A protein marker for differential diagnosis of extranodal natural killer / t-cell lymphoma of the intestine and use thereof

Abstract

This invention discloses a protein biomarker for the differential diagnosis of intestinal extranodal natural killer / T-cell lymphoma and its uses. The protein biomarker is composed of COL6A3 and MFAP2 proteins; or ADAMTS4, COL6A3, and MFAP2 proteins. This invention also provides the use of reagents for detecting the above proteins in the preparation of kits for diagnosing intestinal natural killer / T-cell lymphoma or assessing its risk. This invention effectively solves the technical bottleneck of difficult pathological diagnosis of intestinal ENKTL due to the superficial nature of endoscopic biopsy samples, and can accurately differentiate ENKTL from ulcerative colitis, diffuse large B-cell lymphoma, and colon cancer. This invention is simple to operate; the selected antibody has a strong signal and clear background in IHC detection, and combined with the H-score scoring system, it can obtain quantitative results with good repeatability and high specificity, significantly improving the objectivity and accuracy of pathological diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of biological detection and relates to a disease diagnostic biomarker. Specifically, it is a combination of protein biomarkers for differentiating between extranodal natural killer / T-cell lymphoma (ENKTL), inflammatory bowel disease, and other intestinal tumors, and their applications. Background Technology

[0002] Extranodal natural killer cell (NK) / T-cell lymphoma (ENKTL) is a highly aggressive, Epstein-Barr virus (EBV)-associated non-Hodgkin lymphoma. Its pathogenesis involves the interaction of multiple factors, including host genetic factors and persistent dysregulation of oncogenic signaling pathways. Although ENKTL most commonly occurs in the nasal cavity, primary or secondary intestinal involvement results in a very poor clinical prognosis due to nonspecific symptoms and rapid disease progression.

[0003] Digestive endoscopy and biopsy are crucial for the early detection and definitive diagnosis of intestinal lymphoma. However, due to the deep location of lesions and the risk of perforation, endoscopic biopsies often fail to obtain representative tumor tissue. Therefore, superficial samples typically show only normal tissue, or nonspecific necrosis and inflammation, similar to non-malignant diseases such as inflammatory bowel disease (IBD). This makes it difficult to identify NK / T cell phenotype tumor cells based on positive CD3ε and EBER expression, thus hindering a definitive diagnosis. This diagnostic ambiguity leads to frequent misdiagnosis, delayed treatment, and consequently, a median survival of only a few months. Therefore, establishing reliable biomarkers that can aid in diagnosis even in suboptimal biopsy samples is essential.

[0004] Besides the challenges of sampling, potent diagnostic biomarkers for intestinal ENKTLs remain limited, primarily due to the disease's significant heterogeneity. ENKTLs originate from NK or T cells, exhibit pleomorphic cell morphology, and carry multiple genetic variants, such as HLA-II and IL18RAP. Furthermore, heterogeneity in EBV integration patterns and viral gene expression further increases tumor biodiversity. This intrinsic heterogeneity within tumor cells significantly complicates the identification of stable, specific diagnostic biomarkers.

[0005] To overcome these limitations, research has increasingly shifted its focus to the tumor microenvironment (TME), a complex ecosystem encompassing malignant cells, immune infiltration, blood vessels, fibroblasts, and the extracellular matrix (ECM), which regulates tumor progression, immune escape, and drug resistance. Within this microenvironment, the ECM is particularly crucial. The ECM not only provides structural support but also promotes invasion through dynamic interactions with tumor cells. In ENKTL, a malignant tumor characterized by extensive tissue destruction, ECM remodeling serves not only as a scaffold but also as a key catalyst for infiltration and damage to normal tissues. Therefore, studying ECM-related molecular alterations holds promise for revealing stable biomarkers reflecting disease invasiveness, providing new strategies for early diagnosis.

[0006] Among the many ECM remodeling-related molecules, ADAMTS4, COL6A3, and MFAP2 have shown special biological potential:

[0007] ADAMTS4 (A Disintegrin and Metalloproteinase with Thrombospondin Motifs 4) belongs to the secretory zinc finger metalloproteinase family and is primarily located in the pericellular matrix. Its core biological function is the degradation of proteoglycan-rich extracellular matrix. Its expression is strongly induced by inflammatory cytokines (such as IL-1 and TNF-α) and transforming growth factor-β (TGF-β). Studies have shown that it is highly expressed in various malignant tumors; for example, in colorectal and lung cancer, ADAMTS4 can promote tumor invasion and metastasis by degrading the Versican / Brevican network surrounding tumor cells, thus eliminating physical barriers that restrict cell movement.

[0008] COL6A3 is a crucial component of type VI collagen, playing a vital role in maintaining the structure and elasticity of human connective tissue. The N-terminus of COL6A3 contains an extremely large globular domain composed of up to 10 vWF-A modules, making it a major site for extensive interactions between type VI collagen and surrounding matrix components. In addition to providing structural support, COL6A3 is also involved in complex cellular signaling. Studies of solid tumors such as ovarian cancer have shown that contact with COL6A3 activates the PI3K / Akt signaling pathway, significantly increasing the cell's anti-apoptotic threshold. Furthermore, the matrix enriched with COL6A3 can mimic the physicochemical properties of a stem cell niche, promoting the activation of the Wnt / β-catenin pathway, thereby maintaining the self-renewal capacity of tumor cells.

[0009] MFAP2 (Microfibril Associated Protein 2, also known as MAGP1) is an important component of extracellular matrix microfibrils. MFAP2 participates in the assembly and stability of microfibrils by interacting with fibrillin-1 and fibrillin-2. MFAP2 is not only a structural protein but also a key regulator of various growth factors. Studies have confirmed that it can interact with Notch receptors and regulate the activation of Notch signaling, and the Notch pathway plays a decisive role in T-cell development and the development of T-cell lymphoma.

[0010] In summary, although ADAMTS4, COL6A3, and MFAP2 have been shown to have certain pro-cancer or matrix remodeling effects in some solid tumors or in immune regulation, no existing technology has yet revealed the specific expression patterns of these three proteins in intestinal ENKTL tissues. Furthermore, there is a lack of research and application in the organic combination of these three proteins for the differential diagnosis of intestinal ENKTL in superficial, non-ideal biopsy samples, distinguishing it from ulcerative colitis (UC) and other intestinal tumors (such as colon cancer and DLBCL). Therefore, developing a specific diagnostic tool based on the aforementioned ECM characteristic molecules is of great significance for overcoming the current clinical diagnostic bottleneck of intestinal ENKTL. Summary of the Invention

[0011] This invention provides a protein biomarker for the differential diagnosis of intestinal NK / T cell lymphoma and its application. The protein biomarker and its application aim to solve the technical problems in the prior art, such as the difficulty in diagnosing intestinal ENKTL due to the superficial biopsy tissue and the large amount of necrosis, and the easy confusion with inflammatory bowel disease (IBD) and other intestinal tumors.

[0012] In a first aspect, the present invention provides a protein biomarker for the differential diagnosis of extranodal natural killer / T-cell lymphoma of the intestine, which is composed of COL6A3 protein and MFAP2 protein or;

[0013] Or it may be composed of ADAMTS4 protein, COL6A3 protein, and MFAP2 protein.

[0014] In a second aspect, the present invention provides reagents for detecting COL6A3 and MFAP2 proteins; or reagents for detecting ADAMTS4, COL6A3, and MFAP2 proteins are used in the preparation of diagnostic reagents or kits for diagnosing intestinal natural killer / T-cell lymphoma or assessing its risk.

[0015] Preferably, the reagent is an antibody that specifically binds to the target protein, wherein the protein is ADAMTS4 protein, COL6A3 protein, and MFAP2 protein; or the protein is COL6A3 protein and MFAP2 protein.

[0016] In a third aspect, the present invention provides a kit for diagnosing or assessing the risk of extranodal natural killer / T-cell lymphoma of the intestine, comprising a detection reagent for detecting the expression levels of COL6A3 and MFAP2 proteins; or for detecting the expression levels of ADAMTS4, COL6A3, and MFAP2 proteins.

[0017] The reagents include:

[0018] A combination of antibodies against COL6A3 protein and antibodies against MFAP2 protein;

[0019] Alternatively, a combination of antibodies against ADAMTS4 protein, COL6A3 protein, and MFAP2 protein.

[0020] Furthermore, the kit is used for detection using an immunohistochemical method.

[0021] Furthermore, the kit also describes the following detection method, including the steps:

[0022] (a) Provide a test sample;

[0023] (b) Detect the expression levels of COL6A3 and MFAP2 proteins in the test samples;

[0024] Alternatively, the expression levels of ADAMTS4, COL6A3, and MFAP2 proteins in the sample can be detected and denoted as Y1.

[0025] (c) Compare the protein expression level Y1 with the control reference value Y0;

[0026] Judgment criteria: When the protein expression level Y1 of the tested object is higher than the control reference value Y0, it indicates that the object has a high risk of developing intestinal extranodal natural killer / T-cell lymphoma or is diagnosed with intestinal extranodal natural killer / T-cell lymphoma.

[0027] Furthermore, the test samples are selected from: ex vivo tissue samples (including but not limited to endoscopic biopsy tissue and surgically removed tissue).

[0028] This invention also provides a device for assessing the risk of extranodal natural killer cell / T-cell lymphoma.

[0029] The device includes:

[0030] (a) An input module, which is used to input the expression levels of COL6A3 and MFAP2 proteins of a certain object; or the expression levels of ADAMTS4, COL6A3, and MFAP2 proteins;

[0031] (b) Processing module: The processing module compares the input antibody level Y1 with the control reference value Y0.

[0032] (c) Output module, the output module is used to output the judgment result. When the input expression level data is higher than the preset threshold, it indicates that the subject has a high risk of extranodal natural killer cell / T cell lymphoma or is diagnosed with intestinal extranodal natural killer cell / T cell lymphoma; otherwise, it indicates that the subject has a low risk of extranodal natural killer cell / T cell lymphoma.

[0033] Specifically, the expression level is the H-score score of immunohistochemistry.

[0034] This invention provides a novel, highly specific identification tool. When differentiating ENKTL from ulcerative colitis (UC), DLBCL, or colon cancer, the combination of antibodies against MFAP2 and COL6A3, as well as the combination of antibodies against ADAMTS4 and COL6A3 and MFAP2, exhibits extremely high specificity, with an AUC exceeding 0.99. This effectively overcomes the technical bottleneck of difficult pathological diagnosis of intestinal ENKTL due to the superficial nature of endoscopic biopsy samples, enabling accurate differentiation of ENKTL from ulcerative colitis (UC), diffuse large B-cell lymphoma (DLBCL), and colon cancer.

[0035] Compared with existing technologies, the technical effects of this invention are positive and significant. This invention has strong sample applicability, suitable for routine formalin-fixed paraffin-embedded (FFPE) tissue sections, and especially suitable for small, superficial tissue samples obtained through endoscopic biopsy. Even in cases where tumor cells are scarce, detecting highly expressed target proteins in the background stroma can provide strong auxiliary diagnostic evidence, compensating for the limitations of morphological diagnosis. Furthermore, this invention is simple to operate and yields stable results: the selected antibody provides strong signals and clear background in IHC detection, and the H-score scoring system provides highly reproducible and specific quantitative results, significantly improving the objectivity and accuracy of pathological diagnosis, making it easy to promote in primary pathology departments. Attached Figure Description

[0036] Figure 1 : Flowchart of the implementation process of this invention.

[0037] Figure 2Representative immunohistochemical (IHC) images show the expression patterns of ADAMTS4, COL6A3, and MFAP2 in ENKTL, colon cancer, DLBCL, and UC tissues.

[0038] Figure 3 The violin plot quantifies the H scores of ADAMTS4, OL6A3, and MFAP2 in the four groups.

[0039] Figure 4 ROC curves for distinguishing ENKTL from colorectal cancer using single biomarkers and combined models.

[0040] Figure 5 ROC curves of ENKTL and DLBCL can be distinguished using a single marker and a combined model.

[0041] Figure 6 The ROC curves of ENKTL and UC were distinguished using a single marker and a combined model. Detailed Implementation

[0042] Example 1: Collection of experimental samples and clinical grouping

[0043] This example describes the process of constructing a clinical cohort for screening and validating biomarkers.

[0044] 1. Sample source: Formalin-fixed paraffin-embedded (FFPE) tissue samples were collected from the Department of Pathology, Renji Hospital, affiliated with Shanghai Jiao Tong University School of Medicine.

[0045] All sample collection was approved by the ethics committee and informed consent was obtained from the patients.

[0046] Inclusion criteria: All cases were confirmed by two or more senior pathologists according to the latest WHO classification of lymphoma.

[0047] Rejection criteria: If a biopsy tissue sample is too small or if a fresh tissue specimen is not frozen in time, which seriously affects the results of subsequent experiments, the specimen will be rejected.

[0048] 2. Sample grouping:

[0049] (1) Samples for obtaining proteomics data: 10 cases of extranodal NK / T cell lymphoma (ENKTL), all of which were primary or secondary cases involving the intestine; 6 cases of normal control group, which were obtained from normal intestinal mucosal tissue of non-tumor patients.

[0050] (2) Pathological verification cohort: including 13 cases of ENKTL, 19 cases of ulcerative colitis (UC), 17 cases of diffuse large B-cell lymphoma (DLBCL), and 17 cases of colon cancer.

[0051] Example 2: Screening and identification of intestinal ENKTL interstitial biomarkers

[0052] This embodiment describes the discovery process of the core markers to demonstrate the reliability of the source of the technical solution.

[0053] 1. Extraction of whole protein from tissues and preparation of peptides (FASP method)

[0054] 1.1 Protein Extraction: 50 mg each of intestinal ENKTL tissue and normal control tissue were collected and added to 10 volumes of SDT lysis buffer (4% (w / v) SDS, 100 mM Tris-HCl, 1 mM DTT, pH 7.6). The homogenizer was used for homogenization (6.0 M / s, 60 s, twice). The homogenate was then heated in a boiling water bath for 15 minutes, centrifuged at 14000 g for 40 minutes, and the supernatant was collected.

[0055] 1.2 Quantitative analysis: Protein concentration was determined using the BCA protein quantification kit.

[0056] 1.3 Enzymatic digestion: Take 200 μg of protein, add UA buffer (8 M urea, 150 mM Tris-HCl, pH 8.0), and replace it with a 30 kDa ultrafiltration tube to remove SDS. Add 100 μL of 50 mM iodoacetamide (IAA) and incubate at room temperature in the dark for 30 minutes for alkylation.

[0057] 1.4 Digestion: Wash twice with 100 μL ABC buffer (50 mM NH4HCO3), add trypsin at an enzyme-to-protein ratio of 1:50, and incubate overnight (18 hours) at 37°C.

[0058] 1.5 Desalting: The filtrate was collected, desalted using a C18 Cartridge column, lyophilized, and resuspended in 0.1% (v / v) formic acid solution. The peptide concentration was determined by OD280.

[0059] 2. High-resolution liquid chromatography-tandem mass spectrometry (LC-MS / MS) analysis

[0060] 2.1 High-performance liquid chromatography (HPLC): An Easy-nLC 1200 nanoliter liquid chromatography system was used. The chromatographic column was a C18 reversed-phase analytical column (75 μm × 25 cm, 1.9 μm).

[0061] (1) Mobile phase A: 0.1% (v / v) formic acid aqueous solution.

[0062] (2) Mobile phase B: 0.1% (v / v) formic acid and 80% (v / v) acetonitrile solution.

[0063] (3) Gradient elution program: 0-2 min (2%-8% B); 2-100 min (8%-37% B); 100-110 min (37%-100% B); 110-120 min (100% B). The flow rate was maintained at 300 nL / min.

[0064] 2.2 Mass spectrometry (MS): A Q Exactive HF-X mass spectrometer was used in data-dependent acquisition (DDA) mode.

[0065] (1) Mass spectrometry (MS1): Scan range 350-1500 m / z; resolution 60,000; AGC target 3e6; maximum injection time 20 ms.

[0066] (2) Secondary mass spectrometry (MS2): resolution 15,000; AGC target 1e5; maximum injection time 50 ms; isolation window 1.6 m / z; collision energy (NCE) 28%. Dynamic exclusion time: 30 s.

[0067] 3. Proteomics Data Processing and Differential Analysis

[0068] 3.1 Database Search: Using MaxQuant (v1.6.1.0) software, the UniProt Homo sapiens database (containing 20,422 sequences) was searched.

[0069] 3.2 Parameter Settings:

[0070] (1) Enzyme digestion method: Trypsin / P, allowing a maximum of 2 missed cleavage sites.

[0071] (2) Modification: The fixed modification is cysteine ​​carbamoyl methylation (C); the variable modification is methionine oxidation (M) and acetylation (Acetyl (Protein N-term)).

[0072] (3) Quality tolerance: Precursor tolerance 20 ppm for the first round and 4.5 ppm for the main search; Fragment tolerance 20 ppm.

[0073] (4) Threshold: The false positive rate (FDR) for both peptide and protein levels was set to < 1%.

[0074] Based on the above screening, a total of 4,170 protein expression data were identified.

[0075] 4. Transcriptome data integration and bioinformatics screening of key biomarkers

[0076] 4.1 Transcriptomics Data Integration

[0077] External transcriptome data were obtained from the Gene Expression Comprehensive Database (GEO). The study cohort included 66 ENKTL tumor samples from the GSE90597 dataset and 50 normal intestinal mucosal control samples from the GSE44076 dataset. All computations were performed in the R statistical environment (v4.4.3).

[0078] In the preprocessing stage, non-specific probes mapped to multiple genomic sites were filtered out. Probe identifiers were annotated as gene symbols according to their respective platform annotation files. For genes represented by redundant probes, the single probe with the highest average expression level was retained. After dataset integration, the ComBat algorithm (via the sva package, v3.54.0) was used to eliminate non-biological technical biases. The effectiveness of batch effect removal was validated by principal component analysis (PCA).

[0079] 4.2 Differential analysis of transcriptomic and proteomic data

[0080] Differential expression analysis was performed on transcriptome and proteome datasets using the limma package (v3.62.2). Features satisfying |log2FC| ≥ 1 and corrected p-value < 0.05 were considered statistically significant. Visualization was performed using hierarchical clustering heatmaps and volcano plots.

[0081] To explore functional significance, gene set enrichment analysis (GSEA) was applied to the Hallmark and KEGG pathway sets in MSigDB. Comprehensive gene ontology (GO) annotation was performed using the clusterProfiler package (v4.14.6), covering biological processes (BP), molecular functions (MF), and cellular components (CC) categories. The enrichment results were visualized using enrichment plots, bubble charts, and bar charts.

[0082] Based on the above analytical methods, a total of 624 differentially expressed proteins and 846 differentially expressed genes were identified.

[0083] 4.3 Transcriptome-weighted co-expression network analysis (WGCNA)

[0084] (1) Data preprocessing: Download and integrate external transcriptome datasets (GSE90597 and GSE44076). Calculate the variance of all genes and select the top 5000 genes by variance to construct a co-expression network.

[0085] (2) Network Construction: The WGCNA software package in R was used. The scale-free topology fit index (R) was calculated. 2 ), determine the optimal soft-thresholding power β=6, at which point R 2 >0.85, ensuring the network conforms to scale-free distribution characteristics.

[0086] (3) Module identification: The adjacency matrix is ​​converted into a topological overlap matrix (TOM), and the dissimilarity between genes (1-TOM) is calculated. A hierarchical clustering method is used to construct a clustering tree, and the minimum number of genes in a module (minModuleSize) is set to 30. The merge cut height is set to 0.25 to merge modules with highly similar expression patterns.

[0087] (4) Key module identification: Calculate the correlation between the characteristic genes (Module Eigengene, ME) of each module and the ENKTL phenotype. Screen out the modules that have the strongest positive correlation with the disease phenotype and are functionally enriched in the "extracellular matrix tissue" and "collagen fiber tissue" (i.e., Brown module and Turquoise module).

[0088] 4.4 Machine Learning Feature Optimization and Model Building

[0089] (1) Candidate set generation: The intersection of the DEPs identified by proteomics and the hub genes (the top 30 nodes in the module) in the key module of WGCNA was obtained to obtain 6 core candidate genes including ADAMTS4, COL6A3, and MFAP2, and the other three candidate genes are COL6A2, TIMP3 and DPT.

[0090] (2) To identify key genes associated with the disease, we further evaluated the diagnostic efficacy of different combinations of the above candidate biomarkers using the GSE80632 dataset (n=38, including 25 ENKTL cases and 13 normal samples) through machine learning, in order to define an optimal gene feature that balances diagnostic accuracy and clinical simplicity. The area under the receiver operating characteristic (ROC) curve (AUC) was used to quantify the discriminative performance of the model. Among them, the three-gene feature composed of COL6A3, ADAMTS4, and MFAP2 performed best. Among the four machine learning algorithms evaluated, the Random Forest (RF) and K-Nearest Neighbors (KNN) models achieved an AUC of 0.931, followed by Support Vector Machine (SVM) (AUC=0.903) and Extreme Gradient Boosting (XGBoost) (AUC=0.840). Therefore, this three-gene combination was selected as the final biomarker set for subsequent clinical validation.

[0091] Example 3: Immunohistochemistry (IHC)-based biomarker detection process and expression characterization analysis

[0092] This embodiment details the specific experimental steps for detecting ADAMTS4, COL6A3, and MFAP2 proteins using immunohistochemistry, and elucidates their characteristic expression differences in different intestinal lesion tissues.

[0093] 1. Section preparation and immunohistochemistry

[0094] 1.1 After quality control, another 66 formalin-fixed and paraffin-embedded (FFPE) tissue samples were obtained, including 13 cases of intestinal ENKTL, 19 cases of ulcerative colitis (UC), 17 cases of diffuse large B-cell lymphoma (DLBCL) and 17 cases of colon cancer tissue.

[0095] Cut the paraffin block into continuous sections 4 μm thick, mount them onto a glass slide to prevent detachment, and bake in a 60℃ oven for 1 hour to ensure tight tissue adhesion.

[0096] 1.2 Immunostaining Procedure

[0097] (1) Dewaxing and hydration: The sections were dewaxed in xylene and then hydrated in a series of ethanol (100%, 95%, 85%, 75%) to distilled water.

[0098] (2) Antigen retrieval: Immerse the slides in citrate buffer (pH 6.0) and perform antigen retrieval under heating conditions (such as microwave or high pressure), then allow them to cool naturally to room temperature.

[0099] (3) Blocking: Incubation with 3% (v / v) hydrogen peroxide solution to quench endogenous peroxidase activity; followed by incubation with normal serum blocking solution to reduce nonspecific protein binding.

[0100] (4) Primary antibody incubation: Discard the blocking solution, add the following specific primary antibodies dropwise, and incubate overnight (12-16 hours) in a humidified chamber at 4°C. Rabbit anti-COL6A3 antibody (Proteintech, 19798-1-AP), dilution ratio 1:200;

[0101] Rabbit anti-ADAMTS4 antibody (Affinity Biosciences, DF6986), diluted 1:100;

[0102] Rabbit anti-MFAP2 (MAGP1) antibody (Bioss, bs-18632R), diluted 1:100;

[0103] (5) Color development and counterstaining: After washing with PBS, use a universal HRP-labeled secondary antibody polymer (UltraView) TM Incubate with the Universal DAB Detection Kit, develop DAB staining, and counterstain cell nuclei with hematoxylin.

[0104] 2. Data Quantification and Scoring

[0105] Quantitative assessment was performed using QuPath digital pathology software or optical microscopy. Three representative fields of view (500 μm × 500 μm) were selected from typical lesion areas of each sample to quantitatively determine the expression level of the target protein in each tissue sample, using the "Positive Rate (%)" as the quantitative indicator. The positive rate data of all samples were entered into statistical analysis software to establish a raw database. The results are shown in Table 1 below:

[0106] Table 1

[0107]

[0108]

[0109]

[0110]

[0111] 3. Analysis of staining results and expression characteristics

[0112] The results of testing on 66 independent clinical samples showed that the above biomarkers exhibited significant differences in expression across different lesions. Figure 2 ):

[0113] 3.1 Intestinal ENKTL tissue: ADAMTS4, COL6A3, and MFAP2 all showed consistent strong positive expression, with positive signals mainly located in the cytoplasm and surrounding extracellular matrix (ECM). Quantitative analysis showed that the H-scores of the three proteins in the ENKTL group were significantly higher than those in all other control groups (P < 0.05).

[0114] 3.2 Ulcerative colitis (UC) tissue: The above interstitial markers showed only weak positive or focal patchy staining, and their H-score was significantly lower than that of the ENKTL group.

[0115] 3.3 DLBCL and colon cancer tissue: Samples showed negative or only weak background staining.

[0116] 4. Specificity analysis: To determine the optimal positive interpretation criteria for clinical application, this embodiment uses the Youden Index maximization principle for calculation.

[0117] Calculation formula: J = Sensitivity + Specificity - 1.

[0118] Screening process: All possible positive rate values ​​in the sample are used as candidate cutoff values, and the sensitivity and specificity corresponding to each point are calculated respectively.

[0119] The positive rate value corresponding to the maximum value of the Youden index J was selected as the optimal diagnostic cutoff value. At this cutoff value:

[0120] True Positive (TP): The number of cases that are actually ENKTL and whose test value is higher than the cutoff value.

[0121] True Negative (TN): The number of cases that are not actually ENKTL and whose test value is below the cutoff value.

[0122] Sensitivity calculation: Sensitivity = TP / (TP + FN) × 100%.

[0123] Specificity calculation: Specificity = TN / (TN+FP) × 100%.

[0124] The results are shown in Table 2 below.

[0125] Table 2

[0126]

[0127] Based on the experimental data above, this embodiment systematically verifies the feasibility of using a specific diagnostic cut-off value to assess the sensitivity and specificity of three protein biomarkers in differentiating ENKTL from other disease groups.

[0128] The analysis results show that MFAP2 exhibits the best comprehensive differential diagnostic efficacy. Its optimal cutoff value (86.69%) showed high consistency and stability among different control groups; this means that in clinical practice, there is no need to adjust the judgment criteria for different differential diagnostic subjects, and a single threshold can effectively distinguish between colorectal cancer, DLBCL, and ulcerative colitis (UC), demonstrating a significant broad-spectrum differential diagnostic advantage.

[0129] In contrast, while COL6A3 and ADAMTS4 exhibited extremely high sensitivity in differentiating unrelated tumors (colon cancer, DLBCL), they were limited in specificity in excluding the inflammatory background interference of ulcerative colitis (UC). This suggests that they are more suitable as part of a combined testing suite, or that stratified cutoff values ​​should be set to aid in diagnosis.

[0130] In summary, this embodiment demonstrates that all three proteins have the potential to be used as clinical auxiliary diagnostic biomarkers for ENKTL.

[0131] Example 4: Construction and interpretation of pathological scoring criteria and diagnostic models

[0132] To further explore the efficacy of combined diagnosis using protein biomarkers, this embodiment establishes a standardized quantitative scoring system and further explains how to construct a diagnostic model based on this quantitative data to achieve accurate identification of intestinal ENKTLs.

[0133] 1. Image Acquisition and Region of Interest (ROI) Selection

[0134] 1.1 Instruments and equipment: Observation was performed using a digital pathology slide scanner (such as the Leica or Hamamatsu series) or an optical microscope.

[0135] 1.2 ROI selection: Under low magnification of each slice, avoid necrotic and obviously normal areas, and select 3 typical tumor / inflammation junctions or densely celled areas as regions of interest (ROIs). The area of ​​each ROI is set to 500 μm × 500 μm (equivalent to 400× high magnification field).

[0136] 2. H-score quantitative scoring of mesenchymal markers (ADAMTS4, COL6A3, MFAP2) was performed using a semi-quantitative histochemical scoring (H-score) system, which integrates staining intensity and the proportion of positive cells to more accurately reflect the degree of matrix remodeling.

[0137] 2.1 Staining Intensity Grading

[0138] 0 points (negative): No color development, consistent with the background.

[0139] 1 point (weak positive): The cytoplasm or interstitium shows light yellow granules.

[0140] 2 points (moderately positive): shows a clear brownish-yellow color.

[0141] 3 points (strong positive): appears dark brown or heavily stained.

[0142] 2.2 Positive percentage (P): The percentage (0-100%) of cells / interstitial regions assessed for each intensity level.

[0143] 2.3 Calculation formula: H-score = (1 × P 弱阳性 )+(2×P 中度阳性 )+(3×P 强阳性 ).

[0144] 2.4 Result Range: The total score ranges from 0 to 300. The final result is the arithmetic mean of the three ROIs.

[0145] 3. Construction and interpretation logic of the diagnostic model

[0146] This invention uses immunohistochemistry (IHC) to detect the expression levels (H-score) of the above-mentioned proteins in tissue samples, and found that the H-scores of the three target proteins in ENKTL were significantly higher than those in other disease samples. Figure 3 To achieve accurate multi-dimensional identification, this invention uses binary logistic regression analysis to assign optimal mathematical weights to each biomarker and construct a specific hierarchical diagnostic model.

[0147] It is important to note that in the process of tumor extracellular matrix (ECM) remodeling, diagnostic significance does not solely stem from the absolute abundance of a single protein, but rather depends on the relative proportions and homeostatic imbalances among various matrix components. In our multivariate joint model, certain biomarkers (such as MFAP2) were assigned negative weights (negative coefficients). Statistically, this is a typical example of a "suppressive variable effect"; biologically, it reflects that, under the premise of controlling for high expression of dominant matrix proteins (such as ADAMTS4 and COL6A3), fine-tuning the relative expression of MFAP2 can eliminate background inflammatory noise, thereby defining the most precise boundary of malignant tumors in multidimensional space.

[0148] Based on the above principles, this invention establishes the following three diagnostic models and quantification cutoff values:

[0149] 3.1 Model A: Differentiating ENKTL from Colon Cancer Figure 4 );

[0150] Input metrics: H-scores of ADAMTS4, COL6A3, or MFAP2 (denoted as H respectively) A H C H M ).

[0151] Construction logic: Because the H-scores of these three proteins are extremely low in colorectal cancer samples (usually <30), while they are usually >100 in ENKTL samples. The AUC value of the combined model of the three proteins reaches 1.000.

[0152] Quantitative diagnostic formulas and cutoff values:

[0153] The combined diagnostic index Y = -2.060 + 0.0283 × H A +0.026×H C -0.021×H M .

[0154] Calculate the predicted probability P = 1 / (1+e) -Y ).

[0155] For ease of clinical application, it can also be equivalently converted into a linear diagnostic score: S A =28.39×H A +25.61×H C -21.18×H M

[0156] Interpretation criteria: The optimal cutoff value for the model is P = 0.741 (equivalent to the linear score S). B >3108.8). When this cutoff value is exceeded, it indicates that ENKTL is positive.

[0157] To verify the actual efficacy of the diagnostic model described in this invention in clinical samples, the following is provided: Figure 4 A detailed list of corresponding sample data is provided. This list includes the raw immunohistochemistry (IHC) score (H-score) for each sample, the linear diagnostic score (Score) calculated according to the formula in Model A, the probability of prediction, and the final interpretation result. Data shows that the model's interpretation results for all samples are completely consistent with the clinicopathological diagnoses (100% concordance rate). This confirms the model's superior efficacy in distinguishing between malignant lymphoma and cancer.

[0158]

[0159]

[0160]

[0161] Figure 4 Note: This figure shows the diagnostic efficacy analysis results based on 30 clinical tissue samples (including 13 pathologically confirmed intestinal NK / T-cell lymphomas [ENKTL] as the positive group and 17 colon cancers as the negative control group). The horizontal axis represents the false positive rate (1-specificity), and the vertical axis represents the true positive rate (sensitivity).

[0162] Figure 4 In Figure A: This shows the diagnostic efficacy of the single biomarker COL6A3. Its area under the curve (AUC) reaches 1.000, indicating that in the current sample set, COL6A3 can completely distinguish ENKTL from colorectal cancer.

[0163] Figure 4 B in the diagram shows the diagnostic efficacy of the single biomarker ADAMTS4. Its AUC is also 1.000, demonstrating extremely high specificity and sensitivity.

[0164] Figure 4 C in the figure shows the diagnostic efficacy of the single biomarker MFAP2. Its AUC is 0.964.

[0165] Figure 4 D in the figure represents the performance of the three-marker joint model constructed in this invention. The comprehensive score is calculated according to the formula described in Model A. The results show that the AUC of the joint model is 1.000.

[0166] 3.2 Model B: Differentiating ENKTL from diffuse large B-cell lymphoma (DLBCL) Figure 5 );

[0167] Input metrics: H-scores of ADAMTS4, COL6A3, or MFAP2.

[0168] Construction logic: Utilizing the specific high expression characteristics of three proteins in ENKTL. The combined model of the three proteins achieved an AUC value of 0.991. Among them, ADAMTS4 contributed the most to the single-parameter model, but in the multi-parameter model, the addition of COL6A3 and MFAP2 further enhanced the confidence in interpreting borderline cases.

[0169] Quantitative diagnostic formulas and cutoff values:

[0170] The combined diagnostic index Y = -1.058 + 0.073 × H A +0.020×H C -0.058×H M .

[0171] Calculate the predicted probability P = 1 / (1+e) -Y ).

[0172] Equivalent linear diagnostic score: S B =72.66×H A +19.96×H C -58.47×H M .

[0173] Interpretation criteria: The optimal cutoff value for the model is P = 0.769 (equivalent to the linear score S). B >2262.7). When the threshold is exceeded, it indicates a high risk of being identified as ENKTL positive.

[0174] To verify the actual efficacy of the diagnostic model described in this invention in clinical samples, the following is provided: Figure 5 A detailed list of corresponding sample data is provided. This list includes the raw immunohistochemical (IHC) score (H-score) for each sample, the linear diagnostic score (Score) calculated according to the formula in Model B, the probability of prediction, and the final interpretation result. Data shows that the model successfully excluded all 17 DLBCL samples (100% specificity). In the ENKTL group, only one case (P41) was judged negative due to low ADAMTS4 expression, resulting in an extremely high overall accuracy (29 / 30), demonstrating its clinical value in the differentiation of lymphoma subtypes.

[0175]

[0176]

[0177]

[0178] Figure 5 Note: This figure shows the diagnostic efficacy analysis results based on 30 clinical tissue samples (including 13 pathologically confirmed intestinal NK / T-cell lymphomas [ENKTL] as the positive group and 17 diffuse large B-cell lymphomas [DLBCL] as the negative control group). The horizontal axis represents the false positive rate (1-specificity), and the vertical axis represents the true positive rate (sensitivity).

[0179] Figure 5 A in the figure shows the diagnostic efficacy of the single biomarker ADAMTS4. As the indicator that contributes the most to this model, it is specifically highly expressed in ENKTL, but extremely lowly expressed in DLBCL, demonstrating significant discriminative ability.

[0180] Figure 5 B in the figure shows the diagnostic efficacy of the single biomarker COL6A3. While there is expression overlap in some DLBCL cases, it effectively aids in interpretation in multi-parameter models.

[0181] Figure 5 C in the figure shows the diagnostic efficacy of the single biomarker MFAP2. As a component of the interstitial matrix, it provides important supplementary information for combined diagnosis.

[0182] Figure 5 D in the figure represents the performance of the three-marker joint model constructed in this invention. The comprehensive score was calculated according to the formula described in Model B. The results show that the AUC of the joint model is as high as 0.991, indicating that the combination can distinguish between ENKTL and DLBCL with extremely high accuracy.

[0183] 3.3 Model C: Differentiating ENKTL from Ulcerative Colitis (UC) Figure 6 );

[0184] Input metrics: H-scores (denoted as H) of both COL6A3 and MFAP2 C H M ).

[0185] Construction Logic: Addressing the challenge of morphological similarities between ENKTL and UC, the most significantly different interstitial combinations were selected. The study found that COL6A3 and MFAP2 deposition amounts in the ENKTL matrix were significantly higher than in the inflammatory interstitial matrix of UC. Furthermore, ROC analysis of all combinations showed that the COL6A3 + MFAP2 combination had the highest AUC value of 0.729.

[0186] Quantitative diagnostic formulas and cutoff values:

[0187] The combined diagnostic index Y = -1.186 + 0.003 × H C +0.002×HM .

[0188] Calculate the predicted probability P = 1 / (1+e) -Y ).

[0189] Equivalent linear diagnostic score: S C =2.99×H C +2.36×H M .

[0190] Interpretation criteria: The optimal cutoff value for the model is P = 0.503 (equivalent to the linear score S). C >1197.8). When the overall score of the sample is greater than this cutoff value, it is interpreted as malignant lymphoma (ENKTL).

[0191] To verify the actual efficacy of the diagnostic model described in this invention in clinical samples, the following is provided: Figure 6 A detailed list of corresponding sample data is provided. This list includes the raw immunohistochemistry (IHC) score (H-score) for each sample, the linear diagnostic score (Score) calculated according to the formula in Model C, the probability of prediction, and the final interpretation result. Data shows that the model did not produce any false positives in all 19 UC samples (100% specificity). Although some ENKTL samples had lower scores, the model demonstrated extremely high reliability in excluding benign inflammatory interference, making it suitable for clinical auxiliary screening.

[0192]

[0193]

[0194]

[0195] This model demonstrates that by simply detecting changes in the matrix microenvironment and assigning specific quantitative weights, it is possible to differentiate between malignant lymphomas and benign ulcers to a certain extent, providing a new dimension for pathological diagnosis that is independent of immunophenotype.

[0196] Figure 6 Note: This figure shows the diagnostic efficacy analysis results based on 32 clinical tissue samples (including 13 pathologically confirmed intestinal NK / T-cell lymphoma [ENKTL] as the positive group and 19 cases of ulcerative colitis [UC] as the negative control group). The horizontal axis represents the false positive rate (1-specificity), and the vertical axis represents the true positive rate (sensitivity).

[0197] Figure 6 A in the figure shows the diagnostic efficacy of the single biomarker COL6A3. This indicator reflects the difference in the degree of remodeling between the tumor stroma and the inflammatory stroma.

[0198] Figure 6 B in the figure shows the diagnostic efficacy of the single biomarker ADAMTS4. The results show that its ROC curve is almost diagonal overall, and the AUC is low, suggesting that its ability to distinguish between ENKTL and UC is limited, its diagnostic efficacy is relatively weak, and it may lack sufficient specificity.

[0199] Figure 6 C in the figure shows the diagnostic efficacy of the single biomarker MFAP2. Its deposition in the ENKTL matrix is ​​significantly higher than that in the inflammatory interstitium of UC, demonstrating clear auxiliary diagnostic value.

[0200] Figure 6 D in the figure represents the performance of the dual-marker joint model constructed in this invention. This model eliminates the non-specific ADAMTS4 and calculates the comprehensive score according to the formula described in model C. The results show that the AUC of the joint model is 0.729, and it can effectively eliminate UC interference at the specific cutoff value, indicating a high risk of malignancy.

Claims

1. A protein biomarker for the differential diagnosis of extranodal natural killer / T-cell lymphoma of the intestine, characterized in that, It is composed of COL6A3 protein and MFAP2 protein; or it is composed of ADAMTS4 protein, COL6A3 protein and MFAP2 protein.

2. The use of reagents for detecting COL6A3 and MFAP2 proteins; or reagents for detecting ADAMTS4, COL6A3, and MFAP2 proteins in the preparation of diagnostic reagents or kits, wherein the reagents or kits are used to diagnose intestinal natural killer / T-cell lymphoma or to assess its risk.

3. The use as described in claim 2, characterized in that, The reagent is an antibody that specifically binds to the protein, which is either COL6A3 protein and MFAP2 protein, or ADAMTS4 protein, COL6A3 protein and MFAP2 protein.

4. A kit for diagnosing extranodal natural killer / T-cell lymphoma or assessing the risk of its occurrence, characterized by, The kit includes detection reagents for detecting COL6A3 and MFAP2 proteins; or includes detection reagents for detecting ADAMTS4, COL6A3, and MFAP2 proteins; wherein the detection reagents for detecting COL6A3 and MFAP2 proteins include a combination of antibodies against COL6A3 and MFAP2 proteins; and the detection reagents for detecting ADAMTS4, COL6A3, and MFAP2 proteins include a combination of antibodies against ADAMTS4, COL6A3, and MFAP2 proteins.

5. The kit of claim 4, wherein The kit is used for detection using immunohistochemistry.

6. The kit of claim 4, wherein The kit also describes the following detection method, including the following steps: (a) Provide a test sample; (b) Detect the expression levels of COL6A3 and MFAP2 proteins in the test sample; or the expression levels of ADAMTS4, COL6A3, and MFAP2 proteins; denoted as Y1; (c) Compare the expression level Y1 with the control reference value Y0; When the expression level Y1 of the protein is higher than the control reference value Y0, it indicates that the subject has a high risk of developing intestinal extranodal natural killer / T-cell lymphoma or is diagnosed with intestinal extranodal natural killer / T-cell lymphoma.

7. The kit of claim 6, wherein The test sample is an ex vivo tissue sample.

8. The kit of claim 7, wherein The test samples are intestinal biopsy tissue or surgically removed tissue.

9. An auxiliary diagnostic system for extranodal natural killer / T-cell lymphoma, characterized in that, The system includes: (a) An input module, which is used to input the expression levels of COL6A3 and MFAP2 proteins of a certain object; or the expression levels of ADAMTS4, COL6A3, and MFAP2 proteins; (b) Processing module: The processing module compares the input antibody level Y1 with the control reference value Y0; (c) Output module, the output module is used to output the judgment result. When the input expression level data is higher than the preset threshold, it indicates that the subject has a high risk of extranodal natural killer cell / T cell lymphoma or is diagnosed with intestinal extranodal natural killer cell / T cell lymphoma; otherwise, it indicates that the subject has a low risk of extranodal natural killer cell / T cell lymphoma.

Images