Technical Note
Optimized determination of T cell epitope responses

https://doi.org/10.1016/S0022-1759(02)00423-4Get rights and content

Abstract

Pools of overlapping peptides corresponding to specific antigens are frequently used to identify T cell immune responses to vaccines or pathogens. While the response to the entire pool of peptides provides important information, it is often desirable to also know to which individual peptides within the pool the immune responses are directed. In this report, we analyzed various ways of deconvoluting an immune response to a pool of peptides to determine the number of different peptides to which the T cells are responding. We used a Monte Carlo simulation to optimize the construction of peptide pools that could identify responses to individual peptides using the fewest numbers of assays and patient material. We find that the number of assays required to deconvolute a pool increases by the logarithm of the number of peptides within the pool; however, the optimum configuration of pools changes dramatically according to the number of responses to individual peptides that are expected to be in the sample. Our simulation will help in the design of clinical trials in which the breadth of the response is being measured, by allowing a calculation for the minimum amount of blood that needs to be collected. In addition, our results guide the design and implementation of the experiments to deconvolute the responses to individual peptide epitopes.

Introduction

Recently, there have been great advances in the ability to detect and characterize antigen-specific T cells. Small overlapping peptides that span an entire protein(s) of interest are being routinely used to determine the number of antigen-reactive T cells within clinical samples by intracellular flow cytometry or ELISpot Kern et al., 1998, Kern et al., 1999, Kern et al., 2002, Addo et al., 2001, Altfeld et al., 2001, Betts et al., 2001, Maecker et al., 2001, Yu et al., 2002. Typically, such pools contain peptides that are between 9 and 20 amino acids in length, and overlap to a degree that ensures that every T cell epitope is represented. We have adopted the approach of using peptides that are 15 amino acids long, and overlap by 11 (i.e., starting every 4th amino acid through a protein) Betts et al., 2001, Maecker et al., 2001. For protein such as HIV gag, which is approximately 500 amino acids long, this translates into a set of 120–125 peptides (depending on the strain of HIV from which the gag was derived).

As an alternative, some investigators have used pools of peptides that represent predicted or known HLA class-I-restricted T cell epitopes Dalod et al., 1999, Betts et al., 2000. However, we feel that there are advantages to using the overlapping peptides approach: (1) the HLA type of the patient, and the corresponding dominant epitopes relevant to the patient's haplotype, do not need to be determined in advance; and (2) both CD4+ and CD8+ T cell responses can be assessed Betts et al., 2001, Maecker et al., 2001. Irrespective of these considerations, the methods described here are relevant to deconvoluting peptide responses from pools of overlapping peptides or optimized class I epitopes.

Enumeration of T cell responses to peptide pools by flow cytometric assays usually requires between 0.5 and 1 million cells; ELISpot assays use a few hundred thousand (but are usually done in duplicate or triplicate). Further testing of every individual peptide within a pool to determine to which peptides the T cells are responding would be prohibitive, from both a sample requirement and a reagent requirement. For example, for the HIV gag, one would have to perform 120 separate assays, requiring at least 50 ml of blood and 120 tests. Instead, peptides may be pooled together in sets to determine responses. This has previously been accomplished by creating multiple smaller pools in a “matrix” format, and has been relatively successful for determining individual peptide responses for pools of 100–120 peptides Kern et al., 1999, Betts et al., 2000.

However, there is a need to deconvolute responses from larger pools of peptides. As an example, we are currently planning a clinical trial where we will immunize with DNA vectors that express the HIV envelope protein from each of three clades as well as an HIV gag–pol–nef polyprotein. The combined immunogen expresses four proteins, and the total number of peptides we will use to determine immunogenicity is approximately 800. For some of the vaccinees we will wish to determine the breadth of the response. However, the composition of the “matrix” pools of peptides that would best accomplish this is not self-evident.

Therefore, we wrote a program to model the determination of peptide responses. The software does a complete test of all possible peptide pool configurations against a varying number of potential positive responses from within the immunogen. We were able to determine the optimal configuration of peptide pools for deconvoluting peptide responses, with the goal of minimizing the number of assays and the amount of clinical sample required. Here, we show this optimization for either 120 or 800 peptide libraries; additional results for 64, 480, or 1200 peptide pool libraries are available. The software can also be used to output the optimal peptide pool configuration, and, given the results of a series of assays, can identify which peptides from those pools comprise the response.

Section snippets

Materials and methods

The software used in this analysis, “DeconvoluteThis” version 1.0, was written in C++ using the Metrowerks Codewarrior Pro for Macintosh framework. The application and source code are available from the author by request. The software is implemented for Mac OSX, but may be run on earlier versions of the operating system. The software is capable of both simulating a series of deconvolutions to identify the optimal peptide configuration, as well as aiding in such an experiment by outputting the

Results and discussion

The simplest way to deconvolute the individual responses within a peptide pool is to simply test each of the peptides from the pool separately. From a pool of 100 peptides, however, this would require 100 assays to find the few that are responding: this is highly inefficient. For the purpose of explaining how we approached this problem, we will consider a peptide pool comprising 100 distinct peptides. Typically, the number of peptides that will generate a response from such a pool will be less

References (10)

  • H.T. Maecker et al.

    Use of overlapping peptide mixtures as antigens for cytokine flow cytometry

    J. Immunol. Methods

    (2001)
  • M.M. Addo et al.

    The HIV-1 regulatory proteins Tat and Rev are frequently targeted by cytotoxic T lymphocytes derived from HIV-1-infected individuals

    Proc. Natl. Acad. Sci. U. S. A.

    (2001)
  • M. Altfeld et al.

    Vpr is preferentially targeted by CTL during HIV-1 infection

    J. Immunol.

    (2001)
  • M.R. Betts et al.

    Putative immunodominant human immunodeficiency virus-specific CD8(+) T-cell responses cannot be predicted by major histocompatibility complex class I haplotype

    J. Virol.

    (2000)
  • M.R. Betts et al.

    Analysis of total human immunodeficiency virus (HIV)-specific CD4(+) and CD8(+) T-cell responses: relationship to viral load in untreated HIV infection

    J. Virol.

    (2001)
There are more references available in the full text version of this article.

Cited by (0)

View full text