During the past 10 years significant efforts have been focused on the understanding of the immunity and immunological memory to vaccinia virus and other poxviruses (reviewed in ref. 
. Particularly, it has been shown that both, CD8+ and CD4+ T cells respond vigorously following smallpox vaccination of humans 
. The kinetic analysis of CD4+ versus CD8+ responses suggests that the CD4+ response is lower than the CD8+ at 2 weeks post-vaccination 
, but similar in magnitude at 1 month post-vaccination 
. Furthermore, both vaccinia specific neutralizing antibodies and CD4+ responses are detected in subjects after more than 40 years of Dryvax immunization 
or variola infection 
. In a murine model of vaccinia infection it has been recently demonstrated that there is a strong concordance between CD4+ T cells and antibody protein targets. In fact, this study demonstrated that 11 out of the 18 proteins recognized by CD4+ T cell responses were also detected by antibodies 
. It is also well established that CD4+ T cells are important in primary clearance of vaccinia and in the induction and maintenance of long-term memory and protection from variola challenge. Importantly, Puissant-Lubrano in a recent study showed that the number of residual vaccinia-specific CD4+ lymphocytes (but not CD8+) is inversely associated with the size of the skin lesion formed in response to revaccination in humans 
. Together, these findings clearly support the importance of CD4+ T cells in long term memory to vaccinia infection.
The elucidation of antigen specificity in response to vaccinia immunization has been more difficult for CD4+ than for CD8+ T cells. This could be explained in part by the fact that most CD8+ T cell epitopes have been identified using prediction binding algorithms which have relatively poor prognostic ability for prediction of peptides that bind class II molecules 
. In addition, the use of traditional readouts of CD8+ specificity, such as IFN-γ detection to characterize CD4+ T cell responses, could have underestimated the frequency of CD4+ responses and therefore precluded the understanding of CD4+ specificity. With this in mind, we hypothesized that a comprehensive characterization of the CD4+ T cell response to vaccinia could reveal better readouts of specificity to enumerate and characterize responses that otherwise would have not been detected. In this study, we implemented an approach in which PBMC from human immunized donors were a single time in vitro
stimulated with vaccinia virus, and the resulting CD4+ T cell lines and clones were characterized in terms of their cytokine production (22 cytokines, ). Positive responses were observed for 14 of the 22 cytokines tested, namely GM-CSF, IFN-γ, Il-2, IL-3, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12(p40), IL-13, IP-10, Rantes and TNF-α. Six of these cytokines (GM-CSF, IFN-γ, IL-2, IL-4, IL-13, TNF-α) were selected to characterize both the vaccinia and the peptide responses by CD4+ T cells and all were found to be positive. Our results show that the agonistic activity of the peptides and the response to vaccinia infection is enough to induce a large spectrum of cytokines. Indeed, ICCS revealed that both Th-1 and Th-2 cytokines are being produced simultaneously by the same clonal cells in response to peptide, demonstrating that the clones cannot necessarily be assigned to any particular CD4+ Th-subset. Furthermore, all 4 vaccinia CD4+ T cell clones showed the capacity to kill vaccinia infected targets and vaccinia peptide loaded APCs. However, the concentration of antigen required to trigger cytotoxic T cell responses are higher than those needed to induce cytokine production. Together, our studies show that the vaccinia clones studied here are poly-functional and suggest that the detection levels of each of those functions depend on the amount of antigen used for stimulation.
The detection of GM-CSF in response to mixtures of a positional scanning library was found to be an optimal and reliable readout of CD4+ activation and led to the identification of 4 novel immunogenic vaccinia epitopes derived from proteins prevalently recognized by CD4+ T cells and antibodies upon smallpox vaccination in humans. Previously, GM-CSF was measured in the serum of vaccines in an effort to correlate adverse effects with the presence of cytokines 
but, to our knowledge, this is the first time that GM-CSF is monitored and is shown to be clearly produced by T cells in response to vaccinia.
The determination of cytokine production at decreasing concentrations of peptide clearly showed that the production of GM-CSF requires less peptide than TNF-α (). This finding explains why GM-CSF and not TNF-α was detected in response to mixtures of the positional scanning library composed of billions of different peptides present at very low concentrations. To our knowledge, this study shows for the first time that lower antigen concentrations are required for substantial production of GM-CSF as compared to other cytokines. Its detection after 6 hours and its increase after 48 hours () are in agreement with a report by Abdalla et al in which mRNA of GM-CSF is detectable as fast as 30 minutes after stimulation of CD4+ T cells with a recall antigen (purified protein derivative) and it is maintained at high levels over a period of 96 hours 
. Furthermore, no other evaluated cytokine (IL-2, IL-5, IFN-γ, TNF-α) maintained the fold increase observed for GM-CSF. Topalian et al, also suggested the lower antigen concentration requirement for GM-CSF production 
. In this study, CD4+ T cells produced GM-CSF at significantly higher levels than TNF-α, IL-4, and IFN-γ in response to tumor lysates. While GM-CSF is not a standard cytokine used for monitoring human T cell responses against vaccines or infection, its production by activated T cells is generally accepted. A review by Shi et al in 2006 describes its immunobiology specifically in T cells 
, and more recently a number of studies have demonstrated the advantages of monitoring T cells expressing GM-CSF in the context of infection 
, and vaccine responses to bacterial 
and tumor antigens 
Similar results were obtained either by ICCS () or by measuring the cytokines in the culture supernatants () when the T cell clones were stimulated with their specific vaccinia peptides for 6 hours. In contrast, in response to vaccinia infection (LCLs-Vacc) ICCS revealed that the large majority of the cytokine producing cells produce only TNF-α (52% to 95% depending on the clone) whereas measurement of the same cytokines in the culture supernatants show that GM-CSF, IL-13 and TNF-α are clearly detected. It is therefore important to note that different readouts of activation, antigens, and times of stimulation provide multiple snapshots of the same dynamic process. Indeed, the importance of monitoring multiple cytokines when analyzing the T cell response to vaccination and infection has been previously reported. However, in response to vaccinia very few studies have looked at cytokines other than IFN-γ production. Hammarlund et al 
monitored TNF-α and IFN-γ producing vaccinia specific T cells by ICCS, and Ryan et al 
studied the production of 11 cytokines in response to vaccinia, but they did not include TNF-α, GM-CSF and IL-13 which we identified in the present study to be clearly produced by vaccinia specific T cell clones. In addition, it is clear that a significant percentage of peptide and vaccinia specific T cells in bulk populations can only be detected by ICCS by the measurement of TNF-α production (). Together, these results may explain the intriguing observation 
that in bulk PBMC cultures, the total numbers of IFN-γ secreting cells responding to the individual peptides is higher than the number of IFN-γ secreting cells responding to whole vaccinia virus. Our findings support the idea that quantifying only IFN-γ producing cells underestimates the CD4+ T cell response to vaccinia virus. Furthermore, Jing et al showed that CD4+ responses to a group of vaccinia fragments were considered positive by proliferation but negative by IFN-γ quantification 
. In addition, results with dengue virus also showed that the majority of CD4+ cytokine-positive T cells from donors immunized with the live attenuated vaccine produced either TNF-α alone or TNF-α and IFN-γ when stimulated with heterologous antigen serotypes 
The results presented here show that human vaccinia specific CD4+ T cell clones that emerge following smallpox vaccination produce high levels of IL-13 in response to vaccinia and peptide stimulation. Furthermore, vaccinia specific T cell lines and ex-vivo responses in PBMC from vaccinated donors also respond with IL-13 secretion upon in vitro
vaccinia stimulation ( and data not shown). The role of IL-13 on vaccinia infection has been studied in the context of atopic dermatitis and its effect on vaccinia growth in keratinocytes has been reported 
. In addition, it has been shown that following smallpox immunization 
a small proportion of vaccinia specific CD4+ T cells produce exclusively IL-13 during the peak effector phase of the response (2 weeks following Dryvax vaccination). Huaman et al showed that in humans, immunization with a fragment of Plasmodium falciparum
triggers memory CD4+ T cells that produce IL-13 
. In this regard, it has been suggested that the production of an appropriate amount of IL-13 during infection could moderate the degree of pathogen-induced inflammation. Particularly, data generated in a rat model of parainfluenza type 1 (Sendai) virus infection indicated that appropriate secretion of IL-13 could also serve to limit the extent of virus-induced inflammation 
. Whether IL-13 has a role in long term protection upon vaccination cannot be concluded from these results, but it is clear that quantification of IL-13 secretion by CD4+ T cells could be used to follow up and characterize vaccinia responses upon vaccination.
It is interesting that although all the clones produced IFN-γ in response to PHA and peptide, it was not detected in response to vaccinia infected LCLs when measured in the culture supernatants upon 48 hours of stimulation. In addition, all the clones showed IFN-γ production when measured by ICCS upon peptide stimulation and at a lower level upon vaccinia stimulation (, VRC19-16 and VRC19-29 clones). The most likely explanation for the lack of correlation on the intracellular detection of IFN-γ, but not in the supernatant cultures upon vaccinia stimulation, is the expression of IFN-γ binding proteins (IFN-γ BPs) by vaccinia virus. IFN-γ BPs have been reported to efficiently bind and antagonize soluble IFN-γ 
. An alternative explanation derives from studies reported by Zaunders et al in which they show that IFN-γ production is characteristic of effector CD4+ T cells at very early time-points upon vaccination, but IFN-γ levels decrease as much as 10-fold lower than other indicators of vaccinia specificity during memory T cell differentiation 
In this study vaccinia specific CD4+ T cell clones from human immunized subjects were screened with combinatorial peptide libraries. The screening data was then integrated by a computational analysis, known as positional scanning based biometrical analysis 
, with a vaccinia specific protein database for the prediction of stimulatory peptides. Positional scanning libraries have been extensively and successfully used for the identification of T cell epitopes in a broad range of human diseases 
. The positional scanning library profile for each of the vaccinia clones revealed that the amino acids of the identified peptides in most of the positions correspond to the defined amino acids of the mixtures with the highest stimulatory potency (shown in Table S1
). The high correspondence on amino acids from active mixtures in the active peptides identified for the vaccinia clones is similar to previous results obtained with other pathogen specific T cell clones 
and confirms that the use of positional scanning libraries together with the biometrical analysis provides an efficient unbiased methodology for epitope and antigen identification.
Our previous studies on T cell pathogen specificity have shown that more than one peptide can be recognized by a single pathogen specific T cell clone 
. However, this was not the case in this work in which only one vaccinia epitope was found to stimulate each of the clones. A possible explanation for this finding is that we have used a vaccinia virus (Western Reserve) protein database instead of the complete viral database (Vaccinia: 216 proteins and 59,892 decapeptides versus Viral Genpept 156: 506,176 proteins and 125,055,759 decapeptides). Alternatively, the vaccinia clones presented here are intrinsically not highly cross-reactive and thus, independent of the database used for the biometrical analysis, there is only one (the peptides identified in this study) or very few optimal peptides capable of triggering their activation. Indeed, the biometrical analysis of the viral database revealed, for 3 of the 4 clones, that the identified vaccinia peptide in this work ranked within the top 30 peptides for the 125 million peptides scored (data not shown). Although all these peptides were not synthesized and tested, and thus their possible agonistic activity has not been determined, this finding strongly suggests that the clones are highly specific for the antigens reported here.
Importantly, the epitopes identified by positional scanning peptide libraries and their corresponding proteins are by definition immunogenic, since they have been derived by interrogating vaccinia specific T cell clones expanded in vivo during exposure to the vaccine, and hence relevant to the immune response. Other groups have also reported the clear advantages of using “T cell driven” approaches for the identification of antigens recognized by large size pathogens specific T cells. The T cell driven approach resulted in the identification of immunodominant epitopes for Mycobacterium tuberculosis
with a genome encoding for approximately 4,000 proteins 
. The methodology implemented in our study used a single in vitro
vaccinia virus stimulation of PBMC derived from smallpox immunized subjects in order to expand and clone vaccinia specific T cells. While it is possible that this stimulation may result in the expansion of cross-reactive T cells that were not originally triggered by the in vivo
vaccinia immunization, our data shows that vaccinia specific T cells are only expanded in vaccinia immunized donors and not in unvaccinated subjects ( and data not shown) indicating that the one time in vitro
expansion with the virus expands the repertoire triggered by immunization.
Four novel vaccinia epitopes and their corresponding vaccinia antigens () recognized by CD4+ T cells from humans immunized with the smallpox vaccine are presented in this study. D13L-YID and F13L-DWV proteins are from late expression membrane proteins, E1L-MYT from an early expression enzyme, and A6L-SFW is from a protein with unknown function and late expression. All four proteins have 100% homology within vaccinia and variola strains. In addition, and in agreement with recent findings by Jing et al showing that HLA-DR dominates the presentation of vaccinia antigens to CD4+ T cells 
, all four identified epitopes in this study were found to be HLA-DR restricted. The identification of the epitopes presented here used a T cell driven approach that has not been previously utilized to identify immunogenic pathogen proteins following vaccination in humans. This approach elucidates the peptides and viral antigens recognized by virus specific T cells using an unbiased collection of peptides (positional scanning libraries) presented by autologous LCLs. Thus, no previous knowledge about MHC restriction or antigen specificity is required.
To evaluate the performance efficacy of the T cell driven strategy presented here we compared the extent of recognition by antibodies and CD4+ T cells of each of the 181 vaccinia proteins derived from all reported human studies. Using separately the information from CD4+ T cell 
or antibody responses 
, we calculated a percentage of recognition for each vaccinia protein by dividing the number of subjects that showed protein recognition by the total number of subjects tested (76 or 126 for antibodies and 11 for T cells). Based on the overall distribution of the percentages of recognition and the number of subjects tested we established a cutoff of >40% and >10% for CD4+ T cell or antibody responses, respectively for a protein to be considered predominantly recognized. Using these thresholds 31 proteins were determined as predominantly recognized by CD4+ T cells and 21 by antibodies (data not shown, manuscript in preparation). shows the function, temporal expression 
as well as the percentage of recognition for each of the 10 proteins (4 reported here and manuscript in preparation) that we have identified using T cell driven approaches. Strikingly, 6 out of the 10 identified proteins (D13L, A7L, F13L, A6L, A10L and H5R) are within the 31 predominantly recognized proteins by CD4+ T cells (percentage of recognition >40%) and 5 (D13L, F13L, H5R, A10L and L1R) are within the 21 predominantly recognized proteins by antibodies (percentage of recognition >10%). The significance of our results can be statistically analyzed by comparing the probability of these findings to random sampling. If a random sample of 10 proteins were selected from the 180 proteins analyzed by CD4+ T cell responses, the probability that 6 or more would be among the 31 predominantly recognized proteins would be only 0.21%. Similarly, if a random sample of 10 proteins were selected from the 181 proteins analyzed by antibody responses, the probability that 5 or more would be among the 21 predominantly recognized proteins would be 0.23%. The percentage of predominantly recognized proteins that we identified is about 60%, which is clearly larger than the random probabilities, and indeed the above probabilities indicate that the method used in the present study yields results significantly better than random selection. These findings clearly reveal that the T cell driven approach used in the present study is very efficient in the identification of immunogenic and predominantly recognized proteins. Since the analysis of the whole pathogen proteome to identify relevant antigens might not be feasible for larger pathogens (>1,000 proteins), there is a clear advantage in using positional scanning libraries to elucidate the antigen specificity of the pathogen specific T cells.
Recognition of vaccinia proteins for which peptides have been identified using the “T cell driven” approach presented in this study.
In conclusion, we describe a methodology that can be implemented and applied to determine the specificity of the T cell response upon vaccination or infection with large size pathogens for which other methodologies would be more limiting. Together, the single pathogen in vitro stimulation, the selection of CD4+ T cells specific to the pathogen by limiting dilution, the evaluation of pathogen specificity by detecting multiple cytokines (including GM-CSF), and the screening of the clones with synthetic combinatorial libraries constitutes a novel and valuable approach for the elucidation of the CD4+ T cell specificity in response to large pathogens in human samples. We believe that further efforts on the development of miniature T cell activation assays will be extremely beneficial to reduce the number of clonal CD4+ T cells required to test combinatorial peptide libraries and to be able to define the spectrum of specificities with higher throughput.
Finally, our results support the notion that the cytokines selected to profile CD4+ T cell specific responses upon infection or vaccination should not rely only in IFN-γ secreting cells, but on the assessing of multi-cytokines at different time-points and by using different readout technologies. We believe that the vaccinia antigens identified in this study contribute to the knowledge base of the human immune response to vaccinia immunization and could lead to the development and evaluation of novel and safer smallpox vaccines.