Vigilin Regulates the Expression of the Stress-Induced Ligand MICB by Interacting with Its 5′ Untranslated Region

Natural killer cells are part of the innate immune system, and were recently classified as group 1 innate lymphoid cells (1–3). NK cells secrete cytokines such as IFN-γ and TNF-α, and possess cytotoxic capabilities (4). NK cell activity is governed by the integration of signals received by germ-line encoded inhibitory and activating receptors (5, 6). The inhibitory receptors are stochastically expressed, and their main ligands are MHC class I molecules found on all nucleated cells (7–9). In contrast, the activating receptors are expressed by all NK cells, and their ligands are more diverse in identity and expression pattern (7). Activating ligands comprise both self and non-self sources and are derived from viruses, tumors, damaged and stressed cells, bacteria, fungi, and self-cells (10–17).

A key activating receptor found on all NK cells, NKG2D, binds stress-induced proteins from two families: the human-specific MHC Class I–related chain family (MICA and MICB), and the UL16-binding protein (ULBP) family (ULBP1-6) (18). The NKG2D-ligand interaction is imperative for NK activation against cells undergoing cellular stress, such as viral infection and tumor transformation (18, 19). The critical role of the NKG2D ligands in NK cell activation is essentially exemplified by their downregulation during viral infection and tumor transformation, as a means of immune evasion (19). Healthy cells do not regularly express these ligands; consequently, tight regulation of the NKG2D ligands is important (19).

Nevertheless, the mechanisms responsible for the expression of the NKG2D ligands are not well understood. On the transcriptional level, MICA and MICB were shown to have heat shock response elements in their promoters, occupied by the heat shock transcription factor 1 in heat shocked cells (20). However, the heat shock response element was not necessary for MICA or MICB induction during viral infection or cell proliferation (20). MICA, MICB, and ULBP1 promoters were shown to be bound by Sp-family transcription factors, necessary for their optimal induction in certain cases (20). Additional transcription factors, including NF-κB, p53, and AP-2, were suggested to bind MICA, ULBP1, and ULBP2 (21–23). At the posttranscriptional level, only MICA, MICB, and ULBP3 have been reported to be regulated by cellular and viral microRNAs (miRNAs) (24–30). Additional posttranscriptional regulation was demonstrated on MICB and ULBP2 expression, mediated by RNA-binding proteins (RBPs), which interact with their 3′ untranslated regions (UTRs) (31, 32). Whether the expression of the NKG2D ligands is controlled by RBPs that bind to the 5′UTR is currently unknown.

The 5′UTR region contains numerous regulatory elements, such as upstream open reading frames, secondary structures, and RBP binding sites, known to regulate gene expression (33, 34).

RBPs assemble on nascent and mature mRNAs, regulating all aspects of RNA biogenesis: transcription, splicing, polyadenylation, nuclear export, localization, translation, mRNA stability, and degradation (35). The dysregulation of RBPs can have detrimental consequences on gene expression, which can lead to diseases including cancer, neurodegenerative disorders, muscular atrophies, and autoimmunity (36).

In this study we tested whether RBPs bind to the 5′UTR and regulate the NKG2D ligand, MICB. Consequently, by applying an RNA-affinity purification (RNA-AP) method combined with mass spectrometry analyses we revealed a novel regulatory process of MICB, governed by an uncommon RBP, vigilin.

The cell lines used were RKO, a human colon carcinoma cell line (ATCC CRL-2577), HCT116, a human colorectal carcinoma cell line (ATCC CLL-247), 293T, a human embryonic kidney cell line (ATCC CRL-3216), and BCBL1, a human B cell lymphoma cell line (ATCC CRL-2294). The cells were cultivated at 37°C, >95% humidity and 5% CO₂ in DMEM (Sigma-Aldrich) supplemented with 10% heat inactivated FCS (Sigma-Aldrich).

The 5′UTR of MICB (5′UTR MICB) (sense and antisense orientations) was cloned into the pBSII plasmid using the following primers (restriction sites are underlined): sense forward (BamHI), 5′-CATGGATCCACTGGATAAGCGGTCGCTGAG-3′ and sense reverse (EcoRI), 5′-CATGAATTCGGCCCCTACGTCGCCACC-3′; antisense forward (EcoRI), 5′-CATGAATTCACTGGATAAGCGGTCGCTGAG-3′ and antisense reverse (BamHI), 5′-CATGGATCCGGCCCCTACGTCGCCACC-3′.

The primers were obtained (Sigma-Aldrich) and PCR was performed with cDNA prepared from Jeg3 cells as a template. A shuffled sequence of 5′UTR MICB was generated using http://www.bioinformatics.org/sms2/shuffle_dna.html. Forward (BamHI 5′-GATCCCGACGGGCGATTGTGCTGTTCGATGCAGTGGACCCCGTGACATGTGATCTATGACCAGGAGCTCGGGGGAAAGCTTGGGCTGTTCCGGAGTAGCGGAGGTACTGCGTGACACACG-3′) and reverse (EcoRI 5′-AATTCGTGTGTCACGCAGTACCTCCGCTACTCCGGAACAGCCCAAGCTTTCCCCCGAGCTCCTGGTCATAGATCACATGTCACGGGGTCCACTGCATCGAACAGCACAATCGCCCGTCGG-3′) sequences were ordered (Sigma-Aldrich), annealed, and cloned into pBSII.

Plasmids were subsequently linearized with SalI restriction enzyme (Thermo Scientific [Fermentas]) prior to in vitro transcription. In vitro transcription was performed with the Megascript T7 transcription kit (Ambion) according to the manufacturer’s instructions with modifications. Approximately 50% of the incorporated UTPs were Biotin-16-UTP (GE Healthcare).

The RNA affinity purification assay was performed as previously described (37). Cytoplasmic extract from Jeg3 cells was prepared by harvesting 40 15 cm plates that were ∼90% confluent. The cells were washed twice with cold PBS ×1, followed by resuspension in a five packed cell volume of hypotonic buffer [10 mM Hepes (pH 7.9), 1.5 mM MgCl₂, 10 mM KCl]. Cells were centrifuged and resuspended in a three packed cell volume, followed by 20 min incubation on ice. Cells were subsequently disrupted manually by a glass homogenizer, followed by repeated centrifugation of the supernatant (the cytoplasmic extract) for 15 min at 3300 × g until a clear supernatant solution was observed. The extract was supplemented with 0.11 volumes of cytoplasmic buffer [0.3 M Hepes (pH 7.9), 1.4 M KCl, 0.03 M MgCl₂] and RNase inhibitor (100 U/ml of extract; Promega). Extracts were kept at −80°C until further use. On the day of purification, 50 μl (for every RNA sample) or 200 μl (for preclearing of cytoplasmic extract) of streptavidin-Sepharose beads (GE Healthcare) were preblocked with 1 mg/ml BSA (Amresco), 0.2 mg/ml yeast tRNA (Sigma), and 0.2 mg/ml of glycogen (Ambion) in low salt wash buffer, [20 mM Hepes (pH 7.9), 100 mM KCl, 10 m MgCl₂, 0.01% NP40, 1 mM DTT] for 2.5 h on a turning wheel. After blocking, beads were washed 3 × 1 ml and resuspended in 250 μl of HS-WB-300 [20 mM Hepes (pH 7.9), 300 mM KCl, 10 mM MgCl₂, 0.01% NP40, 1 mM DTT] and incubated either with the biotinylated RNAs or cytoplasmic extract, on a turning wheel for 5 h at 4°C. After coupling of beads to the RNA, unbound RNA was washed away with 4 × 1 ml HS-WB-400 [20 mM Hepes (pH 7.9), 400 mM KCl, 10 mM MgCl₂, 0.01% NP40, 1 mM DTT]. The precleared extract was centrifuged and supernatant was collected and allocated appropriately to the RNA samples, where 5 mg of cytoplasmic extract was used per RNA sample. The RNA bound beads were incubated with the extract overnight at 4°C. The next day, extract was removed and beads were washed with 7 × 1 ml HS-WB-400. Beads were resuspended in 50 μl of 6 M Urea, 0.01% NP40, and 1 mM DTT, and placed in a shaker block at 900 rpm for 30 min at room temperature. Supernatant was then transferred to a new tube for protein precipitation with five volumes of prechilled acetone for 2 h at −20°C. Proteins were then pelleted at 14,000 rpm for 30 min, washed twice with 1 ml of 80% ethanol, dried, and resuspended in 20 μl warm protein sample buffer. Samples were then run on an SDS-PAGE gel. Protein bands were detected by Coomassie brilliant blue G-250 (Sigma-Aldrich). Specific bands were excised and their mass spectrometry analysis was performed by the Smoler Proteomics Center (Technion, Haifa, Israel).

The assay was performed as previously described (38). Briefly, total RKO cell lysates were prepared using ice-cold lysis buffer. The lysate was initially precleared with protein A/G PLUS-agarose beads (sc-2003; Santa Cruz Biotechnology). A polyclonal affinity purified rabbit Ab against vigilin (RN051PW; MBL) and a rabbit IgG control Ab (12-370; Millipore) were incubated with the precleared lysate, followed by addition of protein A/G PLUS-agarose beads. After several washing steps, the bound Abs and associated proteins and RNA were eluted, and the eluents were taken either to Western blot analysis or RNA enrichment analysis. For RNA analysis, RNA was isolated by TRI reagent (T9424; Sigma-Aldrich), and the presence of MICB and MICA transcripts was detected by quantitative RT-PCR (qRT-PCR). Enrichment of MICB/MICA transcripts was calculated relative to the levels of MICB/MICA transcript in the rabbit IgG control sample.

Cell lysates were prepared and run on an SDS-PAGE gel. Proteins were transferred to a nitrocellulose membrane by the tank blot procedure (Bio-Rad), and specific bands were detected using either anti-vigilin (RN051PW; MBL) or anti-GAPDH (SC-32233; Santa Cruz Biotechnology) Abs. Abs were diluted in 5% BSA in PBS. Chemiluminescence by Ab-linked HRP (Jackson ImmunoResearch) was detected using an EZ-ECL detection kit (Biological Industries).

The vigilin knockdown vectors (MISSION short hairpin [sh]RNA clones, SHCLNG-NM_005336) and scrambled vector were present in the pLKO.1-puro plasmids and were purchased from Sigma-Aldrich. As previously described (24), lentiviruses of the vectors were generated in 293T cells using a transient three-plasmid transfection protocol including a plasmid encoding the lentiviral Gag/Pol, and a plasmid encoding the VSV-G. Then 48 h after transfection, the supernatant with the viral particles were collected and used to infect cells. The 293T, BCBL1, HCT, and RKO cells were transduced with the lentiviruses, and selected using puromycin at a concentration of 5 μg/ml for 293 T cells, 2.5 μg/ml for BCBL1 and RKO cells, and 1 μg/ml for HCT cells. The presence of vigilin was detected by Western blot analysis.

For vigilin overexpression, an expression vector containing the vigilin cDNA was purchased from GE Healthcare Life Sciences (MGC human HDLBP sequence-verified cDNA, MHS6278). PCR was performed for gene amplification using the following primers (restriction sites underlined, Kozak sequence italicized): Forward, (NotI) 5′-CCCGCGGCCGCGCCGCCACCATGAGTTCCGTTGCAGTTTTGA-3′; reverse (XhoI) 5′-GGGCTCGAGTTATCGTTTGGGGCCC-3′. Vigilin was then cloned into a pHAGE-dsRED(−)-eGFP(+) lentiviral vector and viral particles were prepared as described above. Transfection efficiency was assessed by Western blot of vigilin.

Flow cytometry was performed with anti-MICA and anti-MICB Abs purchased from R&D Systems. Cells were incubated on ice for an hour with 0.2 μg of Ab per 100,000 cells. Detection was done with a secondary goat-anti-mouse Ab coupled to AlexaFluor 647 (Jackson ImmunoResearch) for 30 min on ice. Analysis was performed using the FACS-Calibur flow cytometer (BD Biosciences) and CellQuest software.

Analysis of CD107a on the surface on NK cells has been previously described (39). Primary activated bulk human NK cells were prepared as previously described (40). Briefly, PBMCs were collected from heparinized blood by centrifugation on Lymphoprep (StemCell Technologies). NK cells were isolated using the EasySep human NK cell enrichment kit (StemCell Technologies). Activated NK lines were generated by culturing the isolated NK cells with irradiated feeder cells (allogeneic PBMCs from two donors and 8866 cells) and 20 μg/ml PHA (Roche). The cultures were maintained in DMEM:F-12 Nutrient Mix (70:30), 10% human serum (Sigma), 2 mM glutamine, 1 mM sodium pyruvate, 1× nonessential amino acids, 100 U/ml penicillin, 0.1 mg/ml streptomycin (all from Biological Industries), and 500 U/ml rhIL-2 (PeproTech). For our assays, the primary activated bulk NK cells were incubated with various targets at a 1:1 E:T ratio . Anti-CD56 PE and anti-CD107a APC Abs (BioLegend) were included in the reaction mixtures. After 2 h at 37°C, the levels of CD107a were assessed by FACS. Blocking NKG2D function was performed by adding 5 μg/ml of anti-NKG2D (R&D Systems) to the NK cells for 1 h on ice, followed by the addition of the target cells and the abovementioned Abs.

The 5′UTR MICB, a shuffled sequence of the 5′UTR MICB (5′UTR ShuffMICB) as described above, and the 5′UTR of MICA were cloned upstream to a Firefly luciferase reporter gene in the pGL3 vector (Promega) as previously described (24). 5′UTR MICB forward (HindIII) 5′-CCCAAGCTTACTGGATAAGCGGTCGCTGAG-3′, reverse (NcoI) 5′-GGGCCATGGGGCCCCTACGTCGCCACCTT-3′; 5′UTR ShuffMICB forward (HindIII) 5′-CCCAAGCTTTCTCAAGACCGGGCGTAAGTG-3′, reverse (NcoI) 5′-GGGCCATGGGATCGAGGCGTTCCAACGCC-3′. Primers were ordered from Sigma-Aldrich and PCR was performed with the appropriate pBSII as a template. For 5′UTR of MICA the sequences were ordered, annealed and cloned into pGL3: forward (HindIII) 5′-AGCTTCACTGCTTGAGCCGCTGAGAGGGTGGCGACGTCGGGGCCC-3′, reverse (NcoI) 5′-CATGGGGCCCCGACGTCGCCACCCTCTCAGCGGCTCAAGCAGTGA-3′. RKO shScramble and RKO shVigilin 1 cells were plated 24 h prior to the assay, 30,000 cells per well in a 24-well plate. A quantity of 100 ng per well of the pGL3 vector, and 25 ng per well of a Renilla luciferase reporter vector (pRL-CMV), were transfected into the cells using the TransT-LT1 reagent (Mirus Bio). Then 48 h post transfection, the cells were harvested and luciferase activity was measured using Dual-Luciferase Reporter Assay System (Promega). The Firefly/Renilla activity ratio was normalized to that in the control cells, and the relative Firefly activity of the reporter was calculated.

RKO shScramble and RKO shVigilin 1 cells were plated the day before the treatment, at a confluency of ∼70%. The next day, the cells were treated with either 5 μg/ml of actinomycin D (ActD) (Sigma-Aldrich) or an equal volume of DMSO (Sigma-Aldrich) as a diluent control. The cells were treated for 8, 16, and 24 h. Cells were subsequently harvested and RNA was extracted, followed by cDNA preparation as detailed below. Using qRT-PCR, MICB abundance was calculated relative to GAPDH.

To detect mRNA levels, RNA was extracted from cell lysates using the QuickRNA Kit (Zymo Research). For the preparation of cDNA, RNA was anchored with Oligo dT primers (Thermo Scientific) followed by the addition of M-MLV reverse transcriptase (Invitrogen). RNA and cDNA preparation were performed according to the manufacturers’ protocols.

For the detection of specific mRNAs, freshly prepared cDNAs were used for SYBR Green-based detection in a QuantStudio 12 K Flex Real-Time PCR cycler (Life Technologies) with primers targeting: GAPDH, MICA, MICB, Renilla, and Firefly luciferase. GAPDH forward, 5′-GAGTCAACGGATTTGGTCGTGAPDH-3′, reverse, 5′-GATCTCGCTCCTGGAAGATG-3′; MICA forward 5′-ATCTTCCCTTTTGCACCTCC-3′, MICA reverse 5′-AACCCTGACTGCACAGATCC-3′; MICB forward, 5′-CTGCTGTTTCTGGCCGTC-3′, reverse, 5′-ACAGATCCATCCTGGGACAG-3′; Renilla, forward, 5′-TGAGGAGTTCGCTGCCTACC-3′, reverse, 5′-TGCGGACAATCTGGACGACG-3′; Firefly, forward, 5′-CATCTTCGACGCAGGTGTC-3′, reverse, 5′-GACTGGCGACGTAATCCAC-3′.

To identify RBPs that bind the 5′UTR of the MHC Class I–related chain family of stress induced ligands, we concentrated on MICB because its 5′UTR is significantly longer than that of MICA (117 bp versus 60 bp, respectively). The 5′UTR of MICA also includes few UTP nucleotides, and could therefore not precipitate properly with the UTP-biotinylation based assay described below. We performed an RNA-AP assay followed by mass spectrometry analysis (37). A schematic representation of the RNA-AP assay is shown in Fig. 1. We initially cloned the 5′UTR of MICB and control sequences into the pBSII plasmid, followed by in vitro transcription with biotinylated UTP (Fig. 1A). Control sequences included a non-biotinylated 5′UTR sequence, the inverse sequence of the 5′UTR, and a shuffled 5′UTR MICB sequence. The biotinylated transcripts were then incubated with streptavidin beads (Fig. 1B), and subsequently incubated with cytoplasmic extract (Fig. 1C). The samples were precipitated, and proteins that bound the transcripts were eluted and run on an SDS PAGE gel (Fig. 1D, 1E). Protein bands that were specific to the 5′UTR transcript were excised and sent for identification by mass spectrometry (Fig. 1E, blue box). From the mass spectrometry results, three known RBPs were chosen for further analysis: vigilin, nucleolin, and hnRNP U. Knockdown of nucleolin and hnRNP U by shRNA did not alter MICB surface expression (data not shown). Thus we concluded that although these proteins might bind the 5′UTR of MICB, they do not affect MICB expression. Therefore, we did not continue work on these proteins.

To determine whether vigilin directly binds MICB endogenously, we performed an RNA immunoprecipitation assay according to a protocol described by Peritz et al. (38) (described in detail in the “2Materials and Methods” section). Whole cell lysates were incubated with an Ab against vigilin or a control Ab. Following precipitation of the Abs, the coimmunoprecipitated RNA was extracted, and enrichment levels of MICB and MICA transcripts were assessed by qRT-PCR. We initially confirmed that we were able to successfully precipitate vigilin with the anti-vigilin Ab (Fig. 2A), followed by the qRT-PCR, which demonstrated an enrichment of MICB mRNA when compared with a control Ab (Fig. 2B). As a control, we also assessed the mRNA levels of MICA, which remained unchanged between the two treatments. Consequently, we confirmed that vigilin does not promiscuously bind mRNAs, but rather possesses binding specificity.

To assess the effect vigilin has on MICB expression we stably transfected RKO cells with a shRNA against vigilin (RKO shVigilin 1) and a scrambled shRNA (RKO shScramble) as a control. The transfectants were tested for vigilin expression by Western blot, and a significant downregulation of vigilin was observed in RKO shVigilin 1 cells as compared with RKO shScramble cells (Fig. 3A). FACS analyses of the RKO transfectants revealed that downregulation of vigilin does not have an effect on MICA, yet it leads to an increase in MICB surface expression (Fig. 3B and 3C, MICB levels quantified in Fig. 3D). Additional shRNAs against vigilin were transfected into RKO cells, and similar effects on MICA and MICB expression were observed (Supplemental Fig. 1A–C, MICB levels quantified in Supplemental Fig. 1D). To further corroborate our findings that vigilin may be a negative regulator of MICB, we transfected HCT116 cells with a different shRNA against vigilin (shVigilin 2) and a scrambled shRNA (shScramble). HCT116 cells transfected with shVigilin 2 displayed a significant downregulation of vigilin expression compared with cells transfected with shScramble (Fig. 3E). Similar to the results seen with RKO cells, vigilin knockdown did not affect MICA expression, but did lead to an increase in MICB surface expression (Fig. 3F, 3G, quantified in Fig. 3H). An additional two cell lines, 293T and BCBL1, were analyzed for MICA and MICB surface expression after vigilin knockdown with shVigilin 2. Knockdown of vigilin in 293T cells (Supplemental Fig. 2A), which only expresses MICA, did not affect MICA expression (Supplemental Fig. 2B). BCBL1, like HCT116, expresses both MICA and MICB. Similar to HCT116, knockdown of vigilin in BCBL1 (Supplemental Fig. 2C) increased MICB expression whereas MICA expression remained unchanged (Supplemental Fig. 2D). Taken together, these results strongly suggest that vigilin is a specific negative regulator of MICB.

In addition to downregulation of vigilin expression we attempted to overexpress vigilin in RKO, HCT, 293T, and BCBL1 cells (Supplemental Fig. 3). Irrespective of the level of the overexpression (ranging from 1.06 to 5-fold overexpression), neither MICB or MICA expression were affected. We inferred that because vigilin is already constitutively expressed in these cells, overexpression of this protein may not influence its activity as the system is already at 100% capacity.

To investigate whether vigilin regulation of MICB affects NK cell function, we performed CD107a degranulation assays. Primary activated bulk human NK cells were incubated with either RKO shScramble or RKO shVigilin 1 cells, at an E:T ratio of 1:1. The levels of CD107a, which are indicative of NK cell activation, were assessed by FACS. As can be seen, NK cells were significantly more activated in the presence of RKO shVigilin 1 cells. To confirm that the increase in NK cell activation was mediated by an increase in MICB expression, the NKG2D receptor on NK cells was blocked prior to the assay. Accordingly, NK cell activation was reduced to similar levels (the two bars to the right, Fig. 4). We were able to validate these results with an additional RKO shVigilin transfectant (Supplemental Fig. 4).

To evaluate whether vigilin specifically binds the 5′UTR MICB we performed a dual-luciferase reporter assay. We initially generated reporter plasmids consisting of the 5′UTR MICB, 5′UTR of MICA, or 5′UTR ShuffMICB (Fig. 5A, white box) fused upstream to the Firefly luciferase gene (light gray box, Fig. 5A). The reporter plasmid and an internal control Renilla luciferase plasmid were transfected into RKO shScramble and RKO shVigilin 1 cells, and the activity ratio of Firefly/Renilla luciferases was measured. As can be seen in Fig. 5B, when either the 5′UTR of MICA or 5′UTR ShuffMICB (white and gray bars, respectively) were upstream to the Firefly gene, the relative activity remained unchanged irrespective of vigilin’s presence. However, in the absence of vigilin, the relative Firefly activity was increased when the 5′UTR MICB was upstream to the Firefly gene. This indicated that vigilin binds to the 5′UTR MICB and hinders expression of the gene downstream.

RBPs exert their function by regulating different aspects of RNA biogenesis, including transcription, translation, and mRNA stability (35). We initially tested whether vigilin regulates the stability of MICB mRNA. We treated RKO shScramble and shVigilin cells with the transcription inhibitor ActD. By inhibiting transcription over time, we can compare the rate of MICB mRNA decay in cells with or without vigilin (RKO shScramble versus shVigilin, respectively). After every time period, each cell line was harvested and qRT-PCR was performed to assess the abundance levels of MICB mRNA. We calculated the MICB mRNA levels relative to GAPDH mRNA, which was used as our reference gene. RKO shScramble or RKO shVigilin 1 cells that were not treated with ActD were denoted as time 0 and set as 1. As can be seen in Fig. 6A, the abundance of MICB mRNA relative to GAPDH mRNA was the same in both RKO shScramble and shVigilin cells treated with ActD for increasing time periods. We therefore concluded that vigilin does not affect MICB mRNA stability; thus, it may regulate its translation. As such, we repeated the dual-luciferase assay as described above, and assessed both Firefly activity and Firefly mRNA expression. As can be seen in Fig. 6B, the absence of vigilin led to an increase in Firefly activity, however, the amount of Firefly mRNA remained unchanged. Consequently, we concluded that vigilin’s binding to the 5′UTR MICB does not affect mRNA expression, but rather regulates translation efficiency. To our knowledge, this is the first time MICB has been shown to be regulated at its 5′UTR; furthermore, by a distinct RBP, vigilin (Fig. 7, red circle).

The need for eight distinct NKG2D stress induced ligands and complete understanding of their regulation both remain conundrums. It has been suggested that different stress ligands are upregulated in response to various stresses, and that redundant stress ligands may have developed in response to different pathogens or tumors selectively downregulating several of the stress ligands (19). Nevertheless, the stress ligands must be tightly regulated under normal conditions to prevent inappropriate immune activation. Different mechanisms exist to control NKG2D ligand expression at the transcriptional and posttranscriptional levels, including transcription factors, miRNAs, and RBPs (19–32).

With regard to MICB it was demonstrated that this stress-induced ligand is subject to intense posttranscriptional regulation at its 3′UTR, which includes cellular and viral miRNAs, as well as RBPs (Fig. 7) (24, 25, 27, 28, 30, 31). There are 10 cellular miRNAs and three viral miRNAs known to downregulate MICB expression (24, 25, 27, 30). A total of six RBPs were found to influence MICB expression at its 3′UTR, five were negative regulators (CUGBP1, FUBP3, HuR, MATR3, and XRN2), and one a positive regulator (IGF2BP2) (31). Only four of the RBPs were documented to have consensus sequences that appeared in the 3′UTR of MICB (41–44). Of the four, only CUGBP1 had one site within MICB’s 3′UTR that was confirmed to be important for its function (31). The additional RBPs listed in Fig. 7 have several predicted binding sites throughout the sequence, where some sites were more essential for RBP function than others (31). However, it is important to note that the linear sequence motifs of RBPs may not be the only determinant of binding, because they are relatively short and occur frequently throughout the genome (45, 46). Accordingly, additional factors such as RNA structural motifs, secondary and tertiary structures, and interactions with other RBPs or miRNAs may heavily influence their activity (47–49).

Interestingly, only one of the RBPs, CUGBP1, has a site that overlaps with one cellular miRNA, miR-10b. The fact that most MICB regulators do not have overlapping binding sites suggests they may cooperate to exert their function, or that they are differentially expressed under diverse circumstances or conditions. Whether they directly interact is yet to be determined. In addition, our group has recently discovered that the RBP IMP3 indirectly downregulates MICB expression (32).

Whether MICB is also regulated at its 5′UTR was previously unknown. Consequently, we set out to discover a new mode of regulation of MICB by searching for unique RBPs that bind to its 5′UTR. To identify unique RBPs that bind to the 5′UTR of MICB we began with an RNA affinity purification assay, which led us to discover vigilin. Vigilin, also known as high density lipoprotein binding protein, is a large protein containing 15 hnRNP K (KH) domains, which interacts with and binds RNA (50, 51). Vigilin is a ubiquitous multifunctional protein, located both in the cytoplasm and nucleus, and is highly conserved from yeast to human (52–57). Its exact biological functions remain to be fully elucidated. Vigilin can act as an HDL receptor, and has been reported to transport cholesterol (58). Alternatively, it has been found to interact with translation machinery, including tRNAs, ribosomes, and elongation factors (59–61). It has been shown to bind A-to-I edited mRNAs, and also to be involved in heterochromatin formation (62, 63). Vigilin has also been suggested as a tumor suppressor in various tumors (64, 65). Interestingly, one of these studies is the only study to present vigilin’s role as an RBP in human cells, where it binds to and represses the translation of the c-fms proto-oncogene mRNA in breast cancer cells (64). Vigilin does not have a consensus binding sequence, but rather a suggested motif of (A)_nCU and UC(A)_n (66). We mutated this site in the 5′UTR MICB, and observed no change in vigilin suppression (data not shown). Thus, in MICB vigilin binds other sites. It was also shown that vigilin requires at least 69 bp for binding (64). Further research regarding its role as an RBP has not been documented until now.

We discovered, using four cell lines, that vigilin negatively regulates the expression of MICB, and that in its absence there is a significant increase in MICB expression, whereas MICA expression remained unchanged. The differences in the extent of MICB upregulation upon vigilin knockdown may be attributed to the differential expression of the various MICB regulators (Fig. 7). It will be interesting to further investigate the interplay between the different MICB regulators and their influence on MICB expression.

When we attempted to overexpress vigilin in the abovementioned four cell lines, we did not observe any effect on either MICA or MICB surface expression. We therefore conjectured that either the overexpression levels were not enough to induce any effect on said proteins, or because vigilin is constitutively expressed in these cells, the protein is already working at 100% capacity. As such, our data would suggest that overexpression of vigilin would not be employed by tumors to evade NK cell recognition. Conversely, it has been reported that vigilin is highly mutated in certain tumors and may act as a tumor suppressor (64, 65). However, we suggest that its absence in these tumors may be a cellular trigger to activate the immune system against the transformed cells because we observed that vigilin downregulation induced MICB upregulation, which resulted in increased NK cell activation. It will be important to further elucidate vigilin’s role in tumor transformation and subsequent immune response.

We verified that vigilin binds MICB endogenously, and that the binding was mediated by MICB’s 5′UTR. This binding did not affect mRNA stability, but did decrease its translation efficiency, irrespective of the gene downstream to MICB’s 5′UTR. Taken together, we have successfully demonstrated a novel mode of MICB regulation, based on a unique RBP binding to its 5′UTR and hindering its expression.

The authors have no financial conflicts of interest.