Our cross-sectional study included 38 immunosuppressed pediatric patients (IP) at University Hospital La Paz, who received a liver graft between March 2018 and November 2022, and 25 EBV seropositive HC. All patients gave informed consent, approved by the ethics committee of our institution (reference PI-4000).

Demographic and clinically relevant information from each patient were collected ( Table 1 ). PTLD diagnosis was based on histopathologic criteria. Transplant indication was categorized in five groups ( Table 1 ) (53).

Specific T-cell response against EBV was assessed at two different timepoints, determined by patient availability (median time 3.7 months interquartile range [IQR] 3.2-4.9 between visits). EBV serology and viral load were measured in parallel with each immune response assessment for every patient.

Based on their serological status at the first visit, patients were categorized into EBV seronegative (IP-S^NEG) and EBV seropositive (IP-S^POS) individuals; the latter group was further classified according to EBV viral load into negative (IP-S^POSVL^NEG) and positive (IP-S^POSVL^POS). At the time of the second visit, patients were reclassified based on their updated serological and viral load status at that time. Due to sample exclusions for technical reasons, the immune response was not measured in all samples. ICS was performed to 38/38 (100%) and 28/38 (74%) patients at first and second visit, respectively. AIMS was performed to 27/38 (71%) and 29/38 (76%) patients at first and second visit, respectively, although results for both timepoints were available for only 21 individuals. Among HC participants, ICS and AIMS were successfully performed to 24/25 (96%) and 20/25 (84%) samples, respectively.

Heparinized blood samples from all individuals were collected to isolate peripheral blood mononuclear cells (PBMCs) by density gradient.

Analysis of specific T-cell response to EBV by ICS was assessed as described by Lovelace and Maecker (54). Details of the method are provided in the Supplementary Material .

T-cell responses were further categorized as monofunctional, when only one response marker was displayed (CD107a, IFNɣ, Interleukin 2 [IL2] or TNFα), and polyfunctional, when more than one response marker was expressed. Integrated median fluorescence intensity (iMFI) was calculated for each response marker, multiplying MFI by the frequency of the corresponding specific population.

Unstimulated PBMCs background was subtracted from all test samples to obtain the frequency of EBV-specific T lymphocytes. Gating strategy is displayed in Supplementary Figure S1A .

Analysis of AIMS was performed by flow cytometry. Details of the method are provided in the Supplementary Material .

CD4+ T lymphocytes were distributed in naïve (Tn, CD27+CD45RO-), effector (Teff, CD27-CD45RO-), central memory (Tcm, CD27+CD45RO+) and effector memory (Tem, CD27-CD45RO+) subpopulations. Memory compartment was calculated by the sum of Tcm and Tem subpopulations.

Unstimulated PBMCs background was subtracted from all test samples to obtain the frequency of EBV-specific T lymphocytes. Gating strategy is displayed in Supplementary Figure S1B .

Immunophenotype of T, B, natural killer (NK) and natural killer T (NKT) lymphocytes was performed by multiparametric flow cytometry. Details of the method are provided in the Supplementary Material .

EBV viral loads were quantified in whole blood by a specific qPCR assay following manufacturer’s instructions (RealStar^® EBV PCR-Kit 1.0, Altona). Results were informed in International Units per milliliter (IU/mL). The negative group for viral load comprised exclusively patients with zero IU/mL.

EBV serological status was determined by a chemiluminescent microparticle immunoassay. Presence of IgM antibodies against viral capsid antigen (VCA) and/or IgG antibodies against VCA and nuclear antigen (EBNA) were measured following manufacturer’s instructions (Abbott, Germany). EBV seropositive status was defined by the positivity of at least one of the analyzed antibodies.

Descriptive data are presented as median with IQR. Categorical data are presented as absolute number and proportion (%). The software package Prism 8 (GraphPad, USA) was used for statistical analysis. Significance of differences comparing frequencies was determined by Pearson χ2-test and by t test or analysis of variance (Mann Whitney or Kruskal-Wallis tests) when comparing median values. Median frequencies between timepoints were compared by Wilcoxon signed-rank test. Correlation between ICS and AIMS results was assessed by linear regression. P-values under 0.05 were considered significant.

A first classification of our cohort (n=38) was performed according to EBV serological status: IP-S^NEG (6/38, 16%) and IP-S^POS (32/38, 84%). A second distribution of EBV-seropositive patients (n=32) was made according to EBV viral loads: IP-S^POSVL^POS (8/32, 25%) and IP-S^POSVL^NEG (24/32, 75%).

First analysis of our cohort ( Table 1 ) showed statistically significant differences when comparing age (HC adults vs. IP) and EBV-serology pre-transplantation (seropositive vs. seronegative). Interestingly, 52% of post-transplant EBV seropositive patients were negative pre-transplantation. When time since transplantation was analyzed, we observed that IP with positive EBV viral loads had been more recently transplanted (IP-S^POSVL^POS 32 months IQR 12-36 vs. IP-S^POSVL^NEG 44 months IQR 35-47, p=0.002). Percentages of T, B and NK subpopulations were similar among groups ( Table 1 ).

At first visit, EBV-specific %CD3+ T cells by ICS was higher in HC than IP groups (IP-S^NEG 0.03% vs. IP-S^POS 0.04% and HC 0.06%; p=0.41 and p=0.24, respectively), but differences did not reach statistical signification. Nonetheless, all three groups showed similar positive-control stimulation (IP-S^NEG 0.75%, IP-S^POS 0.79% and HC 0.45%; p=0.08), indicating that cellular response in vitro was not impaired by immunosuppression. Positive-control responses remained comparable when dividing mono/polyfunctional CD4+ and CD8+ T-cell subpopulations (data not shown).

However, when splitting EBV-specific response between T-cell subpopulations ( Figure 1A ), statistically significant differences were observed. Median frequency of monofunctional CD4+ and CD8+ T cells was higher than polyfunctional cells in all groups ( Figure 1A ). Interestingly, although monofunctional responses were detected in higher frequencies, only polyfunctional CD8+ T cells significantly discriminated EBV seronegative patients from seropositive HC and IP groups (IP-S^NEG 0.00% vs. IP-S^POS 0.04% and HC 0.02%; p=0.01 and p<0.001, respectively; Figure 1A ).

For a more detailed analysis, the different EBV-specific CD8+ T-cell profiles were analyzed ( Figures 1B, C ). Three different CD8+ polyfunctional profiles differentiated seronegative recipients from both HC and IP seropositive individuals: CD8+CD107a+IFNɣ+IL2+TNFɑ+, CD8+CD107a+IFNɣ+IL2-TNFɑ+ and CD8+CD107a-IFNɣ+IL2-TNFɑ+ ( Figure 1B ). Furthermore, seropositive IP had higher frequencies of EBV-specific CD8+ polyfunctional cells than seronegative IP in two other subsets: CD8+CD107a+IFNɣ+IL2-TNFɑ- and CD8+CD107a+IFNɣ-IL2-TNFɑ+. The profile CD8+CD107a+IFNɣ+IL2-TNFɑ- was the most frequent one (0.13%).

Precisely, this CD8+CD107a+IFNɣ+IL2-TNFɑ- subset significantly discriminated patients with positive viral loads from the rest of individuals (IP-S^POSVL^POS 0.43% vs. IP-S^POSVL^NEG 0.07% and HC 0.03%; p=0.03 and p=0.001, respectively; Figure 1C ). IP-S^POSVL^POS patients also showed higher %CD8+CD107a+IFNɣ-IL2+TNFɑ- compared to HC, but not to IP-S^POSVL^NEG group ( Figure 1C ). Remarkably, polyfunctional response was more intense than monofunctional response in seropositive individuals ( Supplementary Table S1 ). All three cytokine markers IFNɣ, IL2 and TNFɑ had significantly higher iMFI values in polyfunctional response, both in CD4+ and CD8+ T cells, whereas CD107a only showed higher intensity in polyfunctional CD8+ T cells ( Supplementary Table S1 ). In line with our previous results, CD8+ polyfunctional subpopulation allowed discriminating seropositive from seronegative status, according to iMFI from all four response markers ( Table 2 ). However, regarding EBV viral load, only total (mono and polyfunctional) CD107a iMFI on CD8+ T cells significantly differentiated IP with positive viral loads from the other groups (IP-S^POSVL^POS 123,398 vs. IP-S^POSVL^NEG 20,708 and HC 21,207; p=0.03 and p=0.01, respectively).

To simplify cytometry panels, we studied whether CD8+ T cells expressing only CD107a and IFNɣ markers could be distinctive, regardless of other cytokines. As expected, %CD8+CD107a+IFNɣ+ T cells were significantly higher among seropositive individuals compared to seronegative IP (IP-S^NEG 0.02% vs. IP-S^POS 0.37% and HC 0.12%; p=0.002 and p=0.04, respectively). However, considering EBV viral load, although the frequency of specific cells was also increased within positive IP group, differences were significant compared to controls, but not to negative patients (IP-S^POSVL^POS 0.70% vs. IP-S^POSVL^NEG 0.19% and HC 0.12%; p=0.10 and p=0.03, respectively).

Finally, we explored whether CD3+IFNɣ+ T cells, the main population targeted by other methods, allowed discrimination of serology and/or viral-load status. While frequency of CD3+IFNɣ+ T cells was insufficient for serology discrimination (IP-S^NEG 0.10% vs. IP-S^POS 0.31% and HC 0.12%; p=0.31 and p>0.99, respectively), iMFI was significantly higher in seropositive than seronegative individuals (IP-S^NEG 182 vs. IP-S^POS 3,628 and HC 1,380; p=0.003 and p=0.01, respectively). When comparing viral-load status no statistically differences were reached either with CD3+IFNɣ+ frequency (IP-S^POSVL^POS 0.37% vs. IP-S^POSVL^NEG 0.24% and HC 0.12%; p>0.99 and p=0.40, respectively) or iMFI (IP-S^POSVL^POS 3,834 vs. IP-S^POSVL^NEG 2,809 and HC 1,380; p=0.62 and p=0.82, respectively).

At first visit, differences were found in EBV-specific T-cell frequencies by AIMS between HC and IP groups ( Figures 2A, B ). As expected, CD4+ Tn EBV-specific subset was the lowest within each group, compared to Teff and memory compartment ( Figures 2A, B ). On the other hand, seropositive groups ( Figure 2A ) had higher frequency of EBV-specific memory compartment than Teff cells (IP-S^POS memory compartment 0.99% vs. Teff 0.04%, p=0.12; HC memory compartment 8.24% vs. Teff 0.37%, p<0.001), although differences were statistically significant only within HC group. Interestingly, seronegative IP showed higher frequencies of Teff than memory EBV-specific cells (memory compartment 0.00% vs. Teff 0.25%, p=0.14), although differences did not reach statistical signification.

Compared to IP seronegative group, percentages of EBV-specific cells were higher in HC group for Tn (0.14% vs. IP-S^NEG 0.00%, p=0.03), Tem (4.31% vs. IP-S^NEG 0.00%, p=0.01) and Tcm (4.33% vs. IP-S^NEG 0.00%, p=0.002) subpopulations ( Figure 2A ). However, comparing frequencies between seropositive and seronegative IP did not yield any statistically significant differences. Again, positive-control stimulation in CD4+ T cells was comparable by AIMS (IP-S^NEG 24.52%, IP-S^POS 32.67% and HC 26.42%; p=0.62).

We next studied seropositive IP grouped by viral-load status ( Figure 2B ) and observed that %CD4+ Tcm cells was significantly higher in HC than seropositive IP with detectable EBV (IP-S^POSVL^POS Tcm 0.13% vs. HC Tcm 4.32%, p=0.02). Interestingly, median %CD4+ Tem cells in seropositive IP-S^POSVL^POS was higher than in IP-S^POSVL^NEG (0.90% vs. 0.34%, respectively; p>0.99), although no significant differences were found.

Finally, we further investigated the potential correlation between the parameters defined by ICS (%CD8+CD107a+IFNɣ+IL2-TNFɑ-) and AIMS (%CD4+CD134+CD25+ Tcm), which effectively distinguished positive from negative individuals. We noted that those parameters exhibited no correlation (data not shown), likely due to the comparison involving distinct T-cell subpopulations. Thus, a potential correlation between EBV-specific CD4+ T cells by ICS (% and iMFI CD4+INFɣ) and AIMS (%CD4+CD134+CD25+ Tcm) techniques was next sought, although we did not observe any correlation, by either frequency or iMFI (data not shown). However, within HC group, after excluding the data from one individual lacking EBV-specific CD4+CD134+CD25+ memory T cells, a significant correlation between iMFI CD4+INFɣ by ICS and %CD4+CD134+CD25+ by AIMS (r² = 0.24 and p=0.04, Figure 3 ) was found.

EBV-specific response was measured on a second visit by both techniques. Serological and viral-load status was re-evaluated and patients were reclassified accordingly. Tacrolimus blood levels remained similar at first and second timepoints in all three groups (data not shown).

At the second timepoint, identical results to those reported at the first visit were found when comparing frequencies of EBV-specific cells detected by ICS and AIMS. We confirmed that only polyfunctional CD8+ specific T cells significantly discriminated EBV seronegative patients from seropositive individuals (IP-S^NEG 0.00% vs. IP-S^POS 0.03% and HC 0.02%; p=0.01 and p=0.02, respectively). Likewise, %EBV-specific cells from HC were higher than those detected in seronegative patients for Tn (0.14% vs. 0.01%, p=0.05), Tem (4.31% vs. 0.00%, p=0.01) and Tcm (4.33% vs. 0.00%, p=0.003) subpopulations.

Regarding IP-S^POSVL^POS group, all recipients with detectable EBV at first timepoint cleared viral loads at the second visit (median time 3.5 months IQR 2.2-5.3 between timepoints). Consequently, median %CD8+CD107a+IFNɣ+IL2-TNFɑ- (first visit 0.26% vs. second visit 0.10%, p=0.81; Figure 4A ) and %CD4+CD25+CD134+ Tcm cells (first visit 0.39% vs. second visit 0.14%, p=0.88; Figure 4B ) decreased at the second timepoint, although no significant differences were found.

On the other hand, only two IP-S^POSVL^NEG patients at first timepoint had detectable EBV at the second visit (black arrows in Figures 4C, D ) (median time 3.7 months IQR 3.3-4.8 between timepoints). Interestingly, significant differences were found by ICS (first visit 0.08% vs. second visit 0.06%, p=0.04; Figure 4C ), but not by AIMS (first visit 0.00% vs. second visit 0.76%, p=0.31; Figure 4D ). Frequency of CD8+CD107a+IFNɣ+IL2-TNFɑ- cells decreased or remained similar for all patients, except for the individual who tested positive for EBV at the second visit, whose frequency increased from 0.06% to 0.09% (black arrow in Figure 4C ). This patient’s frequency of specific CD4+CD25+CD134+ Tcm cells also increased from 0.00% to 1.06% (black arrow in Figure 4D ). Nevertheless, that variation in the specific response by AIMS was not observed in the other patient who tested positive for viral load at the second timepoint (black arrow in Figure 4D ). One IP-S^POSVL^NEG patient had his immunosuppression regimen changed between visits, incorporating mycophenolate to his treatment with corticosteroids and tacrolimus. Interestingly, frequency of EBV-specific T cells by ICS decreased from 0.44% to 0.25%, while frequency of CD4+CD25+CD134+ Tcm cells remained at 0.00% (white arrows in Figures 4C, D ).

Finally, all seronegative patients ( Figures 4E, F ) kept their negative serological status at the second visit (median time of 3.9 months IQR 3.6-4.0 between timepoints). Consequently, no differences in EBV-specific response were found, either by ICS (0.01% vs. 0.00%, p=0.44; Figure 4E ) or AIMS (0.00% vs. 0.00%, p>0.99; Figure 4F ).