ABSTRACT
A detailed characterization of the dynamics and breadth of the immune response to an acute viral infection, as well as the determinants of recruitment to immunological memory, can greatly contribute to our basic understanding of the mechanics of the human immune system and can ultimately guide the design of effective vaccines. In addition to neutralizing antibodies, T cells have been shown to be critical for the effective resolution of acute viral infections. We report the first in-depth analysis of the dynamics of the CD8+ T cell repertoire at the level of individual T cell clonal lineages upon vaccination of human volunteers with a single dose of YF-17D. This live attenuated yellow fever virus vaccine yields sterile, long-term immunity and has been previously used as a model to understand the immune response to a controlled acute viral infection. We identified and enumerated unique CD8+ T cell clones specifically induced by this vaccine through a combined experimental and statistical approach that included high-throughput sequencing of the CDR3 variable region of the T cell receptor β-chain and an algorithm that detected significantly expanded T cell clones. This allowed us to establish that (i) on average, ∼2,000 CD8+ T cell clones were induced by YF-17D, (ii) 5 to 6% of the responding clones were recruited to long-term memory 3 months postvaccination, (iii) the most highly expanded effector clones were preferentially recruited to the memory compartment, and (iv) a fraction of the YF-17D-induced clones could be identified from peripheral blood lymphocytes solely by measuring clonal expansion.
IMPORTANCE The exhaustive investigation of pathogen-induced effector T cells is essential to accurately quantify the dynamics of the human immune response. The yellow fever vaccine (YFV) has been broadly used as a model to understand how a controlled, self-resolving acute viral infection induces an effective and long-term protective immune response. Here, we extend this previous work by reporting the identity of activated effector T cell clones that expand in response to the YFV 2 weeks postvaccination (as defined by their unique T cell receptor gene sequence) and by tracking clones that enter the memory compartment 3 months postvaccination. This is the first study to use high-throughput sequencing of immune cells to characterize the breadth of the antiviral effector cell response and to determine the contribution of unique virus-induced clones to the long-lived memory T cell repertoire. Thus, this study establishes a benchmark against which future vaccines can be compared to predict their efficacy.
INTRODUCTION
During the acute response to a viral infection, viral antigen (Ag)-specific effector CD8+ T cell clones (also known as cytotoxic T lymphocytes, or CTLs) become activated and expand as they recognize and eliminate infected host cells (1, 2). The Ag specificity of a T cell clone is determined by the T cell receptor (TCR), which is encoded by random, RAG-mediated V(D)J recombination. Thus, each T cell clone may be identified by its unique TCRβ CDR3 region, formed from the joining of the V, D, and J gene segments along with deletions and nontemplated insertions at the junctions, with CDR3 being the primary determinant of Ag specificity (3, 4). The identification and tracking of virus-specific CTLs has resulted in the extensive characterization of their phenotype and function (5–8). The identification of virus-specific T cells during the course of an infection has allowed the measurement of the number of unique clones responding to a particular viral epitope (9–11). These studies suggested that the magnitude of the T cell clonal response to different viral Ags is not uniform; for example, in the case of the yellow fever vaccine (YFV), peptide NS4b induces a more robust T cell response than peptide NS5 (9, 12). Moreover, there is extensive variability in the number of unique clones activated by a particular viral epitope (13, 14), which depends both on the quantity of peptide presented (15) and on the microenvironment of the lymph node where the T cell encounters the Ag (7). In addition, responses to chronic and acute viruses seem to be characterized by different patterns of activation and waning of effector cells, as well as different memory cell phenotypes, which might be related to the different patterns of exposures to viral Ags in these two different types of infection (reviewed in reference 16). Finally, major hisotcompatibility complex polymorphisms lead to variable epitope presentation in different individuals (17, 18), complicating the characterization of dominant and nondominant clonal CTL responses.
The formation of virus-specific CD8+ memory T cells is also believed to be dependent on the magnitude of the clonal response to Ag (19, 20). After an acute infection is resolved, the virus-specific effector CD8+ T cell pool contracts (21), and a much smaller number of long-lived memory T cells that are capable of responding to subsequent infections is maintained (22). It is thought that effector T cell clones present in high abundance are recruited to the memory repertoire with higher frequency than less abundant clones (11, 23, 24), but it is not clear whether this simply reflects the limitations of currently available techniques. Therefore, highly sensitive techniques are necessary to establish the contribution of less abundant clones to the memory pool (12). Furthermore, to date it has not been possible to relate the magnitude and diversity of the effector T cell response to the subsequent abundance of individual clones in the memory T cell repertoire. Thus, the detailed characterization of the dynamics of the T cell repertoire in response to an acute viral infection can increase our understanding of the breadth of the immune response, the formation of immunological memory, and how the human immune system responds to acute viral infections and immunization with viral vaccines.
We used vaccination with the yellow fever (YF) virus vaccine YF-Vax, which is based on the YF-17D204 attenuated strain, as a model of acute viral infection. YF-17D harbors only 20 amino acid changes compared to the wild-type strain, most of which are found in the E protein and are thought to result in changes in viral tissue tropism (25). In addition, this attenuated virus is replication competent, so that administration of the YFV results in a mild viral infection that is predicted to elicit an immune response that is almost identical in quality to that induced by wild-type infection (26). Since exposure to YF virus is geographically limited, and YFV is a very effective vaccine that elicits an optimal, long-term protective immune response upon administration of a single dose, this model has been used extensively to explore the human immune response to a controlled, self-resolving acute viral infection (reviewed in references 16 and 27). These seminal studies have shown that (i) the ability of YF-17D to infect dendritic cells and signal through multiple Toll-like receptors may be related to the effectiveness of this vaccine (28); (ii) neutralizing antibodies (nAbs) are the best surrogate marker for protection against YF virus and remain detectable for many years (29, 30); and (iii) CD8+ T cells expand massively before nAbs can be detected (and are thus likely involved in the control of viremia) and persist in the memory compartment for decades (6, 30).
Our understanding of the CD4+ response to YFV is limited. Although helper T cells are clearly required for the production of YFV-specific Abs (including nAbs), different studies have reported variable levels of induction of CD4+ T cells upon vaccination with YFV (30, 31). Some analyses have revealed that cytokine-producing YFV-specific CD4+ T cells can be detected as early as day 2 postvaccination and that they return to baseline by day 28, suggesting that the kinetics of CD4+ T cells precede those of CD8+ T cells (12, 32). Recently, James et al. used class II HLA-DR restricted, YFV-specific tetramers to characterize the CD4+ response to YFV in more depth, showing that all 10 proteins in the YF virus genome contain antigenic epitopes recognized by CD4+ T cells (33). This study also revealed a wide range of frequencies of CD4+ T cells specific for a limited number of YFV epitopes in peripheral blood (from 0 to 100 cells per million CD4+ T cells) and established that YFV-specific T cells, which display a predominant Th1-like memory phenotype, occur at ∼10- to 100-fold-higher frequencies in vaccinated versus unvaccinated individuals, depending on the time point considered (33).
In contrast, there have been several detailed analyses of the kinetics and phenotype of CD8+ T cells induced by vaccination with YFV. For example, Miller et al. (6) showed that activated effector CD8+ T cells (TAE) peak 2 weeks after administration of the YFV and defined the YFV-specific subpopulation of CD8+ CTL cells as CD38+ HLA-DR+ Ki-67+ Bcl-2lo. In addition, this study established a strong correlation between the levels of CD38+ HLA-DR+ CD8+ T cells and the expression of gamma interferon (IFN-γ) by total CD8+ T cells in response to YF virus-infected cells, and it demonstrated that stimulation of CD8+ T cells from YFV-vaccinated volunteers with a comprehensive pool of peptides that span the YF virus polyprotein also induced IFN-γ. Since unrelated memory CD8+ T cells (such as those specific for chronic viruses like Epstein-Barr virus [EBV] and cytomegalovirus [CMV] and therefore presumed to preexist at the time of vaccination with YFV) were not found among the expanded CD8+ T cell population, these observations suggest that, at least in the case of YFV, the bystander effect is minimal, and they also imply that the vast majority of TAE clones observed after administration of YF-17D are YF virus specific. Finally, those authors showed that Ag-specific cells could be identified more than 30 days postvaccination, indicating that the YFV-specific effector CD8+ T cells had waned and also that a certain proportion of them had entered the memory compartment (6). Subsequent work from the same group employed an array of overlapping peptides that spanned the entire YF virus polyprotein to demonstrate that vaccination with YFV induces a broad CD8+ T response that targets several epitopes in each of the 10 viral proteins (9). The use of tetramers carrying an immunodominant epitope from the nonstructural NS4b protein helped define the phenotypes of YFV-specific CD8+ T cells through the expansion, contraction, and memory phases of the immune response, further confirming that CD38+ HLA-DR+ CD8+ T cells dramatically expand after YFV-17D administration and produce cytotoxic effector molecules (9). Similar results were observed by Co et al., who identified YFV-specific proliferation and cytolytic responses on day 14 postvaccination and isolated CD8+ T cell lines that were specific for epitopes from structural and nonstructural YF virus proteins, some of which persisted for up to 19 months postvaccination (10). Again, follow-up data from a tetramer-based approach showed that YFV-specific CD8+ T cells could be identified as early as 7 to 9 days postvaccination, before IFN-γ production was detectable, that memory cells corresponded mostly to a differentiated effector phenotype (CD45RA+ CCR7− CD62L−), and that these peptide-specific responses lasted for at least 54 months (34). A more recent study using a limited set of YF virus HLA-tetramer epitopes suggested that the CD8+ response to YFV is broad and complex and that responses to different epitopes vary in magnitude and duration (12). Those authors also found that YFV-specific effector CD8+ T cells were CD45RAhi CCR7− PD1+ CD27hi and that only some of these cells transition to the T cell memory compartment, at which point they became CD45RA+ CCR7− PD1− CD27lo (12).
In this study, we developed a complementary approach to study the dynamics of the human effector and memory CD8+ T cell repertoire upon acute viral infection. First, we isolated total peripheral blood mononuclear cells (PBMCs) from volunteers who received the YF-17D vaccine prevaccination (on day 0) and on days 14 and 90 postvaccination, and we used flow cytometry to sort a fraction of these samples into CD8+ CD38+ HLA-DR+ activated, effector T cells on day 14 (i.e., at the peak of their abundance) and into memory CD8+ T cells on days 0 and 90. High-throughput sequencing of the rearranged TCRβ locus in each sample, combined with a computational method to identify expanded T cell clones, allowed the characterization of individual, YFV-induced CD8+ T cell clones during the acute phase, to estimate the abundance of each of these clones and to track them into the memory phase of the antiviral response. This synthesis of flow cytometry sorting protocols and high-throughput sequencing enabled the measurement of the T cell response to viral infection at an unprecedented resolution. Finally, we show that our approach allows the identification of many YFV-induced CD8+ T cell clones by assessing clonal expansion directly from peripheral blood samples (i.e., without previous sorting of activated, effector, or memory CD8+ T cells) and that a large proportion of these clones overlap those identified through immunosequencing of the flow cytometry-sorted activated, effector CD8+ T cell population.
MATERIALS AND METHODS
Vaccination and sample collection.Nine volunteers between the ages of 18 and 45 years consented under Fred Hutchinson Cancer Research Center (FHCRC) and the University of Washington Vaccine Research Clinic (UWVRC) IRB protocols to receive the yellow fever single-dose vaccine YF-VAX (based on the YF-17D204 strain of the yellow fever virus [26]) and to have 200 ml of blood drawn at three different time points: immediately before vaccination (day 0), 2 weeks postvaccination (day 14), and 3 months postvaccination (day 90) (Table 1). Written informed consent to use the blood samples in this study was obtained from each subject. The administration of the YF vaccine and all blood draws and were performed at the UWVRC.
Experimental designa
Cell sorting.All cell sorting was performed at FHCRC. Whole-blood samples (200 ml) were collected and PBMCs were isolated by using Histopaque (Sigma-Aldrich, St. Louis, MO) density gradient centrifugation. CD8+ T cells were isolated from total PBMCs by magnetic separation using CD8 MicroBeads and the autoMACs Pro separator (both from Miltenyl Biotec, Auburn, CA), followed by staining with anti-CD3–Alexa Fluor 700, anti-CD8–allophycocyanin (APC)-H7, anti-CD38–phycoerythrin (PE), HLA-DR–fluorescein isothiocyanate, anti-CD14–Pacific Blue, anti-CD19–V450, anti-CD45RO–PE Cy7, anti-CD45RA–APC, anti-CD62L–peridinin chlorophyll protein-Cy5.5, and 4′,6-diamidino-2-phenylindole (DAPI) (all obtained from BD BioSciences, San Jose, CA). T cell subpopulations were sorted using the BD FACSAria II system and FACSDiva v6.1.3 software (BD Biosciences). First, we gated on propidium iodide-negative (PI−) CD14− CD19− to remove dead cells, monocytes, and B cells and then on CD3+ CD8+ to exclude non-T cell lymphocytes and CD4+ T cells. Finally, we isolated four different CD8+ T cell subsets: from the day 0 prevaccination samples we isolated CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ memory T cells (TM-0); from the day 14 postvaccination samples we isolated CD3+ CD8+ CD14− CD19− CD38+ HLA-DR+ Ag-experienced, activated effector T cells (TAE-14); and from the day 90 postvaccination samples we isolated CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ CD62Llo effector memory T cells (TEM-90) and CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ CD62Lhi central memory T cells (TCM-90). To avoid contamination, CD38+ HLA-DR+ cells were excluded from the effector memory and central memory T cell populations. Day 90 samples from three of the volunteers had to be discarded due to contamination.
DNA extraction and immunosequencing.Genomic DNA was purified from total PBMCs and each sorted T cell population sample by using the QIAmp DNA blood minikit (Qiagen). For each sample, DNA was extracted from ∼1 million T cells, and the TCRβ CDR3 regions were amplified and sequenced using ImmunoSEQ (Adaptive Biotechnologies, Seattle, WA) as previously described (35). In brief, bias-controlled V and J gene primers were used to amplify rearranged V(D)J segments for high-throughput sequencing at ∼20× coverage. After correcting sequencing errors via a clustering algorithm, CDR3 segments were annotated according to the International ImMunoGeneTics Collaboration (36, 37) to identify the V, D, and J genes that contributed to each rearrangement. Sequences were classified as nonproductive if it was determined that nontemplated insertions or deletions produced frameshifts or premature stop codons. We used a mixture of synthetic TCR analogs in each PCR to estimate the absolute template abundance (i.e., the number of cells bearing each unique TCR sequence) from sequencing data, as previously described (38).
Identification of expanded and enriched effector T cell clones.We defined a T cell clone as the population of T cells bearing a unique TCRβ rearrangement. To computationally identify those T cell clones whose frequencies differed between samples from a given subject taken at different time points, or between cell populations (e.g., between total PBMCs and a specific sorted T cell population for the same time point), we used the algorithm described below. The input data consisted of the absolute abundance for each TCRβ clone in each sample.
We assumed that the repertoire contains S distinct clones, and their proportional abundances in samples 1 and 2 are given by the multinomial vectors p(1) = {p1(1), p2(1), …, pS(1)} and p(2) = {p1(2), p2(2), … , pS(2)}, with
Next, we assume that the aggregated proportional change of all truly changed clone abundances is small [i.e.,
We then defined s as the number of distinct clones actually observed across the two samples, where in general s ≤ S. Without loss of generality, indices 1 through s of the repertoire clones correspond to the observed clones. After performing the above analysis on each of the s observed clones, we generated a vector of P values: P = {P1, P2, …, Ps}.
To choose a rejection region (thereby identifying a set of significantly changed clones between the two samples under consideration), we used the positive false discovery rate (pFDR) method of Storey (39), which defines the pFDR as the expected proportion of true null hypotheses among all rejected hypotheses:
The second equality follows from Bayes' theorem, with π0 being the prior probability that a hypothesis is null. The last equality follows from the definition of a P value, if the P values themselves are regarded as independently and identically distributed random variables.
For each P value, Pi, the associated Q value, Qi, may be estimated; this is the minimum pFDR that can occur when rejecting P values less than or equal to Pi. By examining the number of significant tests at various Q value thresholds, an appropriate threshold can be selected (see Fig. 1, below). Control of pFDR is preferred for control of the famiwise error rate (FWER), i.e., the probability of one or more false alternative hypotheses. The latter (typically controlled by the Bonferroni method) is overly conservative, as it fails to reject many false null hypotheses in order to attain any nontrivial FWER. The pFDR, on the other hand, rejects these hypotheses at the cost of a specifiably small proportion of rejected true null hypotheses.
The resulting set of significance tests allows the identification of T cell clones whose frequencies are different in the two samples (i.e., dynamic T cell clones). For example, applying this algorithm to the comparison of total PBMCs isolated on day 14 postvaccination to activated CD8+ T cells purified from the same sample identifies a set of enriched, activated CD8+ T cells that are expected to be YFV specific. In contrast, the comparison of total PBMCs obtained from the same volunteer on day 0 (prevaccination) and on day 14 postvaccination identifies a set of putative YFV-reactive clones based on clonal expansion.
RESULTS
It is well established that effector CD8+ T cells expand in response to an acute viral infection (39). Expanded clones can either bind specifically to a pathogen-derived epitope presented by a type I HLA molecule, or they can be induced to expand nonspecifically by cytokines released by other cells, in a process known as the bystander effect (40). In the case of the YFV model, which results in a self-limited, acute viral infection (16, 27) and has thus been extensively used to characterize the human antiviral immune response, activated effector CD8+ T cells peak 2 weeks postvaccination (6, 10) and express a particular set of phenotypic markers, including CD38, HLA-DR, Ki-67, and Bcl-2 (6). The massive expansion of these activated, effector CD8+ cells in response to vaccination with YFV is specific, since these cells have been shown to produce cytokines in response to stimulation with peptides from YF-17D proteins (9, 34), and existing memory CD8+ T cells specific for other viruses, such as CMV or EBV, do not contribute to the activated, proliferating pool of CD8+ T cells (6).
To further explore the dynamics of the T cell repertoire in response to an acute viral infection, we administered a single dose of the live attenuated YFV YF-VAX, based on the YF-17D204 strain of the YF virus (26), to nine healthy volunteers, none of whom reported being previously exposed to the YF virus or having received a YFV. We drew 200 ml of peripheral blood from each subject on day 0 (immediately prior to vaccination) and on days 14 and 90 postvaccination (Table 1). To identify CD8+ T cells present in the memory compartment prior to vaccination, we sorted a fraction of the total PBMCs obtained from all 9 subjects on day 0 into CD8+ memory T cells (TM-0, defined as CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ cells [41]). Similarly, to characterize the activated effector CD8+ T cells induced by vaccination with YFV, we also sorted a fraction of the total PBMCs obtained from all 9 subjects on day 14 postvaccination by selecting CD3+ CD8+ CD14− CD19− CD38+ HLA-DR+ activated effector CD8+ T cells (TAE-14) (6). Finally, to determine which of these clones enter the memory compartment, we sorted PBMCs obtained on day 90 from 6 of the subjects into effector memory (TEM-90) and central memory (TCM-90) CD8+ T cells (respectively, CD8+ CD45RO+ CD62Llo and CD8+ CD45RO+ CD62Lhi) (42). We were unable to characterize the TEM-90 and TCM-90 cell populations from the 3 other subjects because these samples had to be discarded due to contamination.
To identify and quantify YFV-induced T cell clones, we extracted genomic DNA from ∼1 million T cells for either total PBMCs or sorted T cell populations (Table 1). Next, we used PCR amplification and high-throughput sequencing to characterize the CDR3 regions of rearranged TCRβ loci as previously described (35). TCRβ sequences are nearly unique for each clone, so that the data can be used to assess the dynamics of the cellular adaptive immune response both over time and between T cell subpopulations. Additionally, we determined the number of original templates corresponding to each PCR-amplified clonal sequence by assessing the amplification of a set of synthetic templates, thus providing an estimate of the cellular abundance for each clone in each sample (38).
Identification of vaccine-induced clones.To assess the dynamics of the YFV-induced CD8+ T cell repertoire, we determined whether each unique clone (defined by sequencing the CDR3 region of the TCRβ chain) was enriched in the day 14 postvaccination, YFV-induced effector CD8+ T cell compartment (TAE-14), defined by the expression of CD38 and HLA-DR (6), in comparison to the corresponding total PBMC sample from that time point from the same subject. To do this, we developed a statistical method to identify clones that had significant proportional abundance differences between two samples (see Materials and Methods) (Fig. 1A). Our approach controls for the false-positive rate and takes into account experimental errors that result in the presence of false positives in the YFV-induced TAE-14 compartment (i.e., cells that do not have the indicated surface markers). This avoids overstating the number of YFV-induced clones, which would result from a simple enumeration of clones present in the TAE-14 compartment, since it includes low levels of many clones that are no more frequent than in the corresponding total PBMC sample. Instead, we considered a clone to be YFV induced if (i) it was significantly enriched in the TAE-14 compartment with respect to the corresponding total PBMC sample, and (ii) it carried a productive TCRβ rearrangement. Since the subjects who participated in this study had not been previously exposed to either the YF virus or a YFV, we also took into consideration whether each unique CD8+ T cell clone identified was present in the day 0 prevaccination memory cell compartment (TM-0). Based on these criteria, we classified T cell clones into four categories, as follows: YFV-induced clones (i.e., enriched in the TAE-14 compartment versus the day 14 postvaccination total PBMC sample from that individual but absent in the corresponding TM-0 compartment); cross-reacting or bystander clones (i.e., enriched in the TAE-14 compartment versus the corresponding total PBMC sample but present in TM-0), and those not enriched in the TAE-14 compartment that were either present or absent in TM-0 (Fig. 2).
Selection of FDR thresholds. (A) Number of clones classified as YFV induced for various FDR significance thresholds for all subjects. A threshold of 0.01 was selected. (B) Number of clones classified as putatively reactive clones for various FDR significance thresholds for all subjects. A threshold of 0.05 was selected. Each subject is represented by a different tone of gray, as indicated in the legend.
Identification of YFV-induced clones. The graphs show the abundance of unique clones identified by statistical enrichment on the activated effector CD38+ HLA-DR+ CD8+ T cell compartment on day 14 postvaccination (TAE-14) versus those present in the corresponding total PBMC sample from the same time point for subject 1 (A) and for subjects 2 to 9 (B). Clones were classified into four categories based both on their presence in the TAE-14 and the TM-0 compartments, as indicated in the legend. Red clones are present in the TAE-14 compartment, whereas gray clones are not; while clones absent in the TM-0 compartment have a black edge and those present in the TM-0 compartment do not. Darker colors indicate that multiple data points have been superimposed in that particular position. Regions bound by dashed lines indicate clones present in only one sample. YFV-induced clones were significantly enriched in the CD38+ HLA-DR+ CD8+ T cell-sorted population compared to the corresponding total PBMC sample.
For the nine subjects in the study, we detected on average 2,000 clones that were enriched in the TAE-14 compartment compared to the corresponding day 14 postvaccination total PBMC sample from the same individual (2,135 ± 770) (Table 2). This number constitutes a direct estimate of the number of activated, effector CD8+ T cell clones that expand upon binding to HLA:YFV-derived epitope complexes in response to vaccination with YF-17D. In addition, the vast majority of these clones (on average, 91.5% [Table 2]) were absent in the TM-0 population and were thus clearly induced by vaccination with YFV-17D.
Number of YFV-induced clonesa
Characterization of the recruitment of individual clones to immunological memory.Next, we determined which of the YFV-induced clones entered the long-term central and effector memory compartments by analyzing samples obtained from six of the subjects 90 days postvaccination (Table 1). Preliminary studies demonstrated that YFV-induced CD8+ TAE cells return to baseline levels 30 days postvaccination and suggest that YFV Ag-specific cells that are detected beyond this time point correspond to memory cells (6). Therefore, we tracked in the day 90 postvaccination samples YFV-induced clones that were identified as enriched for the TAE-14 compartment but that were absent from the TM-0 compartment (i.e., the putative YFV-specific clones), to determine which were contained in the effector memory compartment (TEM-90, defined as CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ CD62Llo), the central memory compartment (TCM-90, defined as CD3+ CD8+ CD14− CD19− CD45RA− CD45RO+ CD62Lhi), or both. Figure 3A and Table 3 show that 3.1% and 2.5% of YFV-induced clones absent in TM-0 were identified exclusively in the TEM-90 or the TCM-90 compartments, respectively, while 6.7% were identified in both. Moreover, we saw that the degree of expansion of a clone (defined as the abundance of a clone in the day 14 postvaccination total PBMC sample for clones absent in the day 0 prevaccination total PBMC sample) correlated with the efficiency of its recruitment to the memory T cell compartment (Fig. 3B).
Recruitment of YFV-induced clones to immunological memory compartments. (A) Efficiency of recruitment of YFV-induced clones to the effector (TEM+ TCM−) and central (TEM− TCM+) memory compartments, or both (TEM+ TCM+), as a percentage of all clones classified as YFV induced. (B) Efficiency of recruitment to the effector and central memory compartments (or both) for YFV-induced clones absent from the day 0 prevaccination total PBMC samples, classified into categories based on their abundance in the day 14 postvaccination total PBMC samples. Clones with a higher degree of expansion are more efficiently recruited to the memory compartment. The aggregated data for all subjects are shown; subject-wise source data can be found in Table SI in the supplemental material.
Number of YFV-induced clones newly recruited to the TCM-90 and TEM-90 memory compartmentsa
The YFV-induced clones that were newly recruited to the TEM-90 or TCM-90 compartments represent 0.43% and 0.45% (as measured by unique clone counts), or 0.41% and 0.28% (as measured by template abundance) of the corresponding memory compartment aggregated over all samples (Fig. 4A). While the number of templates per unique CD8+ T cell clone in the TEM-90 compartment averaged 8.3, those in the TCM-90 compartment averaged 2.8, indicating that YFV-induced clones recruited to the effector memory compartment are more significantly expanded than those recruited to central memory (Fig. 4B).
Composition of the effector and central memory compartments on day 90 postvaccination. (A) Proportion of YFV-induced clones newly recruited to the effector (TEM-90) and central (TCM-90) memory compartments on day 90 postvaccination, computed both by clone and template counts. (B) Number of templates per YFV-induced clone identified in the TEM-90 andTCM-90 memory compartments. More templates per clone were observed in the TEM-90 compartment, indicating that these clones were more highly expanded. The aggregated data for all subjects are shown; subject-wise source data can be found in Table SII in the supplemental material.
Finally, we were interested in determining whether expanded CD8+ clones that are recruited to the memory compartment possess any particular characteristics that distinguish them from those that become activated upon vaccination but then wane. To address this, we analyzed whether several indicators of specificity (such as CDR3 length or V-J gene usage) correlated with the probability that a given CD8+ T cell clone would be recruited to memory. Although no simple indicator showed an association with recruitment to memory, we found that both the degree of expansion of a clone and the specificity determined by effector sorting (i.e., the fold enrichment in the TAE-14 compartment versus that in the corresponding total PBMC sample from day 14 postvaccination) were positively associated with recruitment.
Concordance between expansion in total PBMCs and enrichment in the activated effector CD8+ T cell compartment.In addition to the data presented above, our approach also allowed the identification of activated, effector CD8+ T cells that expanded massively in response to YFV through the direct comparison of the unsorted, total PBMC samples obtained on days 0 and 14 postvaccination. The statistical method described in detail in Materials and Methods can be applied to the identification of T cell clones that have significantly expanded in a day 14 postvaccination total PBMC sample, compared to the corresponding total PBMC sample from the same individual collected prevaccination (Fig. 1B and 5). Among all the cells present in the day 14 postvaccination sample, we identified a set that was highly expanded but that was not captured by the antiviral-specific TAE-14 flow cytometry sort (i.e., CD38+ HLA-DR+ CD8+ T cells from the day 14 postvaccination). These clones likely corresponded to non-CD8+ T cells that expressed the TCRβ receptor (such as CD4+ T cells), but they could also belong to YF-induced CD8+ T cells that possess different surface markers than those previously reported by Miller et al. (6), or to clonal expansions not induced by the vaccine.
Identification of YFV putatively reactive clones. The graphs show the abundance of unique clones identified by statistical enrichment in the day 14 postvaccination total PBMC sample compared to the prevaccination day 0 total PBMC sample from subject 1 (A) and for subjects 2 to 9 (B). Putatively reactive clones are enclosed by a blue box. Significant enrichment (or expansion) was defined based on a q value threshold, with 1% and 5% expected false-positive rates for YFV-induced and putatively reactive clones, respectively (see Materials and Methods). Clones were classified into four categories based both on their presence in the TAE-14 and the TM-0 compartments, as indicated in the legend. Darker colors indicate that multiple data points are superimposed in that particular position. Regions bound by dashed lines indicate clones present in only one sample.
Finally, to assess how well the expanded CD8+ T cell clones detected in the total PBMC population based only on immunosequencing (i.e., not sorting particular cell populations by flow cytometry) were in concordance with the previously identified TAE-14 clones (i.e., those identified statistically after flow cytometric sorting of CD38+ HLA-DR+ CD8+ T cells), we counted how many expanded CD8+ T cells carrying productive rearrangements identified in the total PBMC sample analysis were classified as YFV-induced through the statistical analysis of the flow cytometry-sorted CD38+ HLA-DR+ CD8+ T cell clones described above. Table 4 shows that a significant proportion of these “putatively reactive” clones—between 25% and 95.2%, depending on the subject—were present in the TAE-14 compartment, suggesting they were induced by YF-17D. In aggregate, 62% of the putatively reactive clones identified as having expanded in the day 14 postvaccination total PBMC sample (compared to the corresponding prevaccination sample) could be classified as YFV induced. These data suggest that our approach has the potential of identifying vaccine-specific responding clones through the characterization of clones expanded in the total PBMC population using exclusively immunosequencing.
Concordance between clones identified as “putatively reactive” in the total PBMC sample and YFV-induced clones identified by their presence in the activated, effector CD8+ T cell compartmenta
DISCUSSION
Using the power of high-throughput immunosequencing, we identified and tracked YFV-induced activated, effector CD8+ T cells as they clonally expanded and underwent phenotypic modification in response to vaccination of human volunteers with the YF-17D vaccine. Previous work using the YFV as a model for an acute, self-resolving viral infection illustrated the general kinetics of the human antibody, CD4+-, and CD8+-based antiviral immune response (6, 9, 10, 12, 16, 30–34). Although some of these studies used either peptide pools or tetramer-based approaches to address the response to a limited number of viral epitopes (6, 9, 33), or calculated the percentage of different immune cellular compartments that represented clones induced by vaccination with YFV (reviewed in reference 16), the clonal breadth and complexity of the response to a viral infection has been beyond the reach of available methods.
In this study, we determined that an average of approximately 2,000 different CD8+ T cell clonal lineages are activated by vaccination with YFV during the acute phase of the immune response and that about 12% of them can be detected in the long-term memory compartment (including both central and effector memory CD8+ T cells). It would be interesting to determine if a similar number of CD8+ T cell clonal lineages are induced by other viral vaccines or by naturally occurring acute viral infections. We also observed that clones that were most expanded on the total PBMC sample from day 14 postvaccination were also more likely to enter the memory compartment 3 months postvaccination, in agreement with previous data (12). Although we were unable to identify other defining characteristics that differentiate CD8+ T cell clones that expand in response to YFV vaccination and are present in the memory compartment on day 90 postvaccination from those that wane during that period, future studies will attempt to characterize these two populations further, including their epitope specificity, since this would yield valuable information that could guide the design of vaccines against other pathogens. Interestingly, almost all of the clones that were markedly expanded in the total PBMC sample from day 14 postvaccination (compared to the corresponding day 0 prevaccination total PBMC sample from the same individual) were classified as YFV-induced CD8+ T cells by the combination of flow cytometry and statistical analysis. In fact, we observed very few clonally expanded T cells in the periphery that were not identified as YFV-induced clones, in agreement with previous reports showing that while CD8+ T cells greatly expand in response to vaccination with YFV, the CD4+ expansion is much less dramatic (6, 30, 32, 33). It is important to consider that the sampling depth used in this study limits the detection of bystander CD8+ cells or of CD4+ T cells that are only modestly expanded. Thus, our current level of detection is likely not sufficient to distinguish CD4+ T cell expansion above the intrinsic system noise.
We did not observe a particular pattern of V(D)J gene usage among the expanded CD8+ T clone repertoire. This result partially agrees with those of a preliminary study of V gene usage performed by Co et al. (34), which used a limited set of anti-human Vβ antibodies. Those authors did not observe a dominant Vβ family that predominated among the tetramer-specific CD8+ T cells in two individuals vaccinated with YFV, but they reported that although gene usage changed over time from the acute to the memory phase, no particular V genes persisted between the acute and memory phases of the antiviral response (34).
Finally, it is noteworthy that many of the CD8+ T cell clones identified as expanded through the comparison of the day 14 postvaccination and the day 0 prevaccination total PBMC samples were classified as likely YFV specific in our initial characterization of clones enriched in the activated effector CD8+ T cells versus the total PBMC sample on day 14 postvaccination. Thus, our approach is capable of identifying a fraction of the highly expanded CD8+ T cells by immunosequencing of total PBMCs prior to infection or vaccination and during the acute response (i.e., 10 to 14 days postvaccination), and our method could be used to ascertain the establishment of long-term memory by sorting memory T cells a few months after infection (or later) and tracking the CD8+ T cells previously identified as being virus induced. Future experiments will address the epitope specificity of the YFV-induced CD8+ clones, by using, for example, tetramer technology to purify clones that bind to previously identified immunodominant YFV epitopes.
A similar strategy could also be applicable to the evaluation of the B cell response to vaccines and viral infections. In conclusion, immunosequencing can be used to characterize the strength and breadth of the B and T cell responses induced by vaccines and viral infections, and it has the potential to be utilized to evaluate novel vaccines in terms of their potential ability to induce effective long-term protective immune responses.
ACKNOWLEDGMENTS
We thank Rafi Ahmed and Rama Akondy for sharing their flow cytometry sorting protocols. H.R. thanks Melissa Alvendia for coordinating the study, and W.D. thanks Bryan Howie for helpful discussions.
This work was funded in part by grant R56 AI0181860 from the NIH to H.R.
H.R. owns stock and receives consulting fees from Adaptive Biotechnologies. W.S.D., R.O.E., M.V., T.M.S., C.D., and C.S. are employees of Adaptive Biotechnologies with salary and stock options.
FOOTNOTES
- Received 9 December 2014.
- Accepted 2 February 2015.
- Accepted manuscript posted online 4 February 2015.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JVI.03474-14.
- Copyright © 2015, American Society for Microbiology. All Rights Reserved.
