FC and JWK conceived and designed the experiments. FC, EH, JW, and JWK performed the experiments. JWK and PM wrote the paper
The authors have declared that no conflicts of interest exist.
The use of peptide libraries for the identification and characterization of T cell antigen peptide epitopes and mimotopes has been hampered by the need to form complexes between the peptides and an appropriate MHC molecule in order to construct a complete T cell ligand. We have developed a baculovirus-based peptide library method in which the sequence encoding the peptide is embedded within the genes for the MHC molecule in the viral DNA, such that insect cells infected with virus encoding a library of different peptides each displays a unique peptide–MHC complex on its surface. We have fished in such a library with two different fluorescent soluble T cell receptors (TCRs), one highly peptide specific and the other broadly allo-MHC specific and hypothesized to be much less focused on the peptide portion of the ligand. A single peptide sequence was selected by the former αβTCR that, not unexpectedly, was highly related to the immunizing peptide. As hypothesized, the other αβTCR selected a large family of peptides, related only by a similarity to the immunizing peptide at the p5 position. These findings have implications for the relative importance of peptide and MHC in TCR ligand recognition. This display method has broad applications in T cell epitope identification and manipulation and should be useful in general in studying interactions between complex proteins.
A baculovirus expression library -- encoding peptides in the context of MHC molecules -- has been developed to identify and characterize unique peptide-MHC complexes that bind specific T cell receptors.
The identification of peptide epitopes associated with particular αβ T cell receptors (αβTCRs) is often still a bottleneck in studying T cells and their antigenic targets in, for example, autoimmunity, hypersensitivity, and cancer. A direct genetic or biochemical attack on this problem can be successful, especially with class I major histocompatibility complex (MHCI)-presented peptides. For example, tumor (
The reward for the labor involved in identifying peptide epitopes directly can often be the identification of the protein source of the peptide, especially as the sequencing of the genomes of many organisms approaches completion. However, in many situations, rather than identifying this precise peptide epitope, it is sufficient to identify a peptide “mimotope.” Mimotopes can be defined as peptides that are different in sequence from the actual peptide recognized in vivo, but that are nevertheless capable of binding to the appropriate MHC molecule to form a ligand that can be recognized by the αβTCR in question. These peptides can be very useful for studying the T cell in vitro, for altering the immunological state of the T cell in vivo (
Mimotopes can sometimes be identified in randomized peptide libraries that can be screened for presentation by a particular MHC molecule to the relevant T cell (
In other applications, another powerful library method has been sequential enrichment/expansion of a displayed library of protein–peptide variants by direct ligand–receptor binding, e.g., using bacterial phage or yeast (also reviewed in
Using fluorescent αβTCRs as probes, we have identified in the library mimotopes for two types of T cells, both originally produced by immunization of mice with the same IAb–peptide combination. One of these T cells was predicted from previous data (
For this study we selected two T cell hybridomas, both prepared from IAb mice immunized with the peptide p3K. This peptide binds well to IAb (
(A) Ribbon structure of the α1 and β1 domains of IAb with a wire-frame representation of the bound p3K peptide (
(B) The figure shows the response of 105 B3K-06 hybridoma cells to various peptides presented by 105 IAb-bearing APCs, LB-15.13.
(C) The figure shows the response of the T cell hybridoma YAe-62 to various MHCII molecules. In each case, 105 hybridoma cells were incubated overnight with MHCII presented in various ways. For IAb-p3K, soluble IAb-p3K was immobilized in the culture well before the addition of the hybridoma cells. In other cases, 106 spleen cells were used directly as APCs without additional peptide antigen. For pEα, the spleen cells came from IAb-pEα/ΔIAβ/ΔIi mice (
The hybridoma YAe-62 was chosen as a representative of broadly allo-reactive T cells present in mice carrying transgenes and gene knockouts that lead to expression of MHCII almost completely occupied by a single peptide (
We previously established methods that used baculovirus-infected insect cells to produce soluble MHC molecules with covalently bound antigenic peptides (
(A and B) Previously described constructions (
(C) The construction was further modified as described in the
(D and E) A degenerate DNA fragment was produced by PCR (D) and cloned into the construct replacing the GFP-encoding sequence (E) as described in the
(A) Sf9 insect cells were infected with baculovirus encoding a membrane-bound form of IAb-p3K. After 3 d, the surface expression of IAb-p3K was detected with an anti-IAb mAb using flow cytometry.
(B) The genes for mouse ICAM (CD54) and B7.1 (CD80) were cloned into an insect cell expression plasmid as described in the
(C) Either Sf9 (open bars) or Sf9-ICAM/B7.1 (closed bars) cells were infected with baculovirus expressing IAb-p3K. After 3 d, the infected insect cells were used as APCs to stimulate IL-2 production from B3K-06 and YAe-62. Uninfected cells were used as negative controls.
Next we prepared fluorescent, soluble αβTCR reagents for use in flow cytometry to detect insect cells displaying the appropriate peptide–MHCII combination. Fluorescent multivalent versions of the soluble αβTCRs of B3K-06 and YAe-62 bound to insect cells displaying the IAb-p3K, but not a control peptide–MHCII combination (
(A) Sf9 insect cells were infected with baculovirus encoding IAb bound either to p3K (filled histogram) or a control peptide (FEAPVAAALHAV) (unfilled histogram). After 3 d, the infected insect cells were incubated with polyvalent, fluorescent soluble αβTCRs from B3K-06 or YAe-62. The binding of each αβTCR was assessed by flow cytometry.
(B) Cells, prepared as in (A), were simultaneously analyzed with fluorescent αβTCRs and a mAb specific for IAb (17–227) that does not interfere with αβTCR–IAb interaction.
(C) The binding of the αβTCRs is shown only for those infected insect cells that bear a high level of surface IAb (dotted region in [B]).
Insect cells displaying IAb-p3K bound the αβTCR reagents very heterogeneously (
Our experiments showed that fluorescent αβTCRs could be used with flow cytometry to identify insect cells infected with a baculovirus encoding a specific peptide–MHC combination. We next tested whether this system could be used to enrich baculoviruses encoding a particular peptide–MHC. Insect cells were infected at an MOI of about 1 with a mixture of baculoviruses. Of these viruses, 1% encoded the IAb-p3K molecule and 99% encoded a control molecule (an αβTCR β chain). The infected cells were stained with fluorescent YAe-62 αβTCR and analyzed by flow cytometry. Although a distinct population of brightly fluorescent cells was not seen, the 1% of the cells with the brightest fluorescence were sorted, as were an equal number of cells that were very dully fluorescent (
(A) Sf9 cells were infected with a mixture of virus, 99% of which encoded a control protein (a TCR β chain linked to the gp64 transmembrane/cytoplasmic tail) and 1% of which encoded IAb-p3K. After 3 d, the infected cells were analyzed as in
(B) The sorted cells were incubated with fresh Sf9 insect cells to allow propagation of the viruses and production of new stocks. The stocks were used to infect new Sf9 cells, and after 3 d the analysis of αβTCR binding was repeated.
The most widely used method for introducing gene constructions into baculovirus involves assembling the construct first in an
We then designed a peptide library based on the structure of p3K bound to IAb (see
A large number of Sf9 insect cells were infected at an MOI of about 1, with baculovirus carrying the IAb–peptide library. After 3–4 d, the cells were analyzed with fluorescent B3K-06- or YAe-62-soluble αβTCR, as described above. Fluorescent cells were sorted and cultured with fresh uninfected Sf9 cells to create new infected cells for analysis and an enriched viral stock. This process was repeated three to four times. In each case, when no clear fluorescent population was apparent, the brightest 1% of the infected cells was sorted. In later rounds the majority of the cells in a clearly distinguishable fluorescent population were sorted.
Sf9 insect cells (1 × 107 to 1.5 × 107) were infected at a MOI of approximately 1 with an aliquot of baculovirus encoding the IAb–peptide library. After 3 d, the infected cells were analyzed for binding the αβTCR of either B3K-06 or YAe-62. Either obviously fluorescent cells or the brightest 1% of the cells were sorted (2 × 104 to 8 × 104 cells) and added to 3 × 106 fresh Sf9 cells to propagate and reexpress the viruses contained in the sorted cells. These infected cells were then reanalyzed and sorted using the fluorescent αβTCRs. This process was repeated until no further enrichment of αβTCR binding was seen. In most cases, the reanalysis was done directly from the cells that were cocultured with the sorted cells. In a few cases, an intermediate viral stock was made and then used to infect additional Sf9 cells. The turn around time per cycle was 4–7 d. The figure shows the reanalysis in a single experiment of the initial viral stocks and all of the various intermediate enriched viral stocks. Sf9 cells were infected at an MOI of less than 1 with the viral stocks and analyzed as in
At the time of the final enrichment, single infected cells binding each of αβTCRs were sorted into individual wells of 96-well culture plates containing fresh Sf9 cells in order to prepare clonal viral stocks. These stocks were used to infect fresh Sf9 cells, which were reanalyzed for binding to the appropriate αβTCR as in
(A) Sf9 cells were infected with stock from four baculovirus clones (B9, B13, B17, and B23) isolated from the virus pool enriched with the αβTCR of B3K-06. After 3 d, an aliquot of cells from each infection was analyzed as in
(B) Same as (A), but using YAe-62 and clones (Y2, Y14, Y28, Y52) derived from the IAb–peptide library using the YAe-62 αβTCR.
aAmino acids homologous to those in p3K are shown in red
bDetermined from mean fluorescence as in
cSorted by frequency and then by level of TCR binding
Given our previous data indicating that the B3K-06 αβTCR interacted with all five of the p3K amino acids varied in this library (
When bound to IAb on ICAM/ B7.1-expressing Sf9 APCs, FEAQRARAARVD was able to stimulate B3K-06 to produce IL-2, but not nearly as well as did p3K. This loss of stimulating activity was caused by one or more of the lysine-to-arginine substitutions and/or the asparagine-to-alanine substitution at p7. Interestingly, the substitution of alanine for asparagine in p3K eliminated the response of B3K-06 to soluble peptide presented by an IAb-bearing mouse APC (see
Consistent with the hypothesis that the αβTCR of YAe-62 would be more peptide promiscuous than that of B3K-06, we found 20 different peptide sequences among the analyzed clones that produced an IAb–peptide combination that bound the YAe-62 αβTCR. It is likely that many more would be identified if more clones were analyzed. Five sequences were found multiple times. Not unexpectedly, these were among those that bound the YAe-62 αβTCR most strongly. There was a 100-fold range in the intensity of αβTCR binding to the different IAb–peptide combinations, ranging from about 4-fold to 400-fold binding above that seen with a negative control peptide. One obvious property of these peptides stands out. There appeared to be a very strong selection for a basic amino acid at position 5. In 16 of 20 of the peptides, a lysine, arginine, or histidine was found at position 5, matching the lysine found in p3K. As a control, we sequenced random clones picked either from the original
There was no strong selection for amino acids homologous to those of p3K at positions p2, p3, p7, or p8. The amino acids at positions p2 and p3 appear nearly random, suggesting little or no essential contact between this part of the peptide–MHC ligand and the receptor, although these positions may contribute to the wide range of apparent αβTCR affinities seen. While not homologous to the asparagine in p3K, leucine was found at p7 in six of 20 (30.0%) of the YAe-62 αβTCR-selected peptides and three of 11 (27.2%) of the IAb-binding peptides that were not bound by the YAe-62 αβTCR, but only two of 17 (11.8%) of the random
The amino acid at position p8 is predicted to be fully surface exposed. In the selected peptides, rather than an amino acid homologous to the lysine of p3K, there may be an overrepresentation of amino acids with small neutral sidechains (threonine, serine, alanine, glycine) at this position. Perhaps this indicates that, in general, larger sidechains can be inhibitory at this position, but again more data would be required to test this idea.
The 12 IAb–peptide combinations that bound the YAe-62 αβTCR most strongly were also the ones that were able to induce IL-2 production from YAe-62. Among these, a number with the very highest apparent affinities stimulated YAe-62 better than did p3K. However, there was not a direct correlation between apparent affinity and the level of IL-2 production; i.e., several peptides that yielded complexes with IAb with about the same apparent affinity for the αβTCR nevertheless stimulated very different levels of IL-2 production from YAe-62. This may be related to the phenomenon of altered peptide ligands (
Overall, our results supported our original prediction that for conventional T cells, such as B3K-06, most of the surface-exposed residues of the peptide would be important in peptide–MHC recognition, while for broadly allo-MHC-reactive T cells, such as YAe-62, peptide recognition would be much more promiscuous.
The peptide degeneracy allowed for a given αβTCR–MHC combination has been a subject of study over many years. While minor changes in the exposed amino acids sidechains of the peptide can often destroy αβTCR recognition, usually at least some variation is tolerated within the predicted footprint of the αβTCR on the peptide–MHC ligand (
We have reported the properties of mice that have been genetically manipulated to express their MHCII molecules virtually completely occupied by a single peptide (
The experiments reported here were designed to test this prediction by comparing the peptide promiscuity of one of these broadly allo-reactive T cells, YAe-62, typical of T cells from these mice, to that of a T cell with the same nominal specificity produced by immunization of conventional mice. The results support the conclusion that the broadly allo-reactive T cell has a much greater peptide promiscuity than does the conventional T cell. This question of T cell promiscuity is an important one in that it addresses the existence of a very large set of TCRs that apparently make it through positive selection, but never see the light of day in normal animals, because they are negatively selected on self-MHC with little input from the MHC-bound peptide. Thus, studying the peripheral fully negatively selected T cell repertoire gives a false impression of the interaction requirements necessary or sufficient for positive selection. These promiscuous T cells may also give us insight into possible evolutionary conserved αβTCR–MHC interactions that have been hard to sort out with conventional T cells.
While perhaps much less frequent than in single peptide–MHC mice, peptide-promiscuous T cells have been described in normal individuals (
In order to study the relationship between peptide sequence and αβTCR recognition, we developed a baculovirus-based display method for rapid identification of peptides that form complexes with MHC that bind a particular αβTCR. Display is one of the most powerful library techniques available. Its underlying principle is that the protein or peptide members of the library are expressed on the surface of organisms that harbor the DNA encoding them. A binding assay that isolates all members of the library with the appropriate properties copurifies the organism and the encoding DNA. The DNA is then amplified and reexpressed and the process repeated as many time as necessary to enrich fully the relevant molecules, whose sequence can be deduced from the copurified DNA. The great advantage of display libraries is that all members of the library that satisfy the screening conditions are enriched simultaneously without the need to identify them one by one.
In order for peptides to be tested for αβTCR binding, they must be complexed with the relevant MHC molecule on a platform suitable for interaction with the T cell and/or its αβTCR. For display libraries, one aspect of this problem has been solved by the ability to express MHC molecules with sequence for a covalently attached antigenic peptide imbedded in the MHC genes (
In these studies, the immunizing peptide (epitope) for the αβTCR was already known. However, this method should be useful as well in identifying mimotopes for αβTCRs whose peptide epitope is not known, provided that suitable peptide anchor residues for MHC binding are known. One limitation of this display method as presented here is the size of the peptide library. The bottlenecks caused by the preparation of the library in an
We have developed this method using IAb as the displayed MHCII molecule carrying the peptide library. However, using the same strategy, we have successfully displayed numerous other MHCII molecules, such as murine IEk and human DR4, DR52c, and DP2 (data not shown). While the leucine zippers that we included in this construct are not strictly required for expression of IAb, they have helped considerably in expression of some of these other MHCII molecules. Moreover, we (
As opposed to methods that use T cell activation as the peptide-screening method, an advantage of display methods that use flow cytometry for screening and enrichment is that the strength of binding of receptor and ligand can be estimated and manipulated. In the results reported here, by limiting the analysis to insect cells bearing a particular level of peptide–MHC, a uniform level of αβTCR binding was seen for an individual peptide sequence, but the strength of binding varied over two orders of magnitude for different peptides, presumably reflecting the relative affinity of the receptor for different IAb–peptide combinations. Thus, depending on whether one was interested in high- or low-affinity ligands for the αβTCR, one could enrich for peptides with these properties directly during the screening of the library. Such an approach has been used with antibody (
It is worth noting that there was not a direct correlation between the strength of αβTCR binding to a particular peptide–MHC combination and the subsequent level of IL-2 secretion seen from the T cell responding to this combination. While in general the best IL-2 secretion was obtained with complexes with the highest apparent affinities, some IAb–peptide combinations with apparent high affinity stimulated IL-2 production poorly. One interesting possibility is that this observation is related to the phenomenon of altered peptide ligands in which amino acid variants of fully immunogenic peptides only partially activate or even anergize the T cell (
In summary, the very properties that have made baculovirus a very successful expression system for complex eukaryotic proteins also make it suitable for library display methods, with potential application not only in T cell epitope/mimotope discovery, characterization, and manipulation, but also in studying a wide variety of other protein–protein interactions.
The peptides pEα (FEAQGALANIAVD), p3K (FEAQKAKANKAVD), and various alanine-substituted variants of p3K were synthesized in the Molecular Resource Center of the National Jewish Medical and Research Center (Denver, Colorado, United States), as were all oligonucleotides used in PCR and DNA sequencing. Automated DNA sequencing was also performed in this facility.
The insect cell lines Sf9 and High Five were obtained from Invitrogen (Carlsbad, California, United States). The IAb-p3K-reactive T cell hybridoma B3K-06 was produced from C57BL/6 mice as previously described (
The T cell hybridoma YAe-62 (
cDNA, prepared from B3K-06 and YAe-62, was used as template in a PCR using oligonucleotides that flanked the Vα and Vβ regions and introduced restriction enzyme sites that allowed cloning of the PCR fragments into a previously described baculovirus expression vector for soluble αβTCRs (
DNA fragments encoding the baculovirus hr5 enhancer element, IE1 gene promoter, and IEI poly(A) addition region were synthesized by PCR using baculovirus DNA as template. The fragments were used to construct an insect cell expression vector (pTIE1) on a pTZ18R (Pharmacia, Uppsala, Sweden) backbone with the hr5 enhancer at the 5′-end, followed by the IE1 promoter, a large multiple cloning site (Esp3I, MunI, SalI, XhoI, BsrGI, HpaI, SpeI, BstXI, BamHI, BspEI, NotI, SacII, XbaI), and the IE1 poly(A) addition region. The complete sequence of the pTIE1 vector has been deposited in GenBank (see Supporting Information). DNA fragments encoding mouse ICAM and B7.1 were cloned between the XhoI and NotI sites of the multiple cloning site. Sf9 cells were transfected with a combination of the plasmids by the standard calcium phosphate method and cells expressing both molecules on their surfaces were cloned without selection at limiting dilution to establish the line Sf9-ICAM/B7.1.
T cell hybridoma cells (105) were added to microtiter wells containing either (1) saturating immobilized peptide–MHC, (2) 10 μg/ml peptide plus 105 LB-15.13 cells, (3) 5 × 104 Sf9-ICAM/B7.1 insect cells infected 3 d previously with baculovirus encoding a displayed peptide–MHC, (4) 106 spleen cells from IAb-pEα single peptide mice, or (5) 106 spleen cells from various knockout or MHC congenic mice. After overnight incubation the culture supernatants were assayed for IL-2 as previously described (
The following mAbs were used in these studies: 17/227, a mouse IgG2a antibody, specific for IAb (
To assemble multivalent fluorescent versions of the soluble αβTCRs, first a biotinylated version of ADO-304 was prepared. In brief, purified ADO-304 at 1–3 mg/ml in 0.1 M NaHCO3 was labeled with Sulfo-NHS-LC-Biotin (Pierce Chemical Company, Rockford, Illinois, United States) at a molar ratio of 2.5:1 (biotin:antibody) for 4 h at room temperature. The reaction was quenched with 0.1 M lysine and the product dialyzed extensively against PBS. The resulting derivative contained about one biotin per molecule of mAb. The biotinylated mAb was complexed in excess with AlexaFlour647–streptavidin (Molecular Probes, Eugene, Oregon, United States). The complex was separated from the free biotin–antibody using Superdex-200 size exclusion chromatography (Pharmacia). In preliminary experiments, the amount of soluble αβTCR required to saturate an aliquot of a large single batch of this reagent was determined. To prepare the multivalent αβTCR, the appropriate amount of soluble αβTCR was mixed with an aliquot of the fluorescent anti-Cα reagent overnight. For staining for flow cytometry, this mix was used without further purification. Each 100 μl sample contained approximately 2 μg of the fluorescent reagent plus 105 Sf9 insect cells. This mixture was incubated at 27°C for 1–2 h. The cells were then washed for analysis. The advantages of this method for preparing fluorescent multimers over using direct enzymatic biotinylation (
Analytical flow cytometry was performed with a FacsCaliber flow cytometer (Becton-Dickinson, Palo Alto, California, United States). For sorting, a MoFlo instrument was used (Dako/Cytomation, Glostrup, Denmark).
For displaying IAb on the surface of baculovirus-infected insect cells, modifications were made as described in
As described in
The GenBank (
This work was supported by United States Public Health Service grants AI-17134, AI-18785, and AI-22295. We thank Amy Marrs and Randy Anselment of the National Jewish Molecular Resource Center for oligonucleotide and peptide syntheses well as automated DNA sequencing. We also thank Shirley Sobus, Josh Loomis, and Bill Townend of the National Jewish Flow Cytometry Facility for help with flow cytometric analysis and sorting.
antigen-presenting cell
third complementarity region
green fluorescent protein
inactivated class II MHC molecule IA β gene
inactivated class II MHC molecule invariant chain gene
monoclonal antibody
major histocompatibility complex
class I MHC molecule
class II MHC molecule
multiplicity of infection
a peptide containing the core sequence FEAQKAKANKAV
a peptide containing the core sequence FEAQGALANIAV
αβ T cell receptor