Modification of Cysteine Residues In Vitro and In Vivo Affects the Immunogenicity and Antigenicity of Major Histocompatibility Complex Class I–restricted Viral Determinants

All media, including cystine-containing and cystine/methionine-free DMEM, were purchased from Biofluids, Inc. The mastocytoma cell line P815 (H-2^d) and the thymoma cell line EL-4 (H-2^b) were maintained in DMEM containing 10% fetal bovine serum (FBS),¹ 5 × 10⁻⁵ M β-ME, antibiotics, and 2 mM glutamine (DME-10). K^d-transfected T2 (T2-K^d) and RMA-S (RMA-S/K^d) cells were cultured in RPMI 1640 containing 10% FBS and the above supplements (RP-10). CTL stimulation and maintenance were performed in RP-10 medium containing 10 U/ml of recombinant human IL-2. In some assays, cells were incubated with IMDM supplemented with 10% FBS (I-10).

8–10-wk-old female BALB/c mice or C57BL/6 mice were injected intraperitoneally with 1 ml of a 1:10 dilution of chicken egg allantoic fluid containing influenza A virus Puerto Rico/8/34 (PR8) or intravenously with 10⁷ PFU of lymphocytic choriomeningitis virus (LCMV) WE strain. Splenocytes were stimulated with peptide-pulsed APCs for 7 d in vitro at least 3 wk after virus priming. Cells were stimulated in RP-10 with 10 U/ml IL-2 in 6-well plates. In brief, 3 × 10⁷ splenocytes were stimulated with 6 × 10⁵ virus-infected or peptide-pulsed (1 nM) APC, which were irradiated with 200 Gy before addition to cultures. After 4 d, live cells were recovered via a Ficoll–Hypaque gradient and recultured with fresh IL-2–containing medium. CTL activities were tested after 7 d in standard ⁵¹Cr-release microcytotoxicity assays. In some cases, short-term CTL lines were used, which were restimulated weekly for up to 5–6 wk.

For PR8 infection, log-phase cells were harvested and washed in serum-free Autopow MEM (Life Technologies), adjusted to pH 6.6, and resuspended in 200 μl of the same medium containing 100 μl PR8-containing allantoic fluid per 10⁶ cells. For vaccinia virus (VV) infection, cells were infected at a multiplicity of infection of 10 in basal salt solution supplemented with 0.1% BSA (wt/vol) at a concentration of 5 × 10⁶ cells/ml. Cells were incubated for 1 h at 37°C in a water bath before 2 ml of prewarmed complete medium was added. Cells were then further incubated for 1–2 h before ⁵¹Cr labeling, unless indicated otherwise.

All peptides were synthesized, HPLC-purified, and analyzed by mass spectrometry by the Biologic Resource Branch, National Institute of Allergy and Infectious Diseases. Peptides were dissolved at 1 mM in DMSO and stored at −20°C unless otherwise indicated. For analytical purposes, peptide masses were determined by matrix assisted laser desorption ionization with time of flight detection (MALDI-TOF) using a Hewlett Packard mass spectrometer (model G2025A) and cyano-4-hydroxycinnamic acid as the matrix. Peptide binding to live cells was determined by protection of class I molecules to melting (6). In brief, T2-K^d cells were cultured for 14–16 h at 26°C. Synthetic peptides, diluted in FBS-free DME in the presence or absence of tris (2-carboxyethyl) phosphine hydrochloride (TCEP; Pierce Chemical Co.) as indicated, were added to cells, which were incubated at 37°C for 2 h to denature K^d molecules not stabilized by peptide binding. Cells were then washed and stained with fluorescein-conjugated SF1-1.1 (PharMingen). Live cells were gated based on scattering and exclusion of ethidium homodimer (Molecular Probes, Inc.) present at 5 μg/ml for 5 min before the last wash. For each histogram, 10,000–20,000 cells were counted on a Becton Dickinson FACScan™, and live cells were analyzed using CELLQuest™ software (Becton Dickinson).

Natural peptides were recovered and analyzed as previously described (7). In brief, cultures of P815 cells were expanded in roller bottles. 5 × 10⁸–10⁹ cells were infected as described above and incubated for 6 h at 37°C before being pelleted, washed twice in PBS, lysed with ice-cold 0.33% TFA/ H₂O, and further disrupted using a TenBroek tissue homogenizer on ice. At this stage, synthetic peptides were added to uninfected P815 cells as control. Lysates were sonicated and centrifuged at 10,000 g for 30 min, and the supernatants were passed through a 3K cutoff filter (Macrosep™ filtron 3K; Pall Filtron Corp.). Samples were dried to a volume >400 μl using a SpeedVac (Savant Instruments, Inc.) and fractionated on a C18 column (Deltapack; Waters) at 1 ml/min on TFA/acetonitrile gradient (7). Either 0.25- or 1-ml fractions were collected.

Generally, 10⁶ target cells were labeled with 100 μCi of Na⁵¹CrO₄ (Dupont) in minimum volume of medium at 37°C for 60 min. For some experiments, RMA-S/K^d cells that had been cultured for 12–14 h at 26°C were labeled at 26°C for the same time period. After two washes, 10⁴ cells were aliquoted into round-bottom, 96-well plates containing serial dilutions of effector T_CD8+. For testing HPLC fractions, target cells in 50 μl of either PBS or FCS-free medium were exposed to 5 μl of fractions for 30 min at 26°C before T_CD8+ were added. In some experiments, TCEP was freshly dissolved in H₂O and used at 200 μM, both at peptide-pulsing and microcytotoxicity assay stages. The radioactivity in supernatants collected after 4–6-h incubation at 37°C was determined using a gamma counter. The percent specific release was then determined as: % specific release = (CTL-induced release − spontaneous release)/(release by detergent − spontaneous release) × 100.

We previously reported that in K^d-restricted responses to PR8 influenza virus nuclear protein (NP), NP_147–155 is the immunodominant determinant, with NP_39–47 and NP_218–226 exhibiting subdominant status (8). This hierarchy is not accounted for by peptide affinity, as NP_147–155 binds to K^d with the lowest efficiency, as determined by a K^d “melting” assay performed either with RMA-S cells expressing K^d from a transfected gene (data not shown) or T2 cells (8). We initially focused on T_CD8+ avidity to explain the immunodominance of NP_147–155, as 10-fold less synthetic NP_147–155 was usually required to sensitize target cells for lysis by T_CD8+ lines raised to the individual peptides under conditions similar to those used for the peptide binding assay (Fig. 1 B). This was observed using either short- or long-term lines stimulated in vitro by synthetic peptides derived from animals immunized either with PR8 or rVV expressing NP or cytosolic or endoplasmic reticulum (ER)-targeted minigene product versions of the determinants. Taking into account the lower efficiency of NP_147–155 binding to K^d, the data in Fig. 1 suggest that <10% of K^d-NP_147–155 complexes are required for T_CD8+ triggering relative to K^d complexed to either of the subdominant determinants.

Several findings, however, suggested that matters might be a bit more complicated. Unlike NP_147–155, the dose–response curves of NP_39–47 and NP_218–226 varied considerably between experiments, depending in part on the manner in which the assay was executed. We also experienced difficulties in stimulating and maintaining T_CD8+ lines to these subdominant determinants, often observing slower growth after restimulation and morphological abnormalities of the cells, which were frequently larger than T_CD8+ specific for NP_147–155. This was not strictly related to the subdominant status of these determinants, as T_CD8+ specific for other subdominant determinants behaved similarly to NP_147–155-specific T_CD8+.

A property shared by NP_218–226 and NP_39–47 is the presence of cysteine (Table I). The report by Meadows et al. (4) demonstrating the dramatic effects of sulfhydryl modification of cysteine-containing residues on T_CD8+ recognition prompted us to examine possible effects of cysteine modification on NP_39–47 and NP_218–226 binding and antigenicity. We first studied the properties of synthetic peptides in which cysteine is replaced by serine or alanine (the most conservative substitutions). For NP_39–47, the cysteine→ serine substitution had no significant effect on peptide binding, whereas the cysteine→ alanine substitution increased peptide potency by  ∼10-fold (Fig. 2 A). For NP_218–226, either substitution reduced peptide potency in stabilizing K^d molecules by ∼10-fold (Fig. 2 B). Each of the substitutions resulted in large increases in antigenicity relative to the wild-type peptides (Fig. 2, C and D). For NP_218–226, the substituted peptides were 1,000–10,000-fold more antigenic on a per-complex basis (assuming that complex formation at the endpoint of peptide titrations is proportional to peptide binding efficiency determined by the melting assay).

These findings prompted us to study the effects of sulfhydryl modification on the antigenicity of NP_39–47 and NP_218–226. In most experiments described below, the two peptides were studied in parallel. Because the results were highly similar, only results with NP_218–226 are shown for the sake of clarity and simplicity.

Cysteine readily forms disulfide bonds at neutral or slightly basic pH in the presence of O₂ at atmospheric tension, and oxidation to the disulfide is stimulated by trace amounts of iron salts that are present in tissue culture media. The disulfide can be reduced to the original thiol form by exposure to reducing agents. To determine whether disulfide formation affected peptide antigenicity, synthetic NP_218–226 was added to cells in the presence of dithiothreitol or TCEP, and cells were tested for lysis by NP_218–226-specific T_CD8+. Either of these reducing agents increased peptide potency by ∼10-fold (Fig. 3 A). Reducing agents did not affect the potency of noncysteine-containing peptides, including NP_147–155 (Fig. 3 B), an LCMV peptide (described below), or the cysteine→ serine- or cysteine→ alanine-substituted NP_39–47 peptides (not shown). Enhancement of cysteine peptide recognition is, as expected, dependent on the concentration of reducing agent, with TCEP being more effective on a molar basis than dithiothreitol (not shown). The optimal concentration for TCEP was 200 μM (used in additional experiments), as higher concentrations (1 mM) sometimes increased spontaneous release values in ⁵¹Cr-release assays. The effect of TCEP on NP_218–226 antigenicity is particularly impressive when considered in view of the 100–1,000-fold decrease in peptide binding to K^d in the presence of TCEP (described below).

In additional experiments, we found that inclusion of reducing agents in the media used to stimulate and propagate T_CD8+ specific for NP_218–226 or NP_39–47 greatly enhanced their growth and altered their appearance, to the extent that these T_CD8+ were indistinguishable from T_CD8+ raised to NP_147–155 or other immunodominant determinants.

These findings indicated that sulfhydryl modification can have major effects on the antigenicity and immunogenicity of synthetic peptides, effects that can lead to erroneous conclusions regarding the nature of their interactions with class I molecules, the affinity of T_CD8+ specific for the peptide, and the growth characteristics of the cells. To broaden these findings, we examined the effect of reducing agents on the in vitro antigenicity of  three oft-studied, D^b-restricted determinants from LCMV: two containing cysteine (GP_33–41 and GP_276–284) and a control determinant lacking cysteine (NP_396–404) (Table I). As with the influenza virus determinants, the cysteine-containing peptides were recognized 10–100-fold less efficiently than the cysteine-free peptide in the absence of reducing agent (Fig. 4, A and B), whereas TCEP had no affect on the antigenicity of NP_396–404 (Fig. 4 C).

Based on these findings, it is clear that sulfhydryl modification of cysteine-containing synthetic peptides can have major effects on T_CD8+ growth and target cell recognition that must be taken into consideration in investigating the biological properties of the corresponding naturally produced determinants and the T_CD8+ they induce.

We turned our attention to why reducing agents enhance the antigenicity of synthetic NP_218–226. The major possibilities were reduction of disulfide-linked peptide dimers and reduction of disulfide-bound species derived from culture media. The most abundant sulfhydryl-containing compound present in DMEM is cystine (cysteine–cysteine dimers). To determine whether NP_218–226 becomes cysteinylated in DMEM, freshly dissolved NP_218–226 was incubated in normal or cystine-free DMEM for 2 h. Peptides present in media were then separated by RP-HPLC, and the masses in peptide-containing fractions were determined by mass spectroscopy (Fig. 5). After incubation in cysteine-free media, the only modification detected was a small amount of dimerization (Fig. 5 B). By contrast, in cysteine-containing media, most of the monomer was converted to a separate eluting form representing cysteinylated peptide (Fig. 5 C). This fraction also contained a minor species with an additional mass of 16 daltons that probably represents oxidation of the neighboring methionine residue in the peptide.

The effect of cysteinylation on NP_218–226 antigenicity was examined by measuring the K^d binding and antigenicity of an unmodified preparation of NP_218–226 incubated for 2 h in cystine-containing DMEM with or without TCEP before addition to cells. As seen in Fig. 6 A, the resulting cysteinylation was associated with an ∼10-fold increase in rescuing K^d molecules from melting and an ∼10,000-fold decrease in capacity to sensitize target cells for lysis by T_CD8+ induced by APCs pulsed with unmodified NP_218–226.

We also examined the effect of NP_218–226 dimerization on antigenicity and K^d binding. Analysis of various stocks by RP-HPLC in conjunction with mass spectrometry revealed that a 1-yr-old stock of peptide in DMSO was >95% dimerized (Fig. 5 D). Using this stock as an NP_218–226 dimer source, we investigated the effect of  2-h incubation at room temperature in cystine-free or cystine-containing DMEM. NP_218–226 dimers were stable under these conditions (Fig. 5, E and F). Having identified a source of dimers and demonstrated the stability of dimers in DMEM, we could examine the K^d binding and antigenicity of dimers (Fig. 6), which revealed that dimers behaved similarly to cysteinylated NP_218–226.

We draw two conclusions from these findings. First, cysteinylation and dimerization of  NP_218–226 is associated with enhanced K^d binding yet greatly reduced antigenicity, using T_CD8+ restimulated by the reduced peptide. Second, in normal DMEM, cysteinylation occurs preferentially to peptide dimerization, even when NP_218–226 is present at relatively high concentrations. As a second-order reaction, dimerization should be greatly disfavored at decreasing peptide concentrations, whereas cysteinylation continues at a first-order reaction rate. Therefore, cysteinylation is probably the major process for modifying cysteine-containing peptides at the concentrations used in K^d-binding and ⁵¹Cr-release assays.

Given the potential for cysteine modification in vitro, we examined whether NP_218–226 produced by PR8-infected cells was modified in vivo. Low M_r peptides present in acid extracts from whole cells were fractionated by RP-HPLC and tested for their abilities to sensitize target cells for lysis by T_CD8+ raised to reduced NP_218–226. TCEP was added to the fractions to reveal the presence of SH-modified forms of peptides rendered nonantigenic by the modification. As shown in Fig. 7 A, antigenic peptides were recovered in fractions eluting from 25–27 min, matching the elution times of cysteinylated (25 min) and unmodified (27 min) NP_218–226. No activity was present in the 29-min fraction, where dimeric NP_218–226 elutes. The amounts of peptide recovered were well below that required for saturation (peak lysis of column fraction of 40% versus 70% with a saturating amount of synthetic peptide; Fig. 7 A, top right), suggesting that a considerable fraction of NP_218–226 recovered from PR8-infected cells is cysteinylated.

Cysteinylation of NP_218–226 might have occurred artefactually during the extraction process. To examine this possibility, cell homogenates were doped with synthetic, unmodified NP_218–226 and then processed identically to virus-infected cells (note that in this and subsequent experiments, to increase the chromatographic resolution, fraction size was reduced from 1 ml to 0.25 ml) (Fig. 7 B). In this case, <1% of the antigenic activity (as determined by titrating fractions; data not shown) was recovered in the cysteinylated form. In this experiment, peptides were tested in the presence or absence of TCEP. Even unmodified NP_218–226 required TCEP treatment, which we attribute to rapid peptide cysteinylation during target cell sensitization.

To examine themaximal potential for posttranslational modification of NP_218–226, we infected cells with an rVV (VV-NP_218–226) that expresses the peptide in the cytosol as a minigene product (with NH₂-terminal methionine to enable translation initiation). As reported previously (7–9), this greatly enhances the number of peptide–class I complexes generated by cells. RP-HPLC fractionation from minigene-expressing cells revealed the presence of material coeluting with unmodified NP_218–226, as well as cysteinylated peptide (Fig. 7 C). For the first time, dimeric peptide was also detected. These activities cannot be attributed to methionine-extended NP_218–226, which, for each form, elutes slightly later than NP_218–226 (data not shown). Titration of the antigenic activities in the fractions (not shown) revealed a ratio of unmodified/cysteinylated/dimeric forms of  ∼6:3:1.

In an additional experiment (data not shown), we examined the K^d dependence of NP_218–226 recovery in HPLC fractions after VV-NP_218–226by using L929 cells and L929 cells expressing K^d from a transgene. Peptides corresponding to cysteinylated and unmodified NP_218–226 were recovered from L929-K^d cells but not L929 cells. Dilution of peak fractions revealed that expression of K^d resulted in at least a 25-fold increase in the recovery of NP_218–226. This confirms numerous prior studies demonstrating the MHC dependence of antigenic peptide recovery (1).

The ultimate demonstration of the biological relevance of cysteinylated NP_218–226 is that specific T_CD8+ are elicited in PR8-infected mice. We could show this by stimulating splenocytes derived from PR8-infected mice with RP-HPLC–purified, cysteinylated peptide. After three to four rounds of stimulation, we obtained T_CD8+ that preferentially recognize cysteinylated peptide (Fig. 8 A). In the same assay, the noncysteinylated peptide is preferentially recognized by T_CD8+ induced in the standard manner (Fig. 8 B). The recovery of  T_CD8+ specific for the cysteinylated peptide could not be attributed to in vitro stimulation of naive T_CD8+, as we failed to obtain any activity using splenocytes from nonimmunized mice (data not shown).

Using T_CD8+ stimulated by the cysteinylated NP_218–226, it was possible to formally demonstrate that the 25-min fraction derived from minigene-expressing cells contained cysteinylated peptide (Fig. 7 D). Indeed, now antigenicity was destroyed by TCEP exposure, in contrast to the enhancing activity observed with other peptides. We also detected an additional peak at 24.5 min. This probably represents the methionine-oxidized form of cysteinylated NP_218–226, which, based on experience with other peptides, often elutes slightly faster than the nonoxidized form, providing evidence that this modification occurs in cells.

In this paper, we confirm and extend the findings of Meadows et al. (4) that modification of the SH group of cysteine-containing peptides has important positive and negative effects on their antigenicity and immunogenicity in vitro and in vivo. Failure to consider these effects can have disastrous consequences for the accurate interpretation of several different types of experiments.

In our own case, the use of synthetic peptides in the absence of reducing agents led us to erroneously favor the idea that the subdominant status of two cysteine-containing peptides was due to a greater number of complexes required for T_CD8+ recognition and correlated with the atypical growth of T_CD8+ in vitro. In other experiments (our unpublished results), it also led us to the incorrect quantitation of peptides recovered from virus-infected cells. A further potential methodological pitfall is that autooxidation of reducing agents can cause additional artifacts. This can be minimized by inclusion of a chelating agent, such as DTPA (diethylenetriaminepentaacetic acid), with the reducing agent.

These errors are probably widespread. We demonstrate that two cysteine-containing LCMV determinants restricted by a different class I molecule are also modified in vitro through their sulfhydryl groups, with a resulting 10–100-fold loss in antigenicity. These determinants were the subject of a recent study (10) focused on the factors involved in the immunodominance hierarchy of the determinants. The failure to add reducing agents during peptide titration probably led to erroneous calculations of peptides present in virus-infected cells, particularly because the peptide titration curves were nearly identical to those we obtained in the absence of reducing agents. Results obtained in this study with in vivo transfer of T_CD8+ lines must also be questioned, as the cells were propagated in vitro with synthetic peptides in the absence of reducing agents. In another recent study of the fine specificity of a T_CD8+ clone for a cysteine-containing peptide, amino acids were substituted for cysteine, many of which enhanced the antigenicity of the peptide (11). Based on our findings, we would predict that simple reduction of the wild-type peptide would have a similar (or greater) effect.

In addition to reducing antigenicity, modification of cysteine can result in the generation of T_CD8+ specific for the modified determinant. We show that PR8-infected mice generate T_CD8+ that prefer cysteinylated NP_218–226. We also demonstrate that PR8-infected cells generate an SH-modified peptide that coelutes with cysteinylated NP_218–226 and is recognized by T_CD8+ specific for cysteinylated NP_218–226. This species almost certainly represents cysteinylated NP_218–226, although definitive evidence requires mass spectroscopy. The recognition of posttranslationally modified peptides by T_CD8+ adds to the already formidable challenge of understanding in vivo T_CD8+ responses but can be ignored only at the peril of the investigator.

We can only speculate where NP_218–226 is cysteinylated during its processing and presentation by virus-infected cells. It is theoretically possible that the cystine derives naturally from a disulfide bond present in NP. It is difficult, though admittedly not impossible, to imagine the proteolytic liberation of cysteinylated NP_218–226. Given the highly reducing environment of the cytosol and nucleus, it also seems unlikely that cysteinylation would occur before peptide translocation into the ER. The ER provides a much more oxidizing environment and possesses resident proteins that catalyze thiol–disulfide interchange, including protein disulfide isomerase (12). There is evidence that exogenous homocysteine is added to HLA class I molecules in an early secretory compartment (13). If cysteinylation occurs in the ER, it may occur before peptide loading onto class I molecules. Alternatively, cysteinylation could occur after peptide binding, particularly in the case of NP_218–226, as the SH must be directed away from the groove (so as to accommodate dimer binding). In this case, it could occur anywhere from the ER to the cell surface. Cysteine is thought to be the major reductant in the endosomal pathway, and although there is no evidence that functional K^d molecules visit these compartments, this remains a possibility.

It was previously reported that disulfide-linked homodimeric peptides could bind to class I molecules (14). We provide another example of this and further demonstrate that this can occur naturally in cells. Recovery of dimers required expression of NP_218–226 as a cytosolic minigene, which results in at least 10–1,000-fold overproduction of peptide–class I complexes relative to expression of peptide in its natural context (7, 9), and the extent to which overexpression of peptide is required remains to be determined. As peptide cross-linking is expected to be a second-order reaction, it implies that NP_218–226 is present at a very high concentration intracellularly. It is hard to imagine this occurring outside of the ER, and it may occur in the vicinity of TAP, which is required for transport of NP_218–226 into the ER (our unpublished results).

These findings have important clinical implications. First, for synthetic peptide vaccines (or other exogenous antigen vaccine preparations with vulnerable cysteine residues in antigenic peptides), modification of the cysteine in vitro or in vivo can obviously have major negative effects on immunogenicity. This can be avoided by modifying the side chain to a nonreactive form. For NP_218–226 and NP_39–47, this is achieved simply by substitution with alanine or serine, which did not detrimentally affect peptide binding or T_CD8+ triggering. This strategy will probably work for most peptides. For others, it is possible that chemical modification of the SH group (e.g., treatment with an alkylating agent or a heavy metal) will do the job. Second, if increased cysteinylation is associated with a disease process, this could lead to autoimmune recognition of cysteinylated self peptides.

ER

endoplasmic reticulum

FBS

fetal bovine serum

LCMV

lymphocytic choriomeningitis virus

NP

nuclear protein

RP-HPLC

reversed-phase HPLC

VV

vaccinia virus