Abstract
Objective
T cell-mediated rejection (TCMR) remains a significant challenge in
organ transplantation. This study aimed to define a TCMR-associated
cytokine gene set and identify drugs to prevent TCMR through drug
repurposing.
Methods
Gene expression profiles from kidney, heart, and lung transplant
biopsies were obtained from the Gene Expression Omnibus (GEO) database.
Differentially expressed genes (DEGs) between TCMR and non-TCMR groups
were identified, and their intersection with cytokine-related genes
yielded an 11-gene TCMR-associated cytokine gene set (TCMR-Cs). To
evaluate the effectiveness of this gene set, a diagnostic predictive
model was constructed using Lasso regression and multivariate logistic
regression, with validation in independent datasets. Connectivity Map
(CMap) analysis was employed to screen drugs targeting TCMR-Cs.
Experimental validation of the identified drug was performed in vitro
using T cell activation and Th1 differentiation assays, and in vivo in
a mouse skin transplant model with survival analysis.
Results
The TCMR-Cs exhibited outstanding predictive performance for TCMR,
achieving an AUC of 0.99 in the training cohorts and maintaining strong
performance in the test cohorts. CMap analysis identified peroxisome
proliferator-activated receptor gamma (PPARγ) agonists as potential
therapeutic candidates. Experimental validation showed that the PPARγ
agonist rosiglitazone significantly suppressed T cell activation and
reduced Th1 differentiation in vitro without cytotoxic effects. The
combination of rosiglitazone and rapamycin significantly prolonged
graft survival.
Conclusions
This study defined a novel TCMR-associated cytokine gene set that
effectively predicts TCMR and identified PPARγ agonists, which prevent
TCMR and improve graft survival when combined with rapamycin.
Keywords: T-cell-mediated rejection, organ transplantation, cytokine
genes, predictive model, drug repurposing, PPARγ agonists,
rosiglitazone
1. Introduction
Organ transplantation is a life-saving treatment for patients with
end-stage organ failure, significantly improving survival rates and
quality of life ([37]1). Kidney, heart, and lung transplantation
procedures benefit tens of thousands of patients worldwide each year.
However, long-term graft survival is often hindered by transplant
rejection, among which T cell-mediated rejection (TCMR) remains one of
the major challenges ([38]2–[39]4). TCMR arises from the activation of
T cells through immunological synapses formed with antigen-presenting
cells (APCs), such as dendritic cells and inflammatory myeloid
cells. These interactions, mediated by the recognition of antigens by
the T cell receptor (TCR), lead to interferon-gamma (IFNG)-driven
inflammatory signaling pathways and the expression of
inflammation-related genes, including IL2, IFNG, and ADAMDEC1.
Together, these immune processes result in severe inflammation and
graft dysfunction, highlighting the critical roles of T cells,
dendritic cells, and macrophages in TCMR pathogenesis ([40]5, [41]6).
Although immunosuppressive therapies, such as cyclosporine, tacrolimus,
and mycophenolate mofetil, have demonstrated efficacy in reducing
short-term rejection episodes ([42]7), their long-term use is
associated with substantial adverse effects, including infections,
malignancies, hypertension, diabetes, and nephrotoxicity
([43]8–[44]12). These complications significantly impair patient
quality of life and limit the applicability of current therapeutic
regimens. Moreover, the diagnosis of TCMR relies heavily on invasive
graft biopsies, which are prone to sampling errors, procedural
complications, and subjective variations in histopathological
evaluations ([45]13, [46]14). These challenges underscore the critical
need for innovative diagnostic tools and safer, more effective
immunomodulatory interventions to address TCMR.
Cytokines, as key regulators of immune responses, represent promising
candidates for improving both the diagnosis and treatment of TCMR.
Cytokine-related genes play central roles in mediating immune
dysregulation during TCMR, making them valuable targets for biomarker
discovery and therapeutic development. To expand their translational
potential, a broader understanding of cytokine dysregulation across
different transplant types is necessary. A multi-organ perspective may
uncover shared pathways underlying TCMR, facilitating the
identification of universal targets for intervention.
Among potential therapeutic targets, the nuclear receptor peroxisome
proliferator-activated receptor gamma (PPARγ) has received attention
for its immunomodulatory and anti-inflammatory properties. Originally
recognized for its role in regulating lipid and glucose metabolism in
diseases such as type 2 diabetes and dyslipidemia ([47]15, [48]16),
PPARγ has now been implicated in modulating immune responses. PPARγ
agonists, such as rosiglitazone, have been shown to attenuate T cell
activation, alter cytokine production, and regulate macrophage
polarization, thereby suppressing excessive inflammatory responses
([49]17). These properties position PPARγ agonists as promising
candidates for addressing immune dysregulation in TCMR and reducing
graft rejection. However, their potential therapeutic value has yet to
be fully explored in the context of organ transplantation.
In this study, we addressed the limitations of traditional diagnostic
and therapeutic approaches to TCMR by identifying a novel
TCMR-associated cytokine gene set (TCMR-Cs) through the integration of
transcriptomic data from kidney, heart, and lung transplant rejection
biopsies. This gene set provided valuable insights into the mechanisms
underlying TCMR and served as the basis for both diagnostic and
therapeutic innovations. A predictive model constructed using the
TCMR-Cs demonstrated its utility in differentiating TCMR from non-TCMR
patients, while drug repurposing analyses identified PPARγ agonists as
promising immunomodulatory agents for TCMR prevention. Experimental
validation further confirmed the efficacy of the PPARγ agonist
rosiglitazone in modulating immune responses, including its ability to
suppress T cell activation and prolong graft survival in combination
with rapamycin. These findings lay a foundation for future precision
medicine strategies in transplantation and highlight the potential of
cytokine-based interventions to overcome the challenges of TCMR.
2. Materials and methods
2.1. Data acquisition and preprocessing
Gene expression microarray datasets for kidney ([50]GSE192444)
([51]18), heart ([52]GSE150059) ([53]19), and lung ([54]GSE150156)
([55]20) transplants were retrieved from the GEO database. All datasets
were log2-transformed and normalized using the “limma” package in R to
ensure consistency across datasets. The biopsy diagnosis for each
sample was defined based on the original annotations provided in the
corresponding dataset. In order to ensure a focused and accurate
analysis, samples were regrouped into two categories: those with a
biopsy diagnosis of “TCMR” were classified into the “TCMR” group, while
those diagnosed with other conditions, such as “ABMR” or “No
Rejection,” were classified into the “non-TCMR” group. To minimize
potential bias and improve the specificity of the analysis, samples
labeled as “Mixed Rejection,” “Possible TCMR,” “Borderline,” or
“Possible ABMR” were excluded, as the inclusion of such heterogeneous
or uncertain diagnoses could confound the identification of molecular
features specific to TCMR ([56]5).
2.2. DEG identification and functional enrichment analysis
DEGs were identified using the “limma” package in R. Linear models
accompanied by empirical Bayes moderation were applied to stabilize
variance estimates. DEGs were defined by a log-fold change (logFC)
threshold of ±1 and a p-value < 0.05. Significant DEGs were then
subjected to functional annotation using the “clusterProfiler” package.
Gene Ontology (GO) analysis was performed to explore biological
processes (BP), molecular functions (MF), and cellular components (CC)
associated with these genes. Kyoto Encyclopedia of Genes and Genomes
(KEGG) pathway analysis identified relevant biological pathways
implicated by the DEGs. Both GO and KEGG analyses used the
“org.Hs.eg.db” as the reference database. Visualization of GO and KEGG
results was performed using bar plots or bubble charts.
2.3. Data visualization
Principal component analysis (PCA) was conducted using the “FactoMineR”
and “factoextra” packages to visualize variance within datasets and
achieve a clear separation between TCMR and non-TCMR groups. Heatmaps
and volcano plots of DEGs were generated using the “EnhancedVolcano”
and “pheatmap” packages, displaying differential expression patterns
visually. To identify genes shared across kidney, heart, and lung
datasets, Venn diagrams were generated using the “VennDiagram” R
package. Additional visualizations were generated using the “ggplot2”
package.
2.4. Definition of the TCMR-Cs
Upregulated DEGs shared across kidney, heart, and lung TCMR datasets
were identified through intersection analysis. These shared DEGs were
further cross-referenced with a cytokine-related gene set curated from
the literature ([57]21), resulting in 11 genes defined as the TCMR-Cs.
Gene interaction networks among the TCMR-Cs were analyzed using
GeneMANIA ([58]http://genemania.org/) ([59]22). This analysis included
co-expression, physical interactions, and shared pathways to support
the inclusion of these genes as a cohesive set.
2.5. Predictive model construction and validation
To validate the significance of the identified TCMR-Cs, a predictive
model was constructed. Normalized gene expression data from the kidney
([60]GSE192444), heart ([61]GSE150059), and lung ([62]GSE150156)
datasets were combined to form the training cohort. Lasso regression
for feature selection was implemented using the “glmnet” R package to
address high-dimensional data. Cross-validation was utilized to
determine the optimal lambda parameter, and lambda.1se was selected to
balance model complexity and performance. Five hub genes (CXCL13, IFNG,
TNFSF13B, CCL3, CCL18) were identified through this process. A
multivariate logistic regression model was then constructed using the
selected genes. Risk scores were calculated for each sample based on
the following formula:
[MATH: Risk Score=∑
i=1n(coefi×expri) :MATH]
where
[MATH:
coefi :MATH]
represents the regression coefficient for gene
[MATH: i :MATH]
, and
[MATH:
expri :MATH]
indicates the expression level of gene
[MATH: i :MATH]
. The model’s performance was assessed through receiver operating
characteristic (ROC) curve analysis. The risk score threshold was
determined using the Youden index. Independent validation datasets from
kidney ([63]GSE98320) ([64]23), heart ([65]GSE124897) ([66]24), and
lung ([67]GSE125004) ([68]25) transplantation were used to evaluate the
generalizability of the model. ROC curves and area under the curve
(AUC) values were calculated for each test cohort. Given the class
imbalance present in the datasets, additional performance metrics such
as precision, recall, and F1 score were implemented to ensure a more
thorough evaluation of the model’s predictive performance. These
metrics were computed for both the training and independent test
datasets to robustly validate the model.
2.6. Immune cell infiltration analysis
The CIBERSORT algorithm ([69]26), utilizing the LM22 signature matrix,
was employed to estimate the proportions of 22 immune cell types,
providing insights into the immune landscape of the datasets.
Differences in immune cell infiltration between TCMR and non-TCMR
groups were visualized using box plots and heatmaps. Statistical
significance was determined using Kruskal-Wallis tests, while
correlation analysis was performed using the “Hmisc” package in R.
2.7. Drug screening
CMap ([70]https://clue.io/) database ([71]27) was used to identify
small molecules with potential therapeutic effects based on
connectivity scores that reflect the ability to reverse TCMR-associated
gene expression. Among the top candidates, PPARγ receptor agonists were
highlighted, and rosiglitazone was selected for further experimental
validation.
2.8. Mouse skin transplantation model
Donor mice were 8-week-old male BALB/c mice, while recipient mice were
8–12-week-old male C57BL/6 mice. Donor mice were euthanized, and their
tail skin was thoroughly disinfected with povidone-iodine. Tail skin
was excised and trimmed into 1.0 × 1.0 cm² grafts. Recipient mice were
anesthetized with isoflurane, and the hair on their dorsal region was
shaved and disinfected with povidone-iodine. Once the disinfectant
dried, a 1.0 × 1.0 cm² graft bed was prepared by excising the skin. The
donor skin graft was placed onto the graft bed and secured at the four
corners and edges using sutures. After surgery, the graft was protected
with a sterile bandage, and the mice were housed in clean cages with
heating pads to maintain body temperature until they fully recovered.
The mice were divided into four groups: control group, rosiglitazone
group, rapamycin group, and combination group (rosiglitazone +
rapamycin). Starting on the day of transplantation, the mice received
daily intraperitoneal injections of the respective treatments. The
rapamycin group received 0.5 mg/kg/day, the rosiglitazone group
received 10 mg/kg/day, and the combination group received both drugs at
the same dosages. The control group received an equivalent volume of
the solvent. Bandages were removed on postoperative day 7. Graft
survival was assessed daily through macroscopic observation, and
photographs were taken for analysis. Graft rejection was defined as
≥90% necrosis of the graft area.
2.9. Mouse CD4^+ T cell culture
CD4^+ T cells were activated in 96-well plates pre-coated with 5 μg/mL
anti-mouse CD3ϵ antibody (BioLegend, Cat# 100340) in PBS, incubated
overnight at 4°C. CD4^+ T cells were isolated from the spleens of
6–8-week-old C57BL/6 mice using the CD4^+ T Cell Isolation Kit
(StemCell, Cat# 19765) following the manufacturer’s protocol. Spleens
were homogenized through a 70 μm cell strainer to obtain single-cell
suspensions, followed by red blood cell lysis using 0.84% ammonium
chloride. Purified CD4^+ T cells were resuspended in RPMI-1640 medium
supplemented with 10% fetal bovine serum (FBS), 1%
penicillin-streptomycin, 50 μM β-mercaptoethanol, 2 μg/mL anti-CD28
antibody (BioLegend, Cat# 102116), and 50 U/mL recombinant mouse IL-2
(BioLegend, Cat# 575402). The cells were seeded in the anti-CD3ϵ-coated
plates and cultured at 37°C in a humidified atmosphere with 5% CO[2].
During the culture period, rosiglitazone was added to the medium at
concentrations of 10 μM or 30 μM, or an equivalent volume of solvent
was added as a control. After 24 hours of incubation, cells were
harvested for flow cytometric analysis.
For Th1 differentiation, CD4^+ T cells were cultured in RPMI-1640
medium supplemented with 10% fetal bovine serum, 1%
penicillin-streptomycin, 50 μM β-mercaptoethanol, recombinant murine
IL-12p70 (10 ng/mL, Propetech, Cat# 210-12-10UG), and neutralizing IL-4
antibody (10 μg/mL, BioLegend, Cat# 504122) under Th1-inducing
conditions. The cells were seeded at a density of 1 × 10^6 cells/mL in
96-well flat-bottom plates and cultured for 72 hours at 37°C in a
humidified atmosphere with 5% CO[2].
2.10. Flow cytometry analysis
To evaluate cell viability and death, CD4^+ T cells treated with
rosiglitazone (10 μM or 30 μM) or solvent control were first stained
using the live-dead Zombie Aqua (BioLegend, Cat# 423102) according to
the manufacturer’s instructions. Stained cells were then washed and
resuspended in staining buffer (PBS containing 1% FBS and 0.1% sodium
azide) for surface marker staining. For surface marker staining, cells
were incubated with FITC-conjugated anti-mouse CD4 (ThermoFisher, Cat#
25-7021-82), APC-conjugated anti-mouse CD69 (BioLegend, Cat# 104513),
and PE-conjugated anti-mouse CD25 (BioLegend, Cat# 102007) antibodies
at 4°C in the dark for 30 minutes. After staining, cells were washed
twice with cold staining buffer and resuspended in 300 μL PBS for
analysis. For intracellular cytokine staining, cells were stimulated in
vitro with 50 ng/mL PMA, 1 μg/mL ionomycin, and 10 μg/mL brefeldin A
for 5 hours in a humidified 37°C incubator. Following stimulation,
cells were fixed and permeabilized using Fixation Buffer (BioLegend,
Cat# 420801) according to the manufacturer’s instructions.
Intracellular IL-2 staining was performed using PE-Cy7-conjugated
anti-mouse IL-2 antibody (ThermoFisher, Cat# 25-7021-82) at 4°C in the
dark for 30 minutes.
For Th1 analysis, cells were collected following culture under
Th1-inducing conditions, washed twice with cold PBS, and stained for
CD4^+ surface markers using FITC-conjugated anti-mouse CD4 antibody at
4°C in the dark for 30 minutes. After surface staining, intracellular
cytokine staining was performed using the intracellular fixation and
permeabilization kit (BioLegend, Cat# 420801). Subsequently, cells were
stained for IFN-γ using anti-mouse IFN-γ-eFluor450 (ThermoFisher, Cat#
48-7311-82) at 4°C in the dark for 30 minutes. After washing, cells
were resuspended in PBS with 1% FBS and analyzed using flow cytometry.
Flow cytometric analysis was conducted using a flow cytometer, and data
were processed with FlowJo software (Version 10.8.1). The gating
strategy was consistent across all samples. Debris and dead cells were
excluded based on FSC/SSC properties during gating, and CD4^+ T cells
were selected for further analysis.
2.11. Statistical analysis
All statistical analyses were performed using R Software (version
4.3.0) and GraphPad Prism (version 9.5.1). Group comparisons were
conducted using a two-tailed unpaired Student’s t-test for two groups
or one-way analysis of variance (ANOVA) for comparisons among multiple
groups. Graft survival in the skin transplantation experiments was
analyzed using Kaplan-Meier survival curves and the log-rank test. The
diagnostic performance of models was evaluated using ROC curves and
AUC, with optimal cutoff points determined by the Youden index.
Statistical significance was defined as a two-tailed p < 0.05.
3. Results
3.1. Identification and functional annotation of DEGs in TCMR patients across
different organ transplants
To investigate the molecular mechanisms underlying TCMR across
different transplanted organs, microarray data from kidney, heart, and
lung transplant biopsy samples were retrieved from the GEO database and
analyzed.
In the kidney transplant cohort ([72]GSE192444), which included 21 TCMR
samples and 242 non-TCMR samples ([73] Figure 1A ), a total of 454
upregulated and 158 downregulated DEGs were identified ([74]
Figures 1B, C ). Similarly, the heart transplant cohort
([75]GSE150059), consisting of 76 TCMR samples and 1032 non-TCMR
samples ([76] Figure 1D ), yielded 587 upregulated and 34 downregulated
DEGs ([77] Figures 1E, F ). In the lung transplant cohort
([78]GSE150156), including 23 TCMR samples and 291 non-TCMR samples
([79] Figure 1G ), 439 upregulated and 351 downregulated DEGs were
identified ([80] Figures 1H, I ). These findings highlight the
significant transcriptomic changes associated with TCMR in each organ
system.
Figure 1.
[81]Figure 1
[82]Open in a new tab
Differential gene expression between TCMR and non-TCMR groups across
different transplant cohorts. PCA plots illustrate the separation of
TCMR and non-TCMR samples in three transplant cohorts: renal transplant
([83]GSE192444) (A), heart transplant ([84]GSE150059) (D), and lung
transplant ([85]GSE150156) (G). Each point represents an individual
sample, with colors distinguishing the TCMR group from the non-TCMR
group. Heatmaps (B, E, H) display the expression patterns of DEGs
between TCMR and non-TCMR groups in the corresponding cohorts
([86]GSE192444, [87]GSE150059, and [88]GSE150156). Rows correspond to
DEGs, and columns represent individual samples. Volcano plots (C, F, I)
present the results of DEG analysis between TCMR and non-TCMR groups
for each cohort. Red dots indicate upregulated genes in the TCMR group,
while blue dots indicate downregulated genes. The x-axis represents
log2 fold change, and the y-axis represents -log10 p-value. Labeled
genes denote the top 10 upregulated or downregulated genes with the
highest absolute log2 fold changes.
Functional annotation of the upregulated DEGs in each cohort using GO
and KEGG pathway analyses revealed significant enrichment in
immune-related biological processes and signaling pathways ([89]
Supplementary Figures S1A–F ). Despite certain organ-specific
differences in DEG profiles, key enriched pathways—such as “immune
response-activating signaling pathway” and “immune receptor
activity”—were consistently observed across kidney, heart, and lung
transplants. This consistency highlights the shared immune processes
underlying TCMR, regardless of the type of transplanted organ, further
supporting the notion that TCMR is driven by conserved molecular
mechanisms.
3.2. Shared molecular features and cytokine network analysis in TCMR patients
To identify conserved molecular signatures across different
transplanted organ types, we conducted an intersection analysis of the
upregulated DEGs shared across kidney, heart, and lung TCMR cohorts.
This approach identified 190 commonly upregulated DEGs (CU-DEGs) shared
across the three organ types ([90] Figure 2A ). GO enrichment analysis
of the CU-DEGs revealed significant associations with immune-related
functional categories, including “positive regulation of cytokine
production,” “immune response-activating signaling pathways,”
“leukocyte activation involved in immune response,” “regulation of T
cell activation,” and “leukocyte cell-cell adhesion” ([91] Figure 2B ).
These results suggest that the shared DEGs play essential roles in
driving immune activation and cell-cell interactions during TCMR.
Further analysis using KEGG pathway enrichment demonstrated that
CU-DEGs were prominently involved in pathways directly linked to
transplant rejection, such as “allograft rejection,” “phagosome,”
“graft-versus-host disease,” “hematopoietic cell lineage,” and
“cytokine–cytokine receptor interaction” ([92] Figure 2C ). These
enriched pathways highlight the interplay between innate and adaptive
immune responses in TCMR pathogenesis and reinforce the critical role
of a cytokine-driven inflammatory cascade in mediating transplant
rejection.
Figure 2.
[93]Figure 2
[94]Open in a new tab
Shared gene characteristics and cytokine network analysis in TCMR
patients. (A) A Venn diagram showing the commonly upregulated
differentially expressed genes (CU-DEGs) among TCMR patients in renal
transplant ([95]GSE192444), heart transplant ([96]GSE150059), and lung
transplant ([97]GSE150156) cohorts, identifying a total of 190 shared
genes. (B) Bubble plot displaying the GO functional enrichment analysis
of CU-DEGs. (C) Bar chart illustrating the KEGG pathway enrichment
analysis of CU-DEGs. (D) A Venn diagram depicting the intersection
between CU-DEGs and cytokine-related genes, identifying 11
TCMR-associated cytokine genes (TCMR-Cs). (E) PPI network of the
TCMR-Cs constructed using the GeneMANIA database. The network reveals
both direct and indirect interactions among the 11 key genes, along
with additional functionally related genes.
By intersecting the identified CU-DEGs with a curated cytokine-related
gene set, we further narrowed down a set of 11 TCMR-associated cytokine
genes (TCMR-Cs): CXCL13, IFNG, TNFSF13B, CCL3, CXCL11, CXCL9, CXCL10,
CCL8, CCL18, LTB, and CCL19 ([98] Figure 2D ). These cytokine and
chemokine genes are well-established as key regulators of immune
responses and leukocyte recruitment, underscoring their importance in
mediating immune dysregulation during TCMR. To explore the functional
relationships among these TCMR-Cs, we constructed a protein-protein
interaction (PPI) network using the GeneMANIA database. The resulting
network revealed extensive direct and indirect interactions both within
the cytokine gene set and with additional functional partners ([99]
Figure 2E ). This PPI network illustrates the intricate and
interconnected roles of these cytokines and chemokines in TCMR,
emphasizing their collective contribution to immune activation,
leukocyte recruitment, and inflammation. Collectively, these findings
define a cytokine-driven axis of immune dysregulation in TCMR, which
may serve as the basis for developing therapeutic strategies aimed at
modulating cytokine activity during transplant rejection.
3.3. Diagnostic predictive model for TCMR using lasso and logistic regression
To further explore the biological significance and diagnostic potential
of the previously identified 11 TCMR-Cs, we used their expression
profiles to construct a predictive model for TCMR. This approach aimed
to evaluate whether these cytokine genes collectively could reliably
distinguish TCMR patients from non-TCMR patients across multiple organ
transplant types.
The 11 TCMR-Cs were used as input features for modeling, and
transcriptomic data from the kidney ([100]GSE192444), heart
([101]GSE150059), and lung ([102]GSE150156) transplant cohorts were
integrated to form a comprehensive multilayer dataset. Patients were
categorized into TCMR (120 samples) and non-TCMR (1565 samples) groups
based on clinical outcomes. To select the most informative genes and
construct the optimal model, Lasso regression analysis was performed
([103] Figure 3A ). After tuning the parameters using the lambda.1se
selection criterion, five genes—CXCL13, IFNG, TNFSF13B, CCL3, and
CCL18—were identified as the key predictors for the final model ([104]
Figure 3B ). These five genes are referred to as TCMR Core Hub Genes
(TCMR-Hubs).
Figure 3.
[105]Figure 3
[106]Open in a new tab
Diagnostic Model for TCMR Using Lasso and Logistic Regression. (A)
Cross-validation plot for the Lasso regression, showing the trend of
mean squared error (MSE) with the logarithmically transformed lambda
parameter during cross-validation. (B) Coefficient path plot of the
Lasso regression, illustrating the shrinkage of candidate gene
coefficients toward zero as the regularization parameter lambda
increases. At the selected lambda value (lambda.1se), five key genes
were retained for constructing the prediction model. (C) Logistic
regression coefficients of the TCMR-Hubs. (D) Forest plot summarizing
the odds ratios (OR) and 95% confidence intervals (CIs) of the five key
genes identified by logistic regression. (E) Box plot depicting the
risk scores for each sample in the training cohort, calculated based on
the regression coefficients of the five key genes. Significant
differences in risk scores are shown between the TCMR and non-TCMR
groups. (F) ROC curve of the risk scores, with the red dot indicating
the optimal threshold determined by the Youden index for distinguishing
TCMR and non-TCMR groups. ***P< 0.001.
To examine the specific contributions of these TCMR-Hubs to the
predictive model, multivariate logistic regression was conducted,
revealing their individual regression coefficients and highlighting
their predictive importance. The coefficients for the selected genes
were as follows: CXCL13 (0.83), IFNG (1.33), TNFSF13B (1.51), CCL3
(-1.19), and CCL18 (0.36) ([107] Figure 3C ). The corresponding odds
ratios (ORs) and 95% confidence intervals (CIs) further validated their
diagnostic relevance ([108] Figure 3D ). Using these coefficients, risk
scores for all patients in the training cohort were calculated,
providing a quantitative basis for stratifying patients by TCMR risk.
Comparative analysis of risk scores between TCMR and non-TCMR patients
showed a pronounced separation, with TCMR patients exhibiting
significantly higher scores ([109] Figure 3E ). ROC curve evaluation
confirmed the predictive performance of the model, achieving an
exceptional AUC of 0.99 in the training cohort ([110] Figure 3F ).
Using the Youden index, the optimal risk score threshold for
classification was determined to be 17.64, effectively distinguishing
high-risk patients from low-risk individuals. These results indicate
the high diagnostic accuracy of the model and emphasize the biological
and predictive importance of the five TCMR-Hubs in capturing immune
features associated with TCMR.
3.4. Validation of the TCMR predictive model
To validate the robustness and generalizability of the predictive
model, we first evaluated its performance within the training cohort.
The distinct expression profiles of the five TCMR-Hubs between TCMR and
non-TCMR samples provided a solid basis for accurate classification
([111] Figure 4A ). At an optimal cutoff of 0.03, the model
demonstrated excellent sensitivity (0.99) and specificity (0.91) in the
training cohort, with an outstanding AUC of 0.99 ([112] Figure 4B ).
Moreover, additional metrics were utilized to provide a more nuanced
assessment of the model’s diagnostic performance in the training data.
Specifically, the precision, recall, and F1 score of the model were
0.999, 0.909, and 0.952 ([113]Figure 4C), respectively, indicating a
strong balance between minimizing false positives and maximizing
detection of TCMR cases. These results reinforced the ability of the
model to reliably identify TCMR within the development dataset.
Figure 4.
[114]Figure 4
[115]Open in a new tab
Evaluation of the predictive performance of the TCMR diagnostic model.
Heatmaps display the expression patterns of TCMR-Hubs in the TCMR and
non-TCMR groups across the training cohort (A), renal transplant
testing cohort ([116]GSE98320, (D)), heart transplant testing cohort
([117]GSE124897, (G)), and lung transplant testing cohort
([118]GSE125004, (J)). Receiver operating characteristic (ROC) curves
for the prediction model are shown for the training cohort (B), renal
transplant testing cohort ([119]GSE98320, (E)), heart transplant
testing cohort ([120]GSE124897, (H)), and lung transplant testing
cohort ([121]GSE125004, (K)). Each ROC curve includes the AUC, optimal
cutoff point, specificity, and sensitivity. Performance metrics,
including precision, recall, and F1 score, are shown for the training
cohort (C), renal transplant testing cohort (F), heart transplant
testing cohort (I), and lung transplant testing cohort (L).
The predictive capability of the model was then validated across
independent test cohorts, representing kidney ([122]GSE98320), heart
([123]GSE124897), and lung ([124]GSE125004) transplants, to confirm its
applicability across diverse transplant types. Additional performance
metrics were calculated for these test datasets to ensure comprehensive
evaluation. In the kidney test cohort, the model displayed strong
diagnostic performance, achieving an AUC of 0.87 ([125] Figures 4D, E
), with precision, recall, and F1 score of 0.979, 0.839, and 0.904
([126] Figure 4F ), respectively. Similarly, in the heart test cohort,
the model maintained exceptional accuracy with an AUC of 0.98 ([127]
Figures 4G, H ), combined with a precision of 1.000, recall of 0.897,
and an F1 score of 0.946 ([128] Figure 4I ). Lastly, the lung
transplant test cohort achieved high predictive power with an AUC of
0.95 ([129] Figures 4J, K ), alongside precision, recall, and F1 score
values of 0.981, 0.963, and 0.972 ([130] Figure 4L ), respectively.
These additional metrics not only confirmed the strong diagnostic
capabilities of the model but also highlighted its robustness in
handling imbalanced datasets, as evidenced by consistently high
precision and balanced recall.
Collectively, these findings indicate that the TCMR predictive model
based on the TCMR-Cs gene set can robustly and accurately assess TCMR
risk across various organ transplant types. By integrating
transcriptomic data and focusing on a core set of cytokine genes, this
model demonstrates potential clinical utility for the early detection
of TCMR.
3.5. Immune infiltration characteristics in TCMR patients
To investigate the immune microenvironment of TCMR, immune cell
infiltration levels were analyzed in TCMR and non-TCMR patients across
kidney, heart, and lung transplant cohorts. Despite organ-specific
differences, the TCMR groups consistently demonstrated elevated
infiltration of several immune cell subsets, including CD8^+ T cells,
activated NK cells, M1 macrophages, M2 macrophages, γδ T cells,
follicular helper T cells (Tfh cells), naive B cells, and plasma cells
([131] Figures 5A, B ; [132]Supplementary Figures S2A–D ).
Figure 5.
[133]Figure 5
[134]Open in a new tab
Immune infiltration characteristics of TCMR patients. (A) Heatmap
showing the infiltration levels of various immune cells in the TCMR and
non-TCMR groups within the renal transplant cohort ([135]GSE192444).
(B) Box plot ranking the abundance of immune cell infiltration in the
TCMR group within the renal transplant cohort ([136]GSE192444). (C) Box
plot comparing the infiltration levels of immune cells between the TCMR
and non-TCMR groups within the renal transplant cohort
([137]GSE192444). (D) Correlation matrix illustrating the relationships
among different immune cells in the TCMR group. Blue indicates negative
correlations, while red indicates positive correlations, with deeper
colors representing stronger correlations. (E) Correlation analysis
between TCMR-Hubs and immune cell infiltration levels in the TCMR
group. Blue indicates negative correlations, while red indicates
positive correlations, with deeper colors representing stronger
correlations. *P < 0.05, **P < 0.01, ***P < 0.001. *p<0.05, **p<0.01,
***p<0.001, ****p<0.0001 and non-significant values (ns) indicate
p≥0.05.
Organ-specific differences in immune infiltration were observed when
comparing TCMR patients to non-TCMR patients. In the kidney transplant
cohort, infiltration of M1 macrophages, γδ T cells, plasma cells, and
Tfh cells was significantly increased ([138] Figure 5C ). Among heart
transplant TCMR patients, higher infiltration levels of CD8^+ T cells,
M1 macrophages, γδ T cells, and Tfh cells were observed ([139]
Supplementary Figure S2E ). Similarly, lung transplant TCMR patients
showed elevated infiltration of M1 macrophages, activated NK cells, γδ
T cells, and activated memory CD4^+ T cells ([140] Supplementary Figure
S2F ). These observations highlight both shared and organ-specific
immune mechanisms in TCMR. In contrast, several immune subsets were
consistently reduced across all three organ types in the TCMR group,
including regulatory T cells (Tregs) and resting mast cells. Moreover,
M2 macrophages displayed lower infiltration levels in both kidney and
heart TCMR cohorts ([141] Figure 5C ; [142]Supplementary Figures S2E, F
). This indicates a shift in the immune environment of TCMR patients
towards a pro-inflammatory state, characterized by an imbalance between
immune activation and regulation.
Correlational analysis between immune cell subsets revealed strong
positive associations between certain immune populations, such as M1
macrophages and activated NK cells ([143] Figure 5D ;
[144]Supplementary Figures S3A, C ). Furthermore, the relationship
between the five TCMR-Hubs and immune cell infiltration was examined.
Positive correlations were identified between TCMR-Hubs and
immune-activated subsets, including γδ T cells, activated memory CD4^+
T cells, plasma cells, activated NK cells, M1 macrophages, and Tfh
cells. Conversely, TCMR-Hubs were negatively correlated with
immune-regulatory or resting subsets, such as Tregs, resting NK cells,
resting mast cells, naive B cells, and resting memory CD4^+ T cells
([145] Figure 5E ; [146]Supplementary Figures S3B, D ).
These findings emphasize the critical role of specific immune subsets,
particularly pro-inflammatory and effector cells, in the progression of
TCMR. The observed negative correlations with regulatory immune
populations further underscore the imbalance within the immune
microenvironment in TCMR. These results suggest that modulating key
immune subsets may provide a therapeutic strategy for mitigating TCMR.
3.6. Immunomodulatory effects of the PPARγ agonist rosiglitazone in TCMR
To identify potential therapeutic agents for TCMR, we focused on
utilizing the TCMR-Cs to explore targeted treatment strategies. Using
the 11 TCMR-Cs and the top 10 upregulated DEGs in TCMR patients, a
drug-repurposing analysis was conducted through the CMap database. The
analysis yielded a ranked list of small-molecule compounds with
potential relevance to TCMR, based on their ability to reverse the
TCMR-associated transcriptional signature. From the top 100 compounds
showing negative CMap scores, the most frequent mechanisms of action
included “PPAR receptor agonist,” “serotonin receptor antagonist,”
“acetylcholine receptor antagonist,” “carbonic anhydrase inhibitor,”
and “HDAC inhibitor” ([147] Supplementary Figure S4 ). Among these,
PPAR receptor agonists stood out due to their established
anti-inflammatory and immunomodulatory properties. To validate their
therapeutic potential, rosiglitazone, a PPARγ agonist, was selected for
in vitro and in vivo functional experiments.
To evaluate its immunomodulatory capacity, CD4^+T cells were isolated
from mouse spleens, activated with anti-CD3, anti-CD28, and IL-2, and
treated in vitro with rosiglitazone at concentrations of 10 μM and 30
μM. After 24 hours of treatment, flow cytometry analysis revealed that
rosiglitazone significantly inhibited CD4^+T cell activation, as
evidenced by reduced expression of activation markers CD69 and CD25
([148] Figures 6A, B ), and suppressed functional activity by
significantly reducing IL-2 production ([149] Figure 6C ). Moreover, at
both concentrations, no significant differences in cell viability and
death were observed between the rosiglitazone-treated and control
groups ([150] Supplementary Figure S5A ), indicating that rosiglitazone
effectively modulates immune responses without inducing cytotoxicity
under the tested conditions. Additionally, the effect of rosiglitazone
on Th1 differentiation was investigated. CD4^+T cells were isolated and
cultured in Th1-inducing conditions (recombinant IL-12 and anti-IL-4
antibody) for 72 hours in the presence or absence of rosiglitazone.
Flow cytometry analysis revealed that rosiglitazone treatment
significantly reduced the proportion of Th1 cells, as indicated by a
decreased percentage of CD4^+IFN-γ^+cells compared to the untreated
Th1-differentiated control ([151] Supplementary Figure S5B ). These
results suggest that rosiglitazone not only inhibits T cell activation
but also suppresses the differentiation of pro-inflammatory Th1 cells.
Figure 6.
[152]Figure 6
[153]Open in a new tab
Immunoregulatory effects of the PPARγ agonist rosiglitazone in TCMR.
CD4+ T cells were isolated from mouse spleens, activated in vitro, and
treated with 10 μM or 30 μM rosiglitazone (Rosi) or an equivalent
volume of solvent control for 24 hours. Flow cytometry was used to
measure the mean fluorescence intensity (MFI) of CD69 (A) and CD25 (B)
in each group. (C) Proportion of IL-2-positive CD4+ T cells measured by
flow cytometry after 24 hours of culture. (D) Representative
photographs of skin grafts from four groups of mice: control group,
rosiglitazone group, rapamycin group (Rapa), and rapamycin combined
with rosiglitazone group. (E) Survival curves of skin grafts from the
four experimental groups. *p<0.05, **p<0.01, ***p<0.001 and ns indicate
p≥0.05.
The therapeutic potential of rosiglitazone was further assessed in vivo
using a murine skin transplant rejection model. BALB/c mouse skin was
transplanted onto the backs of C57BL/6 mice, and the animals were
divided into four groups: control, rosiglitazone (10 mg/kg/day),
rapamycin (50 mg/kg/day), and a combination of rosiglitazone and
rapamycin ([154] Figure 6D ). In the rosiglitazone-only group,
monotherapy did not significantly prolong graft survival compared to
the control group. In contrast, rapamycin-treated mice exhibited
extended graft survival. Importantly, the combination therapy of
rapamycin and rosiglitazone demonstrated a synergistic effect,
significantly prolonging graft survival compared to rapamycin alone
([155] Figure 6E ).
These results highlight the immunomodulatory potential of PPARγ
agonists in TCMR. Rosiglitazone effectively suppressed T cell
activation in vitro, reduced Th1 differentiation, and demonstrated no
cytotoxic effects on T cells under the tested conditions, supporting
its safety and efficacy as an immunomodulatory agent. In vivo,
rosiglitazone enhanced graft survival when combined with rapamycin, as
evidenced by extended survival times, suggesting a synergistic
mechanism. These findings provide robust evidence supporting the
repurposing of PPARγ agonists as adjunctive therapies for TCMR.
Furthermore, they underscore the importance of combining targeted
immunomodulators with existing immunosuppressive regimens to optimize
outcomes in transplantation medicine.
4. Discussion
This study defines a novel TCMR-associated cytokine gene set and
highlights its potential in both understanding the pathogenesis of TCMR
and facilitating drug repurposing efforts to address this significant
clinical challenge in organ transplantation. By integrating multi-organ
transcriptomic data, we established an organ-agnostic approach to
investigate key immune mechanisms underlying TCMR and proposed
innovative therapeutic strategies based on these findings. Early
detection of TCMR and the identification of targeted interventions are
critical for improving long-term transplant outcomes ([156]28),
especially since current immunosuppressive therapies often result in
severe complications such as infections, organ toxicity, and malignancy
with prolonged use ([157]7, [158]29). Responding to the urgent demand
for safer and more sustainable treatments, this study offers a new
framework for identifying actionable biomarkers, constructing
predictive models, and exploring drug repurposing opportunities in
transplantation medicine.
By focusing on the shared immune characteristics of TCMR across kidney,
heart, and lung transplantation, we identified a cytokine-driven gene
signature with significant translational potential. The intersection of
TCMR-related genes with cytokine-associated gene sets yielded a concise
and relevant gene signature (TCMR-Cs) that provides a foundation for
understanding common immune mechanisms and identifying therapeutic
targets. Functional annotation of these genes, through GO and KEGG
pathway enrichment analyses, revealed their central roles in immune
activation pathways, including “Regulation of T cell activation,”
“Leukocyte cell-cell adhesion,” “Allograft rejection,” and
“Graft-versus-host disease.” Notably, these pathways are essential
mediators of rejection and underpin many of the immune dysregulation
processes involved in TCMR. This organ-agnostic approach reinforces the
biological relevance of TCMR-Cs and opens the possibility of targeting
shared molecular processes across multiple organ transplant types,
thereby advancing the development of generalized therapeutic
strategies.
Key cytokines within our gene set emerged as significant contributors
to TCMR pathogenesis, emphasizing their potential as biomarkers and
therapeutic targets. CXCL13, for instance, promotes B cell infiltration
in kidney transplantation, increasing the severity of rejection, and
its elevated levels in serum have been proposed as both a biomarker and
therapeutic target for TCMR ([159]30). Similarly, IFNG plays a pivotal
role in modulating endothelial cell activity and antigen presentation
pathways, exacerbating rejection ([160]31). These mechanistic insights
into cytokine-mediated immune regulation further validate their
potential as molecular targets for therapeutic intervention in TCMR.
Other cytokines, such as CCL3, TNFSF13B, and CCL18, are equally
important in TCMR. Elevated levels of CCL3 have been linked to immune
cell recruitment and activation in both antibody-mediated rejection
(ABMR) and TCMR, indicating its dual relevance ([161]32). Likewise,
TNFSF13B, which encodes the cytokine BAFF, plays a regulatory role in
B-cell activity and rejection-related immune networks ([162]33,
[163]34). Although its direct role in TCMR requires further study, its
centrality in autoimmune and rejection pathways highlights its
therapeutic potential. Finally, CCL18 has been implicated in
accelerating graft rejection by recruiting alloreactive T cells, as
shown in a humanized skin transplant model ([164]35). Taken together,
these findings establish a shared cytokine-mediated axis of immune
dysregulation that is pivotal to TCMR across multiple organ systems,
further validating the importance of our identified gene set as both a
biological and therapeutic resource.
Our predictive model, developed using Lasso regression and multivariate
logistic regression analysis, demonstrated robust performance,
achieving high AUC values in both the training and testing cohorts.
These results not only validate the clinical utility of the TCMR-Cs as
a robust biomarker but also emphasize the model’s reliability in
stratifying TCMR patients based on gene expression profiles. By
enabling the early recognition and risk stratification of high-risk
patients, this model provides a potential tool for optimizing clinical
immunosuppressive regimens, which could ultimately reduce the incidence
of acute rejection and improve long-term transplant outcomes. By
enabling the early identification of high-risk TCMR patients, the model
provides a valuable tool for clinicians to optimize immunosuppressive
regimens and potentially reduce the incidence of acute rejection.
Compared to previous studies that predominantly focus on single-organ
systems ([165]36–[166]38), our multi-organ approach offers a broader
perspective by identifying universal biomarkers and therapeutic targets
shared across different organ transplant types. This comprehensive
strategy underscores the potential for developing organ-agnostic
therapeutic interventions.
Further analysis of immune infiltration patterns in TCMR patients
revealed significant increases in CD8^+ T cells, activated natural
killer (NK) cells, and M1 macrophages. These immune subsets are key
mediators of the effector phase of rejection, contributing to direct
cytotoxicity and amplifying inflammatory cascades. The observed
enrichment of these immune cell populations aligns with previous
studies emphasizing the central role of cellular immunity in TCMR
pathogenesis ([167]39–[168]43). These findings provide further evidence
that cellular immunity plays a pivotal role in TCMR and highlight the
importance of targeting these immune subsets or their upstream pathways
to mitigate rejection.
In addition, drug screening through the CMap database identified
peroxisome PPARγ agonists as promising candidates for TCMR management.
PPARγ agonists are well-known regulators of glucose and lipid
metabolism ([169]15), with established clinical applications in
treating metabolic disorders such as type 2 diabetes, dyslipidemia, and
non-alcoholic fatty liver disease ([170]44). Emerging evidence also
suggests that PPARγ agonists exert immunomodulatory effects on T cells
by inhibiting the differentiation of Th1, Th2, and Th17 effector T
cells, thereby reducing the secretion of associated cytokines such as
IFN-γ, IL-4, IL-13, and IL-17A ([171]45, [172]46). Additionally, PPARγ
agonists enhance the generation and functionality of regulatory T cells
(Tregs), which play a critical role in suppressing effector T cell
activity ([173]46). These mechanisms suggest that PPARγ agonists may
have potential as immunomodulatory agents in transplantation medicine.
In our study, the PPARγ agonist rosiglitazone demonstrated the ability
to suppress T cell activation and IL-2 production in vitro and
significantly prolonged graft survival in vivo when combined with
rapamycin. These findings are consistent with the reported
anti-inflammatory properties of PPARγ agonists ([174]47–[175]49),
supporting their potential as adjunctive therapies in TCMR. However,
the mechanisms underlying the synergy between rosiglitazone and
rapamycin warrant further investigation.
Despite these promising findings, this study has several limitations.
First, the data used to develop the predictive model and identify
therapeutic candidates were derived from publicly available databases,
which may introduce variability due to differences in data collection
and processing methods. Second, the heterogeneity of transplant types
and patient populations poses challenges to the generalizability of the
model across all clinical scenarios. Future investigations should focus
on expanding the sample size and diversity of patient cohorts,
particularly by incorporating longitudinal data to assess temporal
changes in TCMR-related gene expression and immune infiltration.
Finally, while PPARγ agonist rosiglitazone demonstrated
immunomodulatory potential in preliminary analyses, its efficacy in
clinical settings requires validation through larger in vivo and
clinical trials. These efforts are essential for optimizing drug dosing
strategies, minimizing potential off-target effects, and evaluating
combinatorial therapies with existing immunosuppressive regimens.
5. Conclusion
In this study, we defined a novel gene set—TCMR-Cs by analyzing gene
expression profiles from kidney, heart, and lung transplant biopsies.
This gene set formed the basis for a predictive model that demonstrated
high diagnostic accuracy across multiple transplant cohorts, validating
its potential as a tool for risk stratification and clinical
decision-making. Furthermore, using the TCMR-Cs, we identified PPARγ
agonists through CMap-based drug repurposing as promising therapeutic
candidates for TCMR. Experimental validation showed that PPARγ agonist
rosiglitazone effectively suppressed T cell activation in vitro and
prolonged graft survival in vivo when combined with rapamycin. These
findings emphasize the importance of TCMR-Cs in understanding
cytokine-driven immune dysregulation and highlight the potential of
PPARγ agonists as adjunctive therapies in transplantation medicine.
Funding Statement
The author(s) declare financial support was received for the research,
authorship, and/or publication of this article. This study was
supported by the National Natural Science Foundation of China (Grant
No. 82241219, 82127808, 81921004).
Data availability statement
The original contributions presented in the study are included in the
article/[176] Supplementary Material . Further inquiries can be
directed to the corresponding authors.
Ethics statement
The animal study was approved by Ethics Committee of Tianjin First
Central Hospital. The study was conducted in accordance with the local
legislation and institutional requirements.
Author contributions
LH: Conceptualization, Data curation, Formal analysis, Investigation,
Methodology, Validation, Visualization, Writing – original draft,
Writing – review & editing. XZ: Investigation, Methodology, Validation,
Writing – review & editing. WZ: Investigation, Methodology, Validation,
Writing – review & editing. SJ: Methodology, Validation, Writing –
review & editing. JZ: Conceptualization, Project administration,
Writing – review & editing. JMZ: Conceptualization, Methodology,
Project administration, Writing – review & editing. WS:
Conceptualization, Methodology, Project administration, Writing –
review & editing. ZS: Conceptualization, Funding acquisition, Project
administration, Writing – review & editing.
Conflict of interest
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Generative AI statement
The author(s) declare that no Generative AI was used in the creation of
this manuscript.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or claim
that may be made by its manufacturer, is not guaranteed or endorsed by
the publisher.
Supplementary material
The Supplementary Material for this article can be found online at:
[177]https://www.frontiersin.org/articles/10.3389/fimmu.2025.1539645/fu
ll#supplementary-material
[178]DataSheet1.docx^ (4.4MB, docx)
References