Application of cfRNA detection reagent in preparation of AS early diagnosis product

Abstract

The invention relates to the technical field of molecular diagnosis and artificial intelligence, in particular to application of a cfRNA detection reagent in preparation of an AS early diagnosis product. The cfRNA is selected from the group consisting of AGAP5, SERPINF1 and LINC01630; the AS is ankylosing spondylitis. According to the present invention, the plasma cfRNA of the AS early stage patient, the healthy population and other spinal joint disease patients is subjected to the deep sequencing to screen the AS specific cfRNA marker, and the diagnosis model is constructed in combination with the machine learning algorithm so as to achieve the early stage precise diagnosis of the AS, and provide the reliable diagnosis tool for the clinic.

Description

TECHNICAL FIELD

[0001] The application relates to the field of molecular diagnosis and artificial intelligence, and particularly relates to application of a cfRNA detection reagent in preparation of an AS early diagnosis product. BACKGROUND

[0002] Ankylosing spondylitis, abbreviated as AS, is a chronic autoimmune disease mainly characterized by inflammation of sacroiliac joints and spinal attachment points, and is more common in young adults. The disease has an insidious onset, and early symptoms are mainly back pain and morning stiffness, which are easily confused with common diseases such as lumbar muscle strain and lumbar disc herniation, leading to delayed diagnosis. According to statistics, the average interval from the onset of symptoms to the definite diagnosis of AS patients is as long as 5-10 years, and the early stage of the disease is the key window period for intervention and treatment. Delayed diagnosis will lead to irreversible damage such as spinal rigidity and joint deformity, and seriously affect the quality of life of patients. At present, the clinical diagnosis of AS mainly relies on the New York classification criteria for AS revised in 1984, which needs to be combined with clinical symptoms, imaging examination and laboratory indexes. Imaging examinations such as sacroiliac joint X-ray and MRI, and laboratory indexes such as human leukocyte antigen B27, HLA-B27. However, the sensitivity of X-ray examination for early sacroiliac joint lesions is low, and MRI is more sensitive but expensive and has poor popularity; the positive rate of HLA-B27 is about 6%-8% in the general population and about 90% in AS patients, but there are certain false positives and false negatives, which cannot be used as a diagnostic basis alone. Therefore, it is urgent to develop a non-invasive, high-sensitivity and high-specificity early diagnosis technology to make up for the shortcomings of existing diagnostic methods.

[0003] Plasma free RNA, abbreviated as cfRNA, is extracellular RNA existing in peripheral blood plasma, derived from apoptosis, necrosis or active secretion, and can reflect the gene expression state of the body's tissues and organs. With the development of high-throughput sequencing technology, cfRNA sequencing can comprehensively capture the expression profile information of RNA in plasma, and has the advantages of non-invasiveness, easy sample acquisition and dynamic monitoring. In recent years, cfRNA has shown good application prospects in the diagnosis of tumors, autoimmune diseases and other diseases, but there is no related report on cfRNA as an AS early diagnosis marker. SUMMARY

[0004] The application aims to provide application of a cfRNA detection reagent in preparation of an AS early diagnosis product. The application screens AS-specific cfRNA markers by deep sequencing of plasma cfRNA of AS early patients, healthy people and other patients with spinal joint diseases, and constructs a diagnosis model by combining a machine learning algorithm, so that early and accurate diagnosis of AS can be realized, and a reliable diagnosis tool can be provided for clinical use.

[0005] To achieve the above object, the application specifically adopts the following technical scheme: The application of the detection reagent for cfRNA in the preparation of an AS early diagnosis product, wherein the cfRNA is AGAP5, SERPINF1 and LINC01630; The AS is ankylosing spondylitis.

[0006] Further, the detection reagent comprises primers for quantitatively detecting the AGAP5, SERPINF1 and LINC01630.

[0007] Further, the primer sequences for detecting the AGAP5, SERPINF1 and LINC01630 are shown in SEQ ID NO. 1-6.

[0008] Further, the product is a kit.

[0009] Further, the kit further comprises TB Green Mix and enzyme-free water.

[0010] Further, the product is a diagnosis system, and the diagnosis system comprises: a detection device and a detection reagent, which are used for detecting the relative expression amounts of AGAP5, SERPINF1 and LINC01630 and the erythrocyte sedimentation rate of a subject; a comparison device, which is used for receiving the relative expression amounts of AGAP5, SERPINF1 and LINC01630 and the erythrocyte sedimentation rate output by the detection device, analyzing and comparing the detection results of the relative expression amounts of AGAP5, SERPINF1 and LINC01630 and the erythrocyte sedimentation rate of the subject with known grouping information, and judging whether the subject has ankylosing spondylitis.

[0011] Further, the detection device detects the relative expression amounts of AGAP5, SERPINF1 and LINC01630 through real-time fluorescent quantitative PCR.

[0012] Further, the known grouping information comprises the relative expression amounts of AGAP5, SERPINF1 and LINC01630 and the erythrocyte sedimentation rate of ankylosing spondylitis patients, healthy people and other patients with spinal joint diseases, and the other patients with spinal joint diseases are patients with psoriatic arthritis, reactive arthritis and inflammatory bowel disease arthritis.

[0013] Further, the analysis and comparison are based on a known grouping information data set, and a machine learning model is constructed with AGAP5, SERPINF1, LINC01630 relative expression and erythrocyte sedimentation rate as characteristic variables, AGAP5, SERPINF1, LINC01630 relative expression and erythrocyte sedimentation rate of the subject are substituted into the constructed machine learning model, the diagnostic probability of ankylosing spondylitis of the subject is calculated, and whether the subject is ankylosing spondylitis is judged according to the diagnostic probability of ankylosing spondylitis.

[0014] Further, the machine learning model is a logistic regression model, and the diagnostic threshold is set to 0.5, if the diagnostic probability of the subject is greater than or equal to 0.5, it is determined that the subject is an early positive of ankylosing spondylitis, and if the diagnostic probability is less than 0.5, it is determined that the subject is an early negative of ankylosing spondylitis.

[0015] Beneficial effects: The application can effectively distinguish early AS patients from healthy people and other patients with spinal joint diseases by screening AS-specific cfRNA markers through high-throughput sequencing combined with machine learning algorithms, and constructing an LR diagnostic model with an AUC of 0.921, thereby avoiding misdiagnosis and missed diagnosis.

[0016] The application can diagnose early AS patients with a disease course of less than 2 years before the patient has obvious spinal deformity, provide a time window for early clinical intervention, and improve the prognosis of the patient. The differential genes screened by the application can not only be used as diagnostic markers, but also provide clues for the study of the pathogenesis of ankylosing spondylitis and the discovery of therapeutic targets. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 : Volcano plot of differentially expressed cfRNA of normal people and ankylosing spondylitis patients.

[0018] Figure 2 : Volcano plot of differentially expressed cfRNA of ankylosing spondylitis and other patients with spinal joint diseases.

[0019] Figure 3 : Venn diagram of AS-specific candidate cfRNA markers.

[0020] Figure 4 : ROC curve of AGAP5.

[0021] Figure 5 : ROC curve of CSPG4BP.

[0022] Figure 6 : ROC curve of CSPP1.

[0023] Figure 7 : ROC curve of CTSK.

[0024] Figure 8 ROC curve of CUTALP.

[0025] Figure 9 ROC curve of DCAF12L2.

[0026] Figure 10 ROC curve of GNPTAB.

[0027] Figure 11 ROC curve of LINC01485.

[0028] Figure 12 ROC curve of LINC01630.

[0029] Figure 13 ROC curve of NOL10.

[0030] Figure 14 ROC curve of SERPINF1.

[0031] Figure 15 LASSO coefficient distribution diagram.

[0032] Figure 16 LASSO regression variable shrinkage diagram. DETAILED DESCRIPTION

[0033] The present application will be described in detail below in conjunction with the accompanying drawings and specific examples, but should not be understood as limiting the present application. If not specifically stated, the technical means used in the following examples are conventional means well known to those skilled in the art, and the materials, reagents, etc. used in the following examples, if not specifically stated, can be obtained from commercial channels.

[0034] Example 1: Screening of AS-specific cfRNA markers.

[0035] Step 1: Study subject recruitment and sample collection.

[0036] The study subjects of the present application were recruited from January 2024 to December 2024 at the Department of Rheumatology and Immunology and the Health Management Department of the Second Affiliated Hospital of the Air Force Military Medical University. All study subjects signed the informed consent form, and the project has been approved by the hospital ethics committee.

[0037] 1.1 Inclusion and exclusion criteria.

[0038] (1) AS early case group, n=40, 34 males and 6 females.

[0039] 1) Age 18-45 years old, high-risk age group for AS, average age 32.5±6.8 years old, disease duration 1.2±0.5 years.

[0040] 2) AS classification criteria recommended by ASAS working group of international assessment in 2010, and disease duration ≤2 years, defined as early stage.

[0041] 3) No biological agent treatment such as TNF inhibitor, IL-17 inhibitor.

[0042] 4) Signed informed consent, and complete clinical data, including symptoms, signs, HLA-B27 test results and sacroiliac joint imaging report.

[0043] (2) Healthy control group, n=40, 20 males and 20 females.

[0044] 1) Age 18-45 years old, average age 32.5±6.8 years old, disease duration 1.2±0.5 years, matched with gender and age in case group.

[0045] 2) No low back pain, morning stiffness and other symptoms, and no history of autoimmune disease.

[0046] 3) HLA-B27 test negative.

[0047] 4) Signed informed consent, and no recent history of infection, trauma or surgery.

[0048] (3) Other spondylarthritis control group, n=40.

[0049] 1) Age 18-45 years old, 20 patients with psoriatic arthritis, 10 patients with reactive arthritis, 10 patients with inflammatory bowel arthritis, average age 43.2±6.5 years old.

[0050] 2) Disease duration ≤2 years, no biological agent treatment.

[0051] 3) Signed informed consent, complete clinical data.

[0052] 1.2 Plasma sample collection and processing.

[0053] (1) All study subjects were collected 5mL of venous blood in the morning on an empty stomach, using EDTA anticoagulant tube, gently inverted 5-6 times to prevent blood clotting; (2) Within 2 hours after collection, centrifuge at 3000r / min for 10 minutes at 4℃ to separate the upper plasma; (3) Take 2mL of plasma and transfer to a RNAase-free centrifuge tube, centrifuge again at 12000r / min for 15 minutes at 4℃ to remove residual cell debris; (4) The centrifuged plasma was aliquoted into RNase-free cryotubes and immediately stored at -80°C in an ultra-low temperature freezer to avoid repeated freezing and thawing.

[0054] Step 2: Extraction and quality detection of free RNA in plasma.

[0055] 2.1 Extraction of cfRNA.

[0056] From the above samples, 30 cases of early AS patients, 30 cases of healthy people, and 30 cases of other spinal joint diseases were randomly selected as sequencing sample sets. Commercial cfRNA extraction kit Qiagen miRNeasy Serum / Plasma Kit was used, and the specific steps were as follows: (1) Take out the plasma sample stored at -80°C, and slowly thaw on ice.

[0057] (2) Take 200 μL of thawed plasma, add 1000 μL of lysis buffer containing β-mercaptoethanol, and vortex for 1 minute.

[0058] (3) Add 300 μL of anhydrous ethanol, and vortex again.

[0059] (4) Transfer the mixture to the adsorption column, centrifuge at 8000 r / min for 15 seconds at room temperature, and discard the effluent.

[0060] (5) Add 700 μL of washing buffer 1 and 500 μL of washing buffer 2 in turn, and centrifuge at 8000 r / min for 15 seconds at room temperature, respectively, and discard the effluent.

[0061] (6) Transfer the adsorption column to a new RNase-free centrifuge tube, add 30 μL of RNase-free water, and centrifuge at 12000 r / min for 2 minutes after standing at room temperature for 2 minutes to elute cfRNA.

[0062] (7) The eluted cfRNA was immediately stored at -80°C, or directly used for subsequent experiments.

[0063] 2.2 Quality detection of cfRNA.

[0064] (1) Concentration and purity detection: The concentration and purity of cfRNA were detected using Nanodrop 2000 spectrophotometer, the concentration required A260 / A280 ratio 1.8-2, and the purity required A260 / A230 ratio ≥2.0.

[0065] (2) Integrity detection: Agilent 2100 Bioanalyzer was used to detect the integrity of cfRNA using RNA 6000 Pico Kit, and RNA Integrity Number (RIN) was used to evaluate the integrity of cfRNA. The RIN value was required to be greater than or equal to 6.0.

[0066] (3) Contamination detection: Real-time fluorescent quantitative PCR was used to detect genomic DNA contamination. GAPDH gene was used as an internal reference gene. If the Ct value was greater than or equal to 35, it was determined that there was no significant genomic DNA contamination, and the sample could be used for subsequent sequencing.

[0067] Nanodrop detection showed that the concentration of cfRNA was 12-25 ng / μL, A260 / A280=1.9-2.0, and A260 / A230=2.1-2.3. Agilent 2100 detection showed that RIN=6.2-7.5, and there was no significant genomic DNA contamination.

[0068] Step 3: Plasma free RNA sequencing and differential expression analysis.

[0069] 3.1 Library construction and high-throughput sequencing.

[0070] (1) Take the qualified cfRNA with a total amount of greater than or equal to 100 ng, and use NEBNext Ultra II Directional RNA Library Prep Kit to construct cDNA library. The specific steps include RNA fragmentation, reverse transcription to synthesize cDNA, end repair, A tailing, adapter ligation and PCR amplification.

[0071] (2) After the library construction, Agilent 2100 Bioanalyzer and Qubit 4.0 fluorometer were used to detect the fragment size and concentration of the library. The target fragment size was 200-300 bp, and the concentration was greater than or equal to 10 nM.

[0072] (3) Illumina NovaSeq 6000 sequencing platform was used for high-throughput sequencing in double-end 150 bp mode. The sequencing data of each sample was greater than or equal to 10 Gb, which ensured that the sequencing depth met the subsequent analysis requirements.

[0073] The sequencing data of each sample was 10-12 Gb, and Q30 was greater than or equal to 90%. 3.2 Sequencing data preprocessing.

[0074] (1) Raw data filtering: FastQC software was used to evaluate the quality of the raw sequencing data Fastq file, and reads containing adapter sequences, low-quality base Q value <20 and N content >5% were removed.

[0075] (2) Alignment and quantification: The filtered clean reads were aligned to the human reference genome GRCh38.p13 using HISAT2 software for alignment and STAR software for correction; the FeatureCounts software was used for gene expression quantification to obtain the gene expression matrix of each sample, expressed in reads count.

[0076] (3) Data normalization: The gene expression matrix was normalized by the Trimmed Mean of M-values method to correct the sequencing depth difference between samples, and the normalized gene expression value was obtained.

[0077] After preprocessing, the standardized gene expression matrix was obtained, and the expression of 10027 genes was detected.

[0078] 3.3 Differential expression cfRNA screening.

[0079] (1) DESeq2 software was used for differential expression analysis of AS early case group and healthy control group, AS early case group and other spondyloarthritis control group.

[0080] (2) Set the differential expression screening standard: |log2 Fold Change (FC)|>1.5, and adjusted P value<0.05.

[0081] (3) The differential expression cfRNA of AS early case group compared with healthy control group and other spondyloarthritis control group, including mRNA, lncRNA and circRNA, was obtained respectively, and the intersection of the two was taken as the AS-specific candidate cfRNA marker.

[0082] Differential expression analysis showed that there were 154 differential expression cfRNAs in the AS early case group compared with the healthy control group, and 174 differential expression cfRNAs compared with the other spondyloarthritis control group, as shown in Figure 1 and Figure 2 The intersection obtained 11 candidate cfRNA markers, as shown in Figure 3 The gene sequence numbers of the 11 candidate cfRNA markers are shown in Table 1, and the primers of the 11 candidate cfRNA markers are shown in Table 2.

[0083] Table 1: Specific information of 11 candidate cfRNA markers.

[0084] Table 2: Primers of 11 candidate cfRNA markers.

[0085] Example 2: Validation of AS-specific cfRNA markers.

[0086] Step 1: Validation sample set construction.

[0087] Ten independent patients with early-stage AS, ten healthy individuals, and ten patients with other spondyloarthritis were selected as the validation sample set. The sample collection and processing methods were the same as in Example 1.

[0088] Step 2: Real-time quantitative PCR verification.

[0089] 2.1 For the 11 candidate cfRNA biomarkers screened in Example 1, specific primers were designed. The primers were synthesized by Sangon Biotech Co., Ltd., with GAPDH as the internal reference gene.

[0090] 2.2 The cfRNA was reverse transcribed into cDNA using the PrimeScript RT Master Mix kit.

[0091] 2.3 qPCR was performed using the TB Green Premix Ex Taq II kit. The reaction mixture consisted of 20 μL of cDNA, 0.8 μL of each primer, 10 μL of TB Green Mix, and 6.4 μL of enzyme-free water. The reaction conditions were: 95℃ pre-denaturation for 30 seconds, 95℃ denaturation for 5 seconds, and 60℃ annealing extension for 30 seconds, for a total of 40 cycles. 2.4 The relative expression levels of candidate cfRNA biomarkers were calculated using the 2^(-ΔΔCt) method. Receiver operating characteristic (ROC) curve analysis was used to screen cfRNAs with an area under the curve (AUC) > 0.80 as the final seven AS-specific biomarkers: AGAP5, LINC01485, SERPINF1, LINC01630, CSPG4BP, CUTALP, and GNPTAB. Figures 4 to 14 As shown.

[0092] Example 3: Construction and validation of an early diagnosis model for AS.

[0093] Step 1: Data preparation and feature selection.

[0094] 1.1 Using the relative expression levels of the seven AS-specific cfRNA markers verified in Example 2 as feature variables, a multi-dimensional feature matrix was constructed in combination with clinical indicators, including erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP).

[0095] 1.2 Combine the sequencing sample set and the validation sample set, n = 120, and divide them into a training set and a validation set at a ratio of 7:3, with 84 cases in the training set and 36 cases in the validation set. Use stratified random sampling to ensure that the proportions of early-stage AS cases, healthy people and disease controls in the two groups are consistent.

[0096] 1.3 Use LASSO regression to screen the characteristic variables, remove redundant variables, and retain three markers AGAP5, SERPINF1 and LINC0163 and ESR values as characteristic variables that significantly contribute to diagnosis, as shown in Figure 15 and Figure 16 .

[0097] Example 4: Construction and performance verification of the diagnostic model.

[0098] Step 1: Based on the screened characteristic variables, three machine learning models are constructed respectively: 1.1 Logistic regression model, abbreviated as LR model: Use the R language glmnet package, set family = binomial (link = "logit"), and use 5-fold cross-validation to optimize the model regularization parameter λ.

[0099] 1.2 Random forest model, abbreviated as RF model: Use the R language randomForest package, set the number of decision trees ntree = 1000, and optimize the feature subset size mtry = 4 through grid search.

[0100] 1.3 Support vector machine model, abbreviated as SVM model: Use the R language e1071 package, select the radial basis kernel function kernel = "radial", and optimize the penalty parameter C = 10 and the kernel function parameter gamma = 0.1 through grid search.

[0101] Step 2: Model performance verification and comparison.

[0102] 2.1 Performance evaluation index: Use AUC, sensitivity, specificity, accuracy, F1 score and Matthew correlation coefficient MCC as the model performance evaluation index.

[0103] 2.2 Internal verification: Evaluate the performance of the three models in the validation set, repeat the verification 20 times for each model, and take the average value.

[0104] 2.3 Model comparison: Use Z-test to compare the AUC values between models, and use net reclassification improvement index NRI to compare the classification ability of the models.

[0105] 2.4 Optimal model determination: As shown in Table 3, the AUC of LR model in the validation set was 0.921, 95% CI: 0.795-0.934, accuracy = 0.838, sensitivity = 0.712, specificity = 0.845, F1 score = 0.631, and MCC = 0.782, which were significantly better than the AUC of 0.843 of RF model and the AUC of 0.806 of SVM model, thus the LR model was determined as the final early diagnosis model of AS.

[0106] Table 3: Prediction performance of each model.

[0107] Example 4: Clinical application of the early diagnosis method of AS.

[0108] Step 1: Sample detection.

[0109] The plasma samples of 20 patients suspected of AS were collected, and the relative expression amounts of the finally screened 3 cfRNA markers were detected by qPCR according to the method of Example 1, and the ESR value was also detected.

[0110] Step 2: Model prediction.

[0111] The detected cfRNA relative expression amounts and ESR value were substituted into the constructed LR diagnosis model to calculate the AS diagnosis probability of the patient.

[0112] Step 3: Diagnosis determination.

[0113] The diagnosis threshold was set to 0.5, if the diagnosis probability of the patient was ≥ 0.5, it was determined as early AS positive, and if the diagnosis probability was < 0.5, it was determined as early AS negative.

[0114] Step 4: Result interpretation.

[0115] Combined with the clinical symptoms and imaging examination results of the patient, the early diagnosis of AS was provided for the clinician to assist in developing a treatment plan.

[0116] The clinical application test was performed on 20 patients suspected of AS, and the diagnosis coincidence rate of the LR diagnosis model was 89%, the missed diagnosis rate was 10%, and the misdiagnosis rate was 1%, which proved that the model had good clinical application value.

[0117] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to these embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all changes and modifications falling within the scope of the present application.

Claims

1. The application of cfRNA detection reagents in the preparation of early diagnostic products for AS, characterized in that, The cfRNAs were AGAP5, SERPINF1, and LINC01630; The AS mentioned is ankylosing spondylitis.

2. The application of the cfRNA detection reagent according to claim 1 in the preparation of early AS diagnostic products, characterized in that, The detection reagent includes primers for the quantitative detection of AGAP5, SERPINF1, and LINC01630.

3. The application of the cfRNA detection reagent according to claim 2 in the preparation of early AS diagnostic products, characterized in that, The primer sequences for detecting AGAP5, SERPINF1, and LINC01630 are shown in SEQ ID NO.1~6 respectively.

4. The application of the cfRNA detection reagent according to claim 3 in the preparation of early AS diagnostic products, characterized in that, The product in question is a reagent kit.

5. The application of the cfRNA detection reagent according to claim 4 in the preparation of early AS diagnostic products, characterized in that, The kit also includes TB Green Mix and enzyme-free water.

6. The application of the cfRNA detection reagent according to claim 2 in the preparation of early AS diagnostic products, characterized in that, The product is a diagnostic system, which includes: The detection device and reagents were used to detect the relative expression levels of AGAP5, SERPINF1, and LINC01630, as well as the erythrocyte sedimentation rate (ESR) in the subjects. The comparison device is used to receive the relative expression levels of AGAP5, SERPINF1, and LINC01630 and the erythrocyte sedimentation rate (ESR) output by the detection device, analyze and compare the detection results of the relative expression levels of AGAP5, SERPINF1, and LINC01630 and the ESR with known grouping information to determine whether the subject has ankylosing spondylitis.

7. The application of the cfRNA detection reagent according to claim 6 in the preparation of early AS diagnostic products, characterized in that, The detection device described herein uses real-time quantitative PCR to detect the relative expression levels of AGAP5, SERPINF1, and LINC01630.

8. The application of the cfRNA detection reagent according to claim 6 in the preparation of early AS diagnostic products, characterized in that, The known grouping information includes the relative expression levels of AGAP5, SERPINF1, and LINC01630 and the erythrocyte sedimentation rate (ESR) values ​​of patients with ankylosing spondylitis, healthy individuals, and other spondyloarthritis patients. The other spondyloarthritis patients are patients with psoriatic arthritis, reactive arthritis, and inflammatory bowel disease arthritis.

9. The application of the cfRNA detection reagent according to claim 6 in the preparation of early AS diagnostic products, characterized in that, The analysis and comparison are based on a known group information dataset. A machine learning model is constructed using the relative expression levels of AGAP5, SERPINF1, and LINC01630, as well as the erythrocyte sedimentation rate (ESR) value, as feature variables. The relative expression levels of AGAP5, SERPINF1, and LINC01630, as well as the ESR value of the subjects, are substituted into the constructed machine learning model to calculate the probability of ankylosing spondylitis diagnosis for the subjects. The probability of ankylosing spondylitis diagnosis is used to determine whether the subjects have ankylosing spondylitis.

10. The application of the cfRNA detection reagent according to claim 9 in the preparation of early AS diagnostic products, characterized in that, The machine learning model is a logistic regression model, with a diagnostic threshold of 0.

5. If the diagnostic probability of a subject is ≥0.5, it is determined to be an early positive case of ankylosing spondylitis; if the diagnostic probability is <0.5, it is determined to be an early negative case of ankylosing spondylitis.

Images