| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Genzyme Molecular Oncology, Framingham, MA 01701
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CD8+ cytotoxic T lymphocytes recognize short 8- to 10-aa peptides comprising the minimal essential epitopes encoded by tumor Ags. Heteroclitic peptide derivatives of native epitopes achieve reversal of tolerance through the activation of CTL that have the ability to interact productively with the native epitope, indicating that they have escaped thymic deletion and peripheral tolerance, despite their specificity for self Ags (7, 8). The phenomenon of antigenic molecular mimicry whereby a foreign pathogen can break tolerance to a self epitope and give rise to autoimmune aggression represents a naturally occurring example of peripheral or central tolerance reversal by heteroclitic epitope vaccination (9, 10, 11). However, because the molecular definition of the determinants that confer immunogenicity is poorly understood, it is not possible to reliably identify such epitope mimics based on native epitope sequences through rational design (12, 13, 14, 15).
Because epitope mimics that dependably give rise to T cells that effectively cross-react with the corresponding native epitope cannot be rationally designed in a reliable way, we have developed an empirical screening method for the optimization of MHC class I-restricted CTL epitopes. This technique, termed solid-phase epitope recovery (SPHERE),2 allows the identification of individual peptides of desired activities amid tens of millions of peptides. Several groups have reported screening complex combinatorial peptide libraries to identify reactive T cell epitope mimics (16, 17, 18, 19, 20, 21, 22). These approaches use either solid-phase or solution-phase peptide libraries. Groups using solid-phase combinatorial peptide libraries have, for practical reasons, reported screening only a small fraction of their libraries. Although these approaches identify reactive peptides, it is statistically likely that more effective peptide derivatives that were not identified, owing to the fact that the majority of possible library species were not tested. Those methods using complex combinatorial solution-phase libraries seek to identify a motif derived from reactivities present in multiple sublibraries. These motifs require further interrogation by synthesizing and reassaying tens or hundreds of peptides to identify the most desirable ones that conform to the motif. In contrast, we have limited our library complexity by synthesizing a library custom tailored for screening particular MHC-restricted T cell lines or clones. Because it is a prerequisite that strongly immunogenic peptides bind well to their restricting MHC molecule, we have engineered our library with fixed residues at positions previously shown to confer high-affinity binding to the HLA-A2 molecule, thereby reducing the library complexity by nearly 7000-fold compared with a fully degenerate 9-mer library. Consequently, a relatively high percentage of our library peptides can potentially bind well to the HLA-A2 molecule yet still display >47 million different sequences to the TCR of HLA-A2-restricted T cells. The recursive deconvolution strategy we use easily allows screening of an entire library of nearly 50 million peptides yielding both the peptide sequences and their relative activities directly from the screening process. Therefore, our method stands in distinction in that it identifies highly reactive altered peptide ligands with precisely defined structures and with no need for further optimization.
In this study, we apply SPHERE to identify and characterize heteroclitic peptides reactive with a previously reported HLA-A2-restricted T cell line that recognizes an epitope encoded by the human melanoma-associated Ag, gp100 (23).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
TIL1520, TIL620-10, and TIL1235 were generously provided by M. Nishimura (National Institutes of Health, Surgery Branch, Bethesda, MD). These T cell lines were maintained in AIM V medium (Life Technologies, Carlsbad, CA) supplemented with penicillin, streptomycin, 10% human AB serum (Sigma-Aldrich, St. Louis, MO), and 6000 IU/ml human rIL-2 (Proleukin; Chiron, Emeryville, CA). A549 and T2 cells were obtained through American Type Culture Collection (Manassas, VA) and maintained in DMEM/10% FBS and RPMI 1640/10% FBS (JRH Biosciences, Lenexa, TX), respectively.
Peptide synthesis
Peptides were synthesized by standard 9-fluorenylmethyloxycarbonyl (F-moc) chemistry and purified to >95% by reverse-phase HPLC (New England Peptide, Fitchburg, MA). Amino acid analysis and mass spectrometry were performed to accurately quantitate the amount and purity of each peptide. All peptides were resuspended in 100% DMSO (Sigma-Aldrich) at a concentration of 10 mg/ml and used at the indicated dilutions.
Peptide library synthesis
A solid-phase combinatorial peptide library was synthesized by
standard F-moc chemistry according to the split-synthesis method
previously reported (24) using 100-200 Mesh Wang
resin as the solid support (Rapp Polymere, Tübingen, Germany).
The structure of the library was FLXXXXXXV; where X represents any
amino acid except cysteine with the C-terminal valine residue
covalently attached to the beads via an acid cleavable linker. The
library complexity is calculated to be 47,045,881. On average, each
bead contains 
200 pmol of a unique peptide.
Peptide release chemistry
Library preparation: library beads containing the peptides were suspended in dichloromethane/dimethylformamide (1:2) to form a neutral density bead suspension and arrayed in 96-well teflon filter bottom plates (Millipore, Bedford, MA) at a density of 10,000 beads/well. All solvent washes and cleavage reactions were performed in the filter plates using a vacuum manifold apparatus (Millipore). The beads in each well of the filter plate were washed two times with 100 µl dichloromethane, two times with 100 µl diethylether, and air dried. N-terminal F-moc groups were cleaved with 200 µl piperidine/dimethylformamide (1:4) for 20 min and filtered away. The beads were then prepared for release of the peptides into solution and removal of the side-chain protecting groups by washing two times with 200 µl dimethylformamide, two times with 200 µl dichloromethane, two times with 200 µl diethylether, and air dried.
Primary screen peptide release
Partial peptide release from the beads and the
simultaneous removal of the side-chain protecting groups was achieved
by incubation with 200 µl/well trifluoroacetic
acid/acetonitrile/water/anisole (40:55:2.5:2.5) for 60 min at room
temperature resulting in release of 2–5% of the peptide/bead.
Released peptide was captured in a 96-well polystyrene
replica plate and lyophilized under high vacuum. Immediately, the beads
were incubated with 200 µl acetonitrile/0.5 M Tris, pH 7.5 (1:1) for
30 min, washed once with 200 µl acetonitrile/water (1:1), once with
200 µl acetonitrile, once with 200 µl dichloromethane, once with
200 µl diethylether, air dried and stored at -20°C.
Secondary screen peptide release
The 10,000 beads in wells of interest were suspended in dichloromethane/dimethylformamide (1:2) and arrayed evenly over a 96-well filter plate. The release chemistry was performed as before except without the piperidine incubation and the cleavage reaction was for only 30 min.
Tertiary screen peptide release
The 100 beads in wells of interest on the secondary screen
plates were arrayed as before over a new 96-well filter plate yielding
1 bead/well. Peptide release was identical with that of the
secondary screen plates.
Peptide sequencing
Peptide sequencing was performed by Edman degradation directly from the beads at the Harvard Microsequencing Facility (Cambridge, MA).
Library screening
Primary, secondary, and tertiary screens were all performed identically, using a standard 51Cr-release microcytotoxicity assay with the following modifications. A total of 2 µl released peptide was added to V-bottom 96-well plates and T2 cells were added at a density of 1000 cells/well in a total volume of 100 µl/well and incubated at 37°C/5% CO2 for 60 min. A total of 1000 T cells in 100 µl RPMI 1640/10% human AB serum were then added to each well and the plates were returned to the incubator for 4 h. Supernatant was harvested (25 µl) from each well and the amount of released 51Cr quantitated using a Wallac TriLux MicroBeta plate counter (Turku, Finland). Spontaeous 51Cr release was measured in the absence of effector T cells and total 51Cr release was measured by lysing the cells with 0.1% Triton X-100. Percentage of specific lysis was calculated according to the following formula: 100 x (experimental cpm - spontaneous cpm)/(total cpm - spontaneous cpm).
In vitro T cell education studies
Normal donor aphoresis products were obtained from Dana-Farber Cancer Research Institute (Boston, MA). PBMC were isolated by centrifugation over Ficoll (Nycomed, Oslo, Norway). CD8+ T cells were isolated using magnetic beads (Dynal, Oslo, Norway) according to the manufacturer’s instructions. To generate autologous DCs, monocytes were isolated from the PBMC and treated with GM-CSF (Immunex, Seattle, WA) and IL-4 (PeproTech, London, U.K.) for 6 days. One day before establishing the T cell/DC cocultures, the DCs were pulsed with peptides (10 µg/ml) overnight followed by the addition of the previously isolated CD8+ T cells at a T:DC of 10:1. Cultures were restimulated with peptide-pulsed DCs. IL-2 (50 IU/ml) was introduced on day 8 and added every 3–4 days as needed. The bulk cultures were assayed 1 wk after the fifth restimulation. Peptide-pulsed T2 cell targets were prepared as described above and adenovirus-infected target cells were allowed to incubate with the viruses for 48 h at the indicated multiplicity of infection before being used in the CTL assay. All cultures were assayed in quadruplicate using 2E + 3 51Cr-labeled target cells at the indicated E:T for 16 h.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
An overview of the SPHERE methodology is depicted in Fig. 1
. The reactivity of peptides in the
library is measured using a 51Cr-release
microcytotoxicity assay. Parameters such as E:T ratio, tolerance to
DMSO, required peptide concentration, and complexities were optimized
by spiking in known solid-phase peptides under mock screening
conditions (data not shown). The library deconvolution strategy
involves three iterations starting with pools of 10,000 peptides for
the primary screen, pools of 100 peptides for the secondary screen, and
single peptides for the tertiary screen. Because high-affinity MHC
binding is required for good immunogenicity, the nonamer peptide
library was engineered to favor binding to a particular MHC molecule.
For HLA-A2, residues 1, 2, and 9, predicted by functional studies and
x-ray crystallography to contact the MHC, were fixed with amino acids
F, L, and V, respectively, which have been reported to confer
high-affinity MHC binding (25, 26, 27). This serves to limit
the library complexity while maintaining broad diversity among the
residues predicted to contact the TCR. Therefore, the structure of our
dedicated HLA-A2 solid-phase library is
F-L-X6-V-linker-bead, where X is any one of 19 aa
(cysteine excluded) and the linker that tethers the peptides to the
beads is chemically cleavable. Having a fixed amino acid (valine)
coupled to the linker allows uniform kinetics for partial cleavage of
all peptides, and by keeping the cleavage reagent limiting, any desired
aliquot of peptide can be uniformly released from the beads in a manner
that is largely independent of the amount of solid-phase peptide in the
reaction (data not shown). This library has a complexity of 47,045,881.
Because our HLA-A2-dedicated library is biased toward hydrophobic
residues, all peptide and peptide mixture stock solutions were carried
in 100% DMSO so as not to bias against those peptides that are poorly
soluble in aqueous solution.
|
Polystyrene beads representing the entire library were arrayed
into 47 96-well teflon filter-bottom plates at a density of 10,000 per
well. A small portion (5%/bead, 10 pmol) of the peptide from
each bead was released by acid hydrolysis and captured in standard
96-well polystyrene replica plates using a Millipore Multiscreen
apparatus (Millipore) and lyophilized. The released peptides
were resuspended with 500 µl/well DMSO and 2 µl of each 10,000
peptide mixture was pulsed onto 51Cr-labeled T2
cells and subsequently exposed to TIL1520 at an E:T of 1:1. In the
interest of minimizing T cell consumption, we used 1000 target cells
and 1000 effector cells per well. The effector cells used in this study
were TIL1520, having specificity for the 209–217 epitope encoded in
the human gp100 melanoma Ag. Under these conditions, the entire primary
screen consumes only 5 million T cells. Lytic activity was measured
by 51Cr-release as described. Fig. 2 shows the frequency distribution of
observed responses compiled from the entire primary screen. The vast
majority of the peptide pools show little or no lysis induction;
however, there are some that stimulate potent lytic activity from this
T cell line (Fig. 3, a and
d). The highest scoring wells were further analyzed in a
secondary screen. Note that this experiment has been repeated several
times with similar frequency distributions and with consistent
identification of the same high-scoring peptide pools.
|
|
The library beads corresponding to seven of the highest scoring
peptide pools were rearrayed evenly over new 96-well filter plates and
an aliquot of peptide was released into replica plates. The released
peptides, now 100-peptide mixtures, were reassayed as in the primary
screen. In every case, a single pool demonstrated clear lysis above
background from a single peptide pool, indicating that the peptide
responsible for the observed activity in the primary screen was present
in the lower complexity mixtures. Typical secondary screen results are
shown in Fig. 3
, b and e. The peptide pools
showing activity were further analyzed in a tertiary screen.
Tertiary screen
As before, the beads corresponding to each of the positive
100-peptide pools were each rearrayed evenly over new 96-well filter
plates yielding 1 bead/well, and replica plates of solution-phase
peptides prepared. These released peptides were assayed as before and
in each case a single reactive well was observed. Typical tertiary
screen data are shown in Fig. 3, c and f.
Upon completion of the tertiary screen, the reactive peptides were
recovered as solid-phase pure populations and were sequenced directly
from the beads by Edman degradation. The sequences of the seven highest
scoring peptides are shown in Table I.
Two of these peptides, SP-1 and SP-2, share some sequence homology with
the native epitope. Analysis of the homologous sequences reveal a motif
that highlights TCR contact points or structural elements required to
share recognition with the native epitope. Most notably, the amino
acids D, Q, and F at positions 3, 4, and 7 are dominant and appear in
several of the most reactive peptides. Substitutions at other positions
most likely contribute to overall MHC and TCR affinities, but differ
sufficiently from the native sequence that they could not have been
predicted by rational design methods. Consistent with previous reports,
we observed that the majority of the sequences had minimal or no
homology to the wild-type epitope (28, 29). There appears
to be a strong bias toward hydrophobic residues among the non-MHC
anchor residues of these peptides. This may be related to the high
density of hydrophobic side-chain residues that have been shown by
x-ray crystallographic studies to line the peptide-binding groove of
the HLA-A2 molecule.
|
Some of the peptides are so divergent from the native epitope that
we wished to demonstrate whether they would react specifically with
TIL1520. To address this, we resynthesized and tested the reactivity of
some of these peptides in a 51Cr-release assay
with the T cell line used in the screen (TIL1520) or an irrelevant T
cell line (TIL1235) that recognizes a different epitope in another Ag.
Fig. 4 shows the result of this assay and
strongly supports the idea that all of these peptides were correctly
identified legitimate "hits" in our screen and that they do
indeed react specifically with
gp100209–217-reactive TIL. Next, we sought to
determine whether the SPHERE peptides would react with an independently
derived gp100209–217-specific TIL population
(TIL620-10). To this end, a subset of the peptides shown in Table I,
chosen for their sequence diversity, were tested for reactivity with
either TIL1520 or TIL620-10 in a 51Cr-release
assay. These results are shown in Fig. 5, and indicate that the peptides react equally well with TIL620-10,
implying that even the distantly related epitope mimics are
functionally similar to the native epitope in this assay.
|
|
Having demonstrated that the identified SPHERE peptides are
antigenic (i.e., they are recognized by preformed
gp100209–217-specific CTL), we next sought
to determine whether they were also immunogenic (i.e., could prime a de
novo immune response). To this end, we compared their relative
abilities to educate normal donor HLA-A2+ T cells
in vitro. These in vitro T cell education studies were designed to test
the ability of the SPHERE peptides to expand and sensitize T cells to
lyse targets presenting the native epitope or themselves. Because most
of the peptides tend to fall out of solution forming a fine precipitate
when the DMSO stock solutions are added to the aqueous media, we chose
to use autologous immature dendritic cells (DCs) as APC, which
are better equipped to take up particulate matter than are PBMC
(30). We found that the native
gp100209–217 peptide is poorly immunogenic in
our in vitro assay using normal donor T cells. Detection of
peptide-specific reactivity routinely required five rounds of
restimulation, and was best demonstrated using a 16- to 22-h incubation
period with target cells in the CTL assay. Hence, we adopted these
conditions for comparison of the SPHERE peptides with the wild-type
epitope. We found that the SPHERE peptides were capable of eliciting
responses even in those individuals that responded poorly or not at all
to the wild-type peptide (Fig. 6). When
these studies were extended to include a total of 20 normal donors, it
was noted that none of the peptides were immunogenic in all
individuals; however, there were differential responses suggesting that
the T cells each peptide preferentially stimulated may represent
different T cell populations, perhaps with different donor-dependent
precursor frequencies. This idea was further supported by the marked
increase in population coverage observed when the peptides were used in
combination with one another. As shown in Table II, a mixture of three different SPHERE
peptides (SP-1, SP-2, and SP-3) achieved nearly a 90% response rate in
vitro. These data strongly support the utility of a SPHERE peptide
mixture to overcome tolerance to a nonmutated self Ag to which the
majority of individuals would be tolerant.
|
|
Given the divergence of the SPHERE peptide sequences from the
native epitope, we sought to determine the specificity of the T cells
educated with these peptides. To this end, we tested in vitro-educated
T cells for reactivity to a human lung tumor cell line (A549), which
does not express HLA-A2 or gp100 (Fig. 7). When the cell line was converted to
express either HLA-A2 or gp100 by recombinant adenovirus infection, the
SPHERE peptide-educated normal donor T cells still did not lyse them.
However, when the cell line was doubly infected to express both HLA-A2
and gp100, the SPHERE peptide-educated T cells lysed the modified A549
cells. These data demonstrate that even peptides with extreme sequence
diversity can produce functionally indistinguishable HLA-restricted,
Ag-specific responses and that naturally processed and presented
peptide from the native Ag can render tumor cells susceptible to lysis
by these effectors. Note that the levels of HLA-A2 and gp100 proteins
expressed by infected A549 cells is lower than the endogenously
expressed proteins in the
HLA-A2+/gp100+ melanoma
cell line, MEL624, as measured by FACS analysis using specific Abs
(data not shown).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
47 million peptides for those that react with
an HLA-A2-restricted CD8+ T cell line with known
specificity for the 209–217 peptide of the human melanoma Ag, gp100.
We have demonstrated the capability of SPHERE to quickly and
efficiently identify reactive epitope mimics that reliably give rise to
T cells that cross-react well with their native epitope counterpart.
This screening methodology offers several advantages over other epitope
identification strategies: 1) limiting the library complexity by fixing
residues that confer tight binding to the HLA-A2 molecule enables
complete library screening; 2) quantitative ranking of all peptides
during the screening process to readily identify peptides with the most
desirable activity; 3) modest consumption of T cells (1E5 T cells per
1E6 peptides screened); 4) completion of the entire screen in just a
few days; and 5) reusability of a single library (250 times).
Because our primary screen is performed at relatively low peptide
complexity, it is statistically unlikely that more than one reactive
peptide will be present in the same peptide pool, and this was
confirmed upon deconvolution. Therefore, we obtain a quantitative
ranking of all reactive peptides present in the library directly from
the primary screen. Another reason that we maintain low peptide
complexity is to remain statistically confident that a strongly
inhibitory peptide will not mask the activity of a potent agonist. The HLA-A2 library is biased toward hydrophobicity, because every peptide contains at least one F, L, and V. This bias inevitably leads to differences in solubility; and therefore, bioavailability in the assay. To the extent possible, we attempt to minimize these effects by using 100% DMSO to resuspend our solution-phase peptide pools and perform all peptide pulsing in at least 2% DMSO, a concentration tolerated quite well by the cells in our assay. Given the hydrophobic peptide-binding pocket of the HLA-A2 molecule, it was not entirely unexpected that the most reactive peptides identified in the screen were extremely hydrophobic, generally with a preference for hydrophobic residues throughout. We note that although this hydrophobicity complicates synthesis and purification of these peptides, it may ultimately be beneficial for in vivo vaccination. Several groups have reported that covalent modification of peptides with hydrophobic lipid moieties or noncovalent association of peptides with lipid vesicles can markedly improve their immunogenicity (31, 32).
Analyses of the sequences of the reactive epitope mimics identified reveal that the limits of antigenic mimicry are far reaching. Although some of the reactive peptides bear a strong resemblance to the native epitope, none could have been predicted by rational design methods. Other peptides, being unrelated to the native epitope although structurally related to one another, are distinctly nonobvious from a rational design perspective and challenge our current understanding of epitope structure and function. Furthermore, even the structurally divergent epitope mimics identified are reactive with an independently derived T cell clone specific for the same native epitope. We interpret this finding to mean that either these peptides share a common topology with the native epitope when bound to the MHC, or that they bind differentially to the MHC and/or the TCR.
The clinical utility of the epitope mimics identified by SPHERE is
underscored by their ability to induce native epitope-specific CTL
responses from normal donor T cells in vitro. This feature
distinguishes SPHERE peptides from rationally designed anchor-modified
epitope mimics that have been previously reported (12, 13). Structural diversity notwithstanding, it seems likely that
the different SPHERE peptides preferentially engage different, yet
functionally convergent, T cell populations in each normal donor. This
may be the basis by which they can elicit immunologic responses from
normal donors in vitro that seem unable to mount a response to the
wild-type epitope. Analyzing TCR V
usage
within the in vitro-educated bulk cultures of normal donor T cells by
PCR analysis has further supported this hypothesis (data not shown). If
the peptides are acting through different mechanisms, and considering
that not every individual responded to every SPHERE peptide, we
reasoned that there might be a benefit to using them in combination. We
found that we could more reliably induce a wild-type peptide-specific T
cell response from normal donor T cells when a combination of peptides
was pulsed onto the same DC population and used to stimulate them.
These responses were generally similar in magnitude to using any one
peptide, even though the total peptide concentration of the mixture was
the same as the individual peptide concentrations. Recently, Wang et
al. (33) reported an interesting phenomenon that may
account, at least in part, for these observations. Their data showed
that DCs pulsed simultaneously with different peptide epitopes could
prime an immune response in mice more efficiently than either peptide
alone. In fact, they could do so even in the absence of
CD4+ helper activity, which was required when the
peptides were pulsed individually. It was concluded that different MHC
class I-restricted peptides, when simultaneously present on the same
DC, can provide helper function for each other in the absence of
activated CD4+ T cells.
In conclusion, SPHERE can be used to rapidly identify or optimize MHC class I-restricted epitopes. It is a convenient way to reliably identify peptides that represent different "flavors" of the same native epitope which, as we have demonstrated, may allow for the formulation of vaccines capable of circumventing self tolerance or overcoming immunologic ignorance. This technology will find applications in tumor and infectious disease immunotherapy. Additionally, we note that the SPHERE screen can be modified to identify specific epitope antagonists which may find utility in the development of therapeutics for autoimmune disease or allotransplantation.
|
Footnotes |
|---|
2 Abbreviations used in this paper: SPHERE, solid-phase epitope recovery; DC, dendritic cell; F-moc, 9-fluorenylmethyloxycarbonyl.
Received for publication April 2, 2002. Accepted for publication June 26, 2002.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
