Allergic contact dermatitis (ACD) is a T cell-mediated skin disease. It results from the acquisition of skin sensitization to chemicals (metal ions and organic chemicals) encountered at skin surfaces (1). ACD is the most common manifestation of immunotoxicity in humans and can be caused by a wide array of chemicals including nickel (jewelry), fragrances, preservatives, rubber (gloves), dye (hair colorants), adhesives of various kinds, and topical medications, including antibiotics. Little is known about what drives variation in the human immune response to sensitizing chemicals, and consequently a better understanding of the molecular and cellular events that underlie skin sensitization would help improve the safety assessment of the sensitizing properties of new chemicals (2), as well as potentially leading to improved therapeutic strategies. Both CD4+ and CD8+ T cells, interacting with dendritic cells and keratinocytes (3), have been implicated in ACD, but current models suggest CD8+ effector T cells play a major role in the elicitation of contact allergic reactions in mice (4, 5) and probably also in humans (6). Studies in mouse models of skin sensitization have shown that strong skin sensitizers induce an oligoclonal T cell response with a high frequency of CD8 effector T cells (7). The details of the recognition by T cells of chemical allergens are still poorly understood. An exception is the response to divalent metal cations such as nickel, which is a common cause of ACD. Metal ions have been shown to bind directly to major histocompatibility complex (MHC) molecules or to peptide/MHC complexes, thus creating new antigenic determinants that are recognized as foreign by the T cell compartment (8).

As skin-sensitizing chemicals are haptens, and too small to trigger immune responses by themselves, it has been proposed that the vigor of the immune response elicited by different sensitizing chemicals (and thereby their relative potency) might correlate with chemical reactivity and their ability to form stable associations with proteins or peptides (9–11). Greater chemical reactivity has been suggested to translate into a broader range of modified peptides, creating a wider range of potential T cell epitopes and hence driving a broader and stronger T cell response (12). A global analysis of the repertoire of antigen receptors expressed on T cells that respond to a skin sensitizer may provide further insight into the processes, which drive skin sensitization and ACD. In so doing, our aim has been to develop a quantitative, mechanistic understanding of how the dose of chemical sensitizer exposure relates to the likelihood of inducing skin sensitization in humans. It is anticipated that such understanding will inform and improve the safety assessment of chemicals (2, 13).

Recent advances in massively parallel high throughput sequencing have opened the possibility of global analysis of the T cell repertoire, and several protocols have been reported. We have developed a robust single-strand DNA ligation protocol, which tags each molecule of T cell receptor (TCR) αβ messenger RNA in a sample with a unique molecular barcode before PCR amplification and sequencing (14). The most important aspect of our protocol is that it allows us to collect quantitative repertoire data that accurately reflect TCR abundance. Specifically, the molecular barcode allows us effectively to correct both for PCR bias and for sequencing error at the analysis stage (15). This is not currently possible using the commercial DNA-based TCR sequencing platforms. We have also developed a suite of software tools for the analysis of the TCR data obtained, which can be used to process the raw sequence files and assign a unique unambiguous V, J, and complementarity-determining region (CDR)3 to each sequence (16).

In the investigations described here we have applied these experimental techniques to characterize the in vitro response of T cells from individuals sensitized to paraphenylenediamine (PPD), a component of some hair dyes (17) and a strong skin sensitizer (18). PPD–human serum albumin (HSA) conjugate is shown to induce proliferation of a subset of peripheral blood T cells from sensitized individuals. This proliferative response is associated with increased abundance of a small proportion of TCRs, with a highly skewed V and J gene usage. The repertoire of expanded TCRs reflects antigen-driven proliferation of diverse sets of sequence-related TCRs.

The proliferative response of peripheral blood T cells from a cohort of individuals with ACD (as diagnosed by patch test result and clinical history) to PPD was measured in vitro. In preliminary studies we added free PPD to the culture wells, but the chemical proved to be cytotoxic at higher concentrations, and the proliferative response was therefore difficult to quantify reproducibly. Instead, we made conjugates of PPD and HSA and measured the response to the chemical conjugate. As described previously (20), we observed a dose-dependent proliferative response to PPD–HSA, with significantly increased [³H]thymidine incorporation at 10 and 100 µg/ml conjugate (Figure 1A). Further experiments used 100 µg/ml as this gave the maximum proliferation. The proliferation showed a trend to correlation with patch test score (Figure 1B) although there was substantial variability in the proliferation observed in patients with high score, such that the correlation did not reach statistical significance with this number of patients.

We looked in further detail at the phenotypic changes observed in the proliferation cultures. The flow cytometry results on a typical sample with high in vitro antigen-specific proliferation (1 of 12 independent experiments analyzed) is shown in Figure 2A, and the accompanying proliferation to PPD–HSA and to the HSA control are shown in Figure 2B. The CD4 and CD8 cells were analyzed separately and stained with a combination of CD27 and CD45RA to define naive, CM, EM, and EM CD45RA revertants (EMRA) (19). The in vitro culture contained a smaller proportion of fewer CD8+ naïve cells (perhaps because of selective cell death). The addition of antigen (PPD–HSA, 100 µg/ml) caused a small shift from CM to EM in both CD4+ and CD8+ populations, when compared to cells cultured in medium alone. This effect albeit small was seen consistently. The mean decrease in CM CD4+ was 9% (SEM = 2.5, n = 5), and the corresponding increase in EM CD4+ was 9 (SEM = 2.6, n = 5). The mean decrease in CM CD8+ was 2% (SEM = 1.0, n = 5), and the corresponding increase in EM CD4+ was 2% (SEM = 0.8, n = 5). HSA alone caused similar phenotypic changes, suggesting some non-specific carrier-dependent effect. However, thymidine incorporation indicates less cell proliferation in response to HSA, suggesting that HSA may induce phenotypic maturation with less proliferation than PPD–HSA.

Interestingly, we noted that the ex vivo CD4 and CD8 populations showed a profound depletion in the proportion of EM populations in vivo, compared to the values reported previously (19). This observation was consistent among different patients and replicates an observation made previously on PBMC from patients with allergy to methylisothiazolinone (21). Since we did not have access in this study to blood samples taken before the patch test, we could not determine whether the observation related to allergic status or as a response to the exposure to PPD in the diagnostic patch test, so this observation was not pursued further.

We next examined the TCR repertoire of the T cells cultured in vitro. The T cells were collected after culture, pooled to give 0.5–1 × 10⁶ T cells, and processed for TCRα and TCRβ sequencing. The pipeline generates a list of distinct TCR sequences for each sample, together with an abundance indicating the number of times the sequence is found within that sample. The abundance of each distinct TCR found in the ex vivo and in the in vitro stimulated samples (HSA or PPD–HSA) was plotted. We find a clear expansion of TCRα and β sequences following PPD–HSA stimulation that do not expand following HSA stimulation but are present at lower frequency in the HSA only as well as in the ex vivo sample. For each distinct TCR sequence identified after 6 days of culture, we then calculated a stimulation index (SI), defined as the logarithm (base 2, equivalent to the number of cell doublings if cell death is ignored) of the abundance in the in vitro culture relative to the abundance in the ex vivo sample. The abundance of the TCRs that were absent from the ex vivo sample (due to sampling) is estimated as 1 (since abundances of ex vivo samples have a mean of 1.3, median of 1, and TCRs observed in vitro after proliferation must have been present at the start of the proliferation assay with an abundance of at least 1). The SI for cultures with medium alone or in the presence of antigen (PPD–HSA, 100 µg/ml) is shown in Figure 3A. The majority of TCRs show no change or a slight decrease in frequency following in vitro culture. This is consistent with the overall stability of phenotype seen after culture in vitro. However, a small number of sequences are found at much higher abundance after culture in the presence of antigen. The number of distinct TCRs with an SI greater than 3 (eightfold increase or three divisions) in cultures with medium alone, PPD–HSA, or HSA alone is shown in Figure 3B. The qualitative pattern observed is similar to the pattern of thymidine incorporation shown in Figure 2B. The approximately 800 expanded TCRα and TCRβ sequences are equivalent to about 0.8% of the total number of TCRs detected, reflecting the specificity of the proliferative response observed.

We next compared the abundance (Figure 4A) or the frequency (Figure 4B) of the TCRs that were expanded in the PPD–HSA cultures, in the HSA only, medium only (in Figure 4A only), or ex vivo repertoire. In general, the TCRs expanded in PPD–HSA were not observed at high abundance in the other cultures. However, in many cases, the sequence was simply not observed in the corresponding other cultures. This could have been due to differential amplification or sampling (i.e., the relevant precursor was not present in the culture at the start). In order to look more carefully at specificity, we plotted the abundance of all the clones shown in Figure 4A, which were also present at least once in the HSA culture (Figure 4C). The abundance of the TCRs varied quite broadly, with some TCRs showing abundances of >500 after culture in PPD–HSA. In contrast, abundances in the HSA cultures were always less than 10, suggesting that the expansion observed was specific to the presence of the PPD hapten.

We examined whether the TCRs that had expanded in the presence of PPD–HSA (SI > 3) represented a population, which had been pre-expanded in vivo. Figure 5 shows the abundance of the PPD–HSA expanded TCRs in the ex vivo pre-expansion sample. The distribution is compared to the overall abundance distribution in the ex vivo sample from the same donor. The histogram for the expanded set of TCRs shows a clear shift toward higher abundance, with >20% of the expanded TCRs having an abundance of >2 in the PBMC sample. Thus the set of TCRs, which expand in vitro are already present at higher frequencies than average in vivo, consistent with being derived from an allergen-induced memory T cell response.

Several previous studies have reported that antigen-specific responses often favor specific V genes [e.g., Ref. (22)]. We therefore compared the frequency of each V and J α and β gene in the PPD–HSA expanded TCRs with their frequency in the ex vivo pre-culture sample (Figure 6). Since the size of the set of expanded TCRs is rather small (in the order of 800) we selected 100 repeated samples from the ex vivo repertoire to estimate sampling heterogeneity. We then plotted the frequency in the antigen-expanded sample with the frequency in blood. The majority of genes fell in or near the diagonal, indicating a similar frequency in PBMC and in antigen cultures. However, for both V and J genes, one or more gene showed substantial changes after culture. TRAV29/DV5, for example, was present at 10 times its frequency in the matched PBMC sample (Figure 6). Thus in vitro culture with PPD–HSA conjugate drives selective expansion of TCRs with skewed V and J gene usage.

The TCR region with the greatest contact with antigen/MHC complex is the CDR3 loop. The relationship between two CDR3 sequences can be captured by the Levenshtein distance, the number of changes (i.e., substitution, deletion, and addition) required to transform one CDR3 into the other. The amino acid sequences of the 800 expanded TCRs were clustered according to their Levenshtein distance matrix (Figure 7A). The sequences fall into a number of families of closely related sequences. A multisequence alignment of the subtree containing the most abundant TCRβ CDR3 member is illustrated as an example (Figure 7B). The sequences within a family show a high degree of conservation, especially toward the N terminal part of the sequence. However, the overall average Levenshtein distance between all expanded CDRs (α = 9.53, β = 8.61) is only slightly smaller than between a random sample of ex vivo (pre-expanded) TCRS (α = 9.81, β = 9.47). PPD–HSA exposure therefore drives the expansion of families of TCRs, which are strongly conserved within a family but weakly related or unrelated between families.

Paraphenylenediamine is widely used as a hair dye (17) and as a component of black henna tattoos (23), and the overall prevalence of PPD contact allergy was 0.8% in a recent study of European populations (18). The immune response to PPD, therefore, serves as a good model with which to study the immune response to contact allergens and to understand better the nature of allergen recognition. In this study we analyzed the T cell response to PPD at the level of TCR repertoire, highlighting the breadth of TCRs associated with antigen-induced proliferation.

The response to PPD, and to contact allergens in general, is believed to reflect a T cell response, comprising both CD4+ and CD8+ T cells (1). We therefore measured T cell in vitro proliferative responses in PBMC from a cohort of individuals clinically diagnosed as PPD-allergic and who were shown to be patch test positive. A chemical conjugate of PPD with HSA was used as it is less cytotoxic than free PPD and has previously been characterized in detail (20, 24). We first confirmed previous reports that peripheral blood T cells show a dose-dependent antigen-specific response to PPD–HSA, as measured by [³H]thymidine incorporation. In vitro culture was associated with a selective loss of naïve T cells and a small shift from CM to EM in cultures exposed to antigen.

Interestingly, we noted that the starting PBMC population from the PPD-allergic individuals appeared to be depleted of EM cells, compared to the values reported previously for healthy volunteers (19). This might reflect recruitment of effector cells from blood to the skin either due to the patch test carried out 4 days prior to venepuncture or to the allergic condition itself. A similar observation was recently reported for another skin sensitizer, methylisothiazolinone (21). Further longitudinal studies will be required to establish the cause of this relative paucity of EM cells in the peripheral blood and its possible impact on responses measured in vitro.

Our TCRseq protocol incorporates unique molecular identifiers, which allow us to count accurately the number of TCR RNA molecules present in a sample. The relationship between RNA count and cell count depends on the number of TCR RNA molecules per cell. Although we cannot rule out some changes in RNA levels during different phases of T cell activation, previous studies do not suggest substantial changes (25, 26) and would not contribute significantly to the large changes in abundance we observe.

An SI of approximately 9, equivalent to a 500-fold increase in abundance, is consistent with a culture period of 6 days, and T cell replication times of around 24 h for the first generation, and 16–18 h for subsequent generations (27). We do not have dynamic information on T cell numbers during the culture, but since thymidine incorporation is a function of those cells still in S phase during the last day of culture, these results suggest that the observed proliferation can be attributed to a few thousand cells derived from only 2–4 precursors per 100,000 T cells. These TCRs are not expanded in cultures without PPD–HSA, suggesting they represent antigen-specific responses to the PPD conjugate.

Together with a very small number of T cells that continue to proliferate throughout the culture, we also observed many more cells (approximately 0.8% of total), which increase in abundance by a smaller amount. We cannot determine whether these cells represent PPD-specific cells of lower affinity, which respond less robustly to antigen, or perhaps include bystander cells activated in vitro as a result of antigen-presenting cell activation and cytokine secretion.

The sequences of the TCRs expanded in vitro show hallmarks of an antigen-specific response. They exhibit strong skewing of both V and J gene usage (28–30), and they consist of families of TCRs with conserved α and β CDR3s. Nevertheless, the in vitro expanded TCRs show only slightly greater similarity to each other than they do to a random selection of CDR3s from the in vivo repertoire, and the clones showing a high SI are found in several different and quite distinct families of CDR3 sequences.

It is instructive to compare the results observed here with those of an extended series of studies on T cell responses to another strong contact sensitizer, trinitrophenyl (TNP) sulfonic acid [reviewed in Ref. (31)]. Although carried out prior to the era of high throughput sequencing, and in an experimental mouse model, these studies reported remarkably similar qualitative features of the contact sensitizer-specific responses. In particular, the TNP response was associated with a high frequency of hapten-specific T cells (32), was associated with strong preferential usage of certain V and J regions in both alpha and beta chain (33, 34), but was nevertheless very polyclonal involving a wide variety of different TCR sequences (9, 35). These features may therefore represent common features of the sensitizer-specific repertoire in both mouse and man.

The results of our study are consistent with a model in which PPD–HSA conjugate stimulates a broad diversity of TCRs, and with a wide range of stimulation strengths, which manifest as different degrees of in vitro expansion. The reasons for the breadth of response observed remain to be defined. Although PPD has been shown to react principally with a single free cysteine in HSA (20), smaller amounts of alternative conjugate products may also be present, giving rise to a variety of hapten/protein conjugate peptides. Alternatively, a single PPD–HSA-derived peptide may be processed and presented by APC in a variety of ways and recognized by a variety of T cells.

The study begins to characterize the magnitude and breadth of the T cell response to a well-characterized strong skin sensitizer, PPD (36). Further investigations are in progress to separate the contribution of CD4 and CD8 cells and to define the antigen specificity of the responding cells in more detail. The parameters learnt from such studies will help achieve a better understanding of the fundamental characteristics of the T cell response to contact allergens that can in turn be used to improve the clinical relevance of human health risk assessment for chemical skin sensitizers (2, 13).

TO and AP carried out all the T cell culture and T cell receptor sequencing. JW and NG organized patient sample collection and ethical approval for the study. KB, JH, and MI performed the analysis of the T cell receptor sequence data. RD designed the T cell phenotype studies and contributed to the manuscript. GM, IK, and BC designed and supervised the study and wrote the manuscript.

GM and NG are employees of Unilever PLC, which funded the study.