Application of mRNA biomarkers in the early diagnosis of lung signet ring cell carcinoma and reagent kits

Abstract

This invention discloses the application of mRNA biomarkers in the early diagnosis of lung signet ring cell carcinoma and a kit. The biomarkers are a combination of C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10, and THBD. This invention uses FFPE cancer tissue from patients clinically diagnosed with primary lung signet ring cell carcinoma using existing detection methods as samples, and adjacent normal tissue as control samples. RNA sequencing technology is used to detect the mRNA expression profiles of cancer tissue and control tissue. Statistical and machine learning methods are applied to identify 17 differentially expressed mRNA biomarkers. Based on these 17 biomarkers, a risk scoring model with high sensitivity and specificity is constructed. The model has a sensitivity >75% and a specificity >89% for the diagnosis of early lung signet ring cell carcinoma.

Description

Technical Field

[0001] This invention belongs to the field of biological detection technology, specifically relating to the application of mRNA biomarkers in the early diagnosis of lung signet ring cell carcinoma and diagnostic kits. Background Technology

[0002] Lung signet ring cell carcinoma (LSRCC) is a rare and highly aggressive subtype of poorly differentiated adenocarcinoma. It is characterized by tumor cells containing abundant mucus, which pushes the nucleus to the cell edge, creating a "signet ring" appearance. Furthermore, the cancer cells are disorganized, appearing in sheets, islands, or diffusely distributed, and are often accompanied by significant pathological mitotic figures and necrotic foci. Its distinctive pathological morphology is frequently associated with high malignancy and poor prognosis. Signet ring cell carcinoma often presents with no obvious symptoms or signs in its early stages, making it difficult to detect through routine physical examinations, thus hindering early diagnosis based on clinical manifestations. According to statistics from the World Health Organization, due to its insidious nature and rapid progression, most patients with signet ring cell carcinoma are diagnosed at an advanced stage, resulting in a significantly lower 5-year survival rate than other types of cancer. Therefore, identifying highly sensitive and specific early diagnostic biomarkers can help achieve early detection and intervention, thereby significantly reducing mortality, medical resource consumption and the economic burden on patients, and improving their quality of life.

[0003] Currently, endoscopic examination combined with biopsy is the main clinical method for diagnosing lung signet ring cell carcinoma. However, its ability to identify early, small lesions is limited, and it suffers from false negatives and false positives. For example, in terms of cell morphology alone, cells from various benign and malignant lesions can exhibit a "signet ring" morphology, such as those found in gastric xanthoma and pseudomembranous colitis. Therefore, this needs to be taken seriously in differential diagnosis. Furthermore, existing hematological tumor markers such as CEA and CA19-9 do not have ideal sensitivity and specificity for diagnosing signet ring cell carcinoma. Although breakthroughs have been made in genomics, transcriptomics, and proteomics in recent years, research on biomarkers for the early diagnosis of lung signet ring cell carcinoma remains relatively lagging, and stable, highly sensitive, and highly specific biomarkers have not yet been discovered.

[0004] Currently, there are few reports on the use of molecular-level biomarkers in the early diagnosis of lung signet ring cell carcinoma. Two relatively relevant studies are: Study 1 (PMID: 39910169): Researchers used deep visual proteomics to discover ATR signal activation and upregulation of DNA repair proteins such as FANCD2 and RECQL5 in LSRCC, suggesting that replication stress could serve as a potential diagnostic marker. However, this study was based on only one patient, and the generalizability of its results needs further validation. Study 2 (PMID: 20022810) focused on gastric signet ring cell carcinoma. Researchers analyzed 353 gastric samples from two independent patient subgroups in Japan using microRNA microarrays, comparing microRNA expression patterns between non-tumor mucosa and cancer samples. In 160 paired non-tumor mucosa and cancer samples, 22 microRNAs were upregulated and 13 were downregulated in gastric cancer, with 292 (83%) samples correctly distinguished by this feature. However, this study mainly assessed the relationship between microRNA expression and gastric cancer progression and prognosis, without further exploring the potential of differentially expressed genes as diagnostic markers.

[0005] Currently, no suitable biomarkers for the early detection of lung signet ring cell carcinoma have been reported that simultaneously ensure sensitivity and high specificity. Developing a biomarker for the early diagnosis of lung signet ring cell carcinoma has become one of the urgent problems to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to provide an mRNA marker and kit for the early diagnosis of lung signet ring cell carcinoma.

[0007] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0008] This invention provides mRNA markers for the early diagnosis of lung signet ring cell carcinoma, wherein the mRNA markers are a combination of C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10 and THBD.

[0009] The present invention also provides the use of the detection reagent for the expression level of the mRNAs markers in the preparation of compositions or kits for evaluating, diagnosing or monitoring lung signet ring cell carcinoma.

[0010] On the other hand, the present invention provides a kit for the early diagnosis of lung signet ring cell carcinoma, the kit comprising reagents for detecting the expression levels of the mRNA markers.

[0011] As a specific implementation, the reagent obtains the expression values ​​of the mRNA markers in cancer tissues and control tissues respectively using RNA sequencing technology.

[0012] As a specific implementation scheme, the kit also includes a risk scoring model, the formula of which is as follows:

[0013]

[0014] X i P(X) represents the numerical value of the biomarker for sample i. i ) represents the probability of the category corresponding to sample i predicted by the diagnostic model, where 0 represents a healthy individual and 1 represents a patient with lung signet ring cell carcinoma; the sum of the probabilities of all categories equals 1, and the category with the highest probability is taken as the final prediction result of sample i.

[0015] As a specific implementation plan, the risk scoring model uses a model algorithm selected from Gini, Entr, Random Forest, Weighted Ensemble Model, and Linear Regression.

[0016] This invention also provides a method for diagnosing early-stage lung signet ring cell carcinoma using the aforementioned kit, comprising the following steps:

[0017] Step 1: Collect cancer tissue from patients with suspected lung signet ring cell carcinoma based on clinical diagnosis, and use RNA sequencing to obtain the expression levels of mRNA markers in the cancer tissue;

[0018] Step 2: Input the expression values ​​of the obtained mRNA markers into the risk scoring model formula to calculate the probability of the patient being predicted as healthy or as having lung signet ring cell carcinoma;

[0019] Step 3: Select the output result with the highest probability and give the predicted result of the patient's risk of developing lung signet ring cell carcinoma.

[0020] The English abbreviations used in this invention are explained below:

[0021] C1QA: Complement C1q A chain

[0022] CRISPLD2: A cysteine-rich secretory protein containing protein 2 within its LCCL domain.

[0023] ROBO4: Axonal guidance receptor 4

[0024] SPARCL1: Rich in cysteine-containing acidic secretory protein-like protein 1

[0025] SFTPB: Surfactant Protein B

[0026] LAMC3: Laminin subunit γ3

[0027] FN1: Fibronectin 1

[0028] EMP1: Epithelial membrane protein 1

[0029] CD68: Leukocyte differentiation antigen 68

[0030] THBS1: Platelet-Reactive Protein 1

[0031] PRG4: Proteoglycan 4

[0032] ICAM1: Intercellular adhesion molecule-1

[0033] PODXL: podocyte marker protein

[0034] KDM6B: Lysine demethylase 6B

[0035] ENPP2: Exonucleotide pyrophosphatase / phosphodiesterase 2

[0036] KLF10: Krüppel-like factor 10

[0037] THBD: Thrombomodulin

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1. This invention utilizes RNA sequencing technology to detect the mRNA expression profiles of cancer tissues and control tissues from patients with lung signet ring cell carcinoma. Statistical and machine learning methods were applied to identify 17 differentially expressed mRNA markers, specifically: C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10, and THBD.

[0040] 2. Based on the 17 biomarkers discovered, this invention constructs a risk scoring model with high sensitivity and specificity. The model has a sensitivity of >75% and a specificity of >89% for the diagnosis of early lung signet ring cell carcinoma. Attached Figure Description

[0041] Figure 1 This is a diagram of the immunohistochemical staining results in Example 1.

[0042] Figure 2 The ROC curve of the training queue for the Limit Random Tree model algorithm in Example 3 is shown.

[0043] Figure 3The ROC curve is the validation queue curve of the extreme random tree model algorithm in Example 3. Detailed Implementation

[0044] The technical solution of the present invention will be described in detail below with reference to the embodiments. Unless otherwise specified, all reagents and biological materials used below are commercial products.

[0045] Example 1: Screening of mRNA markers

[0046] (1) Study cohort and clinical information

[0047] A total of 22 tissue samples were included, comprising cancerous tissues from 12 patients with lung signet ring cell carcinoma and 10 adjacent normal tissues as healthy controls. Fifteen samples were randomly selected as the training cohort, including 8 lung signet ring cell carcinoma tissues and 7 adjacent normal tissues. The validation cohort consisted of 7 samples, including 4 cancerous tissues and 3 adjacent normal tissues. Specific grouping information is shown in Table 1.

[0048]

[0049] (2) FFPE sample preparation and sectioning

[0050] FFPE sample preparation consists of two parts: formaldehyde fixation and paraffin embedding. Appropriately sized tissue samples are placed in 10% neutral buffered formalin (also known as formaldehyde) fixative and fixed at room temperature for 18-24 hours. After fixation, the samples are removed and rinsed with running water for several minutes to remove residual fixative and any crystals that may have formed. After fixation, the tissue samples need to be dehydrated to ensure sufficient penetration of the embedding medium. A gradient dehydration process is used: the samples are immersed in 75% ethanol, 85% ethanol, 95% ethanol, anhydrous ethanol (I), and anhydrous ethanol (II) for 1 minute each. After dehydration, the samples are immersed twice in xylene for 1 minute each time, then placed in molten paraffin. The paraffin-impregnated tissue samples are placed in a mold, molten paraffin is poured in, and the paraffin is allowed to cool and solidify. In this way, the tissue samples are embedded in paraffin.

[0051] After the embedded tissue block has been fully cooled, fix it on the sample holder of the microtome. Adjust the slice thickness of the microtome to 4-5µm. Use a brush to gently remove the sliced ​​tissue from the blade and place it in warm water (40℃) to spread it out. After spreading, firmly attach the slice to the glass slide and bake it at 60℃ for 2 hours.

[0052] (3) Immunohistochemical detection

[0053] Sections were dewaxed with xylene and then sequentially immersed in anhydrous ethanol (I), anhydrous ethanol (II), 95% ethanol, 85% ethanol, 70% ethanol, and pure water for 5 min each time. Antigen retrieval was performed with 10 mM sodium citrate (pH 6.0), followed by treatment in 3% hydrogen peroxide aqueous solution at room temperature for 10 min, and blocking with 5% BSA at room temperature for 30 min. Primary antibody information: Napsin A (ab73021, abcam), CK7 (ab181598, abcam), and TTF-1 (ab76013, abcam). Antibodies were incubated at room temperature for 1 h, washed thoroughly with TBST, and incubated with rabbit secondary antibody (A0208, Beyotime) or mouse secondary antibody (A0216, Beyotime) at room temperature for 30 min. After washing four times with TBST for 10 min each time, the sections were mounted and observed under a microscope to determine the location and expression level of the antigen in cells or tissues. See also Figure 1 The image shows the results of immunohistochemical staining. The staining results show that Napsin A and CK7 exhibit distinct brownish-yellow staining. Specifically, diffuse brownish-yellow granular deposits are visible in the cytoplasm, while no obvious staining is observed in the intercellular matrix, indicating that Napsin A and CK7 are mainly expressed in the cytoplasm of the cells in this tissue. TTF1 shows distinct punctate brownish-yellow staining and is mainly expressed in the cell nucleus. Figure 1 The results showed that the patient's tumor cells were positive for alveolar epithelial markers Napsin A, CK7, and TTF1, strongly suggesting that the patient had lung adenocarcinoma.

[0054] (4) RNA extraction from FFPE samples

[0055] Following the product instructions, total RNA was extracted from FFPE samples using the ReadiMag FFPE RNA Kit (3DMed, Shanghai), and finally eluted with 60 µL of RNase-free water. RNA purity was assessed using NanoDrop, RNA concentration was quantified using Qubit, and mRNA fragmentation was detected using an Agilent 2100 Bioanalyzer and its accompanying RNA analysis kit (5067-1514, Agilent) (to assess DV200 values).

[0056] (5) mRNA expression level detection

[0057] Following the product instructions, ribosomal RNA was removed using the KAPA RiboErase HMR kit (KK8481, Kapa), and mRNA libraries were constructed using the KAPA RNA Hyper kit (KK8542, Kapa). 400 ng of total RNA was loaded onto each sample, followed by rRNA removal, DNase digestion, first- and second-strand synthesis, adapter ligation, and library amplification. PCR-enriched products were purified using Agencourt Ampure XP-PCR purification beads (A63881, Beckman), and the library DNA was finally washed away with 20 µL of nuclease-free water. DNA concentration was quantified using an Invitrogen Qubit 4.0 fluorometer and the accompanying Qubit dsDNA HS Assay Kit (Q32854, Thermofisher). The distribution of DNA fragments in the libraries was detected using an Agilent 4150 bioanalyzer and its accompanying chips and reagents, the High Sensitivity D1000 ScreenTape & Reagents (5067-5584 & 5067-5585, Agilent). Sequencing was performed using the Illumina NovaSeq platform with the PE150 sequencing strategy, yielding 12 GB of sequencing data per library.

[0058] (6) Sequencing data analysis workflow

[0059] Based on RNA sequencing technology, the expression levels of mRNA in cancerous tissues and adjacent normal tissues of patients with lung signet ring cell carcinoma were obtained. The analysis workflow of the sequencing data is as follows:

[0060] 1) Sequencing data alignment. After removing sequencing adapters from the RNA sequencing data using FASTP software (version: 0.23.4), the sequencing data was aligned to the human reference genome hg38 using STAR software (version: 2.7.11b) (genome download link: http: / / hgdownload.soe.ucsc.edu / goldenPath / hg38 / bigZips / ).

[0061] 2) Quantitative analysis of mRNA expression. The number of reads aligned to different mRNAs was quantitatively counted using RSEM software (version 1.3.3).

[0062] 3) mRNA annotation. mRNAs were annotated using the Gencode v43 database, retaining those annotated as known mRNAs for subsequent analysis.

[0063] 4) mRNA filtering. For the training cohort, mRNAs with at least 10 reads aligned to all training cohort samples are retained for subsequent analysis. For the validation cohort, mRNAs selected from the training cohort are retained for subsequent analysis.

[0064] 5) mRNA expression level normalization. The original mRNA expression levels of the training cohort samples were normalized using the M-value weighted truncated mean (TMM) method, and the same parameters were used to process the validation cohort samples.

[0065] (7) Discovery of biomarkers

[0066] Samples were grouped according to pathological examination results. Based on the expression levels of mRNAs in the training cohort, statistical and machine learning methods were used to discover mRNAs that could distinguish between patients with lung signet ring cell carcinoma and healthy controls as biomarkers. The process is as follows:

[0067] 1) Training cohort grouping. Based on the pathological test results of the samples, the samples in the training cohort were divided into two groups: healthy controls and lung signet ring cell carcinoma.

[0068] 2) Statistical methods were used to screen biomarkers. The statistical methods U-test and T-test were used to calculate the statistical significance of differences in expression of all mRNAs between the two groups. Statistical significance was assessed using the p-value, with a p-value ≤ 0.05 considered statistically significant. First, mRNAs with a median and mean expression values ​​greater than 5 in both groups were screened. Then, mRNAs meeting at least one of the following two screening criteria were retained for subsequent analysis: (a) according to the U-test, the absolute value of the logarithm of the difference in median expression (logFC median) > 1, and the U-test p-value ≤ 0.05; (b) according to the T-test, the absolute value of the logarithm of the difference in mean expression (logFC mean) > 1, and the T-test p-value ≤ 0.05.

[0069] 3) Machine Learning Methods for Biomarker Screening. To determine the final biomarkers used to construct the risk scoring model, a variety of machine learning algorithms were used for biomarker screening, including six algorithms: Linear Model, Random Forest, Extremely Randomized Trees, Neural Network, Gradient Boosting Machine (GBM), and Ensemble Methods. The main process is as follows: (a) Using the mRNAs screened in step 2) above as initial biomarkers, the baseline performance of the biomarkers was evaluated; (b) The feature importance of each biomarker in the model was evaluated, biomarkers with importance > 0 were retained, and the model was retrained; (c) The feature importance of the biomarkers in the new model was evaluated again, and step (b) was repeated until the feature importance of the biomarkers in each model was > 0. These biomarkers were used for subsequent analysis.

[0070] Example 2: Risk Scoring Model Construction

[0071] Using healthy individuals and lung signet ring cell carcinoma as classification and prediction targets, and employing the biomarkers discovered in Example 1, four machine learning algorithms were used: Extremely Randomized Trees, Linear Model, Random Forest, and Ensemble Methods. Different hyperparameters were preset for each algorithm, and diagnostic models using six different machine learning algorithms were trained on the entire training cohort sample data. The final risk scoring model formula is as follows:

[0072]

[0073] X i P(X) represents the numerical value of the biomarker for sample i. i The probability of sample i corresponding to the predicted category by the diagnostic model is denoted as , where 0 represents a healthy individual and 1 represents a patient with lung signet ring cell carcinoma. The sum of the probabilities of all categories equals 1, and the category with the highest probability is taken as the final prediction result for sample i. The trained model is saved on the hard drive as a file. When calling the model, inputting the sample marker expression value will yield the model's prediction result.

[0074] Example 3: Evaluation and Validation of the Performance of the Risk Scoring Model

[0075] In the training queue, 5-fold cross-validation was used to evaluate the model's classification performance and feature weights, while in the validation queue, the model's generalization performance was evaluated. Model evaluation and validation metrics included the area under the receiver operating characteristic curve (AUC, ranging from 0 to 1), accuracy (ranging from 0 to 1), positive predictive value (ranging from 0 to 1), and negative predictive value (ranging from 0 to 1). The model's specificity (ranging from 0 to 1) and sensitivity (ranging from 0 to 1) were also evaluated; higher values ​​indicated better model classification performance.

[0076] Table 2 shows the evaluation results of individual markers in the training and validation queues for the 17 markers: C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10, and THBD.

[0077]

[0078] The data in Table 2 indicate that the evaluation effect of a single marker is poor, and the generalization performance is also poor. A high AUC value in either the training or validation queue alone indicates poor generalization performance. Ideally, the AUC values ​​in the training and validation queues should be similar.

[0079] Seventeen biomarkers were combined, and the classification performance and feature weights of the model were evaluated using 5-fold cross-validation in the training cohort. Combined with pathological examination results, the model's formula and procedure were used to predict the risk of lung signet ring cell carcinoma for each patient. Model evaluation metrics included the area under the receiver operating characteristic curve (AUC, range 0–1), accuracy (range 0–1), positive predictive value (range 0–1), negative predictive value (range 0–1), sensitivity (range 0–1), and specificity (range 0–1). The evaluation results of the 17 biomarker combinations in the training cohort are shown in Table 3. The results indicate that in the training cohort, this risk scoring model has high AUC, accuracy, positive predictive value, negative predictive value, specificity, and sensitivity, demonstrating superior predictive performance. See also Figure 2 , which is the ROC curve of the training queue for the Limiting Random Tree Model algorithm.

[0080]

[0081] To validate the performance of the risk scoring model in predicting lung signet ring cell carcinoma, an independent cohort was selected as the validation cohort to evaluate the model's classification effectiveness and feature weights. Using pathological examination results as the true values, the model evaluation metrics included AUC (range 0–1), accuracy (range 0–1), positive predictive value (range 0–1), negative predictive value (range 0–1), sensitivity (range 0–1), and specificity (range 0–1). The evaluation results of the 17 biomarkers used in the validation cohort are shown in Table 4. The results indicate that in the validation cohort, this risk scoring model has high AUC, accuracy, positive predictive value, negative predictive value, specificity, and sensitivity, demonstrating superior predictive performance. See also Figure 3 , which is the ROC curve of the validation queue for the extreme random tree model algorithm.

[0082]

[0083] Example 4: Application of Risk Scoring Model

[0084] 1) Collect cancer tissue from patients with suspected lung signet ring cell carcinoma based on clinical diagnosis, and use RNA sequencing to obtain the expression levels of biomarkers in the cancer tissue;

[0085] 2) Input the obtained expression values ​​of C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10 and THBD into the trained model formula to calculate the probability of each patient being predicted as healthy or lung signet ring cell carcinoma;

[0086] 3) Select the output result with the highest probability and give the predicted risk of each patient developing lung signet ring cell carcinoma.

[0087] In summary, this invention utilizes RNA sequencing technology to detect the mRNA expression profiles of cancerous tissues from patients with lung signet ring cell carcinoma and healthy control tissues. Statistical and machine learning methods were applied to identify 17 differentially expressed mRNA biomarkers: C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10, and THBD. Based on these 17 biomarkers, a risk scoring model with high sensitivity and specificity was constructed. The model demonstrated a sensitivity >75% and a specificity >89% for the diagnosis of early-stage lung signet ring cell carcinoma.

[0088] The above are merely some preferred embodiments of the present invention, and the present invention is not limited to the contents of these embodiments. For those skilled in the art, various changes and modifications can be made within the scope of the present invention's technical solutions, and any such changes and modifications are within the protection scope of the present invention.

Claims

1. An mRNA marker for early diagnosis of lung signet ring cell carcinoma, characterized in that: The mRNA markers are a combination of C1QA, CRISPLD2, ROBO4, SPARCL1, SFTPB, LAMC3, FN1, EMP1, CD68, THBS1, PRG4, ICAM1, PODXL, KDM6B, ENPP2, KLF10, and THBD.

2. The use of the detection reagent for obtaining the expression level of the mRNA markers of claim 1 in the preparation of a composition or kit for diagnosing lung signet ring cell carcinoma.

3. A reagent kit for the early diagnosis of lung signet ring cell carcinoma, characterized in that: The kit includes reagents for detecting the expression level of the mRNA markers of claim 1.

4. The kit for early diagnosis of lung signet ring cell carcinoma according to claim 3, characterized in that: The reagents were used to obtain the expression values ​​of the mRNA markers in cancer tissues and control tissues, respectively, using RNA sequencing technology.

5. The kit for early diagnosis of lung signet ring cell carcinoma according to claim 3, characterized in that: The kit also includes a risk scoring model, the formula of which is as follows: ； The numerical value of the marker representing sample i Let be the probability of the category corresponding to sample i predicted by the diagnostic model, where 0 represents a healthy individual and 1 represents a patient with lung signet ring cell carcinoma; the sum of the probabilities of all categories is equal to 1, and the category with the highest probability is taken as the final prediction result for sample i.

6. The kit for early diagnosis of lung signet ring cell carcinoma according to claim 5, characterized in that: The risk scoring model uses algorithms selected from Gini, Entr, Random Forest, Weighted Ensemble Model, and Linear Regression.

Images