Noninvasive prediction of operational tolerance in liver transplantation based on a peripheral blood transcriptional biomarker panel

Introduction

Long-term survival with high quality of life among liver transplant recipients closely depends on precise management of immunosuppression (IS) (1). Accompanied by the progress of IS therapy, liver transplantation (LT) has been recognized as the optimal life-saving measure for end-stage liver disease (2). However, the potential co-morbidities resulting from lifelong use of IS therapy could not be overlooked, such as malignancy, infection, cardiovascular events, impaired liver and kidney function (1,2). Therefore, the ultimate goal in the field of transplantation immunology can be to achieve an IS-free state after LT (3).

Given the tolerogenic properties of the liver, increasing evidence indicates that the dosage reduction or even withdrawal of IS therapy is feasible for specially selected liver transplant recipients (4). Up to 20% of liver transplant recipients might spontaneously achieve withdrawal of all IS while maintaining stable allograft status for 1 year or more, which was described as spontaneous “operational tolerance (OT)” (5). However, there is still a lack of reliable tools for identifying those with spontaneous OT among liver recipients undergoing IS therapy (6,7).

Currently, the clinical applicability of clinical parameters or graft biopsies for OT prediction is mainly restrained by unreliable prediction accuracy or risk of complications (8). Exploration of noninvasive and precise methods for OT prediction, particularly the circulating biomarkers, offers a direction to break the deadlock (7,9,10). Martínez-Llordella et al. had made the most commendable efforts, which revealed the association between the peripheral blood mononuclear cell (PBMC) expression profiling and OT in LT (11). Subsequently, the roles of PBMC biomarkers for OT prediction in LT have been explored by several research teams (12-15). However, it is undeniable that the previous studies are generally limited by imperfect study design or unadvanced statistical methods, and a powerful combined prediction model incorporating peripheral blood transcriptional biomarkers with clinical parameters for noninvasive OT prediction is highly anticipated.

In this study, we aimed to make full use of the relevant public datasets (including the GSE28842 and GSE11881 datasets) in a reasonable way to identify more reliable peripheral blood transcriptional biomarkers that are indicative of OT, and to establish an excellent gene-based predictive model. Besides, a meaningful combined model integrating the gene-based model and the clinical feature was also developed. We present this article in accordance with the TRIPOD reporting checklist (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-76/rc).

Methods

Data preparation

The GSE28842 dataset shared by Bohne et al. (13) and the GSE11881 dataset shared by Martínez-Llordella et al. (11) were obtained from the Gene Expression Omnibus (GEO) database. According to the methods described in the original articles, liver transplant recipients were weaned off IS therapy after a tapering-off period (6–9 months) and followed up for more than 12 months. If no rejection episodes occurred within 1 year, the recipients were deemed as operationally tolerant liver transplant recipients (TOL) (TOL group). Those developing rejection within 1 year were classified as non-operationally tolerant liver transplant recipients (non-TOL group). Routine liver biopsy (according to Banff criteria) and allograft dysfunction were used for rejection diagnosis. They were performed in accordance with the classic timeline representation of modern IS withdrawal trials in LT (including LIFT trial and OPTIMAL trial) (16,17). In the GSE28842 dataset, baseline pre-weaning serum samples and important clinical characteristics [such as patients’ age, gender, and time between the LT and IS withdrawal (namely “pre-withdrawal time”)] were available from 45 liver recipients (including 20 TOL recipients and 25 non-TOL recipients) before the initiation of drug minimization. In the GSE11881 dataset, peripheral blood samples were collected from 9 TOL recipients at least 1 year after successful IS discontinuation and from 8 non-TOL recipients at the time of rejection requiring reintroduction of IS. The microarray analyses of PBMC samples in the GSE28842 and GSE11881 datasets were both performed using Affymetrix Human Genome U133 Plus 2.0 arrays. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

OT-related genes screening

Normalization was performed using quantiles. The differentially expressed genes (DEGs) in the TOL group compared with the non-TOL group were first identified before IS drug withdrawal (GSE28842 dataset) or after reaching the corresponding endpoints of follow-up (GSE11881 dataset) separately. Then, by utilizing overlapping analysis, the OT-related core genes were discovered. Differential expression analysis was performed using the R package “limma” (18). The threshold was set as “fold change (FC) >1.3 with P value <0.05”.

Model construction and validation

Based on the GSE28842 dataset, least absolute shrinkage and selection operator (LASSO)-logistic regression was carried out using R package “glmnet” (19). According to lambda.min via 10-fold cross-validation, the optimal gene set for modeling was identified. The coefficients (β values) of the genes were calculated by performing multivariate logistic regression. The formula was as follows: gene-based score = (β₁ * expression of Gene₁) + (β₂ * expression of Gene₂) + ... + (β_n * expression of Gene_n). Besides, by using the R package “rms” (https://CRAN.R-project.org/package=rms), a nomogram combining the gene prediction model with clinical features was constructed. Concordance index (C-index), calibration plot, receiver operating characteristic (ROC) curve analysis, and decision curve analysis (DCA) were performed for comprehensive model evaluation. The DCA and ROC analyses were achieved by utilizing the R packages “rmda” (https://CRAN.R-project.org/package=rmda) and “pROC” (20) separately.

Statistical analysis

Mann-Whitney test or Student’s t-test was used to compare non-normal or normal variables between two groups. The performance of the model was assessed by the area under the curve (AUC). Logistic regression analysis was used to analyze the factors associated with OT. SPSS 22.0, R 4.4.2, and GraphPad Prism 8.0.2 were utilized for statistical analysis and graphing. Statistics with a P value <0.05 were considered significant.

Results

Screening of peripheral blood transcriptional biomarkers related to OT in LT

The Figure 1 visually displayed the study design. Inter-sample normalization of the microarray data enrolled in this study was conducted by using the quantiles method (Figure 2A). According to the screening strategy, 23 dysregulated genes between the TOL group and the non-TOL group before IS drug withdrawal were first screened out in the GSE28842 dataset. To ensure the selected genes were indeed related to the immune status of liver transplant recipients, six core genes related to OT in LT were further identified by performing overlapping analysis using the GSE11881 dataset (Figure 2B,2C). A heat map with the cluster analysis was generated to present the relative expression levels of the six genes between the TOL group and the non-TOL group in the GSE28842 dataset (Figure 2D).

Figure 1 The flowchart describing the overall study design. IS, immunosuppression; LASSO, least absolute shrinkage and selection operator; LT, liver transplantation; ROC, receiver operating characteristic; TOL, operationally tolerant liver transplant recipients.

Figure 2 Screening of peripheral blood transcriptional biomarkers related to OT in liver transplantation. (A) Data normalization of the GSE28842 and GSE11881 datasets using quantiles. (B) Volcano plot representing the screening of differentially expressed genes between the TOL group and the non-TOL group before immunosuppression drug withdrawal (GSE28842 dataset) or after reaching the corresponding endpoints (GSE11881 dataset). (C) Identification of the core genes related to OT using overlapping analysis. (D) Heat map displaying the relative expression levels of the 6 genes in different groups. OT, operational tolerance; TOL, operationally tolerant liver transplant recipients.

Construction and validation of a gene-based predictive model

On the basis of the GSE28842 dataset, the best gene set (including IL2RB, SH2D1B, and KLRC1) was selected for model construction by performing LASSO regression analysis (Figure 3A,3B). Through the univariate logistic regression analysis, the correlations between the expressions of the three genes and the OT status (namely, the TOL group versus the non-TOL group) were confirmed (all P<0.05). Then, the three-gene score was developed based on the results of multivariate logistic regression. Three-gene score = (1.208 * expression of IL2RB) + (0.985 * expression of KLRC1) + (0.231 * expression of SH2D1B) (Table S1).

Figure 3 Construction and validation of a gene-based predictive model for operational tolerance in liver transplantation. (A) Using the LASSO-penalized logistic regression model, the optimal gene set for model construction was chosen by 10-fold cross-validation and lambda.min. (B) LASSO coefficient profile of the genes. (C) Comparison of the three-gene score between the TOL group and the non-TOL group. (D) ROC curves exhibiting the predictive abilities of the three-gene score and each gene. ***, P<0.001. AUC, area under the curve; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic; TOL, operationally tolerant liver transplant recipients.

As shown in Figure 3C, the three-gene score of the TOL group (n=20) was significantly higher than that of the non-TOL group (n=25) (P<0.001). ROC analyses indicated that the three-gene score had a good predictive value for OT following LT (AUC =0.794, P<0.05), which was better than that of the three genes alone (Figure 3D).

Establishment and validation of a nomogram combining the gene-based score with the clinical feature

The clinical characteristics of the patients in the GSE28842 cohort were presented in Table S2. The pre-withdrawal time was longer in the TOL group compared with that in the non-TOL group. Consistent with it, logistic regression analyses suggested that “pre-withdrawal time” was associated with the OT status in LT (Table S3). Importantly, “three-gene score” and “pre-withdrawal time” were identified as the independent predictors of OT after LT (Table S4). Hence, a novel transcriptomic-clinical model combining the “three-gene score” with “pre-withdrawal time” was constructed and visualized as a nomogram (Figure 4A).

Figure 4 Establishment and validation of a genomic-clinical predictive model for OT in LT. (A) The nomogram of the combined model integrating the three-gene score with the clinical feature (“pre-withdrawal time”). Comprehensive analysis of the predictive ability of the nomogram for OT in LT using the calibration curve (B), the receiver operating characteristic curves (C), and the decision curves (D). AUC, area under the curve; LT, liver transplantation; OT, operational tolerance.

The good agreement between the nomogram-predicted probability of OT and the actual probability of OT were observed in the calibration plots (Figure 4B). The C-index (equal to the AUC of ROC curves) of the combined model for OT (AUC =0.850) was significantly superior to that of “pre-withdrawal time” (AUC =0.745) (P<0.05) (Figure 4C). In addition, the combined model had a better net benefit than that of “pre-withdrawal time” from the perspective of clinical application value by generating visual DCA curves (Figure 4D).

Discussion

The ultimate goal in the field of transplantation immunology can be to achieve an IS-free state after LT (1). Up to 20% of liver transplant recipients might achieve OT. However, there is still a lack of reliable tools for identifying the propensity to achieve spontaneous OT among liver recipients undergoing IS therapy (7). Given the drawbacks of the clinical parameters or graft biopsies for OT prediction, exploration of noninvasive and precise methods for OT prediction, particularly the peripheral blood transcriptional biomarkers, offers a direction to break the deadlock (8,10). There are several distinctive features in this study compared to the existing research (11,12,14,21). First, instead of simply identifying genes associated with OT before IS drug withdrawal in the GSE28842 dataset from a mathematical perspective, we introduced the GSE11881 dataset to ensure the dysregulated genes were indeed associated with non-rejection after LT as much as possible. In addition, unlike some previous studies that either used a single indicator or unreasonably chose multiple biomarkers for modeling, an advanced statistical approach (LASSO regression) was employed to determine the optimal gene set to construct the multiple gene-based model for OT prediction. Besides, following the identification of the independently predictive values of the gene-based score and the clinical features for OT in LT, a transcriptomic-clinical model was established to enhance predictive capabilities. Crucially, the combined model was visualized as a nomogram to strengthen the clinical applicability.

On the basis of the above improved study design, the three-gene score based on a peripheral blood transcriptional biomarker panel (including IL2RB, SH2D1B, and KLRC1) was constructed and displayed a good predictive power (AUC =0.794) for the probability of OT following LT. Besides, consistent with the findings in the previous studies, “pre-withdrawal time” was identified as one of the most powerful clinical predictors for successful IS drug withdrawal (8,22). Hence, the predictive capability was upgraded (AUC =0.850) by combining the three-gene score with the clinical feature (“pre-withdrawal time”). In clinical practice, various nucleic acid quantification technologies [including quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR), microarray analysis, and RNA sequencing] are easily accessible in most hospitals for laboratory testing, and the testing cost is relatively acceptable. Importantly, PBMC samples and clinical information of the liver recipients are also easily obtainable. In light of the good performance of our model, it can be adapted as a sensitive tool for noninvasive identification of patients with a high probability of achieving OT and to exclude patients unlikely to tolerate IS withdrawal. Under the guidance of the combined model for post-transplant decision-making in LT, it is feasible to select patients in whom the potential benefits of drug discontinuation outweigh the risks of rejection, and the adverse effects and medical costs of IS could be minimized among the liver recipients (5). More importantly, the long-term graft outcomes could be optimized to the greatest possible extent in the era of personalized medicine (2).

It has been reported that the three key genes (including IL2RB, SH2D1B, and KLRC1) are closely involved in immunoregulation in organ transplantation (23-27). Numerous intrahepatic innate immune cells [including the hepatogenic natural killer (NK) cells and NK-T cells (NKT)] are central to immune homeostasis and tolerance induction, which determines the unique immune characteristics of the liver (4,23). Consistent with our findings, Fang et al. investigated the dynamic features of PBMCs in liver transplant recipients using single-cell transcriptome sequencing and found that a novel cell population (namely the CD8⁺KLRC1⁺NKT cells) may be crucial to the formation of immune tolerance in LT. Dynamic monitoring of CD8⁺KLRC1⁺NKT cells after LT may make it possible to detect the development of OT (25). In addition, IL2RB plays a critical role in the dynamic regulation of immune responses by regulating the T cell proliferation, differentiation, and cytolytic effector activity (24). Zhang et al. identified that patients with IL2RB defects are prone to suffer from immunological diseases and a loss of peripheral tolerance (26). Besides, Hruba et al. observed that NK cell-derived SH2D1B expression level was suppressed by steroids in kidney transplantation. The level of SH2D1B in the peripheral blood was significantly higher in the recipients with OT than that of the recipients suffering from acute kidney allograft rejection, which suggests its effects in the regulation of alloimmune response (27). The mechanisms of the above genes concerning OT in LT are worth further exploration.

Many PBMC-derived biomarkers for OT prediction in LT have been published (13-15). Using the same dataset, two predictive signatures for OT were identified in the original article of the GSE28842 dataset, including a 7-gene set (SLAMF7, KLRF1, CLIC3, PSMD14, ALG8, CX3CR1, and RGS3) (AUC =0.71) and a 3-gene set (NCR1, PDGFRB, and PSMD14) (AUC =0.76) (13). Consistent with our findings, most of them were also NK cell-related genes. Based on the improved study design, our three-gene score (including IL2RB, SH2D1B, and KLRC1) showed a relatively better performance (AUC =0.794). In addition, the excellent powers of biomarkers published in another research have attracted significant attention, including FEM1C (AUC =0.967) and SENP6 (AUC =0.933) (14). Notably, Pérez-Sanz et al. conducted this study in a cohort with a relatively small sample size (n=17, TOL/non-TOL =5/12) and using a traditional testing method (qRT-PCR) (14). Besides, Chruscinski et al. proposed an innovative mathematical approach for OT prediction using PBMC data, namely the ratio between an anti-inflammatory gene (FGL2) and the pro-inflammatory gene (IFNG) (15). They found that the PBMC biomarker (FGL2/IFNG ≥1) enriched for the identification of operationally tolerant liver transplant patients (15). It was a pity that they did not perform a diagnostic test and provide the specific AUC of the PBMC FGL2/IFNG. Thus, the heterogeneity and limitations in the existing studies should be acknowledged, which may be caused by the small sample sizes, the different composition of patient cohorts, the various data analysis methods, and a lack of standardized techniques and protocols for sample collection and testing. In the future, a multicenter large sample cohort study is highly anticipated to independently evaluate the values of the multitude of PBMC biomarkers for OT prediction in LT.

There are still several limitations in the present study. First, given the sample size of the GSE28842 dataset is relatively small, this cohort could not be ideally divided into a training cohort and an internal validation cohort. Besides, except for the GSE28842 dataset, no other suitable dataset was available via a comprehensive search in the GEO database. Therefore, the power of the three-gene score for OT prediction before IS withdrawal is pending external validation with multicenter cohorts in the future. In addition, apart from the clinical characteristics (including patients’ age, gender, and pre-withdrawal time), the detailed clinicopathologic information of the recipients is not available in the GSE28842 dataset. Hence, it is unattainable to extensively evaluate the independent predictive indicators and establish a more attractive transcriptomic-clinicopathologic nomogram combining the three-gene score with other crucial predictive clinicopathological factors.

Conclusions

In conclusion, we developed a valuable three-gene score for predicting the OT after LT. Importantly, the genomic-clinical model combining “three-gene score” with “pre-withdrawal time” showed excellent clinical applicability and predictive efficacy. This model might make it feasible to non-invasively and precisely identify the formation of spontaneous OT among liver recipients undergoing IS therapy and safely achieve IS drug withdrawal. Thus, the number of adverse effects and costs related to IS therapy would be effectively reduced, and the long-term graft outcomes would be optimized in the era of personalized medicine.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-76/rc

Peer Review File: Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-76/prf

Funding: This work was supported by the Fundamental Research Funds for the Central Universities (grant No. 2042023kf0043), the National Natural Science Foundation of China (grant No. 82403534), and the Undergraduate Training Programs for Innovation of Wuhan University (grant No. W202510486413).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-76/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Montano-Loza AJ, Rodríguez-Perálvarez ML, Pageaux GP, et al. Liver transplantation immunology: Immunosuppression, rejection, and immunomodulation. J Hepatol 2023;78:1199-215. [Crossref] [PubMed]
2. Terrault NA, Francoz C, Berenguer M, et al. Liver Transplantation 2023: Status Report, Current and Future Challenges. Clin Gastroenterol Hepatol 2023;21:2150-66. [Crossref] [PubMed]
3. Kenison JE, Stevens NA, Quintana FJ. Therapeutic induction of antigen-specific immune tolerance. Nat Rev Immunol 2024;24:338-57. [Crossref] [PubMed]
4. Burra P, Zanetto A, Schnabl B, et al. Hepatic immune regulation and sex disparities. Nat Rev Gastroenterol Hepatol 2024;21:869-84. [Crossref] [PubMed]
5. Dhayanithy G, Radhakrishnan S, Ann Martin C, et al. Understanding immunological insights of liver transplantation: a practice for attaining operational tolerance. Clin Exp Immunol 2025;219:uxae125. [Crossref] [PubMed]
6. Thomson AW, Vionnet J, Sanchez-Fueyo A. Understanding, predicting and achieving liver transplant tolerance: from bench to bedside. Nat Rev Gastroenterol Hepatol 2020;17:719-39. [Crossref] [PubMed]
7. Eurich D, Schlickeiser S, Ossami Saidy RR, et al. How to Estimate the Probability of Tolerance Long-Term in Liver Transplant Recipients. J Clin Med 2023;12:6546. [Crossref] [PubMed]
8. Appenzeller-Herzog C, Hartleif S, Vionnet J. Clinical parameters and biomarkers predicting spontaneous operational tolerance after liver transplantation: A scoping review. Am J Transplant 2021;21:3312-23. [Crossref] [PubMed]
9. Chichelnitskiy E, Goldschmidt I, Ruhl L, et al. Plasma immune signatures can predict rejection-free survival in the first year after pediatric liver transplantation. J Hepatol 2024;81:862-71. [Crossref] [PubMed]
10. Taner T, Bruner J, Emamaullee J, et al. New Approaches to the Diagnosis of Rejection and Prediction of Tolerance in Liver Transplantation. Transplantation 2022;106:1952-62. [Crossref] [PubMed]
11. Martínez-Llordella M, Puig-Pey I, Orlando G, et al. Multiparameter immune profiling of operational tolerance in liver transplantation. Am J Transplant 2007;7:309-19. [Crossref] [PubMed]
12. Martínez-Llordella M, Lozano JJ, Puig-Pey I, et al. Using transcriptional profiling to develop a diagnostic test of operational tolerance in liver transplant recipients. J Clin Invest 2008;118:2845-57. [Crossref] [PubMed]
13. Bohne F, Martínez-Llordella M, Lozano JJ, et al. Intra-graft expression of genes involved in iron homeostasis predicts the development of operational tolerance in human liver transplantation. J Clin Invest 2012;122:368-82. [Crossref] [PubMed]
14. Pérez-Sanz F, Revilla-Nuin B, Martínez-Alarcón L, et al. Tolerance Biomarkers in Liver Transplantation: Independent External Validation of the Predictive Strength of SENP6 and FEM1C Gene Expression. Transplantation 2019;103:1887-92. [Crossref] [PubMed]
15. Chruscinski A, Rojas-Luengas V, Moshkelgosha S, et al. Evaluation of a gene expression biomarker to identify operationally tolerant liver transplant recipients: the LITMUS trial. Clin Exp Immunol 2022;207:123-39. [Crossref] [PubMed]
16. US National Library of Medicine. ClinicalTrials.gov. Liver Immunosuppression Free Trial (LIFT). Available online: https://clinicaltrials.gov/study/NCT02498977
17. US National Library of Medicine. ClinicalTrials.gov. Operations and Pelvic Muscle Training in the Management of Apical Support Loss: The OPTIMAL Trial (OPTIMAL). Available online: https://clinicaltrials.gov/study/NCT00597935
18. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
19. Tibshirani R, Bien J, Friedman J, et al. Strong rules for discarding predictors in lasso-type problems. J R Stat Soc Series B Stat Methodol 2012;74:245-66. [Crossref] [PubMed]
20. Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77. [Crossref] [PubMed]
21. Li L, Wozniak LJ, Rodder S, et al. A common peripheral blood gene set for diagnosis of operational tolerance in pediatric and adult liver transplantation. Am J Transplant 2012;12:1218-28. [Crossref] [PubMed]
22. Appenzeller-Herzog C, Rosat A, Mathes T, et al. Time since liver transplant and immunosuppression withdrawal outcomes: Systematic review and individual patient data meta-analysis. Liver Int 2024;44:250-62. [Crossref] [PubMed]
23. Prosser AC, Klenerman P, Lucas M. Understanding Liver Transplantation Outcomes Through the Lens of Its Tissue-resident Immunobiome. Transplantation 2025;109:e349-61. [Crossref] [PubMed]
24. Campbell TM, Bryceson YT. IL2RB maintains immune harmony. J Exp Med 2019;216:1231-3. [Crossref] [PubMed]
25. Fang Y, Bian C, Li Z, et al. ScRNA-seq revealed disruption in CD8(+) NKG2A(+) natural killer T cells in patients after liver transplantation and immunosuppressive therapy. Immun Inflamm Dis 2023;11:e990. [Crossref] [PubMed]
26. Zhang Z, Gothe F, Pennamen P, et al. Human interleukin-2 receptor β mutations associated with defects in immunity and peripheral tolerance. J Exp Med 2019;216:1311-27. [Crossref] [PubMed]
27. Hruba P, Tycova I, Krepsova E, et al. Steroid free immunosuppression is associated with enhanced Th1 transcripts in kidney transplantation. Transpl Immunol 2017;42:18-23. [Crossref] [PubMed]