MAJOR ARTICLE
Moving to Human Immunodeficiency Virus
Type 1 Vaccine Efficacy Trials: Defining T Cell
Responses As Potential Correlates of Immunity
Nina D. Russell,1,3 Michael G. Hudgens,2 Richard Ha,1 Colin Havenar-Daughton,1 and M. Juliana McElrath1,3
1
Program in Infectious Diseases, Clinical Research Division, and 2Program in Biostatistics, Public Health Science Division, Fred Hutchinson Cancer
Research Center, and 3Department of Medicine, University of Washington, Seattle
There is evidence in both simian immunodeficiency virus and human immunodeficiency virus (HIV) type 1
infection that class I major histocompatibility complex–restricted CD8+ cytotoxic T lymphocytes play a pivotal
role in controlling infection and, potentially, in protecting by immunization. Progress has been made in
designing strategies to elicit these responses with HIV-1 vaccines, but methods to reproducibly quantify them
have posed difficulties. An interferon-g enzyme-linked immunospot assay, using peptide pools spanning the
HIV-1 genes, was developed and standardized. This method is rapid (2 days), sensitive (threshold of detection,
⭓0.005%), quantitative, feasible using cryopreserved cells, and able to define epitope specificities. When this
assay was applied to 36 HIV-1–seropositive and 10 HIV-1–seronegative subjects, it proved to be robust (specificity, 100%). When responses in natural infection were compared with vaccine-induced responses, vaccine
recipient responses were ⭓1 log lower, which confirms the importance of using this sensitive assay as an
initial screen in vaccine protocols.
There is an urgent global need to develop an effective
human immunodeficiency virus (HIV) type 1 vaccine.
A major challenge in this undertaking has been to better
define the components of immunity that are elicited
by immunization and that may confer protection. It is
presumed that HIV-1 prevention by immunization will
Received 23 July 2002; revised 10 October 2002; electronically published 6
January 2003.
Presented in part: 8th Conference on Retroviruses and Opportunistic Infections,
Chicago, 4–8 February 2001 (abstract 177).
Written informed consent was obtained from all subjects, and the human
experimentation guidelines of the Fred Hutchinson Cancer Research Center
institutional review board were followed.
Financial support: National Institutes of Health (grants AI-46725, AI-48017, AI41535, AI-47806, and AI-27757); Burroughs Wellcome Clinical Scientist Award in
Translational Research (to M.J.M.).
Reprints or correspondence: Dr. M. Juliana McElrath, Program in Infectious
Diseases, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. N, PO Box
19024, Seattle, WA 98109 ([email protected]).
The Journal of Infectious Diseases 2003; 187:226–42
2003 by the Infectious Diseases Society of America. All rights reserved.
0022-1899/2003/18702-0007$15.00
226 • JID 2003:187 (15 January) • Russell et al.
be mediated substantially, although probably not exclusively, by class I major histocompatibility complex
(MHC)–restricted HIV-1–specific CD8⫹ T cells. Evidence to support this is based on the significant contribution of HIV-1–specific CD8⫹ T cells in controlling
infection and, possibly, in protecting against transmission in seronegative persons with repeated high-risk
HIV-1 exposures [1–7]. HIV-1–specific CD8⫹ T cells
exert antiviral activities through cytolysis of infected
cells and by release of cytokines, particularly interferon
(IFN)–g, and chemokines, such as RANTES, macrophage inflammatory protein (MIP)–1a, and MIP-1b.
Thus, an important goal of HIV-1 vaccine clinical trials
is to elicit high-frequency, cytolytic, and IFN-g–secreting memory CD8⫹ T cells that are capable of recognizing multiple epitopes within diverse HIV-1 strains.
Recent studies in nonhuman primate models and in
early phase 1 and 2 clinical trials have indicated that
the administration of live vectors, particularly recombinant pox viruses, can induce class I MHC–restricted
cytotoxic T lymphocytes that recognize epitopes expressed by
the HIV-1 gene inserts. However, the magnitude, breadth, and
persistence of these responses have been variable [8–11]. Clinical trials with other promising viral vectors and DNA constructs are also under way, and a method to systematically
compare these regimens in phase 1 and 2 trials is needed. More
important, the decision to proceed with a large-scale efficacy
trial that uses a vaccine strategy designed to protect primarily
through the induction of CD8⫹ T cell effectors will be based
on the precise identification of the specificities and frequencies
of responses sufficient to render this immune defense. Thus,
it is imperative to establish a standardized, sensitive approach
to measure HIV-1–specific CD8⫹ T lymphocyte activities that
can be employed in field trial settings and used to compare the
relative immunogenicities of various vaccine regimens in phase
1 and 2 trials.
Detection of CD8⫹ cytotoxic T lymphocyte (CTL) activity has
traditionally relied on the lysis of autologous target cells expressing HIV-1 antigens in a chromium release assay, which indicates
target cell membrane disruption resulting from release of perforin
and granzymes by CTLs. This assay has been used in numerous
vaccine studies [12–14] but has several disadvantages. Because
the level of responses are lower in HIV-1–uninfected vaccine
recipients than in HIV-1–infected persons, optimal detection of
antigen-specific CD8⫹ T cells by these methods requires fresh,
rather than cryopreserved, peripheral blood mononuclear cells
(PBMC). Real-time assays become problematic in the conduct
of multicenter international trials. Moreover, quantitation of cytolytic activities currently relies on the use of a limiting dilution
assay, which requires laborious in vitro stimulation and expansion steps that may be less reflective of the in vivo state.
Enumeration of virus-specific CD8⫹ T cells by techniques
such as IFN-g ELISPOT, MHC-peptide tetramer staining, and
flow cytometric analysis of intracellular cytokine production
offers the advantage of improved sensitivity and more-rapid
delineation of class I MHC–restricted epitopes, compared with
that of chromium-release CTL assays. In addition, these approaches are simpler to perform, require fewer cells, and obviate
the need to expand and clone antigen-specific T lymphocytes
and establish autologous B lymphoblastoid cell lines [15–19].
The IFN-g ELISPOT assay offers an advantage over tetramer
staining in its ability to provide a functional analysis of the
HIV-1–specific response in populations spanning multiple class
I HLA types without prior knowledge of epitope recognition.
In addition, advanced operator training and expensive equipment are not necessary for measuring the frequency of responses by ELISPOT, which makes this technique potentially
more feasible in a field setting than flow-based detection of
intracellular cytokine secretion.
For these reasons, we chose to focus on the development
and validation of an IFN-g ELISPOT assay for the screening
and quantitation of HIV-1–specific CD8⫹ T cells induced by
vaccination. Lymphocytes from 6 chronically HIV-1–infected
individuals were used to optimize the assay and to ensure its
sensitivity as a screening test in vaccine studies. Assay performance was then assessed in persons at various stages of HIV-1
infection, as well as in vaccinated and unvaccinated HIV-1–uninfected subjects. Our results indicate that the IFN-g ELISPOT
assay can provide a quantitative measurement of the breadth
of the HIV-specific CD8⫹ T cell response and can define HIV1 epitopes. This highly sensitive and specific approach appears
to be feasible for application in large-scale HIV-1 vaccine clinical trials, to determine the role of vaccine-induced CD8⫹ T
cells in conferring protection against infection.
POPULATION, MATERIALS, AND METHODS
Study population. To establish optimal conditions for the performance of the IFN-g ELISPOT assay in HIV-1 vaccine recipients, fresh and cryopreserved PBMC purified from a single leukapheresis were analyzed from 6 HIV-1–infected subjects. One
patient (MP0471) had chronic HIV-1 infection, and 5 others
(NP1, NP2, NP13, NP14, and NP15) were categorized as longterm nonprogressors (LTNPs), as defined elsewhere [20]. HIV1–specific responses were also evaluated in PBMC from HIV1–uninfected persons participating in a double-blind, placebo-controlled vaccine trial sponsored by the AIDS Vaccine Evaluation Group (AVEG protocol 022), subsequently known as the
HIV Vaccine Trials Network. The immunogen was a recombinant
canarypox vector containing HIV-1 genes encoding Env, Gag,
and protease (vCP205; Aventis Pasteur) with or without a recombinant gp120 boost (rgp120/SF-2, Chiron; rgp120/MN,
VaxGen).
To establish criteria to define a positive response, assays were
performed using PBMC from 36 HIV-1–seropositive and 10
HIV-1–seronegative individuals. There were 6 untreated, primary HIV-1–infected patients whose cells were obtained at a
mean of 71 days after infection, 12 untreated chronically HIV1–infected patients, 8 treated chronically HIV-1–infected subjects, and 10 LTNPs. The 10 seronegative individuals had been
previously enrolled in AVEG protocol 201; prevaccination cryopreserved PBMC from 5 high-risk and 5 low-risk study participants were used for these experiments after ∼8.2 years of
storage in liquid nitrogen [21].
Peptides. The National Institutes of Health (NIH) AIDS
Research and Reference Reagent Program provided synthetic
HIV-1 20-mer peptides with 10-aa overlaps. These included 49
Gag 20-mers based on the HIV-1 subtype B consensus sequence
(HXB2), 80 Env (MN), 100 Pol (HXB2), and 20 Nef (BRU/
LAI). In addition, 122 HIV-1 Gag 15-mers overlapping by 11
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 227
aa (SF2) were generously provided by Dr. Louis Picker (Oregon
Health Sciences University, Portland), as well as the NIH AIDS
Reagent Program (HXB2). Two hundred twelve HIV-1 Env
(MN), 248 Pol (HXB2), and 49 Nef (BRU) 15-mers overlapping
by 11 aa were obtained from SynPep.
Negative control wells contained either media alone (no peptide) or a pool of 5 irrelevant peptides (purchased from Mimotopes) derived from highly conserved regions of actin and HLA
class I a chain precursor. Each peptide was reconstituted in 100%
dimethyl sulfoxide (DMSO) and used at a final concentration
of 2 mg/mL (!1% DMSO), unless otherwise indicated.
IFN-g ELISPOT assays. An ELISPOT assay was used to
detect HIV-1–specific IFN-g–producing T cells from either
fresh or cryopreserved PBMC. Ninety-six–well hydrophobic
polyvinylidene difluoride membrane-bottomed plates (Millipore) were coated with 50 mL of 10 mg/mL anti–IFN-g monoclonal antibody (MAb; 1-D1K, mouse IgG1; Mabtech) overnight at 4C. After washing 3 times with PBS (pH 7.2), plates
were blocked with RPMI/HEPES with 10% fetal bovine serum
(FBS) (R10) at room temperature for 1 h.
PBMC were added in 100 mL of R10 to precoated plates at
concentrations of 1 ⫻ 10 5 or 2 ⫻ 10 5 cells/well. Cryopreserved
PBMC were thawed and incubated overnight in RPMI 1640
medium with 10% FBS before being added to plate. To determine the cell subset producing IFN-g, T cell depletions were
performed using anti-CD8⫹ and anti-CD4⫹ antibody-coated
immunomagnetic beads (Microbeads, Miltenyi Biotec; Dynabeads, Dynal), according to the manufacturers’ instructions.
HIV-1–specific peptides and control peptides were added to
desired wells to give a final concentration of 2 or 5 mg/mL.
Cells stimulated with phytohemagglutinin (1 mg/mL) served as
a positive control. Responses to individual or pooled peptides
were tested in duplicate. For some experiments, a matrix of
Gag peptide pools was used to map the response to the single
peptide level, as described elsewhere [22, 23].
Plates were incubated overnight (16–20 h) at 37C in 5%
CO2, washed with PBS containing 0.05% Tween-20, and incubated at room temperature (RT) for 2 h with a secondary
biotinylated anti–IFN-g MAb at 1 mg/mL (7-B6-1, mouse IgG1;
Mabtech). Avidin biotinylated enzyme complex (Vectastain
ABC Elite Kit, PK-6100; Vector Laboratories) was added at RT
for 1 h, followed by AEC peroxidase substrate (Vectastain).
After developing plates for ∼7 min, reaction was stopped by
washing with water, and plates were air-dried. Colored spotforming cells (SFCs) were counted manually under a stereomicroscope (Stemi 2000; Carl Zeiss) or using an automated
ELISPOT reader (Immunospot; Cellular Technology). Responses were considered to be positive if (1) ⭓10 SFCs were
detected/2 ⫻ 10 5 cells after subtraction of the negative control
and (2) SFCs were ⭓2-fold than those in the negative control
wells.
228 • JID 2003:187 (15 January) • Russell et al.
Statistical analysis. Simple summary statistics, tests, and
plots were used for initial data exploration. To determine the
qualitative agreement between 2 methods (e.g., overnight vs.
2-day incubation time), standard methods for 2 ⫻ 2 contingency tables were used [24]. Specifically, McNemar’s test was
used to assess marginal homogeneity, and the tetrachoric correlation coefficient (TCC) was used to assess the degree of
positive association. Quantitative agreement was assessed using
net responses calculated by subtracting the mean control response from the mean experimental response and converting
to 200,000 cells/well, if necessary. Statistical measures of quantitative agreement focused on the concordance correlation coefficient (CCC) and the total deviation index (TDI) [25]. The
CCC measures agreement along the identity line, in which a
value of 1 represents perfect agreement (Y p X ), a value of ⫺1
represents perfect disagreement (Y p ⫺X ), and a value of 0
represents no agreement. The CCC factors in a product of
accuracy and precision coefficients where the accuracy coefficient is 1 if and only if the marginal distribution of Y and X
are equal and the precision coefficient is simply the Pearson’s
correlation coefficient (r). The accuracy coefficient measures
how far the least-squares line is from the line of equality,
whereas r measures how well the least-squares line fits the data. The TDI is defined such that the absolute value of the
differences between the 2 methods is !TDI1⫺a, with probability
1 ⫺ a. To assess agreement, outcomes for the same PBMC
across different peptide pools were assumed to be independent,
and sensitivity analysis was performed by stratifying the data
according to nonoverlapping peptide pools (pools of peptides
that did not contain the same individual peptide). Mixed-effects
models were also used to account for possible dependency
among repeated observations on the same PBMC [26]. All tests
were computed for a significance level of a p .05. All statistical
computations were performed using S-PLUS (version 6.0; Insightful) or SAS (version 8.1; SAS Institute) software.
RESULTS
To standardize and optimize the identification and quantitation of HIV-1–specific CD8⫹ T cells, 6 HIV-1–infected patients
(NP1, NP2, NP13, NP14, NP15, and MP0471) were chosen for
detailed study with the IFN-g ELISPOT assay. These individuals
had previously demonstrated consistent HIV-1–specific class I
MHC–restricted CTLs recognizing ⭓1 epitopes. Clinical and
virological profiles of the patients included the following: median age, 39 years (range, 33–55 years); median duration of
infection, 13 years (range, 6–16 years); median CD4⫹ cell count,
505 cells/mm3 (range, 366–1225 cells/mm3); and median plasma
HIV-1 RNA load at the time of leukapheresis, 462 copies/mL
(range, !50–39,565 copies/mL). Only 1 (MP0471) of 6 patients
had received antiretroviral therapy at the time of study. Sever-
al technical parameters of the assay were varied to establish
maximal sensitivity. Conditions such as cell concentration, peptide concentration, incubation time, peptide pool size, and peptide length were tested to improve sensitivity while maintaining
a simple, rapid, and relatively low-cost methodology. Assays
were then optimized to identify epitope specificities and the
MHC restricting allele. Subsequent analyses were performed on
HIV-1–infected patients with acute and chronic infection, as
well as HIV-1–uninfected vaccine recipients, to determine the
frequency and range of responses.
Cell number per well. To determine the optimal cell number per well to detect HIV-1–specific CD8⫹ responses in the
low frequency range, PBMC from 4 donors were diluted serially
(8.0 ⫻ 10 5–0 cells/well) and stimulated with single Gag 20-mer
(figure 1A and 1B) or 15-mer peptides (figure 1C and 1D)
containing class I MHC–restricted HIV-1 epitopes identified in
previous experiments. As shown in 4 representative examples
in figure 1, positive responses were commonly detected with
150,000 PBMC/well, and increases, often linear, were observed
up to 400,000–600,000 PBMC/well. At greater cell numbers,
the frequency of HIV-1–specific IFN-g SFCs reached a plateau
and, in some cases, declined in conjunction with increasing
IFN-g SFC frequencies in the control wells (figures 1C and
1D). These findings were confirmed with repeated testing that
used both 20- and 15-mers.
Cell numbers from clinical trial participants will be limiting
if analyses include the identification of specific T cell subsets
recognizing multiple gene products and definition of optimal
epitopes within these regions. We anticipate that ⭐200,000 cells/
well will be available from cryopreserved specimens without exceeding the blood volume restrictions for vaccine trials. Thus,
on the basis of the above results, a comparison between 100,000
and 200,000 cryopreserved PBMC/well from 5 HIV-1–infected
donors was performed after overnight stimulation with pools
containing 50, 25, or 10 Gag 20-mer peptides. Positive responses
were easily detected with both cell concentrations when the frequency of IFN-g SFCs was higher (150 SFCs/well; data not
shown). However, responses clearly positive with 200,000 cells/
well were frequently below or at the threshold for positivity with
100,000 PBMC/well. In repeated experiments in which responses
Figure 1. Titration of cell no./well. Serial dilutions (8.0 ⫻ 105 –0 cells/well) of cryopreserved peripheral blood mononuclear cells (PBMC) obtained
from 4 human immunodeficiency virus (HIV) type 1–infected donors—MP0471 (A), NP13 (B), NP14 (C), and NP15 (D)—were added to the wells. Cells
were stimulated in triplicate with a single Gag peptide, either a 20-mer (A and B) or a 15-mer (C and D) peptide containing a class I major
histocompatibility complex–restricted HIV-1 epitope identified in previous experiments or with a pool of negative control peptides. Mean total of
interferon (IFN)–g spot-forming cells (SFCs) is depicted for each cell concentration.
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 229
Figure 2. Comparison of peptide stimulation times. Cryopreserved peripheral blood mononuclear cells (PBMC) from 2 human immunodeficiency
virus (HIV) type 1–infected donors, MP0471 (A) and NP15 (B), were stimulated with Gag 15-mer peptide pools for 1 (16–20 h) or 2 days (36–40 h),
and the interferon (IFN)–g spot-forming cell (SFC) frequency is shown. Cells were stimulated with either a pool of all 122 Gag 15-mer peptides, pools
of 25 Gag 15-mers, or the negative control peptide pool. Each bar represents the mean (SD) no. of IFN-g SFCs in duplicate wells. C, Scatterplot
of stimulation times for all pools tested in 4 donors. PBMC from MP0471 and NP15 were stimulated with Gag 15-mers, and PBMC from NP1 and
NP13 were stimulated with Gag 20-mers. Thirty-one positive concordant experiments are shown by patient, with least-squares line (solid line) and
line of equality (dashed line).
were at low frequency, 200,000 cells/well were necessary to consistently detect IFN-g SFCs distinct from background responses
in the control wells. Therefore, the use of 200,000 PBMC/well
may be more advantageous in demonstrating vaccine-induced
HIV-1–specific T cells present in lower frequencies than those
recognizing immunodominant epitopes typically found in infected persons.
Duration of stimulation. To establish the time sufficient
to stimulate antigen-specific T cells to secrete IFN-g, we ana230 • JID 2003:187 (15 January) • Russell et al.
lyzed SFC frequencies in 4 patients, using a total of 65 peptide
pools, and compared overnight (16–20 h) incubation with 2day (36–40 h) incubation (figure 2). Of note, there was no
substantial increase in response to the control peptides with
the longer incubation in any of the experiments performed.
When cryopreserved PBMC were tested from MP0471, the response to the pool of all 122 Gag 15-mers decreased by 49
SFCs (20.4%) with the longer stimulation time, whereas the
response to a pool of 25 Gag 15-mers increased by 30 SFCs
(16.5%) (figure 2A). Only 1 of 4 peptide pools recognized by
PBMC from NP15 generated a greater number of spots with
the longer stimulation time (figure 2B, pool 1–25). PBMC from
NP13 and NP1, known low-level responders, were similarly
tested in repeat experiments with Gag 20-mers in pools of 50,
25, and 10 peptides (data not shown).
When data from all 65 peptide pool comparisons were stratified according to positive and negative responses, there was
excellent qualitative agreement between the 2 incubation times.
Only 4 of 65 comparisons yielded discordant results, with
higher frequency responses (125 SFCs/well) detected regardless
of incubation time. For the 30 comparisons that used pools of
10 Gag 20-mer peptides, there was near-perfect agreement, with
only 2 discordant outcomes (TCC, 0.97; 95% confidence interval [CI], 0.90–1.00; P p 1.00, McNemar’s test). Similar results held for other pools. Analysis of quantitative agreement
between the 2 incubation times focused on the 31 comparisons
in which both incubation times yielded positive responses. The
sign and Wilcoxon signed-rank tests on the differences in net
responses gave mixed results (P p .15 and P p .04, respectively). Figure 2C shows a scatterplot of the 31 comparisons
with the solid least-squares line (intercept b0 p 16.1 [95% CI,
2.7–29.5]; slope b1 p 0.7 [95% CI, 0.6–0.8]) and the dotted
45 line of perfect agreement. The CCC equaled 0.85 with 1sided lower 95% confidence limit (CL) of 0.79, indicating good
agreement between the 2 methods. The accuracy and precision
(Pearson’s correlation) coefficients were 0.94 and 0.91, respectively, and the TDI0.9 estimate was 53.7 (1-sided upper 95% CL,
66.5), which suggests that, on average, 90% of the paired experiments differed by !54 SFCs. Quantitative agreement analysis considering only nonoverlapping peptide pools (pools of
peptides that did not contain the same individual peptide) gave
similar results. Taken together, little difference both qualitatively
and quantitatively was observed between the 2 stimulation periods. Thus, for convenience, we chose the shorter overnight
period of stimulation.
Peptide concentration. To assess the optimal concentration of each peptide within a pool, we sought to determine the
lowest concentration that preserves sensitivity but conserves
reagents. Of note, lower IFN-g SFC frequencies were commonly
observed when peptide pools contained 11% DMSO (data not
shown); thus, final DMSO concentrations were subsequently
maintained at !1%, to avoid cellular toxicity. Initial experiments examined responses to 15- and 20-mers titrated from 0
to 20 mg/mL using both individual peptides and pools of peptides containing the individual peptide (figure 3A and 3B, representative examples in 2 donors). In general, IFN-g SFCs were
detected on stimulation with ⭓0.1 mg/mL of peptide. The IFNg SFC frequency rose with increasing concentrations of the
individual peptide, up to 5–10 mg/mL, after which a plateau
was reached (figure 3A and 3B). When responses to the peptide
pools were tested, similar or decreased frequencies were observed with peptide concentrations ⭓5 mg/mL, and a striking
fall in SFC frequencies was noted when 20 mg/mL of each
peptide was used (figure 3A and 3B).
Obvious advantages in using lower peptide concentrations
include the ability to test more peptides in each pool without
increasing DMSO content and to economize on reagents. Thus,
additional studies were performed to determine whether 2 mg/
mL, rather than 5 mg/mL, was sufficient for detection of HIV1–specific T cells. Cryopreserved PBMC from 5 donors were
stimulated with 20-mers and from those 2 donors with 15mers. No benefit in the ability to detect a positive response was
observed with use of the higher peptide concentration (figure
3C and 3D, representative experiments). For the 2 donors tested
with 15-mers, 9 experiments were done for each donor at both
2 and 5 mg/mL. Of these 18 pairs, there was only one discordant
response (positive at 5 mg/mL and negative at 2 mg/mL), which
indicates excellent qualitative agreement between the 2 methods
(TCC p 1.00; P p 1.00, McNemar’s test). For the 13 paired
samples with concordant positive responses, the CCC was 0.98
(1-sided lower 95% CL, 0.97), indicating excellent agreement
between the 2 methods. The accuracy and precision (Pearson’s
correlation) coefficients equaled 0.99 and 1.00, respectively, and
the TDI0.9 estimate was 77.8 (1-sided upper 95% CL, 108.4).
The especially high correlation coefficients are partially attributable to the broad range of the response data (10–851 net
SFC/105 PBMC), for which the intersample variability is much
greater than the intrapair variability. Sign and Wilcoxon signedrank tests were both significant (P ! .01 ), indicating a stronger
response with the higher peptide concentration. Using a mixed
effect model on positive responses, stimulation with the higher
peptide concentration resulted in a small percentage increase
in SFCs/2 ⫻ 10 5 PBMC recognizing pools of 25 15-mers (16%;
P p .03) and individual 15-mers (12%, P p .06). This benefit
was not appreciated, however, after stimulation with the larger
pool of 122 peptides (P p .63). On the basis of these findings,
peptide concentrations of 2 mg/mL are sufficient for the detection of HIV-1–specific IFN-g SFCs, particularly when
screening large peptide pools, although slightly lower SFC frequencies may result when testing individual or smaller peptides
pools.
Peptide pool size. Next, we determined the optimal number of peptides in a pool that could afford detection of responses
to multiple epitopes while conserving cell numbers. The SFC
frequencies from 6 donors were compared using 20-mer peptide pools spanning Gag: 1 large pool of 49 20-mers (1–49), 2
pools of 25-mers (1–25 and 26–49), 5 pools of 10-mers (1–10,
11–20, 21–30, 31–40, and 41–49), and 25 pools of 2 20-mers.
Positive responses to individual 20-mers were generally maintained by stimulation with pools of all sizes (figure 4, representative examples). However, individual peptide responses
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 231
Figure 3. Stimulation with varying peptide concentrations/well. Cryopreserved peripheral blood mononuclear cells (PBMC) from 2 donors, NP15
(A and C) and NP14 (B and D), were thawed and stimulated with Gag 15-mers at varying concentrations. A and B, interferon (IFN)–g spot-forming
cell (SFC) frequencies after stimulation with Gag peptides alone or in combination, with each peptide concentration titrated from 0 to 20 mg/mL. C
and D, Comparison of IFN-g SFC frequencies when PBMC were stimulated with either 2 or 5 mg/mL of peptide. Both individual Gag 15-mers and
pools of 15-mers were tested. Bars represent the mean (SD) nos. of IFN-g SFC in duplicate wells.
were frequently greater (31%–220%) when the 20-mer was
contained in a smaller pool (10 or 2 peptides), compared with
the larger pools (25 or 50 peptides) (figure 4A and 4B). Thus,
stimulation with 10 20-mers within a pool may be optimal to
detect low-frequency responses, although responses were clearly
detectable when larger pools of 25–50 20-mers were used. Similar comparisons were performed with Gag 15-mers in pools
of 122, 25 (spanning the same amino acid sequence as pools
containing 10 20-mers), or individually and revealed fewer differences. However, there tended to be an advantage in the detection of higher SFC frequencies when using pools of 25, rather
than 122, 15-mers (data not shown).
To determine whether there is an inhibitory effect when using
peptides pools containing 11 recognized epitope, we compared
SFC frequencies with pools of 10 20-mers with those detected
using the individual peptides comprising the pool. Experiments
in 2 donors using Env and Pol 20-mer peptides are depicted
in figure 4C and 4D. The SFC frequency measured after stimulation with the peptides mixed in the pool was not necessarily
232 • JID 2003:187 (15 January) • Russell et al.
the sum of the responses seen with the individual peptides, but
it was generally greater than those recognized by the individual
peptides. Thus, there was no evidence of a significant inhibitory
effect by combining 11 responding peptide in a pool.
Amino acid length of peptide. To determine the optimal
peptide length to stimulate and detect both CD8⫹ and CD4⫹
HIV-1–specific T cells, we compared 15-mers overlapping by 11
aa (n p 122) and 20-mers overlapping by 10 aa (n p 49) in
repeated experiments in 5 seropositive donors. We observed up
to ⭓2-fold SFC frequencies after stimulation with the pool of
Gag 15-mers, compared with the pool of 20-mers (figure 5A and
5B, representative examples). Even greater differences were observed with the individual 15-mers, compared with those of the
20-mers containing the same epitope (figure 5A and 5B). Of
note, a response considered to be negative after stimulation with
a 20-mer peptide (6 SFC for 20-mer 15-QMVHQAISPRTLNAWVKVVE) was easily detectable after stimulation with the corresponding, embedded 15-mer peptide (62 SFC for 15-mer 37QAISPRTLNAWVKVV) (figure 5B).
Figure 4. Peptide pool size comparisons. A and B, Pools of 50, 25, 10, and 2 Gag 20-mer peptides were used to stimulate cryopreserved peripheral
blood mononuclear cells (PBMC) from 2 human immunodeficiency virus (HIV) type 1–infected donors, (A) NP13 and (B) MP0471. In these experiments,
only 1 recognized epitope was contained within each of the larger pools. C and D, Interferon (IFN)–g responses to pools of 10 Gag 20-mer peptides
were compared with the responses observed with multiple individual peptides contained within those pools that stimulated a positive response. Bars
represent the mean (SD) no. of spot-forming cells (SFCs) in duplicate wells.
To quantify differences in SFC frequencies after stimulation
with peptides of different lengths, data from all experiments in
5 donors were analyzed. As seen in the scatterplot in figure 5C,
every net response was greater when 15-mers were used. All
responses are included in figure 5C, except one experiment
(NP14), which generated 11000 SFCs/2 ⫻ 10 5 PBMC using 15mer peptides. Over the range of data in figure 5C, the CCC
was 0.60 (1-sided lower 95% CL, 0.40), with precision and
accuracy coefficients of 0.92 and 0.66, respectively, indicating
a good fit regarding the least-squares line but a change in scale
between the 2 assays. Both the sign and Wilcoxon signed-rank
tests were highly significant (P ! .001), indicating higher response rates by stimulation with 15-mers. Similarly, the solid
least-squares regression line in figure 5C had a slope significantly !1 (b1 p 0.64 [95% CI, 0.48–0.80]), which implies that
15-mers tend to give ∼1.6 times the number of spots as 20-
mers. In summary, these data suggest that stimulation with
pools of 15-mer peptides increases the ability to detect lowlevel responses and leads to the detection of higher SFC frequencies in this assay.
Fresh versus frozen PBMC. To discern the sensitivity of
the IFN-g ELISPOT assay using frozen cells, we compared results using fresh and cryopreserved PBMC obtained from a
single leukapheresis in our 6 HIV-1–seropositive patients. For
these and subsequent experiments using cells thawed between
9 and 62 days after cryopreservation, cell recovery averaged
60% and cell viability averaged 90%. Of note, improved spot
quality and higher spot counts were observed by incubating
thawed cells overnight at 37C (data not shown), which led to
a mean increase of 5% viability and mean loss of 13% of cells.
Although the SFC frequencies from some frozen populations
were diminished, compared with the SFC frequencies from
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 233
Figure 5. Comparison of peptide amino-acid length. Individual and pooled 20-mer peptides overlapping by 10 aa and 15-mer peptides overlapping
by 11 aa were compared for stimulation of cryopreserved peripheral blood mononuclear cells (PBMC) in 2 donors, NP14 (A) and NP1 (B). Also included
are individual peptides containing class I major histocompatibility complex–restricted human immunodeficiency virus (HIV) type 1 epitopes recognized
by these donors. Each bar represents the mean (SD) no. of interferon (IFN)–g spot-forming cells (SFCs) in duplicate wells. C, Scatterplot of 15mers vs. 20-mers in 5 donors with line of equality (dashed line) and least-squares regression line (solid line).
fresh populations, positive responses were consistently maintained (figure 6A and 6B). In addition, the overall proportion
of positive responses observed with the fresh PBMC was not
significantly different than that observed with the frozen cells.
For example, of the 144 pools of 2 Gag 20-mer peptides tested
across all 6 donors, 13.9% were positive when using frozen
PBMC, compared with 11.8% positive when using fresh cells
(P p .61, McNemar’s test). Similar results were observed with
the pools of 10 and 25 Gag 20-mer peptides. In addition, there
was a strong positive association between fresh and cryopreserved assays. For example, the TCC was 0.86 (95% CI,
0.63–1.00) for the 30 pools of 10 Gag 20-mer peptides.
To determine the impact of the duration of cryopreservation
on SFC frequency, Gag-specific responses from the same leukapheresis were tested in 5 donors using fresh PBMC and PBMC
thawed after several time intervals spanning 442 days. As seen
in the representative data shown in figure 6C, with the exception of responses to 1 pool of 10 Gag peptides (Gag 11–20), the
SFC frequency was greater when testing fresh PBMC than previously frozen PBMC. A moderate decline in SFC was noted after
4 months of freezing. However, consistently positive responses
were maintained over the 14 months when examining all responding peptide pools. Using a repeated-measures model on
positive responses from all 5 donors, approximately one-third
fewer SFCs resulted with the use of previously cryopreserved
PBMC than fresh PBMC (P p .10). However, the length of cry-
Figure 6. Interferon (IFN)–g responses in fresh vs. cryopreserved peripheral blood mononuclear cells (PBMC). A and B, Comparison between the
no. of IFN-g spot-forming cells (SFCs) observed after stimulation with pools of Gag 20-mer peptides using fresh vs. cryopreserved PBMC obtained
from the same leukapheresis. Bars represent mean (SD) no. of SFCs in duplicate wells. C, IFN-g responses in fresh PBMC after stimulation with
pools of 10 Gag 20-mer peptides were compared with responses in cryopreserved cells thawed at multiple timepoints over a 430-day period after
leukapheresis.
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 235
opreservation did not have a significant effect on the frequency
of SFCs (P p .22).
Interoperator reproducibility and intraoperator variability. Next, we looked at the reproducibility and variability of
responses measured both within the same and across different
operators. To examine the variability in SFC frequency across
multiple assays performed by a single operator, we had one
technician perform 5 identical tests on cells from each of 2
donors. A single PBMC thaw was used for all 5 tests in which
responses to 5 different pools of 25 Gag 15-mers were measured
(data not shown). For 9 of 10 combinations of donor and
peptide pools, the 5 tests were in perfect qualitative agreement;
the 1 discordant result had 4 negative responses and 1 positive
response. Quantitative analysis of positive responses using a
random effects model revealed that only ∼1% of the variability
of the responses was attributable to the different tests—that is,
the ELISPOT procedure itself.
To assess the amount of variability introduced when multiple
operators perform the assay, we compared SFC frequencies
obtained when 5 different technicians performed identical assays in cells from each of 3 donors. Again, a single PBMC thaw
was used for each donor assay, for which responses to pools
of 25 Gag 15-mers were assessed (data not shown). The 5
operators were in perfect qualitative agreement for all 15 combinations of donor and peptide pool (7 negative and 8 positive).
Quantitative agreement between the operators was excellent,
with an analysis of positive responses using a random effects
model revealing that !0.5% of the variability of the responses
was attributable to the different operators. Thus, there was
excellent reproducibility both within and across operators.
IFN-g ELISPOT for fine mapping of HIV-1–specific T cell
responses.
IFN-g ELISPOT assays that use peptide pools
spanning multiple HIV-1 genes and a pool of optimal peptides
as stimulating antigens can be readily applied as a rapid screening tool to assess vaccine-induced T cell responses. Once a
positive response is detected, the assay can be repeated to characterize this response down to the individual 15- or 20-mer or
ideal immunodominant 8–10-mer peptide, as shown in figure
7 in patient MP0471, whose class I HLA type was A24, B*1501,
B27. In this case, responses to 1 Gag peptide (27-IYKRWIILGLNKIVRMYSPT), 1 Env peptide (74-IVELLGRRGWEVLKYWWNLL), 5 Pol peptides (33-KILEPFRKQNPDIVIYQYMD,
36-GQHRTKIEELRQHLLRWGLT, 49-IAEIQKQGQGQWTYQIYQEP, 50-QWTYQIYQEPFKNLKTGKYA, and 99-KIIRDYGKQMAGDDCVASRQ), and the immunodominant, HLA B15–restricted, Nef 9-mer were characterized. Of note, the Gag
20-mer peptide 27 includes an amino acid sequence matching
a HLA-B15 CTL epitope described elsewhere (GLNKIVRMY)
and an HLA-B27–restricted epitope (KRWIILGLNK) [27, 28],
and the Env 20-mer peptide 74 includes an HLA-B27–restricted
epitope (GRRGWEALKY) [29].
236 • JID 2003:187 (15 January) • Russell et al.
Cross-sectional analysis of HIV-1–specific IFN-g responses
in HIV-1–infected subjects.
After optimizing the IFN-g
ELISPOT assay conditions for performance using cryopreserved PBMC, frozen cells from 36 HIV-1–seropositive subjects
and 10 HIV-1–seronegative subjects were tested for both CD8⫹
and CD4⫹ T cell responses to HIV-1 Gag, Env, Pol, and Nef
using pools of 15-mer peptides (table 1). For these experiments,
anti-CD4⫹ antibody–coated immunomagnetic beads were used
to define T cell subsets. CD4⫹ responses were based on the
positively selected CD4⫹ T cell fraction, and CD8⫹ responses
were based on the CD4⫹-depleted PBMC population. The seropositive individuals included 6 untreated primary HIV-1–infected patients whose PBMC were obtained at a mean of 71
days after infection, 10 LTNPs, 12 untreated chronically HIV1–infected patients, and 8 treated chronically HIV-1–infected
subjects. The 10 HIV-1–seronegative individuals were study
participants in AVEG protocol 201, whose PBMC were drawn
at a baseline prevaccination visit. As expected, median plasma
HIV-1 RNA levels in the LTNPs and treated patients were lower
(!1500 copies/mL) than those with untreated primary and
chronic infection (110,000 copies/mL).
All but one individual tested had a CD8⫹ IFN-g response to
at least one of the HIV-1 gene products. By contrast, CD4⫹ T
cell responses were found in fewer patients (42%). With the
exception of the LTNPs, the levels of CD4⫹ T cell responses were
0.5–1.5 log lower than the CD8⫹ T cell responses in the same
patient groups (data not shown). Overall, the response in the
chronically HIV-1–infected subjects undergoing treatment was
of a greater magnitude than the responses observed in the other
groups, with a median total number of SFCs of 6006 cells/106
CD8⫹-enriched PBMC. However, the chronically HIV-1–infected
subjects who were untreated had a broader response, with a
median of 50% of the 15-mer pools recognized. The group of
primary infected subjects had a narrower and less robust response, with only 18% of the 15-mer pools recognized and a
median total number of SFCs of 866 cells/106 CD8⫹-enriched
PBMC. Of the 10 HIV-1–seronegative subjects tested, not one
demonstrated a positive response to any of the HIV-1 gene products tested. The median number of SFCs observed was 0 cells/
106 CD8⫹-enriched PBMC, and the maximum number of SFCs
observed was 4.7 cells/106 CD8⫹-enriched PBMC. On the basis
of these results, the IFN-g assay can easily distinguish positive
CD4⫹ and CD8⫹ T cell responses to multiple HIV-1 epitopes
over a broad range of SFC frequencies.
Observations from these 46 experiments were analyzed to
establish a statistical basis for a criterion for positivity in the
IFN-g ELISPOT assay. A formal statistical criterion (figure 8)
based on an exact binomial test was used to determine whether
the number of SFCs in the experimental wells of these assays
was significantly greater than in the control wells. A Bonferroni
correction was used to control for the statistical false-positive
Figure 7. Interferon (IFN)–g ELISPOT assay using peptide pools can be used to delineate epitopes across multiple human immunodeficiency
virus (HIV) type 1 genes. Cryopreserved peripheral blood mononuclear cells (PBMC) from patient MP0471 were stimulated with pools of 20-mer
peptides and individual 20-mer peptides, in addition to previously described HLA-matched 9-mer peptides from HIV-1 Gag (A), Pol (B), Env (C),
and Nef (D). For the individual 9-mers, only positive responses are shown. Bars represent the mean (SD) no. of spot-forming cells (SFCs) in
duplicate wells.
rate observed with multiple peptide pool comparisons. Two
lines are depicted in figure 8 to represent the different numbers
of peptide pools used for the experiments in HIV-1–seronegative subjects (26 pools) versus HIV-1–seropositive subjects (14
pools). The positive criterion curve was generated by comparing the total number of SFCs in the experimental and control
wells for a particular peptide pool with the proportion of the
total number of SFCs attributable to the experimental wells.
By this analysis, a volunteer is considered to be a positive responder if the proportion of SFCs attributable to at least one
peptide pool is above the curve. For the seronegative subjects
tested, all were negative responders by this criterion (figure 8).
Thus, this statistical criterion can easily distinguish true positive
responses in HIV-1–infected persons from true negative responses in HIV-1–uninfected persons.
A comparison was made at both the subject and peptide pool
level between the results obtained using this statistical criterion
and the standard criterion of twice or greater background and
⭓10 SFCs/2 ⫻ 10 5 cells after subtraction of the negative control
that is described in Materials and Methods. For 93% of the
764 peptide pools tested, the statistical criterion gave the same
result as the standard criterion. Of the 51 discordant peptide
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 237
Table 1. Summary of cross-sectional analysis of human immunodeficiency virus (HIV) type 1–specific interferon (IFN)–g–secreting CD8+ T cells in patients with acute and chronic HIV-1 infection.
Clinical parameter
CD4⫹ T cell count, median cells/mm3
Primary
a
infection
(n p 6)
Long-term
nonprogressors
(n p 10)
Untreated
chronic
infection
(n p 12)
Treated
chronic
b
infection
(n p 8)
738
702
450
565
39,610
1435
11,806
53
100
100
92
100
Pools recognized by responders only, median no. (%)
2.5 (18)
4.5 (32)
7.0 (50)
5.5 (39)
SFCs/106 CD8⫹ cells for responders only, median total
866
1732
3760
Plasma HIV-1 RNA level, median copies/mL
HIV-1–specific CD8⫹ T cell responses
Responders, %
6006
⫹
NOTE. Analyses were performed by IFN-g ELISPOT on cryopreserved peripheral blood mononuclear cells. CD8 responses
were identified by overnight stimulation with pools of peptides spanning HIV-1 Env, Gag, Pol, and Nef. SFC, Spot-forming cell.
a
Cells were obtained from primary HIV-1–infected patients at a mean of 71 days after acquisition of infection, and none were
treated at the time of analysis.
b
Chronically HIV-1–infected patients designated as treated were receiving a stable course of 2–4 antiretroviral agents, including
at least 1 protease inhibitor, at the time point measured.
pool responses, all were positive by the statistical criterion and
negative by the standard criterion, which suggests a possible
increase in the sensitivity of the assay when using the new
statistical method. Subjects were categorized as positive responders when at least one peptide pool was positive, and, at
the subject level, there was perfect agreement between the results obtained using the 2 criteria. Thus, this statistical criterion
provides a sensitive alternative to the standard approach for
the categorization of positive and negative responses by IFNg ELISPOT.
IFN-g ELISPOT response in HIV-1 vaccine recipients.
Finally, vaccine-induced responses defined by the IFN-g ELISPOT assay were examined in cryopreserved PBMC (12 years)
from HIV-1 vaccine recipients identified previously to have had
significant CD8⫹ Gag-specific lysis by chromium release assays.
Responses were evaluated in AVEG protocol 022 recipients of
a recombinant canarypox vector vaccine (vCP205) and rgp120/
SF-2 boost [30, 31]. One vaccine recipient was identified to
have had CD8⫹ Gag–specific lysis at an effector:target ratio of
50:1 of 20% and 50% from the 546- and 728-day time points
(6 and 12 months after the final vaccine boost). Using PBMC
from day 728, borderline-positive IFN-g responses were detected in the pool containing 20-mer peptides 1–10 and the
pool containing peptides 1, 11, 21, 31, and 41 (pools 1–41),
both of which contained Gag peptide 1 as their only peptide
in common (figure 9A). Therefore, by this assay, the response
was mapped to a single HIV-1 Gag 20-mer peptide (1-MGARASVLSGGELDRWEKIR). When the experiment was repeated
using CD4⫹- and CD8⫹-depleted PBMC from the earlier day
546 time point, the PBMC SFC frequency was reduced by 150%
with the depletion of CD8⫹ T cells and thus was confirmed to
be a CD8⫹ T cell response (figure 9B).
Cryopreserved PBMC from a different AVEG protocol 022
238 • JID 2003:187 (15 January) • Russell et al.
vaccine recipient obtained 3 months (day 455) after the final
vaccination were enriched for CD8⫹ T cells and tested for a
response to pools of 10 Gag 20-mer peptides. A low-level positive response of 16 SFCs/2 ⫻ 10 5 CD8⫹-enriched PBMC was
detected to Gag pool 21–30 (figure 9C). This response was
mapped to 15-mer peptide 66 (IYKRWIILGLNKIVR) by using
an overlapping matrix of pools (A–J) of 5 Gag 15-mer peptides
that corresponded to the amino acid sequences contained in
the positive 20-mer pool (figure 9D). All 15 amino acids contained in 15-mer peptide 66 are also contained in 20-mer peptide 27, a peptide found in the Gag 20-mer pool 21–30 shown
in figure 9C, and a peptide to which positive responses were
confirmed in subsequent experiments (data not shown). Thus,
these experiments indicated that the IFN-g ELISPOT assay can
detect, quantify, and determine the specificity of low-frequency
vaccine-induced CD8⫹ T cells in cryopreserved PBMC in HIV1–uninfected clinical trial participants.
DISCUSSION
The challenge to develop an efficacious HIV-1 preventative vaccine is enormous and unprecedented. Heretofore, protective immunity imparted by vaccines that curtail infections of public
health importance largely has been attributed to the induction
of antibodies that are easily measured by standardized serological
testing. The compelling evidence that CD8⫹ T cells in primates
may play a crucial role in vaccine efficacy against simian immunodeficiency virus infection [32–34] in conjunction with
strong support that T cells are vital in controlling HIV-1 infection
calls for vaccine strategies that can induce HIV-1–specific CD8⫹
T cells. Yet, determining the precise T cell responses that correlate
with vaccine effect and relying on these measurements in the
conduct of large-scale trials that will determine vaccine efficacy
Figure 8. Positive criterion analysis in human immunodeficiency virus (HIV) type 1–seropositive and –seronegative patients. Cryopreserved
peripheral blood mononuclear cells from 36 HIV-1–seropositive and 10 HIV-1–seronegative subjects were CD4⫹ depleted and tested for responses
to HIV-1 Gag, Env, Pol, and Nef by using pools of 15-mer peptides. The jagged lines are positive criterion curves with a false-positive rate of
5%. Twenty-six pools of 25 15-mer peptides were used to test the HIV-1–seronegative subjects (thick line), and 14 pools of 50 15-mer peptides
were used to test the seropositive subjects (thin line). The horizontal axis represents the total no. of spot-forming cells (SFCs) from the experimental
and the control wells for a particular peptide pool (N p E ⫹ C ). The vertical axis is the proportion of the total no. of SFCs attributable to the
experimental wells (P p E/E ⫹ C).
and licensure is unparalleled in vaccine development. In addition,
there is increasing recognition that other intracellular pathogens
and tumors may require antigen-specific T cells to prevent or
control disease. To this end, our investigation provides a framework for examining antigen-induced T cell responses in human
trials, particularly with respect to HIV-1 vaccine development,
but it is also relevant to clinical immunotherapeutic and vaccine
studies designed to protect or control other infectious agents and
neoplasms.
Our findings provide evidence that the IFN-g ELISPOT assay
is a simple and sensitive approach to measure CD8⫹ T cell
responses induced by either vaccination or natural HIV-1 infection. We demonstrate that the variability inherent in the
methodology is acceptably low and that the reproducibility
among several operators is high. Therefore, if such an approach
is instituted and standardized among various networks conducting vaccine trials, the data should be comparable across
protocols. This is important in accelerating the evaluation of
multiple vaccine strategies in phase 1–3 trials. Perhaps one of
the weakest links in assay reproducibility lies in the interpretation of the spots formed on the plates as true IFN-g–secreting
cells. In this regard, the use of a commercial image analyzer
offers an attractive alternative to visual enumeration of spots
by light microscopy. However, these units still require a considerable amount of subjective operator input to accurately
distinguish “true“ spots from background. This becomes particularly relevant in the analysis of vaccine-induced T cells from
HIV-1–uninfected persons, which may result in smaller and
less intense spots than those more easily recognized in testing
HIV-1–infected patients. Thus, consistency in the detection and
interpretation of SFCs will be necessary. In addition, to ensure
the ongoing quality control of assay reagents and the standardization of spot-counting parameters and to aid in comparability of assay results across laboratories using samples from
international sites, it will be critical to identify reagents that
can be used as positive controls. A recently described pool of
peptides whose sequences span 23 optimal CTL epitopes within
influenza virus, cytomegalovirus, and Epstein-Barr virus will
be useful to stimulate memory T cells whose IFN-g secretion
patterns can be contrasted to those induced by the vaccines.
These are recognized by ∼85% of the general US population
[35], and studies are in progress to develop similar panels of
epitopic peptides that are recognized globally. Finally, the criterion for positivity formulated here and described in more
detail in forthcoming work can be successfully applied using
data from multiple laboratories, and considerations are built
in to account for the infrequent responder with unusually high
background cytokine secretion, as well as plate-to-plate and
well-to-well variability.
Practical considerations are paramount as well in large-scale
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 239
Figure 9. Interferon (IFN)–g ELISPOT responses in CD8⫹ T cells from AIDS Vaccine Evaluation Group protocol 022 human immunodeficiency
virus type 1 vaccine recipients of vCP205 plus SF-2 rgp120 boost. A, Cryopreserved peripheral blood mononuclear cells (PBMC) obtained 12 months
after final vaccination were stimulated with a matrix of Gag 20-mer peptide pools. B, Cryopreserved PBMC from 6 months after the final vaccination
were CD8⫹ depleted and tested for a response to Gag pool 1–10 as well as the individual Gag peptide 1. C, Cryopreserved PBMC from another
vaccine recipient obtained 3 months after final vaccination were enriched for CD8⫹ T cells and tested for a response to pools of 10 Gag 20-mer
peptides. D, Positive response to Gag pool 21–30 was mapped to 15-mer peptide 66 using an overlapping 5 ⫻ 5 matrix of pools of 5 Gag 15mer peptides (depicted above the graph) that corresponded to the amino acid sequences contained in the positive 20-mer pool. Bars represent
mean (SD) no. of spot-forming cells (SFCs) in duplicate wells.
vaccine trials, and the IFN-g ELISPOT assay satisfies many
criteria for feasibility. We demonstrate excellent concordance
in the ability to detect positive responses in cryopreserved versus freshly isolated CD8⫹ T cells. Although the IFN-g SFC
frequencies observed in cryopreserved CD8⫹ T cells are, on
average, one-third lower than in fresh CD8⫹ T cells obtained
at the same venipuncture, fortunately the duration of cryopreservation does not diminish the SFC frequencies. The im-
240 • JID 2003:187 (15 January) • Russell et al.
portance of timely, meticulous processing of blood and PBMC
cryopreservation cannot be underestimated. The use of PBMC
with viabilities of at least 85% on thawing will be necessary to
consistently detect IFN-g–secreting CD8⫹ T cells, particularly
if the frequencies are low. Another advantage to emphasize with
the IFN-g ELISPOT procedure is the ability to detect responses
using pools of peptides at low concentrations (1–2 mg/mL). In
addition, with acceptable blood volumes (∼50 mL) from one
venipuncture, it is feasible over 3–4 days to screen responses
by stimulating with peptide pools and to define the CD4⫹ or
CD8⫹ epitopic responses to all HIV-1 gene products within a
8–15 aa range. Taken together, these attributes offer a tremendous advantage over the previous use of chromium release
assays in the detection of MHC-restricted CD8⫹ CTL responses.
Our results indicate that the threshold for detection of a
positive response by ELISPOT is ∼50 IFN-g SFCs/106 PBMC
(0.005%). A significant feature in improving the detection of
HIV-1–specific CD8⫹ T cells was the use of 15-mer rather than
20-mer peptides for cell activation. Our results indicate that
there was an average increase in IFN-g SFCs/106 PBMC of
150% with the use of the 15-mers, compared with that of the
20-mers. The sensitivity of the assay is within the range to
detect vaccine-induced responses, as based on the representative
findings in 2 vaccine recipients with concordant CTL responses
(figure 9) and on evidence from previous studies that memory
cells that remain after antigen clearing of infection comprise
!0.1% of PBMC [36]. The ability to identify such low-level
responses by ELISPOT provides an advantage over intracellular
cytokine staining or tetramer binding by flow cytometry, because it is unlikely that the current flow-based methods can
routinely distinguish 0.005% antigen-specific staining cells
from negative or irrelevant control staining cells. Nevertheless,
there remains the possibility that such low-frequency responses
may not be relevant for vaccine protection, because the interval
of time necessary for activation and proliferation of sufficient
numbers of effector cells may exceed the window of time in
which containment of viral replication is required for a vaccine
effect. Certainly, we demonstrate that the total HIV-specific
IFN-g–secreting cells in HIV-1–infected persons, regardless of
stage of infection, are present in frequencies ⭓1 log greater
than those induced by the canarypox vector vaccines. This
difference reflects, in part, the persistence of HIV-1 replication,
which maintains the memory population, but it remains to be
clarified whether this frequency, which provides only partial
and certainly not durable control, is relevant to that needed
for protection.
Understanding the effectiveness of CD8⫹ T cells in vaccine
protection may require information beyond the mere measurement of the frequencies of antigen-specific cells induced at peak
time intervals. The state of activation and differentiation of the
memory T cells may be key to protection invoked by the secondary response, and the effector cells must be capable of homing
to the site of active viral replication. To this end, the flow-based
cytometric methods distinguishing cytokine secretion, as well as
phenotypic characteristics, will be important to use after the
ELISPOT screening. Thus, we propose an algorithm to screen
PBMC for the recognition of HIV-1 epitopes using 15-mer peptides spanning the gene products contained in the vaccine reg-
imen. After this, the epitopes recognized by vaccine-induced T
cells can be identified using individual 15-mers within the peptide
pool, as well as with smaller optimal 8–10-mers. Additional studies to identify the CD4⫹ or CD8⫹ T cell subset mediating the
response, the MHC restricting molecule and the phenotypic
properties (memory, differentiation, activation, and homing molecular expression) can be performed by flow cytometric studies
that incorporate intracellular cytokine staining after 6 h of stimulation with the individual 15-mer or optimal peptide. In this
sequence, there is no necessity to deplete T cell subsets to distinguish the particular one mediating the response, obviating
larger cell requirements and the theoretical possibility that cytokine secretion by CD8⫹ T cells will be suboptimal if the CD4⫹
helper cells and antigen-presenting cells are removed prior to the
ELISPOT assay.
Finally, until there is a vaccine efficacy trial that substantially
induces IFN-g–secreting cells, it remains unclear whether these
are the immune cells that will afford the greatest protection.
Lytic function or secretion of other cytokines (interleukin-2 or
tumor necrosis factor–a) or chemokines may be just as pertinent. It is envisioned that many of these functions can be
routinely measured by flow cytometric analysis once the optimal epitope is identified. Moreover, no matter how high the
frequency and avidity of the T cell response, if only a narrow
response is elicited, this may increase the likelihood for viral
escape soon after infection, which may forfeit any of the early
protective effect afforded by the vaccine. In addition, induction
of T cell responses that are not cross-reactive with the epitopes
in the infecting strain may render the vaccine-induced responses ineffective. The ability to examine diverse epitopic responses throughout the HIV-1 genome, as well as across viral
subtypes, is easily accomplished with the IFN-g assay described
here, and such studies are currently in progress.
In conclusion, the IFN-g ELISPOT assay that we describe
provides a practical, sensitive, and validated instrument for the
assessment of cellular immune responses to HIV-1 immunogens in clinical vaccine trials. It is a simple assay that can be
performed using PBMC that have been cryopreserved for varying lengths of time, so that multiple time points can be examined side by side in a single experiment. The feasibility of
this approach was recently demonstrated when IFN-g ELISPOT
assays were used in a large-scale phase 2 vaccine trial. The data
from this protocol are the subject of a pending publication,
but preliminary results indicate that the ELISPOT was able to
detect low-level CD8⫹ T cell responses, in addition to providing
excellent discrimination between vaccine and placebo recipients. As we move forward to phase 3 clinical trials with more
effective immunogens, the IFN-g ELISPOT assay provides us
with an opportunity to analyze vaccine responses to establish
an immune correlate of protection from HIV-1 infection.
T Cell Responses in HIV Vaccine Trials • JID 2003:187 (15 January) • 241
Acknowledgments
We thank the volunteers for their participation in our study,
Jean Lee for arranging for the leukapheresis of our study subjects, Ya-Lin Chiu for help with data management and statistical
programming, and Alicia Cerna for assistance with manuscript
preparation.
References
1. Koup RA, Safrit JT, Cao Y, et al. Temporal association of cellular immune responses with the initial control of viremia in primary human
immunodeficiency virus type 1 syndrome. J Virol 1994; 68:4650–5.
2. Borrow P, Lewicki H, Hahn BH, Shaw GM, Oldstone MB. Virus-specific CD8⫹ cytotoxic T-lymphocyte activity associated with control of
viremia in primary human immunodeficiency virus type 1 infection.
J Virol 1994; 68:6103–10.
3. Musey L, Hughes J, Schacker T, et al. Cytotoxic-T-cell responses, viral
load, and disease progression in early human immunodeficiency virus
type 1 infection. N Engl J Med 1997; 337:1267–74.
4. Ogg GS, Jin X, Bonhoeffer S, et al. Quantitation of HIV-1–specific
cytotoxic T lymphocytes and plasma load of viral RNA. Science
1998; 279:2103–6.
5. Rowland Jones S, Sutton J, Ariyoshi K, et al. HIV-specific cytotoxic Tcells in HIV-exposed but uninfected Gambian women. Nat Med 1995;1:
59–64.
6. Goh WC, Markee J, Akridge RE, et al. Protection against human immunodeficiency virus type 1 infection in persons with repeated exposure: evidence for T cell immunity in the absence of inherited CCR5
coreceptor defects. J Infect Dis 1999; 179:548–57.
7. Mazzoli S, Trabattoni D, Lo Caputo S, et al. HIV-specific mucosal and
cellular immunity in HIV-seronegative partners of HIV-seropositive
individuals. Nat Med 1997; 3:1250–7.
8. Belshe RB, Gorse GJ, Mulligan MJ, et al. Induction of immune responses to HIV-1 by canarypox virus (ALVAC) HIV-1 and gp120 SF2 recombinant vaccines in uninfected volunteers. NIAID AIDS Vaccine
Evaluation Group. AIDS 1998; 12:2407–15.
9. Evans TG, Keefer MC, Weinhold KJ, et al. A canarypox vaccine expressing
multiple human immunodeficiency virus type 1 genes given alone or
with rgp120 elicits broad and durable CD8⫹ cytotoxic T lymphocyte
responses in seronegative volunteers. J Infect Dis 1999; 180:290–8.
10. Clements-Mann ML, Weinhold K, Matthews TJ, et al. Immune responses to human immunodeficiency virus (HIV) type 1 induced by
canarypox expressing HIV-1MN gp120, HIV-1SF2 recombinant gp120,
or both vaccines in seronegative adults. NIAID AIDS Vaccine Evaluation Group. J Infect Dis 1998; 177:1230–46.
11. Belshe RB, Stevens C, Gorse GJ, et al. Safety and immunogenicity of
a canarypox-vectored human immunodeficiency virus type 1 vaccine
with or without gp120: a phase 2 study in higher- and lower-risk
volunteers. J Infect Dis 2001; 183:1343–52.
12. Ada GL, McElrath MJ. HIV type 1 vaccine-induced cytotoxic T cell
responses: potential role in vaccine efficacy. AIDS Res Hum Retroviruses 1997; 13:205–10.
13. McElrath MJ, Siliciano RF, Weinhold KJ. HIV type 1 vaccine-induced
cytotoxic T cell responses in phase I clinical trials: detection, characterization, and quantitation. AIDS Res Hum Retroviruses 1997; 13:211–6.
14. Pialoux G, Excler JL, Riviere Y, et al. A prime-boost approach to HIV
preventive vaccine using a recombinant canarypox virus expressing
glycoprotein 160 (MN) followed by a recombinant glycoprotein 160
(MN/LAI). The AGIS Group, and l’Agence Nationale de Recherche
sur le SIDA. AIDS Res Hum Retroviruses 1995; 11:373–81.
15. Larsson M, Jin X, Ramratnam B, et al. A recombinant vaccinia virus
based ELISPOT assay detects high frequencies of Pol-specific CD8 T
cells in HIV-1–positive individuals. AIDS 1999; 13:767–77.
242 • JID 2003:187 (15 January) • Russell et al.
16. Tan LC, Gudgeon N, Annels NE, et al. A re-evaluation of the frequency
of CD8⫹ T cells specific for EBV in healthy virus carriers. J Immunol
1999; 162:1827–35.
17. Lalvani A, Brookes R, Hambleton S, et al. Rapid effector function in
CD8⫹ memory T cells. J Exp Med 1997; 186:859–65.
18. Smith JG, Liu X, Kaufhold RM, Clair J, Caulfield MJ. Development
and validation of a gamma interferon ELISpot assay for quantitation
of cellular immune responses to varicella-zoster virus. Clin Diagn Lab
Immunol 2001; 8:871–9.
19. Maeker HT, Dunn HS, Suni MA, et al. Use of overlapping peptide
mixtures as antigens for cytokine flow cytometry. J Immunol Methods
2001; 255:27–40.
20. Malhotra U, Holte S, Dutta S, et al. Role for HLA class II molecules
in HIV-1 suppression and cellular immunity following antiretroviral
treatment. J Clin Invest 2001; 107:505–17.
21. McElrath MJ, Corey L, Montefiori D, et al. A phase II study of two
HIV type 1 envelope vaccines, comparing their immunogenicity in
populations at risk for acquiring HIV type 1 infection. AIDS Vaccine
Evaluation Group. AIDS Res Hum Retroviruses 2000; 16:907–19.
22. Kern F, Surel IP, Faulhaber N, et al. Target structures of the CD8⫹–Tcell response to human cytomegalovirus: the 72-kilodalton major immediate-early protein revisited. J Virol 1999; 73:8179–84.
23. Betts MR, Casazza JP, Patterson BA, 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; 74:9144–51.
24. Agresti A. Categorical data analysis. Hoboken, NJ: John Wiley & Sons,
1990.
25. Lin L, Hedayat AS, Sinha B, Yang M. Statistical methods in assessing
agreement: models, issues, and tools. J Am Stat Assoc 2002; 97:257–70.
26. Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS system for
mixed models. Cary, NC: SAS Institute, 1996.
27. Johnson RP, Trocha A, Yang L, et al. HIV-1 Gag-specific cytotoxic T
lymphocytes recognize multiple highly conserved epitopes: fine specificity of the Gag-specific response defined by using unstimulated peripheral blood mononuclear cells and cloned effector cells. J Immunol
1991; 147:1512–21.
28. Goulder PJ, Phillips RE, Colbert RA, et al. Late escape from an immunodominant cytotoxic T-lymphocyte response associated with progression to AIDS. Nat Med 1997; 3:212–7.
29. Lieberman J, Fabry JA, Kuo MC, et al. Cytotoxic T lymphocytes from
HIV-1 seropositive individuals recognize immunodominant epitopes
in Gp160 and reverse transcriptase. J Immunol 1992; 148:2738–47.
30. Team AVEGP. Cellular and humoral immune responses to a canarypox
vaccine containing human immunodeficiency virus type 1 Env, Gag,
and Pro in combination with rgp120. J Infect Dis 2001; 183:563–70.
31. Gupta K, Hudgens M, Corey L, et al. Safety and immunogenicity of a
high-titered canarypox vaccine in combination with rgp120 in a diverse
population of HIV-1–uninfected adults: AIDS Vaccine Evaluation Group
protocol 022A. J Acquir Immune Defic Syndr 2002; 29:254–61.
32. Barouch DH, Santra S, Schmitz JE, et al. Control of viremia and prevention of clinical AIDS in rhesus monkeys by cytokine-augmented
DNA vaccination. Science 2000; 290:486–92.
33. Robinson HL, Montefiori DC, Johnson RP, et al. Neutralizing antibodyindependent containment of immunodeficiency virus challenges by
DNA priming and recombinant pox virus booster immunizations. Nat
Med 1999; 5:526–34.
34. Evans DT, O’Connor DH, Jing P, et al. Virus-specific cytotoxic Tlymphocyte responses select for amino-acid variation in simian immunodeficiency virus Env and Nef. Nat Med 1999; 5:1270–6.
35. Currier JR, Kuta EG, Turk E, et al. A panel of MHC class I restricted
viral peptides for use as a quality control for vaccine trial ELISPOT
assays. J Immunol Methods 2002; 260:157–72.
36. Callan MF, Fazou C, Yang H, et al. CD8⫹ T-cell selection, function,
and death in the primary immune response in vivo. J Clin Invest 2000;
106:1251–61.