The Journal of Immunology
The Utility and Limitations of Current Web-Available
Algorithms To Predict Peptides Recognized by CD4 T Cells in
Response to Pathogen Infection
Francisco A. Chaves,* Alvin H. Lee,* Jennifer L. Nayak,†Katherine A. Richards,* and
Andrea J. Sant*
The ability to track CD4 T cells elicited in responseto pathogen infection or vaccination is critical because of the role these cells play
in protective immunity. Coupled with advances in genome sequencing of pathogenic organisms, there is considerable appeal for
implementation of computer-based algorithms to predict peptides that bind to the class II molecules, forming the complex recog-
nized by CD4 T cells. Despite recent progress in this area, there is a paucity of data regarding the success of these algorithms in
identifying actual pathogen-derived epitopes. In this study, we sought to rigorously evaluate the performance of multiple Web-
available algorithms by comparing their predictions with our results—obtained by purely empirical methods for epitope discovery
in influenza that used overlapping peptides and cytokine ELISPOTs—for three independent class II molecules. We analyzed the
data in different ways, trying to anticipate how an investigator might use these computational tools for epitope discovery. We come
to the conclusion that currently available algorithms can indeed facilitate epitope discovery, but all shared a high degree of false-
positive and false-negative predictions. Therefore, efficiencies were low. We also found dramatic disparities among algorithms and
between predicted IC50values and true dissociation rates of peptide–MHC class II complexes. We suggest that improved success of
predictive algorithms will depend less on changes in computational methods or increased data sets and more on changes in
parameters used to “train” the algorithms that factor in elements of T cell repertoire and peptide acquisition by class II
molecules.The Journal of Immunology, 2012, 188: 4235–4248.
CD4 T cells play in immunity to complex pathogens. Further
success in identification of the peptides that are the focus of an
adaptive CD4 T cell response is essential for understanding the
mechanisms of protective immunity and the factors that influence
the dynamics and specificity of host–pathogen interactions. CD4
T cell epitope identification is also needed for vaccine evaluation,
tetramer-based studies of T cell phenotype, and for development
of peptide-based vaccines. With increasing success in genome se-
quencing of complex bacterial and viral pathogens (reviewed in
Refs. 1–5), candidate proteins for vaccines are increasing, but
identification of epitopes that are the focus of immune responses
remains a bottleneck in this research.
D4 T cells are known to play a key role in protective
immunity to infectious organisms, and much current re-
search uses epitope-specific probes to study the role that
A number of empirical approaches have historically been used
for epitope discovery—including biochemical isolation and pro-
teolytic fragmentation of antigenic proteins (6, 7), derivation of
genetic constructs that encode all or selected segments of candi-
date pathogen-derived proteins (8–11), elution and sequencing of
peptides from pathogen-infected cells or tumor cells (12–16), and
individual epitope mapping—using arrays of synthetic peptides
(17–22). These approaches, typically coupled with T cell assays to
identify the immunologically active peptide within the candidate
Ag, are time consuming and involve significant expenditure of
effort and resources to be successful. The labor-intensive nature
of these approaches is a particularly large obstacle for complex
pathogens that express hundreds of proteins, of which only a small
fraction may be the target of T cells or B cells or which may serve
a protective role as vaccine candidates.
The considerations of time and expense required for empirical
approaches have led to the development and refinement of algo-
rithms that use different logic bases and sources of data to predict
epitopes that will be presented by particular MHC molecules
(reviewed in Refs. 23–28). Because the major selective force in
peptide binding to MHC involves side chains of amino acids
(“anchors”) in the peptide with depressions (“pockets”) in the
MHC molecule, the algorithms focus on scoring these interac-
tions as a means to predict CD4 epitopes. Some methods such as
matrix-based algorithms operate with the general model that each
amino acid adds or detracts from the binding of the peptide to the
MHC protein in a largely predictable, independent, and quantifi-
able manner (29, 30). Large data sets, or “training data,” are used
to construct and refine the algorithms that ultimately search for the
highest 9-mer core in a peptide and output the predicted binding
affinity of every candidate peptide. Other less rigid algorithms that
operate using such methods as neural networks (31, 32) and par-
ticle swarm optimization (33) have also been developed and used.
*Department of Microbiology and Immunology, David H. Smith Center for Vaccine
Biology and Immunology, University of Rochester Medical Center, Rochester,
NY 14642; and†Department of Pediatrics, University of Rochester Medical Center,
Rochester, NY 14642
Received for publication December 15, 2011. Accepted for publication February 29,
and R01AI51542 (to A.J.S.). J.L.N. was supported by Grant 1K12HD068373-01 from
the National Institutes of Health.
Address correspondence and reprint requests to Prof. Andrea J. Sant, David H. Smith
Center for Vaccine Biology and Immunology, University of Rochester Medical
Center, 601 Elmwood Avenue, Box 609, Rochester, NY 14642. E-mail address:
The online version of this article contains supplemental material.
Abbreviations used in this article: AUC, area under the curve; DDM, n-dodecyl
maltoside; HA, hemagglutinin; HEL, hen egg lysozyme; M1, matrix protein 1;
NA, neuraminidase; NP, nucleocapsid protein; NS1, nonstructural protein 1; ROC,
receiver operating characteristic.
Finally, Sette and coworkers (34) describe a “Consensus” ap-
proach that essentially averages the predicted ranking hierarchy of
a given set of peptides scored by what their studies suggest to be
the best-performing three to four Web-available algorithms.
In general, the predictive algorithms developed for MHC class I
peptides that activate CD8 T cells significantly outperform those
in large part because of the nature of their respective peptide
binding pockets. MHC class I molecules are closed at their pe-
riphery, thus limiting the size of the peptide that binds to 8–10 aa
(35–37). Therefore, the amino acids that contribute the key anchor
positions for pocket interactions are easily identified. In contrast,
the binding pocket of MHC class II is open at its periphery.
Elution and sequencing studies indicate that peptides bound by
class II molecules typically range from 9 to 25 aa in length (38,
39), and long peptides are well presented to CD4 T cells (40, 41).
Often, the “register” of these peptides, the amino acids that
comprise the 9-aa core within the MHC-binding groove and that
dictate MHC binding and T cell recognition, is not known. For
several peptides analyzed in detail, multiple registers are pre-
sented by class II (42–45). Therefore, simple knowledge that
a peptide is presented by a class II molecule does not provide
insight into what amino acids control its presentation. Because
much of the data that train the algorithms are long peptides, it has
been challenging to know how identified antigenic peptides bind
to the class II molecule and thus to predict what peptides within an
uncharacterized set will be presented by the same or related class
II molecules. Databases containing epitopes presented by class II
molecules have steadily accumulated (46–48) and been used to
refine existing algorithms, but the issue of uncertain binding cores
Because CD4 T cells are known to be critical regulators of both
B cell and CD8 T cell responses to pathogens, to provide direct
function for responses against intracellular pathogens, and to be
critical for vaccine success, it is clear that despite the theoretical
challenges, there is considerable appeal in using computation-
based algorithms to identify candidate CD4 T cell epitopes, par-
ticularly for complex pathogenic organisms that express hundreds
of proteins. “Benchmark” studies that assessed the accuracy of the
predictions typically measured the ability of the algorithms to
predict binding to MHC class II molecules using existing or new
data sets (25, 29, 32–34, 49–51). Evaluation of peptide perfor-
mance using functional tests of the actual immunogenicity of the
predicted peptides have been much more limited. The studies
performed are typically restricted to a handful of peptides, are
sometimes tested by immunization with free peptides rather than
a complex Ag, and, often, the presenting class II molecule for
CD4 T cell recognition is not unequivocally identified (20, 34, 52–
55). In recent years, our laboratory has empirically and compre-
hensively investigated the peptide specificity of CD4 T cells eli-
cited in response to primary influenza infection using a completely
unbiased approach involving overlapping peptide libraries and
cytokine ELISPOT assays. Multiple strains of mice have been
studied, including HLA-DR transgenic mice and common inbred
strains of mice (21, 22, 56). We have identified and quantified the
responses to .500 influenza-derived peptides presented by these
different class II molecules that are the focus of CD4 T cell
In this study, we have evaluated the ability of Web-available
algorithms to predict the specificity of CD4 T cells elicited in
response to influenza. We had three goals for this study. The first
was to evaluate the performance of the algorithms for their effi-
ciency in epitope identification by combining our results in epitope
discovery with advances by other groups in developing predictive
algorithms. The second goal was to develop useful strategies to
implement algorithms for epitope discovery for future investi-
gations of the role that CD4 T cells play in protective immunity to
to improve performance of the algorithms for future efforts to
facilitate epitope discovery. To our knowledge, the analyses pre-
sented in this study are the first to comprehensively evaluate the
performance of algorithms in comparison with the results of
empirical and nonbiased epitope discovery of multiple pathogen-
encoded Ags and unrelated class II molecules.
Materials and Methods
The HLA-DR1 and the HLA-DR4 transgenic mice were obtained from
D. Zaller (Merck) through Taconic Laboratories and were maintained in the
specific pathogen-free facility at the University of Rochester according to
institutional guidelines. C57BL/10 mice were purchased from The Jackson
Laboratory (Bar Harbor, ME). Mice were used at 2–5 mo of age.
All animal protocols used in this study adhere to the Association for As-
sessment and Accreditation of Laboratory Animal Care International, the
Animals Guide. All protocols have been approved by the University of
Rochester Committee on Animal Resources (animal welfare assurance no.
A3291-01). The protocols under which these studies were conducted were
approved on March 4, 2006 (protocol no. 2006-030), and April 10, 2008
(protocol no. 2008-023).
Influenza infection of mice
A/New Caledonia/20/99 was produced in allantoic fluid of embryonated
eggs. Briefly, eggs were purchased from SPAFAS (North Franklin, CT) and
incubated at 70˚F and 100% humidity for 9 d, followed by the infection
of the allantoic cavity with 100 ml human (H1N1) influenza virus A/New
Caledonia/20/99 (kindly provided by Dr. John Treanor, University of
Rochester). Virus was collected from the allantoic fluid, and the EID50for
the produced virus was determined. HLA-DR1, HLA-DR4, transgenic, and
C57BL/10 (I-Ab) mice were infected intranasally with A/New Caledonia/
20/99 at 50,000 EID50in 30 ml PBS. Groups of mice were 2–5 mo old at
the time of infection and were anesthetized by intraperitoneal injection
with tribromoethanol prior to infection. Ten to fourteen days postinfection,
the mice were euthanized, and spleens and mediastinal lymph nodes were
isolated and used as sources of CD4 T cells for ELISPOT analyses. Lym-
phocytes were pooled from four to eight mice and depleted of B cells, CD8
cells, and macrophages by negative selection using MACS depletion
(Miltenyi Biotech, Bergisch Gladbach, Germany) according to the manu-
The dissociation experiments were performed using the procedure as
previously described (57–60). Briefly, 50 ml 100 nM purified soluble DR1
molecules expressed in S2 cells were incubated with 1 ml 250 mM
fluorescein-labeled binding peptides at a concentration of 100 nM and
pH 5.3 in McIlvaine’s buffer containing 0.025% NaN3. The binding mix-
ture was incubated at 37˚C during 2–24 h, and then the class II–peptide
complexes were separated from excess labeled peptides at room temper-
ature using microspin columns (Bio-Rad, Hercules, CA). Unlabeled in-
fluenza hemagglutinin (HA) [306–318] (PKYVKQNTLKLAT) peptide
was added at final concentration of 5 mM to prevent possible rebinding of
the dissociating peptide. Soluble I-Admolecules were produced and iso-
lated as previously described (58–60). Briefly, a total of 1 3 1010I-Ad
phosphatidyl inositol-linked CHO cells were solubilized with 1.5 l of 50
mM Tris, 150 mM NaCl, 1 mM n-dodecyl maltoside (DDM), and 0.025%
NaN3 containing protease inhibitors. Class II molecules were isolated
using Ab affinity chromatography. After phosphatidyl inositol cleavage,
the soluble I-Adwas eluted from the column at pH 11, and fractions
containing class II were pooled, dialyzed against PBS (pH 7.4) containing
0.2 mM DDM and 0.025% NaN3(DDM/PBS), and concentrated with
a Centriprep YM-10 device (Millipore, Bedford, MA). Fifty microliters of
100 nM purified soluble I-Admolecules were incubated with 1 ml 250 mM
fluorescein-labeled binding peptides at a concentration of 100 nM and pH
5.3 in McIlvaine’s buffer containing 0.025% NaN3. The complexes were
isolated using microspin columns as described earlier for DR1. Dissocia-
4236PREDICTION OF CD4 T CELL EPITOPES
tion assays were similarly performed by incubating the purified MHC class
II–peptide complexes at 37˚C in McIlvaine’s buffer plus 0.025% NaN3at a
final pH of 5.3 in the presence of either unlabeled influenza HA [306–318]
(PKYVKQNTLKLAT) for DR1 or Ea [52–68] (ASFEAQGALANIAV-
DKK) for I-Adto prevent possible rebinding of the dissociating peptide.
The dissociating peptide was separated from the remaining complex with
a BioSep SEC-S 3000 column (Phenomenex, Torrance, CA) using a Shi-
madzu chromatograph (Shimadzu Corporation, Columbia, MD). The in-
tensity of the remaining class II–peptide complex was determined using
a fluorescence detector (at wavelengths of 495-nm excitation and 525-nm
emission; Shimadzu). The intensity of the peak belonging to the complex
decreases over time, allowing quantification of the dissociation t1/2. The
graph of the intensity of the MHC class II–peptide complex versus the
dissociation time is used to generate the dissociation curve. Baselines
values were measured before each dissociation experiment to ascertain that
there were no traces of previous samples, and the program for elution of
the complexes by HPLC–SEC includes a wash time that ensures complete
removal of all previous materials. Dissociation curves fit to single expo-
nential decay curves with r2. 0.9. The half-life is the time at which the
initial fluorescence intensity of the complexes decays to half, which is
stated as the t1/2dissociation, proportional to the stability of that complex.
17-mer peptides overlapping by 11 aa to cover the entire sequences of the
HA and neuraminidase (NA) proteins from the A/New Caledonia/20/99
influenza virus (H1N1), the nonstructural protein 1 (NS1) sequence from
the A/New York/444/2001 influenza virus (H1N1), and the nucleocapsid
protein (NP) and matrix protein 1 (M1) sequences from A/New York/348/
2003 influenza virus (H1N1) were used. Peptide arrays were obtained from
the National Institutes of Health Biodefense and Emerging Infections
conserved, and the A/New York/444/2001 NS1 aa sequence is .99%
conserved compared with A/New Caledonia/20/99 (H1N1) influenza virus.
The peptides were reconstituted to 10 mM in PBS, with or without added
DMSO for hydrophobic peptides, and 1 mM DTT for cysteine-containing
peptides. Working solutions were prepared in DMEM (Invitrogen, Carls-
bad, CA), filter sterilized, and stored at 220˚C. The final concentration of
the individual peptides used in the ELISPOT assays was 10 mM.
Use of Web-based MHC class II–peptide binding prediction
The primary sequences for HA, NA, NP, NS1, and M1 from A/New
Caledonia/20/99 (H1N1) were obtained from the Protein Data Bank. The
sequences were run using the Web-available predictors NetMHCII 2.2,
NetMHCII-pan 2.1, and SMM-align at http://www.cbs.dtu.dk/services/ and
Consensus, ARB, and Tepitope at http://tools.immuneepitope.org/main/. If
only 15-mer peptides were scored by algorithms (Consensus, ARB, and
Tepitope), the empirically tested peptides were matched to the functionally
assayed 17-mer sequences by alignment with their N termini. The raw
scores, the IC50, or the percentile rank values were used to determine the
peptides that were selected by the algorithms. For methods that report IC50
values such as NetMHCII 2.2 (referred to as NetMHCII hereafter),
NetMHCII-pan 2.1 (referred to as NetMHCIIpan hereafter), SMM-align
(referred to as SMM hereafter), and ARB, the predictors qualify a peptide
as follows: 1) nonbinders for peptides with IC50.500 nM; 2) weak
binders for peptides with IC50between 50 and 500 nM; and 3) strong
binders for peptides with IC50,50 nM. These specific cutoffs were used to
analyze predictive performance in some experiments based on their use by
the developers or users of predictive algorithms (49, 61–64). For Con-
sensus, the percentile rank was used to rank the peptides, whereas for
Tepitope, the raw score was used for ranking. When a given functionally
evaluated peptide was penalized because its predicted binding core was at
either the N or C terminus of the scored peptide epitope, we evaluated each
9-aa core within the adjacent scored peptides having the same core as the
epitope and assigned the score that had the best binding value. When
scoring the selected viral “proteome,” the input sequences were the entire
amino acid sequences of the five proteins joined together, in the order HA,
NA, NP, NS1, and M1, with nonnative peptides at the junctions of the
proteins eliminated from consideration. Only the experimentally tested
peptides were scored for analysis.
Derivation of receiveroperating characteristic curves and area
under the curve scores
Algorithm prediction accuracy was measured via receiver operating char-
acteristic (ROC) curve, and the area under the curve (AUC) values were
calculated as described by others (23–25, 32, 34). Each algorithm’s binding
predictions were compared with the empirically defined set of “epitopes”
and categorized under a binary classification system. At each discrimina-
tion threshold, peptides were grouped into one of four categories: true
positive (TP), predicted binders as defined by the threshold that elicited
response in T cell assays; false positive (FP), predicted binders as defined
by current threshold that did not elicit response in T cell assays; true
negative (TN), peptides predicted relative to threshold that did not elicit
response in T cell assays; and false negative (FN), when peptides predicted
as nonbinding relative to threshold elicited response in T cell assays. The
true-positive rate (TPR) [TPR = TP/(TP + FN)] was plotted against the
false-positive rate (FPR) [FPR = FP/(FP + TN)] for each threshold to form
the ROC curve. The area under each algorithm’s curve was calculated to
give the AUC score.
CD4 T cell epitope discovery
Peptide-specific cytokine ELISPOT assays were performed as previously
described (65) and as represented schematically in Fig. 1. Briefly, 96-well
filter plates (Millipore, Billerica, MA) were coated with 2 mg/ml purified
rat anti-mouse IL-2 (clone JES6-1A12; BD Biosciences, San Jose, CA) in
PBS, washed, and incubated with media to block nonspecific binding. CD4
T cells isolated and purified from previously infected mice (at 100,000 to
300,000 cells per well) were cocultured with fibroblasts expressing the
DR1 or APCs isolated from syngeneic mice and the test peptide at 10 mM.
After overnight coculture, plates were processed to visualize IL-2–pro-
ducing cells as described (21, 22, 56), and cytokine ELISPOTs were
enumerated with an Immunospot reader series 2A using Immunospot
software version 2.
Identification of CD4 T cell epitopes through an entirely
Recently, our laboratory has identified the full repertoire of in-
fluenza-specific CD4 epitopes from five major influenza proteins in
several inbred strains of mice that express distinct class II mole-
cules [SJL, C57BL/10, HLA-DR1 (“DR1”) transgenic mice] by
using a completely empirical approach (21, 22, 65). HLA-DR4
(“DR4”) epitopes were additionally identified for this study. Mice
were infected by intranasal inhalation, which represents the nat-
ural route of infection for this pathogen in humans. At 10–14 d
postinfection, CD4 T cells were isolated from lymphoid tissue,
purified, and assessed directly for their peptide specificity using
synthetic peptides and cytokine ELISPOT assays, allowing direct
ex vivo quantification of Ag or peptide-reactive lymphocytes. CD4
T cell specificity was identified from a sequential, iterative method
of epitope identification using overlapping synthetic 17-mer pep-
tides representing the entire translated sequence of these five viral
proteins, as schematically illustrated in Fig. 1. These selected viral
proteins that were examined (HA, NA, NP, M1, and NS1) differ
with regard to their expression within the influenza virion and in
infected cells, allowing us to evaluate whether these variables
influenced epitope selection or algorithm performance. HA and
NA are expressed as transmembrane proteins in the plasma
membrane of infected cells and in the virion envelope, NP is lo-
calized within the cytosol and nucleus, and M1 is associated with
the inner leaflet of the plasma membrane and is abundant in the
virion. NS1 is notable because it is excluded from the virion
For epitope discovery, the approach depended on the number of
candidate peptides. For small viral proteins represented by ,50
different peptides (M1 and NS1), each peptide was tested indi-
vidually in ELISPOT assays. For larger proteins (HA, NA, and
NP), we used the peptide pooling matrix strategy outlined on the
right in Fig. 1, adapted by us (21, 22, 65) from an approach
originally described by Tobery and coworkers (19, 66). Three
different alleles (HLA-DR1, HLA-DR4, and I-Ab) were the focus
of the current study because they have been studied extensively
and thus we imagined that the performance of the algorithms
The Journal of Immunology4237
would be highest for these. The individual peptide epitopes rec-
ognized by the CD4 T cells from infected mice are listed in
Supplemental Table I, with the responses adjusted and presented to
represent cytokine ELISPOTs elicited by each peptide per million
CD4 T cells tested. Peptides were considered positive if they re-
producibly recruited .30 cytokine-producing cells per million
CD4 T cells, and data presented represents the average value from
at least three and typically more than five independent assays.
Comparison of algorithm predictions versus the empirically
To analyze the relationship between predicted MHC class II
binding peptide and experimentally identified epitopes, we scored
the 17-mer peptides used for empirical identification of epitopes
using the Web-available algorithms. We chose algorithms that
integrate much of the available information regarding peptide
binding to class II molecules and limited our studies to Web-
accessible algorithms. In the analyses described in this study, we
evaluated ARB, SMM, Tepitope, NetMHCIIpan, NetMHCII 2.2,
and the Consensus method. The “score” of each peptide is pro-
vided in different formats for the different algorithms. Some of
the algorithms, such as ARB, NN-align, and SMM (29, 30, 32),
calculate a “predicted IC50” value for each peptide scanned and
recommend a cutoff for strong, weak, and nonbinding peptide,
typically in the range 10–500 nM. Other algorithms, such as
Tepitope (67) and Consensus (34), rank the tested peptides from
which an arbitrary cutoff can be chosen, without any estimate of
absolute affinity. The Consensus method ranks the peptides based
on assimilation of the prediction scores from what their studies
have revealed to be the most accurate algorithms. The choice of
how to analyze the performance of the algorithms in the current
study was driven both by our desire to objectively analyze their
accuracy and also by anticipation of the way these tools might be
used by other investigators.
We began the analyses using the Consensus and the NetMHCII
methods, both of which have unique strengths and have been
widely used or evaluated by others (25, 29, 49, 50, 62, 68–71). The
Consensus method calculates the median percentile rank of each
peptide based on the predictions of three independent, high-
performing algorithms. Because the scoring output is a percentile
rank, there is no absolute value of affinity implied, and candidate
epitopes are identified on inclusion within the top scorers. The
recently developed NetMHCII (32) corrects for biases due to rep-
licate binding cores within the training data and is one of the few
algorithms that factors in effects of the amino acids that flank the 9-
aa core of the peptides. It also has added flexibility in that it allows
evaluation of peptides of different lengths. This method provides
output in the form of predicted IC50values. Both algorithms are
able to predict epitopes for the three alleles of MHC class II
molecules studied here, and both were found by their originators
to outperform other individual state-of-the-art algorithms.
To globally visualize the predicted binding versus the actual
hierarchy of CD4 T cell responses to influenza, the number of
cytokine ELISPOTs elicited by each peptide tested within the five
viral proteins (the “proteome”) were plotted versus the predicted
affinity (NetMHCII) or percentile rank (Consensus). Predicted
values are represented as their reciprocal, so that the height of the
epitope on the y-axis represents a peptide’s actual or predicted
“strength.” The CD4 cytokine ELISPOT data are represented as
spots per million for every peptide analyzed, with an arbitrary
cutoff of 100 spots per million CD4 T cells indicated, as this might
be a minimal frequency (0.01%) useful for such methods at tet-
ramer-based flow cytometry or intracytoplasmic cytokine staining.
These graphs, shown in Fig. 2A, are thus divided into four quad-
rants, where the top-right quadrant represents the “double-posi-
tive” peptides—those peptides both predicted by the algorithm and
that were also true CD4 T cell epitopes, as detected by cytokine
ELISPOT assays. The bottom-left quadrant represents the double-
negative peptides, those neither predicted nor observed. The top-
left quadrant represents the “false-positive” peptides, those pre-
dicted to be above the threshold used, and the bottom-right
quadrant represents the false negatives, those peptides found to
be epitopes in the influenza-specific CD4 T cell response but not
predicted to be high-affinity binders. The number of peptides that
fell into each quadrant for each allele is tabulated in Fig. 2C,
which also indicates the fraction of peptides in each category. Also
quantified in Fig. 2C are minor epitopes, those that recruit low but
detectable numbers of cells .30 but ,100 cells per million. All
regions scored are illustrated schematically in Fig. 2B.
These analyses indicated that both the number and percentage
of each candidate peptide that is in the “double-positive” quadrant
varies significantly with the allele and algorithm. For example,
DR1 has the highest fraction of double-positive peptides chosen
by NetMHCII (16 of 30 total positives, or 53%), most notably the
four highly dominant peptides, recruiting .400 cells per million
CD4 T cells and estimated to be of very high affinity. These true-
proach followed for epitope discovery. To limit the
number of peptides tested, a matrix array was used for
HA, NA, and NP proteins where peptides were pooled
into 8–12 peptides and then arrayed into intersecting
rows and columns as described (27, 29). Only pep-
tides within pools that were positive in both columns
and rows in the matrix, eliciting a cytokine response
$30 spots per 106cells, were considered positive. All
the positive peptides from the matrix were further
tested individually. In the case of M1 and NS1 pro-
teins, which comprised ,50 peptides, all peptides
were tested individually.
Schematic of the experimental ap-
4238PREDICTION OF CD4 T CELL EPITOPES
positive peptides were offset by a total of 61 false-positive pep-
tides. DR4 had only one peptide that was both predicted and found
empirically, whereas I-Abhas none. The range for Consensus
ranged from 25–41% true positives, depending on the allele, and
between 19 and 27 peptides predicted but not found, with DR1
having three peptides that were highly ranked and found to be
immunodominant. In all cases, although the algorithms selected
some epitopes, there were many epitopes that were not predicted
to be MHC binding peptides by either algorithm and many pep-
tides that were predicted to be binders that were not CD4 epitopes.
Because some cellular proteins have preferential access to the
MHC class II presentation pathway (39, 72–74), we analyzed the
individual peptide epitopes for each viral protein separately to
determine if algorithm performance varied among the proteins
examined. Supplemental Figs. 1 and 2 show the results of these
analyses for the three alleles analyzed and the predictions made by
NetMHCII and Consensus. Although there were not striking dis-
parities between predicted versus true epitopes, two rather con-
sistent trends were noted. First, NetMHCII predictions for NS1,
M1, and NP appeared to have a somewhat higher “hit rate” across
alleles and fewer false positives. For example, for DR1, 11 NS1-
derived peptides were predicted to be high-affinity binders, and six
of these were true epitopes. Second, we noted that epitopes within
HA seemed to have an opposite pattern, where 30 peptides were
predicted to be high-affinity binders to this allele, but only eight of
these were true epitopes. Supplemental Fig. 2 shows a similar type
of analysis performed using the Consensus method. By the nature
of the scoring for this algorithm, all protein/allele combinations
will have similar numbers of peptides that fall above and below
any cutoff, independently of whether they are truly high-affinity
binders. Using a 10% cutoff as a “prescreen” would have indeed
allowed identification of some of the major epitopes that were
recognized by CD4 T cells after influenza infection. In the case of
DR1 of the 25 major epitopes, only six were identified to be in the
top 10% rank, two in NS1 and two in M1. For DR4, of the eight
major epitopes, only three were identified as strong binders, and
they were all in NP. For I-Ab, there were 11 immunodominant
epitopes, and of these, three epitopes were within the top 10%
predicted and two were in NP.
The relationship between cutoff choices for selection of peptides
compared with the yield of actual epitopes for both algorithms is
most readily seen in Fig. 3, which represents each peptide con-
tained in the tested proteome plotted in order of decreasing pre-
dicted binding affinity or rank by NetMHCII and Consensus,
respectively, with two potential cutoffs shown, corresponding to
IC50values of 50 or 500 nM for NetMHCII and 10 or 20% pre-
dicted top binders for Consensus. From this analysis, several
conclusions were made. First, the epitopes identified empirically
through functional studies are mostly contained within the top
50% of the 332 peptides scored. This indicates that the actual
epitopes recognized by CD4 T cells are enriched in the high-
scoring group of peptides. Second, although there is clustering
of epitopes in this top half of the peptides, the most robust CD4
T cell specificities are not clustered within the highest of the
predicted affinities. This is particularly apparent for the human
DR4 molecule. Thus, although this type of display allows use of
algorithms to select peptides, the “cutoff” for selection of peptides
for further confirmation needs to be generous if it is hoped to
capture the most dominant of epitopes with the candidate peptides.
Finally, using NetMHCII, the ranking of affinity versus response is
for CD4 T cells isolated from previously infected HLA-DR1, HLA-DR4, or C57BL/10 mice were compared for their response magnitude (represented as
cytokines spots/million on the x-axis) versus their predicted affinity (IC50values, NetMHCII 2.2) or ranking (Consensus), indicated on the y-axis. The
inverse value of the affinity is used for NetMHCII 2.2 so that greater values represent increasing predicted affinity of the peptide for MHC class II. The
threshold indicated by the vertical dotted lines in each panel in (A) represents major epitopes (eliciting .100 spots per million). (B) The grid represents the
areas scored in (A). Area 1 includes peptides predicted but not found, area 3 represents peptides both predicted and found, area 4 represents peptides neither
predicted nor found, and area 6 represents peptides that were not predicted to bind but that were found. Areas 2 and 5 represent the minor epitopes that
recruit .30 but ,100 spots per million CD4 T cells. The values in (C) quantify the number of peptides in each area shown in (B).
Accuracy of NetMHCII 2.2 and Consensus for prediction of influenza-specific CD4 Tepitopes. (A) All of the peptides analyzed functionally
The Journal of Immunology 4239
different for each allele, so selection of a consistently appropriate
cutoff may be difficult. For example, for DR1, the use of a 50 nM
predicted affinity cutoff would select ∼80 of 332 peptides and
would have successfully identified 30% of the true epitopes, in-
cluding the three most immunodominant. The same predicted
affinity for DR4 would select for ,15 peptides and only four
epitopes, none of which are the most immunodominant. For the
DR4 allele, the 500 nM cutoff would be more effective. For I-Ab,
the 50 nM cutoff would select only two peptides, of which none
of the 11 major epitopes would be found, and here the lower
threshold of 500 nM would be more effective. Thus, the algorithms
do not perform equivalently for all alleles, in terms of epitopes
“captured” at any given cutoff. The bottom panel of Fig. 3 shows
Consensus prediction, where the peptides are simply ranked.
Consensus was the most successful for I-Abwhere most of the
peptide epitopes would have been identified with the 20% cutoff
and many identified even with the 10% cutoff. Finally, although
the identified epitopes are generally within the top third, even the
lowest ranked peptides contain some epitopes. In conclusion, both
methods facilitated identification of epitopes, and if only a few
epitopes are needed, a very stringent cutoff will select a subset of
peptides that contains at least a few epitopes, but many epitopes,
including major immunodominant epitopes, will be missed.
Evaluation of the relationship between predicted affinity and
We next extended our analyses to additional algorithms. We had
twogoals:first toassess theagreementamongalgorithms witheach
other and second to examine the relationship between immuno-
dominance and the affinity estimates generated by the algorithms.
Shown in Fig. 4 is this analysis performed with several algorithms
that present estimated affinity in the form of IC50, each of which is
represented by a different symbol. Only SMM, NetMHCII, and
ARB incorporate I-Ab, so the analysis of this MHC class II
molecule was restricted to these three algorithms. For simplicity,
we only included peptide epitopes that recruited at least 50 spots
per million CD4 T cells. For several epitopes, there is good
agreement among all of the algorithms for predicted binding, and
these estimates could allow selection of epitopes. For example, for
DR1, among the 32 dominant peptide epitopes, 20 were predicted
by at least three algorithms to possess high-affinity binding, and
three highly dominant M1 peptides and one NS1 peptide were
predicted to be high affinity by all four algorithms, suggesting that
some peptides have features that are well recognized to promote
binding to this allelic form of human class II.
For DR4 and I-Ab, the relationship between predicted IC50
among algorithms versus epitope dominance was much weaker.
Among the eight major epitopes restricted by DR4 that recruited
at least 100 CD4 T cells per million, two were predicted to bind
with high affinity by more than two algorithms, whereas for I-Ab,
of the 12 major epitopes, only two were selected by at least two
algorithms to have high-affinity binding. This analysis also re-
vealed that a number of epitopes that were identified empirically
were not predicted to be MHC binding by any of the algorithms.
For DR1, 9 of 32 were judged to be nonbinders by at least three
algorithms, for DR4, 4 of 8 were judged to be nonbinders, and for
based on the indicated algorithms, for each of the experimentally identified epitopes. The x-axis shows the relative position of every epitope among the total
of 332 synthetic peptides that span the entire HA, NA, NP, NS1, and M1 proteins. Shown for each peptide is the magnitude of the cytokine response elicited
in recall experiments. The experimentally tested peptides were placed in the hierarchy for each algorithm calculated either as [(1/IC50) 3 100] for
NetMHCII or percentage in the ranking of all peptides in Consensus. Vertical dotted lines within each graph indicate the cutoffs of 50 or 500 nM IC50
values for NetMHCII or 10 or 20% best predicted binding peptides for Consensus.
Ranking of experimentally identified epitopes using the NetMHCII and Consensus algorithms. Representation of the predicted ranking,
4240 PREDICTION OF CD4 T CELL EPITOPES
I-Ab, the vast majority of epitopes were predicted to be poor
binders or nonbinders. There was also a significant range in the
IC50values predicted for individual peptides, as evidenced by the
high degree of spread among the different symbols in Fig. 4.
Analysis of the efficiency of predictive algorithms for response
Because there are many false-positive peptides predicted by each
algorithm, we concluded that secondary screens are important and
next sought to evaluate different thresholds that might be used for
selection of candidate peptides. Of particular interest was the cost
(how many peptides would be tested in secondary screens) versus
benefit (how many epitopes found) of different thresholds. In this
analysis, we selected the subset of algorithms that provides output
in the form of predicted affinity (SMM, NetMHCIIpan, NetMHCII,
and ARB). Fig. 5A shows the fraction of the response predicted or
what would have been “captured” by preselection of epitopes by
the indicated algorithm, at either an IC50value of 50 or 500 nM.
The cutoffs were chosen based on common use in the literature
by those who derive or use the predictive algorithms (49, 61–
64). Although these two values are somewhat arbitrary and other
cutoffs, such as 5 nM for very high affinity or 1000 nM for very
low affinity, could be used, the two cutoffs selected provide
a framework that can be used to compare the predictive efficacy of
the individual algorithms. The number of cytokine-producing cells
elicited by all of the peptides identified by an algorithm were
summed and then divided by the total number of cytokine-
producing CD4 T cells response identified empirically. One can
consider this to be the “benefit” of using the algorithms to select
peptides for further study. Fig. 5B presents the number of peptides
that were selected by each algorithm at the two cutoff values and
thus the number that would need to be made if all of the peptides
predicted to be in that group of binders were tested in functional
Use of a 50 nM IC50value of predicted affinity did capture at
least 30% of the response for DR1, with increasing performance
using SMM , NetMHCIIpan , NetMHCII , ARB. However,
the apparent greater performance came at the “cost” of increasing
numbers of peptides to be tested (Fig. 5B, open bars) for these
algorithms, based on additional peptides estimated to be better
than the 50 nM IC50value. For the less well studied DR4 mole-
cule, ARB outperformed the other algorithms and would have
allowed identification of 40% of the response using the 50 nM
from different algorithms for influ-
enza-specific CD4 T cell epitopes.
Shown by bars for each allele is the
magnitude of the response recalled
by the indicated peptide, represented
as cytokine spots per million CD4
T cells on the right-hand y-axis.
Shown by the symbols are the affinity
predictions from each algorithm for
the peptides relevant to the indicated
allele, with the axis shown in the left-
hand y-axis. The black line indicates
the threshold of 50 nM, where all
symbols above that line are peptides
predicted to bind with high affinity.
The solid red line indicates the thresh-
old of 100 spots/million for dominant
specificity using different affinity cutoffs. The cytokine ELISPOTs for all
epitopes recognized by CD4 T cells in each strain were summed to rep-
resent the total response. Predictions of MHC class II binding peptides
were calculated by each algorithm, and the epitopes that were success-
fully predicted at either 50 nM (open bars) or 500 nM (filled bars) IC50
values were identified. (A) Responses to these two subsets of peptides
were summed and divided by the total response elicited by all of the
peptides to yield “% predicted,” the fraction of the response that would
have been identified by using only the algorithm-selected peptides at the
indicated threshold. (B) The total number of peptides predicted by each
algorithm for each allele at either the 50 or 500 nM threshold is shown.
The total number of peptides that would have been made using no se-
lection is 332.
Algorithm performance in predicting CD4 T cell response
The Journal of Immunology4241
cutoff and would have required testing of only 60 peptides. At the
500 nM cutoff, NetMHCIIpan appeared to outperform the other
algorithms in terms of cost versus benefit, with .60% of the re-
sponse captured and only requiring testing of 90 peptides. For I-Ab,
ARB seemed to be the best performing, with .30% of the re-
sponse predicted at 500 nM predicted affinity cutoff, which would
have selected ∼80 peptides to be tested in secondary screens. This
comparative retrospective analysis shows that it is difficult to
identify algorithms or parameters that will efficiently identify
epitopes for untested class II molecules.
Analysis of algorithm performance by ROC curves
The large discrepancies within the absolute affinities predicted by
to pick a threshold adequately to compare predictive ability be-
tween algorithms. As observed in the previous section, a good
predictive algorithm in a relative sense can hypothetically be too
stringent or lenient with its predicted affinities and be eclipsed by
an algorithm with less predictive ability but more moderate affinity
estimate. To circumvent this caveat, the binary classifier system—
sorting predictions into true positive, false positive, true negative,
and false negative—was imposed at all possible thresholds, and
for each threshold, a true-positive rate and false-positive rate were
calculated and plotted as ROC curves for each algorithm (Fig. 6),
a method that is commonly used when multiple algorithms are
compared for predictive performance (23–25, 32, 34) and that is
generally used to measure the distinguishing capability of a clas-
sification or diagnostic system (75). The AUC values were cal-
culated to score the overall predictive ability of each algorithm
without the need for a binding cutoff and compared with what
would have been predicted by chance alone. From the given
analysis, NetMHCIIpan and NetMHCII seem to score consistently
higher than the mean AUC value. However, aside from Tepitope’s
relatively poor performance in DR1, there is no algorithm that
significantly outperforms or underperforms relative to the other
algorithms. Furthermore, the differences between algorithms are
increased for more well studied MHC class II alleles. The range
and SD between AUC values is largest in DR1 and near negligible
Potential advantage of percentile ranking versus predicted
affinity for selection of peptides as candidates for further
The overall inconsistency in using affinity estimates to predict
epitopes, as well as the tendency of some algorithms to estimate
high affinity to many peptides (ARB) and other to estimate low
possibility that the affinity estimates can be inaccurate in an ab-
solute sense but might be useful in a relative sense. We therefore
asked if ranking of peptides in a hierarchy of affinity rather than
absolute affinity estimates would lead to more consistent results
and more efficient use of the algorithms while allowing us still to
estimate the actual fractional response successfully predicted. To
evaluate this, we simply ranked all the peptides by their estimated
IC50values and then used a simple percentile cutoff rather than
affinity threshold to select peptides (Fig. 7). Using this method of
ranking, we were also able to extend our studies to Consensus and
Tepitope, which simply rank peptides. When this fixed percentile
ranking rather than estimated affinity was used to select peptides,
a much better “performance” was apparent for all alleles and
algorithms, with as much as 40–60% of the response predicted at
even the 10% cutoff, depending on the allele. There was a par-
ticularly striking agreement when comparing the different algo-
rithms for DR1 using this method. This result indicates that
algorithm outputs of IC50values for peptide–class II interactions
are the most useful for simple ranking of candidate peptide epit-
opes, rather than for use as absolute values for selection. This
conclusion from our studies is consistent with recently published
approaches that assimilate the rankings of peptides, rather than
their estimated affinities, to select peptides for further empirical
study (34, 71, 76).
An additional and quite striking observation from this analysis
is that although irregular in shape, the lines that describe the pre-
diction effectiveness of each algorithm in capturing the response
essentially overlap. They all have a fairly sharp slope to the 10%
cutoff (successfully predicting ∼20–30% of the response), then
a diminished slope from ∼10–60% cutoff, and then approach
a plateau after the 60–70% cutoff, representing the low yield of
additional epitopes with increasing peptides tested. The conclu-
sion from these analyses is that ranking of peptides by different
algorithms normalizes their prediction efficiency, and when
compared in this way, they all perform similarly to each other. The
conclusion is in agreement with the ROC analyses discussed
earlier. The individual peptides “captured” at each cutoff varies
somewhat for each algorithm (data not shown), but any given
algorithm would apparently allow identification of ∼25% of the
response by selection of 32 peptides of 332, at least for these three
alleles of MHC class II molecules.
predictive ability of SMM, NetMHCIIpan, NetMHCII, ARB, Tepitope, and
Consensus are shown in terms of ROC curves. The “standard” (y = x) line
represents the curve formed had peptides been chosen at random. Pre-
dictive ability is gauged by how far the points on the curve fall to the upper
left of the standard line. Because better performing algorithms encompass
more area under their respective curves, AUC values are a numerical
representation of algorithm efficiency, with 1 representing a perfect pre-
dictor and 0.5 representing a nonpredictive or random algorithm.
ROC analysis and AUC values of predictive algorithms. The
4242 PREDICTION OF CD4 T CELL EPITOPES
Kinetic measure of class II–peptide affinity as a predictor of
One issue that is raised by the data in Figs. 2 and 4 is why some
peptides that are not predicted to bind to the relevant MHC class II
at all do in fact successfully elicit CD4 T cell responses to in-
fluenza infection. One possibility to explain these false negatives
is that some peptides can bind very weakly to the “presenting”
peptide MHC class II molecule but nonetheless have properties
that allow elicitation of a CD4 T cell response; for example, TCR
contacts that recruit a tremendously high number of reactive CD4
T cells, a parameter that is not currently evaluated by available
algorithms. The other possibility is that the IC50values estimated
by the algorithms do not accurately assess the true affinity of some
MHC class II–peptide complexes. Our laboratory has previously
shown that dissociation kinetics of MHC class II–peptide com-
plexes rather than competition assays are the best predictor of
immunogenicity, both in responses to complex proteins and in
multipeptide-based vaccines (57, 77–80). In analyses of class II–
peptide complexes with dissociation rates that differ by seven
orders of magnitude, McConnell’s group (81) showed that the
association rates were essentially the same, indicating that disso-
ciation kinetics provide accurate measurements of the affinities of
peptide–class II complexes. To evaluate whether this parameter
is a better indicator of immunogenicity for influenza-specific
responses, we tested the dissociation rates of several DR1 pep-
tide complexes using purified class II and purified fluoresceinated
peptides. Two peptides that elicited greater than 100 CD4 T cells
per million but that were not predicted by any of the algorithms
to possess high-affinity binding to DR1 were chosen for this
biochemical study. One peptide was derived from HA (375-
SGYAADQKSTQNAIN), and one was from NP (73-ERRNKY-
LEEHPSAGK). Two control peptides analyzed in earlier work
(56) were studied in parallel. Dissociation half-times of the pep-
tide–class II complexes (t1/2), cytokine ELISPOTs, and predicted
IC50values for these peptides are shown in Table I. It is clear from
this study that all of the immunogenic peptides tested have very
high affinity and stable interactions with the presenting class II
molecule. The HA [375–389] and NP [73–88] peptides that were
not predicted to bind with high affinity to DR1 by any of the
algorithms had dissociation half-times of 600 and 500 h, respec-
tively, indicating that they persist on DR1, even at acidic pH, for
.20 d. Only the one HA peptide previously examined by our
group (ELLVLLENERTLDFHD, t1/2350 h) was predicted to be
a high-affinity binder (low IC50) by most of the algorithms.
Strikingly, among the four peptides, there was no correlation be-
tween the dissociation rates and predicted IC50value. The two
peptides that had the most rapid dissociation rates (HA-162 and
HA-440 with t1/2of 85 and 350 h, respectively) had the highest
estimated affinity (mean of 106 and 148 nM IC50, respectively),
whereas the peptides that had the most stable interactions with
DR1 (HA-375 and NP-73, t1/2of 600 h and 500 h, respectively)
had the lowest mean estimated affinity (3354 and 1300 nM, re-
spectively). Table II shows a similar type of analyses performed
with the murine I-Admolecule using the algorithms that can be
used to score antigenic peptides presented by I-Adand predict
IC50values (NetMHCII, SMM, and ARB). In these studies, known
immunodominant or cryptic epitopes from model or pathogenic
origin, whose binding register and dissociation kinetics were
previously determined by us (57, 59, 79, 80, 82), were studied. We
also compared peptide variants of these peptides whose binding
stability to the class II molecule could be modulated by changes in
anchor residues, as we have described (57, 59, 79, 82). Again, as
was seen with some of the epitopes presented by HLA-DR1,
immunodominant peptides such as LACK [161–173] and an HA
Relationship between the predicted IC50values, kinetic stability, and the immunodominance of influenza-derived peptides presented by
PeptideSequence Cytokine Spotst1/2(h) pH 5.3SMM IC50
Shown are the number of CD4 T cells recruited per million, the t1/2dissociation values for the DR1–peptide complexes, measured as described in Materials and Methods, and
affinity (IC50value) of the peptide for DR1 predicted by the indicated algorithm.
aPeptide with a sequence extension to improve peptide solubility but which did not change the affinity value predicted by the algorithm.
predicted by individual algorithms. The cytokine ELISPOTs for all of the
epitopes recognized by CD4 T cells in each strain of mice were summed to
represent the total response. Prediction of MHC class II binding peptides
were estimated by each algorithm, and epitopes were ranked by each al-
gorithm in the order of diminishing predicted affinity. Cytokine ELISPOT
responses elicited at each incremental rank of the peptides selected by each
algorithm were summed and divided by the total response elicited by all of
the peptides to yield the number of peptide reactive CD4 T cells suc-
cessfully predicted at the indicated threshold. Shown by the dotted line is
the point at which 50% of the response would have been predicted or
“captured” through the use of the indicated algorithm.
Percentile rank cutoff versus fraction of CD4 response
The Journal of Immunology 4243
peptide [126–138, T128 . V] designed to improve its anchor
interaction with the P1 pocket (57, 59, 80, 82) were predicted by
the algorithms to have variable but generally weak IC50values that
were not substantially different from peptides that very poorly
recruit CD4 T cells in vivo such as hen egg lysozyme (HEL) [11–
25] or HA [126–138] (57, 82). In all cases, however, the measured
off-rates of the peptides from the class II molecule correlated with
their immunodominance. Table III additionally shows that in
many cases, incorrect or differing binding registers were predicted
by the different algorithms from what our studies have shown to
be the correct binding register (57, 59, 82). We conclude from
these studies that the ability to bind strongly to the presenting
class II molecule is indeed a strong predictor of immunodomi-
nance and that peptide dissociation assays accurately measure this
parameter, but that algorithms sometimes fail to recognize the
features of high-affinity binding.
Inthisstudy,we havecombined theresults ofanunbiasedapproach
for influenza CD4 T cell epitope discovery with implementation of
predictive algorithms to select immunodominant peptide epitopes.
Advances in CD4 epitope discovery are becoming increasingly
important for the tracking and functional characterization of CD4
Tcells duringorafterinfectionorvaccination andare alsoessential
for derivation of peptide-based vaccines. Purely empirical strate-
gies for epitope discovery, using either the method we have used or
other approaches such as tetramer-based epitope mapping (83–86),
are daunting because of their cost and labor-intensive nature. Also,
for pathogens with large genomes, purely empirical approaches
are impractical. Therefore, the potential benefits of computer-
based algorithms to facilitate identification of CD4 T cell epito-
pes are significant. The increasing depth of available databases for
peptides presented or bound by class II molecules (46, 87–89),
coupled with enhanced sophistication of computer-based algo-
rithms, has led to increasing optimism toward use of predictive
tools as a starting point for epitope discovery. We therefore sought
to evaluate the “state-of-the-art” performance of a number of these
algorithms to predict peptides that are the focus of the CD4 T cell
response to a live influenza infection.
Several important conclusions were possible from the analyses
epitope discovery. With all of the three allelic forms of class II
molecules tested in this study, the identified epitopes were highly
represented in the top third to half of the peptides predicted by the
different algorithms. Although the degree of enrichment varied
with the allele and the algorithm, this result does suggest that use of
predictive algorithms can be a good first step toward epitope
identification, at least for the alleles analyzed in this study and for
pathogens with limited genome sizes or for protein vaccines.
Specific strategies to use this enrichment step for these types of
applications will depend on the goals of the investigator. For ex-
ample, if one seeks to identify only a few epitopes for tracking
a response to a particular pathogen, fairly stringent selection
(i.e., selection of the top 5–10% of peptides) will likely be suc-
cessful. If more criteria need to be fulfilled (highly immunodo-
minant peptides or CD4 epitopes within a particular protein),
lower thresholds may need to be used, with the increasing “yield”
balanced by increasing “cost” of greater numbers of synthetic
peptides to test. Agreement of algorithms and prediction effi-
ciencies were enhanced through use of a ranking system rather
than predicted IC50value to select peptides.
Second and quite striking for each of the algorithms tested, there
is a high number of “false-positive” peptides identified: peptides
that were predicted to be high-affinity binders to the host class II
molecule but that were not recognized by CD4 T cells elicited in
response to influenza infection. The number of peptides in this
category varied, depending on the specific allele and algorithm
and the cutoff used for selection of peptides. For example, with
high stringency (either 50 nM for algorithms that predict IC50
values or 10% for algorithms that rank peptides), the range in false
positives for DR1 ranged from 20 to 65 peptides for the 332-
peptide “proteome,” consisting of five viral proteins tested. When
the threshold included all peptides that are predicted to bind to the
relevant MHC class II molecule, there were .100 false-positive
peptides. From a practical standpoint, the high false-positive rate
for many class II–protein combinations indicates in general that
the algorithms should be used only as a first step in epitope dis-
covery. We conclude that secondary screens are essential to
Correlation of predicted IC50values with the experimentally determined half-lives of peptide/I-Adcomplexes and the
NameSequence Immune Responset1/2(h)NetMHCII 2.2SMM-alignARB
anchor-substituted peptides and dissociation kinetics
Accuracy of the peptide binding cores predicted by algorithms for I-Adcompared with cores determined experimentally with
SMM-alignBinding Core ARBKnown Binding Core
Bold text indicates the known binding core.
4244PREDICTION OF CD4 T CELL EPITOPES
identify the true epitopes, after which focused studies can be pur-
sued. For large-genome organisms, however, secondary screens of
large numbers of candidate peptides may be unfeasible using the
currently available computational strategies.
positive peptides. The first and simplest explanation is that these
peptides have been “chosen” correctly by the algorithm to be high-
affinity binders for the MHC molecules expressed in the host but
lack some property needed to be a CD4 T cell epitope. One ele-
ment that could be lacking in MHC-binding peptides that do not
elicit CD4 T cells could be their ability to be processed and pre-
sented by the class II molecule, due to the source protein abun-
dance or the location of the peptide within the viral protein. A
second feature that might diminish a peptide’s antigenicity despite
high-affinity binding to class II molecules is that it possesses few
amino acid residues in solvent-exposed positions that are potent in
recruitment of CD4 T cells. The predicted affinity of some peptides
may be an overestimate, even for those peptides that have “at-
tractive” TCR contacts and are abundantly released and available
for peptide binding during Ag presentation. We do know that the
reproducibility in responses among individual animals is reason-
ably good (Supplemental Fig. 3), suggesting that we have not
missed epitopes due to pooling of lymphocytes from multiple
animals, each of which might display greatly varied responses.
Pooling of CD4 T cells from multiple mice is a necessary exper-
imental strategy because of the number of different peptide epi-
topes tracked in parallel in our studies and the low yields of CD4
T cells from individual mice. Finally, from a purely technical
standpoint, it is possible that poor peptide quality or solubility
limits our T cell detection for occasional peptides, thus accounting
for false positives. The peptides used in our studies were HPLC
purified, but it is possible that peptide quantity or purity was
overestimated by the supplier or alternatively that peptide solubility
was limited, thus diminishing the signal in the ELISPOT assays.
Our analyses also revealed a high false-negative rate; peptides
be MHC binders by most of the algorithms. The analyses of this
at least 25% of the peptides that elicit strong CD4 T cell responses
were not predicted by most of the algorithms to bind to the MHC
molecule in question. Although in the animal studies reported here,
we have studied the immunodominance pattern at the peak of the
immune response (day 10–14), we have found that the hierarchies
in peptide-specific CD4 T cells remain in the memory phase of the
immune response (56). These results suggest that epitopes dis-
covered in our studies are representative of those persisting in the
memory phase and in human circulating CD4 T cells. In fact,
some of the HLA-DR epitopes discovered in our transgenic
models predicted to be of very low affinity by all of the algorithms
tested, including NP-73 and NP-402 for DR1 and M1-97 for DR4,
have been reported in the literature from human subjects (84, 90).
Finally, we are confident that the false-negative rate is not due to
incorrect assignment of the class II restriction element in the mice.
Only the HLA-DR1 mice express an additional class II molecule,
and for these peptide epitopes, MHC restriction was unambigu-
ously assigned through the use of transfected APC in ELISPOT
assays, as we have described (21).
The false-negative score for many epitopes suggests that the
algorithms in general fail to recognize certain types of peptides that
well studied, such as those in this study. In favor of this possibility
are results of our studies with the I-Adclass II molecule, which has
been extensively analyzed for its peptide binding and cocrystal-
lized with antigenic peptide (91). I-Adwas widely thought to have
a “promiscuous” binding pocket with specificity primarily dictated
by the P4 and P6 pockets (39, 91, 92), but we recently found the
P1 pocket to be a strong determinant of peptide binding particu-
larly with glutamic acid at this position (59, 82). Our studies
suggest that this charged reside binds via a novel salt bridge within
this pocket. Because of this alternative type of interaction in the
P1 pocket, the “value” of glutamic acid in antigenic peptides was
One unexpected finding revealed by our analyses is that the
the algorithms tested, even if ranking is used to order peptides. It is
remarkable that algorithms developed recently do not outperform
others developed more than a decade ago when the efficiency of
prediction was analyzed in this way. This conclusion prompts us to
question whether the approaches used for development and re-
finement of computation tools have fundamental limitations that
will not improve even with more sophisticated computational
methods. We suggest that rather than focus on modifications of
computational strategies, increased performance of computational
aids to epitope prediction may require a serious re-evaluation of
the data that are used to “train” the algorithms. One improvement
that might significantly enhance performance would be dispropor-
tionately to weigh data that provide the most information on the
peptide core and stability of binding to the host MHC molecule.
Most of the newly developed methods are trained on data derived
primarily from competitive binding assays. Because of their rel-
atively low cost and high-throughput capacity, these assays are
very useful for identification of peptides that do or do not bind to
a given class II allele. However, the quantitative ability of peptides
to compete for binding may reflect properties that are distinct from
those that promote accumulation onto class II molecules, clearly
a key event for immunogenicity (57, 77, 80, 82). From the sim-
plest point of view, IC50values are measures of the ability of
a given peptide to diminish accumulation of a labeled “indicator
peptide.” Such competitive assays are based on the assumptions
that data arise from a simple equilibrium between single, ho-
mogenous monovalent receptor and ligand and follow the laws of
mass action (93–96). It is now clear from the work of many groups
that the reactions between peptide and MHC II proteins involve
long-lived intermediates, heterogeneous initial states, and a pre-
equilibrium between active and inactive forms of the class II
protein and that binding reactions compete with inactivation of the
MHC protein, an event that is potentiated by formation of unstable
MHC–peptide complexes (97–101). The complexities in class II–
peptide interactions, coupled with the possibilities of multiple
binding cores of antigenic peptides, indicate that predictions based
solely on competitive binding assays will be limited in their ability
quantitatively to assess the ability of the test peptides to bind
stably to the class II molecule. Dissociation assays and elution and
sequencing or crystallographic studies together provide key in-
sight into the amino acids that promote accumulation of peptides
onto the class II molecules. Although these assays are cost and
labor prohibitive as routine methods, more investment in genera-
tion of these data and more weight given to these data could
enhance the predictive capacity of computational strategies.
Incorporation of data from these assays might also improve ac-
curacy of the 9-aa core predicted to bind in the MHC class II
pocket. We hypothesize that limitations in our understanding of
peptide–class II interactions have led to incorrect core assign-
ments, a hypothesis recently supported by discrepancies between
predicted cores and cocrystallization of peptide–class II com-
plexes (25, 63) and by our own data shown in this study. Faulty
core designation may “contaminate” the prediction accuracy of
preferences in binding going forward.
The Journal of Immunology 4245
True breakthroughs in algorithm performance for prediction of
CD4 T cell epitopes is also likely to be facilitated by considerations
of peptide sequences that contribute to recruitment of diverse
TCRs. Although strength of MHC binding is the keyparameter that
determines a peptide’s immunogenicity, solvent-exposed residues
within the peptide likely determine the upper range of the mag-
nitude of a CD4 T cell response that can be recruited by a peptide.
Recent theoretical and functional studies and detailed consider-
ation of the molecular events in thymic positive and negative se-
lection events support the view that particular solvent-exposed
amino acids within MHC-bound peptides dramatically influence
TCR repertoire development and elicitation (102–108). Notably,
recent syntheses of computation and functional studies suggest
that the best antigenic epitopes will have TCR contact amino acids
that both survive negative selection and have the ability to recruit
diverse TCR, and these studies quantify the amino acids that tend
to favor this “equation” (109, 110). Incorporation of these char-
acteristics of TCR recruitment into computational methods of
epitope prediction will likely increase the efficiency of computer-
based peptide epitope selection. We suggest that more explicit
experimental evaluation of these issues, coupled with the com-
putational power of many groups of investigators, will dramati-
cally enhance the performance of computation-based predictors,
a key step needed for efficient epitope discovery.
We thank Michael (Rusty) Elliot for comments on the manuscript and fac-
ulty at the David H. Smith Center for Vaccine Biology and Immunology
for helpful discussions.
The authors have no financial conflicts of interest.
1. De Groot, A. S., and R. Rappuoli. 2004. Genome-derived vaccines. Expert Rev.
Vaccines 3: 59–76.
2. Herrera-Najera, C., R. Pin ˜a-Aguilar, F. Xacur-Garcia, M. J. Ramirez-Sierra, and
E. Dumonteil. 2009. Mining the Leishmania genome for novel antigens and
vaccine candidates. Proteomics 9: 1293–1301.
3. Sette, A., and R. Rappuoli. 2010. Reverse vaccinology: developing vaccines in
the era of genomics. Immunity 33: 530–541.
4. Chaudhuri, R., S. Ahmed, F. A. Ansari, H. V. Singh, and S. Ramachandran.
2008. MalVac: database of malarial vaccine candidates. Malar. J. 7: 184.
5. Dudek, N. L., P. Perlmutter, M. I. Aguilar, N. P. Croft, and A. W. Purcell. 2010.
Epitope discovery and their use in peptide based vaccines. Curr. Pharm. Des.
6. Shimonkevitz, R., S. Colon, J. W. Kappler, P. Marrack, and H. M. Grey. 1984.
Antigen recognition by H-2-restricted T cells. II. A tryptic ovalbumin peptide
that substitutes for processed antigen. J. Immunol. 133: 2067–2074.
7. Allen, P. M., D. J. Strydom, and E. R. Unanue. 1984. Processing of lysozyme
by macrophages: identification of the determinant recognized by two T-cell
hybridomas. Proc. Natl. Acad. Sci. USA 81: 2489–2493.
8. Villarreal-Ramos, B., J. Sanchez-Garcia, N. Stoker, E. Timms, D. Chomer,
K. Raff, and N. A. Mitchison. 1991. Screening gene expression libraries for
epitopes recognized in Mycobacterium leprae by mouse T cells. Eur. J.
Immunol. 21: 2621–2624.
9. Kurokohchi, K., T. Akatsuka, C. D. Pendleton, A. Takamizawa, M. Nishioka,
M. Battegay, S. M. Feinstone, and J. A. Berzofsky. 1996. Use of recombinant
protein to identify a motif-negative human cytotoxic T-cell epitope presented
by HLA-A2 in the hepatitis C virus NS3 region. J. Virol. 70: 232–240.
10. Turner, M. J., C. S. Abdul-Alim, R. A. Willis, T. L. Fisher, E. M. Lord, and
J. G. Frelinger. 2001. T-cell antigen discovery (T-CAD) assay: a novel tech-
nique for identifying T cell epitopes. J. Immunol. Methods 256: 107–119.
11. Valentino, M. D., Z. J. Maben, L. L. Hensley, M. D. Woolard, T. H. Kawula,
J. A. Frelinger, and J. G. Frelinger. 2011. Identification of T-cell epitopes in
Francisella tularensis using an ordered protein array of serological targets.
Immunology 132: 348–360.
12. Reinhold, B., D. B. Keskin, and E. L. Reinherz. 2010. Molecular detection of
targeted major histocompatibility complex I-bound peptides using a probabi-
listic measure and nanospray MS(3) on a hybrid quadrupole-linear ion trap.
Anal. Chem. 82: 9090–9099.
13. Purcell, A. W., and J. J. Gorman. 2001. The use of post-source decay in matrix-
assisted laser desorption/ionisation mass spectrometry to delineate T cell
determinants. J. Immunol. Methods 249: 17–31.
14. Karunakaran, K. P., J. Rey-Ladino, N. Stoynov, K. Berg, C. Shen, X. Jiang,
B. R. Gabel, H. Yu, L. J. Foster, and R. C. Brunham. 2008. Immunoproteomic
discovery of novel T cell antigens from the obligate intracellular pathogen
Chlamydia. J. Immunol. 180: 2459–2465.
15. Tan, C. T., N. P. Croft, N. L. Dudek, N. A. Williamson, and A. W. Purcell. 2011.
Direct quantitation of MHC-bound peptide epitopes by selected reaction
monitoring. Proteomics 11: 2336–2340.
16. Strug, I., J. M. Calvo-Calle, K. M. Green, J. Cruz, F. A. Ennis, J. E. Evans, and
L. J. Stern. 2008. Vaccinia peptides eluted from HLA-DR1 isolated from virus-
infected cells are recognized by CD4+ T cells from a vaccinated donor. J.
Proteome Res. 7: 2703–2711.
17. Hiemstra, H. S., J. W. Drijfhout, and F. Koning. 2000. Synthetic peptide li-
braries for T-cell epitope identification. Methods Mol. Med. 41: 89–96.
18. Geginat, G., S. Schenk, M. Skoberne, W. Goebel, and H. Hof. 2001. A novel
approach of direct ex vivo epitope mapping identifies dominant and subdom-
inant CD4 and CD8 T cell epitopes from Listeria monocytogenes. J. Immunol.
19. Tobery, T. W., and M. J. Caulfield. 2004. Identification of T-cell epitopes using
ELISpot and peptide pool arrays. Methods Mol. Med. 94: 121–132.
20. Maciel, M., Jr., S. N. Kellathur, P. Chikhlikar, R. Dhalia, J. Sidney, A. Sette,
T. J. August, and E. T. Marques, Jr. 2008. Comprehensive analysis of T cell
epitope discovery strategies using 17DD yellow fever virus structural proteins
and BALB/c (H2d) mice model. Virology 378: 105–117.
21. Richards, K. A., F. A. Chaves, and A. J. Sant. 2009. Infection of HLA-DR1
transgenic mice with a human isolate of influenza a virus (H1N1) primes
a diverse CD4 T-cell repertoire that includes CD4 T cells with heterosubtypic
cross-reactivity to avian (H5N1) influenza virus. J. Virol. 83: 6566–6577.
22. Nayak, J. L., K. A. Richards, F. A. Chaves, and A. J. Sant. 2010. Analyses of
the specificity of CD4 T cells during the primary immune response to influenza
virus reveals dramatic MHC-linked asymmetries in reactivity to individual viral
proteins. Viral Immunol. 23: 169–180.
23. Zhang, H., P. Wang, N. Papangelopoulos, Y. Xu, A. Sette, P. E. Bourne,
O. Lund, J. Ponomarenko, M. Nielsen, and B. Peters. 2010. Limitations of Ab
initio predictions of peptide binding to MHC class II molecules. PLoS ONE 5:
24. Wang, P., J. Sidney, Y. Kim, A. Sette, O. Lund, M. Nielsen, and B. Peters. 2010.
Peptide binding predictions for HLA DR, DP and DQ molecules. BMC Bio-
informatics 11: 568.
25. Lin, H. H., G. L. Zhang, S. Tongchusak, E. L. Reinherz, and V. Brusic. 2008.
Evaluation of MHC-II peptide binding prediction servers: applications for
vaccine research. BMC Bioinformatics 9(Suppl 12): S22.
26. Lin, H. H., S. Ray, S. Tongchusak, E. L. Reinherz, and V. Brusic. 2008.
Evaluation of MHC class I peptide binding prediction servers: applications for
vaccine research. BMC Immunol. 9: 8.
27. Peters, B., H. H. Bui, S. Frankild, M. Nielson, C. Lundegaard, E. Kostem,
D. Basch, K. Lamberth, M. Harndahl, W. Fleri, et al. 2006. A community re-
source benchmarking predictions of peptide binding to MHC-I molecules.
PLOS Comput. Biol. 2: e65.
28. Nielsen, M., O. Lund, S. Buus, and C. Lundegaard. 2010. MHC class II epitope
predictive algorithms. Immunology 130: 319–328.
29. Nielsen, M., C. Lundegaard, and O. Lund. 2007. Prediction of MHC class II
binding affinity using SMM-align, a novel stabilization matrix alignment
method. BMC Bioinformatics 8: 238.
30. Bui, H. H., J. Sidney, B. Peters, M. Sathiamurthy, A. Sinichi, K. A. Purton,
B. R. Mothe ´, F. V. Chisari, D. I. Watkins, and A. Sette. 2005. Automated
generation and evaluation of specific MHC binding predictive tools: ARB
matrix applications. Immunogenetics 57: 304–314.
31. Zhang, G. L., A. M. Khan, K. N. Srinivasan, J. T. August, and V. Brusic. 2005.
Neural models for predicting viral vaccine targets. J. Bioinform. Comput. Biol.
32. Nielsen, M., and O. Lund. 2009. NN-align. An artificial neural network-based
alignment algorithm for MHC class II peptide binding prediction. BMC Bio-
informatics 10: 296.
33. Zhang, W., J. Liu, and Y. Niu. 2010. Quantitative prediction of MHC-II binding
affinity using particle swarm optimization. Artif. Intell. Med. 50: 127–132.
34. Wang, P., J. Sidney, C. Dow, B. Mothe ´, A. Sette, and B. Peters. 2008. A sys-
tematic assessment of MHC class II peptide binding predictions and evaluation
of a consensus approach. PLOS Comput. Biol. 4: e1000048.
35. Falk, K., O. Ro ¨tzschke, S. Stevanovic ´, G. Jung, and H. G. Rammensee. 1991.
Allele-specific motifs revealed by sequencing of self-peptides eluted from
MHC molecules. Nature 351: 290–296.
36. Deres, K., T. N. Schumacher, K. H. Wiesmu ¨ller, S. Stevanovic ´, G. Greiner,
G. Jung, and H. L. Ploegh. 1992. Preferred size of peptides that bind to H-2 Kb
is sequence dependent. Eur. J. Immunol. 22: 1603–1608.
37. Matsumura, M., D. H. Fremont, P. A. Peterson, and I. A. Wilson. 1992.
Emerging principles for the recognition of peptide antigens by MHC class I
molecules. Science 257: 927–934.
38. Chicz, R. M., R. G. Urban, W. S. Lane, J. C. Gorga, L. J. Stern, D. A. Vignali,
and J. L. Strominger. 1992. Predominant naturally processed peptides bound to
HLA-DR1 are derived from MHC-related molecules and are heterogeneous in
size. Nature 358: 764–768.
39. Hunt, D. F., H. Michel, T. A. Dickinson, J. Shabanowitz, A. L. Cox,
K. Sakaguchi, E. Appella, H. M. Grey, and A. Sette. 1992. Peptides presented to
the immune system by the murine class II major histocompatibility complex
molecule I-Ad. Science 256: 1817–1820.
40. Lippolis, J. D., F. M. White, J. A. Marto, C. J. Luckey, T. N. Bullock,
J. Shabanowitz, D. F. Hunt, and V. H. Engelhard. 2002. Analysis of MHC class
4246PREDICTION OF CD4 T CELL EPITOPES
II antigen processing by quantitation of peptides that constitute nested sets. J.
Immunol. 169: 5089–5097.
41. Srinivasan, M., S. Z. Domanico, P. T. Kaumaya, and S. K. Pierce. 1993. Pep-
tides of 23 residues or greater are required to stimulate a high affinity class II-
restricted T cell response. Eur. J. Immunol. 23: 1011–1016.
42. Landais, E., P. A. Romagnoli, A. L. Corper, J. Shires, J. D. Altman,
I. A. Wilson, K. C. Garcia, and L. Teyton. 2009. New design of MHC class II
tetramers to accommodate fundamental principles of antigen presentation. J.
Immunol. 183: 7949–7957.
43. Bankovich, A. J., A. T. Girvin, A. K. Moesta, and K. C. Garcia. 2004. Peptide
register shifting within the MHC groove: theory becomes reality. Mol. Immu-
nol. 40: 1033–1039.
44. Robertson, J. M., P. E. Jensen, and B. D. Evavold. 2000. DO11.10 and OT-II
T cells recognize a C-terminal ovalbumin 323-339 epitope. J. Immunol. 164:
45. McFarland, B. J., A. J. Sant, T. P. Lybrand, and C. Beeson. 1999. Ovalbumin
(323-339) peptide binds to the major histocompatibility complex class II I-A(d)
protein using two functionally distinct registers. Biochemistry 38: 16663–
46. Peters, B., J. Sidney, P. Bourne, H. H. Bui, S. Buus, G. Doh, W. Fleri,
M. Kronenberg, R. Kubo, O. Lund, et al. 2005. The immune epitope database
and analysis resource: from vision to blueprint. PLoS Biol. 3: e91.
47. Peters, B., J. Sidney, P. Bourne, H. H. Bui, S. Buus, G. Doh, W. Fleri,
M. Kronenberg, R. Kubo, O. Lund, et al. 2005. The design and implementation
of the immune epitope database and analysis resource. Immunogenetics 57:
48. Zhang, G. L., H. H. Lin, D. B. Keskin, E. L. Reinherz, and V. Brusic. 2011.
Dana-Farber repository for machine learning in immunology. J. Immunol.
Methods 374: 18–25.
49. Nielsen, M., S. Justesen, O. Lund, C. Lundegaard, and S. Buus. 2010.
NetMHCIIpan-2.0—improved pan-specific HLA-DR predictions using a novel
concurrent alignment and weight optimization training procedure. Immunome
Res. 6: 9.
50. Dimitrov, I., P. Garnev, D. R. Flower, and I. Doytchinova. 2010. MHC class II
binding prediction—a little help from a friend. J. Biomed. Biotechnol. 2010:
51. Hu, X., W. Zhou, K. Udaka, H. Mamitsuka, and S. Zhu. 2010. MetaMHC:
a meta approach to predict peptides binding to MHC molecules. Nucleic Acids
Res. 38(Web Server issue): W474-9.
52. Mustafa, A. S., F. A. Shaban, R. Al-Attiyah, A. T. Abal, A. M. El-Shamy,
P. Andersen, and F. Oftung. 2003. Human Th1 cell lines recognize the My-
cobacterium tuberculosis ESAT-6 antigen and its peptides in association with
frequently expressed HLA class II molecules. Scand. J. Immunol. 57: 125–
53. Zhao, B. P., L. Chen, Y. L. Zhang, J. M. Yang, K. Jia, C. Y. Sui, C. X. Yuan,
J. J. Lin, and X. G. Feng. 2011. In silico prediction of binding of promiscuous
peptides to multiple MHC class-II molecules identifies the Th1 cell epitopes
from secreted and transmembrane proteins of Schistosoma japonicum in
BALB/c mice. Microbes Infect. 13: 709–719.
54. Foo, D. G., P. A. Macary, S. Alonso, and C. L. Poh. 2008. Identification of
human CD4 T-cell epitopes on the VP1 capsid protein of enterovirus 71. Viral
Immunol. 21: 215–224.
55. Fonseca, C. T., E. Cunha-Neto, A. C. Goldberg, J. Kalil, A. R. de Jesus,
E. M. Carvalho, R. Correa-Oliveira, J. Hammer, J. Sidney, A. Sette, and
S. C. Oliveira. 2005. Identification of paramyosin T cell epitopes associated
with human resistance to Schistosoma mansoni reinfection. Clin. Exp. Immu-
nol. 142: 539–547.
56. Richards, K. A., F. A. Chaves, and A. J. Sant. 2011. The memory phase of the
CD4 T-cell response to influenza virus infection maintains its diverse antigen
specificity. Immunology 133: 246–256.
57. Lazarski, C. A., F. A. Chaves, S. A. Jenks, S. Wu, K. A. Richards, J. M. Weaver,
and A. J. Sant. 2005. The kinetic stability of MHC class II:peptide complexes
is a key parameter that dictates immunodominance. Immunity 23: 29–40.
58. Chaves, F. A., and A. J. Sant. 2007. Measurement of peptide dissociation from
MHC class II molecules. Curr. Protoc. Immunol. Chapter 18: Unit 18.14.
59. Chaves, F. A., K. A. Richards, A. Torelli, J. Wedekind, and A. J. Sant. 2006.
Peptide-binding motifs for the I-Ad MHC class II molecule: alternate pH-
dependent binding behavior. Biochemistry 45: 6426–6433.
60. Chaves, F. A., P. Hou, S. Wu, and A. J. Sant. 2005. Replacement of the
membrane proximal region of I-A(d) MHC class II molecule with I-E-derived
sequences promotes production of an active and stable soluble heterodimer
without altering peptide-binding specificity. J. Immunol. Methods 300: 74–92.
61. Nelson, P. A., M. Khodadoust, T. Prodhomme, C. Spencer, J. C. Patarroyo,
M. Varrin-Doyer, J. D. Ho, R. M. Stroud, and S. S. Zamvil. 2010. Immuno-
dominant T cell determinants of aquaporin-4, the autoantigen associated with
neuromyelitis optica. PLoS ONE 5: e15050.
62. Akhoon, B. A., P. S. Slathia, P. Sharma, S. K. Gupta, and V. Verma. 2011. In
silico identification of novel protective VSG antigens expressed by Trypano-
soma brucei and an effort for designing a highly immunogenic DNA vaccine
using IL-12 as adjuvant. Microb. Pathog. 51: 77–87.
63. Bordner, A. J., and H. D. Mittelmann. 2010. Prediction of the binding affinities
of peptides to class II MHC using a regularized thermodynamic model. BMC
Bioinformatics 11: 41.
64. Nielsen, M., C. Lundegaard, T. Blicher, B. Peters, A. Sette, S. Justesen,
S. Buus, and O. Lund. 2008. Quantitative predictions of peptide binding to any
HLA-DR molecule of known sequence: NetMHCIIpan. PLOS Comput. Biol. 4:
65. Richards, K. A., F. A. Chaves, F. R. Krafcik, D. J. Topham, C. A. Lazarski, and
A. J. Sant. 2007. Direct ex vivo analyses of HLA-DR1 transgenic mice reveal
an exceptionally broad pattern of immunodominance in the primary HLA-DR1-
restricted CD4 T-cell response to influenza virus hemagglutinin. J. Virol. 81:
66. Tobery, T. W., S. Wang, X. M. Wang, M. P. Neeper, K. U. Jansen,
W. L. McClements, and M. J. Caulfield. 2001. A simple and efficient method
for the monitoring of antigen-specific T cell responses using peptide pool arrays
in a modified ELISpot assay. J. Immunol. Methods 254: 59–66.
67. Sturniolo, T., E. Bono, J. Ding, L. Raddrizzani, O. Tuereci, U. Sahin,
M. Braxenthaler, F. Gallazzi, M. P. Protti, F. Sinigaglia, and J. Hammer. 1999.
Generation of tissue-specific and promiscuous HLA ligand databases using
DNA microarrays and virtual HLA class II matrices. Nat. Biotechnol. 17: 555–
68. Cantor, J. R., T. H. Yoo, A. Dixit, B. L. Iverson, T. G. Forsthuber, and
G. Georgiou. 2011. Therapeutic enzyme deimmunization by combinatorial T-
cell epitope removal using neutral drift. Proc. Natl. Acad. Sci. USA 108: 1272–
69. Yauch, L. E., T. R. Prestwood, M. M. May, M. M. Morar, R. M. Zellweger,
B. Peters, A. Sette, and S. Shresta. 2010. CD4+ T cells are not required for the
induction of dengue virus-specific CD8+ T cell or antibody responses but
contribute to protection after vaccination. J. Immunol. 185: 5405–5416.
70. Chen, C., C. Dow, P. Wang, J. Sidney, A. Read, A. Harmsen, J. E. Samuel, and
B. Peters. 2011. Identification of CD4+ T cell epitopes in C. burnetii antigens
targeted by antibody responses. PLoS ONE 6: e17712.
71. Arens, R., P. Wang, J. Sidney, A. Loewendorf, A. Sette, S. P. Schoenberger,
B. Peters, and C. A. Benedict. 2008. Cutting edge: murine cytomegalovirus
induces a polyfunctional CD4 T cell response. J. Immunol. 180: 6472–6476.
72. Chicz, R. M., R. G. Urban, J. C. Gorga, D. A. Vignali, W. S. Lane, and
J. L. Strominger. 1993. Specificity and promiscuity among naturally processed
peptides bound to HLA-DR alleles. J. Exp. Med. 178: 27–47.
73. Sant, A. J. 1994. Endogenous antigen presentation by MHC class II molecules.
Immunol. Res. 13: 253–267.
74. Loss, G. E., Jr., C. G. Elias, P. E. Fields, R. K. Ribaudo, M. McKisic, and
A. J. Sant. 1993. Major histocompatibility complex class II-restricted presen-
tation of an internally synthesized antigen displays cell-type variability and
segregates from the exogenous class II and endogenous class I presentation
pathways. J. Exp. Med. 178: 73–85.
75. Swets, J. A. 1988. Measuring the accuracy of diagnostic systems. Science 240:
76. Calvo-Calle, J. M., I. Strug, M. D. Nastke, S. P. Baker, and L. J. Stern. 2007.
Human CD4+ T cell epitopes from vaccinia virus induced by vaccination or
infection. PLoS Pathog. 3: 1511–1529.
77. Sant, A. J., F. A. Chaves, S. A. Jenks, K. A. Richards, P. Menges, J. M. Weaver,
and C. A. Lazarski. 2005. The relationship between immunodominance, DM
editing, and the kinetic stability of MHC class II:peptide complexes. Immunol.
Rev. 207: 261–278.
78. Sant, A. J., F. A. Chaves, F. R. Krafcik, C. A. Lazarski, P. Menges, K. Richards,
and J. M. Weaver. 2007. Immunodominance in CD4 T-cell responses: impli-
cations for immune responses to influenza virus and for vaccine design. Expert
Rev. Vaccines 6: 357–368.
79. Weaver, J. M., C. A. Lazarski, K. A. Richards, F. A. Chaves, S. A. Jenks,
P. R. Menges, and A. J. Sant. 2008. Immunodominance of CD4 T cells to
foreign antigens is peptide intrinsic and independent of molecular context:
implications for vaccine design. J. Immunol. 181: 3039–3048.
80. Weaver, J. M., F. A. Chaves, and A. J. Sant. 2009. Abortive activation of CD4
T cell responses during competitive priming in vivo. Proc. Natl. Acad. Sci. USA
81. Kasson, P. M., J. D. Rabinowitz, L. Schmitt, M. M. Davis, and
H. M. McConnell. 2000. Kinetics of peptide binding to the class II MHC
protein I-Ek. Biochemistry 39: 1048–1058.
82. Lazarski, C. A., F. A. Chaves, and A. J. Sant. 2006. The impact of DM on MHC
class II-restricted antigen presentation can be altered by manipulation of MHC-
peptide kinetic stability. J. Exp. Med. 203: 1319–1328.
83. Grotenbreg, G. M., N. R. Roan, E. Guillen, R. Meijers, J. H. Wang, G. W. Bell,
M. N. Starnbach, and H. L. Ploegh. 2008. Discovery of CD8+ T cell epitopes in
Chlamydia trachomatis infection through use of caged class I MHC tetramers.
Proc. Natl. Acad. Sci. USA 105: 3831–3836.
84. Roti, M., J. Yang, D. Berger, L. Huston, E. A. James, and W. W. Kwok. 2008.
Healthy human subjects have CD4+ T cells directed against H5N1 influenza
virus. J. Immunol. 180: 1758–1768.
85. Yang, J., E. A. James, L. Huston, N. A. Danke, A. W. Liu, and W. W. Kwok.
2006. Multiplex mapping of CD4 T cell epitopes using class II tetramers. Clin.
Immunol. 120: 21–32.
86. Reijonen, H., and W. W. Kwok. 2003. Use of HLA class II tetramers in tracking
antigen-specific T cells and mapping T-cell epitopes. Methods 29: 282–288.
87. Salimi, N., W. Fleri, B. Peters, and A. Sette. 2010. Design and utilization of
epitope-based databases and predictive tools. Immunogenetics 62: 185–196.
88. Zhang, Q., P. Wang, Y. Kim, P. Haste-Andersen, J. Beaver, P. E. Bourne,
H. H. Bui, S. Buus, S. Frankild, J. Greenbaum, et al. 2008. Immune epitope
database analysis resource (IEDB-AR). Nucleic Acids Res. 36(Web Server is-
89. Chung, J. L., J. Sun, J. Sidney, A. Sette, and B. Peters. 2010. IMMUNOCAT-
a data management system for epitope mapping studies. J. Biomed. Biotechnol.
90. Ge, X., J. A. Gebe, P. L. Bollyky, E. A. James, J. Yang, L. J. Stern, and
W. W. Kwok. 2010. Peptide-MHC cellular microarray with innovative data
The Journal of Immunology4247
analysis system for simultaneously detecting multiple CD4 T-cell responses. Download full-text
PLoS ONE 5: e11355.
91. Scott, C. A., P. A. Peterson, L. Teyton, and I. A. Wilson. 1998. Crystal struc-
tures of two I-Ad-peptide complexes reveal that high affinity can be achieved
without large anchor residues. [erratum appears in Immunity 1998;8(4): 531]
Immunity 8: 319–329.
92. Sette, A., S. Buus, S. Colon, C. Miles, and H. M. Grey. 1988. I-Ad-binding
peptides derived from unrelated protein antigens share a common structural
motif. J. Immunol. 141: 45–48.
93. Hulme, E. C., and M. A. Trevethick. 2010. Ligand binding assays at equilib-
rium: validation and interpretation. Br. J. Pharmacol. 161: 1219–1237.
94. Zettner, A. 1974. Some theoretical aspects of competitive binding assays. Hum.
Pathol. 5: 6–8.
95. Zettner, A. 1973. Principles of competitive binding assays (saturation analysis).
1. Equilibrium techniques. Clin. Chem. 19: 699–705.
96. Cheng, Y., and W. H. Prusoff. 1973. Relationship between the inhibition con-
stant (K1) and the concentration of inhibitor which causes 50 per cent inhibition
(I50) of an enzymatic reaction. Biochem. Pharmacol. 22: 3099–3108.
97. Beeson, C., and H. M. McConnell. 1994. Kinetic intermediates in the reactions
between peptides and proteins of major histocompatibility complex class II.
Proc. Natl. Acad. Sci. USA 91: 8842–8845.
98. de Kroon, A. I., and H. M. McConnell. 1994. Kinetics and specificity of
peptide-MHC class II complex displacement reactions. J. Immunol. 152: 609–
99. Beeson, C., and H. M. McConnell. 1995. Reactions of peptides with class II
proteins of the major histocompatibility complex. J. Am. Chem. Soc. 117:
100. Zarutskie, J. A., R. Busch, Z. Zavala-Ruiz, M. Rushe, E. D. Mellins, and
L. J. Stern. 2001. The kinetic basis of peptide exchange catalysis by HLA-DM.
Proc. Natl. Acad. Sci. USA 98: 12450–12455.
101. Sadegh-Nasseri, S., and H. M. McConnell. 1989. A kinetic intermediate in the
reaction of an antigenic peptide and I-Ek. Nature 337: 274–276.
102. Turner, S. J., K. Kedzierska, H. Komodromou, N. L. La Gruta, M. A. Dunstone,
A. I. Webb, R. Webby, H. Walden, W. Xie, J. McCluskey, et al. 2005. Lack of
prominent peptide-major histocompatibility complex features limits repertoire
diversity in virus-specific CD8+ T cell populations. Nat. Immunol. 6: 382–389.
103. Hogquist, K. A., T. A. Baldwin, and S. C. Jameson. 2005. Central tolerance:
learning self-control in the thymus. Nat. Rev. Immunol. 5: 772–782.
104. Chen, J. L., G. Stewart-Jones, G. Bossi, N. M. Lissin, L. Wooldridge,
E. M. Choi, G. Held, P. R. Dunbar, R. M. Esnouf, M. Sami, et al. 2005.
Structural and kinetic basis for heightened immunogenicity of T cell vaccines.
J. Exp. Med. 201: 1243–1255.
105. Stewart-Jones, G. B., A. J. McMichael, J. I. Bell, D. I. Stuart, and E. Y. Jones.
2003. A structural basis for immunodominant human T cell receptor recogni-
tion. Nat. Immunol. 4: 657–663.
106. Rudolph, M. G., R. L. Stanfield, and I. A. Wilson. 2006. How TCRs bind
MHCs, peptides, and coreceptors. Annu. Rev. Immunol. 24: 419–466.
107. Chu, H. H.,J. J. Moon, A. C. Kruse, M.Pepper, and M.K. Jenkins. 2010. Negative
selection and peptide chemistry determine the size of naive foreign peptide-MHC
class II-specific CD4+ T cell populations. J. Immunol. 185: 4705–4713.
108. Obar, J. J., K. M. Khanna, and L. Lefranc ¸ois. 2008. Endogenous naive CD8+
T cell precursor frequency regulates primary and memory responses to infec-
tion. Immunity 28: 859–869.
109. Kosmrlj, A., A. K. Jha, E. S. Huseby, M. Kardar, and A. K. Chakraborty. 2008.
How the thymus designs antigen-specific and self-tolerant T cell receptor
sequences. Proc. Natl. Acad. Sci. USA 105: 16671–16676.
110. Kosmrlj, A., E. L. Read, Y. Qi, T. M. Allen, M. Altfeld, S. G. Deeks, F. Pereyra,
M. Carrington, B. D. Walker, and A. K. Chakraborty. 2010. Effects of thymic
selection of the T-cell repertoire on HLA class I-associated control of HIV
infection. Nature 465: 350–354.
4248PREDICTION OF CD4 T CELL EPITOPES