We selected 8 cases based on histopathological and clinical features ( Table 1 ), representing 4 diagnostic groups: 1) non-rejection conditions; 2) Active AMR; 3) Acute TCMR; 4) Chronic active AMR. The clinical diagnosis is interpreted by our renal pathologists based on the 2018 Banff Criteria (22).

We performed 10x Genomic Visium spatial transcriptomics analysis on H&E - stained sections from archived FFPE core needle biopsies of human kidney allografts following Visium Spatial Gene Expression for FFPE workflow (Graphic Abstract). 1) Sample preparation and RNA quality control: section FFPE tissues onto charged glass slides. 2) Assess RNA integrity using methods Distribution Value 200 (DV200): DV200 represents the percentage of RNA fragments that are longer than 200 nucleotides in a sample. This method is particularly useful for evaluating the quality of degraded RNA samples, such as those extracted from FFPE tissue. Only samples with a DV200 value equal to or greater than 30% were processed. 3) Performed standard H&E staining directly on the glass slides. 4) Evaluated H&E staining slides to select areas of interest for 6.5 x 6.5 mm capture areas. 5) Probe Hybridization with whole transcriptome probe panels. 6) Used the Visium CytAssist instrument to precisely transfer bound probes onto the Visium slide. The Visium slide contains 6.5 x 6.5 mm capture areas with 55 μm barcoded squares. 7) Generated gene expression libraries from each tissue section (library preparation). 8) Sequenced the libraries on compatible Illumina sequencers, such as NovaSeq X series systems. 9) Employed Space Ranger software for data processing, applied standard quality control metrics to filter out low-quality spots, and combined all eight samples into a unified dataset (23–25). 10) Utilized Loupe Browser for interactive data exploration, integrating whole transcriptome analysis with precise spatial information from archived FFPE samples.

All bioinformatics analysis was performed utilizing the BioTuring Lens platform (https://bioturing.com) (26, 27). 10X Visium spots were clustered via the Louvain method (principal component analysis (PCA) Resolution=1). Uniform Manifold Approximation and Projection (UMAP) visualization or t-distributed stochastic neighbor embedding (t-SNE) dimension reduction were generated via PCA of gene expression with no batch correction (n_neighbors=30). Segmentation analysis was applied to acquire 4–7 unsupervised clusters in each diagnostic category ( Supplementary Figure 1 ). Cell types and subtypes per Louvain-derived cluster were predicted using the HaiTam algorithm (https://talk2data.bioturing.com). Spots that were not confidently characterized into a single cell type (i.e., undefined) were omitted from the analysis. UMAP-based visualization displayed clusters with annotated labels, which were obtained based on histopathologic features and known marker genes associated with kidney structures ( Supplementary Figure 1 ) (28). Differential expression of genes (DEG) among spots in each case was calculated via the Venice algorithm (p<0.05) treating each spot as an individual sample data point. Hierarchical clustering heatmaps of the DEGs were generated and organized via a dendrogram of the cases and cluster plots of marker genes per cluster. Expression of specific genes per spot was measured and overlayed onto the UMAP or t-SNE.

To evaluate the consistency between our findings and existing research, we compared the transcriptomic signatures of AMR and TCMR from our FFPE tissue analysis with RNA signatures derived from frozen tissue bulk transcriptome microarrays, as reported by Halloran et al. in 2018 and 2024. This comparison was visualized using a Venn diagram, highlighting similarities and differences between the two approaches.

To analyze the gene networks, canonical, and bio-functional pathways, we applied Gene Ontology (GO) Enrichment Analysis tools to the lists of differentially expressed genes (ShinyGo v0.66, http://bioinformatics.sdstate.edu/go/) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (29).

To identify DEGs in each of the rejection types with respect to non-rejection conditions, we used the Venice algorithm. Hierarchical clustering of the DEGs revealed distinct gene expression pattens among 8 cases ( Figure 1 ). Similar transcriptomic profiles patterns were observed among two cases in the same diagnostic groups (non-rejection cases, acute TCMR and chronic active AMR), except for the active AMR group. The C4d-positive active AMR case demonstrated significantly different transcriptomic signatures compared to the C4d-negative active AMR case, despite both being positive for donor-specific antibodies (DSA). Moreover, C4d-negative active AMR case showed a closer pattern to chronic active AMR cases. Furthermore, chronic active AMR cases shared some overlapping features with acute TCMR, which is consistent with recent study published by Shah et, al (30). These results demonstrated that the transcriptomic signatures from FFPE core needle biopsy tissues have the potential to aid in distinguishing between different types of rejection and may also enable further subclassification of AMR.

To evaluate the concordance between DEG derived from our FFPE tissue transcriptomic signatures with top transcripts associated with rejection by MMDX (16, 31), we compared two gene sets and observed some overlapping between our FFPE tissue transcriptomic signatures associated with active AMR and acute TCMR, and the MMDX transcripts linked to universal rejection (16) ( Supplementary Figures 2A–C ). In addition, our FFPE tissue transcriptomic signatures associated with active AMR and acute TCMR showed some overlapping with the top 20 transcripts linked to AMR, TCMR, and injury- and rejection-associated transcripts as reported by Halloran et al. in their 2024 MMDX study ( Supplementary Figures 2D–F ).

Furthermore, our analysis of the top 30 transcriptomic signatures in FFPE tissue from rejection groups ( Supplementary Table 1 ) revealed additional important genes that are associated with transplant rejection. For example, in C4d-positive active AMR case, S100A8 and S100A9 were significantly upregulated. These calcium-binding proteins, primarily expressed in monocytes, play a crucial role in kidney transplant rejections, and high expression levels of S100A8 and S100A9 in myeloid cells during kidney transplant rejections have been linked to favorable outcomes (32). In acute TCMR, the expression of FCGR3A gene, which encodes the Fc gamma receptor IIIA (FcγR IIIA or CD16), was significantly increased, with its specific role to be elaborated upon later. Additionally, Interferon Regulatory Factor 4 (IRF4) was significantly upregulated. Similar to IRF1, this transcription factor is critical for immune regulation, particularly in T and B cells, and plays a significant role in transplant rejection by regulating genes involved in inflammation and lymphocyte activation (33). IRF4 not only regulates adaptive immune responses but also plays a crucial role in the function and differentiation of innate immune cells such as monocytes and macrophages (34). For example, IRF4 negatively modulates proinflammatory cytokine production by macrophages following Toll-like receptor stimulation, underscoring its vital regulatory role in innate immunity (35). Moreover, the expression of complement component C3 was significantly increased. C3, part of the complement system that is frequently activated in acute AMR (33), was also significantly increased in acute TCMR.

Acute rejection poses a significant threat to allograft survival. It is crucial to identify the specific cell populations that play key roles in various forms of acute rejection. Understanding these cellular dynamics is essential for developing potential innovative targeted therapies and improving long-term transplant outcomes. Therefore, we performed a joint visualization of spots in all cases using t-SNE dimension reduction method. The cell type composition of each case ( Figure 2A ) was generated by referencing the expression profiles of 10X Visium bins against a published meta-database of characterized kidney cells using BioTuring. Acute TCMR cases demonstrated a prominent tissue-resident macrophage population ( Figure 2B ). These tissue-resident macrophages (markers: CD68 and CD163) exhibited high expression of FCGR3A ( Figure 2C ). The “Monocyte category” includes classical (FCGR3A- and CD14+), intermediate monocytes (FCGR3A+ and CD14+) and non-classical monocytes (FCGR3A+ and CD14-), while the “classical monocyte” category specifically represents the classical monocytes (36). The analysis revealed that the C4d-positive active AMR case showed a significant population of non-classical and intermediate monocytes, which was the highest among and significantly different from all other cases ( Figure 2B ). This distinct subcluster of monocytes (markers: CD14 and CD68) demonstrated a high FCGR3A expression ( Figure 2D ). Spatial transcriptomics data analysis of FCGR3A expression using UMAP visualization for each case is shown in Supplementary Figure 3 . FCGR3A is involved in cellular cytotoxicity and is thought to play a significant role in acute rejection (18, 37). Our findings echo those of Lamarthée et al, who demonstrated a specific association between recipient-derived FCGR3A+ monocytes and NK cells, and the severity of intragraft inflammation. Their study utilized different technologies - scRNA-seq and multiplexed immunofluorescence (MILAN) - on different sample types (human frozen kidney biopsy tissues).

To identify the spatial locations of these distinct monocyte/macrophage subclusters, the expression of monocyte/macrophage markers and FCGR3A was mapped onto the biopsy H&E images using Loupe Browser ( Figure 3 ). In C4d-positive AMR, clusters over representative areas of peritubular capillaritis (PTCitis) and glomerulitis showed enrichment in both monocyte/macrophage markers and FCGR3A expression ( Figures 3A, B ). In acute TCMR, both grade 1B ( Figures 3C, D ) and grade 2A ( Figures 3E,F ) cases demonstrated enrichment of monocyte/macrophage markers and FCGR3A expression in clusters over representative areas of tubulitis and interstitial inflammation. Additionally, inflammatory cells in the intimal arteritis (V1 lesion) of the acute TCMR grade 2A case exhibited high co-expression of monocyte/macrophage markers and FCGR3A ( Figures 3E, F ).

To identify enriched functional pathway associated with DEG, we performed functional pathway analysis of the DEGs using GO enrichment analysis and KEGG analysis (29). GO analysis revealed top perturbed GO biological process pathways enriched in all rejection groups, with key pathways associated with metabolic changes in trained immunity ( Figures 4A–D ). For instance, carboxylic acid catabolic, amino acid and fatty acid metabolic process pathways were upregulated in C4d-negative active AMR ( Figure 4A ). Intermediates from these process can enter glycolysis and the tricarboxylic acid (TCA) cycle, linking these pathways together (38). In chronic active AMR, there was an increase in aerobic glycolysis and mitochondrial oxidative metabolism (such as oxidative phosphorylation, respiratory electron transport chain, and Adenosine triphosphate (ATP) synthesis) ( Figure 4D ). In contrast to C4d-negative active AMR and chronic active AMR, we observed several key immune-related pathways in C4d-positive active AMR ( Figure 4B ). These included pathways involved in activating and regulating immune responses, as well as those regulating innate immune responses and NF-kappa B signaling. These findings parallelled our observations in acute TCMR, where we also identified upregulation of pathways associated with mononuclear cells (lymphocytes and monocytes/macrophages) differentiation, immune response-activating signaling pathways, phagocytosis, and regulation of innate immune response ( Figure 4C ).

KEGG analysis supported the GO analysis findings, revealing similar upregulation of metabolic pathways in both C4d-negative active AMR and chronic active AMR ( Figures 4E–H ). Moreover, both conditions exhibited increased ROS production. In addition to the immune-related pathways identified in the GO analysis, KEGG analysis uncovered upregulation of additional rejection-associated damage and macrophage response to transplant allografts pathways in TCMR, including ROS production, leukocyte trans-endothelial migration and FcγR-mediated phagocytosis ( Figure 4G ). Furthermore, KEGG analysis revealed upregulation of antigen processing and presentation pathways in chronic active AMR ( Figure 4H ).

Innate allorecognition, which allows innate immune cells to discriminate between self and non-self, is one of the most important mechanisms of innate immune activation during acute transplant rejection (39). CD47, leukocyte immunoglobulin-like receptor A (LILRA), and signal-regulatory protein-α (SIPRα) are key markers associated with monocytes/macrophage activation and function in both transplant rejection and trained immunity within the innate alloantigen recognition pathway (40). The LILR family consists of 11 innate immunomodulatory receptors, primarily expressed on lymphoid and myeloid cells. Based on their signaling domains, LILRs are classified as either activating (LILRA) or inhibitory (LILRB). LILRA1–2 and LILRA4-6, with the exception of the soluble LILRA3, mediate immune activation, whereas LILRB1–5 primarily inhibit immune responses and promote tolerance (41). On allograft tissues, SIPRα and MHC class I antigens are expressed and are recognized by CD47 and LILRA that are expressed on host monocytes, respectively. The UMAP visualization ( Figures 5A–E ) and violin plots of log2 fold changes ( Figures 5F–I , Supplementary Figures 4A–D ) illustrated significantly higher expression of CD47 (p<0.05) ( Figures 5D, H , Supplementary Figure 4C ) and notably higher expression of SIRPα in C4d-positive active AMR ( Figures 5E, I , Supplementary Figure 4D ). CD47 and SIPRα expression are also upregulated in acute TCMR cases. However, this upregulation is not as pronounced as in the C4d-positive active AMR case. The interaction between FCGR3A and LILRA is believed to play important roles in monocytes/macrophage activation and function during transplant rejection (40, 42, 43), and we observed significant upregulation of FCGR3A in C4d-positive active AMR and acute TCMR cases ( Figures 5B, F , Supplementary Figure 4A ). However, we did not observe significant LILRA1–6 expression upregulation among these cases ( Figures 5C, G , Supplementary Figures 4B, E ). Although LILRB1–5 generally suppress immune responses and promote tolerance, LILRB2 expression is notably increased in C4d-positive active AMR case. This may be explained by recent findings that LILRB2 activation is associated with macrophage recruitment and an inflammatory macrophage phenotype, as observed in non-alcoholic steatohepatitis (NASH) (44).

The expression of key metabolic gene markers across different groups for trained immunity, including the key genes involved in glycolysis and mitochondrial oxidative metabolism were depicted as bubble plot ( Figure 6 ). The bubble plot also included genes that encode metabolic intermediates, which are believed to induce epigenetic changes, such as fumarase (FH) gene and succinate dehydrogenase complex (SDHA/SDHB/SDHC/SDHD). This analysis revealed distinct expression patterns between groups experiencing rejection and those without rejection. Non-rejection conditions, such as acute calcineurin inhibitor (CNI) toxicity and subtle acute tubular injury (ATI), showed elevated activity in the mTOR pathway, glycolysis, and mitochondrial oxidative metabolism. In contrast, all rejection groups exhibited more pronounced elevations in glycolysis and mitochondrial oxidative metabolism activities than mTOR pathway activity. Notably, within glycolysis-related genes, Enolase 1 (ENO1) showed a significant increase in non-rejection conditions and C4d-negative active AMR, while Pyruvate kinase (PKM) was significantly elevated in acute TCMR groups and chronic active AMR. C4d-positive active AMR displayed significant increases in both genes. Additionally, clusters associated with acute TCMR and chronic active AMR showed evidence of increased levels of metabolic intermediates, SDHA/SDHB, which are thought to induce epigenetic changes.