Using intein catalysis to probe the origin of major histocompatibility complex class I-presented peptides

Abstract

All vertebrate nucleated cells generate peptides from their expressed gene products and then display them at the cell surface bound to MHC class I molecules. This allows CD8⁺ T cells to detect and eliminate abnormal cells that are synthesizing foreign proteins, e.g., from viruses or mutations. To permit the immune system to more uniformly monitor a cell's proteins, regardless of their half-life or location, it has been thought that the products of rapid degradation of the mistakes of protein synthesis (defective ribosomal products, DRiPs) preferentially contribute to the class I-presented peptides. However, using intein catalysis to generate peptide sequences exclusively by posttranslational splicing of mature proteins, we show here that presented peptides can be generated from fully folded and functional proteins. Remarkably, the presentation of peptides from two model mature proteins is just as efficient as from newly synthesized proteins subject to errors in translation or folding. These results indicate that for the constructs we have analyzed, DRiPs are not a more efficient source of class I peptides for antigen presentation than the turnover of mature functional proteins. Accordingly, our data suggest that one of the major ways the immune system evaluates the health of cells is by monitoring the breakdown products of the proteome.

All proteins in cells are turned over, albeit at markedly different rates. Truncated, misfolded, or damaged polypeptides are degraded very rapidly, whereas normal proteins have half-lives that can vary from minutes to days (1). The majority of these normal and abnormal cellular proteins are catabolized by the ubiquitin–proteasome pathway (2). It is generally accepted that the vast majority of presented peptides are generated by proteasomes (3). It was originally assumed that these presented peptides derived from the normal turnover of the proteome, although it was never formally proven. However, it was subsequently suggested that protein turnover was unlikely to be the major source of peptides on the theoretical grounds that this would impair the ability of the immune system to rapidly detect viral infections (4). This is because many viral proteins are quite stable and it was reasoned that the generation of presented peptides from these antigens would be very slow. It was therefore suggested that there must be alternative sources of presented peptides (4, 5).

The fact that peptides from stable proteins were rapidly presented after viral infection suggested that another mechanism was operative, wherein the kinetics of the generation of peptides was independent of the half-life of the source protein. Accordingly, it was hypothesized that defective ribosomal products (DRiPs) might be a major source of presented peptides. The idea was that there was a constant rate of errors in the synthesis of all proteins (because of truncations, misfolding, etc.) and that polypeptides containing mistakes were rapidly degraded. By focusing on these errors in protein synthesis, the antigen presentation pathway would sample all newly synthesized proteins more uniformly, because the defective species of any protein would be degraded rapidly regardless of the stability of the corresponding mature forms (4).

In support of the DRiPs hypothesis, it was found that cells do in fact make mistakes in protein synthesis, although the exact frequency of these errors is controversial (6, 7). In addition, in some cases it was found that the generation of peptide–MHC class I complexes from an antigen rapidly ceased after protein synthesis was terminated (8–10). These results were interpreted to indicate not only that the degradation of newly synthesized, rapidly degraded proteins does contribute to class I presentation, but more importantly that DRiPs were a preferential source of peptides for antigen presentation, because despite encompassing a minor fraction, <1–30% (6, 7) of all synthesized proteins, the abrogation of new protein synthesis had a major impact on class I presentation (7). It was later reported that the generation of antigenic peptides from DRiPs was from 25-fold to 65-fold more efficient than from “retirees” (11). These observations led to the speculation that there was a specialized biosynthetic pathway and/or compartmentalization that might supply DRiPs directly to the antigen presentation pathway (12, 13).

This DRiPs model has been accepted by many different research groups in the field and numerous papers and reviews have described DRiPs as a major source of presented peptides, whereas in contrast, the turnover of mature proteins via the ubiquitin–proteasome pathway has been suggested to play a minor role in generating antigenic peptides (14–20).

More recently a mechanism has been described whereby small amounts of newly synthesized protein are degraded immediately after a “pioneer” round of nuclear translation before mRNA export to the cytosol for bulk translation (18). Although no mention was made to the folding status of these pioneer translation products, they are believed to be very rapidly degraded and to constitute a major source of antigenic peptides for class I presentation; accordingly, they share the hallmarks of DRiPs (20).

The DRiPs model predicts that presented peptides should be generated much more efficiently from defective and rapidly degraded polypeptides immediately after synthesis (4, 12). Because the question of how the immune systems monitors cellular proteins is fundamental for understanding immune surveillance, and also because it has practical implications for how antigens might be optimally designed or delivered in vaccines, we have attempted to critically test this model. To do this we compared the efficiency of antigen presentation between a construct in which the presented peptide could come from DRiPs, to a second construct in which the epitope would only form in a folded functional protein and therefore could not be generated from DRiPs.

Results

Construct Description and Splicing Kinetics.

Inteins are proteins that perform a self-splicing reaction wherein they make two internal cleavages and then precisely ligate these cleavage products (21). The residues that form the intein's catalytic and splicing sites are far apart in the primary sequence but are brought together to generate an active site when the protein is fully folded, as shown in Fig. 1 A and B for the RecA intein from Mycobacterium tuberculosis. This same mechanism is used by other inteins such as the one in PRP8 from Penicillium chrysogenum, with the active site residues highly conserved (Fig. 1A).

Fig. 1.

Intein structure–function and construct description. (A) Clustal W alignment of the amino acid sequences of the M. tuberculosis and P. chrysogenum mini-inteins used. Regions involved in the splicing catalysis are highlighted in black. (B) Diagram of the crystal structure of the M. tuberculosis full-length RecA intein (Protein Data Bank ID code 2IMZ). Highlighted in black are the terminal residues and the histidine side chain in the N3 motif, all of which are involved in catalysis (27). (C–F) Schematic of lentiviral construct inserts (note that the product of the prespliced construct is the same as generated by intein splicing). Positions of the class I epitopes SIINFEKL (S8L), KCSRNRQYL (K9L), and ASNENMETM (A9M) are shown relative to GFP, HA tag, and the two inteins.

We reasoned that if an antigenic peptide sequence (epitope) were split between the two splice sites of the intein (Fig. 1 C and E), then the intact epitope could not exist in a DRiP, but would form only after the protein fully folded and became catalytically active; we will refer to such epitopes as “splice dependent.” Therefore, the generation of the epitope sequence is posttranslational and requires the protein to be in its native state. Because DRiPs by definition (4) are derived from nonnative species, epitopes produced by intein splicing should not be considered DRiPs. Moreover, if a construct containing a splice-dependent epitope also had elsewhere in its sequence a second intact epitope, whose formation did not require splicing, then one could compare the efficiency of presentation of this splice-independent epitope, which could be made from DRiPs, to that of the splice-dependent one that could not be made from DRiPs (Fig. 1 C and E). Finally, by removing the intein sequence from the cDNA, we prepared another construct directly encoding the final spliced intein product (which we will refer to as “prespliced”); in this situation both epitopes would be synthesized intact and could come from DRiPs (Fig. 1 D and F).

We identified epitopes from the influenza nucleoprotein (NP, A9M epitope) and murine Jarid1d protein [minor histocompatibility antigen on the Y chromosome (HY), K9L epitope] that could be split between the splice sites of the PRP8 and RecA inteins, respectively, while retaining the key residues needed for splicing and ligation. Constructs were made with the split epitopes and a second intact epitope from ovalbumin (S8L), and also green fluorescent protein (GFP) and hemagglutinin (HA) tags to follow their expression (Fig. 2). The constructs were placed under the control of a tetracycline-on promoter for inducible expression. We found that the expression of these constructs was proportional to the amount of doxycycline added over a wide dynamic range (Figs. 3A and 4A), allowing us to generate dose–response curves in subsequent experiments.

Fig. 2.

Intein splicing kinetics. (A) Autoradiogram after SDS/PAGE and anti-HA immunoprecipitation of lysates from GFP-RecA-HY lentivirus-infected E36 cells pretreated with epoxomicin (5 μM, 2 h) subjected to [³⁵S]-Met/Cys pulse label followed by a “cold” chase. Unspliced intein is indicated with a black arrowhead, and the spliced version with an open arrowhead. (B) Same as A except using lysates from GFP-PRP-NP lentivirus-infected E36 cells. (C and D) Arrows as in A, densitometry quantification of the bands on the autoradiograms above. See also Figs. S1, S3, and S4.

Fig. 3.

Presentation at steady state from A9M constructs. E36 cells transduced with lentiviral doxycyline (Dox)-inducible GFP-NP (prespliced), GFP-PRP-NP (intein), or N3 (splicing defective). GFP-PRP-NP mutant constructs were induced with the indicated amounts of Dox and the following measurements were performed in A–D: (A) GFP mean fluorescence intensity (MFI) after 24 h. (B) Cell surface presentation of D^b-A9M complexes measured by coculture in the presence of brefeldin A (to stop further transport of MHC class I complexes) of 12.64-CD8αβ-Luc T-cell hybrids (35) with E36 cells induced for 24 h. (C) D^b-A9M complexes assayed as in B but with 6 h GFP-PRP-NP E36 Dox induction in the presence or absence of epoxomicin (5 μM). (D) Cell surface presentation of K^b-S8L complexes at 24 h postinduction was measured with the 25-D1.16 antibody (39). Data from A–C were replotted as follows: (E) Presentation of K^b-S8L epitope relative to GFP MFI. (F) Presentation of D^b-A9M epitope relative to GFP MFI. (G) Linear regression of presentation of D^b-A9M versus K^b-S8L epitope for the various constructs.

Fig. 4.

Presentation at steady-state from K9L constructs. Similar to Fig. 3, except the E36 cells were transduced with constructs carrying GFP-K9L (prespliced), GFP-RecA-K9L (intein), or N3 (splicing defective) mutant GFP-RecA-HY and the D^b-K9L epitope was measured instead of D^b-A9M using HY TCR-transgenic T cells: (A) GFP mean fluorescence intensity (MFI) after 24 h. (B) Cell surface presentation of D^b-K9L complexes quantified by intracellular cytokine staining assays for TNFα production by coculture in the presence of brefeldin A (to stop further transport of MHC class I complexes) of HY TCR-transgenic T cells with E36 cells induced for 24 h. (C) D^b-K9L complexes assayed as in B but with 4 h GFP-RecA-HY E36 Dox induction in the presence or absence of epoxomicin (5 μM). (D) Cell surface presentation of K^b-S8L complexes at 24 h postinduction, quantified by assaying the production of luciferase from RF33-Luc cells (35) cocultured with the E36 cells for 16 h in the presence of brefeldin A (1 μg/mL). Data from A–C were replotted as follows: (E) Presentation of K^b-S8L epitope relative to GFP MFI. (F) Presentation of D^b-K9L epitope relative to GFP MFI. (G) Linear regression of presentation of D^b-K9L versus K^b-S8L epitope.

Inteins are only found in primitive, mostly single cell organisms and it was unknown whether the PRP8 and RecA inteins would actually splice in mammalian cells. Therefore, we expressed these constructs in an antigen-presenting cell (22) (E36, Materials and Methods), and using pulse-chase experiments, we evaluated whether protein splicing occurred. As shown in Fig. 2 A and C, the GFP-RecA-HY construct splices with a half-time of 52 min. The GFP-PRP8-NP construct also splices and does so more rapidly, with a half-time of 8 min (Fig. 2 B and D). Splicing does not occur in either intein (Fig. 1) if one of their key N3 motif (Fig. S1) active site residues is mutated (His-73 to Leu) (21), indicating that catalysis is required for this reaction (Fig. S1). We confirmed that the splicing reaction generated the expected product for the RecA construct by analyzing the splice product(s) by mass spectrometry (Fig. S2) and antigen presentation assays for the spliced epitope (below). The overall half-life of the spliced GFP-RecA-HY product was 12 h, indicating that the spliced products are relatively stable proteins.

Intein Splicing Can Generate MHC Class I Epitopes.

We next induced the expression of the GFP-PRP8-NP construct for 24 h and analyzed the presentation of the splice-dependent A9M epitope on D^b MHC class I molecules (Fig. 3B). Presentation was blocked by the proteasome inhibitor epoxomicin (Fig. 3C) and therefore the epitope was being generated through the expected antigen presenting pathway. This presentation requires splicing because the catalytically inactive construct stimulates the A9M-specific T cell very poorly if at all (Fig. 3G), although this latter construct does efficiently generate S8L-MHC class I complexes, which do not require splicing (Fig. 3G). Therefore, these data indicate that an antigenic epitope can be presented from a mature and functional protein.

Our system allows us to compare the efficiency of presentation of the same epitope when it is derived exclusively from mature protein (splice dependent) or when it is derived from either mature or newly synthesized protein (prespliced). Remarkably, the presentation of the A9M epitope from the split intein construct is very similar to that from the prespliced construct at all concentrations of doxycycline tested (Fig. 3F). When the presentation of the A9M versus S8L epitope is plotted, the slopes of the lines are essentially the same (Fig. 3G). These results indicate that A9M is presented as efficiently from mature protein species as it is from newly synthesized protein species.

Because the PRP8 intein construct begins to splice rapidly, it was formally possible that the domain containing the epitope was still abnormal (a DRiP) but not yet degraded when splicing occurs. If this were to occur, the splice product might still be a DRiP and rapidly degraded. This seemed unlikely because at early time points when proteasomes were inhibited, we did not see increases in the spliced products nor evidence of polyubiquitinated forms as would have been expected if there were DRiPs (Fig. 2 A and B). Nevertheless, this possibility was further addressed using the RecA intein construct because its splicing rate is much slower (t_1/2 > 50 min, Fig. 2C) and DRiPs would be degraded quickly (because the t_1/2s of DRiPs are a few minutes, they would be gone before splicing occurred) (6). Consistent with the results above, the split K9L epitope is presented from the RecA intein (Fig. 4 B and F) and this requires splicing because the catalytically inactive construct is not presented (Fig. 4G). The presentation of the K9L epitope is proteasome dependent (Fig. 4C). Again the presentation of the HY epitope was similar from the splice-dependent intein and prespliced constructs when normalized to GFP expression (Fig. 4F), as were the slopes of the K9L versus S8L lines (Fig. 4G). Therefore, even after a delay during which all DRiPs would have been degraded, the mature epitope was as efficiently presented when epitope formation required substrate protein maturation as when it did not. Moreover, these results extend our findings to two different inteins and two different antigens/epitopes.

Analysis of Presentation at Early Time Points.

DRiPs might be an important source of presented peptides at early times after the start of antigen synthesis, before cells accumulate substantial levels of mature protein. To evaluate the possibility of preferential utilization of DRiP-derived epitopes during the approach to steady state, we compared the presentation of the splice-dependent and splice-independent epitopes at various times after induction. For both constructs the presentation of the splice-independent SL8 epitope was greater than the splice-dependent peptide at very early time points, with the difference being greater for RecA-K9L (Fig. 5). This difference could be due to the intrinsic delay in splice kinetics (i.e., at early time points there is less spliced intein product compared with the unspliced version; Fig. S3B) and/or to a contribution from DRiPs. To distinguish between these possibilities we modeled the expected differences in presentation in silico (Materials and Methods). The model takes into account synthesis, splicing, processing, and presentation steps (Fig. 6A), with rates derived from the literature, our experiments (Fig. 2 and Fig. S3), or a range of values derived from curve fitting (Table S1). For both GFP-RecA-HY and GFP-PRP-NP, the lag in presentation of the splice-dependent epitope shortly after induction could be entirely accounted for by the delay in splicing (Fig. 6 B and C). Moreover this model suggested that a slower kinetic of early presentation would be less apparent with an intein construct that spliced more rapidly, as observed (compare GFP-RecA-HY, Fig. 6B, with the faster-splicing GFP-PRP8-NP, Fig. 6C). Taken together, these analyses find no evidence of a substantial contribution from DRiPs to the presentation of epitopes from the RecA and PRP8 constructs, even at early time points.

Fig. 5.

Time course of epitope presentation. (A) Same as Fig. 4G, except presentation was measured at 2, 4, 12, and 24 h. (B) Same as Fig. 3G, except presentation was measured at 3, 6, 12, and 24 h.

Fig. 6.

Model and comparison between predictions and observations. (A) Diagram of the processes modeled in silico (Materials and Methods). (B and C) Presentation from cells expressing the prespliced vs. the intein constructs (GFP-RecA-HY and GFP-PRP-NP, respectively) as a function of Dox induction time. Observed points were calculated as the quotient of the slopes of the regression lines fitted in Fig. 5 A and B and plotted as mean ± SEM of at least three independent experiments. Predicted curves were obtained from a KinTek Explorer (41) simulation experiment (Materials and Methods, Table S1, and Fig. S3).

The DRiPs model would predict that an epitope that exists only in a functional protein should not be presented well (4, 12, 16). Our data show that this was not at all the case, at least for the two different antigen constructs we tested. Moreover, as opposed to earlier studies, we were able to compare the efficiency of presentation of the same epitope from the same protein when it was derived exclusively from a mature substrate (A9M or K9L) versus when it could be derived from DRiPs (S8L). The presentation efficiency of these two epitopes (i.e., splice-dependent and independent ones) that form at different times after synthesis of the same polypeptide chain was comparable. This analysis clearly showed that epitopes generated by splicing exclusively after protein folding and maturation were presented as well as ones that were generated during translation. The differences observed in the presentation efficiency at very early time points were not a result of a predominance of DRiPs, because they were accounted for by the intrinsic lag due to splicing kinetics. Taken together our data strongly argue that for our model substrates, the MHC class I antigen presentation pathway is fed by the degradation products of mature functional proteins in an efficient way, and that DRiPs are not preferentially contributing to the pathway as proposed in the DRiPs hypothesis (4).

Discussion

There has not been any prior demonstration that mature proteins are the greatest peptide contributor to MHC class I direct presentation. It was this lack of direct evidence that prompted Boon and coworkers in 1989 (5) and, once the involvement of the proteasome had been determined (3), Yewdell and coworkers in 1996 (4) to propose alternative sources of peptides for class I presentation. Our present findings are consistent with earlier observations wherein mature proteins were loaded into the cytosol of cells and their peptides were rapidly presented on MHC class I (23–25). By themselves, these earlier observations did not exclude a DRiP-dominant mechanism because the protein was added exogenously and the amount of protein needed to generate peptides was not quantified (and therefore presentation from loaded protein could have been extremely inefficient) and it was possible that the presented peptides were coming from denatured protein (DRiP-like) in the material used as an “inoculum.”

Is it possible that a substantial fraction of the products from our constructs are DRiPs? For example, could the intein moiety have completed splicing while the fused GFP domain is misfolded and therefore be rapidly degraded and presented? This is very unlikely because DRiPs are degraded extremely rapidly (t_1/2 of <10 min) (26) and therefore should be gone before splicing even begins to occur (≥30 min for RecA constructs). Another possibility is that some of the intein splice product is abnormal and becomes a rapidly degraded DRiP. This is also unlikely because the intein splicing mechanism is quite precise (27) and we show that the resulting splice product is not unstable. Nevertheless, to further address these possibilities we evaluated whether DRiPs could be detected biochemically either before or after splicing of our intein constructs. DRiPs were originally demonstrated and quantified by measuring the accumulation of polyubiquitinated species of pulse biosynthetically labeled proteins after treatment of cells with proteasome inhibitors (26). Therefore, we assayed whether polyubiquitin laddering of our constructs could be detected in cells treated with a proteasome inhibitor and pulse labeled with [³⁵S]-Met/Cys for times as short as 3 min. Using this approach, polyubiquitin laddering of our constructs before or after splicing was below the limit of detection (Fig. 2 and Fig. S3). Therefore, all together our findings argue that presentation in our system is not coming from DRiPs but rather from functional protein.

Our results do not argue against either the mere existence of DRiPs or their potential utilization as a source for presented peptides. However, they do strongly argue that DRiPS are not the major source of presented peptides in any of the systems examined here. Moreover, the fact that we do not detect a contribution from DRiPs for relatively stable antigens and at times well before the mature proteins have reached their steady-state levels, would be consistent with the synthetic error rate being relatively low, as suggested by Vabulas and Hartl (7). Whereas others have argued that even a small fraction of DRiPs (11) or of rapidly degraded pioneer translation products (18) would preferentially contribute most of the antigenic peptides, our observations make us conclude otherwise. Nevertheless, we cannot rule out the possibility that DRiPs are more important sources of presented peptides for some other antigens and/or situations, for example during certain viral infections, such as Epstein-Barr virus (28).

What about the teleological argument that by focusing on mature proteins the immune system would be inefficient, because proteins that turn over slowly would be less likely to yield peptides for presentation than proteins with a short half-life (4, 5)? It is possible that this is true in some circumstances. However, this argument does not take into account how proteins are actually degraded. Specifically, normal (non-DRiP) proteins with long half-lives are nevertheless continuously degraded and this process starts immediately after synthesis. In other words, some antigenic peptides will be available almost immediately after synthesis. These may be sufficient to stimulate T cells, which can recognize cells displaying very few peptide–MHC complexes (29). Moreover, at later time points the rate of peptide generation should be dependent on an antigen's rate of synthesis and independent of its half-life; this is because at steady state, a protein's synthesis and degradation rates are equal.

If the mature proteome is the predominant source of presented peptides, then why did inhibition of protein synthesis rapidly block antigen presentation in previous studies (9, 10) (this was considered to be the major evidence for DRiPs)? One problem is that this evidence for DRiPs was indirect and the drop in presentation observed after inhibiting protein synthesis could have been due to effects on other processes, such as a depletion of a limiting component of the class I pathway or a compensatory cellular response. Among other possibilities, MHC class I presentation clearly requires synthesis of MHC heavy and light chain and depletion of these nascent molecules leads to the degradation of TAP1 and tapasin (30), which are necessary for loading peptides onto class I molecules. Also in support of this possibility, inhibition of protein synthesis was found to block the cross presentation of exogenous antigens (9, 31), which obviously does not require antigen synthesis.

In the few examples where an inducible promoter was used to selectively terminate antigen synthesis instead of using an inhibitor that shut down total protein synthesis, antigen presentation did continue in some cases (32) and in some other cases, where presentation was diminished, the degradation of the long-lived mature protein was subsequently found to occur by autophagy (33), which may not contribute to MHC class I antigen presentation.

One of the complications of the model where DRiPs were the dominant source-presented peptides is that it required some additional mechanism to promote presentation of peptides from DRiPs, although excluding those from mature proteins (4). A number of such mechanisms have been postulated, including a specialized synthetic pathway for DRiPs (12) or a specialized degradative compartment (13) for these products, for which no firm evidence yet exists (16). Given our findings it is not necessary to postulate such mechanisms, and in fact our data strongly argue against a linked synthesis-presentation mechanism for our constructs, because epitopes that form posttranslationally are still presented efficiently.

Overall, the present findings add a different experimental point of view to the fundamental issue of what exactly is the source of the peptides being presented on MHC class I. Moreover, they are also important because the two mechanisms (DRiPs versus mature protein) have very different implications on how to develop optimal vaccine antigens and how immunodominance might be affected.

Materials and Methods

Cell Lines and T Cells.

E36 cells expressing H-2D^b and ICAM1 (E36.17.3 or E36/D^b) (34) were infected with a lentivirus encoding H-2K^b to yield E36/K^bD^b (hereafter E36). The AN9-D^b–specific 12.64-CD8αβ-LUC (hereafter 12.64) and SL8-K^b–specific RF33-Luc (hereafter RF33) T-cell hybridomas, which produce luciferase when stimulated, were described previously (35). Spleens from female HY transgenic mice (36) (Jarid1d, also called Smcy, a kind gift of Raymond Welsh, Department of Pathology, University of Massachusetts Medical School) were used as a source of HY-D^b–specific CD8 T cells. All mice were maintained in accordance with the guidelines of the Institutional Animal Care and Use Committee of the University of Massachusetts Medical School.

Molecular Cloning.

The RecA mini-intein encoding plasmid was a generous gift from Henry Paulus (Boston Biomedical Research Institute, Watertown, MA) (37). The PRP8 mini-intein encoding plasmid was kindly provided by Stephanie Pöggeler (Georg August Universität, Göttingen, Germany) (38). The GFP encoding sequence was amplified from pEGFP-C1 (Clontech), which encodes a modified Aequorea victoria GFP. Fusion constructs containing antigenic epitopes, GFP, and hemaglutinin (HA) tags ± inteins in Fig. 1C were generated by PCR, cloned into pCDNA 3.1 (Invitrogen), and subsequently moved into a modified doxycycline-inducible pTRIPZ lentiviral vector (Open Biosystems). Briefly, the (red fluorescent protein) RFP and miRNA cassette in pTRIPZ was replaced with the ORF containing the construct and a bovine growth hormone poly(A) signal from pCDNA 3.1. The heavy chain of H-2K^b was cloned into the pCDH lentivirus vector (System Biosciences).

Lentivirus Packaging.

Vesicular stomatitis virus G protein (VSV-G) pseudotyped lentivirus particles were generated by cotransfecting the lentiviral vector with the Open Biosystems lentivirus packaging mix into LentiX 293T cells (Clontech). 72 h after transfection, supernatants were collected and concentrated by 20,000 × g ultracentrifugation in an SW28 rotor (Beckman Coulter). Pellets were resuspended in 0.5 mL PBS.

Infection and Cell Purification.

E36 cells were cultured overnight in RPMI-1640 10% (vol/vol) FCS in six-well plates. Infection was performed in the presence of Polybrene (Sigma) at 10 μg/mL. Seventy-two hours later, infected cells were purified to at least 90% purity by positive selection after overnight induction with 1 μg/mL doxycycline. For positive selection, cells were detached and stained with 25-D1.16 antibody (39) in PBS containing 2% (vol/vol) FCS and 2 mM EDTA followed by a wash step and binding to Dynal beads conjugated to goat antimouse IgG according to the manufacturer’s instructions (Dynal Pan mouse IgG; Invitrogen). Cells were washed twice using a Dynamag magnet according to the manufacturer’s instructions (Invitrogen) and returned to culture. The E36 cells were maintained in G418, monitored for K^b, D^b, and ICAM-1 expression, and when necessary sorted for high uniform expression of all of the transfected genes.

Radiolabeling Experiments.

Pulse chase.

E36 cells were cultured overnight with 2 μg/mL doxycycline in RPMI-1640 10% (vol/vol) FCS in T150 flasks. Cells were trypsinized, washed with PBS, and resuspended in RPMI-1640 without l-methionine or l-cysteine (Sigma). Cells were starved in this medium for 20 min and [³⁵S]-l-Met/Cys mix was added (100 μCi/million cells; EasyTag; Perkin-Elmer) for 3 min. Cells were transferred to an ice bath and centrifuged at 300 × g for 5 min at 4° C, labeled media was removed, and cells were resuspended in chase media [RPMI-1640 10% (vol/vol) FCS with excess l-methionine and l-cysteine]. Aliquots were collected at the indicated times, centrifuged at 300 × g, and pellets lysed in TBS 0.5% DOC 1% Nonidet P-40 with a protease inhibitor mixture (Complete mini; Roche). Cold lysates were centrifuged at 14,000 × g in a tabletop microfuge and supernatants were passed through a 0.22-μm column (SpinX; Corning). Filtrates were subject to immunoprecipitation overnight at 4 °C with 15 μL Dynabeads protein G (Invitrogen) bound to HA-7 monoclonal anti-HA tag antibody (300 ng/sample; Sigma). Beads were washed three times with lysis buffer, resuspended in loading buffer (XT buffer with reducing agent; BioRad), and run on 10% Bis-Tris SDS/PAGE with Mops running buffer. Gels were dried and exposed to a phosphor storage screen (GE Healthcare). Screens were scanned using a Typhoon trio imager (GE Healthcare), and bands were analyzed with ImageJ software [National Institutes of Health (NIH), Bethesda, MD]. Prespliced and spliced bands contained fewer methionine and cysteine residues than the full-length intein bands, hence their density measurements were adjusted accordingly.

Continuous labeling.

Uninduced E36 cells were starved as described above, and grown in the presence of label (25 μCi/million cells [³⁵S]-l-Met/Cys mix; EasyTag; Perkin-Elmer) and doxycyline. Aliquots were collected at the indicated times and transferred to tubes containing excess methionine and cysteine in an ice bath. Samples were processed as described above.

Antigen Presentation Assays.

Expression of the lentivirally transduced antigen constructs in E36 cells was induced at different times by adding serially diluted doxycycline to cells growing in 96-well plates at 10⁵ cells per well in triplicate. Antigen presentation was stopped by transferring cells to an ice bath (for FACS staining) or (for hybridoma assays) by adding brefeldin A (1 μg/mL; GolgiPlug; BD Biosciences)

Immunofluorescent Staining.

Cells were trypsinized and stained with 300 ng 25-D1.16 in 30 μL PBS containing 2% (vol/vol) FCS 0.01% sodium azide (flow cytometry buffer) on ice followed by a wash step and staining with Alexa Fluor 647 goat antimouse IgG (Invitrogen). Cells were analyzed on a FACSCalibur (BD Biosciences) flow cytometer. The results were analyzed with FlowJo software (Tree Star).

Hybridoma Assays.

A total of 10⁵ hybridoma cells per well were added to 96-well plates containing E36 cells induced for different times. Cells were cocultured in the presence of brefeldin A for 16 h, harvested, and processed as described (35).

Cell Surface Presentation of K9L-D^b Complexes.

Presentation was quantified by coculturing female HY TCR-transgenic splenocytes with E36 cells induced for different lengths of time in the presence of brefeldin A and then quantifying TNFα production by intracellular cytokine staining assays (40). The response was measured as percent TNFα⁺ gated on CD8⁺ T3.70⁺ cells.

Clustal W Alignment in Fig. 1A.

Alignment parameters are as follows: EBLOSUM62 matrix, gap penalty 12, extend penalty 2. Results: 23.1% identity, 34.8% similarity.

Mathematical Analysis.

GraphPad Prism was used to fit exponential decay curves to the densities of the gel bands from pulse-chase experiments (Fig. 2). The splicing rate constants on Table S1 were derived from these curves. KinTek Explorer (41, 42) was used to model the kinetics of expression, splicing, epitope processing, and transport, using the model shown in Fig. 6A. All steps except k₅ were modeled as first order processes. k₁ and k₆ represent the translation rate for intact intein and prespliced constructs. Values were obtained by fitting densitometry values or GFP fluorescence values (Fig. S3). A process with rate equal to k₁ replenishing the mRNA pool was included to simulate a constant mRNA supply. k₂ represents the splicing rate, determined from the data in Fig. 2. k₃ represents the degradation rate of spliced or prespliced protein to epitope peptide. The value was obtained from fitting the densitometry or GFP fluorescence data (Fig. S3). k₄ represents the rate of peptide epitope generation after degrading the source protein and MHC I loading. Its value is unknown, although there are in vitro studies in the literature (43). We were able to vary this value over a wide range (0.001–1 h⁻¹) with little effect on the prespliced:intein presentation ratio. k₅ represents the delay due to MHC–peptide trafficking from the endoplasmic reticulum to the cell surface and is modeled as a fixed time delay, added to the time scale of the kinetic simulation before plotting in Fig. 6 B and C. Varying k₅ within a range of literature values (Table S1) did not significantly change the predicted presentation ratios shown in Fig. 6. The loss of surface MHC I complexes by reinternalization was found to be around 0.02 h⁻¹ by pulse-chase in a study using dendritic cells (44). Varying this surface complex turnover rate (k₈ in Fig. 6A) from 1.5 to 5 h⁻¹ in our model did not significantly affect the predicted prespliced:intein presentation ratios shown in Fig. 6 B and C, and hence it was not explicitly included in the model.