| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
*
Institut National de la Santé et de la Recherche Médicale, Unité 25, Paris, France;
The Walter and Eliza Hall Institute, Royal Melbourne Hospital, Victoria, Australia;
Istituto di Medicina Interna, Università di Perugia, Perugia, Italy; and
§
Roche Milano Ricerche, Milan, Italy
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The dominant pathway of Ag processing for presentation by MHC class I molecules includes at least three steps: degradation of protein to peptide fragments in the cytosol by the proteasome, transport of the peptides into the endoplasmic reticulum (ER)3 by TAP, and peptide assembly with newly synthesized and TAP-associated MHC class I molecules in the ER (2). Little is known about the quantity and quality of antigenic peptides generated and handled by these processes. As a consequence, it is unclear 1) whether and to what extent these steps are selective; 2) whether an abundant or limiting supply of peptides is provided to MHC class I molecules; and 3) whether processing serves all MHC class I molecules with equal efficiency. Indirect evidence suggests that peptide supply by processing may be limiting, at least for some MHC class I molecules. A part of the pool of newly synthesized HLA class I molecules appears to be degraded (3), possibly because of the lack of cognate peptides, and the kinetics of assembly with peptides are known to vary between HLA class I alleles (4). Moreover, free MHC class I heavy chains can be detected on the surface of nonmutant cell lines, and surface expression of rodent and human MHC class I molecules can be increased by cell incubation with cognate peptide ligands, suggesting that peptide supply may be limiting under physiologic circumstances (5, 6, 7).
Selective peptide transport by TAP represents one of the principal candidate mechanisms that may limit or modulate the supply of peptides to HLA class I molecules. It is clear that selectivity of peptide transport by TAP can have a dramatic effect on peptide presentation by MHC class I molecules: extensive polymorphism in the TAP2 subunit of the rat transporter leads to transport of distinct peptide sets into the ER, which is reflected in the nature of peptides presented by MHC class I molecules (8, 9, 10, 11, 12). However, evaluation of a potential functional role of peptide selection by human TAP has been more difficult, since human TAP is functionally monomorphic (13, 14) so that analysis of the effects of TAP by comparing peptides presented in HLA-identical cells carrying different TAP alleles is not feasible.
Peptide transport by TAP is a sequential process in which interaction of peptide with a putative binding site appears to be followed by a translocation step that requires cleavage of ATP (15, 16, 17, 18, 19). According to all available evidence, peptide selection by TAP occurs exclusively in the initial binding step. This conclusion was first suggested by comparisons of peptide binding to TAP in human or insect cells and peptide transport in human cells (20, 21, 22). More recently, we have undertaken an extensive analysis of peptide binding to rat transporters and found that the well-characterized dramatic effect of TAP polymorphism on transport selectivity and peptide presentation by rat MHC class I molecules is based on strong differences in peptide binding affinity of polymorphic rat TAP complexes (43). These studies also demonstrated that cofactors absent from insect cells cannot be important for peptide binding selectivity of TAP, as already suggested by an earlier comparison of peptide transport in human cells with transport in insect cells expressing human TAP complexes (20).
Selectivity of TAP can be studied in assays that measure either peptide binding to TAP or TAP-dependent peptide accumulation in the ER (17, 18). Studies employing the latter assay have suggested that human TAP is much more permissive than mouse TAP or the restrictive rat TAP complex TAP1/TAP2u (10, 23), a conclusion confirmed by our recent comparison of peptide binding to rat and human TAP complexes (43). While these studies suggested that human TAP may not select peptides at all, a subsequent analysis demonstrated that when substitutions were performed in another sequence context, human TAP transported preferentially peptides with hydrophobic aromatic or positively charged C-terminal residues (14). These studies compared ER accumulation of substituted peptides after prolonged cell incubation at 37°C, a technique whose sensitivity may in some cases be limited due to the high peptide throughput in TAP-mediated transport (17, 24) and active peptide depletion from the ER by an unknown mechanism (25, 26). However, these and other reports (27) suggested that human TAP may transport at least a subset of peptides with low efficiency.
By overexpressing TAP in insect cell microsomes, we previously developed an assay to monitor the initial step of peptide binding to human TAP (18). When we analyzed the influence of peptide sequence on TAP binding affinity, we found that peptide binding affinity for the human TAP complex varies strongly according to the nature of the residues in the three amino-terminal positions and at the carboxyl terminus and can be summarized in a peptide binding motif (22). More recently, these observations have been confirmed in a study measuring competition for binding to human TAP by combinatorial peptide libraries (28). We also observed that peptide selection by human TAP focuses on positions that generally represent anchors for binding to HLA class I molecules and noted that the TAP preferences are in conflict with the preferences of some HLA class I molecules (22). As a result of these conflicts, it was conceivable that ligands for some HLA class I molecules were preferred by the human transporter, which might lead to more efficient peptide supply to these alleles.
In this study, we examined TAP affinities of ligands for a large number of HLA class I molecules. To this end, we developed artificial neural networks (ANNs) that allowed us analyze the TAP affinities of 408 HLA class I ligands. Using ANNs, we determined whether 1) peptides naturally presented by HLA class I molecules show evidence of selection by the transporter; and 2) the human transporter is likely to supply peptide ligands to all HLA class I molecules with equal efficiency.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Most peptides used in this study were synthesized on an Advanced
ChemTech396 multiple peptide synthesizer (Advanced ChemTech,
Louisville, KY). Purity was assessed by HPLC analysis and was generally
>80%. Besides the library of 163 polyAla peptides, 235 peptides were
synthesized and tested in TAP binding assays. These included 17
peptides of various length (experiment shown in Fig. 2
), 100 9-mer
peptides with random sequences for testing predictions, and 218 other
9-mer peptides used for training ANNs. Among these, 35 had been
designed for a chemometric analysis of TAP binding (D. Riganelli,
unpublished observations) and were composed exclusively of Ala, Asp,
Ile, Lys, Phe, Ser, Tyr, and Val; 13 other peptides had previously been
used for a chemometric analysis of peptide binding to HLA-B27 (29). One
hundred and seventy peptides used for ANN training had sequences
derived from natural proteins; 156 of these peptides are known to bind
to various HLA class I molecules, and 81 have also been identified as
natural CTL epitopes or by elution from HLA class I molecules.
Sixty-eight of the HLA class I binding peptides have been shown to bind
to HLA-A2, and 28 have been shown to bind to HLA-B27; the remaining 60
peptides bind to the following HLA class I molecules: A29, six
peptides; B40 and B58, five peptides each; A11, B7, B44, and Cw7, four
peptides; A3, A24, A31, A33, B61, and B62, three peptides; A1, A68, and
B38, two peptides; and B37, B52, B53, and B57, one peptide.
|
Peptide binding to human TAP complexes overexpressed in Sf9 insect cell microsomes was measured as previously described (18), using 300 nM iodinated reporter peptide R-9-L (Arg-Arg-Tyr-Asn-Ala-Ser-Thr-Glu-Leu) in the presence of increasing concentrations of unlabeled competitor peptide (0.01- to 1000-fold molar excess). In each experiment, duplicate samples without competitor peptide and a series of dilutions of unlabeled reference competitor peptide R-9-L or, for the polyAla library, unsubstituted peptide Ala-Ala-Ala-Ser-Ala-Ala-Ala-Ala-Tyr were included. For training of ANNs, all IC50 values (concentrations required for 50% inhibition of specific binding) were normalized with respect to the IC50 of reporter peptide R-9-L (typically 210–270 nM). Unless otherwise specified, the IC50 values reported here are normalized values.
Ala matrix-based prediction of TAP binding affinity
Normalized IC50 values for substituted polyAla
peptides were transformed in natural logarithms to assign a numerical
value to each natural amino acid (except cysteine) in each position of
9-mer peptides. To predict the affinities of test peptides, logarithmic
values for the nine amino acids of test peptides were added. For better
graphical representation in Figure 2
, predicted values were adjusted to
a scale from 0 to 10.
ANN training and predictions
ANN simulations were performed with the PlaNet software, version 5.6 (30). Binding affinity predictions were based on the average of four simulations with a nonlinear (architecture 180-2-1) ANN configuration. ANNs were trained using backpropagation training algorithms (31) in up to 300 cycles on a set of 381 9-mer peptides of known TAP binding affinity (see under Peptides for composition of the training set). Each amino acid was encoded as a 20-bit string with a unique position set at 1; all other positions were set at 0. Peptides were represented as binary strings of 180 bits. Each prediction of binding affinity of a peptide represents the average of four predictions obtained with four independent ANNs trained on the same set of peptides. Binding affinity was scored on a scale from 0 to 10. A binding score of 0 corresponded to a normalized IC50 of >1000, whereas a score of 10 corresponded to an IC50 of <0.03. Initially, 1000 random 9-mer peptides were generated, and their TAP binding affinities were predicted by a preliminary ANN trained with a simple binder/nonbinder classification of a subset of 350 peptides. Due to its low predictive performance, this ANN was later discarded. However, selection of 100 random peptides for evaluation of predictive performances was based on the preliminary ANN; for each of the scores (0 to 9) 10 peptides were selected.
The coefficient of correlation between ANN predictions and measured values was 0.732 (p = 5 x 10-18). The performance of ANN and Ala matrix predictions was also assessed by relative operating characteristics (ROC) analysis, a useful method for true performance of prediction systems that eliminates biases due to the distribution of data or the setting of arbitrary classification thresholds (32). ROC areas on the order of 0.5 represent predictions that are equivalent to chance alone, areas between 0.7 and 0.9 represent useful predictions, and areas greater than 0.9 represent high accuracy predictions. ROC area values of ANN (values for Ala matrix are in parentheses) predictions were 0.78 (0.79) for a score of 1, 0.81 (0.79) for a score of 2, 0.82 (0.84) for a score of 3, 0.86 (0.85) for a score of 4, 0.91 (0.85) for a score of 5, 0.95 (0.89) for a score of 6, 0.96 (0.90) for a score of 7, 0.97 (0.99) for a score of 8, and 1 (0.98) for a score of 9.
For simulated experiments on peptides binding to HLA class I molecules
shown in Table II, all HLA class I molecules were considered for which
at least 20 9-mer ligands (binders) meeting the following criteria were
found in the MHCPEP database: 1) no polyAla or polyGly peptide, 2) no
single-substituted analogue of another peptide in database, 3) no
eluted peptide or CTL epitope, 3) intermediate or high binding affinity
for relevant HLA class I molecule reported, and 4) if no details on
binding affinity reported, accept only if canonical anchor residues are
present. For the experiment shown in Table I, the three HLA class I molecules were
selected for which, in addition to at least 20 binders, at least 20
ligands identified by peptide elution and sequencing were contained in
MHCPEP. To exclude identified or putative signal sequence-derived
peptides from the groups of eluted peptides and control binders, source
proteins (where known) were retrieved for all 408 9-mer sequences from
a database (nonredundant GenBank, PIR, Swissprot), and all hydrophobic
peptides derived from the N-terminus (residues 1–25) were removed.
|
|
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To develop a strategy for prediction of TAP binding of 9-mer
peptides that would allow us to analyze an unlimited set of peptides,
we first undertook a detailed analysis of the effects of sequence
variations in 9-mer peptides on peptide binding to the human TAP
complex. Initially, we determined the effects on binding to TAP of
substitutions by 19 natural amino acids (except cysteine) in the fixed
sequence environment of a polyAla peptide. This confirmed and extended
our previous observation (22) that some, but not all, positions in
short peptides have major effects on binding affinity (Fig. 1). According to the magnitude of the
effects, positions can be ranked as follows: P9 >
P2/P3 > P1 > P7 > P4/P5/P6/P8. In the four most
important positions, changes in IC50 of >3 logs (P9)
or of >2 logs (P1, P2, and P3) can result from single substitutions.
Among these four positions, all but P1 showed similar preferences for
large hydrophobic and basic residues, while Gly, acidic residues
(except Glu in P2), and Pro (except in P3) are detrimental for binding.
Note that there was a marked preference for hydrophobic aromatic
residues in P3 and P9, and Ser at the C-terminus was highly
unfavorable. Only at the NH2-terminus were large
hydrophobic residues almost as unfavorable as acidic residues and Pro,
while basic and smaller residues (especially Ala) were preferred.
|
TAP can bind and transport peptides of 16 or more residues (15, 18, 21, 25, 26). We wondered whether the binding efficiency of longer peptides also depends on terminal positions independent of peptide length ("end rule"); alternatively, the four positions with the greatest importance for binding may need to be located at a fixed interval of approximately six residues (interval between P3 and P9 in a 9-mer) even in longer peptides ("interval rule"). To identify the mechanism of binding of longer peptides, we inserted a Pro-Asp or Pro-Asp-Ala-Pro-Asp sequence at various positions of a 9-mer peptide with high TAP binding affinity, thus creating a series of 11- and 14-mer peptides. The insertion comprised residues with marked unfavorable effects on binding when present in any of the critical positions in 9-mer.
The results shown in Figure 2 are
compatible with the end rule, but not with the interval rule.
Replacement of P2 by the insertions had the most dramatic deleterious
effect on binding affinity, followed by the insertions in P9 and P1.
This order was expected according to the effects of single
substitutions (Fig. 1). Note that the insertions in P1 and P2 each
resulted in unfavorable substitutions in two critical positions. Thus,
the binding affinity of peptides with a length of 9 to 14 residues
appears to depend mainly on interactions of their termini with the TAP
complex. However, the magnitude of the effects of the insertions was
smaller than expected according to the effects of substitutions in
9-mer, suggesting that binding of longer peptides to TAP may be less
susceptible to destabilizing effects of unfavorable residues.
Prediction of binding of 9-mer peptides to TAP
Based on the analysis of peptide sequence parameters determining TAP binding affinity, we set out to develop and test two strategies for prediction of TAP binding of 9-mer peptides. The first strategy used a matrix of natural logarithmic values representing normalized IC50 values of substituted polyAla peptides (Ala matrix) and was based on the assumption that the effects of substitutions in the polyAla peptide are independent of sequence context. The second approach was designed to recognize and take into account sequence environment-restricted effects and involved affinity prediction using ANNs. An ANN is a computer simulation of a system of interconnected processing units. An ANN can be trained to extract and remember a pattern present within a set of data; it can subsequently recognize that pattern when presented with new data. ANN performance can incrementally be upgraded by training of the ANN with additional datasets (33, 34, 35).
While the first strategy was based exclusively on the information provided by the set of 163 polyAla peptides, ANNs were trained with binding data of 218 9-mer peptides in addition to the set of polyAla peptides; most of these have been described to bind to various HLA class I molecules (see Materials and Methods for details). IC50 data used for ANN training as well as ANN-predicted TAP binding affinities were expressed in a scale from 0 to 10, representing a 5-log range of normalized IC50 values from >1000 (score 0) to <0.03 (score 10), with a score increment of 1 corresponding to a threefold smaller IC50. Based on preliminary Scatchard plot data of selected peptides, we estimate that this 5-log scale represents peptide binding Kd values ranging at least from 10-8 to 10-5 M (S. Daniel and P. M. van Endert, unpublished observations).
To evaluate the correctness of Ala matrix and ANN predictions of TAP
binding, we determined TAP affinities of a set of 100 9-mer peptides
with random sequences. Both strategies were able to predict
experimental values of the majority of peptides (Fig. 3). To compare the performances of matrix
and ANN-based predictions, matrix-predicted affinity scores were
adjusted to the scale of 0 to 10 used for ANN training and predictions,
and the precision of predictions for individual scores was evaluated
separately for the two approaches (see Materials and Methods
for details). For both strategies, predictive performance was
satisfactory for low affinity peptides (scores 0–4) and was high for
intermediate to high affinity binders (scores 5–10). Although the
number of peptides scoring at the individual levels did not allow
statistical comparison of the two strategies for prediction, ANNs
appeared to be superior for intermediate to high affinity binders.
Twenty-seven percent of the ANN-predicted scores were exact (factor of
1–3 between observed and predicted IC50 values), 65% of
the observed scores differed by not more than 1 from the predicted
scores (factor of 3–10), and 87% of the observed scores differed by
not more than 2 (factor of 10–30). The corresponding numbers for Ala
matrix predictions are 20, 54, and 76%. Relatively precise prediction
of affinities of >50% of peptides based on the matrix suggests that
context-specific effects are small or absent in many peptides. However,
the superior performance of ANNs suggests that consideration of these
effects by predictive tools is possible and necessary for universally
applicable prediction of TAP affinity. For further analyses, we
therefore used ANN-based predictions.
|
Having validated ANN performance, we used ANNs to search for evidence of selection of epitopes by TAP in vivo. We analyzed TAP binding affinities of peptides sequenced after elution from HLA class I molecules, reasoning that these naturally processed peptides are relatively abundant ligands of HLA class I molecules and that their TAP binding affinities should therefore be higher than those of control peptides with similar binding affinities for the same HLA class I molecules; because of their generally TAP-independent access to the ER, signal peptide-derived ligands were excluded from the groups of eluted and control peptides. In view of previously obtained evidence that peptides binding to some HLA class I molecules may possess higher mean TAP binding affinities than peptides binding to other HLA class I molecules (22), this analysis was conducted separately for peptides known to bind to individual HLA class I molecules. Sequences of eluted peptides as well as control peptides were chosen in the MHCPEP database (36). As controls, we chose MHCPEP peptides that have been reported to bind to selected HLA class I molecules with intermediate or high affinity, as defined by direct peptide/HLA binding assay; most control peptides had been identified by screening of protein sequences for peptides binding to HLA class I molecules. Although MHCPEP peptides known to be CTL epitopes were not used as control peptides, it cannot be ruled out that some or even many of the selected control peptides can be processed and presented naturally. Conclusive characterization of peptides as not being naturally presented would require high sensitivity elution and sequencing of peptides from HLA class I molecules of cells expressing the protein from which the peptides are derived, a task that is practically unfeasible to accomplish for significant numbers of peptides. Nevertheless, we considered it highly probable that peptides eluted from HLA class I molecules comprise a significantly higher percentage of sequences that are processed and presented with high efficiency than control peptides selected as described.
For three HLA class I molecules, the MHCPEP database contained at least
20 peptides that could be assigned to each of the naturally
processed and control groups (Table I). In all three cases, eluted
peptides possessed significantly higher predicted TAP binding
affinities than the control peptides, as measured in nonparametric
Mann-Whitney tests. The mean predicted scores of peptides eluted from
three molecules differed by 0.8 to 1.9 from control scores,
corresponding to 2.5- to 10-fold lower IC50 values.
Interestingly, TAP binding of peptides eluted from HLA-A2, which binds
peptides with the lowest mean TAP affinity, differed the least from
binding of control peptides. In summary, natural peptide presentation
by HLA class I molecules is associated with higher mean TAP binding
affinity.
TAP binding of peptides binding to HLA class I molecules
Next, we applied ANNs to analyze the relationship between peptide
selectivity of HLA class I molecules and human TAP complexes. First, we
simulated a binding experiment on 5000 random sequence 9-mer peptides
and 210 peptides eluted from various HLA class I molecules (Fig. 4A). Random peptide
sequences were generated using a computer program, with the frequencies
of amino acids representing the codon frequencies in the GenBank
database. Sequences of eluted peptides were retrieved from the MHCPEP
database. High affinity binders (scores 7–10) are very rare in the
random group (1.2%) and were frequent among eluted peptides (17.6%),
while the opposite was true for peptides with very low scores (27.7 vs
6.7% with scores of 0 or 1). The mean IC50 of eluted
peptides was >10-fold lower than that of random peptides (mean score,
4.6 vs 2.5, corresponding to normalized IC50 values of 4.3
and 50). Experimentally determined affinities of 80 natural ligands
also were significantly higher than predicted affinities of random
peptides (Fig. 4B), supporting the conclusion that
TAP affinities of peptides binding to HLA class I molecules are
generally superior to those of random peptides.
|
, C–E, and Table II); overall variation between TAP
affinities of ligands for individual HLA class I molecules was highly
significant (approximate p < 0.0001, by Kruskal-Wallis
2 test). HLA-A2 and B35 bind ligands with mean TAP
binding scores that are only slightly higher than those of random
peptides (Fig. 4C). HLA class I molecules belonging
to the A3 supertype (37) or binding to HLA-A68 bind ligands with mean
TAP binding scores of approximately 4, corresponding to
IC50 values that are 6-fold lower than that of an
average random sequence 9-mer. (Fig. 4, C–E).
Some ligands for these class I molecules possess high transporter
affinities that are rarely found among random sequences. Finally,
HLA-A24 and HLA-B27 binding peptides display high TAP binding
affinities, with a mean score corresponding to an average
IC50 that is 12-fold lower than that of A2 binders (Fig. 4E).
In one-by-one statistical comparisons of the differences in TAP binding
affinities between ligands for individual HLA molecules, most
differences reached statistical significance (Table II). With two
exceptions (A24 vs A31 and A33), mean score differences of 0.8 or more
were associated with statistical significance. TAP binding affinities
of random sequence peptides were significantly different from TAP
affinities of ligands for all HLA class I molecules but HLA-A2.
Differences between TAP binding affinities of HLA-B27 ligands and
ligands for all other HLA class I molecules except HLA-A24 reached
statistical significance. Conversely, HLA-A2 ligands had lower TAP
affinity than ligands for all other molecules except B35. For two HLA
class I molecules, HLA-A2 and B27, the availability of reagents allowed
us to confirm simulated experiments by actual measurements on naturally
presented peptides: a mean score of 2.4 for 10 A2 ligands contrasts
with a mean score of 7.3 for 18 B27-presented peptides (Fig. 4B). This difference corresponds to a 300-fold lower
normalized IC50 for B27-binding peptides compared with
A2-binding peptides.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
can be
interpreted as a more comprehensive peptide binding motif that
complements the previously proposed motif. We have now also
investigated 11- and 14-mer peptides and find that binding affinity of
these peptides may similarly depend mainly on the N- and C-terminal
positions referred to above. To evaluate TAP binding affinities of a large number of peptide ligands, we established a tool for prediction of TAP affinity. As may be expected, sequence context-dependent effects limited the predictive power of the Ala matrix. Using the ANN strategy, we increased predictive performance so that IC50 values of 87% of peptides could be predicted with a maximum factor of error of 10 to 30. We consider this performance sufficient for valid prediction of affinities of large groups of peptides, since errors are likely to neutralize each other. However, for more reliable prediction of TAP affinities of individual peptides, it can be envisaged to increase predictive performance by retraining ANNs with additional datasets.
An important finding of this study is the considerable variation in TAP binding affinity between ligands for different HLA class I molecules. This results from the different levels of compatibility between the preferences of TAP and individual HLA class I molecules. Thus, due to the dominant effects of the same peptide sequence positions on peptide binding to HLA class I molecules and peptide transport by TAP, the range of TAP affinities of potential ligands for an HLA class I molecule is generally defined by its peptide binding preferences. However, within this range, peptides with higher TAP affinities are more likely to be presented by HLA class I molecules. Eight of nine HLA class I molecules investigated in this study bind ligands whose mean TAP affinities are higher than those of random sequence peptides, demonstrating a general coordination of TAP and HLA class I preferences. The poor adaptation of HLA-A2 and TAP preferences is striking and may increase the chance for signal peptides to bind to and be presented by HLA-A2. Since signal peptides may frequently possess low TAP affinities, such sequences were excluded from the groups of eluted peptides and control binders used for ANN predictions in this study.
Due to the limited number of peptides tested for ANN training as well as listed in MHCPEP as HLA class I ligands, both ANN training and predictions focused on ligands for certain HLA class I molecules. One hundred and fifty-six of the 381 peptides used for ANN training bind to HLA class I molecules; among these, HLA-A2 binders (n = 68) and B27 binders (n = 28) were the most abundant, so that ANN performance for ligands for these molecules may be slightly superior to that for other less well-represented molecules.
Are the observed differences in TAP affinities likely to have functional consequences for peptide presentation by individual HLA class I molecules in vivo? Two considerations argue for an importance of these findings in vivo. First, an effect of the selectivity of human TAP on peptide presentation in vivo is suggested by higher transporter affinities of HLA class I-eluted peptides. Second, our parallel analysis of peptide binding to the rat transporters has demonstrated that measurements of peptide binding to insect cell-expressed TAP complexes reflect epitope selection by TAP in vivo (43). The major biologic effect of TAP selection in the rat is a strongly reduced presentation of peptides with positively charged residues at the C-terminus by MHC class I molecules in cells expressing the TAP2u allele, compared with that in cells expressing TAP2a (12). In peptide binding assays, binding affinities (i.e., IC50 values) of various peptides with C-terminal Arg or Lys for TAP1/TAP2u complexes were 10- to 100-fold lower than those for TAP1/TAP2a complexes (43). Differences between TAP affinities of typical ligands for HLA class I molecules reach similar levels, so that they are likely to have a functional impact.
As a corollary of the differences in TAP binding between ligands for individual HLA class I molecules, peptides binding to molecules whose selectivity is well adapted to that of TAP should be transported efficiently at lower cytosolic concentrations than peptides binding to poorly adapted molecules. In other words, potential A2 ligands need to be more abundant in the cytosol than B27 or A24 ligands to obtain TAP-mediated access to the ER. While the Kd for a typical B27 ligand (peptide R-9-L) is in the high nanomolar concentration range (4.1 x 10-7 M), typical ligands for some other alleles may possess Kd values in the 10-6 or even 10-5 M range.
Definite assessment of the biologic consequences of this finding will depend on a better understanding of two issues: the cytosolic concentrations of TAP-transportable peptides, and the extent of post-translocational trimming of peptides in the ER (38, 39). Assuming that peptide trimming in the ER plays a minor role (as suggested by the finding of higher TAP affinities of eluted 9-mer peptides), two possibilities can be envisaged. If a large variety of abundant potential ligands for all HLA class I molecules is generated in the cytosol, all HLA class I molecules will receive an equally rich variety of peptides. If, on the other hand, concentrations of most cytosolic peptides are relatively low, the variety of peptides bound by individual HLA class I molecules should vary according to the degree of adaptation of the molecule to the TAP preferences. In this case, HLA-B27 and A24 can be predicted to present a particularly rich choice of antigenic peptides and may therefore represent especially efficient HLA class I alleles.
There is some evidence that HLA-B27 may belong to the efficient HLA class I molecules. HLA-B27 assembles more rapidly in (presumably peptide-loaded) dimers with ß2m than some other alleles (4). In addition, one study found that HLA-B27 dominates in eliciting CTL responses to several viruses whenever it is present in heterozygous individuals (40). More recently, HLA-B27 has been suggested to provide the highest degree of protection from progression of HIV infection to AIDS (41). Particularly efficient presentation of cytosolic peptides with low concentrations may also play a role in the implication of HLA-B27 in ankylosing spondylitis.
Why could evolution have favored a situation in which most HLA class I molecules (including very frequent ones such as HLA-A2) display less than optimal adaptation to the TAP preferences? If presentation by most HLA class I molecules requires relatively high cytosolic peptide concentrations that are not reached by many self peptides in a normal cell, peptides derived from abundant viral proteins in an infected cell should have an important advantage for presentation. Thus, human TAP may skew the HLA class I-associated system of Ag processing and presentation to its main task, the display of abundant non-self peptides derived from viral or bacterial sources.
|
Footnotes |
|---|
2 Address correspondence and reprint requests to Dr. Peter van Endert, Institut National de la Santé et de la Recherche Médicale, Unité 25, Hopital Necker, 161 rue de Sèvres, 75743 Paris Cedex 15, France. E-mail address:
3 Abbreviations used in this paper: ER, endoplasmic reticulum; ANN, artificial neural network; IC50, concentration for 50% inhibition of specific binding; ROC, relative operating characteristics.
Received for publication November 11, 1997. Accepted for publication March 12, 1998.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Motozono, S. Yanaka, K. Tsumoto, M. Takiguchi, and T. Ueno Impact of Intrinsic Cooperative Thermodynamics of Peptide-MHC Complexes on Antiviral Activity of HIV-Specific CTL J. Immunol., May 1, 2009; 182(9): 5528 - 5536. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. E. Dugan and E. W. Hewitt Structural and Functional Dissection of the Human Cytomegalovirus Immune Evasion Protein US6 J. Virol., April 1, 2008; 82(7): 3271 - 3282. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. M. Schatz, B. Peters, N. Akkad, N. Ullrich, A. N. Martinez, O. Carroll, S. Bulik, H.-G. Rammensee, P. van Endert, H.-G. Holzhutter, et al. Characterizing the N-Terminal Processing Motif of MHC Class I Ligands J. Immunol., March 1, 2008; 180(5): 3210 - 3217. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Lundegaard, O. Lund, C. Kesmir, S. Brunak, and M. Nielsen Modeling the adaptive immune system: predictions and simulations Bioinformatics, December 15, 2007; 23(24): 3265 - 3275. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Montserrat, B. Galocha, M. Marcilla, M. Vazquez, and J. A. Lopez de Castro HLA-B*2704, an Allotype Associated with Ankylosing Spondylitis, Is Critically Dependent on Transporter Associated with Antigen Processing and Relatively Independent of Tapasin and Immunoproteasome for Maturation, Surface Expression, and T Cell Recognition: Relationship to B*2705 and B*2706 J. Immunol., November 15, 2006; 177(10): 7015 - 7023. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. T. Loffredo, J. Sidney, S. Piaskowski, A. Szymanski, J. Furlott, R. Rudersdorf, J. Reed, B. Peters, H. D. Hickman-Miller, W. Bardet, et al. The High Frequency Indian Rhesus Macaque MHC Class I Molecule, Mamu-B*01, Does Not Appear to Be Involved in CD8+ T Lymphocyte Responses to SIVmac239 J. Immunol., November 1, 2005; 175(9): 5986 - 5997. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Merino, V. Montserrat, A. Paradela, and J. A.L. de Castro Two HLA-B14 Subtypes (B*1402 and B*1403) Differentially Associated with Ankylosing Spondylitis Differ Substantially in Peptide Specificity but Have Limited Peptide and T-cell Epitope Sharing with HLA-B27 J. Biol. Chem., October 28, 2005; 280(43): 35868 - 35880. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Doytchinova, S. Hemsley, and D. R. Flower Transporter Associated with Antigen Processing Preselection of Peptides Binding to the MHC: A Bioinformatic Evaluation J. Immunol., December 1, 2004; 173(11): 6813 - 6819. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Peters, S. Bulik, R. Tampe, P. M. van Endert, and H.-G. Holzhutter Identifying MHC Class I Epitopes by Predicting the TAP Transport Efficiency of Epitope Precursors J. Immunol., August 15, 2003; 171(4): 1741 - 1749. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Fruci, G. Lauvau, L. Saveanu, M. Amicosante, R. H. Butler, A. Polack, F. Ginhoux, F. Lemonnier, H. Firat, and P. M. van Endert Quantifying Recruitment of Cytosolic Peptides for HLA Class I Presentation: Impact of TAP Transport J. Immunol., March 15, 2003; 170(6): 2977 - 2984. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Yusim, C. Kesmir, B. Gaschen, M. M. Addo, M. Altfeld, S. Brunak, A. Chigaev, V. Detours, and B. T. Korber Clustering Patterns of Cytotoxic T-Lymphocyte Epitopes in Human Immunodeficiency Virus Type 1 (HIV-1) Proteins Reveal Imprints of Immune Evasion on HIV-1 Global Variation J. Virol., July 29, 2002; 76(17): 8757 - 8768. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. K. Lemberg, F. A. Bland, A. Weihofen, V. M. Braud, and B. Martoglio Intramembrane Proteolysis of Signal Peptides: An Essential Step in the Generation of HLA-E Epitopes J. Immunol., December 1, 2001; 167(11): 6441 - 6446. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Korber, B. Gaschen, K. Yusim, R. Thakallapally, C. Kesmir, and V. Detours Evolutionary and immunological implications of contemporary HIV-1 variation Br. Med. Bull., September 1, 2001; 58(1): 19 - 42. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Lucchiari-Hartz, P. M. van Endert, G. Lauvau, R. Maier, A. Meyerhans, D. Mann, K. Eichmann, and G. Niedermann Cytotoxic T Lymphocyte Epitopes of HIV-1 Nef: Generation of Multiple Definitive Major Histocompatibility Complex Class I Ligands by Proteasomes J. Exp. Med., January 17, 2000; 191(2): 239 - 252. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Paradela, I. Alvarez, M. Garcia-Peydro, L. Sesma, M. Ramos, J. Vazquez, and J. A. Lopez de Castro Limited Diversity of Peptides Related to an Alloreactive T Cell Epitope in the HLA-B27-Bound Peptide Repertoire Results from Restrictions at Multiple Steps Along the Processing-Loading Pathway J. Immunol., January 1, 2000; 164(1): 329 - 337. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Lauvau, K. Kakimi, G. Niedermann, M. Ostankovitch, P. Yotnda, H. Firat, F. V. Chisari, and P. M. van Endert Human Transporters Associated with Antigen Processing (TAPs) Select Epitope Precursor Peptides for Processing in the Endoplasmic Reticulum and Presentation to T Cells J. Exp. Med., November 1, 1999; 190(9): 1227 - 1240. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Lauvau, B. Gubler, H. Cohen, S. Daniel, S. Caillat-Zucman, and P. M. van Endert Tapasin Enhances Assembly of Transporters Associated with Antigen Processing-dependent and -independent Peptides with HLA-A2 and HLA-B27 Expressed in Insect Cells J. Biol. Chem., October 29, 1999; 274(44): 31349 - 31358. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Saveanu, S. Daniel, and P. M. van Endert Distinct Functions of the ATP Binding Cassettes of Transporters Associated with Antigen Processing. A MUTATIONAL ANALYSIS OF WALKER A AND B SEQUENCES J. Biol. Chem., June 15, 2001; 276(25): 22107 - 22113. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
