Evaluation of T-activated proteins as recall antigens to monitor Epstein–Barr virus and human cytomegalovirus-specific T cells in a clinical trial setting

Epstein–Barr virus (EBV) and human cytomegalovirus (HCMV) are ubiquitous human herpesviruses, which establish persistent infections in the human population, while maintaining the capacity for reactivation to lytic infection [1‐3]. Primary infection with EBV and HCMV is in most of the cases asymptomatic, but some individuals develop infectious mononucleosis (IM) (EBV) [4, 5] or a mononucleosis-like illness (HCMV) [6]. Both, EBV and HCMV can cause severe disease in immunocompromised individuals in whom the protective antiviral T-cell response is diminished [1], e.g. 80% of posttransplant lymphoproliferative disorders are associated with EBV infection [7]. EBV infection is also implicated in the pathogenesis of several life-threatening or disabling malignant and non-malignant human diseases [8], like Burkitt and Hodgkin lymphoma [9], nasopharyngeal carcinoma [10, 11], as well as autoimmune disorders such as multiple sclerosis [12]. Regarding HCMV, a sufficient risk stratification for solid organ transplantation is crucial to find the adequate dosage of immunosuppression to prevent recurrent HCMV reactivations on the one hand, but to avoid graft rejection on the other hand [13‐15].

Hence, highly sensitive and validated immune monitoring assays are needed to assess the EBV- and HCMV-specific cell-mediated immunity (CMI) in e.g. transplant patients or as an immunological endpoint in immunotherapy trials to monitor secondary or exploratory immune response endpoints (i.e. to measure vaccine immunogenicity or summarize several longitudinal immunogenicity responses) and to allow for therapy decisions and correlation of assay results with clinical outcome. Immunoassays suitable to determine the frequency of antigen-specific CD4 and CD8 T cells (e.g. enzyme-linked immunosorbent assay (ELISA), enzyme-linked immunospot (ELISpot)- and flow cytometry-based assays) require an antigenic component ideally able to (i) reflect the in vivo situation including interaction of antigen presenting cells (APC) and effector cells, (ii) ensure highly sensitive monitoring while maintaining a high specificity, (iii) induce major histocompatibility complex (MHC) class I and II restricted T-cell responses, (iv) induce different effector functions, and (v) including all clinically relevant immune cells and their dynamic interaction.

The choice of immunogenic stimulants for virus-specific CD8 T-cell monitoring is hampered by the fact that CD8 T-cell reactivation requires antigen processing via the natural endogenous class I presentation pathway [16]. Exogenous full-length (viral) proteins and virus lysates (e.g. HCMV lysate) are usually processed by APC via the exogenous antigen processing pathway and presented by MHC class II molecules, and therefore they can only be used for CD4 T-cell monitoring [16]. In contrast, soluble peptides with an appropriate length of amino acids (aa), can directly bind to MHC class I molecules located on the surface of cells without prior antigen processing [16]. Therefore, most immune monitoring approaches are based on the use of overlapping synthetic peptide pools (PP), spanning the entire aa sequence of the desired viral protein. However, functionality of APC to capture and process viral antigens for MHC class II and I restricted epitope presentation to viral-specific T cells is not considered in this approach [16]. Pools of overlapping peptides have been used for epitope mapping [17, 18], investigating T-cell responses in vaccine development [19‐21], and immunotherapies [22, 23]. The use of PP, however, has several drawbacks: commonly used PP contain 15–20 mers, which is at least for MHC class I restricted epitopes a suboptimal length [24, 25]. Additionally, aa length of the single peptides as well as the position of the T-cell epitope within a peptide can greatly influence the assay sensitivity [26]. The high number of peptides within one pool may also affect the outcome of T-cell monitoring as it is not known to what extent individual peptides interfere with each other in large pools of > 100 peptides [27].

As synthetic peptides can directly bind to MHC molecules without requiring antigen processing, impaired antigen processing and presentation, which may appear in certain patient cohorts is not taken into account. This is extremely important when dealing with immunocompromised patients (e.g. transplant recipients), because it is well known that immunosuppressive drugs interfere with maturation and functionality of APC [28‐31]. PP-based immune monitoring in such patient cohorts may result in an overestimation of responding T cells in vitro, probably not reflecting the in vivo situation in patients [32]. These facts suggest that synthetic PP probably do not reflect the naturally processed peptide repertoire and may be a sub-optimal antigenic component for antiviral immune monitoring approaches of certain patient cohorts.

Specifically engineered recombinant EBV- and HCMV-derived T-activated proteins (TP) are protein antigens that have been urea-formulated (T-activation technology, Lophius Biosciences, Regensburg, Germany) to enter the endogenous antigen presentation pathway more efficiently [33]. Thus, TP-derived peptides are not only presented by MHC class II, but also due to cross-presentation by MHC class I molecules [33]. Due to their natural antigen processing, TP overcome some limitation of PP-based immune monitoring and by activating all clinically relevant immune cell populations they provide an innovative antigenic component for different kinds of immunoassays [33].

Just recently, it was shown that HCMV-derived IE-1 and pp65 TP represent suitable stimulants to monitor functionality of HCMV-specific immunity [34‐36]. However, a comprehensive comparative study of TP and PP induced CMI has not yet been performed.

The magnitude of virus-specific CD4 and CD8 T-cell responses and their diversity (i.e. the cytokine expression pattern) are considered significant parameters in evaluating the efficacy of an immune response [37‐39]. In order to assess suitability and comparability of currently used immunogenic stimulants for immune monitoring of EBV- and HCMV-reactive T cells in clinical trial settings (e.g. to assess EBV- and/or HCMV-specific cell mediated immunity after solid organ or allogenic hematopoietic stem cell transplantation [40, 41]), we determined the number and cytokine expression pattern of CD4 and CD8 T cells upon stimulation with two EBV- and two HCMV-derived TP and their corresponding overlapping PP. Further, we determined the specificity of TP and the TP- and PP-induced response rates when using different specimen material [i.e. freshly isolated versus cryopreserved peripheral blood mononuclear cells (PBMC)].

Our study provides important insight into the validity of these new immunogenic virus-specific stimulants for endpoint monitoring in immunotherapy trials, confirming TP as an innovative tool for antiviral immune monitoring. Considering the natural antigen processing pathway is advantageous for immune monitoring of immunocompromised patient cohorts, where impaired antigen processing needs to be considered.

The authors acknowledge the concept of the Minimal Information About T-cell Assays (MIATA) framework [42, 43]. Therefore, detailed information is provided as structured in the five modules proposed by MIATA (miataproject.​org/​).

Monitoring of EBV-/HCMV-specific cell-mediated immunity (CMI) is co-decisive for antiviral therapy in transplant settings and is often determined as an immunological endpoint in clinical trials [13, 14, 48]. Assessment of HCMV-specific T-cell responses was reported to be an adequate tool to decide on the appropriate level of immunosuppressive therapy in transplant patients to avoid HCMV-reactivation but also graft rejection [14, 15, 35]. Both approaches require highly sensitive and validated immunoassays, and optimal antigenic stimulants to monitor virus-specific CMI [49].

We hypothesize that T-activated proteins (TP) may be superior to commonly used stimulatory viral antigen preparations (i.e. overlapping peptide pools, PP), because they enable an immune monitoring close to the patients in vivo situation by considering natural antigen processing pathways and activation of all clinically relevant effector cell populations [34]. TP as stimulatory antigens may improve immunological endpoint assays, monitoring virus-specific CMI in a clinical trial setting [35, 36]. In this study, we performed comparative analyses of frequencies, phenotype, and cytokine expression pattern of T cells stimulated by EBV-BZLF1, -EBNA3A, HCMV-IE1, and -pp65 TP or the corresponding overlapping PP. In addition, we investigated the suitability of TP as antigenic component in immunoassays with regard to specificity and stimulatory capacity.

Currently used stimulatory antigens are mainly full-length recombinant proteins or corresponding synthetic PP [50, 51]. Both are useful in addressing CD4 or CD8 related T-cell responses, however, PP-based monitoring may consider certain key functions of CMI (i.e. antigen processing/presentation by APC and the communication between immune cells) insufficiently [16]. The debilitating influence of immunosuppressive drugs on T-cell immunity and APC function has been extensively investigated [28‐31]. Using overlapping PP for detection of virus-specific T-cell responses does not necessarily require peptide processing by APC [16]. It is known that extracellular proteases can trim peptides to an optimal length and therefore overcome the natural intracellular processing step of viral proteins by APC [16, 52]. Hence, a diminished function of APC caused by immunosuppressive therapies may not be considered by a PP-based immune monitoring approach. In contrast, exogenous antigens like whole viral proteins and viral TP, need to undergo natural antigen processing pathways and are presented by APC [34, 53].

One of our main findings was that the overall magnitude of ex vivo detectable antigen-reactive CD4 T cells was comparable, irrespective of the recall antigen confirming the usability of TP to monitor frequencies of antigen-reactive CD4 T cells. In contrast, we observed significant higher median frequencies of EBV-EBNA3A, but significant lower median frequencies of HCMV-pp65 reactive CD8 T cells upon TP vs. PP stimulation. The used PP contain 15 mers, which is a suboptimal length for MHC class I restricted epitopes [24, 54], whereas peptides bound by MHC class II typically range from 12 to 20 aa in length [55, 56]. The amino acid length of single peptides, as well as the position of the T-cell epitope within a peptide, can greatly influence results of T-cell based immunoassays [26]. Since extracellular proteases can trim peptides to a more optimal length fitting MHC class I molecules [52], the standard format of 15 mer peptides could represent a good compromise for stimulating both CD8 and CD4 T-cell responses and is widely used for the monitoring of antigen-reactive CD8 T cells. However, it is likely, that peptide trimming by proteases may result in peptides differing in length other than intracellular processing and that these differently generated peptides address various subsets of memory T cells.

TP are cross-presented to CD8 T cells via MHC class I molecules [33]. Thus, immunoassays based on peptides resulting either from natural processing (TP) or trimming (PP) may address and activate different subsets of virus-specific memory T cells. Indeed, at least our results for EBV-EBNA3A and HCMV-pp65 reactive CD8 T cells seem to confirm this hypothesis. This is in line with clinical reports, showing that a certain peptide based ex vivo monitoring of virus-specific T cells did not correlate with the patient´s clinical outcome [32, 57]. EBV-EBNA3A and -BZLF1 TP are currently evaluated as one of three antigenic components in an ongoing, prospective multicentre clinical observational study (Munich infectious mononucleosis (IMMUC) study). The IMMUC study aims at the identification of biomarkers and causative factors of complicated and/or protracted EBV-associated IM. Correlation analysis of detected EBV TP- and PP-triggered T-cell responses with serological and clinical data will provide data on the clinical relevance of a TP-based immune monitoring.

Interestingly, for some individuals we determined a high T-cell response only upon stimulation with one of the two types of antigenic stimulants (either PP or TP), which was more pronounced for the CD8 T-cell subset. Based on the differences of individual TP and PP processing and trimming, we speculate that on a single subject level TP-derived peptides can significantly differ from the corresponding PP composition. For many viral antigens it is yet unknown what proportion of the processed peptides is able to trigger immune responses, because not all peptides presented on MHC class I or II molecules are immunogenic [58]. Profiling of a certain viral immunopeptidome by mass spectrometry to identify MHC ligands and testing of their immunogenicity in vivo, as recently shown for vaccinia virus-derived peptide ligands, would promote the selection of optimal antigenic stimulants [58].

Another important observation of our study was that for all tested EBV- and HCMV-derived antigens, TP and PP trigger significantly different cytokine expression pattern of antigen-reactive CD8 T cells, whereas CD4 T-cell cytokine expression pattern was only different for EBV-EBNA3A reactive CD4 T cells. T-cell cytokine profiling is often determined by immunological endpoint assays monitoring virus-specific T cells in a clinical trial setting to identify correlates of protection [13, 37‐39, 59, 60]. There are first hints that a PP-based immune monitoring cannot predict the clinical outcome in certain patient cohorts. Results of a HCMV PP-based CD8 T-cell monitoring of patients under corticosteroid therapy do not correlate with clinical outcome and monitored CD8 T cells do not seem to be protective against viral reactivation [32]. Also La Rosa et al. did not detect any association between HCMV-pp65 and -IE1 PP-triggered T-cell responses and HCMV-associated disease and viremia in liver transplant patients 3 months after engraftment [57]. However, Sester et al. [61] reported of a correlation and predictive potential of HCMV-specific CD4 T cells with virus control and HCMV-associated disease, using HCMV antigen for ex vivo analysis. This is in line with our results showing comparable cytokine expression profiles of HCMV-pp65, -IE-1, and EBV-BZLF1 reactive CD4 T cells upon stimulation with PP and TP.

Testing the clinical validity (as performed in our current IMMUC study for EBV-specific T-cell responses) will investigate whether TP-based T-cell cytokine profiling is superior in reflecting the in vivo situation in patients and is suitable to identify correlates of protection.

The evaluation of antigens as stimulatory components for T-cell monitoring also includes proving for specificity (true negative rate) or presence of cross-reactive T-cell responses [44, 50, 62, 63]. We determined a specificity of 100% for EBV-BZLF1, HCMV-IE1, and -pp65 TP and 86% for EBV-EBNA3A TP. T-cell responses against EBV-EBNA3A detected in one out of seven EBV-seronegative individuals does not match necessarily a false positive response. Savoldo et al. [64] reported of EBV-seronegative subjects who had sporadically detectable EBV DNA loads and EBV-reactive T cells. They assumed that most likely those subjects failed to produce EBV-specific antibodies, but were able to establish EBV-specific memory T cells [64]. Existence of heterologous T cells cross-reactivity between EBV and other antigens has already been described [65, 66] and may explain the detection of EBV-EBNA3A reactive T cells in a single EBV-seronegative subject. To exclude false positive responses reliably, we would recommend determining a positivity cut-off based on data of a sufficiently large cohort of individuals seronegative for the respective pathogen.

Immunological endpoint assays monitoring antigen-reactive T-cell responses can be performed on freshly isolated or cryopreserved specimen. Analysis of fresh PBMC, although logistically challenging, is often preferred because cryopreservation might affect functional and phenotypic properties of cells [67, 68]. Previous reports showed that cryopreservation has no negative influence on the overall percentage of CD4 and CD8 T-cell subsets [67] as well as on certain CD4 T-cell subpopulations, like naïve (CD4+ CD45RA+ CD95−) and activated (CD4+ CD38+ HLA-DR+) CD4 T cells [69]. Other T-cell subpopulations, like activated CD8 T cells (CD8+ CD38+ HLA-DR+) or memory CD4 T cells (CD4+ CD45RO+) were reported to be decreased in cryopreserved compared to fresh PBMC [69]. Contradictory data exist on alterations of antigen-specific T-cell responses upon cryopreservation [67, 70, 71]. However, contradictions might be explained by different sources (healthy individuals vs. patients) and processing procedures (freezing/thawing SOPs) of biosamples.

Besides reported changes in the immunophenotype and alterations in antigen-specific T-cell responses, the usage of cryopreserved PBMC in a central immune monitoring unit minimize inter-assay and inter-laboratory variation both being well-known problems in many multi-centre clinical trials with local sample processing [70, 71].

To evaluate the influence of the sample material we determined the overall response rates to EBV-BZLF1, -EBNA3A, and HCMV-IE1, and -pp65 TP or their corresponding overlapping PP with freshly isolated versus cryopreserved PBMC in parallel.

With freshly isolated PBMC, we observed a comparable high response rate for both TP- and PP-based assay set-ups. Furthermore, the concordance with HCMV-serology of the TP-based assay was in line with previously published data on HCMV-specific antigens generated with freshly isolated PBMC [35, 36, 50, 72]. For both antigenic stimulants, we observed lower response rates when using cryopreserved PBMC. Thus, a proportion of TP-reactive T cells will not be considered when using cryopreserved PBMC, but the same holds true to a lesser extent for a PP-based monitoring. One explanation for the decreased stimulatory capacity of TP on cryopreserved PBMC could be that (cross-) presentation of MHC class I restricted peptides implies a fully functional processing of proteins by APC, which might be affected in cryopreserved material [16, 73]. However, based on our results, we can rule out a generally impaired APC function due to cryopreservation as HCMV-pp65 TP triggered CD4 T-cell responses in 95% of all tested individuals irrespective of whether freshly isolated or cryopreserved PBMC were used.

Based on our evaluation, we conclude that EBV- and HCMV-derived T-activated proteins are suitable antigenic components for an ex vivo monitoring of antigen-reactive CD4 and CD8 T cells and with restrictions also when using cryopreserved PBMC.