This observational exploratory study prospectively enrolled consecutive adult LT recipients at the routine month-3 visit at a single tertiary transplant center. Clinical stability was defined as absence of biopsy-proven acute rejection, clinically significant infection, or graft dysfunction requiring diagnostic or therapeutic intervention; no additional selection criteria were applied. All eligible recipients were included (n = 14); none declined or was unable to participate. Standard early post-transplant immunosuppression adjustments were allowed. Given the exploratory design and small sample, analyses were biologically structured to prioritize signal coherence; no formal sample size calculation was performed.

Peripheral blood samples for immunophenotyping were collected at predefined post-transplant time points as part of a prospective translational research protocol embedded in routine clinical practice. Primary cross-sectional analyses used month-3 samples (∼90 days after LT), after most AR episodes occur [15] and beyond the conventional threshold between early and late rejection [16]; restricted to recipients without biopsy-proven rejection, this captures a population beyond the highest early immunological risk while remaining clinically informative. Longitudinal exploratory analyses assessing temporal stability of immune phenotypes between months 3 and 6 were performed in recipients with paired samples available at both time points. Missing data were not imputed, and analyses were performed using available-case data.

The study was conducted in accordance with the Declaration of Helsinki and approved by the local institutional ethics committee. All participants provided written informed consent.

Donor–recipient HLA epitope mismatch burden was primarily quantified using the PIRCHE-II score, which estimates the number of donor-derived HLA peptides predicted to be processed and presented by recipient HLA class II molecules. The global PIRCHE-II score was defined a priori as the primary exposure variable and interpreted as a modeling-based surrogate of indirect alloimmune burden, not a direct functional measure of immune activation. Complete donor and recipient HLA typing data were available for all participants.

As a complementary analysis, the Snow score was calculated as a secondary measure of donor–recipient HLA incompatibility based on B-cell epitope–level surface mismatch, using a previously described computational framework [17]. Because Snow captures class I B-cell epitope mismatch rather than global indirect allorecognition, it was analyzed as a contextual comparator, not a co-primary exposure.

Peripheral blood immunophenotyping was performed using multiparameter flow cytometry to characterize innate and adaptive immune cell populations. Immune subsets were primarily expressed as ratios relative to their respective parent populations to capture qualitative immune differentiation while minimizing confounding due to interindividual variability in absolute leukocyte counts. The selection of immune ratios was predefined based on biological plausibility and prior immunological knowledge, rather than data-driven variable selection, in line with recommended analytical approaches for high-dimensional immune profiling [14, 18].

Four immune parameters were defined a priori as prespecified primary immune endpoints based on biological plausibility and prior evidence: i) the ratio of nonclassical to total monocytes, reflecting skewing toward a patrolling and vascular-surveillance monocyte phenotype [19]; ii) the ratio of CD56⁺CD16^high to CD56⁺CD16⁺ natural killer (NK) cells, representing NK-cell functional differentiation [20]; iii) the ratio of CD8⁺CD38⁺ to total CD8⁺ T cells, as a marker of cytotoxic T-cell activation [21]; and iv) the ratio of IgM⁺IgD⁺ to total B-cells, capturing compositional remodeling within the naïve-like B-cell compartment [22]. These four endpoints constituted the prespecified primary outcomes; remaining variables were analyzed within predefined exploratory immune families and interpreted descriptively. Detailed immunophenotyping procedures, immune ratio definitions, and analytical grouping strategies are provided in the Supplementary Methods. These ratios were selected a priori on biological grounds; one interpretable ratio per lineage (monocyte, NK-cell, T-cell, B-cell) spanned innate and adaptive compartments while limiting primary tests.

Serum samples were analyzed for preformed and de novo DSA using single-antigen bead assays following standard clinical protocols. Analyses relating DSA status to immune phenotypes were exploratory, descriptive, and intended to generate hypotheses rather than to establish causal relationships.

Associations between immune phenotypic variables and HLA epitope mismatch scores were evaluated using Spearman’s rank correlation coefficient (ρ), given the small sample size and non-normal distribution of variables. The primary analysis tested associations between the global PIRCHE-II score and the four prespecified immune endpoints, with Benjamini–Hochberg false discovery rate (FDR) correction applied across these four tests [23].

Secondary exploratory analyses evaluated broader immune ratio sets within predefined families, with within-family FDR correction and descriptive interpretation of signal distribution across immune compartments, without redefining primary conclusions. FDR-adjusted q values <0.05 were considered statistically significant, and q values <0.10 were considered suggestive of association.

Additional exploratory analyses related global PIRCHE-II scores and the four prespecified immune ratios to liver biochemistry and an alloimmune clinical endpoint. Associations with continuous variables were assessed using Spearman’s rank correlation coefficient, and comparisons between groups were performed using rank-sum tests. Given the exploratory nature of these analyses and the limited sample size, results were reported descriptively without formal correction for multiple testing.

All analyses were conducted using available-case data without imputation. Statistical analyses were performed using Stata version 16.0 (StataCorp, College Station, TX). Data visualization and figure generation were performed using R software (version 4.5.3; R Foundation for Statistical Computing, Vienna, Austria).

Fourteen adult LT recipients were included. Median age at transplantation was 62.3 years (IQR 56.5–64.3), and the median MELD-Na score was 15 (IQR 13–18). Most grafts were obtained from donors after brain death, and all patients received tacrolimus-based immunosuppression according to institutional protocols. Baseline recipient, donor, and transplant characteristics are summarized in Table 1. No consistent clinical gradients across the range of global PIRCHE-II scores were formally identified.

The global PIRCHE-II score showed substantial interindividual variability, with a median value of 584.5 (IQR 382.0–689.0). Both HLA class I and class II epitopes contributed to the overall mismatch burden. Among class I loci, contributions from HLA-A, -B, and -C were observed, whereas within class II loci, the highest epitope load was consistently identified at the HLA-DQ locus, followed by DRB and DP. The distribution of global PIRCHE-II scores and class- and locus-specific subscores is detailed in Supplementary Table S1. The median Snow score was 13.0 (IQR 11.0–21.0).

Associations between global PIRCHE-II epitope mismatch burden and the four prespecified immune ratio endpoints are shown in Figure 1. After Benjamini–Hochberg correction across the four prespecified endpoints, higher global PIRCHE-II scores were associated with prespecified immune phenotypes across innate and adaptive compartments within the primary analytical framework.

Specifically, global PIRCHE-II scores were correlated with the proportion of nonclassical monocytes among total monocytes (Spearman ρ = 0.735; FDR-adjusted q = 0.017). A similar association was observed for the ratio of CD56⁺CD16^high to CD56⁺CD16⁺ NK cells (ρ = 0.670; FDR-adjusted q = 0.024). Within the adaptive immune compartment, higher PIRCHE-II scores showed directionally consistent associations with increased proportions of IgM⁺IgD⁺ B-cells (ρ = 0.626; FDR-adjusted q = 0.022) and CD8⁺CD38⁺ phenotypically activated T cells (ρ = 0.600; FDR-adjusted q = 0.023), although these adaptive associations were comparatively less robust and less consistent across exploratory analyses than those observed in innate compartments. All associations were directionally consistent across prespecified endpoints, with the strongest and most robust signals observed in innate immune compartments.

To further contextualize the specificity of these findings, we compared PIRCHE-II with an alternative epitope mismatch framework based on B-cell epitope accessibility (Snow score). As illustrated in Figure 2, PIRCHE-II demonstrated directionally concordant associations across prespecified immune endpoints, whereas the Snow score showed weaker and less consistent associations.

To assess the internal consistency of the prespecified findings, exploratory analyses were performed within biologically defined immune families using PIRCHE-II as the exposure variable, with FDR correction applied within each lineage. Detailed results are provided in Supplementary Table S2.

Within the monocyte and NK-cell families, the associations between PIRCHE-II and the nonclassical-to-total monocyte ratio and the CD56⁺CD16^high to CD56⁺CD16⁺ NK-cell ratio, respectively, remained statistically significant after within-family correction. In contrast, no dendritic-cell ratio met the within-family correction threshold. Within the T-cell and B-cell families, several associations were directionally concordant with the prespecified endpoints but did not remain statistically significant after correction for multiple testing.

These exploratory analyses supported the primary findings by showing that the strongest and most consistent associations localized to monocyte and NK-cell compartments, whereas adaptive immune signals were less stable across broader lineage-specific candidate sets.

Figure 3 summarizes the direction and magnitude of associations across immune families. Notably, the strongest and most consistent PIRCHE-II–associated signals were concentrated within innate immune compartments, particularly monocytes and NK cells, whereas adaptive compartments showed weaker and less consistent associations.

Supplementary Tables S1–S3 provide additional exploratory analyses and detailed numerical results.

Exploratory analyses evaluating associations between global PIRCHE-II scores and absolute immune cell counts did not reveal consistent or systematic correlations across innate or adaptive lineages. Although isolated associations reached nominal significance, they were heterogeneous and not robust after multiple-testing correction. No consistent lineage-wide pattern comparable to immune cell ratios was identified. Detailed results are provided in Supplementary Table S3.

Several cytokines classically associated with acute T-cell–mediated immune activation—namely interferon-γ, tumor necrosis factor-α, interleukin-2, and interleukin-17A—were undetectable in the majority of patients and were not included in downstream correlative analyses.

Among the remaining analytes, including IL-4, IL-6, IL-10, ICAM-1, VCAM-1, and E-selectin, associations with immune cell ratios and with global PIRCHE-II scores were heterogeneous and generally modest, without evidence of a consistent lineage-wide pattern and without a systemic inflammatory signature comparable to that observed for immune cell ratios (Supplementary Tables S4, S5).

DSA assessment at month 3 was available in 10 recipients, of whom 2 (20%) exhibited detectable DSA. At month 6, DSA assessment was available in 7 recipients, with DSA detected in 2 (28.6%). Among the six recipients with DSA assessments available at both time points, one patient exhibited persistent DSA from month 3 to month 6, whereas another developed de novo DSA at month 6; no cases of DSA disappearance were observed. All DSA detected at month 3 were directed against HLA class II antigens. At month 6, a more heterogeneous humoral profile was observed, with one recipient exhibiting isolated class I DSA and the other displaying combined class I and class II DSA. Among DSA-positive cases, the maximal mean fluorescence intensity of class II DSA was 5325 (IQR 1650–9000) at month 3 and 4800 (IQR 2600–7000) at month 6, whereas class I DSA reached a maximal MFI of 4000.

Recipients with detectable DSA at month 3 showed higher PIRCHE-II scores compared with DSA-negative recipients (median 779 [IQR 708–850] vs. 589 [IQR 467–647.5]). Notably, both DSA-positive recipients exhibited PIRCHE-II values above the interquartile range of DSA-negative patients, suggesting directional separation despite the very small number of DSA-positive cases. Given this limited sample size, this comparison was interpreted as descriptive and hypothesis-generating rather than inferential, despite a nominal p value (p = 0.037). The only biopsy-confirmed rejection was a mild acute cellular rejection on day 6 post-transplant, before month-3 sampling and without detectable DSA; one recipient (7.1%) had an early surgical site infection, also before month-3 sampling.

In exploratory analyses, Snow scores showed substantial overlap between DSA-positive and DSA-negative recipients at month 3 (median 17 [IQR 13–21] vs. 13 [IQR 9.5–20.5]; p = 0.60).

Exploratory analyses of predefined antigen-presenting cell–related immune ratios according to DSA status at month 3 suggested a consistent directional pattern. Compared with DSA-negative recipients, those with detectable DSA tended to exhibit lower proportions of ILT3⁺ILT4⁺ dendritic cells relative to total dendritic cells (median 0.24 [IQR 0.09–0.38] vs. 0.40 [IQR 0.19–0.45]) and lower plasmacytoid-to-conventional dendritic cell ratios (median 0.25 [IQR 0.21–0.29] vs. 0.59 [IQR 0.17–0.67]), whereas the proportion of CD83⁺ conventional dendritic cells showed substantial overlap between groups. In addition, DSA-positive recipients displayed higher monocyte-associated CD64-to-CD14 ratios (median 14.28 [IQR 0.78–27.78] vs. 0.23 [IQR 0.06–0.76]). All these comparisons were descriptive and should be interpreted as hypothesis-generating.

To provide additional clinical context for the observed immunophenotypic patterns, exploratory analyses were performed relating global PIRCHE-II burden and the four prespecified immune ratios to graft biochemistry and short-term longitudinal immune phenotype trajectories between months 3 and 6.