Harmonisation of short-term in vitro culture for the expansion of antigen-specific CD8⁺ T cells with detection by ELISPOT and HLA-multimer staining

Accurate assessment of both frequency and functionality of T cells to allow reproducible immune monitoring has been the focus of many laboratories in a wide range of fields [1‐4]. The methods used for T-cell enumeration, i.e. ELISPOT assay and multimer staining, have been the subject of a number of harmonisation studies, in particular in the fields of infectious diseases [1, 5] and cancer, where large proficiency panels have been conducted both in Europe and the United States within the remit of the Association for Cancer Immunotherapy Immunoguiding Programme (CIMT-CIP) [6] and the Cancer Immunotherapy Consortium of the Cancer Research Institute (CIC-CRI) [7], respectively. Such panels have investigated factors that influence the frequency of responses detected by ex vivo IFNγ ELISPOT [6, 8, 9] and multimer staining [10, 11].

More recently laboratories have incorporated a period of in vitro stimulation (IVS) prior to assay by ELISPOT or multimer staining. Several studies have shown that ex vivo IFNγ ELISPOT and cultured ELISPOT measure different subsets of effector and memory T cells [12, 13]. Whilst the ex vivo ELISPOT quantifies effector cells, the cultured ELISPOT can measure T cells of central memory phenotype, which are able to proliferate and consequently acquire effector function [14]. IVS for as little as 1 day, when followed by peptide stimulation for 24–48 h in an ELISPOT assay, has been shown to enhance cellular responses to M. tuberculosis [4]. We have also shown in a DNA vaccine clinical trial in patients with prostate cancer that vaccine peptide-specific responses detectable by ELISPOT could be increased by 33 % following IVS for 9 days prior to assay; 6/30 versus 16/30 responders for ex vivo and cultured ELISPOT, respectively [15]. Hence, IVS provides an effective method to enhance the detection of antigen-specific T-cell populations.

Since laboratories have developed their in-house IVS techniques independently, a diverse set of assay parameters are in use regarding cell concentration/density, length of culture, peptide concentration and type and number of exogenous cytokines, among others [16]. An understanding of how comparable and robust individual IVS methods, including a correlation of results post-culture with ex vivo frequencies, is required to better enable the interpretation of data generated following IVS of PBMCs.

Here, we describe our findings from a 2 stage harmonisation process that examined the robustness and variability of short-term in vitro culture for the expansion of antigen-specific T cells of varying ex vivo frequencies. We first evaluated the ability of 5 distinct IVS protocols, originating from 5 laboratories across Europe, to detect pre-defined antigen-specific responses in multiple donors by ELISPOT assay and multimer staining. Features of the “best-performing” IVS method(s) were integrated to establish a harmonised protocol that was then used in each centre to further evaluate inter-assay (for each centre) and inter-laboratory variation in a second phase of the study.

The following Materials and Methods section is MIATA compliant (www.​miataproject.​org) [17]; further details of each centre’s reagents and protocols are provided in Supplementary MIATA Information.

All centres completed the required assays and returned the raw data to the organising centre for central analysis, which included all replicates unless specifically stated (Supplementary MIATA Information, ELISPOT and Multimer Module 4A).

Data generated from this multi-centre exploratory panel have enabled a better understanding of the impact of IVS on the detection of antigen-specific T cells, including an insight into performance characteristics of different IVS protocols. In Phase I, we compared 5 different IVS techniques, and despite the large number of variables between protocols, all centres were able to expand and detect antigen-specific T cells above a pre-defined criteria for positivity. FLU-specific T cells were found in all three donors, but at varying frequencies. The two donors that displayed the highest frequencies ex vivo also had the highest frequencies following IVS, and in parallel the greatest detection rate (often 100 %). The third donor displayed a lower frequency of FLU-specific T cells and this corresponded to a reduced rate of detection, both for ex vivo and post-IVS assay. This was more pronounced for ELISPOT where ‘missed’ responses were in the main due to the number of IFNγ SFC falling short of 2 SD above background (HIV stimulated wells), and thus failing to achieve the criteria for a positive response. We had made recommendations to participating centres on the number of cells per well to plate following IVS based on our own experience of the responses generated by these donors during pre-screening assays. In hindsight, this recommendation was not optimal since the degree of T-cell expansion achieved by each of the different IVS protocols varied greatly. For some centres, T-cell expansion after IVS was lower and, therefore, the detection of an antigen-specific T-cell response may have benefitted from plating more cells during assay set up. In situations where the presence and level of reactivities are unknown, it may also be beneficial to plate cells at two concentrations to increase accurate enumeration of the antigen-specific T-cell frequency. We identified this also as the cause for responses missed by multimer staining, where too few CD8⁺ T cells were acquired (below 100,000 events). This factor has already been identified and reported as key for detecting antigen-specific T cells by multimer [10, 11].

A harmonised IVS protocol was designed based on the data from Phase I for further testing. Both donors investigated in Phase II had high/intermediate frequencies for EBV- and low for FLU-specific T cells ex vivo (pre-screening), and these were expanded well and reliably across all 6 centres using the harmonised IVS protocol. High detection rates were achieved, with consistent 100 % detection by multimer staining. As with Phase I, positive responses ‘missed’ by ELISPOT were associated with a lower frequency of antigen-specific T cells (FLU), which may be attributed to plating too few cells at set up. Overall, the higher the ex vivo frequency of the response the less variability in expansion was observed across the centres. As expected, the use of the harmonised IVS protocol reduced inter-laboratory variation.

Examination of the T-cell reactivities of our donors by both ELISPOT and multimer staining has enabled us to interpret the ex vivo and post-IVS assays with more clarity. In Phase I, FLU-specific T cell responses detected by ELISPOT ex vivo compared to post-IVS showed a correlation (P < 0.0001), but with a low R ² value. Furthermore, when these responses were dissected and examined for individual donors, significance in this correlation was maintained for only 1/3 donors tested. The correlation was therefore donor dependent and must be a reflection of the phenotype of the cells present in vivo at the time samples were taken. Other studies in HIV [13], EBV [20], Hepatitis C [21] and malaria [22] have found that there is no correlation between ex vivo and cultured ELISPOT when comparing T-cell frequencies to the same antigen. These studies compared groups of donors and did not dissect the data at the level of the individual. Furthermore, our data show that ex vivo ELISPOT and multimer did not correlate and illustrate that these assays measure different T-cell populations within a heterogeneous population of antigen-specific T cells.

However, post-IVS, our data from Phase I and II combined demonstrated that a significant correlation became visible between T cell responses detected by ELISPOT and multimer staining (P < 0.0001 and R ² = 0.5113). A likely explanation is that antigen-specific memory T cells proliferate and acquire an effector phenotype that can readily produce IFNγ for detection by both ELISPOT and multimer staining during IVS [23]. This is also supported by a study by Todryk et al. [12] that showed diminished IFNγ antigen-specific T-cell responses post-IVS when central memory cells are depleted pre-culture.

An important aspect of this study concerned the confidence by which a low level, borderline response can be affirmed to be a true positive or a true negative, a common quandary when testing for T-cell reactivities of unknown but anticipated low magnitude in immunotherapy clinical trials. In the past, we have unambiguously detected PSMA-specific T cells in vaccinated patients following, but not before, IVS by ELISPOT, suggesting that IVS can indeed be a useful tool for revealing reactivities that are undetectable or ambiguous ex vivo [15, 24]. However, in this study, the ambiguous CMV and WT1 responses were not resolved either ex vivo or following IVS despite a large number of assays performed within as well as across centres and by two methods. An increase in the number of assay replicates is generally beneficial in confirming a result; however, where there is inconsistency, as in this study, the decision making process is not helped. With limited patient samples available in trials, a large number of assay repeats may further not be feasible. In these cases, a complementary assay to confirm the presence or absence of antigen-specific response may instead be a viable alternative to assay repeats.

However, care must be taken not to disregard a positive response should the result from multiple assays be conflicting, particularly if the assays measure different T cell qualities such as structural presence of an antigen-specific TCR (multimer) as opposed to a functional ability of antigen-specific T cells (ELISPOT). There are two extra considerations that may lend weight to a clear response in the multimer assay. In contrast to the ELISPOT, the number of cells evaluated can be increased during the experiment, limited only by the total number of cells available. Therefore, the ability to detect low frequency T-cell responses ex vivo may be improved by multimer analysis over ELISPOT. Additionally, multimer staining captures T cells independently of the cell’s functional characteristics and is not influenced by the presence of immunosuppressive cells that may also circulate in patients and that will also be included in the test assay.

Our overall conclusion is that IVS can be a robust and reproducible tool that can be applied even in the multi-centre setting. A several log fold expansion of cells is achievable and that opens extra opportunity for further study, for example, for TCR rescue and cloning. Clearly IVS is not the final solution in every case to resolve the ambiguity of a borderline T-cell reactivity measured ex vivo as we demonstrate here. The most robust discrimination appears to result from a combination of a structural and functional evaluation in parallel. This is the approach we are taking for our future studies by building sample collection in a way that takes this into account.