Prediction of steroid demand in the treatment of patients with ulcerative colitis by immunohistochemical analysis of the mucosal microenvironment and immune checkpoint: role of macrophages and regulatory markers in disease severity

Ulcerative Colitis, LAIR1 and TOX2 Expression, and Colorectal Cancer Deep Learning Image Classification Using Convolutional Neural Networks

BACKGROUND

Ulcerative colitis is a chronic inflammatory bowel disease of the colon mucosa associated with a higher risk of colorectal cancer.

OBJECTIVE

This study classified hematoxylin and eosin (H&E) histological images of ulcerative colitis, normal colon, and colorectal cancer using artificial intelligence (deep learning).

METHODS

A convolutional neural network (CNN) was designed and trained to classify the three types of diagnosis, including 35 cases of ulcerative colitis (n = 9281 patches), 21 colon control (n = 12,246), and 18 colorectal cancer (n = 63,725). The data were partitioned into training (70%) and validation sets (10%) for training the network, and a test set (20%) to test the performance on the new data. The CNNs included transfer learning from ResNet-18, and a comparison with other CNN models was performed. Explainable artificial intelligence for computer vision was used with the Grad-CAM technique, and additional LAIR1 and TOX2 immunohistochemistry was performed in ulcerative colitis to analyze the immune microenvironment.

RESULTS

Conventional clinicopathological analysis showed that steroid-requiring ulcerative colitis was characterized by higher endoscopic Baron and histologic Geboes scores and LAIR1 expression in the lamina propria, but lower TOX2 expression in isolated lymphoid follicles (all p values < 0.05) compared to mesalazine-responsive ulcerative colitis. The CNN classification accuracy was 99.1% for ulcerative colitis, 99.8% for colorectal cancer, and 99.1% for colon control. The Grad-CAM heatmap confirmed which regions of the images were the most important. The CNNs also differentiated between steroid-requiring and mesalazine-responsive ulcerative colitis based on H&E, LAIR1, and TOX2 staining. Additional classification of 10 new cases of colorectal cancer (adenocarcinoma) were correctly classified.

CONCLUSIONS

CNNs are especially suited for image classification in conditions such as ulcerative colitis and colorectal cancer; LAIR1 and TOX2 are relevant immuno-oncology markers in ulcerative colitis.

Molecular Activity of Inflammation and Epithelial-Mesenchymal Transition in the Microenvironment of Ulcerative Colitis

Background/Aims

: The genetic expression in the active inflammatory regions is increased in ulcerative colitis (UC) with endoscopic activity. The aim of this study was to investigate the molecular activity of inflammation and tissue remodeling markers in endoscopically inflamed and uninflamed regions of UC.

Methods

: Patients with UC (n=47) and controls (n=20) were prospectively enrolled at the Seoul National University Bundang Hospital. Inflamed tissue was obtained at the most active lesion, and uninflamed tissue was collected from approximately 15 cm above the upper end of the active lesion via colonoscopic biopsies. The messenger RNA expression levels of transforming growth factor β (TGF-β), interleukin (IL)-1β, IL-6, IL-17A, E-cadherin, olfactomedin-4 (OLFM4), leucine-rich repeat-containing G protein-coupled receptor 5 (LGR5), vimentin, fibroblast-specific protein-1 (FSP1), and α-smooth muscle actin (SMA) were evaluated. Mucosal healing (MH) was defined according to a Mayo endoscopic score of 0, 1 or non-MH (Mayo endoscopic score of 2 or 3).

Results

: The messenger RNA expressions of TGF-β, IL-1β, OLFM4, FSP1, vimentin, and α-SMA were significantly higher, and that of E-cadherin was significantly lower in inflamed and uninflamed regions of patients with UC than those in controls. In the inflamed regions, patients in the non-MH group had significantly increased genetic expression of TGF-β, FSP1, vimentin, and α-SMA compared to patients in the MH group. Similarly, the non-MH group had significantly higher genetic expression of TGF-β, IL-1β, IL-6, vimentin, and α-SMA than the MH group in the uninflamed regions.

Conclusions

: Endoscopic activity in UC suggests inflammation and tissue remodeling of uninflamed regions similar to inflamed regions (ClinicalTrials.gov, NCT05653011).

Urinary soluble CD163 is a putative non-invasive biomarker for primary sclerosing cholangitis

Soluble CD163 (sCD163) is a selective marker of macrophages whose circulating levels have been found to be induced in patients with active inflammatory bowel disease (IBD). Urinary proteins are emerging as non-invasive diagnostic biomarkers, and here, sCD163 levels were measured in the urine of 18 controls and 63 patients with IBD by enzyme-linked immunosorbent assay. Urinary sCD163 levels did, however, not differentiate IBD patients from controls. Analysis of sCD163 in the serum of 51 of these patients did not show higher levels in IBD. Primary sclerosing cholangitis (PSC) is often associated with IBD, and sCD163 was higher in the urine of the 21 patients and in the serum of the 13 patients with PSC compared to patients with IBD. Of clinical relevance, urinary sCD163 levels were higher in PSC patients compared to those with other chronic liver diseases (n = 16), while serum sCD163 levels were comparable between the two groups. Serum sCD163 of IBD and PSC patients positively correlated with serum C-reactive protein. Serum creatinine and glomerular filtration rate, surrogate markers for renal function, did not significantly correlate with urinary or serum sCD163 levels in IBD or PSC patients. Moreover, urinary sCD163 was not related to fecal calprotectin levels whereas serum sCD163 of IBD patients showed a positive trend. PSC associated with IBD and PSC without underlying IBD had similar levels of urinary sCD163 while serum sCD163 tended to be higher in the latter group. In PSC patients, urinary sCD163 did not correlate with serum aminotransferase levels, gamma glutamyl transferase, alkaline phosphatase, bilirubin or the Model for End Stage Liver Disease score. Ursodeoxycholic acid was prescribed to our PSC patients and fecal levels of ursodeoxycholic acid and its conjugated forms were increased in PSC compared to IBD patients. Otherwise, fecal bile acid levels of IBD and PSC patients were almost identical, and were not correlated with urinary and serum sCD163 in PSC. In summary, our study identified urinary sCD163 as a potential biomarker for PSC.

Artificial Intelligence Analysis of Ulcerative Colitis Using an Autoimmune Discovery Transcriptomic Panel

Ulcerative colitis is a bowel disease of unknown cause. This research is a proof-of-concept exercise focused on determining whether it is possible to identify the genes associated with ulcerative colitis using artificial intelligence. Several machine learning and artificial neural networks analyze using an autoimmune discovery transcriptomic panel of 755 genes to predict and model ulcerative colitis versus healthy donors. The dataset GSE38713 of 43 cases from the Hospital Clinic of Barcelona was selected, and 16 models were used, including C5, logistic regression, Bayesian network, discriminant analysis, KNN algorithm, LSVM, random trees, SVM, Tree-AS, XGBoost linear, XGBoost tree, CHAID, Quest, C&R tree, random forest, and neural network. Conventional analysis, including volcano plot and gene set enrichment analysis (GSEA), were also performed. As a result, ulcerative colitis was successfully predicted with several machine learning techniques and artificial neural networks (multilayer perceptron), with an overall accuracy of 95–100%, and relevant pathogenic genes were highlighted. One of them, programmed cell death 1 ligand 1 (PD-L1, CD274, PDCD1LG1, B7-H1) was validated in a series from the Tokai University Hospital by immunohistochemistry. In conclusion, artificial intelligence analysis of transcriptomic data of ulcerative colitis is a feasible analytical strategy.

Artificial Intelligence Analysis of Gene Expression Predicted the Overall Survival of Mantle Cell Lymphoma and a Large Pan-Cancer Series

Mantle cell lymphoma (MCL) is a subtype of mature B-cell non-Hodgkin lymphoma characterized by a poor prognosis. First, we analyzed a series of 123 cases (GSE93291). An algorithm using multilayer perceptron artificial neural network, radial basis function, gene set enrichment analysis (GSEA), and conventional statistics, correlated 20,862 genes with 28 MCL prognostic genes for dimensionality reduction, to predict the patients' overall survival and highlight new markers. As a result, 58 genes predicted survival with high accuracy (area under the curve = 0.9). Further reduction identified 10 genes: KIF18A, YBX3, PEMT, GCNA, and POGLUT3 that associated with a poor survival; and SELENOP, AMOTL2, IGFBP7, KCTD12, and ADGRG2 with a favorable survival. Correlation with the proliferation index (Ki67) was also made. Interestingly, these genes, which were related to cell cycle, apoptosis, and metabolism, also predicted the survival of diffuse large B-cell lymphoma (GSE10846, n = 414), and a pan-cancer series of The Cancer Genome Atlas (TCGA, n = 7289), which included the most relevant cancers (lung, breast, colorectal, prostate, stomach, liver, etcetera). Secondly, survival was predicted using 10 oncology panels (transcriptome, cancer progression and pathways, metabolic pathways, immuno-oncology, and host response), and TYMS was highlighted. Finally, using machine learning, C5 tree and Bayesian network had the highest accuracy for prediction and correlation with the LLMPP MCL35 proliferation assay and RGS1 was made. In conclusion, artificial intelligence analysis predicted the overall survival of MCL with high accuracy, and highlighted genes that predicted the survival of a large pan-cancer series.

Anti-HBV drug entecavir ameliorates DSS-induced colitis through PD-L1 induction

PD-L1-mediated signaling is one of the major processes that regulate local inflammatory responses in the gut. To date, protective effects against colitis through direct Fc-fused PD-L1 administration or indirect PD-L1 induction by probiotics have been reported. We have previously shown that the anti-HBV drug entecavir (ETV) induces PD-L1 expression in human hepatocytes. In the present study, we investigated whether ETV induces PD-L1 expression in intestinal cells and provides a protective effect against DSS-induced colitis. ETV induced PD-L1 expression in epithelial cells, rather than T and B cells, improving the symptoms of colitis. In the mechanistic analysis, Th17 cell differentiation was inhibited and B cell infiltration into the lamina propria was reduced. In addition, PD-L1 expression was positively correlated with Foxp3 or CSF1-R. In conclusion, ETV upregulated PD-L1 expression in epithelial cells and ameliorated inflammation in DSS-induced colitis. These results suggest that ETV may be a potential therapeutic agent as a PD-L1 enhancer for the treatment of human IBD.

High PTX3 expression is associated with a poor prognosis in diffuse large B-cell lymphoma

Tumor‐associated macrophages (TAMs) are associated with a poor prognosis of diffuse large B‐cell lymphoma (DLBCL). As macrophages are heterogeneous, the immune polarization and their pathological role warrant further study. We characterized the microenvironment of DLBCL by immunohistochemistry in a training set of 132 cases, which included 10 Epstein–Barr virus‐encoded small RNA (EBER)‐positive and five high‐grade B‐cell lymphomas, with gene expression profiling in a representative subset of 37 cases. Diffuse large B‐cell lymphoma had a differential infiltration of TAMs. The high infiltration of CD68 (pan‐macrophages), CD16 (M1‐like), CD163, pentraxin 3 (PTX3), and interleukin (IL)‐10‐positive macrophages (M2c‐like) and low infiltration of FOXP3‐positive regulatory T lymphocytes (Tregs) correlated with poor survival. Activated B cell‐like DLBCL was associated with high CD16, CD163, PTX3, and IL‐10, and EBER‐positive DLBCL with high CD163 and PTX3. Programmed cell death‐ligand 1 positively correlated with CD16, CD163, IL‐10, and RGS1. In a multivariate analysis of overall survival, PTX3 and International Prognostic Index were identified as the most relevant variables. The gene expression analysis showed upregulation of genes involved in innate and adaptive immune responses and macrophage and Toll‐like receptor pathways in high PTX3 cases. The prognostic relevance of PTX3 was confirmed in a validation set of 159 cases. Finally, in a series from Europe and North America (GSE10846, R‐CHOP‐like treatment, n = 233) high gene expression of PTX3 correlated with poor survival, and moderately with CSF1R, CD16, MITF, CD163, MYC, and RGS1. Therefore, the high infiltration of M2c‐like immune regulatory macrophages and low infiltration of FOXP3‐positive Tregs is associated with a poor prognosis in DLBCL, for which PTX3 is a new prognostic biomarker.

Targeting of CD163+ Macrophages in Inflammatory and Malignant Diseases

The macrophage is a key cell in the pro- and anti-inflammatory response including that of the inflammatory microenvironment of malignant tumors. Much current drug development in chronic inflammatory diseases and cancer therefore focuses on the macrophage as a target for immunotherapy. However, this strategy is complicated by the pleiotropic phenotype of the macrophage that is highly responsive to its microenvironment. The plasticity leads to numerous types of macrophages with rather different and, to some extent, opposing functionalities, as evident by the existence of macrophages with either stimulating or down-regulating effect on inflammation and tumor growth. The phenotypes are characterized by different surface markers and the present review describes recent progress in drug-targeting of the surface marker CD163 expressed in a subpopulation of macrophages. CD163 is an abundant endocytic receptor for multiple ligands, quantitatively important being the haptoglobin-hemoglobin complex. The microenvironment of inflammation and tumorigenesis is particular rich in CD163+ macrophages. The use of antibodies for directing anti-inflammatory (e.g., glucocorticoids) or tumoricidal (e.g., doxorubicin) drugs to CD163+ macrophages in animal models of inflammation and cancer has demonstrated a high efficacy of the conjugate drugs. This macrophage-targeting approach has a low toxicity profile that may highly improve the therapeutic window of many current drugs and drug candidates.

A Single Gene Expression Set Derived from Artificial Intelligence Predicted the Prognosis of Several Lymphoma Subtypes; and High Immunohistochemical Expression of TNFAIP8 Associated with Poor Prognosis in Diffuse Large B-Cell Lymphoma

Objective: We have recently identified using multilayer perceptron analysis (artificial intelligence) a set of 25 genes with prognostic relevance in diffuse large B-cell lymphoma (DLBCL), but the importance of this set in other hematological neoplasia remains unknown. Methods and Results: We tested this set of genes (i.e., ALDOB, ARHGAP19, ARMH3, ATF6B, CACNA1B, DIP2A, EMC9, ENO3, GGA3, KIF23, LPXN, MESD, METTL21A, POLR3H, RAB7A, RPS23, SERPINB8, SFTPC, SNN, SPACA9, SWSAP1, SZRD1, TNFAIP8, WDCP and ZSCAN12) in a large series of gene expression comprised of 2029 cases, selected from available databases, that included chronic lymphocytic leukemia (CLL, n = 308), mantle cell lymphoma (MCL, n = 92), follicular lymphoma (FL, n = 180), DLBCL (n = 741), multiple myeloma (MM, n = 559) and acute myeloid leukemia (AML, n = 149). Using a risk-score formula we could predict the overall survival of the patients: the hazard-ratio of high-risk versus low-risk groups for all the cases was 3.2 and per disease subtype were as follows: CLL (4.3), MCL (5.2), FL (3.0), DLBCL not otherwise specified (NOS) (4.5), multiple myeloma (MM) (5.3) and AML (3.7) (all p values < 0.000001). All 25 genes contributed to the risk-score, but their weight and direction of the correlation was variable. Among them, the most relevant were ENO3, TNFAIP8, ATF6B, METTL21A, KIF23 and ARHGAP19. Next, we validated TNFAIP8 (a negative mediator of apoptosis) in an independent series of 97 cases of DLBCL NOS from Tokai University Hospital. The protein expression by immunohistochemistry of TNFAIP8 was quantified using an artificial intelligence-based segmentation method and confirmed with a conventional RGB-based digital quantification. We confirmed that high protein expression of TNFAIP8 by the neoplastic B-lymphocytes associated with a poor overall survival of the patients (hazard-risk 3.5; p = 0.018) as well as with other relevant clinicopathological variables including age >60 years, high serum levels of soluble IL2RA, a non-GCB phenotype (cell-of-origin Hans classifier), moderately higher MYC and Ki67 (proliferation index), and high infiltration of the immune microenvironment by CD163-positive tumor associated macrophages (CD163+TAMs). Conclusion: It is possible to predict the prognosis of several hematological neoplasia using a single gene-set derived from neural network analysis. High expression of TNFAIP8 is associated with poor prognosis of the patients in DLBCL.