WO2018197874A1

WO2018197874A1 - Clostridium difficile biomarkers and uses thereof

Info

Publication number: WO2018197874A1
Application number: PCT/GB2018/051090
Authority: WO
Inventors: Dr Tanya M. MONAGHAN; Dr Ola H. NEGM
Original assignee: University of Nottingham
Current assignee: University of Nottingham
Priority date: 2017-04-28
Filing date: 2018-04-25
Publication date: 2018-11-01
Anticipated expiration: 2019-10-28
Also published as: GB201706786D0

Abstract

The present invention provides a method of determining the severity of a Clostridium difficile infection in a subject, the method comprising the steps of: (i) obtaining a sample from a subject having or suspected of having Clostridium difficile infection; (ii) detecting one or more biomarkers in the sample, wherein the one or more biomarkers are selected from the group consisting of antibodies to the Clostridium difficile toxin A or a fragment thereof, antibodies to the Clostridium difficile toxin B or a fragment thereof, antibodies to a Clostridium difficile surface layer protein or a fragment thereof, and IFNγ; and (iii) determining, from the results in (ii), whether the subject has a mild Clostridium difficile infection or a severe Clostridium difficile infection.

Description

CLOSTRIDIUM DIFFICILE BIOMARKERS AND USES THEREOF

The present invention relates to novel biomarkers that are able to distinguish between different clinical states of Clostridium difficile infection, and methods of using the same.

Clostridium difficile is a significant nosocomial and community-acquired pathogen and Clostridium difficile infection is amongst the most serious healthcare complications. Clostridium difficile infection causes a spectrum of clinical presentations, ranging from an asymptomatic carrier state to more severe diarrhoea and fatal fulminant colitis. Clostridium difficile infection is the leading worldwide cause of hospital-acquired and antibiotic-associated diarrhoea, imposing a significant financial burden on health service providers. In Europe alone, the annual estimated costs for the management of Clostridium difficile infection amount to€3 billion.

Management of Clostridium difficile infection is complicated due to antimicrobial resistance, recurring infections, and strikingly the inability to reliably differentiate between acute infection and those who may be simply carrying a toxigenic strain of Clostridium difficile but may have diarrhoea attributed to another cause . For example, in inflammatory bowel disease, prevalence rates of Clostridium difficile infection are reportedly higher but it remains unclear if Clostridium difficile is contributing to the inflammatory insult or whether it is acting only as a bystander. Pathogenic strains of Clostridium difficile produce two major exotoxins A and B, which are the primary virulence factors contributing to the pathogenesis of Clostridium difficile infection. Presently, standard diagnosis of Clostridium difficile infection is by detection of these toxins in stool samples. The current approach for treating most Clostridium difficile infection involves administration of antibiotics, typically metronidazole as a starting point and vancomycin in more severe cases of infection. Accurate diagnosis of Clostridium difficile infection as well as risk stratification of infected patients is critical for effective management. However, there are currently no reliable tools to determine and characterise the severity of Clostridium difficile infection in patients, meaning that patients are not always provided with the most suitable care and/or therapy. Ultimately this can result in increased antimicrobial resistance, recurring infections and the need for more emergency invasive treatment. Current conventional laboratory and clinical parameters do not accurately and consistently predict clinical outcomes. No biomarkers are currently being used for predictive purposes in standard clinical practice. In fact, there are no MHRA or FDA- approved seroprognostic or other laboratory-based tests that can reliably distinguish between mild and severe infection in Clostridium difficile infection. The lack of evidence-based prognostication strategies limits effective patient management and use of healthcare resources.

There is thus an unmet clinical need for novel biomarkers that have prognostic value to facilitate the implementation of more targeted management strategies that could improve overall clinical outcomes, especially survival. Predicting the severity for Clostridium difficile infection is particularly important since treatment strategies are stratified based on disease severity. Specifically, oral metronidazole is indicated for mild Clostridium difficile infection, vancomycin for severe Clostridium difficile infection, with addition of intravenous metronidazole for severe-complicated disease. Compared with the costs of the cheapest standard-of-care antibiotic (metronidazole), vancomycin is 20 times more expensive . From a health economics perspective, this fold difference is even more striking if one considers the costs of newer treatments such as Fidaxomicin ( 160-fold difference) and monoclonal antitoxin antibodies such as Bezlotoxumab (greater than 300-fold difference).

It is therefore an aim of the present invention to provide biomarkers that may be used to give an indication of the prognosis of Clostridium difficile infection in a patient and/or to select patients for a particular therapy. Such prognostic biomarkers will allow therapies to be targeted to those most in need and those most likely to benefit. Such biomarkers will also enable stratification of patients to identify those that will respond to a particular course of treatment and those that will not. This will allow patients to be given the most appropriate treatment quickly, and avoid the administration of antibiotics which will not be effective .

Accordingly, in a first aspect of the invention there is provided a method of determining the severity of a Clostridium difficile infection in a subject, the method comprising the steps of:

(i) obtaining a sample from a subject having or suspected of having a Clostridium difficile infection; detecting one or more biomarkers selected from the group consisting of antibodies to Clostridium difficile toxin A or a fragment thereof, antibodies to Clostridium difficile toxin B or a fragment thereof, antibodies to a Clostridium difficile surface layer protein or a fragment thereof, and IFNy, in the sample; and

determining from the results in (ii) whether the subject has a mild

Clostridium difficile infection or a severe Clostridium difficile infection. In particular, the method of the invention allows a subject with mild Clostridium difficile infection to be distinguished from a subject with severe Clostridium difficile infection. In a further embodiment the method of the invention may be used to predict the likely 30 day all-cause mortality in a subject with a confirmed Clostridium difficile infection. Prior to undertaking the method of the invention, a subject may have been diagnosed with a severe Clostridium difficile infection by one or more of the following parameters, having peripheral leucocytosis > 15xl 0⁹/L; having an increase in serum creatinine > 1.5 x above baseline; the presence of pseudomembranous colitis; having a toxic megacolon; having an intestinal perforation; having septic shock requiring intensive care admission or the need for colectomy.

In an embodiment of the invention, in step (ii) at least one biomarker selected from the group consisting of IFNy; antibodies to Clostridium difficile toxin A or a fragment thereof; and antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof; is detected. The absolute levels of the one or more biomarker may be determined, or the relative levels may be determined. If relative levels are determined these may be represented as arbitrary units, which may be denoted as standardised signals. In a further embodiment of the invention, in step (ii) at least antibodies to the Clostridium difficile toxin A or a fragment thereof are screened for. The detection of antibodies to the Clostridium difficile toxin A or a fragment thereof may be sufficient to determine whether or not a subject has a mild or a severe Clostridium difficile infection. In one embodiment only antibodies to the Clostridium difficile toxin A or a fragment thereof are screened for. In some embodiments of the invention, the peak white cell count (WCC) of a subject may also be utilised in a method of the invention. The method may include the step of determining the peak WCC. The peak WCC may be used to determine which subjects should be analyzed further to determine the severity of the Clostridium difficile infection.

In an embodiment, if the peak WCC is less than about 14x l 0⁹/L then a subject may be considered to have a mild Clostridium difficile infection and should be treated accordingly. If a subject has a peak WCC of greater than about 17xl 0⁹/L then a subject may be considered as having a severe Clostridium difficile infection, and should be treated accordingly. However, for subjects with a peak WCC between about 14xl 0⁹/L and about 17xl 0⁹/L, it is more difficult to determine if the subject has a mild Clostridium difficile infection or a severe Clostridium difficile infection, in this case the detection of antibodies to the Clostridium difficile toxin A or a fragment thereof in the sample may be used to determine if a subject has a mild or a severe Clostridium difficile infection. The cut off level for a mild infection may be a peak WCC of about 14.3xl 0⁹/L and be low. The cut off level for a severe infection may be a WCC of about 16.7xl 0⁹/L and above . A peak WCC level of between about 14.3xl 0⁹/L and about 16.7xl 0⁹/L may mean that the detection of antibodies to the Clostridium difficile toxin A or a fragment thereof in the sample is needed to determine if a subject has a mild or a severe Clostridium difficile infection. The WCC may be determined per litre of blood. In an embodiment, only antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof may be detected, or they may be detected in combination with another marker, for example in combination with the peak WCC, and this may be used to determine the severity of the Clostridium difficile infection. In an embodiment, IFN-γ levels may be detected in combination with another marker, for example in combination with peak WCC, and this may be used to determine the severity of a Clostridium difficile infection. In another aspect the invention provides a method to predict the 30 day all-cause mortality of individuals identified to have a Clostridium difficile infection, the method comprising the steps of:

(i) obtaining a sample from a subject having or suspected of having Clostridium difficile infection;

(ii) detecting one or more biomarkers selected from the group consisting of antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof, antibodies to the Clostridium difficile surface layer protein SLP002 or a fragment thereof, antibodies to the Clostridium difficile protein pCDTb or a fragment thereof, and antibodies to the Clostridium difficile toxin B or fragment thereof: and

(iii) predicting, based on the observations in (ii), whether the subject is likely to survive for 30 days or more. In the method of the invention, antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof may be screened initially. Thereafter, one or more of antibodies to the Clostridium difficile surface layer protein SLP002 or a fragment thereof, antibodies to the Clostridium difficile toxin B or a fragment thereof and antibodies to the Clostridium difficile protein pCDTb or a fragment thereof, may also be screened for. The results of these screens may be used in conjunction with screening for antibodies to SLP001 to predict whether the subject is likely to survive for 30 days or more .

All whole C. difficile surface layer proteins described herein, derived from C. difficile ribotypes 001 , 002 and 027, may be purified from C. difficile anaerobic broth cultures in accordance with the methods described by Ryan et al, PLoS Pathogens 201 1 ; 7(6) : e l 002076 and Lynch et al, BMC Evol Biol 2017; 17: 90. Lynch et al, also detail the GenBank accession numbers for the sip A gene of each ribotype. pCDTb is a precursor form of the B fragment of the Clostridium difficile binary toxin. pCDTb may be produced in Escherichia coli from a wholly synthetic recombinant gene construct; the amino acid sequence may be based on the published sequence for 027 ribotype http:www.uniprot.org/A8DS70. The steps involved in the cloning, expression and purification of pCDTb are described in Sundriyal et al, Protein Expr Purif 2010; 74 ( 1) : 42-8. The method to predict the 30 day all-cause mortality may be undertaken on all individuals with Clostridium difficile infection. In an embodiment, higher levels of antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof, and higher levels of antibodies to the Clostridium difficile surface layer protein SLP002 or a fragment thereof, and higher levels of antibodies to the Clostridium difficile protein pCDTb may be predictive of a 30 day or less mortality in a subject.

The method of the invention may be used to offer personalised treatments to individuals depending on the severity of the Clostridium difficile infection. Where a mild Clostridium difficile infection is indicated the therapy offered may be oral administration of metronidazole . Where a severe Clostridium difficile infection is indicated vancomycin, and/or fidaxomicin, may be offered, together with optionally intravenous metronidazole. Intravenous immunoglobulin (IVIG) may also be administered if a severe infection is identified.

The method of the invention may be used, for example, for any one or more of the following: to advise on the prognosis for a subject with a Clostridium difficile infection; to monitor infection progression; to advise on treatment options; and to monitor effectiveness or response of a subject to a treatment for Clostridium difficile infection. The sample may be a bodily fluid, such as, blood, serum, plasma, urine, spinal fluid, synovial fluid, sputum, mucus or lymph. The sample is preferably a blood sample, for example a whole blood sample, blood serum sample, or blood plasma sample .

The method of the invention may include the step of taking the sample from the subject. Alternatively, the method of the invention may not include in step of taking the sample, but instead the sample may be provided after it has been taken from the subject.

In one embodiment, the subject may already have been diagnosed with Clostridium difficile infection before the method of the invention is undertaken. The diagnosis of the Clostridium difficile infection may be based on an assessment of one or more of clinical presentation, pathology and other biomarker expression levels, for example the presence of Clostridium difficile toxin proteins in a stool sample. The subject may be a human or a non-human animal. In one embodiment, the subject is a human. A non-human animal may include dogs, cats, horses, cows, pigs, sheep and non-human primates.

The level of a specific biomarker in a sample, be it an antibody or a protein, may be determined by any suitable method, for example, by immunohistochemistry, spectrometry, ELISA, immunoprecipitation, western blot, or dot blot assay, protein microarray, radioimmunoassay (RIA), fluoroimmunoassay, an immunoassay using an avidin-biotin or streptoavidin-biotin system, and combinations thereof. These methods are well known to the person skilled in the art and other similar methods may also be used.

The level of one or more biomarkers may be determined using targeted tandem mass spectrometry (MS) methods. Examples of such methods include the: accurate inclusion mass spectrometry (AIMS), and quantitative selection reaction monitoring (Q-SRM).

The antibodies detected in a method of the invention may be total IgA antibodies (and IgA subclasses, including IgA l and IgA2), total IgG antibodies (and IgG subclasses including IgG l , IgG2, IgG3, IgG4), total IgM antibodies, or combinations thereof.

The anti-toxin A antibodies detected in a sample may be IgA antibodies (and IgA subclasses), IgG antibodies (and IgG subclasses), IgM antibodies, or combinations thereof. In one embodiment, the anti-toxin A antibodies detected are IgG antibodies. The anti-toxin B antibodies detected in a sample may be IgA antibodies (or IgA sublasses) or IgG antibodies (or IgG subclasses).

The anti-SLPOO l antibodies detected may be IgG antibodies (or IgG subclasses). The anti-SLP002 antibodies detected may be IgG antibodies (or IgG subclasses). The anti-pCDTb antibodies detected may be IgG antibodies, for example IgG2 antibodies or other IgG subclasses. The one or more antibodies may be detected, for example, by using a protein array, wherein the array carries the antibody target and by detecting the level of binding to the array at a particular position (corresponding to the antibody target) an arbitrary level of a particular antibody can be assigned to a particular sample. Whilst the level assigned may be arbitrary (denoted as a standardised signal), it is comparable with other targets in the array and between samples which have used the same array under the same conditions.

The method of the invention may be carried out in vitro.

According to a further aspect, the invention provides a panel of biomarkers, or a panel of agents for detecting a panel of biomarkers, for use in determining the severity of a Clostridium difficile infection in a subject, the panel of biomarkers comprising at least one of:

(i) antibodies to the Clostridium difficile toxin A or a fragment thereof;

(ii) antibodies to the Clostridium difficile toxin B or a fragment thereof,

(iii) IFNy;

(iv) antibodies to a Clostridium difficile surface layer protein or a fragment thereof; and

(v) antibodies to the Clostridium difficile protein pCDTb or a fragment thereof.

In order to predict the severity of a Clostridium difficile infection in a subject the panel preferably comprises at least antibodies to the Clostridium difficile toxin A or fragment thereof.

In order to predict the 30 day all-cause mortality in a subject with a severe Clostridium difficile infection, the panel preferably comprises at least antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof, and optionally one or more of antibodies to the Clostridium difficile surface layer protein SLP002 or a fragment thereof, and antibodies to the Clostridium difficile protein pCDTb or a fragment thereof.

In another aspect, the invention provides means to detect the biomarkers in the biomarker panel of the invention. This means that a panel of proteins or peptide targets may be used to detect the levels of antibodies in a sample which recognize these proteins or peptides. By determining the level of the biomarkers in a sample, either there absolute or relative levels, the severity or 30 day all-cause mortality in a subject with a Clostridium difficile infection can be determined.

In another aspect of the invention, there is provided a use of a method of the invention to detect biomarkers in a sample to determine the prognosis for a subject with a Clostridium difficile infection. According to another aspect to the invention there is provided a method of choosing the most appropriate treatment for a subject with a Clostridium difficile infection by performing the method of the invention on a sample from the subject and choosing, and optionally administering, treatment based on the biomarkers detected in the sample.

In another aspect of the invention, there is provided a method of treating a Clostridium difficile infection, the method comprising the steps of:

(i) detecting anti-toxin A antibody in a sample from a subject; and

(ii) administering treatment for the Clostridium difficile infection based on the anti-toxin A antibody levels.

The skilled person will appreciate that preferred features of any one embodiment and/or aspect of the invention may be applied to all other embodiments and/or aspects of the invention.

There now follows, by way of example only, a detailed description of the present invention with reference to the accompanying figures, in which:

Figure 1 - is a representative analysis of Clostridium difficile antigens using silver stain. Antigens were run on 4-20% precast polyacrylamide gels with a broad range of protein markers. The 7 antigens were then silver stained to check for purity and showed bands at the expected molecular sizes.

Figure 2 - lists the immunoglobulins used in this study.

Figure 3 - lists the cytokines used in this study.

Figure 4 - is a univariate analysis for severe Clostridium difficile infection. Figure 5 - is a univariate analysis for 30-day all-cause mortality.

Figure 6 - is a regression tree for Clostridium difficile infection severity.

Figure 7 - is a regression tree for Clostridium difficile infection 30 day all- cause mortality.

Figure 8 - shows performance metrics for severity and 30-day all-cause mortality. Figure 9 - is an 'Easy read' version of the severity decision tree.

Figure 10 - is an 'Easy read' version of the 30-day all-cause mortality decision tree Figure 11 - a simplified clinical version of the severity decision tree.

Figure 12 - a simplified clinical version of the 30-day all-cause mortality decision tree. MATERIALS AND METHODS

Data from a cohort of inpatients with Clostridium difficile infection at Nottingham University Hospitals NHS Trust was analysed from 2009 to 2013. All patients with Clostridium difficile infection had diarrhoea (defined as a change in bowel habit with 3 or more unformed stools per day for at least 48h) and a positive stool Clostridium difficile (enzyme immunoassay) toxin test. Severe Clostridium difficile infection was defined as peripheral leukocytosis > 15xl 0⁹/L or an increase in serum creatinine > 1.5 times above baseline, or pseudomembranous colitis, megacolon, intestinal perforation, need for colectomy or septic shock requiring intensive care unit admission.

Demographics, co-morbidities (including Charlson co-morbidity index), history, laboratory data [peak white cell count (WCC), eosinophil count, C-reactive protein (CRP), minimum albumin], imaging and endoscopy data, previous history of Clostridium difficile infection and 30-day all-cause mortality were recorded. Serum subclass and strain-specific antitoxin and anti-surface layer protein (SLP) antibody responses were determined by an established and validated Clostridium difficile antigen-specific protein microarray. Anti-toxin A and B neutralising antibodies (NAb) were assessed using a Caco-2 cell-based neutralisation assay. An antibody microarray assay was used for the simultaneous quantification of serum interleukin-(IL-) 1 -a, IL- 1 β, IL-6, IL-8, IL- 12/23, IL-27, tumour necrosis factor-a (TNF a)-, granulocyte- macrophage colony stimulating factor (GM-CSF), interferon-gamma (IFN-γ), and transforming growth factor-β (TGF- β) .

Antigen Microarray

A panel of seven Clostridium difficile antigens were included in this study: highly purified Clostridium difficile whole toxins A and B (toxinotype 0, strain VPI 10463, ribotype 087), toxin B from a Clostridium difficile toxin-B only expressing strain (CCUG 20309), precursor form of B fragment of binary toxin, pCDTb (produced from a wholly synthetic recombinant gene construct) and purified native whole ribotype- specific (001 , 002, 027) surface layer proteins (SLPs). In addition, two antigens of common pathogens as positive control targets were also included: tetanus toxoid (National Institute for Biological Standards and Control (NIBSC)) and lysates from Candida albicans (Jena Bioscience). Moreover, 10-point two-fold serial dilutions of purified human immunoglobulin matching the tested antibody isotype were printed in each array as a standard curve, 4 types were used; IgG (Sigma Aldrich, UK), IgG2, IgA, IgA l (Athens Research & Technology, Inc, USA) .

Assessment of the quality of the antigens

The quality of the Clostridium difficile antigens was assessed by electrophoresis and silver stain. For electrophoresis, an equal volume of NovexTm 2x SDS Tris-Glycine Loading buffer with 40mM DTT was added to each antigen ( lug) and following heating at 90°C for 5min, antigens were separated on Novex 4-20% Tris glycine polyacrylamide gels (Invitrogen Life Technologies, USA) . Silver stain was performed as stated in the Pierce ' Silver Stain Kit manufacturer's instructions (Thermo Scientific, UK) . Briefly, the gel was fixed twice using 2 x 25mL of a 30% Ethanol, 10% Glacial Acetic Acid and 60% De-ionised water for 15min each, then washed twice in 2 x 25mL of a 10% ethanol solution for 5min each followed by washing twice in ultrapure water for 5 minutes each. After lmin incubation in a sensitizer working solution, the gel was washed twice in ultrapure water for 1 min each. A stain working solution was added for 30 min, followed by washing twice in ultrapure water for 20 seconds each. The developer working solution was added immediately for 2-3min and replaced with the stop solution after the first band appeared (Figure 1).

Microarray procedure

The applicants have previously established and validated a Clostridium difficile multiplex protein microarray system to measure the antibodies in patient sera to specific Clostridium difficile antigens (Negm et al, Clinical Exp Immunology, 2017, Negm et al, Clin Vaccine Immunol, 2015, 22(9): 1033 - 1039) . Each antigen was diluted in PBS-Tween-Trehalose (50mM) at the optimum concentration which was tested before running the patient sera; toxins A (200 μg/mL) and B ( 100 μg/mL), toxin B CCUG 20309 (90 μg/mL), pCDTb (200 μg/mL) and purified SLPs (200 μg/ml); Candida and Tetanus ( 100 μg/mL) . Diluted antigens were loaded onto a 384 well plate (Genetix, UK), and printed in quadruplicate onto aminosialine-coated glass slides (Schott, Germany) using a contact arrayer (MicroGridll; Digilab, Marlborough, MA, USA) and a silicon contact pin (Parallel Synthesis Technologies, USA). Printed slides were blocked with fresh 5% BSA for one hour and washed three times with 0.05% PBS-Tween washing buffer. Serum samples were appropriately diluted in antibody diluent according to the immunoglobulin used (Figure 2) for 1 h with shaking. Following three times washing with PBS-Tween 0.05%, slides were incubated with the relevant biotinylated secondary antibodies for one hour with shaking. Finally, slides were incubated with streptavidin-conjugated cy5 diluted 1 :2000 in fresh 5% BSA (E- Biosciences, UK). Probed slides were scanned by GenePix scanner at 635nm and 400 PMT. The images were processed with Genepix Pro-6 Microarray Image Analysis software (Molecular Devices Inc.). Protein signals were determined after background subtraction though customized modules in the R statistical language to generate general mean of signal levels. Specific isotype responses were interpolated against the internal isotype standard curve for each sample .

Cytokine/Antibody microarray

Cytokines were analysed in the serum samples from Clostridium difficile patients by antibody microarray. DuoSet paired antibody kits (R&D Systems, Minneapolis, MN) were used for the detection of 12 cytokines (Figure 3): Human IL-27, BAFF/TNFSF 13, Human IL- /IL- lF2, Human IL-6, Human IL-a/IL- lF l , Human IFN- Y, Human TNF-a, Human GM-CSF, Human CXCL8/IL-8, Human TGF-β Ι , Human APRIL/TNFSF 13 and Human IL- 12/IL-23 P40. The array-based methodology has been previously established (Selvarajah et al, Mediators Inflamm, 2014, 2014: 820304) Briefly, 100 μg/mL of capture antibodies were diluted in PBS-50 mM Trehalose and printed onto aminosialine-coated glass slides (Schott, Germany) using a contact arrayer (MicroGridll; Digilab, Marlborough, MA, USA) and a silicon contact pin (Parallel Synthesis Technologies, USA). The slides were blocked with 5% BSA blocking buffer for one hour at room temperature with shaking. After washing the slides three times with PBS-Tween 0.05%, a cocktail of recombinant protein standards was prepared containing each cytokine at the maximum recommended concentration for standard curve generation and diluted twofold across eight dilutions. The standards and samples were added to the relevant blocks for one hour at room temperature with frequent shaking followed by washing with PBST three times. A mixture of biotinylated detection antibodies was added to each block for 1 h with shaking. Reaction was detected by streptavidin-conjugated cy5 diluted 1 : 1000 in fresh 5% BSA (E-Biosciences, UK) . The slides were scanned at 635 nm with a Genepix 4200AL scanner (Axon GenePix®) and fluorescence intensities were quantified using Axon Genepix Pro-6 Microarray Image Analysis software. The concentration of each cytokine in different samples was extrapolated from the standard curve values and presented in pictograms per mL (pg/mL).

Toxin neutralisation assay

A Caco-2 cell-based assay for anti-toxin A and anti-toxin B neutralising antibodies was used as previously published. Briefly, Caco-2 cells (HTB-37; ATCC) were maintained in minimal essential medium (MEM) plus 20% fetal calf serum, 2mM glutamine and non-essential amino acids at 37°C. Serum samples were diluted in the assay medium at three dilutions ( 1 : 10, 1 : 100 and 1 : 1000), then premixed with toxin A or B (at a 50% lethal dose [LD50]) for 1 h at 37°C before 50 of this mixture was transferred to the cells and incubated for 96 h. Following aspiration of the medium, 50 methylene blue (0.5% [wt/vol] dissolved in 50% [vol/vol] ethanol) was added to the cell culture and incubated for 1 h at room temperature. Then, the cells were washed gently with tap water (to remove excess stain) and air dried. The cells were then lysed by adding 100 1 % (vol/vol) N-lauryl-sarcosine and incubated on a shaker for 15 min at room temperature. The cell biomass was determined by measuring the absorbance of each well on a BioTek Synergy2 (BioTek, USA) plate reader at 405nm. Toxin activity and working LD50 concentrations were defined empirically in preliminary experiments and for each individual batch/lot of toxin used.

Statistical analysis

Descriptive statistics for demographics and outcomes are reported as median (range) or frequency (percent). Univariate and multivariate logistical regression models examined a set of clinical, demographic and immunological variables. These models were used to assess the association of the candidate variables with severe outcomes and 30-day mortality.

A machine-learning approach (regression Tree Model) was used for constructing decision tree algorithms tailored to protein microarray and clinical data sets to predict factors associated with severe Clostridium difficile infection and all-cause 30-day mortality. The same variables evaluated by decision tree analysis were also entered as potential predictors of severity and 30-day all-cause mortality in separate multivariable logistic regression models. Covariates were selected using correlation matrix among predictor variables for possible interactions. A cut-off value of 0.05 was used to include/exclude covariates in the binary multivariable logistic regression models.

The performance of both models was assessed by the following metrics: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), diagnostic accuracy and the Receiver operating Curve (ROC) value . The chi-square test compared the performance metrics between the models. The version of decision tree used was CHAID (Chi-squared Automatic Interaction Detection). CHAID chooses the independent (predictor) variable that has the strongest interaction with the dependent variable . Categories of each predictor are merged if they are not significantly different with respect to the dependent variable . The performance of the decision tree models were internally validated using the k-fold cross-validation (also named one versus all) approach. Using this statistical technique, the original data is randomly partitioned into k subsamples of equal sizes. Of the k subsamples, k- 1 samples are used for model training while one subsample is retained for model validation. The whole process is repeated for each of the k-folds, with each of the k subsamples used exactly once as the validation data. Finally, the results from different k folds can be merged to produce a single estimation. Discrimination of the original and cross-validated models was evaluated through the generation of receiver operating characteristic (ROC) curves and calculation of C-statistics in R. RESULTS

Of 15 1 subjects (52.3% female, median age 72 y, range 19-98 y), severe Clostridium difficile infection was present in 32% and 30-day all-cause mortality was 8.6%, respectively. In univariate analyses, factors independently associated with severe Clostridium difficile infection were: peak WCC, minimum albumin, and CRP (Figure 4); for 30-d all-cause mortality: peak WCC, SLPOO lIgG (Figure 5).

Separate regression tree algorithms were constructed for Clostridium difficile infection severity and 30-day all-cause mortality. These are illustrated in Figures 6 and 7.

For severity (Figure 6), the first splitting parameter is peak white cell count ( 10⁹/L); partitioned into 3 categories. Anti-toxin A IgG is partitioned into 2 categories. Using these two factors alone, a clinical prediction rule can be produced which correctly categorized 93.3% of the cohort into mild or severe infection. Accuracy can be further improved by examining IFN-γ, anti-SLPOO l IgG and anti-toxin B IgG2.

For 30-d all-cause mortality (Figure 7), anti-SLPOO l IgG, anti-SLP002 IgG and anti- pCDTb IgG2 were each portioned into 2 categories. Using these 3 factors, a predictive tool for 30-day all-cause mortality was able to correctly categorize 92. 1 % of the cohort. Predictive accuracy can be further improved by examining anti-toxin B IgG, anti-toxin B IgA and anti-SLPOO l IgA.

The corresponding performance metrics (sensitivity, specificity, accuracy, positive predictive value, negative predictive value and ROC value) for severity and 30-day all-cause mortality are presented in Figure 8.

Two easy-read versions of the above algorithms (Figures 9 and 10) were produced as well as simplified prediction algorithms or tools (Figure 1 1 and 12) for predicting disease severity and 30-d all-cause mortality based on the Decision Tree Model analysis.

Following implementation of the binary multivariable logistic regression model, the performance metrics for predicting severity of Clostridium difficile infection was as follows: Sensitivity = 75.8%, specificity = 97.4%, Accuracy = 90.9%, Positive prediction value (PPV) = 92.6%, negative prediction value (NPV) = 90.4%.

CONCLUSIONS

Presented herein are two novel microarray-based clinical risk stratification tools for patients with a Clostridium difficile infection who may benefit from receiving early, more aggressive and personalized treatment interventions. These decision tree prediction models correctly categorize severity as well as 30-day all-cause mortality in 91.9%, 95.4% of this patient cohort respectively.

Claims

1. A method of determining the severity of a Clostridium difficile infection in a subject, the method comprising the steps of:

(i) obtaining a sample from a subject having or suspected of having

Clostridium difficile infection;

(ii) detecting one or more biomarkers in the sample, wherein the one or more biomarkers are selected from the group consisting of antibodies to the Clostridium difficile toxin A or a fragment thereof, antibodies to the Clostridium difficile toxin B or a fragment thereof, antibodies to a Clostridium difficile surface layer protein or a fragment thereof, and IFNy; and

(iii) determining, from the results in (ii), whether the subject has a mild Clostridium difficile infection or a severe Clostridium difficile infection.

2. The method of claim 1 wherein the method allows a subject with a mild Clostridium difficile infection to be distinguished from a subject with a severe Clostridium difficile infection.

3. The method of claim 1 or 2 which is further able to predict the likely 30 day all-cause mortality in a subject with a confirmed Clostridium difficile infection.

4. The method of any preceding claim wherein in step (ii) at least one biomarker selected from the group consisting of IFNy; antibodies to Clostridium difficile toxin A or a fragment thereof; and antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof; is detected.

5. The method of any preceding claim wherein in step (ii) at least antibodies to the Clostridium difficile toxin A or a fragment thereof are screened for.

6. The method of any preceding claim further comprising the step of determining the peak WCC.

7. The method of claim 6 wherein a peak WCC of less than about 14xl 0⁹/L is diagnostic of a mild Clostridium difficile infection

8. The method of claim 6 wherein a peak WCC of greater than about 17xl 0⁹/L is diagnostic of a severe Clostridium difficile infection

9. The method of claim 6 wherein both the peak WCC is determined and antibodies to the Clostridium difficile toxin A or a fragment thereof are screened for.

10. The method of any of claims 1 to 4 wherein only antibodies to the Clostridium difficile surface layer protein SLP001 are screened for.

1 1. The method of any of claims 1 to 9 wherein IFN-γ levels are screened for in combination with another marker.

12. A method to predict the 30 day all-cause mortality of individuals identified to have a Clostridium difficile infection, the method comprising the steps of:

(i) obtaining a sample from a subject having or suspected of having

Clostridium difficile infection;

(iii) predicting based on the observations in (ii) whether the subject is likely to survive for 30 days or more.

13. A method of choosing the most appropriate treatment for a subject with a Clostridium difficile infection by performing the method of any of claims 1 to 12 on a sample from the subject and choosing, and optionally administering, treatment based on the biomarkers detected in the sample.

14. A method of treating a Clostridium difficile infection, the method comprising the steps of:

(i) detecting anti-toxin A antibody in a sample from a subject; and

15. The method of any preceding claim wherein the sample is blood, serum, plasma, urine, spinal fluid, synovial fluid, sputum, mucus or lymph.

16. The method of any preceding claim wherein the subject is a human.

17. The method of any preceding claim wherein the one or more antibodies are detected by using a protein array.

18. A panel of biomarkers for use in determining the severity of a Clostridium difficile infection in a subject, the panel comprising at least one of:

(i) antibodies to the Clostridium difficile toxin A or fragment thereof;

antibodies to the Clostridium difficile toxin B or fragment thereof, (ii) IFNy;

(iii) antibodies to a Clostridium difficile surface layer protein or fragment thereof; and

(iv) antibodies to the Clostridium difficile protein pCDTb or a fragment thereof.

19. The panel of claim 18 comprising antibodies to the Clostridium difficile toxin A or fragment thereof and at least one of (ii), (iii) or (iv).

20. The panel of claim 18 comprising antibodies to the Clostridium difficile surface layer protein SLP001 or a fragment thereof, and optionally one or more of antibodies to the Clostridium difficile surface layer protein SLP002 or a fragment thereof, and antibodies to the Clostridium difficile protein pCDTb or a fragment thereof.