Constitutive Display of Cryptic Translation Products by MHC Class I Molecules

See allHide authors and affiliations

Science  05 Sep 2003:
Vol. 301, Issue 5638, pp. 1367-1371
DOI: 10.1126/science.1085650


Major histocompatibility complex (MHC) class I molecules display tens of thousands of peptides on the cell surface, derived from virtually all endogenous proteins, for inspection by cytotoxic T cells (CTLs). We show that, in normal mouse cells, MHC I molecules present a peptide encoded in the 3′ “untranslated” region. Despite its rarity, the peptide elicits CTL responses and induces self-tolerance, establishing that immune surveillance extends well beyond conventional polypeptides. Furthermore, translation of this cryptic peptide occurs by a previously unknown mechanism that decodes the CUG initiation codon as leucine rather than the canonical methionine.

The set of peptides displayed by MHC class I molecules comprises the information available to CTLs, which detect the presence of viral, bacterial, or otherwise foreign peptides in this complex mix. Recent evidence has questioned the conventional wisdom that most of these peptides are derived from cellular protein turnover (14). Instead, it appears that a major source of these peptides is newly synthesized polypeptides (5, 6). Independently, examples of antigenic peptides derived from “noncoding” regions of mRNA 5′ and 3′ of the open reading frame, and in alternate reading frames, have accumulated (3). These reports suggest that an additional source of information is available to CTLs surveying for a foreign presence inside the cell (7). We developed a transgenic mouse model to determine whether such “cryptic” peptides derived from noncoding regions were anomalies found in rare tumor or virally infected cells or whether cryptic translation contributes to the pool of available antigenic peptides within normal cells.

The bicistronic transgene we used encoded two peptides. The first peptide, WI9, derived from the Y-chromosome Uty gene (8), was in a conventional translational context with an up-stream inframe AUG initiation codon. The second peptide, LYL8, derived from the H60 histocompatibility gene (9), was placed in the “untranslatable” region downstream of the stop codon and with CUG as its initiation codon (Fig. 1A) (10). Transcription of the transgene was regulated by the Kb MHC I promoter to ensure wide tissue expression (11). A T cell assay was used to detect expression of the conventional WI9 peptide and the “cryptic” LYL8 peptide by transformed tail fibroblasts derived from eight founder mice. In each cell line where the conventional WI9 peptide was detected by the WI9-Db–specific 11p9Z T cell hybridoma, the LYL8 peptide was detected at varying levels by the LYL8-Kb–specific BCZ103 T cell hybridoma (Fig. 1B). Furthermore, cells from primary and secondary lymphoid organs of transgenic but not wild-type mice stimulated the LYL8-Kb–specific BCZ103 T cells (Fig. 1C). These responses were specific, because they were blocked by antibodies to the appropriate MHC molecules.

Fig. 1.

Cryptic LYL8 peptide–Kb MHC I complexes are expressed in transgenic mouse cells. (A) The WI9*LYL8 DNA construct used to generate transgenic (Tg) mice. (B) Response of LYL8-Kb–specific BCZ103 or WI9-Db–specific 11p9Z T cell hybridomas to transformed tail fibroblasts from eight founder mice. (C) Response of T cell hybridomas with the indicated specificities to splenocytes (top) and thymocytes (bottom) from transgenic and nontransgenic mice. Tg indicated by ⚫; Tg+α-Kb, ◼; Tg+α-Db, ▲; B6, ◯; B6+α-Kb, ▢; B6+α-Db, △.

Because of cryptic peptides' potential to elicit an immune response, we determined which specific cell types presented LYL8. We began with dendritic cells (DCs), because these are key to initiating a T cell response (12). Activated DCs isolated by flow cytometry from transgenic mice were efficiently recognized by the LYL8-Kb–specific BCZ103 T cells (Fig. 2A). Although this result suggested that cryptic translation could be an important source of antigenic peptides, the high sensitivity with which DCs were detected posed a problem for assessing other cell types' ability to translate the LYL8 peptide because of potential DC contamination.

Fig. 2.

Distinct lymphoid and nonlymphoid cell types express the cryptic LYL8 peptide–Kb complex. (A) Response of LYL8-Kb–specific BCZ103 T cells to CD11c+B7.2+ bone marrow–derived DCs. (B) Specific lysis of transgenic splenocytes by LYL8-Kb–specific CTL. The left contour plot shows B220-PE-Cy5 compared with YoPro1 staining of target cells after lysis at a 2:1 effector/target ratio. The right contour plot is identical, but the cells were incubated with unconjugated antibodies against B220 before staining with PE-Cy5–conjugated antibodies against B220. (C) Specific lysis of transgenic tail fibroblasts by LYL8-Kb–specific CTL. The left contour plot compares staining by PE-Cy5–labeled antibodies against CD105 with YoPro1 staining of target cells after lysis at a 201 effector/target ratio. The right contour plot shows staining of transgenic fibroblasts, nontransgenic fibroblasts, and DCs with antibodies against Ab.

To overcome this problem, we developed an assay to identify cells translating the LYL8 peptide at an individual level. A mixture containing the cell type to be tested was first stained with the membrane-binding dye PKH26. Unlike 51Cr, which labels only proliferating cells, PKH26 can label both resting and dividing cells. LYL8-Kb–specific cytotoxic T cells were then added to the culture, where they lysed cells expressing the LYL8 peptide. After the T cells had been allowed to kill, the mix was stained with the DNA binding dye Yo-Pro1, which is membrane-impermeant and can therefore mark only dead and dying cells. Antibodies to cell-specific markers were then used to distinguish cell types of interest. Analysis of the mixture by three-color flow cytometry could reveal for each cell whether it was a potential target (PKH26+), whether it had been killed and hence expressed LYL8 (YoPrc+), and whether it expressed the surface marker of interest (Antibody+).

With the use of this method, we assessed expression of the cryptic LYL8 peptide by non-DCs. In a 3-hour incubation, LYL8-Kb–specific CTLs lysed ∼30% of spleen cells from transgenic mice but failed to kill cells from nontransgenic mice. The dead target cells included those expressing the B cell marker B220 (Fig. 2B). Similarly, tail fibroblasts from transgenic but not nontransgenic mice were lysed by LYL8-Kb–specific CTLs (Fig. 2C). The dead fibroblasts expressed CD105, a transforming growth factor–β receptor commonly present on stromal and endothelial cells (13). The absence of DCs was confirmed by failure to detect MHC II–expressing cells in the fibroblast cultures (Fig. 2C). Thus, DCs, B cells, and fibroblasts were capable of translating the cryptic LYL8 peptide.

Next, we assessed the impact of the cryptic LYL8 peptide on the CTL repertoire. Nontransgenic female mice were immunized with spleen cells from syngeneic male transgenic mice to provide CD4 T cell help by epitopes encoded within the male Y chromosome (14). Host splenocytes were then examined for the presence of activated LYL8- and WI9-specific CD8 T cells by intracellular cytokine staining (Fig. 3). Significantly greater numbers of CTLs specific for LYL8 and WI9 were detected as compared with numbers for the negative control peptides AFNV8 and NP. This result is particularly notable because the response to WI9 was about threefold larger than the response to LYL8, whereas on average transgenic spleen cells expressed over 100 copies of WI9 and only one copy of the cryptic LYL8 peptide (10). To test whether the T cell repertoire of the transgenic mice was influenced by the cryptic LYL8 peptide, we immunized female transgenic and nontransgenic mice with spleen cells from male BALB.B mice (Fig. 3). BALB.B mice share the same MHC molecules with B6 mice but differ in the H60 and H28 histocompatibility loci that, respectively, encode the LYL8 and IFL8 peptides in a conventional translational context (9, 15). Nontransgenic B6 mice responded very strongly to the immunodominant LYL8 peptide, less strongly to the IFL8, and poorly to the WI9 peptide, as seen earlier (15, 16). In contrast, LYL8-specific IFN-γ+CD8+ (where IFN-γ indicates interferon-γ) cells were undetectable in the spleens of transgenic mice, whereas the fraction of IFL8-specific IFN-γ+CD8+ T cells remained unchanged (Fig. 3, B and C). Thus, the cryptic LYL8 peptide was capable of eliciting a CTL response and inducing self-tolerance.

Fig. 3.

The cryptic LYL8 peptide is immunogenic and induces tolerance. The two-dimensional flow cytometry plots show expression of CD8 and intracellular IFN-γ by recipient spleen cells immunized as indicated and restimulated in the presence of the indicated peptide. Numbers are the percentage of CD8+ cells that stained positive for IFN-γ. The data are typical of 10 to 14 mice in at least four different experiments.

Two apparent obstacles exist to translation of LYL8: First, it is 3′ of the stop codon that terminates WI9, and, second, it begins with CUG rather than AUG. It is believed that cells express only one class of initiator tRNA, which is specific for AUG and is always loaded with methionine (17). Rare cases of initiation at non-AUG codons are thought to be caused by “wobble” in the pairing of the non-AUG codon with the anticodon of the methionyl initiator tRNA. This mispairing results in incorporation of the methionine residue (18, 19). If wobble were responsible for generation of the cryptic peptide, the first amino acid would be expected to be methionine rather than leucine, normally specified by the CUG codon. Although the BCZ103 T cells recognize LYL8 with leucine as the first residue and cross-react with the MTFNYRNL (MYL8) peptide with methionine as the first residue, high-performance liquid chromatography (HPLC) fractionation can clearly resolve LYL8 from MYL8 (Fig. 4A) (20). We prepared a peptide extract from transgenic spleen cells and fractionated it by HPLC. Each fraction was tested for the presence of LYL8 and MYL8 with the use of BCZ103 T cells and Kb-expressing antigen-presenting cells (APCs) (Fig. 4A). Remarkably, the single peak of antigenic activity detected in the transgenic spleen cell extract co-eluted with LYL8 rather than MYL8. The same LYL8 peptide was also found in extracts of DCs and transformed tail fibroblasts (fig. S1) (10). The initiating CUG was decoded as leucine.

Fig. 4.

The CUG initiation codon is decoded as leucine rather than methionine. (A) Synthetic peptides and transgenic spleen cell extracts were separated by reversed-phase HPLC. Fractions were tested for BCZ103 T cell–stimulating activity with Kb-L cells as APCs. (B) Extracts from COS-7 cells transfected with cDNA encoding Kb and the indicated constructs were analyzed as in (A). (C and D) Lmtk cells transfected with the indicated construct and the appropriate MHC molecule were tested for LYL8-Kb expression with the use of the BCZ103 T cell hybridoma and WI9-Db expression with the use of the 11p9Z hybridoma. (E to G) Lmtk cells transfected with the indicated constructs, Kb, and B7.2 were tested for LYL8-Kb expression with the use of the BCZ103 T cell.

The apparent use of leucine as the initiating amino acid may have been because of a posttranslational modification of the MYL8 peptide, which caused it to co-elute with LYL8 on the HPLC. However, when cells were transfected with DNA constructs encoding the peptide with ATG or CTG as the initiation codon, only the appropriate MYL8 and LYL8 peptides were detected in the peptide extracts (Fig. 4B). The cells' ability to make MYL8 would therefore rule this out as an explanation for the use of leucine.

We next explored the possibility that leucine was used as the first amino acid because the ribosome may have read through the stop codon terminating the WI9 peptide (Fig. 1A). If readthrough were responsible, one would predict that expression of the downstream LYL8 would be proportional to expression of the upstream WI9. In transient transfection experiments, shifting the bicistronic WI9*LYL8 cassette out of frame with the AUG that initiates WI9 translation substantially lowered the level of WI9 but did not change the level of LYL8 (Fig. 4, C and D). Thus, generation of LYL8 was independent of upstream initiation events. Furthermore, if readthrough were responsible, one would predict that LYL8 expression would be inversely proportional to the number of stop codons preceding the LYL8 coding sequence. However, DNA constructs containing one, three, or even six inframe stop codons were equivalent in the level of LYL8 expression by transfected cells (Fig. 4E and fig. S2). In these constructs, there was no open reading frame upstream of LYL8, meaning that any readthrough would have had to result from an alternate upstream initiation codon. Finally, if readthrough and decoding of the CUG codon during elongation were responsible for LYL8 expression, the choice of a specific leucine codon would not affect expression of LYL8. Initiation, by contrast, is known to be efficient only with specific codons (20). Changing the initiating CUG codon to alternate leucine-encoding CUC, CUA, or UUG codons dramatically reduced the expression of LYL8 (Fig. 4F). However, when the codon for the final leucine residue of the LYL8 peptide was changed to CUG, CUC, CUA, or UUG, there was no change in LYL8 expression (Fig. 4G). This result also argues against RNA modification that somehow introduced an upstream AUG as a mechanism to explain the leucine start, because it would require the presence of both the C and G of the CUG codon. Furthermore, the 5′ untranslated region can affect the level of LYL8, suggesting that the mRNA is intact (fig. S3). Taken together, these independent experiments provide strong evidence that leucine is used as a genuine initiating amino acid.

We show that translation of “noncoding” regions occurs constitutively in normal cells. These polypeptides could be particularly valuable for immune surveillance in the frequent case where the conventional protein had few epitopes that could squeeze through the bottle-necks in the antigen-processing pathway (14). In addition, the translation initiation mechanism for generating these peptides can use a leucine rather than a methionine residue. Understanding this mechanism should illuminate not only the expression of otherwise hidden antigenic peptides but also a growing list of non-AUG–initiated genes in the sequenced genomes.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

References and Notes

View Abstract

Navigate This Article