HNRNPR Regulates the Expression of Classical and Nonclassical MHC Class I Proteins

Major histocompatibility complex class I proteins are one of the major immune system modulators. They present peptides to T cells and activate CTLs of the adaptive immune system, and, in contrast, serve as inhibitory ligands for NK cell inhibitory receptors (1). NK cells are innate lymphocytes that, besides supporting developmental processes such as fetal growth during pregnancy (2), are best known for their ability to discriminate between self and altered self by killing virally infected, transformed, and damaged cells. NK cell activity is governed by integrating signals derived from a panel of activating and inhibitory receptors. The inhibitory receptors are expressed stochastically on NK cells, and their main inhibitory ligands are MHC class I molecules (3, 4). There are several families of MHC class I inhibitory receptors; among them, the KIR2DL family shows a unique affinity to HLA-C subtypes. Practically all of the HLA-C alleles can be divided into two groups, which are recognized by KIR2DL1 or by KIR2DL2 inhibitory NK receptors. Generally, the KIR2DL1 receptor recognizes HLA-C subtypes (C2 group), having lysine at position 80, whereas KIR2DL2 receptors recognize HLA-C subtypes (C1 group), having asparagine at position 80 (5).

MHC class I proteins are divided into classical and nonclassical. In humans, the classical proteins are named HLA-A, -B, and -C, each containing hundreds of different alleles. The nonclassical proteins are significantly less polymorphic and include HLA-E, -F, -G, and HFE (6).

HLA-G is one of the most interesting nonclassical MHC class I proteins. It has a very unique expression pattern, in which it is upregulated on cancer cells, whereas on normal tissues its expression is mainly restricted to the extravillous cytotrophoblasts of the placenta, which is of fetal origin (7). Although HLA-G mRNA can be found in several tissues, HLA-G protein is absent, suggesting for regulation of HLA-G expression at the mRNA level.

The transcription processes regulating MHC gene expression were broadly investigated (8), and a few posttranscription regulators were discovered as well (9, 10). However, whether RNA-binding proteins (RBPs) control the expression of classical and nonclassical MHC class I proteins is largely unexplored.

RBPs play a part in every aspect of RNA biogenesis, including transcription, pre-mRNA splicing, polyadenylation, RNA modification, transport, localization, translation, turnover, and immune activities (11–15). One major family of RBPs is the heterogeneous nuclear ribonucleoproteins (HNRNPs), which are among the most abundant proteins in the eukaryotic nucleus, taking part in processing of precursor mRNA (16). In this study, we identify a member of this family, heterogeneous nuclear ribonucleoprotein R (HNRNPR), as a general regulator of classical and nonclassical MHC class I expression.

JEG-3, U87, and Mel1074 cells were maintained in DMEM. The 721.221, Jurkat, Bjab, and U937 cells were maintained in RPMI 1640 medium. All media were supplemented with 10% FCS.

The interactions between RNA and RNA-binding proteins were analyzed by RNA affinity purification, as previously described (17). In short, the 3′ untranslated regions (UTRs) of HLA-G (sense and antisense orientation, National Center for Biotechnology Information Reference Sequence: NT_167249.2) and a UTR of a control gene with similar length and GC content (GPR112) were cloned into the pBSII plasmid using the primers (restriction sites are underlined), as follows: HLA-G sense forward (NotI), 5′-ACGCGGCCGCATTGAAAGGAGGGAGCTA-3′ and HLA-G sense reverse (XbaI), 5′-TGTCTAGAAAAGTTCTCATGTCTTCCATTTAT-3′; HLA-G antisense forward (XbaI), 5′-ACTCTAGAATTGAAAGGAGGGAGCTA-3′ and HLA-G antisense reverse (NotI), 5′-GTGCGGCCGCAAAGTTCTCATGTCTTCCATTTAT-3′; and GPR112 forward (NotI), 5′-TAGCGGCCGCTTTGTGAAGTTGTGCCTAAT-3′ and GPR112 reverse (XbaI), 5′-AGTCTAGAAGGAGAATTATTCTGACTTTAATATTTATC-3′.

In vitro transcription into RNA was performed using the MEGAscript T7 transcription kit (Life Technologies) after linearization of the plasmids with PspOMI restriction enzyme (Thermo Scientific [Fermentas]). About 10% of totally incorporated UTPs were biotin-16-UTPs (GE Healthcare). The biotinylated RNAs were coupled to streptavidin-Sepharose beads (GE Healthcare) and incubated with cytoplasmic extracts prepared from 80% confluent JEG-3 cells overnight at 4°C. After purification and elution of proteins that bound specifically to the RNAs, a SDS gel analysis was performed and specific bands were detected with Coomassie brilliant blue G-250 (Sigma-Aldrich). Specific bands were excised and analyzed by mass spectrometry. Analysis was performed by the Smoler Proteomics Center (Technion, Haifa, Israel).

The various RBP-knockdown (KD) vectors and scrambled vector are in the pLKO.1-puro plasmids and were purchased from Sigma-Aldrich. Transduced JEG-3, 721.221, Jurkat, Bjab, and U937 cells were grown in the presence of 1 μg/ml, 2 μg/ml, 1 μg/ml, 2 μg/ml, and 7 μg/ml puromycin, respectively. Cloning of the HNRNPR and HLA-G 3′ UTRs containing 14-bp deletion and insertion was performed into the pHAGE-DsRED(−)-eGFP(+) lentiviral vector, which also contains GFP. Lentiviruses were generated in 293T cells using a transient three-plasmid transfection protocol, as previously described (18). Transduction efficiency into the cells was assessed by GFP expression, and only cell populations with >90% efficiency were used for experiments. Primer sequences are as follows: HNRNPR forward (NotI), 5′-ATGCGGCCGCACCATGAAGACCTACAGGCAGAG-3′ and HNRNPR reverse (AgeI), 5′-TGACCGGTCTAATGGTGATGGTGATGGTGCTTCCACTGTTGCCCA-3′; HLA-G 14-bp deletion and insertion forward (NotI), 5′-GCGGCGGCCGCGCCGCCACCATGGTGGTCATGGCG-3′; and HLA-G 14-bp deletion and insertion reverse (AgeI), 5′-GCGACCGGTAAAGTTCTCATGTCTTCCATTTA-3′. The 14-bp deletion mutations in HLA-G 3′ UTR 14-bp insertion were generated by PCR-based site-directed mutagenesis using the 5′ primer 5′-AGTGGCAAGTCCCTTTGT-3′ and the 3′ primer 5′-ACAAAGGGACTTGCCACT-3′.

The assay was performed as previously described (19). Briefly, Mel1074 cells expressing HIS-tagged HNRNPR cells were used for the validation of the HNRNPR–HLA–mRNA interactions. Total cell lysate was prepared using an ice-cold lysis buffer. The lysate was precleared using protein A/G PLUS-agarose beads sc-2003 (Santa Cruz Biotechnology). Then the cleared lysate was incubated at 4°C overnight with HIS-tag Ab (R&D Systems; catalogue MAB050) or Ig control (eBioscience; catalogue 14-4714-85). Protein A/G PLUS-agarose beads were then added for an additional 4 h. After several washing steps, RNA was isolated using TRI reagent (T9424; Sigma-Aldrich), and the presence of the HLA transcripts was detected via quantitative RT-PCR (qRT-PCR). Enrichment was calculated relative to the levels of the specific transcript in the Ig control samples.

Total RNA was extracted with TRI reagent (T9424; Sigma-Aldrich) and was treated with Turbo-DNase (Ambion), and a poly(A) tail was added using the poly(A) kit (Ambion). For generation of cDNA libraries, the Moloney murine leukemia virus reverse transcriptase (Invitrogen) was used for reverse transcription (according to the manufacturer’s instructions), in the presence of an adapter primer. Detection of the various transcripts was performed with quantitative real-time PCR (see below).

For quantitative real-time PCR, freshly prepared cDNAs were used for SYBR Green-based detection in a QuantStudio 12k Flex real-time PCR cycler (Life Technologies) with primers targeting GAPDH, MICB, TAP1, TAP2, Tapasin, HLA-A, HLA-B, HLA-C, and HLA-G, as follows: GAPDH forward, 5′-GAGTCAACGGATTTGGTCGT-3′; GAPDH reverse, 5′-GATCTCGCTCCTGGAAGATG-3′; MICB forward, 5′-CTGCTGTTTCTGGCCGTC-3′; MICB reverse, 5′-ACAGATCCATCCTGGGACAG-3′; TAP1 forward, 5′-TCAGGGCTTTCGTACAGGAG-3′; TAP1 reverse, 5′-TCCGGAAACCGTGTGTACTT-3′; TAP2 forward, 5′-ACTGCATCCTGGATCTCCC-3′; TAP2 reverse, 5′-TCGACTCACCCTCCTTTCTC-3′; Tapasin forward, 5′-AAGCTCAAGTCCAGCAGAGC-3′; Tapasin reverse, 5′-CAGCAGGAGCCTGTTCTCAT-3′; HLA-A forward, 5′-GGGTCATATGTGTCTTGGGG-3′; HLA-A reverse, 5′-GCAGTTGAGAGCCTACCTGG-3′; HLA-B forward, 5′-CTCATGGTCAGAGATGGGGT-3′; HLA-B reverse, 5′-TCCGCAGATACCTGGAGAAC-3′; HLA-C forward, 5′-GTGGCCTCATGGTCAGAGAG-3′; HLA-C reverse, 5′-TCCGCAGATACCTGGAGAAC-3′; HLA-G forward, 5′-AGTCAAAGACAGGGTGGTGG-3′; and HLA-G reverse, 5′-GGAGTGGCTCCACAGATACC-3′.

KIR2DL1-Ig and KIR2DL2-Ig fusion proteins were generated in 293T cells and were purified on a protein G column, as previously described (20). The fusion proteins used in this work were regularly assayed by SDS protein gels to ensure the proteins were not degraded. Protein purity of all Ig fusion proteins used in this study was ∼100%.

Lysates of the cells were prepared, and SDS gel electrophoresis was executed. Proteins were transferred onto a nitrocellulose membrane with the tank blot procedure, and specific protein bands were detected using Abs detecting HNRNPR (sc-16541, Santa Cruz Biotechnology [1:200]; 15018-1-AP, Proteintech [1:1000]), DDX3X (09-860, Millipore [1:1000]), PABPC4 (14960-1-AP, Proteintech [1:1000]), or GAPDH (SC-32233, Santa Cruz Biotechnology [1:1000]) as loading control. All was diluted in 5% BSA in PBS. Chemiluminescence caused by detection Ab-linked HRP (Jackson ImmunoResearch Laboratories) was detected.

Flow cytometry for the following Abs was used: HLA-G (purchased from AbD Serotec; catalogue MCA2044), HLA-A*02 (mAb clone BB7.2), HLA-B*07 (mAb clone BB7.1), MHC class I (mAb clone W6/32), ULBP1 (mAb clone 170818), ULBP2 (mAb clone 165903), ULBP3 (mAb clone 166510), ULBP4 (mAb clone 6E6), ICAM-1 (mAb clone HCD54), CD48 (mAb clone 4H9), MICA (mAb clone 159227), MICB (mAb clone 236511), KIR2DL1, KIR2DL2, and CD56 (purchased from BioLegend; catalogue 339504, 312604, and 318310, respectively). Staining was performed with 0.2 μg mAbs per 100,000 cells. KIR2DL1-Ig and KIR2DL2-Ig fusion protein staining was performed with concentrations ranging between 1 μg and 10 μg/well. Binding was detected by the appropriate secondary Ab (Alexa Flour 647 catalogue 115-606-062 or allophycocyanin catalogue 709-136-098). All staining was analyzed by FACS using the CellQuest software.

NK cells were isolated from healthy donors using a MACS separation kit (Miltenyi Biotec) and grown in the presence of IL-2 (PeproTech). Target cells were incubated overnight in the presence of ³⁵S-methionine added to a methionine-free media (Sigma-Aldrich). NK cells were incubated or not with 1 μg anti-KIR2DL2 Ab for blocking for 1 h on ice and then added to the target cells. The level of ³⁵S release was measured after 5 h of incubation with effectors using a beta counter TopCount (Packard).

The locked nucleic acid oligonucleotide 300500 (/5TYE665/ sequence 5′-TCGCTCTGGTTGTAGTAGC-3′; Exiqon) probe that recognizes all MHC class I proteins was used. The assay was performed according to the manufacturer’s instructions for in situ detection of mRNA in fixed cells. Results were analyzed using Nikon 90i confocal microscope and photographed with Nikon D-eclipse C1 camera.

Little is known about whether RBPs regulate class I MHC protein expression. Comparison of the 3′ UTRs of various classical and nonclassical MHC class I proteins revealed that the 3′ UTRs of the MHC class I proteins are highly similar except for the 3′ UTR of HLA-F (Fig. 1A). This suggests that MHC class I proteins might be regulated by shared, posttranscriptional regulation mechanisms.

To identify RBPs that regulate MHC class I expression, we transcribed in vitro the 3′ UTR of HLA-G, using biotinylated oligonucleotides. We then incubated the biotinylated 3′ UTR of HLA-G with protein lysates that were derived from the HLA-G–expressing cell line JEG-3 (21), followed by incubation with streptavidin beads. The unbound proteins were washed, and the bound proteins were eluted and run using SDS-PAGE. Specific protein bands, which did not appear in the control experiments, were excised and sent to mass spectrometry analysis for identification (Fig. 1B). Three candidates were chosen from the mass spectrometry results (Fig. 1C), based on their statistical significance score and known function: DDX3X, a RNA helicase; PABPC4, a poly(A)-binding protein; and HNRNPR, which associates with pre-mRNAs (22).

To test whether the identified RBPs indeed regulate MHC class I expression, the three candidate RBPs were knocked down in JEG-3 cells using short hairpin RNA. KD efficiency was verified using Western blot (WB) assays (Fig. 2A). MHC class I expression on the surface of various JEG-3 KD cells was assayed by FACS using pan anti-MHC class I mAb W6/32. Reduced MHC class I expression was observed only in cells with HNRNPR KD, whereas little or no alteration of MHC class I expression was observed in the other KDs (Fig. 2B). These findings led us to conclude that HNRNPR, but not DDX3 and PABPC4, is a positive regulator of MHC class I expression. To further confirm this, we screened various tumor cell lines for HNRNPR expression using WB (Fig. 2C). Then, we overexpressed HNRNPR in the U87 cell line that is negative for HNRNPR expression (Fig. 2C) and observed elevation of MHC class I (Fig. 2D).

In parallel, we also stained the various JEG-3 KD cells for HLA-G expression using an anti–HLA-G–specific mAb and observed reduced expression of HLA-G following HNRNPR KD (Fig. 3A). HLA-G contains in its 3′ UTR 14-bp insertion polymorphism that was shown to play a role in regulation of HLA-G expression (23, 24). To test whether a 14-bp insertion/deletion plays a role in HNRNPR regulation, we expressed HLA-G with its 3′ UTR that includes either the 14-bp insertion or deletion in 721.221 cells. We then knocked down HNRNPR and tested HLA-G expression. As can be seen in Fig. 3B, KD of HNRNPR led to reduced HLA-G expression, irrespective of whether the 3′ UTR included the 14-bp insertion/deletion.

To confirm that the KD HNRNPR is specific, we stained Bjab cells for an array of NK ligands before and after KD of HNRNPR. We used Bjab cells and not Jeg3 cells, because they express many NK cell ligands, whereas Jeg3 cells are negative for expression of most NK ligands. No change in NK ligand expression was observed (Fig. 3C). In contrast, HNRNPR KD led to a decrease in MHC class I expression (Fig. 3C).

We next analyzed a Gene Expression Omnibus dataset (GDS2241) of various choriocarcinoma cell lines (25) and observed a direct correlation between HNRNPR expression and the expression of HLA-A, HLA-B, HLA-C, and HLA-G (Fig. 4A). To further investigate whether HNRPNR endogenously binds the 3′ UTRs of MHC class I proteins, we performed RNA immunoprecipitation assays. Because none of the commercially available anti-HNRNPR mAbs were able to precipitate this protein, we cloned HNRNPR fused to a HIS-tag. Cloning was done into lentivirus vector that also expresses GFP. We then expressed the his-tag HNRNPR in Mel1074 cells that express little or no endogenous HNRNPR (Fig. 2C). Expression was verified by FACS, using GFP as indicator for the transduction efficiency (Fig. 4B) and by WB (Fig. 4C). We next precipitated HNRNPR using an anti–HIS-tag Ab and performed qRT-PCR on the HNRNPR-bound RNA. A significant enrichment for both HLA-A and HLA-G mRNAs was detected (8- and 5.5-fold, respectively; Fig. 4D, 4E). These combined results indicate that HNRNPR endogenously binds the mRNA of classical and nonclassical (HLA-G) MHC class I.

To test whether HNRNPR affects the expression of each of the classical MHC class I proteins, HLA-A, HLA-B, and HLA-C, we knocked down HNRNPR in BJAB and Jurkat cells that express HLA-A*02 and HLA-B*07, respectively (26, 27). Staining of the appropriate cells with specific Abs against HLA-A*02 and HLA-B*07 revealed that both proteins were downregulated in HNRNPR KD cells (Fig. 5A, 5B).

To the best of our knowledge, no reliable commercial anti–HLA-C–specific mAb is available. Thus, to test whether HNRNPR affects HLA-C expression, we used a fusion protein composed of the extracellular portion of KIR2DL2 fused to human IgG1, named KIR2DL2-Ig. As mentioned in the introduction, HLA-C proteins can be divided into two groups in terms of NK receptor recognition. The inhibitory receptor KIR2DL1 recognizes members of the C2 group, such as, HLA-C*02,*04,*05, and *06, whereas KIR2DL2 recognizes members of the C1 group, such as, HLA-C*01,*03, *07, and*08 (Fig. 5C). To test whether HNRNPR can affect HLA-C expression, we haplotyped the U937 cell line and determined that both HLA-C proteins expressed by this cell line are ligands for NK inhibitory receptor KIR2DL2 (Fig. 5D). When we knocked down HNRNPR in U937 cells, we observed a reduction in MHC class I protein expression (from 1232.51 to 804.83 median fluorescent intensity; Fig. 5E).

Next, we used the KIR2DL2-Ig protein (which recognizes HLA-C*01 and *07; Fig. 5C) (28) to stain the U937 cells. Although the control U937 cells were stained with KIR2DL2-Ig, this staining was abolished in the HNRNPR KD cells (Fig. 5F), indicating that HNRNPR also affects HLA-C expression.

To test whether reduced HLA-C expression observed in HNRNPR KD cells will affect NK cell function, we isolated NK clones expressing the KIR2DL2 receptor. These clones were identified by double staining for CD56 together with anti-KIR2DL1 or anti-KIR2DL2 mAbs (Fig. 6A). Three of the abovementioned clones were then used in killing assays with two targets: wild-type U937 cells and HNRNPR KD U937 cells. As can be seen in Fig. 6B, we observed an increase in killing of HNRNPR KD U937 cells. To demonstrate that the increased killing is due to HLA-C downregulation, we blocked the KIR2DL2 receptor of NK clone D using a specific Ab and observed that the killing of the wild-type cells was restored to the level observed in the HNRNPR KD cells. In contrast, blocking of KIR2DL2 receptor on the HNRNPR KD cells had no effect (Fig. 6C).

Finally, to determine the mechanism by which HNRNPR positively regulates MHC class I expression, we performed a mRNA stability assay. In this procedure, cells were treated with the transcription inhibitor, actinomycin D, and the level of mRNA degradation was determined using qRT-PCR at two time points. We found that the mRNAs of all HLAs were less stable in HNRNPR KD cells compared with the controls (Fig. 7A), indicating that HNRNPR stabilizes the mRNAs of MHC class I molecules. HLA-A seemed to be more stable than the other MHC class I proteins, but it was still less stable under HNRNPR KD conditions. The RNA stability of MICB, which was set as control, was not affected by KD of HNRNPR (Fig. 7A). Further conformation of the specificity of the HNRNPR was the observation that the RNA expression of components of the MHC class I processing/loading pathway such as TAP1, TAP2, and Tapasin was not affected by the HNRNPR KD (Fig. 7B).

Finally, to determine whether some of the HNRNPR stabilization mechanism involves mislocalization of the MHC class I mRNA, we performed RNA fluorescence in situ hybridization assays. As can be seen, the majority of the MHC class I mRNA is perinuclear and the KD of HNRNPR, as expected, decreased the mRNA expression but did not alter its localization (Fig. 7C).

The MHC class I proteins can be divided into classical and nonclassical proteins. Although the classical HLA-A, -B, and -C proteins, and some of the nonclassical ones, such as HLA-E, are expressed on all nucleated cells, other nonclassical MHC class I proteins, such as HLA-G, manifest a restricted expression pattern (6). Despite years of investigation, the factors controlling the ubiquitous expression of classical MHC class I proteins and those controlling the expression of HLA-G are far from being completely understood. Discovering these factors is essential, as it may lead to a better understanding of MHC class I deficiencies (29) and MHC manipulation by cancer or viruses (1, 30).

To identify RBP candidates that might regulate MHC class I expression, we used a RNA-AP method and identified HNRNPR. We discovered that HNRNPR is a general positive regulator of classical MHC class I proteins and that it controls HLA-G expression irrespective of whether the HLA-G 3′ UTR contains 14-bp deletion/insertion. We verified that HNRNPR endogenously binds to the 3′ UTRs of MHC class I and also observed a strong positive correlation between HNRNPR expression and all of the HLA subtype expression using a Gene Expression Omnibus dataset of choriocarcinoma cell lines, which may hint at a functional effect of HNRNPR in cancer.

In the Gene Expression Omnibus database, HLA-C seems to be less affected; however, in our stabilization assays or in the KD it was very efficiently downregulated. Therefore, at the present we cannot determine whether some HLA alleles are more susceptible to HNRNPR regulation.

The binding site of HNRNPR is unknown, but the binding site of its closest relatives, HNRNPQ and HNRNPA/B, was suggested to be a certain triple RNA-recognition motif (RRMx3) (31, 32). These motifs are not found in the 3′ UTR of the MHC class I proteins. Therefore, it is still possible that HNRNPR indirectly affects MHC class I expression. We think, however, that this is unlikely because other components of the MHC class I processing pathway did not appear in the mass spectrometry data of the RNA-AP results and because no change was observed in the RNA levels of some MHC class I processing components in the HNRNPR KD.

Thus, HNRNPR is a general positive regulator of both classical and nonclassical MHC class I. Indeed, the 3′ UTR of classical MHC class I molecules and HLA-G share a striking resemblance. It therefore seems as if an evolutionary conservation maintained a high degree of identity of the 3′ UTR regulatory sequence of the classical MHC class I proteins and HLA-G. Because HLA-E carries the same resemblance, we predict that HNRNPR would influence its expression as well. In contrast, the 3′ UTR of HLA-F is considerably different that all other UTRs, and therefore it does not carry that resemblance and most probably would not be affected by HNRNPR.

It is expected that HNRNP interacts with additional mRNAs, similar to probably all other RBPs that bind several mRNA (33). Still, we show in this work that other NK ligands are not affected by HNRNPR and the expression of several components of the class I processing pathway are also not affected. Furthermore, HNRNPR KD, which led to reduced expression of specific HLA-C subtypes, led to an increased killing by NK cells.

The endogenous mechanisms controlling HNRNPR expression are currently unknown. In many types of transformed and virally infected cells, downregulation of MHC class I expression occurs to evade the adaptive immune system (1, 30). It will be interesting and important to investigate in the future whether virus infection leads to reduced HNRNPR expression, and consequently to a reduction of MHC class I expression, and subsequent escape from CTL attack. Conversely, downregulation of MHC class I also exposes these cells to NK cell–mediated elimination, so HNRNPR can theoretically be manipulated in the opposite direction, to facilitate evasion from NK cells.

The authors have no financial conflicts of interest.