Efficient Internalization of MHC I Requires Lysine-11 and Lysine-63 Mixed Linkage Polyubiquitin Chains

Increased endocytosis of MHC I following polyubiquitination of the membrane proximal cytosolic tail lysine by K5

The downregulation of MHC I in HeLa cells is more marked in the presence of K3 (HeLa-K3) than K5 (HeLa-K5) (Figure 1A). Comparison of MHC I immunoprecipitations (IPs) from radiolabelled HeLa, HeLa-K3 and HeLa-K5 cells revealed higher molecular weight bands above the 44 kDa class I heavy chain, which were identified as ubiquitinated MHC I heavy chain species by reprecipitation with both MHC I and ubiquitin-specific antibodies (Figures 1B and S1A). MK3 acts on MHC I in the endoplasmic reticulum (ER), while K3 and K5 affect cell surface MHC I (1). We wanted to ascertain whether any of the ubiquitinated class I species in HeLa-K5 cells were endoglycosidase-H (EndoH) sensitive, suggesting ubiquitination could occur in the ER or in a premedial Golgi compartment. Pulse-chase analysis on radiolabelled HeLa-K5 cells shows that all the ubiquitinated class I is EndoH resistant, suggesting that K5 only affects MHC I in a post-ER compartment (Figure S1B). While the monoubiquitinated MHC I species appears predominant in HeLa-K5 cells, this is not seen with HeLa-K3 cells. This dominance of the monoubiquitinated class I species in HeLa-K5 cells was further exaggerated following IP of cell surface compared with total MHC I (Figure 1B). In HeLa-K3 cells, little ubiquitinated MHC I is detected at the cell surface, reflecting the more rapid endocytosis of MHC I in these cells compared with HeLa-K5 (Figure 1C). Together, these results show that in HeLa-K5 cells, MHC I is readily monoubiquitinated, but this provides a poor internalization signal, suggesting the ubiquitin chain extension in HeLa-K5 cells is relatively inefficient, leaving many monoubiquitinated MHC I molecules stuck at the cell surface.

MHC I heavy chains contain two conserved cytosolic tail lysine residues. We made mutations in these residues in the HLA-A*0201 (HLA-A2) MHC I molecule, which has the two conserved lysines at positions 335 and 340 and an additional lysine at position 364. The conserved (K₃₄₀) lysine residue in the cytoplasmic tail of HLA-A2 is the preferred acceptor site for K3-mediated polyubiquitination (6,18,23). In HeLa-K5 cells, downregulation of the HLA-A2 (A2_KRA) mutant, expressing only the single membrane proximal lysine (K₃₃₅), was as efficient as downregulation of wild-type HLA-A2 (A2_KKK) (Figure 1D,E). In contrast, the lysine at position 364 (A2_RAK) afforded no MHC I downregulation, whereas the conserved lysine at position 340 (A2_RKA) afforded minimal MHC I downregulation –only twofold higher than a mutant HLA-A2 lacking all cytosolic lysine residues (A2_RRA) (Figure 1E). The membrane proximal lysine (K₃₃₅) is therefore the dominant ubiquitin acceptor and this was confirmed biochemically, as the single lysine HLA-A2_KRA showed an identical polyubiquitin pattern compared with that seen with the wild-type HLA-A2 (Figure 1F). As removal of all cytosolic lysine residues prevents the K5-mediated downregulation of MHC I (Figure 1E), we can exclude contributions from thio- or oxyester ubiquitin chain conjugations via cysteine, serine or threonine residues (24). Thus, K5 differs from K3 in targeting the conserved membrane proximal lysine residue (K₃₃₅) in the cytosolic tail of MHC I for polyubiquitination.

K5-mediated polyubiquitination of MHC I requires ubcH5 and ubc13

K3-induced polyubiquitination of MHC I requires ubcH5 and ubc13 (6). We therefore tested whether K5 also uses these E2 enzymes. MHC I was rescued back to the cell surface in HeLa-K5 cells depleted of ubcH5c and ubc13 by short interfering RNA (siRNA) (Figure 2A). Western blot analysis confirmed effective depletions (Figure 2B). The different effects of the E2 enzymes' depletion were readily distinguishable on biochemical analysis of ubiquitinated MHC I. UbcH5c depletion abrogated K5-mediated MHC I ubiquitination (Figure 2C). In contrast, depletion of ubc13 showed a collapse of the MHC I polyubiquitin ladder to the predominantly monoubiquitinated form.

Quantitative mass spectrometry identifies Lys11 and Lys63 ubiquitin chain linkages on MHC I from HeLa-K5 cells

To examine the ubiquitin chain linkage required for K5-mediated downregulation of MHC I, we used the ubiquitin-AQUA (absolute quantification of ubiquitin) quantitative MS-based approach (25). The use of eight isotope-labelled internal standards to quantify tryptic peptides, derived from the digestion of monoubiquitin and polyubiquitin chains bound to MHC I, allows a direct determination of the levels of ubiquitinated species. Denatured MHC I was immunoprecipitated from membrane fractions of HeLa and HeLa-K5 cells, with the class I heavy chain-specific (HC10) monoclonal antibody (mAb). Radiolabelled samples were run on the same gel to directly compare Coomasie stains with radioimmune precipitations (Figure 3A), and allow alignment of MHC I heavy chain and the ubiquitinated MHC I species.

Following trypsinization and reverse-phase liquid chromatography (LC) of peptides extracted from the indicated gel slices (Figure 3A), the resulting peptides were analysed by MS/MS. Good coverage of MHC I-derived peptides was achieved. The sequence of HLA-A28, the HLA-A allele in HeLa cells is shown (Figure 3B) and gave 49% coverage (44 peptides) of the HLA-A allele and 46% coverage (24 peptides) of the HLA-B allele. From HeLa-K5 cells, a 2.6-fold increase in total ubiquitin was detected in the region above monoubiquitinated class I, as compared with the corresponding region of HeLa cells (Figure 3A, gel slices #3–6, Figure 3C,D). Polyubiquitin chain linkage analysis (Figure 3C,D) identified an enrichment of Lys11 as well as Lys63 linkages in samples prepared from HeLa-K5 cells with a 13-fold increase in Lys11-linked polyubiquitin chains and a 2.2-fold increase in Lys63-linked polyubiquitin chains compared with HeLa cells. Other ubiquitin chain linkages were either detected at very low levels or not at all (Figure 3C,D).

Lys63 and Lys11 of ubiquitin are necessary for K5-mediated MHC I endocytosis

As the LC-MS/MS identified an enrichment of Lys11- as well as Lys63-linked ubiquitin chains in the presence of K5, we wanted to identify whether these polyubiquitin chains also played a functional role in K5-mediated MHC I downregulation. To facilitate this, epitope-tagged Lys-to-Arg ubiquitin-green fluorescent protein (GFP) mutants were overexpressed. These mutants are particularly informative as cotranslational cleavage of the C-terminal GFP frees the terminal glycine of ubiquitin for subsequent conjugation (26). The cleaved GFP provides both a quantitative surrogate marker of mutant ubiquitin introduced into the cells, and also allows gating of the ‘GFP high’ population, so that only those cells in which mutant ubiquitin outcompetes endogenous ubiquitin are analysed.

When a ubiquitin lysine residue critical for K5 function is mutated, increasing levels of GFP expression on the x-axis of the dot plots result in increasing MHC I expression on the y-axis, i.e. rescue of MHC I to the cell surface. The effect of ubiquitin mutants was determined on the MHC I molecule HLA-A2_KRA (A2_KRA), which only contains the single K₃₃₅ cytosolic tail lysine residue. We can therefore exclude a role for different ubiquitin chains on different ubiquitin acceptor lysine residues. Overexpression of the ubiquitin mutant R63 Ub-GFP (Lys63 mutated to Arg63) rescues A2_KRA surface expression, confirming the requirement for Lys63-linked ubiquitin chains on the single membrane proximal lysine during MHC I downregulation (Figure 4A). Lys63-linked polyubiquitin chains are also required for efficient endocytosis as MHC I internalization was slower in HeLa-K5 cells expressing high levels of R63 Ub-GFP (Figure 4B). Overexpression of the R6 and the R11 Ub-GFP mutants also rescued A2_KRA surface expression (Figure 4A). An identical rescue was also seen with R6, R11 and R63 Ub-GFP mutants on total MHC I (Figure S2). Like the R63 mutant, R11 interfered with MHC I endocytosis (Figure 4B), confirming the requirement for Lys11 in K5-mediated MHC I endocytosis, while MHC I internalization was not affected by the R6 mutant.

Lys6, Lys11 and Lys63 of ubiquitin are necessary and sufficient for K5-mediated MHC I endocytosis

Having established that Lys11 and Lys63 were necessary for K5-mediated downregulation of MHC I, we asked whether they were also sufficient. On the backbone of a lysineless ubiquitin mutant (KoUb-GFP), we replaced the indicated arginine residues back to lysines (K63) to determine which residues restored K5 function and prevented cell surface HLA-A2_KRA rescue (Figure 5A). As predicted, A2_KRA surface expression was rescued following Ko ubiquitin expression because polyubiquitin chains cannot be generated. Neither the single (Ko, K6, K11, K63, K48) mutants nor the double mutants (K11/63, K6/11, K6/63, K48/63) were sufficient to prevent a cell surface rescue of A2_KRA expression (Figure 5A). Only the triple mutant (K6/11/63) restored K5 function for both A2_KRA(Figure 5A) and total MHC I (Figure S3), an effect confirmed with internalization assays (Figure 5B). Thus, Lys6, Lys11 and Lys63 of ubiquitin are necessary and sufficient for K5-mediated endocytosis of MHC I.

Cell lines

All cell lines were routinely grown in RPMI-1640 (PAA), 10% foetal bovine serum (PAA) and penicillin/streptomycin (SIGMA).

Antibodies

The following antibodies were used: mAb w6/32, mAb HC10, mAb P4D1 anti-ubiquitin (Santa Cruz), mAb BB7.2 recognizes conformational HLA-A2 , MR24 (a kind gift of E. Wiertz) recognizes the ER luminal epitope of HLA, anti-ubc13 (Zymed), anti-ubcH5 (Boston Biochem), anti-calnexin (AF8 mAb) (a kind gift of M. Brenner), Cy5 and horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson).

Lentivirus propagation, transduction and mutagenesis

The lentivirus expression system used was as described (22). The lentivirus expression plasmid vectors used were kind gifts of Y. Ikeda. Virus preparation and cell transduction have been described elsewhere (22). Site-directed mutagenesis (QuickChange®, Stratagene) was sequence verified.

Flow cytometry

Samples were processed on a FACSCalibur as described (18) and analysed on flowjo software. The FL2 channel was used to analyse GFP expression in cells transduced with ubiquitin mutants.

Internalization assays

Sixty to 72 h post-transduction with lentiviruses, cells were harvested by trypsinization and controls were set aside for Cy5 staining. The remaining cells were stained with w6/32 at 4°C and washed extensively with cold PBS. Samples were warmed to 37°C and at each time-point aliquots were harvested into cold PBS to stop endocytosis. Following the last time-point, all samples were stained with secondary antibody, and processed for flow cytometry. The geomean of w6/32 was determined and the data were analysed in Microsoft Excel. Starting levels of MHC I expression were normalized to 100% and the percent change in geomean (±SEM) expressed as a function of time.

Radiolabelling, immunoprecipitation and polyacrylamide gel electrophoresis

Cells were pulse labelled for 10 min and chased with non-radioactive complete medium for 45 min. Cells were lysed and precleared of non-specific reactivity on protein A-agarose (Sigma) and/or immunoglobulin G (IgG) Sepharose (GE Healthcare) and then w6/32 or BB7.2 plus protein A-agarose was used for a primary IP of conformational MHC I. Following extensive washing, proteins were eluted from the protein A-agarose with 1% SDS in tris-buffered saline (TBS) pH 7.4 at 70°C for 10 min. The SDS was then sequestered with 10 v of 1% non-ionic detergent, the beads were spun out, and the supernatant was transferred to fresh tubes for the reIP with either HC10 (for the w6/32 primary IP) or MR24 (for the BB7.2 primary IP). To detect cell surface MHC I (Figure 1B), radiolabelled cells were washed and stained with w6/32 prior to detergent lysis. Postlysis, samples were split and either precipitated directly with protein A followed by HC10 reIP (S for surface MHC I), or additional w6/32 was added to precipitate the total w6/32 reactive MHC I pool (T for total MHC I). Samples were then subjected to SDS–PAGE and autoradiography. Cells used for MS analysis were harvested by scraping and lysed in a hypotonic buffer/freeze thaw procedure (36) in the presence of protease inhibitors. The membrane fraction was pelleted at 100 000 ×g for 1 h and washed twice with 1 m sodium carbonate prior to solubilization with 1% SDS in TBS at 70°C for 15 min. The SDS was then sequestered with 10 v of 2% C12E9 (Calbiochem) in TBS pH8 and the solubilized membrane proteins were precleared of non-specific binding proteins with IgG Sepharose. MHC I was immunoprecipitated with HC10 (purified in house) covalently conjugated to protein A beads (Pierce Protein A IgG Plus Orientation Kit, Thermo Scientific), extensively washed in 0.1% C12E9 in TBS and eluted at 25°C for 5 min in the elution buffer supplied with the Pierce kit. Beads were excluded by centrifugation through Mobicol spin columns (MoBiTec) and the eluate was neutralized before loading buffer with DTT was added, samples were heated to 70°C and the products were separated on 9% polyacrylamide gels. The gels were fixed then stained with GelCode Blue Coomasie blue stain as per the manufacturer's instructions (Thermo Scientific). Gels were scanned and either dried (radioactive portion) and subjected to phosphorimager analysis, or bands were excised and stored at −80°C prior to MS.

siRNA depletion

Ubc13 (D-003920-01, Dharmacon) and ubcH5c (6) siRNAs were reverse transfected into HeLa and HeLa-K5 using Dharmafect 1 (Dharmacon). Cells were incubated at 37°C/5% CO₂ for 60 h prior to assay.

Western blot analysis

Cells were lysed in 1% TX-100 in TBS plus protease inhibitors (Roche), and diluted into 1 × Laemli buffer prior to SDS–PAGE. Samples were transferred to polyvinylidene difluoride membranes (PVDF, Millipore), the membranes were blocked overnight at 4°C, then incubated with primary and secondary antibodies as noted. Membranes were developed in West Dura Extended Chemiluminescent substrate.

Mass spectrometry

The proteins in the gel slices were digested by trypsin with the addition of eight stable isotope-labelled peptides as internal standards, including all seven −GG peptides and one ubiquitin peptide (Thr55-Lys63). The resulting peptide mixtures were analysed by reverse-phase LC-MS/MS on an LTQ-Orbitrap hybrid mass spectrometer. The instrument was operated to monitor the eight peptides and their related native counterparts by selective reaction monitoring (SRM). The quantification analysis was carried out using xcalibur software

We also applied protein identification with the same samples on the same LC-MS/MS platform. Peptide samples were eluted in a 60-min gradient with buffer B from 4 to 30% (buffer A, 0.4% acetic acid, 0.005% heptafluorobutyric acid and 5% acetonitrile; buffer B, 0.4% acetic acid, 0.005% heptafluorobutyric acid and 95% acetonitrile). The eluted peptides were detected in a precursor ion MS scan by Orbitrap after accumulation to a target value of 1 × 10⁶ in the linear ion trap (400–1600 m/z, 60 000 resolution at m/z 400, 1 µscan), followed by data-dependent MS/MS scans for the eight most abundant ions from each survey scan with an ion intensity above 500 counts and all decharged ions (isolation width of 2 m/z, 35% normalized collision energy, 1 µscan, target value of 5000, 60 seconds dynamic exclusion and preview mode applied). The acquired MS/MS spectra were searched with the Sequest-Sorcerer algorithm on a Sorcerer 2 IDA (Sage-N-Research) (37) against a composite target/decoy database to estimate the false discovery rate (38,39). The target proteins included human proteins supplemented with human ubiquitin sequence without methionine and common contaminant proteins, such as porcine trypsin and human keratins. The decoy proteins were generated from pseudo-reversed sequences of all target proteins. Searching parameters consisted of semitryptic restriction, and dynamic modification of oxidized or acetylated Met (+15.9949 and 42.011 Da, respectively). The initial mass search window was set to 20 ppm After that, the peptide matches were further filtered by dynamically increasing XCorr and ΔCn cutoffs until the global protein false discovery rate of 1% was reached. If the identified proteins only contained a shared peptide, they were merged into the same protein group.