Identification and Characterization of CD4⁺ T Cell Epitopes after Shingrix Vaccination

Understanding the T cell profile associated with the protection observed in elderly vaccinees following Shingrix vaccination is relevant to the general definition of correlates of vaccine efficacy. Our study enables these future studies by clarifying the patterns of immunodominance associated with Shingrix vaccination, as opposed to natural infection or Zostavax vaccination. Identification of epitopes recognized by Shingrix-induced CD4 T cells and their associated HLA restrictions enables the generation of tetrameric staining reagents and, more broadly, the capability to characterize the specificity, magnitude, and phenotype of VZV-specific T cells.

ABSTRACT

Infections with varicella-zoster virus (VZV) are associated with a range of clinical manifestations. Primary infection with VZV causes chicken pox. The virus remains latent in neurons, and it can reactivate later in life, causing herpes zoster (HZ). Two different vaccines have been developed to prevent HZ; one is based on a live attenuated VZV strain (Zostavax), and the other is based on adjuvanted gE recombinant protein (Shingrix). While Zostavax efficacy wanes with age, Shingrix protection retains its efficacy in elderly subjects (individuals 80 years of age and older). In this context, it is of much interest to understand if there is a role for T cell immunity in the differential clinical outcome and if there is a correlate of protection between T cell immunity and Shingrix efficacy. In this study, we characterized the Shingrix-specific ex vivo CD4 T cell responses in the context of natural exposure and HZ vaccination using pools of predicted epitopes. We show that T cell reactivity following natural infection and Zostavax vaccination dominantly targets nonstructural (NS) proteins, while Shingrix vaccination redirects dominant reactivity to target gE. We mapped the gE-specific responses following Shingrix vaccination to 89 different gE epitopes, 34 of which accounted for 80% of the response. Using antigen presentation assays and single HLA molecule-transfected lines, we experimentally determined HLA restrictions for 94 different donor/peptide combinations. Finally, we used our results as a training set to assess strategies to predict restrictions based on measured or predicted HLA binding and the corresponding HLA types of the responding subjects.

IMPORTANCE Understanding the T cell profile associated with the protection observed in elderly vaccinees following Shingrix vaccination is relevant to the general definition of correlates of vaccine efficacy. Our study enables these future studies by clarifying the patterns of immunodominance associated with Shingrix vaccination, as opposed to natural infection or Zostavax vaccination. Identification of epitopes recognized by Shingrix-induced CD4 T cells and their associated HLA restrictions enables the generation of tetrameric staining reagents and, more broadly, the capability to characterize the specificity, magnitude, and phenotype of VZV-specific T cells.

INTRODUCTION

Varicella-zoster virus (VZV), or human herpesvirus 3 (HHV3), is a double-stranded DNA (dsDNA) virus of about 125 kb and belongs to the Herpesviridae family. The corresponding proteome encompasses 69 proteins (1), of which 37, including the envelope, tegument, and capsid proteins, are structural proteins and 32 are nonstructural (NS) proteins involved in multiple functions, such as viral replication, innate host immunity evasion, and transcriptional regulation connected to the viral latency state (1). Three protein layers comprise a VZV particle: a nucleocapsid containing the dsDNA genome, a tegument layer, and an envelope lipid bilayer containing viral glycoproteins (1).

Primary infection generally occurs in children prior to adolescence in temperate climates but can occur at any age and clinically manifests as varicella or chicken pox, a mild to moderate disease, with symptoms appearing 10 to 21 days after exposure (2). The characteristic symptom of varicella is a pruritic vesicular rash appearing on the head, torso, and extremities, accompanied by additional symptoms, including headache, loss of appetite, tiredness, and fever (2). The host immune system controls, but does not eliminate, the virus, as VZV enters a lifelong latency state mainly located in the sensory neurons of the trigeminal and dorsal root ganglia but also in other sensory neurons, as well as autonomic and other cranial nerve and enteric ganglia (1). The reactivation of latent virus is the cause of a second disease state of herpes zoster (HZ), or shingles, characterized by a more localized infection (1, 3, 4). Postherpetic neuralgia (PHN) is a very common neurological complication of HZ that can persist for years and occurs in 26 to 32% of cases. Age is the most important risk factor for PHN, with the risk increasing rapidly after 50 years of age (5). It has been estimated that half of the subjects reaching age 85 years will experience HZ reactivation at least once in their lifetime (2). While this estimation may change depending on the population studied, overall, this highlights the need for efficient control of this disease.

The mechanism of protection from HZ reactivation is unclear, but cell-mediated immunity (CMI) has been shown to be integral to HZ prevention. Additionally, HZ severity has been clinically observed to be associated with a reduced magnitude of VZV-specific effector and effector memory T cells (6).

Three vaccines are currently licensed in the United States for use in preventing VZV-associated diseases. Varivax, a live attenuated vaccine for varicella immunity, was first licensed in the United States in 1995 and is part of the recommended childhood vaccine schedule. In 2005, two doses of this vaccine during childhood were recommended in the United States (2). Two vaccines have been developed to protect against herpes zoster (HZ): Zostavax, a live attenuated vaccine derived from the Oka virus strain, and Shingrix, a vaccine based on recombinant gE protein. Zostavax was licensed in 2006 for people age 60 years and older (2). While vaccine efficacy is often a function of the age of the vaccinee (vaccine efficacy is less if the vaccinee is older), waning of vaccine-induced immunity after vaccination (regardless of age) also occurs over time. Some effectiveness studies have shown less of an age effect on the initial response and less waning over time. In this context, several studies have evaluated the efficacy of Zostavax in preventing HZ, with some initially hypothesizing that the response was reduced as a function of the age of the vaccinee. Indeed, an age effect ranging from 70% efficacy in vaccinees ages 50 to 59 years to a mere 18% efficacy in vaccinees 80 years old or older has been noted. Further, more recent studies have shown an overall waning of immunity after 4 years, regardless of age (7), and a very recent study performed in Sweden showed that the effectiveness of Zostavax was reduced to 34%, and protection, observed in the 61- to 75-year-old age group, was largely absent in individuals age 75 years old and older (8). Overall, Zostavax vaccination was shown to be significantly less effective than Shingrix vaccination, and for this reason, starting in July 2020, Zostavax was no longer available (6, 9). The Shingrix vaccine was recently licensed for use in adults age 50 years and older, and efficacy trials in adults have demonstrated protection regardless of age, with 91% efficacy being observed in vaccinees age 80 years old or older (10). The efficacy of Shingrix at older ages is yet to be fully understood, and additional studies dissecting the responses triggered by vaccination with this vaccine are important for developing successful vaccines against other pathogens for the elderly population.

The recombinant VZV gE protein is coformulated with the novel ASB01 adjuvant to induce both a strong CD4⁺ T cell response and a gE-specific antibody response (10, 11). VZV-specific CD4⁺ T cell responses are associated with positive vaccination outcomes (12–15). The reasons for the remarkable difference in efficacy between Shingrix and Zostavax may involve a capacity for Shingrix to elicit a stronger and/or a more focused response against the gE antigen, a broader epitope repertoire, or an intrinsic difference in the quality and durability of the responses elicited. Several studies have indicated changes in the quality of the specific CD4⁺ T cell response toward a predominantly and persistently polyfunctional response after vaccination (10, 16, 17).

While studies have been conducted to assess the immunogenicity of different VZV proteins (18) and the immunological responses to VZV vaccinations have been described elsewhere (14, 19–21), little is still known regarding the epitopes recognized by T cells after VZV vaccination. Identifying immunogenic epitopes and designing tetramers specific to VZV epitopes will allow further characterization of the polyfunctional response observed in Shingrix vaccinees and may help establish correlates of protection for HZ patients.

Here, we assessed the repertoire of CD4⁺ T cell responses induced by Shingrix vaccination and define a set of epitopes broadly recognized in Shingrix vaccinees. Further, we used both predictive binding methods and reductive HLA restriction assays with identified VZV-specific CD4⁺ T cell epitopes to identify potential tetrameric reagents and immunomonitoring strategies to address the mechanisms of cell-mediated immunity and effective vaccination strategies.

RESULTS

Recruitment of the childhood exposure, postshingles, and Zostavax and Shingrix vaccinee donor cohorts.

To characterize CD4⁺ T cell responses following Shingrix vaccination, we enrolled 18 Shingrix vaccinees. These vaccinees were further subclassified, depending on their status before Shingrix vaccination, as those receiving a Shingrix vaccination only (n = 10), a Shingrix vaccination after shingles (n = 2), and a Shingrix vaccination after Zostavax vaccination (n = 5); an additional donor who had recently been vaccinated with Shingrix but who had previously been vaccinated with Zostavax and who had developed shingles was also included. To put our observations in a broader context of other HZ-related conditions, we also enrolled three independent cohorts of donors to represent individuals with childhood exposure (n = 18), Zostavax vaccinees (n = 18), and individuals with clinically diagnosed shingles with a lack of vaccination (in childhood or otherwise) (n = 18). The childhood exposure cohort encompassed donors who, based on their date of birth, did not receive a childhood vaccination and who were likely to have experienced clinical chicken pox; as such, they represented individuals who had had a natural exposure to VZV and who had never experienced shingles. A total of 72 donors participated in this study. Ethnicity, gender, average age, and the time from the most recent clinical event (Shingrix or Zostavax vaccination or HZ reactivation) are summarized in Table 1. The cohorts enrolled in the present study were based on a pool of volunteers in the San Diego, CA, area. Despite efforts to enroll a cohort as balanced as possible in terms of gender, ethnicity, and age range, some biases toward males and Caucasians and a significant difference in age (P < 0.0001, Kruskal-Wallis test) were observed across cohorts. Because Zostavax vaccination was not available as of July 2020 and a recommendation toward Shingrix vaccination has been in place since early 2019, it was not possible to analyze individuals who had received one of the two HZ vaccinations at the same time. This, as a result, generated a significant difference between the time of exposure and the time of sample collection between individuals with Zostavax vaccinations and individuals with Shingrix vaccinations (P < 0.0001, Mann-Whitney test).

TABLE 1.

General characteristics of donor cohorts

Cohorta	Caucasian (%)	Female (%)	Avg ± SD age (yr)	Avg ± SD time from exposure (days)
Childhood exposure	61	8	51 ± 17	NAb
Shingles	44	13	44 ± 12	1,686 ± 2,305
Shingrix vaccinees	72	10	62 ± 8	123 ± 92
Zostavax vaccinees	100	12	69 ± 7	1,849 ± 1,112

^aEach cohort contained 18 individuals.

^bNA, not applicable.

General features of VZV responses in the various cohorts.

The study of responses in different settings and geographical locations is important in the context of vaccination targeting different human populations, as different HLA alleles influencing T cell immune responses might be prevalent in different populations. Additionally, to be able to analyze unaltered T cell phenotypes, it is important to study T cell reactivity ex vivo. However, testing T cell responses ex vivo against each single peptide, while ensuring comprehensive protein and HLA coverage, requires a volume of blood not often available for human studies. To meet these challenges, we previously developed the megapool (MP) approach, based on the pooling of large numbers of peptides, in which the MP is formulated to consider sequential lyophilization. These MPs have been used in multiple studies in the context of a number of different indications, from allergy to infectious diseases and vaccination (22–26).

Shingrix vaccination is based only on the recombinant gE protein, while in the context of natural exposure, as well as after a vaccination with a live attenuated vaccine, an individual is exposed to the entire VZV proteome. To address the differences in terms of T cell reactivity across the different cohorts, we designed megapools targeting the gE protein specifically versus the rest of the VZV proteome. In the case of the gE protein, 123 peptides (15-mers overlapping by 10 residues) spanning the gE protein were arranged into a single megapool. The gE protein sequence of the Dumas strain is 100% identical to the gE protein of the Oka strain used in Zostavax.

An additional 302 peptides from the remaining non-gE proteins, derived from the entire VZV proteome, were investigated to identify CD4⁺ T cell reactivity outside the gE protein. These included 44 epitopes from the published literature curated in the Immune Epitope Database (IEDB; www.IEDB.org) (27) and 258 predicted HLA class II dominant epitopes from the Dumas strain reference sequence (28). To ensure that the pools of peptides tested were based on a similar number of peptides, the non-gE peptides were arranged into two different megapools, with one corresponding roughly to the structural (S) proteins (but excluding gE) and the other corresponding to the nonstructural (NS) proteins, as previously described (29). The full list of peptides contained in the non-gE megapools (S and NS proteins), as well as additional information regarding the corresponding peptide start position, protein of provenance, and open reading frame (ORF) of origin, is provided in Table S1 in the supplemental material.

VZV CD4⁺ T cell responses in natural infection and following HZ vaccination have been shown to elicit a broad spectrum of cytokines (20, 30, 31). This suggests that a more comprehensive cytokine-independent strategy is necessary to dissect CD4⁺ T cell-specific responses across cohorts and to compare the antigens recognized. For this purpose, we applied a T cell receptor-dependent activation-induced marker (AIM) assay (24, 32) by analyzing the combined membrane markers for expression of 4-1BB (CD137; TNFRS9), an activation-induced costimulatory molecule, and OX40 molecules within the T cell’s CD4⁺ compartment after 24 h of stimulation with the gE, S, and NS megapools (MP). Figure 1 shows in detail the gating strategy approach applied.

FIG 1.

Representative gating strategy. The numbers near the bottoms of the panels represent the percentages of OX40⁺ CD137⁺ CD4⁺ T cells as a function of the cohort studied and the specific stimulation. SSC-A, side scatter area; FSC-A, forward scatter area; SSC-H, side scatter height; FSC-H, forward scatter height; LD, live/dead.

Strong responses to NS proteins were noted in the case of the childhood exposure and shingles cohorts (P = 0.0004 and P < 0.0001, respectively, compared to the unstimulated [dimethyl sulfoxide {DMSO}-treated] control determined by a nonparametric, paired Wilcoxon test). These responses were higher than those observed when S proteins were considered (P = 0.0040 and 0.0002 for the childhood exposure and shingles cohorts, respectively) (Fig. 2A).

FIG 2.

VZV-specific CD4⁺ T cell reactivity in individuals with childhood exposure, shingles, and HZ vaccinations. The frequency of CD137⁺ OX40⁺ CD4⁺ T cells (see Fig. 1 for the gating strategy) was assessed for donor/MP stimulation combinations and compared to that for a paired DMSO-treated negative control. (A) Results for the childhood exposure (black circles, n = 18) and shingles (blue circles, n = 18) cohorts. (B) Results for the Shingrix (orange circles, n = 18) and Zostavax (gray circles, n = 18) vaccinee cohorts. Data are expressed as the geometric mean and geometric standard deviation (SD). All reported statistical analyses were performed using a nonparametric, paired Wilcoxon test. (C) Frequency of CD137⁺ OX40⁺ CD4⁺ T cells from which the background value was subtracted, analyzed to compare the gE, structural (S), and nonstructural (NS) MPs across the cohort. The comparison across cohorts was performed using the Kruskal-Wallis test. (D) Spearman correlation of the frequency of CD137⁺ OX40⁺ CD4⁺ T cells from which the background was subtracted, based on donor age after stimulation with structural (S) and nonstructural (NS) MPs. ns, not significant; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

When the responses of the two vaccinated cohorts were examined, Zostavax and Shingrix vaccinations were found to be associated with significant responses to the various peptide pools. As expected, the responses toward gE in Shingrix vaccinees were remarkably vigorous (geometric mean = 0.18 for Shingrix vaccinees [P < 0.0001] compared to that by the unstimulated [DMSO-treated] control group) (Fig. 2B). These results confirmed that Shingrix vaccination induces a strong CD4⁺ T cell response directed against the gE antigen. We then investigated whether we could observe a difference in protein reactivity across the cohorts analyzed (Fig. 2C). Comparisons across cohorts were performed with background data subtracted, and, interestingly, no systematically significant differences between the structural and NS protein pools were observed (for the S protein pool, P = 0.1491; for the NS protein pool, P = 0.8708 [P values were determined by the Kruskal-Wallis test]), while, as expected, a significant difference across cohorts was observed in the case of the gE protein, driven by Shingrix vaccination (for gE, P < 0.0001 by the Kruskal-Wallis test) (Fig. 2C). When we specifically analyzed the S and NS proteins in Shingrix and Zostavax vaccinees to determine whether these responses were toward similar epitope repertoires, no significant differences were observed (for Shingrix versus Zostavax vaccinees P = 0.6337 for S proteins and P = 0.4381 for NS proteins; P values were determined by the Mann-Whitney test).

To understand if the similarity in protein reactivity observed across cohorts was due to the fact the cohorts had a significantly different age range, we performed a correlation of the CD4⁺ T cell reactivity against structural and NS MPs with the age of the subjects and found no significant correlation (for the structural protein MP, R = 0.1844 and P = 0.1209; for the nonstructural proteins MP, R = 0.0659 and P = 0.5826) (Fig. 2D).

Immunodominant gE regions recognized specifically by Shingrix vaccinee donors.

Next, we defined the gE epitopes recognized in the Shingrix vaccinee cohort described above (Table 1) and performed donor HLA typing (Table S2) to enable determination of the HLA restriction of any responding epitope. As we wanted to ensure that our cohort was representative of the general population, we compared the frequencies of the main HLA class II alleles observed in our cohort with those observed in our repository of over 3,500 donors (Fig. 3A and Table 2). This repository is inclusive of donors from a variety of clinical studies and representative of a diverse set of ethnicities and populations ranging from the United States to South and Central America, Asia, South Africa, and Europe and therefore is a reasonable representation of the general population, with the further advantage of being HLA typed using the same methodology. Of the 28 different HLA class II alleles with phenotypic frequencies of >10% in our cohort, 20 (71%) were also present in the general population with frequencies of >10%. Additionally, 24 of the 27 (89%) alleles included in a reference panel of the most common and representative class II alleles in the general population (33) were also present in the current cohort (Fig. 3A).

FIG 3.

Epitope mapping of Shingrix vaccinees. (A) HLA frequency of the Shingrix vaccinee cohort (n = 18) compared to the HLA frequency of the general population (n = 3,500). (B) Experimental scheme. The results for a representative donor with mesopool reactivity at day 14, peptide deconvolution of one of the positive mesopools at day 17, and relative FluoroSpot images for IFN-γ, IL-5, and TNF-α are shown. P1 to P12, passages 1 to 12, respectively; PHA, phytohemagglutinin; LED, light-emitting diode. (C) Immunodominant regions of the gE protein showing the total magnitude of responses and the number of responding subjects.

TABLE 2.

HLA class II allele frequency in Shingrix vaccinee cohort compared to the general population

HLA class II locus	Allele	Phenotype frequency
		Shingrix vaccinee cohort	General population
DR	DRB1*01:01	11.8	9.1
	DRB1*01:02	11.8	3.6
	DRB1*03:01	11.8	11.1
	DRB1*03:02	5.9	1.8
	DRB1*04:01	11.8	6.2
	DRB1*04:02	5.9	1.7
	DRB1*04:05	11.8	2.7
	DRB1*07:01	47.1	29.0
	DRB1*08:01	5.9	1.2
	DRB1*09:01	11.8	3.7
	DRB1*11:01	5.9	8.8
	DRB1*13:01	11.8	9.3
	DRB1*13:03	5.9	1.3
	DRB1*14:01	11.8	2.3
	DRB1*15:01	23.5	18.4
	DRB3*01:01	23.5	14.6
	DRB3*02:02	23.5	33.5
	DRB4*01:01	70.6	55.1
	DRB5*01:01	23.5	25.0
DQ	DQB1*02:01	11.8	12.3
	DQB1*02:02	35.3	18.7
	DQB1*03:01	29.4	29.0
	DQB1*03:02	11.8	19.8
	DQB1*03:03	17.6	12.0
	DQB1*04:02	11.8	7.4
	DQB1*05:01	17.6	20.2
	DQB1*05:03	11.8	10.7
	DQB1*06:02	23.5	15.8
	DQB1*06:03	11.8	10.2
DP	DPB1*01:01	5.9	17.0
	DPB1*02:01	35.3	23.4
	DPB1*03:01	23.5	11.6
	DPB1*04:01	35.3	47.5
	DPB1*04:02	35.3	18.2
	DPB1*05:01	5.9	15.0
	DPB1*14:01	5.9	0.2
	DPB1*15:01	5.9	1.0
	DPB1*17:01	5.9	6.5

The gE epitope identification strategy is summarized in Fig. 3B and C. To identify individual epitopes, peripheral blood mononuclear cells (PBMCs) were expanded with the gE MP for 14 to 17 days. On day 14, the gE megapool and smaller pools composed of 10 or 11 peptides each (mesopools) were tested for cytokine reactivity. The individual peptides of the positively identified mesopools were tested for cytokine reactivity on day 17. On both days 14 and 17, reactivity was assessed using the FluoroSpot assay (Fig. 3B). It has previously been shown that the ASB01 adjuvant, in general, induces a mixed Th1/Th2 T cell phenotype (34). Therefore, to comprehensively identify the epitopes induced after Shingrix vaccination, while disregarding T cell polarization, we measured gamma interferon (IFN-γ), tumor necrosis factor alpha (TNF-α), and interleukin-5 (IL-5) levels in the FluoroSpot assays by counting the single, double, and triple producers only once, to avoid inflating our results.

Overall, we identified 89 epitopes recognized in at least one donor. When the response frequency (RF) and magnitude data were mapped to the gE sequence, several immunodominant regions were identified (Fig. 3C). The responses were found to be dispersed throughout the protein, with the exception of the 110 C-terminal residues, which were recognized poorly, if at all. A full list of the epitopes identified in this study is available in Table S3.

Comparison with the repertoire of epitopes recognized in different donor cohorts.

The data presented in Fig. 2 indicate that Shingrix vaccinees had a stronger gE response than other cohorts of VZV-exposed/vaccinated individuals, and several epitopes were recognized per donor (Fig. 3C). While one explanation for this observation is that the difference in the amount of time that had elapsed since vaccination may also impact the magnitude of the immune response, those two points raised the question of whether the higher Shingrix response was due to a stronger response to a similar number of epitopes or responses of a similar magnitude directed against a larger number of epitopes. To address this, we conducted pilot deconvolution experiments in two donors from the Zostavax vaccinee cohort and three donors from the childhood exposure cohort.

First, we inspected the overlap in repertoires between the Shingrix vaccinee and Zostavax vaccinee/childhood exposure donors. Of the 89 epitopes identified in samples from the Shingrix vaccinee cohort, 42 (47%) were also recognized in the samples from the Zostavax vaccinee/childhood exposure cohort. Conversely, only 5 (5%) of the peptides not recognized in samples from Shingrix vaccinees were recognized in Zostavax vaccinee/childhood exposure samples (P > 0.0001 by the chi-square test) (Fig. 4A). A Venn diagram is shown in Fig. 4B to better illustrate instances of overlap and a lack of overlap. This indicated that the epitope repertoire recognized in the Shingrix and Zostavax/childhood exposure donors is similar in nature, although Shingrix vaccination appears to induce a broader epitope repertoire for the gE protein, but, again, heeding the caveat that differences in the response breadth and magnitude may be impacted by differences in the amount of time that has elapsed since vaccination. In support of this observation, Shingrix vaccinees responded to an average of 18 ± 10 epitopes, with an average magnitude of 2,041 ± 1,673 spot-forming cells (SFC)/10⁶ PBMCs per response. In contrast, the Zostavax vaccinee/childhood exposure donors responded to 10 ± 8 epitopes, with an average magnitude of 1,353 ± 1,080 SFC/10⁶ PBMCs per response (Fig. 4C and D). Thus, although these differences are not significant, Shingrix vaccinee donors tended to have a broader epitope repertoire and responses of a higher magnitude than the Zostavax vaccinee/childhood exposure cohort.

FIG 4.

Shingrix-vaccinated patients show a trend for recognizing more epitopes and producing a stronger response than the other patients. (A) Number of recognized epitopes over nonrecognized ones in the Shingrix vaccinee donors, Zostavax vaccinee/childhood exposure donors, or both groups of donors (****, P < 0.0001, chi-square test). (B) Venn diagram showing the numbers of epitopes overlapping across the Shingrix vaccinee, Zostavax vaccinee, and childhood exposure donors. (C) Number of recognized epitopes per donor cohort (ns, not significant [P = 0.3095]). (D) Magnitude of responses per donor cohort (ns, not significant [P = 0.7743]). (C and D) The Mann-Whitney test was applied for statistical comparisons across cohorts.

Breadth and magnitude of gE-specific CD4⁺ T cells.

As mentioned above, of the 123 gE peptides tested, 89 were recognized in at least one Shingrix vaccinee, while only 34 were not recognized in any context (Fig. 4A). Of the 89 epitopes inducing responses, 26 epitopes were recognized in only one donor, 19 epitopes were recognized in two donors, and the remaining 44 were recognized in three or more donors (Fig. 5B). In terms of the magnitude of the response, the top 34 epitopes recognized in four or more donors were found to account for 80% of the total SFC response (Fig. 5C).

FIG 5.

Shingrix vaccination is associated with a broad CD4⁺ T cell response. (A) The numbers of responding donors are plotted against the number of epitopes recognized. (B) The total numbers of epitopes recognized are plotted against the percentage of the total magnitude of the response (sum of total SFC for total responding donors).

Table 3 lists these top 34 epitopes eliciting a response in four or more donors. Table 3 also details the number of positive responses, the sum of peptide-specific SFC across donors, the corresponding percentage of total SFC detected in the cohort, the cumulative number of SFC, and the response frequency score, calculated as described in Materials and Methods. Overall, these data indicate that while Shingrix vaccinees are associated with a broad and diverse repertoire, a limited number of epitopes are dominant and account for the preponderance of the response. When we compared the 34 immunodominant epitopes characterized in Shingrix vaccinee donors (Table 3), we found that seven were also recognized by Zostavax vaccinees, although they were not among the most immunodominant epitopes.

TABLE 3.

Most commonly recognized gE epitopes among Shingrix vaccinees

Start position	gE sequence	No. of positive donors/no. of donors tested	Total no. of SFCa	% total SFC	Cumulative % of SFC	RFb score
281	IEPGVLKVLRTEKQY	14/17	51,227	5.5	5.5	0.603
191	IYGVRYTETWSFLPS	13/18	63,696	6.8	12	0.522
186	APIQRIYGVRYTETW	12/18	32,754	3.5	16	0.474
291	TEKQYLGVYIWNMRG	11/17	17,568	1.9	18	0.452
421	AQRVASTVYQNCEHA	11/18	57,444	6.2	24	0.427
356	TFSLAMHLQYKIHEA	10/17	38,384	4.1	28	0.402
286	LKVLRTEKQYLGVYI	10/17	33,867	3.6	32	0.402
196	YTETWSFLPSLTCTG	10/18	32,942	3.5	35	0.380
486	VEAVAYTVVSTVDHF	9/17	43,390	4.6	40	0.353
476	YVFVVYFNGHVEAVA	8/17	30,807	3.3	43	0.304
51	YEPYYHSDHAESSWV	8/17	17,507	1.9	45	0.304
446	SHMEPSFGLILHDGG	8/18	27,209	2.9	48	0.287
31	SVLRYDDFHIDEDKL	8/18	23,577	2.5	50	0.287
341	FHMWNYHSHVFSVGD	7/17	19,080	2.0	52	0.256
386	TCQPMRLYSTCLYHP	6/17	21,414	2.3	55	0.209
496	TVDHFVNAIEERGFP	6/17	19,377	2.1	57	0.209
141	DDRHKIVNVDQRQYG	6/17	16,362	1.8	59	0.209
181	PFTLRAPIQRIYGVR	6/18	15,142	1.6	60	0.197
26	NPVRASVLRYDDFHI	6/18	13,123	1.4	62	0.197
36	DDFHIDEDKLDTNSV	6/18	9,080	1.0	63	0.197
221	LKHTTCFQDVVVDVD	5/17	27,263	2.9	65	0.163
336	PRGAEFHMWNYHSHV	5/17	16,907	1.8	67	0.163
391	RLYSTCLYHPNAPQC	5/17	12,968	1.4	69	0.163
146	IVNVDQRQYGDVFKG	5/17	7,440	0.8	69	0.163
351	FSVGDTFSLAMHLQY	5/17	4,745	0.5	70	0.163
416	TSPHLAQRVASTVYQ	5/18	6,103	0.7	71	0.154
501	VNAIEERGFPPTAGQ	4/17	11,977	1.3	72	0.118
491	YTVVSTVDHFVNAIE	4/17	10,803	1.2	73	0.118
481	YFNGHVEAVAYTVVS	4/17	9,727	1.0	74	0.118
136	PTLNGDDRHKIVNVD	4/17	8,903	1.0	75	0.118
471	SLSGLYVFVVYFNGH	4/17	3,957	0.4	75	0.118
41	DEDKLDTNSVYEPYY	4/18	21,923	2.3	78	0.111
436	DNYTAYCLGISHMEP	4/18	9,900	1.1	79	0.111
176	VEENHPFTLRAPIQR	4/18	5,198	0.6	79	0.111

^aSFC, spot-forming cells.

^bRF, response frequency.

HLA-binding capacity of dominant epitopes.

To characterize the dominant epitopes in more detail, we conducted HLA/epitope binding assays, using purified HLA molecules in vitro, as described in Materials and Methods. Table 4 presents the results of an analysis of a set of 19 common HLA class II alleles selected on the basis of worldwide frequencies (33). Actual binding data were generated for 28 of 34 of the epitopes in Table 3. Six additional epitopes (epitopes 136, 146, 176, 221, 471, and 481) were not available when the binding assays were performed, and in those cases, HLA binding predictions were used instead, using the NetMHCIIpan algorithm (35). The use of HLA binding predictions for those peptides was supported by the strong correlation detected for predicted versus actual measured binding (Fig. 6) in the cases in which both experimental and predicted binding values were determined.

TABLE 4.

Observed in vitro binding capacity of 19 common HLA alleles to epitopes commonly recognized among Shingrix vaccinees^a

Start position	Sequence	No. of bound peptidesb	IC50 (nM)c
			DQA1/B1				DRB1											DRB3/4/5
			5:01/2:01	3:01/3:02	1:01/5:01	1:02/6:02	1:01	3:01	4:01	4:05	7:01	8:02	9:01	11:01	12:01	13:02	15:01	B3*01:01	B3*02:02	B4*01:01	B5*01:01
26	NPVRASVLRYDDFHI	8	3,623	2,232	65	110	10,732	9,222	5,950	11,471	1,964	—d	417	—	8,460	454	2.8	494	34,700	414	462
31	SVLRYDDFHIDEDKL	5	368	2,547	9	13,217	—	3,824	178	2,410	—	36,884	7,252	—	—	1,996	13	495	—	1,140	16,474
36	DDFHIDEDKLDTNSV	3	1,314	29,358	586	—	—	2,960	1,393	189	—	—	35,428	—	—	18,728	1,555	1,276	—	155	—
41	DEDKLDTNSVYEPYY	1	705	3,321	14,664	14,865	10,147	14,990	8,611	8,484	—	34,714	—	—	—	6,424	3,750	18,708	—	7,851	35,316
51	YEPYYHSDHAESSWV	7	249	11,453	1,048	7,333	620	10,926	21	10,156	2,476	972	150	4,873	—	—	553	164	10,958	—	9,457
136	PTLNGDDRHKIVNVD	1	13,356	16,063	18,153	12,144	4,878	235	6,667	9,636	8,633	8,658	8,278	4,653	11,337	1,636	7,104	1,147	6,835	3,630	3,540
141	DDRHKIVNVDQRQYG	6	16,483	—	—	36,976	28,740	1,473	934	12,276	20,608	7,017	119	13,429	358	976	952	2,208	8,380	170	3,539
146	IVNVDQRQYGDVFKG	1	6,345	11,676	6,380	9,447	4,206	159	6,015	6,701	8,397	6,221	5,678	5,966	5,079	3,388	4,922	1,957	11,176	2,153	5,339
176	VEENHPFTLRAPIQR	5	4,968	5,186	7,261	1,746	246	3,585	1,975	2,184	1,209	1,398	997	600	4,161	1,385	1,933	7,537	4,013	769	156
181	PFTLRAPIQRIYGVR	11	4,367	—	14,657	5,264	56	17,788	1,421	5,434	845	48	60	949	593	55	21	5,565	996	96	11
186	APIQRIYGVRYTETW	8	174	10,539	7,410	15,811	1,403	11,307	3,525	6,552	183	1,567	880	5,105	194	684	5	8,514	11,715	447	117
191	IYGVRYTETWSFLPS	11	245	3,883	2,471	8,109	102	27,309	215	625	19	15,656	13	5,158	1,522	174	108	211	10,142	355	115
196	YTETWSFLPSLTCTG	9	634	4,134	381	8,210	20	38,807	40	149	154	2,654	49	1,259	1,855	—	574	26,572	19,164	14,325	428
221	LKHTTCFQDVVVDVD	3	263	1,227	851	1,850	2,141	1,718	2,097	1,528	2,659	6,991	3,029	9,804	5,953	2,808	5,264	547	9,101	3,230	9,423
281	IEPGVLKVLRTEKQY	11	19,213	18,164	—	2,334	82	21,908	83	202	9,613	223	447	95	464	5,947	250	—	980	1.7	100
286	LKVLRTEKQYLGVYI	10	677	8,649	2,165	6,510	197	4,210	58	1,137	1,265	452	491	1,062	260	2,171	25	9,580	882	1.3	837
291	TEKQYLGVYIWNMRG	10	35,859	1,650	958	1,071	543	—	230	240	669	2,172	602	1,159	—	10,843	0.87	933	2,633	295	124
336	PRGAEFHMWNYHSHV	5	—	—	2,913	—	1,067	—	710	3,232	845	1,008	227	4,167	—	703	2.1	1,997	4,486	1,307	1,288
341	FHMWNYHSHVFSVGD	7	—	4,505	1,254	30	1,120	—	833	1,538	2.5	104	4.2	1,213	—	15,002	6.7	3,858	—	1,316	809
351	FSVGDTFSLAMHLQY	7	7,174	4,629	5,112	95	1,310	29,944	4,939	20,004	673	243	20	349	13,509	—	26	8,016	—	99	6,730
356	TFSLAMHLQYKIHEA	8	7,477	—	44	3,091	1,221	—	1,118	1,928	29	274	11	500	13,602	22,315	6.3	1,729	32,786	204	158
386	TCQPMRLYSTCLYHP	3	—	—	1,561	2,895	526	—	1,065	1,495	1,986	4,466	382	29,468	4,733	—	14	3,375	15,526	2,117	5,306
391	RLYSTCLYHPNAPQC	2	12,424	24,467	1,263	3,646	2,078	—	431	1,640	8,632	10,285	2,133	11,791	—	28,343	452	2,515	3,744	—	9,370
416	TSPHLAQRVASTVYQ	8	10,103	32,230	—	270	76	—	72	485	193	2,756	128	—	36,480	1,948	99	—	1,911	380	2,390
421	AQRVASTVYQNCEHA	1	6,622	20,639	—	4,719	21,756	—	5,348	5,717	12,948	—	10,260	—	—	1,886	260	—	—	—	21,713
436	DNYTAYCLGISHMEP	4	904	2,657	723	3,245	1,782	—	1,758	124	8,313	1,209	164	10,985	—	—	1,189	29,400	—	7,038	5,268
446	SHMEPSFGLILHDGG	7	3,667	11,444	2,527	547	223	5,435	176	6,685	193	1,099	12	3,601	—	968	9.1	1,577	19,700	3,678	1,015
471	SLSGLYVFVVYFNGH	2	3,316	6,809	1,382	1,926	696	13,342	4,047	2,121	1,643	4,501	1,901	4,324	2,346	4,869	811	7,255	14,123	3,739	5,531
476	YVFVVYFNGHVEAVA	5	1,152	5,046	1,251	1,519	134	—	195	1,017	2,044	4,550	85	15,029	—	138	1.3	3,018	12,800	9,167	10,118
481	YFNGHVEAVAYTVVS	4	1,036	1,303	3,416	262	127	7,849	1,276	1,553	283	2,085	208	3,879	2,968	1,445	2,064	3,488	4,579	2,657	1,882
486	VEAVAYTVVSTVDHF	10	894	2,086	1,586	2,693	1,108	31,488	72	53	96	696	148	8,329	—	138	95	1,581	10,737	153	531
491	YTVVSTVDHFVNAIE	10	515	604	69	16,277	388	12,518	930	69	109	3,761	243	—	—	6,708	107	8,010	10,222	912	4,935
496	TVDHFVNAIEERGFP	11	63	249	598	343	103	1,661	63	369	5,244	1,608	230	724	2,987	6,998	433	10,694	7,314	219	1,666
501	VNAIEERGFPPTAGQ	3	184	6,672	10,872	9,640	3,732	—	6,381	16,505	—	—	12,975	—	—	—	455	—	—	872	—

^aPredictive data are shown for italicized rows.

^bThe number of bound peptides per each HLA allele with an IC₅₀ of <1,000 nM.

^cBoldface data indicate bound peptides per each HLA allele with an IC₅₀ of <1,000 nM.

^d—, IC₅₀ of >40,000 nM.

FIG 6.

Correlation of predicted versus experimentally measured binding. Experimental (measured) HLA binding (x axis) compared to the predicted HLA binding (NetMHCIIpan [version 3.2] algorithm [35]) is shown, where binding was determined from the IC₅₀ (in nanomolar). The Pearson correlation was applied (r = 0.9689, P < 0.0001).

Most of the dominant epitopes were found to be promiscuous HLA class II binders, with 50% binding to 7 or more alleles in the panel of 19 common specificities considered and the binding of 6.1 ± 3.4 alleles, on average. In contrast, negative peptides (i.e., peptides not recognized in any donor) bound an average of 2.1 ± 3.6 alleles in the same panel of HLA variants.

HLA restriction determinations.

Next, we determined HLA restrictions for selected epitopes by the use of single HLA-transfected cell lines. Briefly, T cell lines (TCL) generated by 14 day of in vitro restimulation with an individual peptide were tested in antigen presentation assays, where different cell lines expressing a single HLA class II molecule were used as antigen-presenting cells (APCs) (36), as detailed in Materials and Methods. Representative results are shown Fig. 7A to C. When a TCL derived from a donor responding to the peptide from positions 446 to 460 of gE (referred to as the gE 446–460 peptide; sequence, SHMEPSFGLILHDGG) was tested with single HLA-transfected APC lines matching the HLA type of the donor, a good response was observed in the case of the HLA DPB1*0402 APCs. No response was noted for any of the other APC lines matching the donor’s HLA type. Accordingly, the gE 446–460 peptide was determined to be DPB1*0402 restricted in this donor. Similarly, the gE 86–100 and 486–500 peptides (sequences, NDYDGFLENAHEHHG and VEAVAYTVVSTVDHF, respectively) were identified to be putatively restricted by DQB1*0302 and DRB1*0701 in other representative donors (denominated “positive restrictions”). Conversely, the HLA/epitope combinations tested in the same assays and found to be negative were denominated “negative restrictions.”

FIG 7.

Donor HLA/epitope restrictions identified by HLA restriction assay. (A to C) Representative examples showing determination of HLA restriction for three epitopes in three donors. (D) Summary of positive and negative epitope/HLA restrictions identified.

The results of the experiments are presented in Table 5 and summarized in Fig. 7D. Overall, we determined positive restriction for 38 peptides and 94 donor/peptide combinations (listed in Table S3). Most positive restrictions were associated with the DRB1 and DQ loci, consistent with the findings of previous studies (37). Not all restrictions could be determined unequivocally, as APCs corresponding to some of the HLA alleles were not available because of limitations in cell availability or other technical issues. In addition to testing as many of the peptides in Table 3 as possible, we also tested certain peptides that were positive in only one or a few donors but that yielded vigorous responses. Table 5 presents a summary of the restrictions determined, including the sequence, the percentage of the response, the number of donors recognized, and the associated HLA restriction(s).

TABLE 5.

Summary of determined epitope/HLA restrictions

Start position	gE sequence	Sum of SFCa	Sum of % of response	Restricted HLA class II allele(s)
186	APIQRIYGVRYTETW	11,446.67	1.59	DRB1*07:01, DRB3*02:02
421	AQRVASTVYQNCEHA	33,653.33	4.67	DRB1*07:01, DPA1*01:03/DPB1*04:01, DRB1*15:01, DPA1*01:03/DPB1*02:01, DRB4*01:01, DPA1*03:01/DPB1*04:02, DRB5*01:01
36	DDFHIDEDKLDTNSV	600.00	0.08	DRB1*13:03
41	DEDKLDTNSVYEPYY	1,406.67	0.19	DPA1*01:03/DPB1*02:01
76	DHNSPYIWPRNDYDG	386.67	0.05	DRB1*13:01
436	DNYTAYCLGISHMEP	24,133.33	3.35	DRB1*09:01
61	ESSWVNRGESSRKAY	63,200.00	8.76	DQA1*05:01/DQB1*03:01
341	FHMWNYHSHVFSVGD	966.67	0.13	DRB4*01:01, DPA1*01:03/DPB1*02:01, DQA1*03:01/DQB1*03:02
91	FLENAHEHHGVYNQG	4,226.67	0.59	DQA1*04:01/DQB1*04:02, DQA1*01:01/DQB1*05:01
351	FSVGDTFSLAMHLQY	293.33	0.04	DRB1*09:01
56	HSDHAESSWVNRGES	16,520.00	2.29	DQA1*05:01/DQB1*03:01, DRB1*13:01
281	IEPGVLKVLRTEKQY	11,980.00	1.66	DRB1*04:05, DRB1*14:01
216	IQHICLKHTTCFQDV	1,393.33	0.19	DPA1*03:01/DPB1*04:02, DQA1*01:02/DQB1*06:02
146	IVNVDQRQYGDVFKG	46,026.67	6.38	DQA1*01:01/DQB1*05:01, DQA1*05:01/DQB1*02:01, DRB1*03:01
191	IYGVRYTETWSFLPS	33,373.33	4.63	DRB1*07:01, DPA1*02:01/DPB1*05:01
376	LEWLYVPIDPTCQPM	3,360.00	0.47	DQA1*01:01/DQB1*02:01
296	LGVYIWNMRGSDGTS	33,386.67	4.63	DQA1*01:01/DQB1*05:01, DQA1*04:01/DQB1*04:02
221	LKHTTCFQDVVVDVD	35,686.67	4.95	DQA1*01:01/DQB1*02:01, DQA1*01:01/DQB1*05:01
286	LKVLRTEKQYLGVYI	3,833.33	0.53	DRB1*14:01
316	LVTWKGDEKTRNPTP	6,893.33	0.96	DPA1*02:01/DPB1*05:01, DPA1*01:03/DPB1*04:01, DRB1*13:03
86	NDYDGFLENAHEHHG	6,446.67	0.89	DQA1*04:01/DQB1*04:02, DQA1*03:01/DQB1*03:02
336	PRGAEFHMWNYHSHV	16,120.00	2.23	DRB1*15:01, DPA1*01:03/DPB1*02:01
136	PTLNGDDRHKIVNVD	33,353.33	4.62	DQA1*01:01/DQB1*02:01, DRB1*03:02, DRB1*03:01
391	RLYSTCLYHPNAPQC	4,400.00	0.61	DPA1*01:03/DPB1*04:01
446	SHMEPSFGLILHDGG	50,273.33	6.97	DRB1*07:01, DRB1*09:01, DQA1*05:01/DQB1*03:01
31	SVLRYDDFHIDEDKL	37,026.67	5.13	DRB1*15:01, DQA1*01:01/DQB1*05:01
386	TCQPMRLYSTCLYHP	9,213.33	1.28	DPA1*01:03/DPB1*04:01, DRB4*01:01
291	TEKQYLGVYIWNMRG	566.67	0.08	DRB1*03:02, DQA1*04:01/DQB1*04:02, DRB1*13:01
496	TVDHFVNAIEERGFP	7,053.33	0.98	DQA1*05:01/DQB1*02:01
311	TYATFLVTWKGDEKT	44,200.00	6.13	DPA1*02:01/DPB1*05:01
466	VDTPESLSGLYVFVV	986.67	0.14	DRB1*07:01
486	VEAVAYTVVSTVDHF	39,460.00	5.47	DRB1*07:01, DQA1*05:01/DQB1*03:01, DRB3*02:02, DRB4*01:01
176	VEENHPFTLRAPIQR	29,973.33	4.16	DQA1*01:02/DQB1*06:02
441	YCLGISHMEPSFGLI	24,680.00	3.42	DQA1*01:02/DQB1*06:02
51	YEPYYHSDHAESSWV	11,153.33	1.55	DRB1*04:02, DQA1*03:01/DQB1*03:01, DPA1*01:03/DPB1*04:01, DQA1*05:01/DQB1*02:01, DRB4*01:01
481	YFNGHVEAVAYTVVS	5,700.00	0.79	DQA1*05:01/DQB1*03:01
196	YTETWSFLPSLTCTG	5,573.33	0.77	DQA1*01:01/DQB1*05:01, DQA1*03:01/DQB1*03:02, DRB1*07:01
476	YVFVVYFNGHVEAVA	62,426.67	8.65	DQA1*05:01/DQB1*03:01
Total		721,373.333	100

^aSFC, spot-forming cells.

Predictive performance of different methodologies.

The data set described above provided an opportunity to explore potential strategies to predict restrictions, without performing HLA restriction assays or performing in vitro HLA binding measurements. In a first determination, we explored the relative merit of four different methodologies to discern positive and negative restriction: the actual measured binding affinity, the predicted binding affinity, the percentile rank of predicted affinity, and the response frequency (RF) determination.

In terms of epitope prediction, different bioinformatics-based algorithms utilizing a broad array of computational approaches and platforms have been developed in the past 30 years. Several of the best-performing algorithms, as determined by community benchmarking studies (38), are available in the IEDB Analysis Resource (IEDB-AR). Two different prediction metrics have been considered in this study: (i) predicted major histocompatibility complex (MHC) binding affinity (or the 50% inhibitory concentration [IC₅₀], in nanomolar) and (ii) relative percentile rank score, which is also based on the IC₅₀ prediction but which ranks peptides as a function of their predicted affinity against a large random library of ligands, thereby allowing the identification of the top percentage of epitopes, as required. While several studies have been performed to identify predicted affinity thresholds for individual MHC class I alleles, in general, a prediction threshold of an IC₅₀ of <500 nM or a percentile rank score of 2% or better has been found to cover >95% of the known class I epitopes; a thorough evaluation of MHC class II alleles and predictions is ongoing (39).

To investigate this point further in the context of class II, we performed HLA predictions for the epitopes identified here and the HLA class II alleles expressed in each Shingrix vaccinee donor in our cohort, extracting both IC₅₀ values and percentile scores, and used the results of the HLA restriction experiments to assess the efficacy of various prediction thresholds. An alternative approach that was considered was to identify predicted restriction by calculating the relative frequency of specific HLA allele-epitope combinations identified across the Shingrix vaccinee donors (40, 41). The area under the curve (AUC) values observed in the receiver operating characteristic (ROC) curves were used to quantify the predictive efficacy across the four methodologies analyzed.

As shown in Fig. 8A and as expected, determination of the actual measured binding was the most effective (AUC = 0.7179), followed by determination of the prediction rank (AUC = 0.6600). Since experimental binding determination is laborious and time-consuming, for the subsequent analyses we focused on predicted rank as the most promising approach from an operational point of view. When the predictive efficacy was examined as a function of the different loci, we found that DRB1 (AUC = 0.7245; Fig. 8B) and DQ (AUC = 0.7542; Fig. 8C) restrictions were predicted with reasonable accuracy. Conversely, the performance was much lower in the case of the DP (AUC = 0.5425; Fig. 8D) and the DRB3/4/5 (AUC = 0.6377; Fig. 8E) loci, perhaps reflecting that these restrictions were relatively few and infrequent. Indeed, by considering only DRB1 and DQ loci as putative restriction elements and utilizing the 20th percentile rank to define binding, we would have correctly called HLA restrictions for 36 of 69 instances, with 26 incorrect restrictions (false positives). In other words, a cutoff of the 20th percentile is more specific and reduces the number of false positives, as, utilizing this criterion, we would be correct about 52% of the time and capture about 58% of the epitope restrictions.

FIG 8.

ROC curves of measured and predictive systems to infer HLA restriction. (A) Comparison of different methodologies. (B) HLA prediction based on rank percentile scores and as a function of the different HLA class II loci analyzed.

DISCUSSION

Here we report a detailed characterization of the epitope targets of the CD4 response of human subjects vaccinated with Shingrix. The results will enable mechanistic studies of vaccine efficacy, provide an example of a systematic analysis of epitope responses to a vaccine antigen, and have some methodological implications for future analyses and HLA restriction determinations.

Shingrix has shown increased efficacy in elderly and other groups compared to the Zostavax vaccine (10, 13, 42). The reasons for this remarkable difference may involve intrinsic differences in the quality and durability of the responses elicited and also may perhaps be because of inclusion of the AS01B adjuvant. We believe that our results will facilitate mechanistic studies to directly address this point, by enabling HLA class II tetramer studies and epitope pool-based isolation of reactive T cells.

In terms of enabling tetramer studies (43, 44), our results experimentally defined a total of 94 HLA/epitope restrictions, which allows an overwhelming majority of the human population, irrespective of ethnicity, to be covered. This set thus allows a limitation of studies that use tetrameric reagents to capture and study antigen-specific T cells to be overcome. Namely, this limitation is that the tetrameric reagents generally available cover only limited allelic diversity and are typically inclusive only of those alleles frequent in Caucasian populations. Our results also enable an alternative approach based on the use of pools of experimentally defined epitopes as bait to capture and characterize antigen-specific T cells (26). Tetrameric reagents incorporating these epitopes may provide information on whether and how specific epitopes affect polyfunctionality. Several studies have started to address these issues. Laing et al. (18) reported that CD4⁺ T cells respond to peptides derived from different VZV ORFs and that the response increases after Zostavax vaccination. Zostavax-specific CD4⁺ T cells have a mixed Th1/Th2/Th21 phenotype (20). Increases in T follicular helper cells (Tfh) and IL-21 in CD4s, in general, and a mixed Th1/Th2 response with both IFN-γ and IL-4 release were reported from PBMCs (20). Levin et al. (17) reported that the profiles and magnitude of Th1 T cell memory responses differentiate Shingrix and Zostavax.

Shingrix vaccination, unsurprisingly, induces a vigorous gE-specific response, in agreement with the findings of previous studies (34, 17, 45). Interestingly, we were also able to detect a significant response against nonstructural (NS) proteins, as well as other structural proteins, in Shingrix vaccinees. These responses were most likely derived by previous natural exposure or chicken pox vaccination. When we compared the responses across Zostavax vaccinees, individuals exposed in childhood, and HZ patients, the three cohorts were found to have a similar response profile, with a trend toward stronger responses against nonstructural (NS) proteins. Specifically, a significantly stronger response against NS proteins than structural proteins was observed for individuals exposed in childhood and HZ patients but not for Zostavax vaccinees, where only a trend was observed. A significantly stronger reactivity against the nonstructural proteins was observed for the childhood exposure and shingles cohorts (for S versus NS proteins, P = 0.0458 for the childhood exposure cohort and P = 0.0016 for the shingles cohort [P values were determined by the nonparametric paired Wilcoxon test]), while a nonsignificant positive trend was observed for the Zostavax and Shingrix vaccinee cohorts (for S versus NS proteins, P = 0.1454 for Shingrix vaccinees and P = 0.1175 for Zostavax vaccinees [P values were determined by the nonparametric paired Wilcoxon test]). This also suggests that the pattern of VZV-specific CD4⁺ T cell protein immunodominance after natural exposure is, in general, against NS proteins. This applied to the whole study cohort, and it is not surprising that the Shingrix recipients had responses against NS proteins since they very likely had childhood exposure. In this sense, Zostavax vaccination mirrors the natural infection profile, while Shingrix skews the natural protein immunodominance toward gE, the only component of the vaccine. Indeed, the T cell reactivity against gE induced by Shingrix is so strong that it is significantly higher than that against the nonstructural (NS) counterpart. This is also in contrast to the heterogenic pattern of CD4⁺ T cell reactivity observed in young adults after chicken pox vaccination, where no clear immunodominance against NS proteins is observed. Instead, a quite diverse pattern of reactivity is observed in different donors, with responses against both structural and NS proteins, suggesting that priming by natural infection, as opposed to chicken pox vaccination in the young, induces a skewed response against NS proteins only and may contribute to the reduced response against Zostavax observed in the elderly (14, 29).

We positively identified a total of 89 epitopes recognized across 18 Shingrix vaccinee donors. Of these, the 34 most responsive captured 80% of the total (SFC) response against all identified epitopes. Thus, while Shingrix vaccinees are associated with a broad and diverse repertoire, a limited number of epitopes are dominant and account for the majority of responses. These epitopes will help build an immunogenic profile of VZV, as currently only 25 VZV gE-derived T cell epitopes recognized in humans have been reported in the published literature, as captured by the IEDB. Of the top 34 epitopes identified in our study, 7 (20%) nest or overlap previously identified epitopes, while the remaining 27 (80%) are novel.

The results in terms of epitope breadth and coverage can be interpreted in the context of other studies from our laboratories that have used a similar methodology to define the epitope repertoire of different antigens by scanning with sets of overlapping peptides. These studies included mycobacterial antigens (46), pollen (47), cockroach (48) and dust mite (25) allergens, and pertussis (48) and tetanus (22) antigens. The pattern of 20 to 30 epitopes encompassing the vast majority of the responses is generally applicable, and in this respect, the gE antigen does not present unusual characteristics that may explain the remarkable vaccine efficacy. A lack of reactivity was generally observed in the terminal end of the gE protein in the Shingrix vaccinees analyzed in this study. This can be explained by the fact that those residues are not included in the recombinant protein used for the Shingrix vaccination to facilitate protein preparation.

Our results indicate that Shingrix vaccination may result in a repertoire marginally broader than provided by Zostavax vaccination and natural exposure and a somewhat higher magnitude of responses with differences in the epitope immunodominance hierarchy. Indeed, broader responses have been invoked as a correlate of protection in the case of hepatitis C virus and other viral infections (49–51). However, in the case of Shingrix versus Zostavax, differences in the breadth of epitopes recognized were not statistically significant, and any differences should also be interpreted with caution, since the samples from the Shingrix vaccinees were generally obtained closer to the time of vaccination than the samples from the herpes zoster cohort.

It is of interest to address whether the epitopes identified for gE are specifically targeted following Shingrix vaccination. The current data set is based on only limited donors, and further studies will have to investigate this issue further. Also, the cohorts studied are reflective of the general volunteer donor pool that answered our recruitment in the San Diego area. Despite every effort made to recruit women and minorities, there was still a bias toward males and Caucasians. As a further caveat, it should be considered that the cohort size utilized in the current study is somewhat limited. For all of the reasons presented above, the results should be interpreted with caution, and larger studies will be required to confirm our conclusions.

Finally, our data, in line with those from previous studies (26, 36), validate the use of single HLA-transfected target cells as a valid methodology to experimentally determine HLA restrictions. The methodology is powerful and accurate, but somewhat laborious. For this reason, we explored the possibility of applying publicly available prediction approaches to infer the HLA restriction of experimentally defined epitopes. In this context, our data set also allowed evaluation of the efficacy of three different methodologies to predict HLA restrictions in the absence of experimental data. We found that the percentile rank metric is the best approach to predict restrictions for the DRB1 and DQ loci. Using a 20th percentile cutoff, we were able to accurately predict 52% of epitope/HLA combinations for DRB1 and DQ alleles identified in our data set. This conclusion has two potential applications: the first is to assign, with a high likelihood, the HLA restriction of experimentally defined epitopes, and the second is to apply prediction algorithms before running experiments to selectively test predicted candidates only in the specific donor(s) with the corresponding HLA type. The second application can help narrow down targets of investigation prior to performing assays and increase the speed and coverage of identifying HLA/epitope combinations, while also conserving precious biological samples.

MATERIALS AND METHODS

Study subjects, PBMC isolation, and HLA typing.

The Shingrix vaccinee cohort consisted of healthy donors: adult male and nonpregnant female volunteers, age 50 years and over, that were enrolled after receiving Shingrix vaccination (n = 18) with the approval of the La Jolla Institute for Immunology (LJI) IRB committee (program VD-184). PBMCs derived from the Shingrix vaccinee cohorts were processed at LJI by density gradient sedimentation using Ficoll-Paque (Lymphoprep; Nycomed Pharma, Oslo, Norway). Isolated PBMCs were cryopreserved in heat-inactivated fetal bovine serum (FBS; HyClone Laboratories, Logan UT) containing 10% dimethyl sulfoxide (DMSO; Gibco) and stored in liquid nitrogen until use in the assays.

HLA typing was performed by an ASHI-accredited laboratory at Murdoch University (Perth, Western Australia, Australia) for class II (DRB1, DRB3/4/5, DQA1/DQB1, DPB1), as previously described (26, 52, 53). Table 1 describes the general cohort characteristics, including ethnicity, race, gender, and age. Table S2 in the supplemental material lists the HLA types of the Shingrix vaccinees. Based on the age of collection and geographical location, there is a high likelihood that all the donors in this study had a childhood exposure to VZV.

Peptides, mesopools, and megapool.

Fifteen-mers overlapping by 10 residues were derived from gE protein of the strain Dumas reference sequence (GenBank accession number NC_001348.1), available in the Viral Expasy Database (https://viralzone.expasy.org/), for a total of 123 peptides. Peptides were synthesized as crude material (A&A, San Diego, CA) and resuspended in DMSO at a concentration of 20 mg/ml.

Part of the peptides synthesized were either pooled in small mesopools, containing 10 peptides each, or pooled with all the gE peptides and sequentially lyophilized as previously described to generate a gE megapool (54). The gE epitopes identified in this study were mapped to the gE Dumas reference sequence (GenBank accession number NC_001348.1) to identify immunodominant regions, using the Immunobrowser tool (55).

Activation-induced marker (AIM) assay.

Cells were cultured in the presence of either gE, structural (S), or nonstructural (NS) protein MPs (2 μg/ml), DMSO (2%), or phytohemagglutinin (PHA; 10 μg/ml) for 22 to 24 h. After stimulation, cells were stained with surface markers for 30 min at 4°C. Detailed information for all the antibodies used for the flow cytometry experiments in this study can be found in Table 6. Surface marker proteins were quantified via flow cytometry (LSRII flow cytometer; BD Biosciences) and analyzed using FlowJo software (version 10.5.3; TreeStar Inc., Ashland, OR). The gating strategy is schematically represented in Fig. 1. Statistical analyses were performed using GraphPad Prism software (version 8; GraphPad Software, San Diego, CA). Responses of paired samples across different stimuli were tested using a Wilcoxon matched pair test.

TABLE 6.

AIM flow cytometry panel^a

Antibody	Fluorochrome	Vol (μl/well)	Vendor	Catalog no.	Clone
CD3	AF700	2	BioLegend	317340	OKT3
CD14	V500	1	BD	561391	M5E2
CD19	V500	1	BD	561121	HIB19
Live/Dead	ef506/Aqua	0.5	eBioscience	65-0866-18
CD8	Cy5.5	2	eBioscience	45-0088-42	RPA-T8
CD4	Pacific Blue	2	BD	558116	RPA-T4
OX40	PE-Cy7	2	BioLegend	350012	Ber-ACT35
CD137	APC	2	BioLegend	309810	4B4-1

^aThe antibody panel used in flow cytometry experiments to identify both the magnitude and the quality of the CD4⁺ T cell response. PE, phycoerythrin; APC, allophycocyanin.

Epitope identification by FluoroSpot assay.

PBMCs were expanded in the presence of the gE megapool (MP) at a concentration of 1 μg/ml for 14 days. After culturing, the cells were collected and restimulated in triplicate at a density of 5 × 10⁴ cells/well with gE MP (1 μg/ml), smaller mesopools containing 10 peptides each (1 μg/ml), PHA (10 μg/ml), and DMSO (0.1%) in 96-well plates that had previously been coated with anticytokine antibodies for IFN-γ, TNF-α, and IL-5 (monoclonal antibodies [MAbs] 1-D1K, MT25C5, and TRFK5, respectively; Mabtech, Stockholm, Sweden) at a concentration of 10 μg/ml.

After 20 h of incubation at 37ºC in 5% CO₂, the cells were discarded and FluoroSpot assay plates were washed and further incubated for 2 h with cytokine antibodies (MAbs 7-B6-1-BAM, 5A10 biotinylated, and MT20D9-WASP, respectively; Mabtech, Stockholm, Sweden). Subsequently, the plates were washed again with phosphate-buffered saline/0.05% Tween 20 and incubated for 1 h with fluorophore-conjugated antibodies (anti-BAM-490, SA-550, and anti-WASP-640). Computer-assisted image analysis was performed by counting the fluorescent spots using an AID iSPOT enzyme-linked immunosorbent spot (ELISpot) assay reader (AIS-Diagnostika, Germany).

Each mesopool was considered positive compared to the background reading based on the following criteria: 100 or more spot-forming cells (SFC) per 10⁶ PBMCs after subtraction of the background reading for each cytokine analyzed, a stimulation index (SI) greater than 2, and a result statistically significantly different from that for the background (P < 0.05) in either a Poisson test or a t test. Mesopools eliciting the strongest cytokine response were deconvoluted at day 17 into the corresponding peptides at a concentration of 10 μg/ml and analyzed using the same criteria applied for the mesopools.

Single HLA-transfected cell lines.

Single HLA-transfected fibroblast (DAP.3) or B lymphocyte (RM3) cell lines were generated and maintained as previously described (26, 36). Based on the HLA typing of the donor, we assigned the HLA-transfected cells considering only matches for the HLA class II DR, DP, and DQ beta chains. When multiple cell lines corresponded to the same beta chain, the alpha chain was matched based on the typing as well.

Short-term T cell line (TCL) cultures were set up using donor PBMCs cultured individually in the presence of single epitopes previously identified in the epitope identification experiments at a 1-μg/ml concentration for 14 days. At day 13, fibroblast-derived APC lines were induced with butyric acid (100 μg/ml) to favor HLA expression on the cell membrane.

On the following day, all APC lines were harvested, and viability (>75% for all APC lines) was determined using trypan blue. Each APC line was pulsed individually with peptide at a concentration of 10 μg/ml for 1 h at 37ºC in 5% CO₂. TCLs were stimulated in triplicate with either a peptide-pulsed APC line, the APC line alone (as a control), peptides (1 μg/ml), PHA (10 μg/ml), or DMSO (0.1%) in 96-well plates that had previously been coated with anticytokine antibodies for IFN-γ, TNF-α, and IL-5 at a concentration of 10 μg/ml. The FluoroSpot assays carried out and the analysis criteria used were as described above for the epitope identification experiments. Computer-assisted image analysis was performed by counting the fluorescent spots using Mabtech’s IRIS FluoroSpot and ELISpot assay reader (Mabtech, Sweden).

HLA binding assay.

Classical competition assays to quantitatively measure peptide binding to purified MHC class II molecules were performed as previously described (56). Briefly, a high-affinity radiolabeled peptide was coincubated at room temperature or 37°C with purified MHC, a cocktail of protease inhibitors, and various concentrations of unlabeled inhibitor peptide. Following a 2-day incubation, the MHC-bound radioactivity was determined by capturing MHC/peptide complexes on antibody-coated Lumitrac 600 plates (Greiner Bio-One, Frickenhausen, Germany) and measuring the number of bound counts per minute using a TopCount (Packard Instrument Co., Meriden, CT) microscintillation counter. The concentration of peptide yielding 50% inhibition of the binding of the radiolabeled peptide was calculated. Under the conditions utilized, where the label concentration is less than the MHC concentration and the IC₅₀ is greater than or equal to the MHC concentration, the measured IC₅₀ values are reasonable approximations of the true (K_d [dissociation constant]) values. Each competitor peptide was tested at six different concentrations covering a 100,000-fold range and in three or more independent experiments. As a positive control, the unlabeled version of the radiolabeled probe was also tested in each experiment.

RF score calculation and HLA epitope prediction.

T cell epitope prediction analyses, as well as relative frequency (RF) score calculations, were performed using IEDB analysis tools (57). The RF score is a combined metric that accounts for the positivity rate (how frequently an epitope elicits a positive response) and the number of independent assays. Specifically, RF = [(r – sqrt(r)]/t, where r is the total number of responding donors, sqrt is square root, and t is the total number of donors tested (54). The RF score was calculated using the RATE tool (40, 41). HLA class II binding predictions were performed following two different strategies: one using the NetMHCIIpan (version 3.2) algorithm (58) only and extracting predicted IC₅₀ values and the other one using the IEDB-recommended 2.22 methodology and extracting rank percentile values. The IEDB-recommended 2.22 methodology is based on a combination of seven prediction algorithms with priority on the consensus method (59), which includes the comblib (60), SMM (61), and NN (62) algorithms, followed by the NetMHCIIpan (version 3.2) algorithm (58).

Data availability.

A complete list of the epitopes identified in this study is available in Table S3.