Periprosthetic joint infection biomarkers

Abstract

In vitro methods of diagnosing a periprosthetic joint infection and monitoring the treatment thereof, comprising the sequential steps of providing a synovial fluid sample from a joint, determining the expression level(s) of microRNA(s), and classifying the expression level(s) into one of the categories "septic" and "aseptic", by comparing the expression levels to the levels observed in a reference group.

Description

[0001] JB0001 P

[0002] -1-

[0003] PERIPROSTHETIC JOINT INFECTION BIOMARKERS

[0004] FIELD OF THE INVENTION

[0005] The present invention relates to the field of diagnosing periprosthetic joint infections.

[0006] BACKGROUND OF THE INVENTION

[0007] Periprosthetic joint infection (PJI) and aseptic loosening remain the most common reasons for the need for revision surgery after total joint replacement surgery (TJR) of the knee and hip. Especially PJIs are still a major challenge in revision arthroplasty with high morbidity and mortality. To successfully treat these patients, an early diagnosis of PJI is crucial, but the differentiation between aseptic loosening and septic PJI can be difficult. This is particularly important, as the treatment strategy differs in the total number of revision surgeries, antibiotic treatment, time of inpatient hospitalization as well as treatment costs.

[0008] The diagnosis of PJI is currently based on various clinical and laboratory parameters in serum and synovial fluid as well as on microbiological results (McNally et al., 2021 ; Parvizi et al., 2018). Especially detection of PJI related microorganisms and their antibiotic resistance patterns can be difficult. Recently, new pathogen specific diagnostic methods based on polymerase chain reaction (PCR) techniques or molecular technologies such as next-generation sequencing (NGS) have been developed (Huang et al., 2020; Indelli et al., 2021 ; Luftinger et al., 2021 ; Pascual et al., 2024). However, there is still a high number of culture negative PJIs described in the literature (Goh et al., 2022; Kalbian et al., 2020). Currently, there are only a few synovial fluid markers described in the literature with insufficient sensitivity and specificity to achieve clinical utility (Sharma et al., 2020; Yilmaz et al., 2023).

[0009] MicroRNAs (miRNAs), small endogenous non-coding RNAs regulating gene expression in human cells, have already drawn much attention as potential diagnostic and prognostic biomarkers in other medical fields.

[0010] WO201 5134551 A1 discloses a kit for diagnosing a neurological disorder, neurological condition, or a neural injury by determining the level of at least one miRNA in a sample of a patient and comparison with a healthy patient’s sample using a softwareclassification-algorithm. One disclosed miRNA is miR-338-5p. JB0001 P

[0011] -2-

[0012] WO20151 12382A1 discloses a method for detecting an infection or infectious agent associated disease condition in a subject based on biomarkers present in exosomes in a bodily fluid sample, wherein the biomarkers may be viral microRNAs.

[0013] WO2018136936A1 describes the use of small non-coding RNA molecules as markers for a disease state in a subject.

[0014] WO201 8076015A1 describes small non-coding RNAs as markers for a disease in a subject.

[0015] Omar et al. (2017) describe differences in gene transcription profiles in periprosthetic tissue samples from patients suffering from chronic periprosthetic hip infections and aseptic hip prosthesis loosening.

[0016] Paksoy et al. (2024) disclose a method to assess periprosthetic joint infections, which relies on the measurement of microRNA biomarkers circulating in the bloodstream. However, low-grade infections do not have the virulence to induce significant differences in blood-circulating microRNA expression, thus severely limiting the sensitivity of said method.

[0017] Deng etal. (2024) identified differentially expressed miRNAs in samples of plasma, bone and tissue from patients who suffer from osteolysis and aseptic loosening following total joint replacement. Analysis revealed that miR-1246 and miR-6089 could be relevant.

[0018] It was shown that miRNAs could be measured in synovial fluid using nextgeneration sequencing (NGS) technology and real-time quantitative PCR (RT-qPCR) (Khamina et al., 2022). Due to changes of expression in response to physiological stimuli and pathological processes (Gutmann et al., 2022), levels of miRNAs may differ in synovial fluid of patients undergoing septic and aseptic knee and hip revision TJR, making them possible biomarkers in the diagnosis of PJL

[0019] Therefore, sensitive and reliable diagnostic methods for identifying PJI are needed. Especially, the diagnostic accuracy and sensitivity of synovial fluid markers delivering early results to choose the right treatment strategy is of great interest.

[0020] SUMMARY OF THE INVENTION

[0021] It is the object of the present invention to provide improved methods to diagnose a periprosthetic joint infection or to monitor the treatment thereof, based on the analysis of microRNA expression levels from synovial fluid samples from a joint without the need of a standard microRNA. The object is solved by the subject matter of the present invention. JB0001 P

[0022] -3-

[0023] According to the invention, there is provided an in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of:

[0024] (a) determining the expression level(s) of

[0025] (i) 1 , 2, 3, 4, or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial fluid sample of said subject;

[0026] (ii) optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p; and

[0027] (b) comparing the expression level(s) of (a) with a predetermined reference range of expression levels of the same miRNA(s) from an aseptic reference group, wherein the expression level(s) of (a) are classified as category septic if the expression level(s) of (a) are different from the reference range, and wherein the expression level(s) are classified as category aseptic if the expression level(s) of (a) are within the reference range.

[0028] According to a further embodiment of the invention, there is provided an in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of:

[0029] (a) determining the expression level(s) of

[0030] (i) 1 , 2, 3, 4 or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial fluid sample of said subject;

[0031] (ii) optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p;

[0032] (b) determining the expression level of a standard miRNA having low expression stability and high average abundance;

[0033] (c) determining a ratio of the expression level of said standard miRNA of (b) and the miRNA(s) of (a);

[0034] (d) determining a ratio of the expression level of the same standard miRNA from an aseptic reference group and the expression level(s) of the same miRNA(s) from an aseptic reference group; and JB0001 P

[0035] -4-

[0036] (e) comparing said ratios, wherein the ratio of (c) is classified as category septic if it is different from ratio of (d), and wherein the ratio of (c) is classified as category aseptic if it is similar to (d).

[0037] According to a specific embodiment of the invention, the expression stability and average abundance is identified with a mathematical model of gene expression; specifically, the standard miRNA is selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b- 3p, miR-221 -3p, miR-25-3p, miR-21-5p, and miR-93-5p.

[0038] According to a specific embodiment of the invention, the synovial fluid sample is from a joint of a subject on which arthroplasty was performed.

[0039] According to a further embodiment of the invention, the synovial fluid sample is from a subject with joint pain.

[0040] According to a specific embodiment of the invention the classification is performed using a classification model.

[0041] According to a further embodiment of the invention, the classification is performed using a software comprising a classification model.

[0042] According to a specific embodiment of the invention, said software executes steps of determining and comparing.

[0043] According to a specific embodiment of the invention, the classification model is selected from the group consisting of multivariate classification models, logistic regression models, support vector machine models, and decision tree models.

[0044] According to a further embodiment of the invention, there is provided a method for monitoring the treatment of a periprosthetic joint infection using the method described above.

[0045] According to a specific embodiment of the invention, there is provided a kit-of-parts comprising:

[0046] (a) detection reagents for detecting the expression level of at least 1 , 2, 3, 4 or 5 microRNAs selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial sample;

[0047] (b) optionally detection reagents for detecting the expression level of one or more of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p, JB0001 P

[0048] -5-

[0049] (c) optionally detection reagents for detecting the expression level of one or more of standard miRNAs, selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p; and

[0050] (d) a software comprising one or more classification model(s) for classification of the subject’s sample into the categories “septic” and “aseptic”.

[0051] According to a specific embodiment of the invention, the classification model provides a classification probability score for classifying the sample into the categories “septic” and “aseptic”, specifically the classification probability score of > 75 % relates to the category “septic” and the classification probability score of < 25 % relates to the category “aseptic”.

[0052] According to a specific embodiment of the invention, the category “septic” comprises the sub-categories “clearly septic” having a classification score of > 90 % and “likely septic” for a classification probability greater than or equal to 75 and less than 90 %; the category “aseptic” comprises the sub-categories “clearly aseptic” having a classification score of < 10 % and “likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; and optionally a further category “unclear” can be determined for a classification probability between 25 to 75 %.

[0053] According to a specific embodiment of the invention, there is provided a method of using the kit-of-parts for diagnosing a periprosthetic joint infection or monitoring the treatment of a periprosthetic joint infection.

[0054] According to a specific embodiment of the invention, a data set is generated, comprising the expression levels of the miRNAs from the aseptic reference synovial fluid samples.

[0055] According to a further embodiment of the invention, the classification model comprises the subsequent steps of:

[0056] (a) performing a multiple regression analysis of the expression level;

[0057] (b) calculating a classification probability score from the multiple regression analysis;

[0058] (c) optionally comparing the classification probability score with the classification probability scores observed in the aseptic reference synovial fluid samples; and

[0059] (d) classifying the sample based on its classification probability score into the categories “septic” and “aseptic”. JB0001 P

[0060] -6-

[0061] According to a specific embodiment of the invention, the category “septic” comprises the sub-categories “clearly septic” having a classification score of > 90 %” and “likely septic” for a classification probability greater than or equal to 75 and less than 90 %; the category “aseptic” comprises the sub-categories “clearly aseptic” having a classification score of < 10 % and “likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; and optionally a further category “unclear” can be determined for a classification probability between from 25 to 75 %.

[0062] According to a further embodiment, herein described is also a method of treating a subject suffering from complications following total joint replacement surgery, said method comprises the sequential steps of providing a synovial fluid sample from said subject, a) determining the expression level(s) of i. 1 , 2, 3, 4, or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in the synovial fluid sample; ii. optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p; and b) comparing the expression level(s) of (b) with a predetermined reference range of expression levels of the same miRNA(s) from an aseptic reference group, wherein the expression level(s) of (b) are classified as category septic if the expression level(s) of (b) are different from the reference range, and wherein the expression level(s) are classified as category aseptic if the expression level(s) of (b) are within the reference range, c) performing a revision surgery and / or antibiotic treatment when the sample is classified as septic.

[0063] FIGURES

[0064] Fig. 1 : NGS-based analysis of microRNA biomarkers in synovial fluid from aseptic and septic knee arthroplasty. Volcano plot depicting Iog2 fold change on the x- axis, and negative log 10 transformed adjusted p-value expressed as false-discovery rate (FDR) on y-axis for the comparison of microRNA levels in synovial fluid from septic (n=20) and aseptic (n=20) knee arthroplasty. The FDR = 0.05 threshold is shown as dashed JB0001 P

[0065] -7- horizontal line. miRNAs with FDR < 0.05 are shown as filled black dots. A subset of 18 miRNAs selected for further validation is highlighted using crosses (*) as symbols.

[0066] Fig. 2: Clustering of aseptic, septic, and osteoarthritic synovial fluid samples based on a microRNA expression pattern. Heatmap based on reads per million (RPM) values of 18 miRNAs selected from the NGS analysis. Septic (n=20, black), aseptic (n=20, white) and osteoarthritic (n=6, grey) synovial fluid samples are shown.

[0067] Fig. 3: Replication of NGS data for 18 microRNAs by RT-qPCR. Comparison of area-under-the-curve (AUC) values for the classification of septic vs aseptic synovial fluid samples obtained by RT-qPCR (y-axis) in comparison to NGS data (x-axis). A dashed line indicates 100 % concordance.

[0068] Fig. 4: Development of multivariate microRNA models for diagnosis of periprosthetic joint infections in knee and hip. Pipeline for the development of classification models: for each uni- or multivariate model, the data were partitioned into a training (60 % or 101 samples) and test (40 % or 72 samples) set. The training set was used to develop a classification model using cross-validation to increase robustness. Optimized and model parameters were then fixed and applied to the independent test set to assess performance (ROC analysis).

[0069] Fig. 5: Development of multivariate microRNA models for diagnosis of periprosthetic joint infections in knee and hip. The classification performances of the probability scores obtained from uni- (single miRNA) and multivariate (models with 2, 3, or 4 miRNAs) classification models on the test data set are shown as area-under-the- curve (AUC) from ROC analysis.

[0070] Fig. 6: Diagnostic performance of bivariate miRNA models for PJL Performances of bivariate classification models in the test data set are shown as AUC values from ROC analysis. A horizontal line indicates AUC = 0.9. For each miRNA, its univariate AUC-value (black dot, Single miR) as well as the AUC-values obtained from pairing the miRNA with each other of the 17 remaining miRNAs (box plot, Paired model) are shown. Results are based on the non-normalized (raw) expression levels estimated by RT-qPCR analysis.

[0071] Fig. 7: The classification performance of the probability score obtained from a specific bivariate miRNA model consisting of miR-142-5p and miR-338-5p for the independent test data set is shown as ROC curve.

[0072] Fig. 8: : Several reported biomarkers of blood-based sepsis are not differentially expressed in synovial fluids from periprosthetic joint infections. Volcano plot illustrating JB0001 P

[0073] -8- the Iog2-transformed fold change (x-axis) of miRNAs between septic and aseptic joints, as well as the Iog10-transformed significance level (false-discovery rate, FDR). A set of miRNAs, which was previously reported to be associated with blood-based sepsis is labelled. Filled black dots highlight miRNAs with significant differences between septic and aseptic synovial fluid. Empty circles highlight miRNAs that did not show a significant difference between septic and aseptic synovial fluid samples). The FDR = 0.05 threshold is shown as dashed horizontal line.

[0074] Fig. 9: The impact of assay normalization on the diagnostic performance of uni- and bivariate miRNA models. Performances of bivariate classification models in the test data set are shown as AUC values from ROC analysis. A horizontal line indicates AUC = 0.9. For each miRNA, its univariate AUC-value (black dot, Single miR) as well as the AUC-values obtained from pairing the miRNA with each other of the 17 remaining miRNAs (box plot, Paired model) are shown. Results are based on the normalized expression levels estimated by RT-qPCR analysis of miR-22-3p as standard microRNA.

[0075] Fig. 10: The impact of assay normalization on the diagnostic performance of uni- miRNA models. A side-by-side comparison of the non-normalized (raw) and miR-22 normalized AUC-values for a univariate classification model. Skipping the normalization works surprisingly well.

[0076] Fig. 11 : The impact of assay normalization on the diagnostic performance of bivariate miRNA models. A side-by-side comparison of the non-normalized (raw) and miR-22 normalized AUC-values for a bivariate classification model. Skipping the normalization works surprisingly well.

[0077] DETAILED DESCRIPTION

[0078] This invention is based on a concept and reduction to practice, for which the inventors have made a significant contribution, thereby providing a novel and inventive solution to the problem underlying the invention. Some aspects described herein may have been developed with the assistance of artificial intelligence (Al), yet the inventors being natural persons have contributed to the prior art by the subject matter described herein that is significantly more than any such Al-assisted aspects.

[0079] Unless indicated or defined otherwise, all terms used herein have their usual meaning in the art, which will be clear to the skilled person. Reference is for example made to the standard handbooks, such as Sambrook et al, "Molecular Cloning: A Laboratory Manual" (4th Ed.), Vols. 1 -3, Cold Spring Harbor Laboratory Press (2012); JB0001 P

[0080] -9-

[0081] Krebs et al., "Lewin's Genes XI", Jones & Bartlett Learning, (2017), and Murphy & Weaver, "Janeway's Immunobiology" (9th Ed., or more recent editions), Taylor & Francis Inc, 2017.

[0082] The subject matter of the claims specifically refers to artificial products or methods employing or producing such artificial products.

[0083] The terms “comprise”, “contain”, “have” and ‘Include” as used herein can be used synonymously and shall be understood as an open definition, allowing further members or parts or elements. “Consisting” is considered as a closest definition without further elements of the consisting definition feature. Thus “comprising” is broader and contains the “consisting” definition.

[0084] The term “about” as used herein refers to the same value or a value differing by + / - 5 % of the given value.

[0085] As used herein and in the claims, the singular form, for example “a”, “an” and ‘the” includes the plural, unless the context clearly dictates otherwise.

[0086] As used herein, “patient”, “individual”, or “subject” includes mammalian organisms, such as human and non-human mammals, for example, but not limited to, rodents, mice, rats, nonhuman primates, companion animals such as dogs and cats as well as livestock, e.g., sheep, cow, horse, etc. Therefore, for example, although the described embodiments illustrate use of the present methods on humans, those of skill in the art would readily recognize that these methods and compositions could also be applied to veterinary medicine as well as on other animals.

[0087] The terms “in vitro” or “in vitro method” refer to activities or methods performed outside an organism such as experimentation, or measurements done in or on living tissue or cells, in an artificial environment outside the organism, preferably with minimum alteration of the natural conditions. Specifically, such in vitro method is therefore not performed on the living organism; it is particularly not performed on humans.

[0088] The term ‘infection” is well known in the art and refers to the invasion of tissues of an organism (host) by pathogens, the multiplication of said pathogens in said tissues, and the reaction of host tissues to the infectious pathogens and the toxins they produce. In a specific embodiment the infection is an infection with virus, fungus, protozoa, plasmodia, worms, insects, or bacteria. The toxins released by an infectious pathogen can produce shock and sepsis.

[0089] Periprosthetic joint infection (PJI) is a disease and is clinically distinct from native bone or joint infection. PJI involves interactions between microorganisms, on the JB0001 P

[0090] -10- one hand, and the implant and host immune system, on the other. A periprosthetic joint infection (P JI) is well known in the art (Parvizi et al., 2011) and exists when: there is a sinus tract communicating with the prosthesis; or a pathogen is isolated by culture from at least two separate tissue or fluid samples obtained from the affected prosthetic joint; or four of the following six criteria exist:

[0091] Elevated serum erythrocyte sedimentation rate (ESR) and serum C-reactive protein (CRP) concentration,

[0092] Elevated synovial leukocyte count,

[0093] Elevated synovial neutrophil percentage (PMN%),

[0094] Presence of purulence in the affected joint,

[0095] Isolation of a microorganism in one culture of periprosthetic tissue or fluid, or

[0096] Greater than five neutrophils per high-power field in five high-power fields observed from histologic analysis of periprosthetic tissue at *400 magnification.

[0097] PJI may be present if fewer than four of these criteria are met.

[0098] In a specific embodiment, PJI diagnostic protocol is established using institutional criteria according to European Bone and Joint Infection Society (EBJIS) to identify PJI (McNally et al., 2021). Specifically, a PJI is classified as low-grade if no signs of acute inflammatory symptoms such as fever, erythema, and warmth are present. The presence of acute signs or symptoms of local inflammation and / or sinus tract with evidence of communication to the joint or visualization of the prosthesis is considered as a high-grade PJI. More specifically, routine laboratory values such as serum CRP levels, leukocyte counts, and the proportion of polymorphonuclear leukocytes in aspiration, as well as microbiological and histopathological findings at the time of revision surgery, are used in PJI diagnostic. In a specific embodiment, at least five periprosthetic tissue cultures from various suspicious surgical sites, and at least one specimen for histopathological analysis, in addition to a sonication analysis of retrieved implants, are obtained intraoperatively. In a more specific embodiment, a histopathological analysis is performed.

[0099] Herein, the term PJI encompasses both acute and chronic infection. In general, acute PJI is characterized by symptoms occurring for less than 4 weeks, whereas in case of a chronic infection symptoms occur for 4 or more weeks. JB0001 P

[0100] -11-

[0101] According to the method of the invention, the above terms low-grade and highgrade are termed septic.

[0102] The term “microorganism” as used herein refers to microorganisms as commonly understood in biology, including but not limited to viruses, bacteria, yeasts and molds. It is well known in the art that microorganisms are ubiquitous, and therefore, various microorganisms are present in various samples, such as animal samples, plant samples and environmental samples.

[0103] As used herein, the term “septic” refers to the occurrence of a periprosthetic joint infection. Specifically, it is characterized by the presence of microorganisms, specifically the presence of bacteria, such as but not limited to staphylococcus aureus.

[0104] As used herein, the term “aseptic” refers to lack of PJI. Specifically, freedom, to a specified degree, from microorganisms, e.g. bacteria.

[0105] As used herein, the term “sample” generally refers to a synovial fluid sample originating from a synovial joint.

[0106] The close proximity of synovial fluid to a potentially infected joint leads to a high signal to noise ratio for detection of microRNAs. Thus, PJIs with a low virulence, such as low-grade PJIs, induce differences in microRNA expression significant enough to allow a diagnosis with confidence. Said differences are not achieved with blood-circulating microRNAs, or with microRNAs from periarticular tissue samples.

[0107] As used herein, the term “providing a synovial fluid sample” refers to the provision of a synovial fluid sample for the inventive method. The synovial fluid sample can be obtained by aspiration of synovial fluid from a joint under sterile conditions in a first or previous step. The synovial fluid sample may have been stored until it is further used for the inventive method.

[0108] The terms “reference synovial fluid samples”, “aseptic reference synovial fluid sample”, or “reference sample” are used interchangeably and are to be understood as a control sample used for comparison with the sample, i.e. the subject’s synovial fluid sample for diagnosing a PJI, specifically a subject with joint pain. The reference synovial fluid sample may include a sample obtained from a biobank, a healthy subject, a subject which did not undergo arthroplasty, a subject without joint pain, or a subject without a joint prosthesis, wherein the reference sample originated is equivalent to the joint from which the (experimental) sample originated. Additionally, a control may also be a standard reference value or range of values. JB0001 P

[0109] -12-

[0110] As used herein, the term “aseptic reference group” refers to a pool of aseptic reference samples from two or more aseptic subjects. In specific embodiments, the aseptic reference group is composed of at least about 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100 subjects, preferably 50 subjects.

[0111] As used herein, the term “joint pain” includes, for example, acute joint pain and chronic joint pain. Generally, pain is felt differently from one individual to the next ranging from mild to severe and varying in type. In a specific embodiment, pain is measured using a pain scale, specifically the pain scale is a numerical rating scale, visual analog scale, Color analog pain scale, categorical scale, Wong-Baker faces pain scale, FLACC pain scale (FI.ACC stands for Facial expression, Leg tension or relaxation, Activity, Crying, Consolability), CRIES pain scale, COMFORT pain scale, McGill pain questionnaire, Mankoski pain scale, or Brief Pain Inventory (BPI) worksheet. In a further specific embodiment, pain is measured using a numerical rating scale, wherein 0 (zero) is no pain, 1-3 is indicative of mild pain, 4-6 is indicative of moderate pain, and 7-10 is indicative of severe pain (10 = maximum imaginable pain). In another specific embodiment, pain is measured using a color analog pain scale, wherein a gradient between two colors (e.g. blue and red) is used to measure the level of pain, specifically the two extreme colors of the gradient are respectively indicative of no pain (e.g. blue) or maximum imaginable pain (e.g. red), and wherein the gradient indicative of the relative level of pain ranging from no pain to maximum imaginable pain.

[0112] The present invention provides selected microRNAs for use in a method for the diagnosis of a PJI in a synovial fluid sample.

[0113] Specifically, said microRNAs are 1 , 2, 3, 4 or 5, of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615, or isoforms or variants thereof, and optionally at least one of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p, or isoforms or variants thereof.

[0114] In a specific embodiment 1 , 2, 3, 4 or 5, microRNA(s) is or are selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615.

[0115] In a further specific embodiment, the microRNA is miR-338-5p.

[0116] In another specific embodiment, the microRNAs miR-338-5p, miR-223-3p, and miR-505-3p are used in the inventive method.

[0117] In another specific embodiment, the microRNAs are selected from, but not limited to, the combinations JB0001 P

[0118] -13- miR-338-5p, miR-223-3p; miR-338-5p, miR-505-3p; miR-223-3p, miR-505-3p; or miR-338-5p, miR-223-3p, miR-505-3p.

[0119] In another specific embodiment, the microRNAs are selected from, but not limited to, the combinations miR-338-5p, miR-142-5p; miR-338-5p, miR-3615; miR-338-5p, miR-223-3p, miR-505-3p, miR-142-5p; miR-338-5p, miR-223-3p, miR-505-3p, miR-3615; miR-338-5p, miR-223-3p, miR-505-3p, miR-142-5p, miR-3615; miR-338-5p, miR-505-3p; miR-142-3p, miR-338-5p; miR-151 -3p, miR-338-5p; miR-191 -5p, miR-223-3p, miR-425-3p; miR-142-5p, miR-191-5p, miR-425-3p; miR-151 a-3p, miR-223-3p, miR-424-3p; miR-30d-5p, miR-214-3p, miR-338-5p, miR-3615; or miR-142-5p, miR-191-5p, miR-424-3p, miR-3615.

[0120] As used herein, the terms “microRNA”, “miRNA”, or “miR” are used interchangeably and designate a non-coding RNA molecule of between 17 and 25 nucleotides which hybridizes to and regulates the expression of a protein-coding messenger RNA. Nucleotide sequences of mature miRNAs and their respective precursors are known in the art and available from the database miRBase at mirbase.org / index.shtml or from Sanger database at microrna.sanger.ac.uk / sequences / ftp.shtml.

[0121] For the purpose of the invention, “isoforms or variants” (which have also be termed “isomiRs”) of a miRNA include trimming variants (5’ trimming variants in which the 5’ dicing site is upstream or downstream from the miRNA sequence; 3’ trimming variants: the 3’ dicing site is upstream or downstream from the miRNA sequence), or variants having one or more nucleotide modifications (3' nucleotide addition to the 3' end of the miRNA; nucleotide substitution by changing nucleotides from the miRNA precursor), or the complementary mature microRNA strand including its isoforms and variants (for example for a given 5’ mature microRNA the complementary 3’ mature microRNA and JB0001 P

[0122] -14- vice-versa). With regard to nucleotide modification, the nucleotides relevant for RNA / RNA binding, i.e. the 5'-seed region and nucleotides at the cleavage / anchor side are exempt from modification.

[0123] As used herein, if not otherwise stated, the term “miRNA” encompasses 3p and 5p strands and optionally also its isoforms and variants.

[0124] Specifically, the identifiers “miR-<number>-5p” as used herein also may encompass its complementary 3p miRNA and vice versa.

[0125] In specific embodiments, the miRNAs of interest are detected using a nucleotide that hybridizes, preferably under stringent conditions, with said miRNA of interest and measuring the hybridization signal.

[0126] As used herein, the term “level”, when relating to a microRNA, generally refers to the expression level of said microRNA.

[0127] The microRNA “expression levels” can be determined by any of the methods described herein.

[0128] In one embodiment, the “expression level” of microRNAs of interest is determined by next-generation sequencing. The term “next-generation sequencing (NGS)”, also known as high-throughput sequencing, is used herein to describe a number of different modern sequencing technologies that allow to sequence and quantify levels of DNA and RNA much more quickly and cheaply than the previously used sequencing methods such as Sanger sequencing. It is based on micro- and nanotechnologies to reduce the size of sample, the reagent costs, and to enable massively parallel sequencing reactions. It can be highly multiplexed which allows simultaneous sequencing and analysis of millions of samples. NGS includes first, second, third as well as subsequent Next Generations Sequencing technologies. Non limiting examples are the nanopore or semiconductor technologies (e.g. Oxford Nanopore Technologies®, United Kingdom) or the Illumina® smalIRNA-Seq Platform or electron detection-based methods such as Thermo Fisher’s Ion Torrent®.

[0129] Specifically, small RNA sequencing libraries can be generated using library preparation kits well known in the art, such as the CleanTag SmalIRNA library preparation kit (TriLink BioTechnologies, LLC, USA), or the RealSeq® Biosciences small RNA- sequencing kit. Usually, RNA is first ligated to a 3’ and 5’ adapter, and is then reverse transcribed into DNA followed by PCR amplification. PCR amplification can be performed using barcoded primers, such as the barcoded primers from Illumina for small RNA sequencing. Before sequencing, PCR products are purified using protocols well known in JB0001 P

[0130] -15- the art. For example, PCR products can be purified using the QiaQuick® protocol from QIAGEN® and can subsequently be size checked by suitable methods, such as e.g. capillary electrophoresis. The performance of small RNA sequencing workflows can be assessed by the addition of synthetic oligonucleotides also referred to as “spike-ins”. Specifically, spike-ins with mathematically randomized 5’ and 3’ ends as well as defined molar concentrations are useful to assess the analytical performance of small RNA sequencing experiments and estimate absolute microRNAs concentrations.

[0131] Specifically, next-generation sequencing can be performed on any suitable platform, such as e.g. the Illumina® NextSeq® 2000. Sequencing reads are usually adapter-trimmed, quality checked and edited according to bioinformatics methods well known in the art to prepare for further use.

[0132] In a further embodiment, the “expression level” of the microRNAs of interest is determined by “polymerase chain reaction” (PCR). PCR methods are well known in the art and widely used. They include quantitative real time PCR, semi-quantitative PCR, multiplex PCR, digital PCR, or any combination thereof. In a particularly preferred embodiment, the levels of microRNAs are determined by quantitative real time PCR (qRT- PCR). Methods of determining the levels of microRNAs using qRT-PCR are known in the art, and are usually preceded by reverse transcription of a microRNA into a cDNA.

[0133] In the PCR methods useful in the present invention, the primers are usually based on the mature microRNA molecule, but may include chemical modifications to optimize hybridization behavior. qRT-PCR methods may determine an absolute level of expression of a microRNA. Alternatively, qRT-PCR methods may determine the relative quantity of a microRNA.

[0134] The relative quantity of a microRNA may be determined by normalizing the level of the microRNA to the level of one or more internal standard nucleic acid sequences or “standard microRNA (miRNA)”, also known in the art as housekeeping gene / microRNA.

[0135] In general, such standard microRNA should have a constant and high expression level in the analyzed sample. The analysis of said expression level with a mathematical model of gene expression yields the expression stability and average abundance of said standard microRNAs. Non-limiting examples of software comprising such a mathematical model are the NormFinder software package (https: / / www.moma.dk / software / normfinder), geNorm (https: / / genorm.cmgg.be), Bestkeeper applet, RefGenes, Genevestigator, or REST — Relative Expression Software Tool (Technical University Munich and QIAGEN®). Generally, a standard microRNA has JB0001 P

[0136] -16- a low variability and a high average abundance in said sample, specifically the microRNA has low intra- and inter-group variability, which is expressed as an expression stability value < 1 and an average abundance 10000 reads per million or copies / pl sample.

[0137] In addition, non-limiting examples of microRNAs that have constant and high levels in synovial fluid and may be used as „standard miRNAs" in the inventive method are let-7g-5p, let-7f-5p, let-7a-5p, let-7i-3p, let-7b-3p, let-7d-3p, let-7i-5p, let-7e-5p, let-7c-5p, let-7f-2-3p, let-7b-5p, let-7a-3p, miR-10b-3p, miR-10a-5p, miR-10b-5p, miR-10a-3p, miR-100-5p, miR-101-3p, miR-101-5p, miR-103a-3p, miR-103a-2-5p, miR-10396a-5p, miR-10396b-5p, miR-10399-5p, miR-10527-5p, miR-106b-5p, miR-106b-3p, miR-106a-5p, miR-107, miR-1180-3p, miR-12136, miR-1246, miR-1247-5p, miR-125b-1-3p, miR-125b-5p, miR-125a-3p, miR-125b-2-3p, miR-125a-5p, miR-126-3p, miR-126-5p, miR-1260b, miR-1260a, miR-127-3p, miR-1271-5p, miR-128-3p, miR-1290, miR-130b-3p, miR-130b-5p, miR-130a-3p, miR-1301-3p, miR-1304-3p, miR-1306-5p, miR-1307-5p, miR-1307-3p, miR-132-5p, miR-132-3p, miR-133a-3p, miR-134-5p, miR-136-5p, miR-136-3p, miR-139-5p, miR-140-5p, miR-140-3p, miR-143-5p, miR-143-3p, miR-144-5p, miR-144-3p, miR-145-3p, miR-145-5p, miR-146a-5p, miR-146b-5p, miR-148a-3p, miR-148a-5p, miR-148b-3p, miR-15a-5p, miR-15b-5p, miR-15b-3p, miR-150-5p, miR-150-3p, miR-151a-5p, miR-151 b, miR-152-3p, miR-1537-3p, miR-155-5p, miR-16-5p, miR-16-2-3p, miR-17-5p, miR-17-3p, miR-18a-3p, miR-181a-5p, miR-181d-5p, miR-181c-5p, miR-181a-2-3p, miR-181a-3p, miR-181c-3p, miR-181 b-5p, miR-182-5p, miR-183-5p, miR-1843, miR-185-5p, miR-185-3p, miR-186-5p, miR-19b-3p, miR-190b-5p, miR-1908-5p, miR-191-3p, miR-192-5p, miR-193a-3p, miR-193a-5p, miR-193b-3p, miR-193b-5p, miR-195-5p, miR-196a-5p, miR-196b-5p, miR-197-3p, miR-199b-5p, miR-199b-3p, miR-199a-3p, miR-199a-5p, miR-20b-5p, miR-20a-5p, miR-200c-3p, miR-203a-3p, miR-205-5p, miR-21-3p, miR-21-5p, miR-210-3p, miR-210-5p, miR-2110, miR-212-3p, miR-218-5p, miR-22-5p, miR-22-3p, miR-221-5p, miR-221-3p, miR-222-3p, miR-222-5p, miR-223-5p, miR-224-5p, miR-23a-5p, miR-23a-3p, miR-23b-3p, miR-23c, miR-2355-3p, miR-2355-5p, miR-24-2-5p, miR-24-3p, miR-25-3p, miR-25-5p, miR-26b-5p, miR-26a-5p, miR-26b-3p, miR-27b-3p, miR-27a-5p, miR-27a-3p, miR-28-5p, miR-28-3p, miR-29b-3p, miR-29c-3p, miR-29a-3p, miR-296-5p, miR-30a-5p, miR-30e-3p, miR-30a-3p, miR-30c-5p, miR-30e-5p, miR-30b-5p, miR-301a-3p, miR-31-5p, miR-3168, miR-32-5p, miR-320b, miR-320d, miR-320a-3p, miR-320c, miR-323b-3p, miR-323a-3p, miR-324-3p, miR-324-5p, miR-326, miR-328-3p, miR-33a-5p, miR-330-3p, miR-331-3p, JB0001 P

[0138] -17- miR-331-5p, miR-335-5p, miR-337-5p, miR-338-3p, miR-339-5p, miR-339-3p, miR-34c-5p, miR-34b-5p, miR-34a-5p, miR-340-3p, miR-340-5p, miR-342-3p, miR-342-5p, miR-3605-3p, miR-361-5p, miR-361-3p, miR-3613-5p, miR-3613-3p, miR-362-5p, miR-362-3p, miR-363-3p, miR-365b-3p, miR-365a-3p, miR-3651 , miR-369-5p, miR-369-3p, miR-371 b-5p, miR-374a-3p, miR-374b-5p, miR-376a-3p, miR-376c-3p, miR-377-3p, miR-377-5p, miR-378a-3p, miR-378c, miR-378i, miR-378a-5p, miR-381-3p, miR-382-3p, miR-382-5p, miR-3909, miR-3960, miR-409-3p, miR-409-5p, miR-410-3p, miR-411-5p, miR-421 , miR-423-3p, miR-423-5p, miR-424-5p, miR-4286, miR-432-5p, miR-433-3p, miR-4448, miR-4454, miR-4485-3p, miR-4488, miR-4492, miR-4508, miR-451 a, miR-4516, miR-452-5p, miR-454-5p, miR-454-3p, miR-455-3p, miR-455-5p, miR-4707-3p, miR-4732-3p, miR-483-5p, miR-484, miR-485-3p, miR-486-3p, miR-486-5p, miR-487b-3p, miR-487a-5p, miR-489-3p, miR-494-3p, miR-497-5p, miR-500a-3p, miR-501-5p, miR-501-3p, miR-502-3p, miR-503-5p, miR-504-5p, miR-5100, miR-532-5p, miR-532-3p, miR-539-5p, miR-539-3p, miR-542-3p, miR-543, miR-548k, miR-548w, miR-548ae-5p, miR-548aj-5p, miR-548x-5p, miR-548e-5p, miR-548e-3p, miR-548c-5p, miR-548am-5p, miR-548ak, miR-548l, miR-548ad-5p, miR-548h-5p, miR-548g-5p, miR-548au-5p, miR-548o-5p, miR-550a-5p, miR-550a-3-5p, miR-574-5p, miR-574-3p, miR-576-3p, miR-576-5p, miR-582-5p, miR-582-3p, miR-589-5p, miR-598-3p, miR-615-3p, miR-618, miR-619-5p, miR-625-5p, miR-625-3p, miR-628-3p, miR-642a-3p, miR-642a-5p, miR-6503-5p, miR-6503-3p, miR-651-5p, miR-652-3p, miR-654-3p, miR-656-3p, miR-660-5p, miR-664a-5p, miR-671-3p, miR-671-5p, miR-7-1-3p, miR-708-5p, miR-744-5p, miR-766-3p, miR-769-5p, miR-7704, miR-7847-3p, miR-874-3p, miR-877-5p, miR-889-3p, miR-9-5p, miR-92b-5p, miR-92b-3p, miR-92a-3p, miR-92a-1-5p, miR-93-3p, miR-93-5p, miR-941 , miR-95-3p, miR-96-5p, miR-98-5p, miR-99a-5p, miR-99b-3p, or miR-99b-5p.

[0139] In a preferred embodiment, the standard microRNA is selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p. In a further preferred embodiment, the standard microRNA is miR-22-3p.

[0140] In addition, synthetic RNA sequences added in an equimolar amount during RNA isolation or cDNA synthesis may be used as references for relative quantification of specific microRNAs. JB0001 P

[0141] -18-

[0142] Alternatively, the relative logarithmic difference between two microRNAs can be calculated to form self-normalizing microRNA pairs. Thereby the need for standard microRNAs can be circumvented.

[0143] Primers for detection of microRNAs are commercially available, e.g. as microRNA miRCURY LNA™ PCR primer sets from Qiagen.

[0144] Since microRNAs are relatively short molecules, it may be useful to lengthen them by adding adenosine monomers to the strand (a technique known as polyadenylation) before reverse transcription and amplification. Briefly, the RNA may be extracted from the sample by a suitable reagent (e.g. Trizol reagent), polyadenylated in the presence of ATP and poly(A) polymerase, reverse transcribed into cDNA using a poly(T) adapter and 5' RACE sequence, and amplified using a forward primer derived from the 3' end of the microRNA and a reverse RACE primer. Improvements of this technique include designing the RACE primer with a nucleotide at its 3' end (constituting an A, C, or G, but not a T, so to exclude priming anywhere on the polyA sequence and enforce priming on the microRNA sequence) or RACE primers which are anchored at the 3’ cDNA end of a specific microRNA using 2, 3, 4, or more nucleotides with or without chemical modification.

[0145] The detection of a microRNA may also be achieved by other methods known in the art, e.g. those described in WO201114476, like by the deep sequencing method, bead-based quantification, e.g. Illumina® bead-arrays, hydrogel-particle based quantification, e.g. Firefly™, by microarray technology, e.g. the Ncode™ human microRNA array available from Invitrogen, chip arrays available from Affymetrix, Agilent, or microarrays which employ LNA-backbone capture probes (miRCURY LNA™ arrays) or microarrays which employ native or chemically modified capture probes (e.g. from Toray).

[0146] The difference in microRNA levels can also be determined using multiplex chemiluminescence-based nucleic acid assays such as Panomics, or reporter plasmid assays (“biosensors”) containing reporter proteins with microRNA-complementary regulatory sites, or other hybridization-based techniques known in the art. For example, the detection of microRNAs can be based on the formation of DNA-RNA hybrids and antibody-based detection, or biotinylated DNA probes specific to miRNAs.

[0147] As used herein, the term “reference range of expression levels” refers to a range of miRNA expression levels determined from an aseptic reference group or reference sample. Specifically, the reference range of expression levels for a given miRNA may be JB0001 P

[0148] -19- defined as the 95 % confidence interval (Cl). The 95 % Cl is calculated based on the mean expression value and standard deviation observed in the reference group. In a specific embodiment, the reference range of expression levels is predetermined.

[0149] As used herein, the term “comparing the expression levels” refers to the direct comparison of the relationship between the expression levels of the same microRNA from two or more different samples. In a specific embodiment, the expression levels of a microRNA from a sample are compared to the expression levels of the same microRNA from an aseptic reference sample or aseptic reference group.

[0150] As used herein, the term “ratio” refers to the mathematical operation of dividing one value by a second value of the same type. In a specific embodiment, the ratio of the expression levels of a microRNA from a sample is divided by the expression levels of the same microRNA from an aseptic reference group. In another specific embodiment, the ratio of the expression levels of a microRNA from a sample is divided by the expression levels of a standard microRNA.

[0151] According to a specific embodiment, a data set is generated, comprising the expression levels of the microRNAs from the aseptic reference synovial fluid samples.

[0152] As used herein, “software” includes, but is not limited to, one or more computer- readable and I or executable instructions that cause a computer or other electronic device to perform functions, actions, and I or behaviors in the intended manner. Instructions can be embodied in various forms, such as routines, algorithms, modules, or programs, including independent applications or code from dynamic link libraries. The software may also be implemented in various forms, such as stand-alone programs, function calls, servlets, applets, instructions stored in memory such as memory 418, part of the operating system, or other types of executable instructions. Those skilled in the art should recognize that the form of the software depends on, for example, the requirements of the intended application, the environment in which it is running, and I or the designer I programmer's wishes. In a specific embodiment, the classification of synovial fluid sample, specifically the microRNA expression levels, is performed using a software comprising a classification model.

[0153] A classification technique (or classifier) is a systematic approach to building classification models from a data set, such as microRNA expression levels of a sample. Examples include logistic regression models, support vector machine models, decision tree models, rule-based classifiers, neural networks and naive Bayes classifiers. Each technique employs a learning algorithm to identify a model that best fits the relationship JB0001 P

[0154] -20- between the attribute set and class label of the input data. The model generated by a learning algorithm should both fit the input data well and correctly predict the class labels of data it has never seen before. Therefore, a key objective of the learning algorithm is to build models with good generalization capability, i.e. models that accurately predict the class labels of previously unknown data.

[0155] As used herein, the term “classification model” refers to a mathematical model that allows the analyzing and classification of miR expression levels from a synovial fluid sample.

[0156] Specifically, a “classification probability score” is calculated from a multivariate (multivariable) analysis of the expression levels of microRNAs. Herein, the term multivariate and multivariable can be used interchangeably. Specifically, a “classification probability score” can be calculated from a multivariate analysis of the expression levels of microRNAs based on the following steps: 1) Model Specification: the classification model is defined using an equation with an intercept and coefficients for each predictor (microRNA) variable. 2) Logit Transformation: the combination of predictors (microRNA) and coefficients is transformed into the log-odds of the probability using the logit function. 3) Inverse Logit Function: the log-odds is converted back to a probability using the logistic function, resulting in the probability score. Specifically, the multivariate analysis is performed with a multivariate classification model. Non-limiting examples of multivariate classification models are multiple regression analysis model, multivariate logistic regression model, logistic regression models, support vector machine models, and decision tree models. In a specific embodiment, the calculated classification probability scores of the sample are compared with the classification probability scores observed in an aseptic reference synovial fluid sample or a data set comprising expression levels or classification probability scores from aseptic reference synovial fluid sample, specifically the detection of an increase or decrease of the probability scores of the sample compared to the probability scores of the reference synovial fluid sample, more specifically, the comparison relates to a decrease or increase in the level of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and / or miR-3615 in a synovial fluid sample, or isoforms or variants thereof, can be a specific indicator of PJL The synovial fluid sample is classified into one of at least two categories “septic” and “aseptic”, wherein the classification is based on the calculated classification probability score(s).

[0157] Further described herein are clinically useful cut-offs for diagnosing a PJI and “classification” of a subject’s synovial fluid sample into one of at least two categories JB0001 P

[0158] -21-

[0159] “septic” and “aseptic” based on detected expression levels of microRNAs, specifically, the classification into one of at least two categories “septic” and “aseptic” is based on a classification probability score.

[0160] Specifically, a classification probability score of > 75 %, specifically > 75 %,

[0161] > 76 %, > 77 %, > 78 %, > 79 %, > 80 %, > 81 %, > 82 %, > 83 %, > 84 %, > 85 %,

[0162] > 86 %, 87 %, > 88 %, > 89 %, or > 90 %, preferably 90 % relates to the category “septic” and the classification probability score of < 25 %, specifically < 25 %, < 24 %, < 23 %, < 22 %, < 21 %, < 20 %, < 19 %, < 18 %, < 17 %, < 16 %, < 15 %, < 14 %, < 13 %, < 12 %, < 11 %, or < 10 %, preferably < 25 % relates to the category “aseptic”. One or more further sub-categories may be “clearly septic”, “likely septic”, “likely aseptic”, “clearly aseptic”, or “unclear”.

[0163] Specifically a sub-category “clearly septic” for a classification probability of > 76 %,

[0164] > 80 %, 81 %, > 85 %, or > 90 %, preferably of 90 % is determined.

[0165] Specifically, a sub-category “likely septic” for a classification probability greater than or equal to 50 %, 51 %, 52 %, 53 %, 54 %, 55 %, 56 %, 57 %, 58 %, 59 %, 60 %, 61 %, 62 %, 63 %, 64 %, 65 %, 66 %, 67 %, 68 %, 69 %, 70 %, 71 %, 72 %, 73 %, 74 %, or 75 % and less than 90 %, or greater than or equal to 50 %, 51 %, 52 %, 53 %, 54 %, 55 %, 56 %, 57 %, 58 %, 59 %, 60 %, 61 %, 62 %, 63 %, 64 %, 65 %, 66 %, 67 %, 68 %, 69 %, 70 %, 71 %, 72 %, 73 %, 74 %, or 75 % and less than 85 %, or greater than or equal to 50 %, 51 %, 52 %, 53 %, 54 %, 55 %, 56 %, 57 %, 58 %, 59 %, 60 %, 61 %, 62 %, 63 %, 64 %, 65 %, 66 %, 67 %, 68 %, 69 %, 70 %, 71 %, 72 %, 73 %, 74 %, or 75 % and less than 80 %, preferably greater than or equal to 75 and less than 90 %.

[0166] Specifically, a sub-category “likely aseptic” for a classification probability greater than 10 and less than or equal to 25 %, 26 %, 27 %, 28 %, 29 %, 30 %, 31 %, 32 %, 33 %, 34 %,

[0167] 35 %, 36 %, 37 %, 38 %, 39 %, 40 %, 41 %, 42 %, 43 %, 44 %, 45 %, 46 %, 47 %, 48 %,

[0168] 49 %, or 50 %, or greater than 15 and less than or equal to 25 %, 26 %, 27 %, 28 %,

[0169] 29 %, 30 %, 31 %, 32 %, 33 %, 34 %, 35 %, 36 %, 37 %, 38 %, 39 %, 40 %, 41 %, 42 %,

[0170] 43 %, 44 %, 45 %, 46 %, 47 %, 48 %, 49 %, or 50 %, or greater than 20 and less than or equal to 25 %, 26 %, 27 %, 28 %, 29 %, 30 %, 31 %, 32 %, 33 %, 34 %, 35 %, 36 %, 37 %, 38 %, 39 %, 40 %, 41 %, 42 %, 43 %, 44 %, 45 %, 46 %, 47 %, 48 %, 49 %, or

[0171] 50 %, preferably greater than 10 and less than or equal to 25 %.

[0172] Specifically, a sub-category “clearly aseptic” for a classification probability of < 20, < 15, or < 10, preferably < 10 is determined.

[0173] Specifically, a further category “unclear” for a classification probability 10 to 90 %, 11 to JB0001 P

[0174] -22-

[0175] 89 %, 12 to 88 %, 13 to 87 %, 14 to 86 %, 15 to 85 %, 16 to 84 % , 17 to 83 %, 18 to 82 %, 19 to 81 %, 20 to 80 %, 21 to 79 %, 22 to 78 %, 23 to 77 %, 24 to 76 %, 25 to 75 %, preferably between 25 to 75 %.

[0176] According to the method provided herein, samples that are classified in either of the (sub-) categories “septic”, “clearly septic”, “likely septic”, or “unclear” have or may be indicative of a P JI, specifically may be indicative of the presence of microorganisms or microorganism contamination in the joint the sample originated from. In contrast, samples that are classified into the category “aseptic” are not indicative of a PJI and are free from microorganisms, hence by extension, the joint the sample originated from are free from a microorganism contamination.

[0177] In a specific embodiment, the expression levels of 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100, preferably 50 aseptic reference samples are determined and used to define a reference interval that will be evaluated as aseptic. Specifically, the reference interval may be defined as the 95 % confidence interval (Cl), which is estimated based on the mean expression value and standard deviation of a reference group using the equation: Cl = x"±t(ns), where t is the t-value for the 95 % confidence interval, which originates from the t- distribution and depends on the sample size n (the degrees of freedom are n-1).

[0178] When applying the inventive method to samples with an unknown diagnosis, the determined expression levels are compared with the reference interval. If the determined expression levels are within the reference interval of variation of aseptic reference samples, the sample will be classified as aseptic (and vice versa).

[0179] In a specific embodiment, no reference sample is needed for classifying the synovial fluid sample from a joint.

[0180] According to a specific embodiment, the expression level of a standard microRNA is determined and the ratio of the expression levels of each microRNA from the sample and a standard microRNA is calculated to normalize all expression levels to the standard microRNA. The subsequent classification is based on the ratios of the expression levels rather than the non-normalized expression levels. Specifically, the ratios of the expression levels are used as input for the classification model.

[0181] As used herein, the term “diagnosing a periprosthetic joint infection” refers to an in vitro method of diagnosing PJI in a sample of a subject, wherein said sample has been taken from the subject previously and is not returned to the subject. Another term JB0001 P

[0182] -23- for a sample being taken for diagnosis from a subject is “providing a sample”. Herein described is an in vitro method for diagnosing a PJI in a synovial fluid sample of a subject.

[0183] As used herein, the term “treatment of a periprosthetic joint infection” refers to procedures performed on a subject suffering from PJI and aiming to treat, cure, or heal the PJI, which can be, but are not limited to, a revision surgery, anti-inflammatory treatment, antibiotic treatment, and the like.

[0184] Herein described is a method for monitoring the treatment of a PJI using the inventive method. The herein described method is not a method for treatment of PJI in a subject, instead, the herein described method refers to monitoring of the treatment, specifically the herein described method is an in vitro method comprising determining and comparing expression levels of miRNAs in synovial fluid samples of the subject. As used herein, the term “complications following total joint replacement (TJR) surgery” refers to medical complications arising following a TJR, which can arise within weeks to months to years following the TJR, specifically with symptoms sustaining for less than 4 weeks in case of acute infection or equal to 4 weeks or longer than 4 weeks in case of chronic infection, and can include one or more of the following symptoms or classic signs of infection e.g. pain, redness, warmth, swelling at the surgical site, synovial inflammation, wound complications or persistent and / or purulent drainage.

[0185] Herein, complications following total joint replacement surgery can be a periprosthetic joint infection.

[0186] Herein described is also a method of treating a subject suffering from complications following total joint replacement surgery, wherein said complications are a periprosthetic joint infection determined by miRNA expression levels of selected inventive miRNAs in synovial fluid of the subject, specifically miRNA expression levels classifying the sample as septic.

[0187] As used herein, treatment of a subject suffering from said complications can include standard medical and surgical procedures for periprosthetic joint infections, which can be, but are not limited to, revision surgery, and / or antibiotic treatments. Specifically treatment can be selected from the group consisting of debridement, administration of antibiotics, implant retention (DAIR), 1 -stage and 2-stage exchange procedures, amputation, device removal without reimplantation (resection arthroplasty), and arthrodesis (fusion). Commonly used antibiotics include cephalosporins like cefazolin, cioxacillin, amoxicillin, piperacillin / tazobactam, carbapenems, aztreonam, aminoglycosides, fluoroquinolones, doxycycline, vancomycin, linezolid, daptomycin, JB0001 P

[0188] -24- clindamycin, trimethoprim / sulfamethoxazole, fosfomycin, rifampin, dalbavancin, oritavancin, or flucioxacillin, depending on the specific pathogen and infection type. A person skilled in the art is familiar with said procedures and knows that the choice of surgical procedure is further influenced by the duration of symptoms, the offending microorganism, and patient comorbidities.

[0189] Further provided herein is a kit-of-parts comprising: a) detection reagents for detecting the expression level of at least 1 , 2, 3, 4, or 5 microRNAs selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial sample; b) detection reagents for detecting the expression level of one or more of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p, c) optionally detection reagents for detecting the expression level of one or more of reference miRNAs, selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p; d) software comprising one or more classification model(s) for classification of the subject’s sample into one of at least two categories “septic” and “aseptic” based on detected expression levels of microRNAs.

[0190] Specifically, the kit-of-parts provided herein can be a qRT-PCR kit comprising reagents for RNA extraction, cDNA synthesis and fluorescence-based amplification of specific microRNA target sequences, specifically at least miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, or miR-3615. Fluorescent reporter probes detect only the DNA containing the sequence complementary to the probe; therefore, use of the reporter probe significantly increases specificity, and enables performing the technique even in the presence of other dsDNA. Using different-colored labels, fluorescent probes can be used in multiplex assays for monitoring several target sequences in the same tube. These fluorescent reporter molecules include sequence specific probes as detection reagents such as Molecular Beacons, FRET Hybridization Probes, Scorpion Primers® or TaqMan® Probes. Various methods of detecting miRNA using qRT-PCR are known to the person skilled in the art.

[0191] Specifically, the kit-of-parts provided herein can be a microarray kit comprising a high-density or low-density array with capture probes for specific microRNA sequences, JB0001 P

[0192] -25- chemicals for microRNA labeling and hybridization and wash buffers to increase stringency and specificity of the hybridization reaction. microRNA labeling can be achieved using biotin probes such as for example Biotin-16-UTP from Lucigen.

[0193] Specifically, the kit-of-parts provided herein can be a next-generation sequencing kit comprising reagents required for 5’ and 3’ adapter ligation to microRNAs, reverse transcription and PCR amplification to obtain sequencing libraries suitable for nextgeneration sequencing.

[0194] The present invention also encompasses the following embodiments:

[0195] 1. An in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of: a) determining the expression level(s) of

[0196] (i) 1 , 2, 3, or 4 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, and miR-505-3p in a synovial fluid sample of the subject;

[0197] (ii) optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, miR-629-5p, and miR-3615; and b) comparing the expression level(s) of (a) with a predetermined reference range of expression levels of the same miRNA(s) from an aseptic reference group, wherein the expression level(s) of (a) are classified as category septic if the expression level(s) of (a) are different from the reference range, and wherein the expression level(s) are classified as category aseptic if the expression level(s) of (a) are within the reference range.

[0198] 2. An in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of: a) determining the expression level(s) of

[0199] (i) 1 , 2, 3, or 4 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, and miR-505-3p in a synovial fluid sample of the subject;

[0200] (ii) optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, miR-629-5p, and miR-3615; JB0001 P

[0201] -26- b) determining the expression level of a standard miRNA having low expression stability and high average abundance; c) determining a ratio of the expression level of said standard miRNA of (c) and the miRNA(s) of (b); d) determining a ratio of the expression level of the same standard miRNA from an aseptic reference group and the expression level(s) of the same miRNA(s) from an aseptic reference group; and e) comparing said ratios, wherein the ratio of (c) is classified as category septic if it is different from ratio of (d), and wherein the ratio of (c) is classified as category aseptic if it is similar to (d).

[0202] 3. The method of embodiment 2, wherein the expression stability and average abundance is identified with a mathematical model of gene expression; specifically, the standard miRNA is selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p.

[0203] 4. The method of any one of embodiments 1 to 3, wherein the synovial fluid sample is from a joint of a subject on which arthroplasty was performed.

[0204] 5. The method of any one of embodiments 1 to 4, wherein the synovial fluid sample is from a subject with joint pain.

[0205] 6. The method of any one of embodiments 1 to 5, wherein the classification is performed using a classification model.

[0206] 7. The method of any one of embodiments 1 to 6, wherein the classification is performed using a software comprising a classification model.

[0207] 8. The method of embodiment 7, wherein said software executes steps (a) and (c) of embodiment 1 or steps (b) to (e) of embodiment 2.

[0208] 9. The method of any one of embodiments 6 to 8, wherein the classification model is selected from the group consisting of multivariate classification models, logistic regression models, support vector machine models, and decision tree models.

[0209] 10. A method for monitoring the treatment of a periprosthetic joint infection using the method of any one of embodiments 1 to 9.

[0210] 11 . A kit-of-parts comprising: a) detection reagents for detecting the expression level of at least 1 , 2, 3, or 4 microRNAs selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, and miR-505-3p; JB0001 P

[0211] -27- b) optionally detection reagents for detecting the expression level of one or more of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, miR-629-5p, and miR-3615, c) optionally detection reagents for detecting the expression level of one or more of standard miRNAs, selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p; and d) a software comprising one or more classification model(s) for classification of the subject’s sample into the categories “septic” and “aseptic”

[0212] 12. The kit-of-parts of embodiment 11 , wherein the classification model provides a classification probability score for classifying the sample wherein the category “septic” has a classification score of > 75” and “aseptic” has classification score of < 25 %.

[0213] 13. The kit-of-parts of embodiment 12, wherein one or more further subcategories are:

[0214] “clearly septic”: for a classification probability > 90 %;

[0215] “likely septic” for a classification probability greater than or equal to 75 and less than 90 %;

[0216] “likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; or

[0217] “clearly aseptic” for a classification probability < 10 %;

[0218] “unclear” for a classification probability between 25 to 75 %.

[0219] 14. A method of using the kit-of-parts of any one of embodiments 11 to 13 for diagnosing or monitoring the treatment of a periprosthetic joint infection in a subject.

[0220] 15. The method of embodiment 14, wherein a data set is generated, comprising the expression levels of the miRNAs from aseptic reference synovial fluid samples.

[0221] 16. The method of embodiment 14 or 15, wherein the classification model comprises the subsequent steps of: a) performing a multiple regression analysis of the expression level; b) calculating a classification probability score from the multiple regression analysis; and d) classifying the sample based on its classification probability score into one of at least two categories “septic” and “aseptic”. JB0001 P

[0222] -28-

[0223] 17. The method of embodiment 16, comprising an additional step of comparing the classification probability score with the classification probability scores observed in the aseptic reference synovial fluid samples to re-evaluate the reference values for the respective categories.

[0224] 18. The method of embodiment 16 or 17, wherein a classification probability score of > 75 % relates to the category “septic” and a classification probability score of < 25 % relates to the category “aseptic”.

[0225] 19. The method of any one of embodiments 16 to 18, wherein one or more further sub categories are:

[0226] “clearly septic”: for a classification probability > 90 %;

[0227] “likely septic” for a classification probability greater than or equal to 75 and less than 90 %;

[0228] “likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; or

[0229] “clearly aseptic” for a classification probability < 10 %;

[0230] “unclear” for a classification probability between 25 to 75 %.

[0231] EXAMPLES

[0232] The examples described herein are illustrative of the present invention and are not intended to be limitations thereof. Different embodiments of the present invention have been described according to the present invention. Many modifications and variations may be made to the techniques described and illustrated herein without departing from the scope of the invention. Accordingly, it should be understood that the examples are illustrative only and are not limiting upon the scope of the invention.

[0233] Example 1

[0234] Materials and Methods

[0235] Study population and experimental design

[0236] After institutional review board approval, the Orthopaedic Biobank Vienna Speising (O.B.V.S.) was retrospectively analyzed. Patients who underwent revision hip or knee TJR due to PJI or aseptic loosening between 06 / 2019 and 06 / 2022 and had pre- or intraoperatively collected synovial fluid stored in O.B.V.S. were analyzed. A total of 179 well characterized samples retrieved from 179 revision arthroplasties (157 knee / 22 hip) were further included. For the NGS-based untargeted and genome-wide discovery of miRNAs in synovial fluid, 46 samples were selected (20 septic, 20 aseptic and 6 JB0001 P

[0237] -29- osteoarthritic). Revision surgeries were classified as infected or non-infected as per the International Consensus meeting’s (ICM) 2018 infection criteria for PJI (Parvizi et al., 2018). On the same set of samples, a RT-qPCR analysis was set out to replicate the NGS results. Thereafter, to validate the candidates identified by small RNA-sequencing the established RT-qPCR protocol was applied to a total of 133 additional synovial fluid samples including 46 septic (35 %) and 87 aseptic (65 %).

[0238] Synovial fluid sample collection

[0239] Synovial fluid samples were taken pre- or intraoperatively and kept in an ice-filled transporter until they were frozen stored at -80 °C. Synovial fluid samples were aliquoted to 500 pL volume in 2 mL sterile microcentrifuge tubes without adding any cryoprotectants until further evaluation. The time for transportation and labelling had to be less than 30 minutes.

[0240] Total RNA extraction

[0241] Frozen synovial fluid samples were thawed to room temperature. After thawing, to remove cellular debris, samples were centrifuged at 12000 x g for 5 minutes at 4 °C. Total RNA was extracted from 100 pL synovial fluid using the miRNeasy Mini kit (Qiagen). All samples were diluted to 200 pL with NFW. Samples were then homogenized with 1000 pL Qiazol and rigorous mixing followed by incubation at room temperature for 15 minutes. 200 pL chloroform were added. Lysates were mixed vigorously and left at room temperature for 3 minutes. The lysates were then centrifuged at 12000 x g for 15 minutes at 4 °C. Precisely 600 pL aqueous phase were transferred to fresh tubes and 50 pL NFW were added. To improve precipitation, 7 pL glycogen (5 mg / mL) were added to each sample. A QIAcube liquid handling robot was used for binding to RNeasy Mini Spin Columns and washing steps. Total RNA was eluted in 30 pL NFW and stored at -80 °C until further analysis.

[0242] Small RNA-sequencing

[0243] Small RNA libraries were prepared using the RealSeq-Biofluids library kit for Illumina sequencing (RealSeq Biosciences) in accordance with the recommendations of the manufacturer. A total of 1 .5 ng RNA was diluted in 8.5 pL and 1 pL of miND® spike-ins were added. Adapter-ligated libraries were generated and subsequently circularized, reverse transcribed, and PCR-amplified using 21 cycles across all samples. DNA library size and yield was quantified with the DNA 1000 kit (Agilent Technologies). Libraries were diluted, pooled to equimolar concentrations and size selected with the BluePippin system using a 3 % agarose cassette (Sage Science). The multiplexes were quantified with the JB0001 P

[0244] -30- high Sensitivity DNA kit (Agilent Technologies). Small RNA-Seq was performed by the Next Generation Sequencing Facility at Vienna BioCenter Core Facilities (VBCF), member of the Vienna BioCenter (VBC), Austria on an Illumina HiSeqV4 with 50 bp single-end reads.

[0245] RT-qPCR analysis

[0246] Reverse transcription was carried out using the miRCURY RT kit (Qiagen), following the manufacturer’s instructions using 2 pL of synovial fluid total RNA as input . Quantitative PCR (qPCR) was performed using the miRCURY SYBR® Green Master Mix and commercial LNA-enhanced miRNA assays (Qiagen). With this setup, only mature miRNAs are detected. The final cDNA dilution was 1 :100. To ensure the quality of the generated data, synthetic spike-ins (Qiagen) were added in equimolar amounts before the respective step in the workflow to assess the efficacy of RNA isolation (UniSp4), reverse transcription (cel-miR-39-3p), and PCR (UniSp3). qPCRs were performed on a Light Cycler 480 II (Roche) with the following settings: 95 °C for 2 minutes (activation), 45 cycles of 95 °C for 10 seconds, and 56 °C for 60 seconds. Melting curves were generated using continuous acquisition between 55 and 99 °C. Cq values were calculated with the 2nd derivative maximum method and normalized using hsa-miR-22-3p.

[0247] Small RNA-sequencing data analysis

[0248] Raw sequencing reads derived were adapter trimmed and filtered for low-quality reads (Q < 30). miRNA identification and annotation were per- formed using the miRNA NGS Discovery (miND) pipeline. Read counts were normalized to the total number of miRNAs reads detected per sample to obtain the reads per million (RPM) for each miRNA and sample. Exploratory data analysis was performed using unsupervised cluster analysis, and differential expression analysis was performed using the statistical methods available under the edgeR package in R / Bioconductor. Independent filtering was performed using DESeq2 to optimize cut-offs for filtering low-abundant miRNAs. A false- discovery rate (FDR) cut-off of 0.05 was applied to identify statistically significant differences in miRNAs levels between septic and aseptic samples.

[0249] Multivariate data analysis

[0250] Multiple logistic regression analysis was performed using the R package“Classification And REgression Training (caret)”. For each set of biomarkers (i.e. combination of miRNAs), input data were preprocessed by selecting complete sets without missing measurements. In the next step, the dataset was partitioned into a training (60 %) and test set (40 %) using createDataPartition() function using a fixed seed JB0001 P

[0251] -31- by set.seed(). For each set of biomarkers, a model was trained by i) evaluating the effect of model tuning parameters on performance using “resampling” using 10-fold cross- validation, ii) choosing the optimal model across these parameters (i.e. selection of an equation), and iii) estimating model performance from the training set. The resulting model parameters were applied to the independent validation set to determine area- under-the-curve (AUC) from ROC analysis.

[0252] Results

[0253] NGS-based discovery of synovial fluid derived microRNA biomarkers of periprosthetic joint infection (PJI) Small RNA-sequencing was performed for untargeted and genome-wide analysis of microRNAs in synovial fluid samples from aseptic (n=20), septic (n=20), and osteoarthritic (n=6) knees. This “discovery” cohort included 43 % male individuals and median age was 73 years with no significant differences between the infected and control groups (Table 1). Median BMI in the cohort was 28.2 kg / m2and matched between groups.

[0254] JB0001 P

[0255] -32-

[0256] Table 1 : Discovery Cohort Demographics

[0257] Age N Ratio (%) Mean SD Median IQR Min Max Range

[0258] 46 100 % 70.3 9.6 73.0 13.551.089.038.0 aseptic2043 % 70.0 7.9 73.0 9.5 56.080.024.0

[0259] Infection septic 2043 % 70.4 11.474.0 15.551.089.038.0

[0260] OA 6 13 % 70.7 10.072.0 9.0 53.081.028.0

[0261] Knee 46 100 % 70.3 9.6 73.0 13.551.089.038.0

[0262] Joint

[0263] Hip 0 0 % male 2043 % 71.4 8.7 73.0 9.0 53.083.030.0

[0264] Gender female 2657 % 69.4 10.471.5 18.351.089.038.0

[0265] BMI N Ratio (%) Mean SD Median IQR Min Max Range

[0266] 46 100 % 29.8 6.1 28.2 8.0 19.1 50.231.1 aseptic2043 % 29.3 5.0 29.0 6.8 19.1 37.8 18.7

[0267] Infection septic 2043 % 30.7 7.6 27.4 8.7 21.850.228.4

[0268] OA 6 13 % 28.5 3.8 27.6 3.4 24.034.9 10.9

[0269] Knee 46 100 % 29.8 6.1 28.2 8.0 19.1 50.231.1

[0270] Joint

[0271] Hip 0 0 % male 2043 % 28.7 2.8 28.2 3.6 24.1 34.09.9

[0272] Gender female 2657 % 30.6 7.7 28.2 9.9 19.1 50.231.1

[0273] Quality of small RNA-seq data was assessed based on total read count, microRNA read count as well as the number of detected microRNAs, and RNA spike-in and NGS (miND®) spike-in recovery. The median total read count was 6 million reads per sample and did not differ between groups. The median microRNA read count was 450 thousand reads per sample. Sequencing depth allowed detection of up to 800 miRNAs per sample (median 540), of which 250 miRNAs consistently showed read counts > 10. No significant difference in the number of detectable miRNAs was observed between aseptic, septic, and osteoarthritic synovial fluids. The recovery of RNA spike-ins as well as NGS spikeins (miND®) was high for synovial fluid.

[0274] Due to the consistently high NGS data quality all replicates were included for statistical comparison of septic (n=20) vs aseptic (n=20). After adjusting for multiple testing (Benjamini-Hochberg false discovery rate (FDR) < 0.05), a total of 132 microRNAs were found differentially expressed (Fig. 1). Of these, 62 (47 %) showed significant up- JB0001 P

[0275] -33- regulation in infected joints, while 70 (53 %) were down-regulated. MicroRNAs up- regulated in response to infection showed effect sizes of >20-fold (miR-223-3p, log2FC=4.54), while down-regulations were moderate with a maximum of 6-fold reduction (miR-4448, log2FC=-2.65) compared to aseptic synovial fluid.

[0276] A subset of 18 microRNAs with FDR < 0.05, a log2FC >|1 |, and logCPM>4 was selected (Fig. 1) and used for further analysis: clustering of all septic and aseptic samples as well as 6 osteoarthritic synovial fluid samples was performed using this set of 18 miRNAs. The resulting heatmap revealed distinct grouping of most septic samples based on their miRNA profile, while osteoarthritic and aseptic synovial fluids formed a second homogeneous cluster (Fig. 2).

[0277] Fig. 1 and Fig. 2 show an NGS-based analysis of microRNA biomarkers in synovial fluid from aseptic and septic knee arthroplasty. Fig. 1 shows a volcano plot depicting Iog2 fold change on the x-axis and adjusted p-value expressed as false-discovery rate (FDR) on the y-axis for the comparison of microRNA levels in synovial fluid from septic (n=20) and aseptic (n=20) knee arthroplasty. miRNAs with FDR < 0.05 are labelled as black dots. A subset of 18 miRNAs selected for further validation is highlighted (x used as symbol). Fig. 2 shows a clustering of aseptic, septic, and osteoarthritic synovial fluid samples based on a microRNA expression pattern from the NGS-based analysis of microRNA biomarkers in synovial fluid from aseptic and septic knee arthroplasty. Heatmap based on reads per million (RPM) values of 18 miRNAs selected for further validation. Septic (n=20), aseptic (n=20) and osteoarthritic (n=6) synovial fluid samples are shown.

[0278] JB0001 P

[0279] -34-

[0280] Table 2: Validation Cohort Demographics

[0281] Age N Ratio (%) Mean SD Median IQR Min Max Range

[0282] 133 100 % 68.8 11.570.0 17.029.092.063.0 aseptic87 65 % 66.7 11.968.0 18.529.087.058.0

[0283] Infection septic 46 35 % 72.8 9.4 72.0 13.051.092.041.0

[0284] Knee 111 83 % 68.8 10.469.0 15.043.092.049.0

[0285] Joint

[0286] Hip 22 17 % 68.7 16.1 72.0 25.029.087.058.0 male 53 40 % 68.6 11.768.0 15.029.092.063.0

[0287] Gender female 80 60 % 68.9 11.471.0 17.343.087.044.0

[0288] BMI N Ratio (%) Mean SD Median IQR Min Max Range

[0289] 133 100 % 30.2 6.4 29.0 7.1 17.754.036.3 aseptic87 65 % 29.9 5.7 29.0 6.3 17.753.335.6

[0290] Infection septic 46 35 % 30.7 7.5 28.7 8.6 20.554.033.5

[0291] Knee 111 83 % 30.3 6.3 28.5 6.8 18.854.035.2

[0292] Joint

[0293] Hip 22 17 % 29.7 6.6 29.4 8.9 17.743.1 25.4 male 53 40 % 30.3 5.9 28.7 5.1 20.554.033.5

[0294] Gender female 80 60 % 30.2 6.7 29.1 8.1 17.753.936.2

[0295] RT-qPCR enables successful technical replication of NGS results.

[0296] Based on the promising results obtained from biomarker discovery, the identical set of 46 samples was analyzed with RT-qPCR. For this, miR-22-3p was identified as a robust endogenous reference for the standardization of RT-qPCR data. The delta Cq- values (dCq) obtained after miR-22-3p normalization ranged of 15 logs (Iog2) with miR-223-3p and miR-19a-3p being the most abundant miRNAs in aseptic synovial fluids, while miRNAs of the miR-548 family (miR-548d-5p and miR-548ay-5p) showed lowest abundance. Ranking of miRNAs based on their baseline abundance in aseptic synovial fluid mostly, but not always coincided with the NGS data, illustrating that each analytical method may be limited by selection bias such as ligation- or amplification-bias. Next, the area-under-the-curve (AUC) from ROC analysis obtained by NGS was correlated vs qPCR for each of the 18 miRNAs (Fig. 3) to assess the reproducibility of NGS results by qPCR. A high correlation was observed as well as agreement of the absolute values for all miRNAs, which confirmed the suitability of RT-qPCR for the performance of further independent validation studies. JB0001 P

[0297] -35-

[0298] Fig. 3 shows the replication of NGS data for 18 microRNAs by RT-qPCR, specifically Fig. 3 shows the comparison of area-under-the-curve (AUC) values for the classification of septic vs aseptic synovial fluid samples obtained by RT-qPCR (y-axis) in comparison to NGS data (x-axis).

[0299] MicroRNA PJI biomarker candidates may be released from infiltrating immune cells or joint tissue.

[0300] To better understand the mechanism underlying the observed differences in synovial fluid miRNAs between septic and aseptic joints, their cell-type dependent expression patterns were analyzed. The publicly available Fantom5 data repository was used to obtain quantitative NGS (logCPM) data for the miRNAs of interested for 21 different cell types, including innate immunity (CD14 Monocytes, Dendritic cells, Macrophages, Neutrophils, and Mast Cells), adaptive immunity (CD19+ B-Cells, NK-cells, CD4+ and CD8+ T-cells), and cell types with presence in joint tissue (endothelial cells, synoviocytes, chondrocytes, adipocytes, preadipocytes, osteoblasts, fibroblasts, myoblasts, mesenchymal stem cells, and skeletal muscle cells). Clustering of miRNAs based on their cell-type dependent expression revealed three distinct clusters corresponding to miRNAs enriched in adaptive immune cell populations (7 miRNAs), innate immunity cell populations (8 miRNAs) and joint tissue (3 miRNAs). These data suggest that both the infiltration of different immune cell populations during the onset and progression of joint infection as well as the response of host cells in the joint result in the altered presence of cell-free miRNAs.

[0301] Independent validation of biomarker candidates by RT-qPCR

[0302] To validate the candidates identified by small RNA-sequencing the established RT- qPCR protocol was applied to a total of 133 additional synovial fluid samples including 46 samples (35 %) that were classified as septic (Table 2). Samples were obtained from patients with median age of 70 years, both genders (40 % male), hip and knee joints (83 % knee). No difference was observed between infection groups for age, type of joint, and gender. BMI showed a median level of 29 kg / m2 and did not differ between infection groups.

[0303] Following analysis of 18 miRNA candidates in the independent samples, the effect size (Iog2 fold change between septic vs aseptic) and adjusted p-value (Kruskal-Wallis with Benjamini-Hochberg correction) were obtained and compared side-by-side to results from the discovery cohort (Table 3). It was found that all miRNAs could be successfully validated with similar effects, sizes and significance levels. JB0001 P

[0304] -36-

[0305] Table 3: microRNA biomarker validation results mirrnRNA inlo9FC Discoverylo9FCValidation p.adj p.adj

[0306] (septic / aseptic, n=20 / 20) (septic / aseptic, n=46 / 87) Discovery Validation hsa-miR-142-3p 2.39 2.23 0.0001 <0.0001 hsa-miR-142-5p 2.68 2.99 <0.0001 <0.0001 hsa-miR-151a-3p -1.38 -1.27 0.0006 <0.0001 hsa-miR-191-5p 1.49 1.45 0.0012 <0.0001 hsa-miR-19a-3p 1.64 0.97 0.0001 0.0006 hsa-miR-214-3p -2 -2.15 0.0002 <0.0001 hsa-miR-223-3p 4.41 4.93 0.0001 <0.0001 hsa-miR-30d-5p 0.68 0.47 0.007 0.0106 hsa-miR-338-5p 3.7 4.66 0.0002 <0.0001 hsa-miR-345-5p 1.57 1.72 <0.0001 <0.0001 hsa-miR-3615 2.48 2.29 0.0001 <0.0001 hsa-miR-424-3p 2.11 -1.73 <0.0001 <0.0001 hsa-miR-425-3p 1.99 2.19 0.0002 <0.0001 hsa-miR-425-5p 1.79 1.42 <0.0001 <0.0001 hsa-miR-505-3p 2.21 2.54 <0.0001 <0.0001 hsa-miR-548ay-5p2.41 1.68 <0.0001 <0.0001 hsa-miR-548d-5p 2.53 1.98 0.0001 <0.0001 hsa-miR-629-5p 2.16 1.94 <0.0001 <0.0001 logFCs and p values calculated with non-parametric tests from qPCR data (dCp to mir-22)

[0307] Multivariate analysis to improve the diagnostic performance of individual microRNA biomarkers for PJI Following successful independent validation of miRNA biomarker candidates, the diagnostic performance of combinations of smaller groups of microRNAs (so-called “models”) selected from the 18 candidates was assessed, since it was observed that not all miRNA candidates were highly correlated in the entire cohort consisting of 173 samples (Table 4,). To generate the classification models a biostatistical workflow was established (Fig. 4), which consisted of the following steps for each uni- and multivariate model that was tested: i) data was randomly partitioned into a training (60 %) and test (40 %) data set; ii) the training data were used to build a multiple logistic regression model and define model parameters, which were then iii) applied to the independent test set to determine the performance using ROC analysis (AUC-values). JB0001 P

[0308] -37-

[0309] Table 4: Total Cohort Demographics

[0310] Age N Ratio (%) Mean SD Median IQR Min Max Range

[0311] 179 100 % 69.2 11.071.0 16.029.092.063.0 aseptic 10760 % 67.3 11.369.0 18.029.087.058.0

[0312] Infection septic 66 37 % 72.0 10.1 72.5 13.851.092.041.0

[0313] OA 6 3 % 70.7 10.072.0 9.0 53.081.028.0

[0314] Knee 15788 % 69.2 10.270.0 15.043.092.049.0

[0315] Joint

[0316] Hip 22 12 % 68.7 16.1 72.0 25.029.087.058.0 male 73 41 % 69.4 11.070.0 13.029.092.063.0

[0317] Gender female 10659 % 69.0 11.1 71.0 16.843.089.046.0

[0318] BMI N Ratio (%) Mean SD Median IQR Min Max Range

[0319] 179 100 % 30.1 6.3 28.9 7.2 17.754.036.3 aseptic 10760 % 29.8 5.6 29.0 6.5 17.753.335.6

[0320] Infection septic 66 37 % 30.7 7.5 28.1 8.9 20.554.033.5

[0321] OA 6 3 % 28.5 3.8 27.6 3.4 24.034.9 10.9

[0322] Knee 15788 % 30.1 6.3 28.4 7.0 18.854.035.2

[0323] Joint

[0324] Hip 22 12 % 29.7 6.6 29.4 8.9 17.743.1 25.4 male 73 41 % 29.8 5.2 28.4 4.9 20.554.033.5

[0325] Gender female 10659 % 30.2 6.9 29.0 9.1 17.753.936.2

[0326] Fig. 4-6 show the development of multivariate microRNA models for diagnosis of periprosthetic joint infections in knee and hip. Fig. 4 shows the pipeline for the development of classification models: for each uni- or multivariate model, the data were partitioned into a training (60 % or 101 samples) and test (40 % or 72 samples) set. The training set was used to develop a classification model using cross-validation to increase robustness. Optimized and model parameters were then fixed and applied to the independent test set to assess performance (ROC analysis). Fig. 5 shows the uni- (single miRNA) and multivariate (models with 2, 3, or 4 miRNAs) performance of classification models on the test data set shown as area-under-the-curve (AUC) from ROC analysis.

[0327] The assessment of multivariate classification models utilizing 1 to 4 miRNAs selected from the group of 18 miRNAs showed that the median AUC could be improved from < 0.9 for univariate models to > 0.9, even > 0.95, for multivariate models consisting of 3 or 4 miRNAs (Fig. 5). To reduce the risk of overfitting and because very good performance (AUC > 0.9) was already observed for models consisting of 2 miRNAs, JB0001 P

[0328] -38- further bivariate models were investigated. Fig. 6 visualizes the AUC values from the test data sets obtained for bivariate models (i.e. consisting of 2 miRNAs) including one fixed and one variably selected miRNA candidate. Univariate performances for the fixed candidate are superimposed as a black dot. This analysis enables the identification of key miRNAs required to achieve good classification performance such as miR-338-5p, miR-223-3p, miR-505-3p, and miR-142-5p, which consistently achieve performances of > 0.9 AUC in the test data sets.

[0329] Fig. 6 shows the performances of bivariate miRNA models for P JI. Performances of bivariate classification models in the test data set are shown as AUC values from ROC analysis. For each miRNA, its univariate AUC-value (blue dot, single-miR) as well as the AUC-values obtained from pairing the miRNA with each other of the 17 remaining miRNAs (red box, paired model) are shown.

[0330] Fig. 7 shows the ROC curve obtained for the bivariate classification model consisting of miR-142-5p and miR-338-5p for the test set.

[0331] The combination of miR-338-5p, which is predominately expressed in neutrophils, and miR-142-5p, which is predominately expressed in B- and T-cells gave a performance of 0.99, which was identical to the performance of a-Defensin in the same data set.

[0332] Example 2

[0333] Results

[0334] Regulation of previously reported sepsis-associated miRNAs in synovial fluids from periprosthetic joint infections and aseptic controls

[0335] Fig. 8 shows several reported biomarkers of blood-based sepsis are not differentially expressed in synovial fluids from periprosthetic joint infections. Volcano plot illustrating the Iog2-transformed fold change (x-axis) of miRNAs between septic and aseptic joints, as well as the significance level (false discovery rate, FDR). A set of miRNAs, which was previously reported to be associated with blood-based sepsis is highlighted and labelled. Filled black dots highlight miRNAs with significant differences between septic and aseptic synovial fluid.

[0336] Example 3

[0337] Results

[0338] The classification of microRNAs from synovial fluid samples works surprisingly well without normalizing the expression levels to a reference miRNA.

[0339] The analysis performed in Example 1 was repeated, specifically (univariate and multivariate classification models in parallel for the miR-22 normalized data and the raw JB0001 P

[0340] -39- data. The data were subsequently compared with each other. The reanalysis of both expression data (miR-22 and raw data) was necessary to ensure that the division of the samples into training (60 %) and test set (40 %) is identical (in the case of smaller data sets, random effects due to the group division can play a role).

[0341] One can clearly see that classification models without normalization of the expression levels to a reference miRNA work very well, equivalent, possibly even better (Fig. 10 and Fig. 11).

[0342] Fig. 9-11 show the impact of assay normalization on the diagnostic performance of uni- and bivariate miRNA models. Fig. 9 shows the classification performances in the form of AUC values determined in the test set of bivariate miRNA classification models. For each miRNA, its individual (univariate) AUC-value (black dot, Single miR) as well as the AUC-values obtained from pairing the miRNA with each other of the 17 remaining miRNAs (black box, Paired model) are shown (Fig. 9). Furthermore, a side-by-side comparison of the non-normalized (raw) and miR-22 normalized AUC-values for univariate (Fig. 10) and bivariate (Fig. 11) classification models was performed.

[0343] Example 4

[0344] Materials and Methods

[0345] Sample Selection and Data Preparation

[0346] Feature selection was performed using a dataset (n=170; 92 aseptic and 78 septic) comprising samples with definitive post-operative diagnoses (aseptic or septic periprosthetic joint infection). Synovial fluid levels of 18 microRNA biomarkers were analyzed by RT-qPCR and expression values were normalized using the ACq method with hsa-miR-22-3p as the endogenous reference control, followed by standardization (zero mean, unit variance) using the step_normalize() function from the recipes package to enable cross-feature comparisons. Initial data splitting into 80% training and 20% test data was performed using in itial_split() from the rsample package with stratified sampling to maintain class balance.

[0347] Model-Specific Feature Selection Strategy

[0348] Since different machine learning algorithms utilize features through fundamentally different mechanisms, a model-specific feature selection approach rather than applying uniform feature selection across all models was implemented. Feature set sizes ranging from 1 to 6 miRNAs were evaluated for each algorithm to balance clinical practicality with predictive performance. Prior to model-specific selection, candidate miRNAs underwent univariate statistical screening using the Wilcoxon rank-sum test (wilcox.testQ from base JB0001 P

[0349] -40-

[0350] R) with Benjamini-Hochberg false discovery rate correction (p.adjust() with method = "BH", adjusted p < 0.05).

[0351] Linear Models

[0352] For linear algorithms (logistic regression, LASSO, ridge regression, and elastic net), feature importance was determined by coefficient magnitudes following model training and hyperparameter optimization. Models were implemented using logistic_reg() from the parsnip package with the „glm“ engine for simple logistic regression and „glmnet“ engine for regularized methods. For regularized models (LASSO, ridge, elastic net), coefficients were extracted from the optimal model identified through cross-validationbased hyperparameter tuning using tune_grid() from the tune package. The tidy() function from the broom package was used to extract coefficient estimates, and features were ranked by absolute coefficient values, with the top-performing features selected for each specified panel size.

[0353] Tree-Based Models

[0354] For tree-based algorithms (random forest, XGBoost, and decision trees), feature importance was calculated using algorithm-specific impurity-based metrics. Random forest models were implemented using rand_forest() from parsnip with the „ranger“ engine, with importance assessed using Gini impurity decrease (importance = „impurity“). XGBoost models used boost_tree() with the „xgboost“ engine, employing gain-based importance measures, and decision trees utilized decision_tree() with the „rpart“ engine for node impurity reduction. Following model training with optimal hyperparameters identified via tune_grid(), feature importance scores were extracted using the vip::vi() function from the vip package. Features were ranked by their respective importance scores and the highest-ranking features were selected for each panel size.

[0355] Other Algorithms

[0356] For remaining algorithms (support vector machines, neural networks, k-nearest neighbors, linear discriminant analysis, quadratic discriminant analysis, and naive Bayes), which do not provide inherent feature importance measures, features were selected based on univariate statistical rankings. Support vector machines were implemented using svm_rbf() with the „kernlab“ engine, neural networks using mlp() with the „nnet“ engine, k-nearest neighbors using nearest_neighbor() with the „kknn“ engine, discriminant analysis using discrim_linear() and discrim_quad() with the „MASS“ engine, and naive Bayes using naive_Bayes() with the „naivebayes“ engine. The statistically JB0001 P

[0357] -41- significant miRNAs were ranked by adjusted p-values from the Wilcoxon rank-sum test, and the top-performing features were selected for each specified panel size.

[0358] Cross-Validation and Model Training

[0359] Feature selection was performed using 10-fold cross-validation repeated 3 times on the training dataset to ensure stability and prevent overfitting. Cross-validation resampling was implemented using vfold_cv() from the rsample package. Workflows were constructed using workflow() from the workflows package, combining preprocessing recipes created with recipe() from the recipes package and model specifications from parsnip. For each model type and feature set size combination, the selected features were later used to train the respective algorithm with hyperparameter optimization via grid search using tune_grid(), optimizing area under the receiver operating characteristic curve (AUC) calculated with roc_auc() from the yardstick package as the primary performance metric. Final model selection was performed using select_best() from the tune package.

[0360] Feature Usage Analysis

[0361] Feature usage frequency was calculated across all model types and panel sizes to identify consistently important miRNAs across different algorithmic approaches.

[0362] Results

[0363] Example 4 describes an analysis performed to identify consistently important miRNAs across different algorithmic approaches. This analysis revealed miRNAs that demonstrated robust discriminatory capacity independent of the specific machine learning approach employed, resulting in the most stable biomarkers for clinical translation. miR-338-5p emerged as the most robust feature, selected in 67 of 78 possible model-size combinations (86 %). It was consistently picked by LASSO (5 / 6 sizes), Ridge (5 / 6), Elastic Net (4 / 6), Random Forest (5 / 6), XGBoost (6 / 6) and across all other classifier types, underscoring its broad discriminatory power. miR-223-3p followed closely, with 61 total selections (78 %), showing balanced representation across linear models (4-5 picks each) and strong inclusion by tree-based methods (3-5 picks). miR-505-3p was chosen 51 times (65 %), with particularly high preference in Random Forest (6 / 6) and moderate uptake by linear and non-linear classifiers (3-4 picks each). JB0001 P

[0364] -42- miR-142-5p (43 / 78, 55 %) demonstrated a bias toward linear approaches — being selected by all six sizes in logistic regression, LASSO, and Elastic Net - but was less favored by Random Forest (0 / 6) and shallow decision trees (2 / 6). miR-3615 (26 / 78, 33 %) contributing robustly across both linear and tree-based pipelines (1-4 picks per algorithm). miR-424-3p (10 / 78, 13 %), miR-151a-3p (7 / 78, 9 %), and miR-191-5p (3 / 78, 4 %) were emphasized by regularized linear models or individual non-parametric classifiers. Table 5 provides an overview of the tested miRNAs.

[0365] Taken together, from a set of 18 miRNAs that discriminated between septic and aseptic samples, three miRNAs - miR-338-5p, miR-223-3p, and miR-505-3p - demonstrated both high selection frequency and cross-algorithm consistency when tested with 13 different algorithms, marking them as prime candidates for a minimal yet robust diagnostic signature. Two additional miRNAs hsa-miR-142-5p and hsa-miR-3615 were most relevant in specific model types.

[0366] JB0001 P

[0367] -43-

[0368] Table 5. Feature Usage Matrix, demonstrating the number of times each miRNA was selected by each model across all feature set sizes (1-6 features): JB0001 P

[0369] -44-

[0370] REFERENCES

[0371] Deng, Y. et al. (2024). Aseptic loosening around total joint replacement in humans is regulated by miR-1246 and miR-6089 via the Wnt signalling pathway. Journal of Orthopaedic Surgery and Research, 19(1), 94.

[0372] Goh et al. (2022). Diagnosis and Treatment of Culture-Negative Periprosthetic Joint Infection. The Journal of Arthroplasty, 37(8): 1488-1493.

[0373] Gutmann et al. (2022). Association of cardiometabolic microRNAs with COVID-19 severity and mortality. Cardiovascular Research, 118(2): 461-474.

[0374] Huang et al. (2020). Metagenomic next-generation sequencing of synovial fluid demonstrates high accuracy in prosthetic joint infection diagnostics. Bone & Joint Research, 9(7): 440-449.

[0375] Indelli et al. (2021). Next generation sequencing for pathogen detection in periprosthetic joint infections. EFORT Open Reviews, 6(4): 236-244.

[0376] Kalbian etal. (2020). Culture-negative periprosthetic joint infection: prevalence, aetiology, evaluation, recommendations, and treatment. International Orthopaedics, 44(7): 1255-1261.

[0377] Khamina et al. (2022). A MicroRNA Next-Generation-Sequencing Discovery Assay (miND) for Genome-Scale Analysis and Absolute Quantitation of Circulating MicroRNA Biomarkers. International Journal of Molecular Sciences, 23(3): 1226.

[0378] Luftinger et al. (2021). Predictive Antibiotic Susceptibility Testing by Next-Generation Sequencing for Periprosthetic Joint Infections: Potential and Limitations. Biomedicines, 9(8): 910.

[0379] McNally et al. (2021). The EBJIS definition of periprosthetic joint infection. The Bone & Joint Journal, 103-B(1): 18-25.

[0380] Omar et al. (2017). Transcriptome-Wide High-Density Microarray Analysis Reveals Differential Gene Transcription in Periprosthetic Tissue From Hips With Chronic Periprosthetic Joint Infection vs Aseptic Loosening. The Journal of Arthroplasty, 32(1), 234 - 240

[0381] Paksoy et al. (2024). MicroRNA expression analysis in peripheral blood and soft-tissue of patients with periprosthetic hip infection. Bone & Joint Open, 5(6): 479-488.

[0382] Parvizi et al. (2018). The 2018 Definition of Periprosthetic Hip and Knee Infection: An Evidence-Based and Validated Criteria. The Journal of Arthroplasty, 33(5): 1309- 1314. e2. JB0001 P

[0383] -45-

[0384] Parvizi et al. (2011 ). New Definition for Periprosthetic Joint Infection: From the Workgroup of the Musculoskeletal Infection Society. Clinical Orthopaedics & Related Research, 469(11): 2992-2994.

[0385] Pascual etal. (2024). Potential value of a rapid syndromic multiplex PCR forthe diagnosis of native and prosthetic joint infections: a real-world evidence study. Journal of Bone and Joint Infection, 9(1): 87-97.

[0386] Sharma et al. (2020). Comparative analysis of 23 synovial fluid biomarkers for hip and knee periprosthetic joint infection detection. Journal of Orthopaedic Research, 38(12): 2664-2674. Yilmaz et al. (2023). Diagnosis of Periprosthetic Joint Infection: The Utility of Biomarkers in 2023. Antibiotics, 12(6): 1054.

Claims

JB0001 P-46-CLAIMS1 . An in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of: a. determining the expression level(s) of i. 1 , 2, 3, 4, or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial fluid sample of said subject; ii. optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p; and b. comparing the expression level(s) of (a) with a predetermined reference range of expression levels of the same miRNA(s) from an aseptic reference group, wherein the expression level(s) of (a) are classified as category septic if the expression level(s) of (a) are different from the reference range, and wherein the expression level(s) are classified as category aseptic if the expression level(s) of (a) are within the reference range.

2. An in vitro method of diagnosing a periprosthetic joint infection in a subject comprising the sequential steps of: a) determining the expression level(s) of(i) 1 , 2, 3, 4, or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial fluid sample of said subject;(ii) optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p; b) determining the expression level of a standard miRNA having low expression stability and high average abundance; c) determining a ratio of the expression level of said standard miRNA of (b) and the miRNA(s) of (a);JB0001 P-47- d) determining a ratio of the expression level of the same standard miRNA from an aseptic reference group and the expression level(s) of the same miRNA(s) from an aseptic reference group; and e) comparing said ratios, wherein the ratio of (c) is classified as category septic if it is different from ratio of (d), and wherein the ratio of (c) is classified as category aseptic if it is similar to (d).

3. The method of claim 2, wherein the expression stability and average abundance is identified with a mathematical model of gene expression; specifically, the standard miRNA is selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p.

4. The method of any one of claim 1 to 3, wherein the synovial fluid sample is from a joint of a subject on which arthroplasty was performed.

5. The method of any one of claims 1 to 4, wherein the synovial fluid sample is from a subject with joint pain.

6. The method of any one of claims 1 to 5, wherein the classification is performed using a classification model.

7. The method of any one of claims 1 to 6, wherein the classification is performed using a software comprising a classification model.

8. The method of claim 7, wherein said software executes steps (a) and (b) of claim 1 or steps (a) to (e) of claim 2.

9. The method of any one of claims 6 to 8, wherein the classification model is selected from the group consisting of multivariate classification models, logistic regression models, support vector machine models, and decision tree models.

10. A method for monitoring the treatment of a periprosthetic joint infection using the method of any one of claims 1 to 9.JB0001 P-48-11. A kit-of-parts comprising: a) detection reagents for detecting the expression level of at least 1 , 2, 3, 4, or 5 microRNAs selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in a synovial sample; b) optionally detection reagents for detecting the expression level of one or more of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p, c) optionally detection reagents for detecting the expression level of one or more of standard miRNAs, selected from the group consisting of miR-22-3p, miR-92a-3p, miR-15a-5p, miR-103a-3p, miR-320a-3p, miR-16-5p, miR-23a-3p, miR-19b-3p, miR-221-3p, miR-25-3p, miR-21-5p, and miR-93-5p; and d) a software comprising one or more classification model(s) for classification of the subject’s sample into the categories “septic” and “aseptic”.

12. The kit-of-parts of claim 11 , wherein the classification model provides a classification probability score for classifying the sample wherein the category “septic” has a classification score of > 75” and “aseptic” has classification score of < 25 %.

13. The kit-of-parts of claim 12, wherein one or more further sub-categories are: “clearly septic”: for a classification probability > 90 %;“likely septic” for a classification probability greater than or equal to 75 and less than 90 %;“likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; or“clearly aseptic” for a classification probability < 10 %;“unclear” for a classification probability between 25 to 75 %.

14. A method for diagnosing a periprosthetic joint infection or monitoring the treatment of a periprosthetic joint infection using the kit-of-parts of any one of claims 11 to 13.

15. The method of claim 14, wherein a data set is generated, comprising the expression levels of the miRNAs from aseptic reference synovial fluid samples.JB0001 P-49-16. The method of claim 14 or 15, wherein the classification model comprises the subsequent steps of: a) performing a multiple regression analysis of the expression level; b) calculating a classification probability score from the multiple regression analysis; and c) classifying the sample based on its classification probability score into one of at least two categories “septic” and “aseptic”.

17. The method of claim 16, comprising an additional step of comparing the classification probability score with the classification probability scores observed in the aseptic reference synovial fluid samples to re-evaluate the reference values for the respective categories.

18. The method of claim 16 or 17, wherein a classification probability score of > 75 % relates to the category “septic” and a classification probability score of < 25 % relates to the category “aseptic”.

19. The method of any one of claims 16 to 18, wherein one or more further sub categories are:“clearly septic”: for a classification probability > 90 %;“likely septic” for a classification probability greater than or equal to 75 and less than 90 %;“likely aseptic”, for a classification probability greater than 10 and less than or equal to 25 %; or“clearly aseptic” for a classification probability < 10 %;“unclear” for a classification probability between 25 to 75 %.

20. A method of treating a subject suffering from complications following total joint replacement surgery, said method comprises the sequential steps of a) providing a synovial fluid sample from said subject, b) determining the expression level(s) of i. 1 , 2, 3, 4, or 5 microRNAs (miRNAs) selected from the group consisting of miR-338-5p, miR-142-5p, miR-223-3p, miR-505-3p, and miR-3615 in the synovial fluid sample;JB0001 P-50- ii. optionally at least one further miRNA, selected from the group consisting of miR-19a-3p, miR-30d-5p, miR-142-3p, miR-151a-3p, miR-191-5p, miR-214-3p, miR-345-5p, miR-424-3p, miR-425-3p, miR-425-5p, miR-548d-5p, miR-548ay-5p, and miR-629-5p; and c) comparing the expression level(s) of (b) with a predetermined reference range of expression levels of the same miRNA(s) from an aseptic reference group, wherein the expression level(s) of (b) are classified as category septic if the expression level(s) of (b) are different from the reference range, and wherein the expression level(s) are classified as category aseptic if the expression level(s) of (b) are within the reference range, d) performing a revision surgery and / or antibiotic treatment when the sample is classified as septic.

Images