwon, Seoul, Republic of Korea. 6 Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, Washington, USA. Correspondence should be addressed to K.A. (ksahn@snu.ac.kr). Received 2 June; accepted 27 July; published online 4 September 211; doi:1.138/ni VOLUME 12 NUMBER 1 OCTOBER 211 nature immunology

2 211 Nature America, Inc. All rights reserved. Figure 1 Expression of HCMV results in less ERAP1b mrna and protein but not less ERAP1a mrna or protein. (a c) Change in cellular mrna by viral (a), mir-us5-1 (b) and mir-us5-2 (c) versus mrna expression in response to control mirna (sigfp) in U373MG cells; red dot, ERAP1 expression. (d,e) Quantitative real-time PCR analysis of ERAP1a mrna (d) and ERAP1b mrna (e) in HEK293T cells treated with,, a or b, presented relative to the expression of GAPDH mrna (encoding glyceraldehyde 3-phosphate dehydrogenase). P <.5, versus control mirna (two-tailed Student s t-test). (f) Immunoblot analysis (IB) of ERAP1 and GAPDH (loading control), RNase-protection assay (RPA) of expression, and RNA blot analysis (RNA) of hsa-mir-16 expression (loading control) in HEK293T cells transfected with control vector or vector expressing, or (M), followed by selection for 1 week with 2 µg puromycin. (g) Immunoblot analysis (IB) of ERAP1 and GAPDH, and RNA blot analysis of the expression of and hsa-mir-16, in HEK293T cells transfected with control mirna (C) or increasing concentrations of vector encoding (2 µg, 4 µg and 1 µg; wedge). Data are representative of three independent experiments (mean ± s.d. in d,e). In this study, we demonstrate that HCMV specifically targeted ERAP1 and thereby inhibited the trimming of precursors to the MHC class I presented mature epitopes, which resulted in the inhibition of cytotoxic T lymphocyte (CTL) immune responses. Our findings expand understanding of the strategy of the host-virus arms race. They also reinforce the proposal that the US2 US11 region in the HCMV genome has evolved into a reservoir of genes that encode viral immunoevasive molecules that act against the elimination of HCMV-infected cells by CD8 + T cells of the host immune response. RESULTS Identification of ERAP1 as a host target of We found 3 mirnas (, mir-us5-1 and mir-us5-2) among the total 14 HCMV mirnas encoded by the presumptive US4 and US5 loci in the US2 US11 region 13,14 (Supplementary Fig. 1a). Given that many products of this region can block the MHC class I antigenpresentation pathway, we hypothesized that some of these mirnas might be able to inhibit MHC class I mediated antigen presentation. To identify potential cellular targets of HCMV mirnas encoded by the US2 US11 region, we investigated changes in the host-cell transcriptome in response to the expression of HCMV, mir- US5-1 and mir-us5-2. We synthesized the three mirnas, as well as small interfering RNA (sirna) targeting the open reading frame of sequence encoding green fluorescent protein (sigfp) or protein disulfide isomerase as a positive control for evaluating transfection efficiency. We transfected these synthetic RNAs twice into U373MG cells, an HCMV-permissive human astrocytoma cell line 25, at intervals of 48 h to maximize transfection efficiency. Immunoblot analysis showed that the expression of protein disulfide isomerase was much lower in cells transfected with sirna specific for this protein (Supplementary Fig. 1b), which confirmed our expectation that the transfection efficiency was sufficient to proceed to microarray analysis of the three HCMV mirnas. Among the top-ranked genes with the most diminished expression, ERAP1 mrna was downregulated more than 9% by (Fig. 1a), whereas mir-us5-1 and mir- US5-2 did not substantially alter the expression of ERAP1 mrna (Fig. 1b,c). On the basis of these results and published studies showing that ERAP1 has a critical role in regulating the antigenic peptide pool d ERAP1a mrna (relative) a 1 b c HCMV mirna a b e ERAP1b mrna (relative) ERAP HCMV mir-us mirna a b f IB RPA RNA ERAP in the endoplasmic reticulum 4, we hypothesized that among these three HCMV mirnas, targets ERAP1 to inhibit MHC class I mediated antigen presentation. Downregulation of ERAP1 mrna and protein by Human ERAP1 produced at least two mrna isoforms, isoform a (ERAP1a) and isoform b (ERAP1b; Supplementary Fig. 2a). Although the amino acid sequences of both isoforms were almost identical, the 3 UTR sequence of ERAP1a differed from that of ERAP1b. Because the base pairing between an mirna and its target mrna generally occurs in the 3 UTR of the target mrna in animal cells 26, we investigated which isoform was targeted by. To confirm the microarray data, we expressed control mirna, and three sirnas as positive controls for knockdown of ERAP1:, which targets the open reading frames of both isoforms; a, which targets the 3 UTR of ERAP1a; and sier- AP1b, which targets the 3 UTR of ERAP1b. We expressed all three sirnas as forms of small hairpin RNA (shrna) in the human embryonic kidney cell line HEK293T. We did quantitative real-time PCR analysis with primers that bind specific regions in the 3 UTR of each isoform (Supplementary Fig. 2a). As expected, expression of a or b resulted in lower expression of ERAP1a or ERAP1b mrna, respectively, and downregulated both isoforms. ERAP1a mrna was not downregulated by expression of (Fig. 1d), whereas ERAP1b mrna was more than 6% lower in abundance after expression of (Fig. 1e). These results indicated that specifically downregulated ERAP1b mrna but not ERAP1a mrna. Next we determined whether the downregulation of ERAP1b mrna led to less ERAP1b protein. As a control, we constructed a vector ((M)) with mutation of the seed nucleotides (nucleotides 2 8 at the 5 end of the mature mirna; Supplementary Fig. 2b) because the seed sequence of an mirna is considered a critical determinant of the recognition and base pairing of target mrna by mirna, and most mirnas bind to their target mrna 3 UTRs with complete Watson-Crick base pairing in the seed region 27. We expressed control mirna (sigfp),, and (M) in HeLa human cervical cancer cells. We investigated the expression of HCMV mir-us5-2 (M) α-erap1 α-gapdh hsa-mir g IB RNA mirna C ERAP1 1 4 α-erap1 α-gapdh hsa-mir-16 nature immunology VOLUME 12 NUMBER 1 OCTOBER

3 211 Nature America, Inc. All rights reserved. a ERAP1a 3 UTR mutant : AC ERAP1a 3 UTR : 5 G A C 3 CCCUCU CA AUGUUG GGGGGA GU UACAGC : 3 UA C GCAGG kcal/mol d Luminescence (relative) 1..5 C 3 UTR : ERAP1a ERAP1a Mut C C C ERAP1b ERAP1b Mut by RNase-protection assay (Fig. 1f) or RNA blot analysis (Fig. 1g). Immunoblot analysis showed that ERAP1 expression was lower after expression than after expression of control mirna, but (M) did not affect ERAP1 protein expression (Fig. 1f). The ability of to downregulate ERAP1 protein seemed to be as effective as that of (Fig. 1f). ERAP1b mrna was times more abundant than ERAP1a mrna in the cells used in these experiments (Supplementary Fig. 3a). To rule out the possibility that direct comparison of ERAP1a and ERAP1b mrna might have been biased because of differences in primer efficiency, we confirmed that both primers amplified each targeted ERAP1 isoform equally (Supplementary Fig. 3b,c). Thus, our findings suggest that the amount ERAP1b protein accounts for most of the ERAP1 content in the cell. In dose-dependent experiments, the amount of ERAP1 protein was inversely proportional to expression (Fig. 1f). These results demonstrated that specifically downregulated ERAP1b mrna, thereby diminishing the overall amount of cellular ERAP1 protein. Direct targeting of the ERAP1b mrna 3 UTR by To confirm that the diminished ERAP1b mrna and ERAP1 protein was due to direct targeting by and to identify potential target sites in the 3 UTR of ERAP1 mrna, we did bioinformatic scanning of the 3 UTRs of ERAP1a and ERAP1b with the RNAhybrid algorithm 28 and rna22 algorithm 29. We selected the most likely target site in the 3 UTR sequences of ERAP1a mrna (Fig. 2a) and ERAP1b mrna (Fig. 2b), focusing on base pairing in the seed sequence of the mirna and its binding energy, because the seed sequence is considered a critical determinant in the recognition of target mrna by mirna 1. e (nt) 3-2- b c ERAP1b 3 UTR mutant : GA ERAP1b 3 UTR : 5 U AAA C U 3 GUCC CCCUGUAC CCAUGUU UAGG GGGACGUG GGUACAG 1. : 3 CA C kcal/mol.5 Mock NT-NR NT (M) (M) AGO f α-flag α-flag α-flag α-gapdh (M) Figure 2 HCMV targets the 3 UTR of ERAP1b and physically binds to ERAP1b mrna in the RISC. (a,b) Predicted binding site for in the 3 UTR of ERAP1a mrna (a) and ERAP1b mrna (b). Bold indicates the expected seed region interaction site; red, nucleotide sequence replaced with the sequence indicated by the arrow in mutant 3 UTRs; blue, HCMV. Bottom right, minimum free energy of hybridization. (c) Dual-luciferase assay of HEK293T cells transfected with (M) AGO Ab HC Lysate g ERAP1a mrna (relative) Luminescence (relative) 3 UTR : ERAP1a (M) (M) Total RISC IP ERAP1a Mut ERAP1b ERAP1b Mut We mutated two nucleotides in the predicted seed binding sites of the 3 UTR of each isoform (Fig. 2a,b), then did a luciferase reporter assay. Notably, we observed that significantly downregulated the expression of firefly luciferase fused to the ERAP1b wild-type 3 UTR, whereas (M) did not (Fig. 2c). In the reciprocal experiment in which we transfected cells with the firefly luciferase vector fused to the mutated ERAP1b 3 UTR, neither wild-type nor (M) had an effect on the expression of firefly luciferase (Fig. 2c). The expression of the firefly luciferase vector containing either the wild-type or mutated ERAP1a 3 UTR was not affected by wild-type or (M) (Fig. 2c). Given that the minimum free energy for the hybridization of to ERAP1b ( 3.5 kcal/mol; Fig. 2b) was greater than that for ERAP1a ( 2.4 kcal/mol; Fig. 2a), the insensitivity of ERAP1a to might have been attributable to insufficient base pairing between these. In addition, we examined the dose-dependent effect of on luciferase reporters containing the ERAP1 3 UTR derivatives. After confirming the dose-dependent expression of by RNA blot analysis (Supplementary Fig. 4), we transfected luciferase reporter 3 UTR vectors into HEK293T cells with dose-dependent expression of and did a luciferase reporter assay. We observed that the expression of firefly luciferase from the vector containing wild-type ERAP1b 3 UTR was inversely proportional to the expression of, whereas the expression of firefly luciferase from vector containing wild-type or mutated ERAP1a 3 UTR or mutated ERAP1b 3 UTR was not affected by (Fig. 2d). Given these observations, we concluded that targeted only the 3 UTR of ERAP1b mrna, but not the 3 UTR of ERAP1a mrna, in a seed region binding dependent manner. h ERAP1b mrna (relative) (M) 5 µg vector expressing mirna (key), 1 ng luciferase reporter vector for wild-type () or mutated (Mut) 3 UTR of ERAP1a or ERAP1b (horizontal axis) and 5 ng renilla vector (control); results are presented relative to the luminescence of renilla luciferase. (d) Dual-luciferase assay as in c except for the use of 1 µg, 2 µg or 5 µg (wedge) of vector expressing and 5 µg vector expressing control mirna (C). (e) RNase-protection assay to detect in HEK293T cells transfected with vector for or (M) and empty vector (Mock) or vector for human AGO1 AGO4 tagged with Flag at the N terminus (AGO), followed by RISC immunoprecipitation 48 h after transfection and extraction of total RNA. NT-NR, no target RNA or RNase; NT, no target RNA;, undigested probe;, HCMV. (f) Immunoblot analysis of aliquots of the RISC immunoprecipitates and cell lysates in e, probed with antibody to Flag (α-flag). Ab HC, heavy-chain immunoglobulin. (g,h) Quantitative real-time PCR analysis of ERAP1a mrna (g) and ERAP1b mrna (h) among RNA extracted from RISC immunoprecipitates or total samples of the cells in e, presented relative to GAPDH mrna. P <.5 and P <.1 (c,d,h), versus control mirna (two-tailed Student s t-test). Data are representative of three independent experiments (mean ± s.d. in c,d,g,h). Mock AGO Total RISC IP 986 VOLUME 12 NUMBER 1 OCTOBER 211 nature immunology

4 211 Nature America, Inc. All rights reserved. Physical interaction of with ERAP1b mrna Although many targets of mirnas have been established by several target-prediction algorithms, such algorithms can occasionally result in false-positive results for cellular targets. To obtain further evidence for the targeting of ERAP1b by, we examined the physical interaction between ERAP1b mrna and by an RNAinduced silencing complex (RISC) immunoprecipitation assay. This assay was developed to extract Argonaute (AGO)-bound mrnas targeted by mirnas 3,31. To determine whether enrichment of the known target mrnas of a specific mirna can be achieved by RISC immunoprecipitation, we chose Homo sapiens mir-21 (hsa-mir-21; called mir-21 here) as a model mirna for the RISC immunoprecipitation assay because mir-21 is known to have a variety of human target genes, such as PDCD4, SPRY2, TPM1, MASPIN, RECK and PTEN 32. After confirming that the product of the vector expressing sh-mir-21 (an shrna-expressing plasmid engineered to express the mature mirna sequence of endogenous mir-21) was of the same sequence and size as endogenous mir-21 (Supplementary Fig. 5a), we transfected HEK293T cells with vector expressing sh-mir-21 along with vectors for human AGO1, AGO2, AGO3 or AGO4 tagged with Flag at the N terminus to confirm the hypothesis that Flag-tagged human AGO proteins formed a complex with mir-21 expressed from sh-mir-21. Although human mirnas, including mir-21, were generally able to form a complex with each human AGO protein (AGO1 AGO4) in the RISC (Supplementary Fig. 5b) 33, we used a combination of four Flag-tagged human AGO proteins to exclude the possibility that a specific mirna targets the specific mrna by one or more AGO proteins. At 48 h after transfection, we confirmed the expression of Flag-tagged human AGO proteins in total cell lysates and the enrichment of Flag-tagged human AGO proteins in RISC immunoprecipitates (Supplementary Fig. 5c). As predicted from the results of published studies 32, there was enrichment for all known target mrnas of mir-21 under mir-21-overexpression conditions, which confirmed the fidelity of our assay (Supplementary Fig. 5d,e). Using the same experimental conditions, after confirming the expression of and enrichment for (Fig. 2e) and Flagtagged human AGO proteins (Fig. 2f) in whole-cell lysates and immunoprecipitates, we examined the physical interaction of with ERAP1b mrna by quantitative real-time PCR. We did not detect enrichment for ERAP1a mrna in the presence of (Fig. 2g). In contrast, there was about 3.8-fold enrichment for ERAP1b mrna solely when was overexpressed (Fig. 2h). As expected, there was enrichment for neither ERAP1 isoform when (M) was expressed (Fig. 2g,h). Together these results demonstrated that physically interacted with ERAP1b mrna in the RISC. Downregulation of ERAP1 during HCMV infection We next investigated whether ERAP1 was downregulated by during infection with HCMV. Because the pre-mirna or mature mirna sequence of is located in the 5 UTR of putative transcripts encoding US5, complete deletion of the entire pre-mirna sequence from the HCMV genome might affect the expression of neighboring genes near the sequence. Thus, to exclude the possibility of any side effects of deletion of the full sequence of, we generated a mutant of laboratory HCMV strain AD169 (HCMV US4) in which processing of the primary transcript by the RNase Drosha was defective by substituting only six nucleotides around the predicted cropping site that contains three nucleotides of the 5 -end of the mature (Supplementary Fig. 6). To confirm the hypothesis that an intact virus was generated without disruption of the entire HCMV genome, we constructed a control revertant virus whose genome sequence was theoretically identical to that of wildtype HCMV AD169 by using HCMV US4 as a template, and then infected human foreskin fibroblast (HFF) cells with wild-type HCMV, HCMV US4 or revertant HCMV, followed by an RNase-protection assay of expression. For infection with wild-type or revertant HCMV, expression began at about 24 h after infection and increased until 72 h after infection, whereas was not expressed in HFF cells infected with HCMV US4 (Fig. 3a). Immunoblot analysis of lysates of the infected cells showed that the overall amount of ERAP1 protein decreased after infection with wild-type or revertant HCMV, whereas it remained unchanged in cells infected with HCMV US4 (Fig. 3b). The expression of did not affect expression of the irrelevant targets IE1-IE2 and GAPDH (Fig. 3b), which indicated had a specific effect on ERAP1 expression. We next analyzed the alteration in the abundance of ERAP1 mrna isoforms during the course of infection. We detected no difference between infection with wild-type or revertant HCMV in the amount of ERAP1a mrna throughout the course of infection (Supplementary Fig. 7a). Whereas infection with HCMV US4 had little effect on the amount of ERAP1b mrna, infection with wild-type HCMV resulted in less ERAP1b mrna, which reached about 1% of that in uninfected HFF cells at 72 h after infection. Expression of did not affect the expression of IE1 mrna (Supplementary Fig. 7b), which indicated that had a specific effect on the expression of ERAP1b mrna. These results suggested that the downregulation of ERAP1 protein resulted mainly from -mediated downregulation of ERAP1b mrna. Inhibition of ovalbumin precursor trimming by As part of our effort to determine the biological relevance of mir- US4-1, we initially assessed the effect of on the production of peptide generated from ovalbumin (OVA) amino acids (SIINFEKL; also known as OVA8) from an ovalbumin-derived precursor through the use of the well-known OVA8 experimental system 5,29. a b HCMV AD169 Time after NT US4 Mut Rev infection (h) : NR NT U (nt) Time after (h) : α-erap1 α-ie1/2 α-gapdh HCMV AD169 US4 Mut Rev U Figure 3 Downregulation of ERAP1 in HCMV-infected HFF cells. (a) RNase-protection assay (as in Fig. 2e) to measure in HFF cells left uninfected (U) or infected for 1 h with wild-type HCMV AD169 (), HCMV US4 ( US4 Mut) or revertant HCMV (Rev) at a multiplicity of infection of 5, followed by extraction of total RNA h, 24 h, 48 h or 72 h after infection., undigested probe;, HCMV. (b) Immunoblot analysis of ERAP1, IE1-IE2 (IE1/2) and GAPDH (loading control) in aliquots of the cells in a. Data are representative of three independent experiments. nature immunology VOLUME 12 NUMBER 1 OCTOBER

5 Figure 4 HCMV mir-us4 inhibits the trimming of OVA8 peptide from OVA precursor peptide by ERAP1. (a) Quantitative real-time PCR analysis of ERAP1a and ERAP1b mrna in H-2K b -expressing HeLa cells transfected with control mirna,, or mir- US4-1(M), followed by 1 week of selection with puromycin; results are presented relative to GAPDH. (b,c) Activity of β-galactosidase (lacz) in H-2K b -expressing HeLa cells (target cells) treated as in a, then transfected for 48 h with vector for N5OVA8 (b) or OVA8 (c) and cultured ( HeLa cells) for 16 h with B3Z cells a mrna (relative) (M) ERAP1a ERAP1b b lacz activity (relative) (M) c lacz activity (relative) (M) (effector cells) at an effector/target ratio of 1 ( B3Z cells), 3 ( B3Z cells) or 9 ( B3Z cells); β-galactosidase production assessed with the lacz substrate CPRG is presented relative to that obtained with control mirna. P <.1, versus control mirna (two-tailed Student s t-test). Data are representative of three independent experiments (mean ± s.e.m.). 211 Nature America, Inc. All rights reserved. We transfected mouse H-2K b expressing HeLa cells with vector encoding OVA8 or N5OVA8, an OVA8 peptide with an extension of five amino acids (LEQLESIINFEKL), along with mirna or sirna. Because each peptides was fused to the C-terminal end of ubiquitin to allow post-translational cleavage of the peptide from cytosolic ubiquitin by ubiquitin C-terminal hydrolase 34 and ERAP1 can trim the N5OVA8 precursor to mature OVA8 peptide 29, we monitored the generation of OVA8 by measuring the response of the B3Z86/9.14 (B3Z) mouse T cell hybridoma. As B3Z T cells are from a CD8 + T cell hybridoma line generated by fusion of the OVA H-2K b specific cytotoxic clone B3 with a β-galactosidase-inducible derivative of the mouse thymoma BW5147 and engineered to secrete β-galactosidase when its T cell antigen receptor engages an H-2K b OVA8 complex 35,36, the B3Z T cell response can be measured by quantification of β-galactosidase production. In confirming the specific and sufficient knockdown of ERAP1b mrna by to the amount in -treated HeLa cells (Fig. 4a), we found that expression in mouse H-2K b expressing HeLa cells transfected to express N5OVA8 led to a lower B3Z response than did expression of control mirna (sigfp) in these cells (Fig. 4b). We also found that expression of (M) did not produce a lower the B3Z response, whereas as a control was able to inhibit the B3Z response (Fig. 4b). In a control experiment, we observed that expression of, or (M) in cells expressing OVA8 had no effect on the B3Z responses (Fig. 4c). In addition, the expression of did not affect the steady-state concentration of components of the MHC class I antigen-presentation machinery, such as MHC class I heavy chain, ERp57, protein disulfide isomerase or tapasin (Supplementary Fig. 8). In accordance with published observations obtained with ERAP1-deficient mice 34,37,38, our findings indicated that was able to inhibit trimming of the OVA8 precursor to OVA8 mature peptide by targeting ERAP1. Subversion of HCMV-specific CTL responses by To directly demonstrate the physiological importance of, we did CTL assays in the presence or absence of in the context of HCMV infection using CD8 + HCMV-specific CTL clones from HCMV-seropositive human donors (Table 1). Initially, we investigated whether ERAP1 is involved in the generation of HCMV-derived CTL epitopes through the use of an in vitro ERAP1 trimming assay. We incubated each synthetic peptide precursor (with a two amino acid extension at the N terminus of the mature form) with various amounts of recombinant ERAP1. We monitored the generation of mature epitopes by analytical high-performance liquid chromatography (Fig. 5a) and calculated the rate of mature epitope generation (Table 1). All epitope precursors tested were effectively trimmed to their mature epitopes (UL23 amino acids (UL23(34 42)), IE1(88 96), UL28( ), UL16(162 17) and UL15( )). In the case of UL15( ), the rate of generation of mature epitope was slower than that of other epitopes, and it was further trimmed into short peptides by ERAP1 (Fig. 5a). These in vitro results suggest that ERAP1 may be involved in the generation of a broad spectrum of HCMV epitopes. The US2 US11 region of wild-type HCMV AD169 encodes several immunoevasion proteins that downregulate the cell surface expression of MHC class I molecules 1. To overcome the limitations imposed by these immunoevasion proteins and to test the exclusive effect of mir- US4-1 on CTL response, we used HCMV RV798, a version of a mutant laboratory strain of HCMV with deletion of the US2 US11 genome region. Infection with HCMV RV798, unlike infection with wild-type HCMV or HCMV US4, slightly upregulated the surface expression of MHC class I molecules (Supplementary Fig. 9). We transfected synthetic into all autologous dermal fibroblasts (fibroblasts CP, RT and DK, which were derived from HCMV-seropositive patients and correspond to CTL clones in Table 1) twice before infection. At 24 h after the second transfection, we infected fibroblast cells Table 1 Data-collection relationships for in vitro ERAP1 trimming and CTL assays CTL clone HCMV epitope Mature epitope epitope Generation rate Restricted HLA allele Fibroblast donor ERAP1 dependency of CTL response 1B12-24 UL23(34 42) WPKDRCLVI LVWPKDRCLVI 12. B511 RT + 1F8-68 IE1(88 96) QIKVRVDMV VKQIKVRVDMV 1.94 B81 CP + 2A1-3 B81 CP + 1A4-1 UL28( ) FRCPRRFCF FWFRCPRRFCF 1.3 Cw72 DK + 9G4-17 UL16(162 17) YPRPPGSGL SLYPRPPGSGL.45 B72 DK + 1A8-8 UL15( ) YADPFFLKY SLYADPFFLKY.42 A11 CP For HCMV empitopes, the amino acids included in the epitope are in parentheses; bolding in the precursor epitope indicates the two amino acid N-terminal extension; the generation rate is presented as the moles of mature peptide generated per mole of enzyme per second; fibroblast donors are identified by initials; +, dependent;, not dependent. 988 VOLUME 12 NUMBER 1 OCTOBER 211 nature immunology

6 a UL23(34 42) IE1(88 96) UL28( ) UL16(162 17) UL15( ) 5 ng, 5 min 5 ng, 45 min 2 ng, 5 min 2 ng, 45 min 1 ng, 15 min 5 ng, 45 min 1 ng, 9 min 1 µg, 45 min 1 µg, 5 min 1 µg, 15 min 211 Nature America, Inc. All rights reserved. b Clone 1B12-24 for epitope UL23(34 42) WPKDRCLVI LVWPKDRCLVI with HCMV RV798 and confirmed by RT-PCR analysis the expression of viral genes encoding proteins containing epitopes used in the in vitro ERAP1 trimming assay (Supplementary Fig. 1). The efficiency of knockdown of ERAP1b by was as much as 65 8% in all donor fibroblasts at 48 h after infection (Supplementary Fig. 11). The -mediated knockdown of ERAP1 inhibited the lysis of antigen-specific target cells by five CTL clones (1B12-24, 1F8-68, 2A1-3, 1A4-1 and 9G4-17; Fig. 5b). CTL clone IA8-8, specific to the epitope of UL15( ), was efficiently able to lyse the target cells regardless of ERAP1 activity (Fig. 5b), which indicated that not all HCMV epitopes are generated in an ERAP1-dependent manner. Overall, these data provide direct evidence for a critical role of in the immunoevasion of CTL responses by targeting ERAP1, a key component of the antigen-processing machinery. Furthermore, our work supports the proposal of an important in vivo role for ERAP1 in antiviral CTL immune responses in humans. DISCUSSION In this study, we have shown that HCMV targeted ERAP1b mrna but not ERAP1a mrna, which resulted in the overall downregulation of ERAP1 protein. We also found that specifically bound to ERAP1b mrna in the RISC complex, and the 3 UTR of ERAP1b mrna was targeted by in a manner that destabilized or degraded mrna and eventually led to less ERAP1b mrna. We demonstrated that during HCMV infection, the amount of ERAP1b mrna and overall ERAP1 protein expression were inversely proportional to expression. Furthermore, we showed that inhibited the lysis of infected target cells by CTLs, which provided evidence of its physiological relevance. Thus, our results have identified a previously unknown mirna-based immunoevasion strategy in which VKQIKVRVDMV QIKVRVDMV Clone 1F8-68 for epitope IE1(88 96) FRCPRRFCF FWFRCPRRFCF Clone 2A1-3 for epitope IE1(88 96) YPRPPGSGL SLYPRPPGSGL Clone 1A4-1 for epitope UL28( ) YADPFFLKY SLYADPFFLKY Clone 9G4-17 for epitope UL16(162 17) Clone 1A8-8 for epitope UL15( ) Figure 5 HCMV inhibits the generation of HCMV-derived antigenic peptides and CD8 + CTL responses. (a) Chromatograms of epitope production in vitro by ERAP1: top, precursor alone; middle, precursor plus a moderate amount of enzyme; and bottom, precursor plus more enzyme., peak corresponding to the mature epitope; right margin, reaction conditions; below, precursor sequence, with N-terminal extension of two amino acids underlined. (b) Chromium-release assay of the lysis of autologous fibroblasts transfected twice with control mirna, or and then, 24 h later, infected for 1 h with HCMV RV798 (multiplicity of infection, 2) and, at 48 h after infection, collected and incubated for 2 h in a 37 C in a CO 2 incubator with 51 Cr, then used as target cells ( ) incubated with CTLs (clone identification at top) as effector cells at an effector/target ratio of 1 ( CTLs), 5 ( CTLs), 2.5 ( CTLs) or 1.25 ( CTLs). P <.1, versus control mirna (two-tailed Student s t-test). Data are representative of three independent experiments (mean ± s.e.m.). interferes with MHC class I mediated antigen presentation by targeting ERAP1, a key enzyme involved in catalyzing the production of antigenic peptides in the endoplasmic reticulum. ERAP1 is critical in establishing immune responses to some viral epitopes. ERAP1 deficiency in mice affects the peptide pool during infection with mouse choriomeningitis virus 39 or lymphocytic choriomeningitis virus 37, as well as CD8 + T cell responses to antigens derived from lymphocytic choriomeningitis virus 38 or influenza virus 37. In ERAP1- deficient mice, many unstable MHC class I bound peptides, probably with N-terminal extension, are presented to the cell surface 8. In our experiments with HCMV-specific CD8 + CTL clones derived from HCMV-seropositive donors, the generation of four epitopes (UL23(34 42), IE1(88 96), UL28( ) and UL16(162 17)) among five HCMV epitopes tested was affected by ERAP1. On the basis of published studies of ERAP1- deficient mice and our findings here, we conclude that downregulation of ERAP1 influences the production of many HCMV-derived antigenic peptides during viral infection, which results in evasion of the recognition of viral antigen by CD8 + T cells during the host immune response. In addition to ERAP1, two other aminopeptidases involved in antigenic peptide processing, ERAP2 (L-RAP) and PLAP (IRAP), are expressed by human cells. ERAP2 can efficiently trim some kinds of epitopes that are distinct from ERAP1-trimmed epitopes 4, and it can physically bind to ERAP1 (ref. 6), which suggests that ERAP1 and ERAP2 might function in concert to produce a variety of MHC class I epitopes in the endoplasmic reticulum. However, not only is ERAP1 expressed more ubiquitously than ERAP2, but its overall expression is higher than that of ERAP2 in human cell lines 6,41. In addition, although PLAP can produce distinct antigenic peptides, it is specialized to localize to intracellular vesicles 42. Therefore, ERAP1 seems to be the main enzyme that trims antigenic peptides in the endoplasmic reticulum of human cells. In this context, nature immunology VOLUME 12 NUMBER 1 OCTOBER

7 211 Nature America, Inc. All rights reserved. it is plausible that HCMV has evolved a strategy to preferentially target ERAP1 rather than other aminopeptidases to inhibit the presentation of epitopes derived from viral gene products under the selective pressure of the host immune response. Of course, it is still possible that the expression or function of ERAP2 and PLAP might also be regulated by HCMV or other viruses for their survival. For a virus to survive, it is crucial that the virus block the presentation of viral antigenic peptides to CD8 + T cells at the immediate-early or early stage of infection. The US2 US11 region of the HCMV genome encodes immunoevasion proteins that can inhibit antigen presentation by MHC class I (ref. 1). The physiological importance of the US2 US11 region has been emphasized by the finding that superinfection of rhesus cytomegalovirus infected rhesus macaques requires evasion of CD8 + T cell immunity by homologs of the HCMV glycoproteins US2, US3, US6 and US11 (ref. 43). Given that mir-us4 is expressed together with these immunoevasion proteins at the early stage of infection, mir-us4 probably acts together and simultaneously with these immunoevasion proteins for evasion of the immune response. Unlike the US glycoproteins, viral mirnas are not immunogenic, and thus viral mirnas might offer better options for viruses to establish lifelong latent infection. Our work should expand the scope of immunoevasive gene products beyond viral glycoproteins to viral mirnas. The identification of cooperation between viral mirna and viral proteins in immunoevasion emphasizes need to incorporate additional determinants in the design of vaccines against HCMV. Methods Methods and any associated references are available in the online version of the paper at Note: Supplementary information is available on the Nature Immunology website. Acknowledgments We thank M. Tsujimoto (RIKEN, Wako) for recombinant human ERAP1 cdna and polyclonal antibody to ERAP1; P. van Endert and L. Saveanu (Institut National de la Santé et de la Recherché Médicale, Paris) for monoclonal antibody 6H9 to ERAP1; I. York (Michigan State University) and K. Rock (University of Massachusetts Medical School) for pug1-based vectors and technical help; N. Shastri (University of California, Berkeley) for the B3Z T cell hybridoma; T. Shenk (Princeton University) for pad/cre BAC; and C.-Y. Kang and T.-K. Kim for technical help. Supported by the Creative Research Initiatives Program of the Ministry of Science and Technology and the Korea Science and Engineering Foundation (K.A.), the Brain Korea 21 project of the Ministry of Education, Science and Technology of the Republic of Korea (S.K., S.L. and D.K.) and the US National Institutes of Health (CA1829 and AI53193 to S.R.R.). AUTHOR CONTRIBUTIONS S.K., D.K., Y.-K.K. and V.N.K. designed and did biochemical and cell biological experiments; J.S. and Y.K. did microarray experiments; S.L., Y.-E.K. and J.-H.A. generated HCMV mutants; I.E. and E.S. did in vitro ERAP1 trimming assays; S.R.R. cloned the HCMV-specific CTLs; and S.K. and K.A. designed the overall study and wrote the paper. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. Published online at Reprints and permissions information is available online at reprints/index.html. 1. Hansen, T.H. & Bouvier, M. MHC class I antigen presentation: learning from viral evasion strategies. Nat. Rev. Immunol. 9, (29). 2. Vyas, J.M., Van der Veen, A.G. & Ploegh, H.L. The known unknowns of antigen processing and presentation. Nat. Rev. Immunol. 8, (28). 3. Nguyen, T.T. et al. Structural basis for antigenic peptide precursor processing by the endoplasmic reticulum aminopeptidase ERAP1. Nat. Struct. Mol. Biol. 18, (211). 4. Haroon, N. & Inman, R.D. Endoplasmic reticulum aminopeptidases: Biology and pathogenic potential. Nat. Rev. Rheumatol. 6, (21). 5. Saric, T. et al. An IFN-γ-induced aminopeptidase in the ER, ERAP1, trims precursors to MHC class I presented peptides. Nat. Immunol. 3, (22). 6. Saveanu, L. et al. Concerted peptide trimming by human ERAP1 and ERAP2 aminopeptidase complexes in the endoplasmic reticulum. Nat. Immunol. 6, (25). 7. Kanaseki, T. & Shastri, N. Endoplasmic reticulum aminopeptidase associated with antigen processing regulates quality of processed peptides presented by MHC class I molecules. J. Immunol. 181, (28). 8. Hammer, G.E., Gonzalez, F., James, E., Nolla, H. & Shastri, N. In the absence of aminopeptidase ERAAP, MHC class I molecules present many unstable and highly immunogenic peptides. Nat. Immunol. 8, (27). 9. Chang, S.C., Momburg, F., Bhutani, N. & Goldberg, A.L. The ER aminopeptidase, ERAP1, trims precursors to lengths of MHC class I peptides by a molecular ruler mechanism. Proc. Natl. Acad. Sci. USA 12, (25). 1. Fabian, M.R., Sonenberg, N. & Filipowicz, W. Regulation of mrna translation and stability by micrornas. Annu. Rev. Biochem. 79, (21). 11. Pfeffer, S. et al. Identification of virus-encoded micrornas. Science 34, (24). 12. Cullen, B.R. Viral and cellular messenger RNA targets of viral micrornas. Nature 457, (29). 13. Pfeffer, S. et al. Identification of micrornas of the herpesvirus family. Nat. Methods 2, (25). 14. Grey, F. et al. Identification and characterization of human cytomegalovirus-encoded micrornas. J. Virol. 79, (25). 15. Grey, F., Meyers, H., White, E.A., Spector, D.H. & Nelson, J. A human cytomegalovirusencoded microrna regulates expression of multiple viral genes involved in replication. PLoS Pathog. 3, e163 (27). 16. Grey, F. et al. A viral microrna down-regulates multiple cell cycle genes through mrna 5 UTRs. PLoS Pathog. 6, e1967 (21). 17. Stern-Ginossar, N. et al. Host immune system gene targeting by a viral mirna. Science 317, (27). 18. Stern-Ginossar, N. et al. Analysis of human cytomegalovirus-encoded microrna activity during infection. J. Virol. 83, (29). 19. Rehm, A. et al. Human cytomegalovirus gene products US2 and US11 differ in their ability to attack major histocompatibility class I heavy chains in dendritic cells. J. Virol. 76, (22). 2. Machold, R.P., Wiertz, E.J., Jones, T.R. & Ploegh, H.L. The HCMV gene products US11 and US2 differ in their ability to attack allelic forms of murine major histocompatibility complex (MHC) class I heavy chains. J. Exp. Med. 185, (1997). 21. Gruhler, A., Peterson, P.A. & Fruh, K. Human cytomegalovirus immediate early glycoprotein US3 retains MHC class I molecules by transient association. Traffic 1, (2). 22. Lehner, P.J., Karttunen, J.T., Wilkinson, G.W. & Cresswell, P. The human cytomegalovirus US6 glycoprotein inhibits transporter associated with antigen processing-dependent peptide translocation. Proc. Natl. Acad. Sci. USA 94, (1997). 23. Park, B., Spooner, E., Houser, B.L., Strominger, J.L. & Ploegh, H.L. The HCMV membrane glycoprotein US1 selectively targets HLA-G for degradation. J. Exp. Med. 27, (21). 24. Gustems, M. et al. Regulation of the transcription and replication cycle of human cytomegalovirus is insensitive to genetic elimination of the cognate NF-kappa B binding sites in the enhancer. J. Virol. 8, (26). 25. Wu, J., O Neill, J. & Barbosa, M.S. Transcription factor Sp1 mediates cell-specific trans-activation of the human cytomegalovirus DNA polymerase gene promoter by immediate-early protein IE86 in glioblastoma U373MG cells. J. Virol. 72, (1998). 26. Chekulaeva, M. & Filipowicz, W. Mechanisms of mirna-mediated post-transcriptional regulation in animal cells. Curr. Opin. Cell Biol. 21, (29). 27. Lewis, B.P., Burge, C.B. & Bartel, D.P. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microrna targets. Cell 12, 15 2 (25). 28. Rehmsmeier, M., Steffen, P., Hochsmann, M. & Giegerich, R. Fast and effective prediction of microrna/target duplexes. RNA 1, (24). 29. Miranda, K.C. et al. A pattern-based method for the identification of MicroRNA binding sites and their corresponding heteroduplexes. Cell 126, (26). 3. Wang, W.X., Wilfred, B.R., Hu, Y., Stromberg, A.J. & Nelson, P.T. Anti-Argonaute RIP-Chip shows that mirna transfections alter global patterns of mrna recruitment to microribonucleoprotein complexes. RNA 16, (21). 31. Nonne, N., Ameyar-Zazoua, M., Souidi, M. & Harel-Bellan, A. Tandem affinity purification of mirna target mrnas (TAP-Tar). Nucleic Acids Res. 38, e2 (21). 32. Moore, L.M. & Zhang, W. Targeting mir-21 in glioma: a small RNA with big potential. Expert Opin. Ther. Targets 14, (21). 33. Peters, L. & Meister, G. Argonaute proteins: mediators of RNA silencing. Mol. Cell 26, (27). 34. York, I.A., Brehm, M.A., Zendzian, S., Towne, C.F. & Rock, K.L. Endoplasmic reticulum aminopeptidase 1 (ERAP1) trims MHC class I-presented peptides in vivo and plays an important role in immunodominance. Proc. Natl. Acad. Sci. USA 13, (26). 35. Karttunen, J., Sanderson, S. & Shastri, N. Detection of rare antigen-presenting cells by the lacz T-cell activation assay suggests an expression cloning strategy for T-cell antigens. Proc. Natl. Acad. Sci. USA 89, (1992). 99 VOLUME 12 NUMBER 1 OCTOBER 211 nature immunology

8 36. Shastri, N. & Gonzalez, F. Endogenous generation and presentation of the ovalbumin peptide/kb complex to T cells. J. Immunol. 15, (1993). 37. Firat, E. et al. The role of endoplasmic reticulum-associated aminopeptidase 1 in immunity to infection and in cross-presentation. J. Immunol. 178, (27). 38. Yan, J. et al. In vivo role of ER-associated peptidase activity in tailoring peptides for presentation by MHC class Ia and class Ib molecules. J. Exp. Med. 23, (26). 39. Blanchard, N. et al. Endoplasmic reticulum aminopeptidase associated with antigen processing defines the composition and structure of MHC class I peptide repertoire in normal and virus-infected cells. J. Immunol. 184, (21). 4. Tanioka, T. et al. Human leukocyte-derived arginine aminopeptidase. The third member of the oxytocinase subfamily of aminopeptidases. J. Biol. Chem. 278, (23). 41. Fruci, D. et al. Altered expression of endoplasmic reticulum aminopeptidases ERAP1 and ERAP2 in transformed non-lymphoid human tissues. J. Cell. Physiol. 216, (28). 42. Georgiadou, D. et al. Placental leucine aminopeptidase efficiently generates mature antigenic peptides in vitro but in patterns distinct from endoplasmic reticulum aminopeptidase 1. J. Immunol. 185, (21). 43. Hansen, S.G. et al. Evasion of CD8 + T cells is critical for superinfection by cytomegalovirus. Science 328, (21). 211 Nature America, Inc. All rights reserved. nature immunology VOLUME 12 NUMBER 1 OCTOBER

9 211 Nature America, Inc. All rights reserved. ONLINE METHODS Cell lines. U373MG cells, HEK293T cells, HFF and HFF-TEL cells were obtained from American Type Culture Collection. U373MG cells, HEK293T cells, HeLa cells expressing mouse H-2K b, HFF and HFF-TEL cells were cultured in Dulbecco s modified Eagle s medium (Gibco) supplemented with 1% (vol/vol) FBS, 2 mm l-glutamine and 5 U/ml of penicillin. Dermal fibroblast cells were generated from skin biopsies and were propagated in Waymouth s media (Gibco) with 15% (vol/vol) FBS, 2 mm l-glutamine and 5 U/ml of penicillin. B3Z 86/9.14 (B3Z) hybridoma T cells were cultured at 37 C in the presence of 5% CO 2 in RPMI-164 medium (Gibco) supplemented with 2 mm l-glutamine, 1 mm pyruvate, 5 mm β-mercaptoethanol, 1 U/ml of penicillin, 1 mg/ml of streptomycin and 1% (vol/vol) FBS. CTLs were cultured in HEPES-buffered RPMI medium supplemented with 1% (vol/vol) human AB serum, 4 mm l-glutamine, penicillin and streptomycin. Viral mutagenesis. Approximately 3 base pairs of the region flanking mir-us4 was cloned into the vector pcdna3.1 (Invitrogen). Six nucleotides in the hairpin of mir-us4 were mutated by site-directed PCR (primers, Supplementary Table 1). Mutagenesis of a bacterial artificial chromosome of the HCMV AD169 strain was done with a BAC Modification Kit (Gene Bridges). A kanamycin-selective cassette containing the US4 homologous arm was amplified by PCR (primers, Supplementary Table 1). The PCR products of mutant mir-us4 (~2-ng) were transfected by electroporation into Escherichia coli (containing pad/cre) for recombination. A revertant of the mir-us4 mutant bacterial artificial chromosome was generated according to the method described above. Mutagenesis was confirmed by DNA sequencing of the bacterial artificial chromosome. T cell cloning. CD8 + T cell clones specific for HCMV were prepared from peripheral blood mononuclear cells of donors as described 44. Approval for these studies was obtained from the Fred Hutchinson Cancer Research Center Institutional Review Board. Informed consent for the collection and analysis of blood samples was provided according to the Declaration of Helsinki. Chromium-release assay. CTL clones were assayed for cytotoxic activity in a chromium-release assay with 51 Cr-labeled fibroblasts as target cells infected for various times with HCMV RV798. Target cells were plated in triplicate at a density of cells per well in 96-well round-bottomed plates, and effector cells were added at various ratios of effector cells to target cells. After 6 h of incubation, supernatants were collected and release of 51 Cr was measured by counting of γ-irradiation. Specific CTL lysis was calculated by the following formula: percent release = ([specific release spontaneous release] / [total release spontaneous release]) 1. Spontaneous release of less than about 5% of the total release was observed in all assays. B3Z assay. B3Z assays were done as described 36. The β-galactosidase activity in B3Z cells was determined by with a BetaRed β-galactosidase Assay kit according to the manufacturer s instructions (EMD Biosciences). In vitro ERAP1 trimming assay. Human recombinant ERAP1 was expressed via the infection of insect cells with baculovirus as described 45. Active ERAP1 (with a C-terminal six-histidine tag) was collected and purified by affinity chromatography (nickel nitrilotriacetic acid column). Protein purity was confirmed by SDS-PAGE and activity was assessed by an established fluorigenic assay 46. Enzymatic assays were done as described 42. For rate calculations, the total surface area of the epitope peak was measured and calibrated relative to the control peptides for calculation of the amount of epitope produced. Constructs. The vector psuperretro (OligoEngine) was used for expression of sirna or mirna. The constructs sirna-gfp (control mirna), mir-us4, mir-us4(m), mir-us5-1,, a and b were generated according to the manufacturer s instructions (OligoEngine; primers, Supplementary Table 1). For luciferase assays, the 3 UTRs of ERAP1a and ERAP1b were amplified (primers, Supplementary Table 1) and the products of amplification were inserted downstream of the luciferase gene in the vector pgl3-cmv, a modified form of pgl3-basic Vector (Promega). The 3 UTRs of ERAP1a and ERAP1b were mutated by site-directed mutagenesis with Pfu DNA Polymerase (Stratagene). The pug1 Mock, pug1-ova8 and pug1-n5ova8 vectors have been described 34. Antibodies. Monoclonal antibody W6/32 recognizes the complex of MHC class I heavy chain and β 2 -microblogulin. Anti-GAPDH (LF-PA18) and anti-erp57 (YF-PA12134) were from AbFrontier. Anti-tapasin (ab13518) was from Abcam. Fluorescein isothiocyanate conjugated goat antibody to mouse immunoglobulin G ( ) and horseradish peroxidase conjugated goat antibody to mouse or rabbit immunoglobulin G ( and , respectively) were from Jackson ImmunoResearch Laboratories. Polyclonal antibody to ERAP1 was generated by cloning of truncated cdna encoding the C-terminal amino acids of ERAP1b into the vector pet28a (primers, Supplementary Table 1) and expression of the cloned DNA in E. coli. The recombinant protein was affinity-purified and used to raise rabbit anti-erap1. Microarray. RNA was assayed for quality control with a Bioanalyzer 21 (Agilent), then the RNA was labeled with a Low RNA Input Linear Amplification Kit Plus (Agilent) and hybridized together with labeled Universal Reference RNA (Stratagene) to a Human Genome Whole FourPlex Microarray (Agilent). Image files of arrays were scanned on GenePix 4B (Axon). Data were extracted with Feature Extraction Software (Agilent) and analyzed with GX software (Genespring). Quantitative real-time PCR. Total RNA was collected with TRIzol (Invitrogen). M-MLV Reverse Transcriptase was used for reverse transcription according to the manufacturer s protocol (Invitrogen). SYBR Premix Ex Taq (Perfect Real Time) was used for amplification and detection according to the manufacturer s protocol (TaKaRa Bio). Luciferase assay. HEK293T cells were seeded in six-well plates 1 d before transfection. For cotransfection, 1 ng firefly luciferase and 5 ng renilla luciferase reporter plasmids were transiently transfected into cells with 2 µg psuper vectors. After h, luciferase activity was measured with a Dual- Luciferase Reporter Assay System (Promega). Firefly luciferase activity was normalized to renilla luciferase activity. Flow cytometry. Cell surface expression of MHC class I was assessed with a FACSCalibur (Becton Dickinson Biosciences). Cells were collected, washed twice with cold PBS containing 1% (vol/vol) BSA and incubated for 1 h at 4 C with a saturating concentration of antibodies. Cells were then washed twice with cold PBS containing 1% (vol/vol) BSA and stained for 3 min at 4 C with fluorescein isothiocyanate conjugated goat antibody to mouse immunoglobulin G ( ; Jackson ImmunoResearch Laboratories). A total of 1, gated events were collected and analyzed with CellQuest software (Becton Dickinson Biosciences). Immunoblot analysis, RNA blot and RNase-protection assay. Cells were lysed for 1 h at 4 C with 1% (vol/vol) Nonidet P-4 in PBS with a proteaseinhibitor cocktail, then were separated by SDS-PAGE, transferred onto nitrocellulose membranes, blocked and probed overnight with the appropriate antibodies. Membranes were washed and incubated for 1 h with horseradish peroxidase conjugated goat antibody to mouse or rabbit immunoglobulin G ( or , respectively; Jackson ImmunoResearch Laboratories). Proteins on immunoblots were visualized with ECL Western Blotting Substrate (Pierce). RNA blots were done according to a published study 47. A mirvana mirna Detection Kit was used for RNase-protection assays according to the manufacturer s recommendations (Ambion). 44. Manley, T.J. et al. Immune evasion proteins of human cytomegalovirus do not prevent a diverse CD8 + cytotoxic T-cell response in natural infection. Blood 14, (24). 45. Evnouchidou, I. et al. The internal sequence of the peptide-substrate determines its N-terminus trimming by ERAP1. PLoS ONE 3, e3658 (28). 46. Niles, A.L., Moravec, R.A. & Riss, T.L. In vitro viability and cytotoxicity testing and same-well multi-parametric combinations for high throughput screening. Curr Chem Genomics 3, (29). 47. Lee, Y. et al. The nuclear RNase III Drosha initiates microrna processing. Nature 425, (23). nature immunology doi:1.138/ni.297