WO2011064000A1 - Method for monitoring, identifying or diagnosing a bacterial infection - Google Patents

Abstract

The present invention relates to a method for monitoring, identifying or diagnosing an infection in an individual, a method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic, a method for assessing the efficacy of a treatment of an infection, an apparatus and a computer-usable medium configured to perform said methods. The method comprises obtaining a body fluid sample and/or a tissue sample from an individual, analysing the metabolite composition of said sample to determine a test metabolite profile of said sample, and comparing said test metabolite profile with a control metabolite profile. The method may comprise applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS), hierarchical multivariate curve resolution (H-MCR), and a multivariate statistical calculation, e.g. an orthogonal partial least squares discriminant analysis (OPLS-DA), of the separation between said test metabolite profile and said control metabolite profile.

Description

METHOD FOR MONITORING, IDENTIFYING OR DIAGNOSING A BACTERIAL INFECTION

Field of the invention

The present invention relates to a method for monitoring, identifying or diagnosing an infection in an individual. It also comprises a method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic and a method for assessing the efficacy of a treatment of an infection. Further, the invention relates to a computer means adapted for performing computing steps of said methods.

Background of the invention

An infection is a detrimental colonisation of a host organism by a foreign species resulting in a disorder or disease condition. Species that cause infections in humans or animals may for example be of a bacterial, viral, fungal or insecticidal origin.

Infections may play a role in a large number of diseases. Acute gastrointestinal infections may sometimes be associated with subsequent disease manifestations from several organ systems that may require hospitalisation of patients. For example, during 1997-2004, microbiologically confirmed gastrointestinal infections were reported for 101 855 patients in Sweden. Among patients who had Salmonella infection, an increased risk was found for aortic aneurism within three months after infection and an elevated risk for ulcerative colitis within one year after infection. This elevated risk for ulcerative colitis was also found for ulcerative colitis among

Campylobacter infections. Within one year, an increased risk was found for reactive arthritis among patients with Yersinia enteritis, Salmonella infection and Campylobacter infection¹.

Inflammatory bowel disease (IBD), including Crohn's disease and ulcerative colitis, involves a chronic, non-specific inflammation of the gastrointestinal tract. Several other studies indicate that IBD may be triggered, at least in part, by a Yersinia infection^2,3. Reactive arthritis is an autoimmune rheumatic disease that is characterised by an asymmetrical inflammation in peripheral and axial joints. Sometimes it is additionally manifested by conjunctivitis, urethritis, or mucocutaneous symptoms. Reactive arthritis is triggered by infections elsewhere in the body, usually by microbes from the gastrointestinal tract or urogenital tract. Triggering infections may for example be Campylobacter, Salmonella, and Yersinia (as mentioned above), as well as Shigella and E.coli. The joint pain in reactive arthritis may be positively correlated to the severity of symptoms of the preceding gastrointestinal infection⁴.

Another disease condition which may be associated with bacterial infection is atherosclerosis. Chlamydia pneumoniae has frequently been found in atherosclerotic tissue and less frequently in normal artery tissue. However, the results of serologic studies of chronic infection by C.

pneumoniae up until now are inconclusive⁵.

An infection may cause symptoms that make it difficult to distinguish clinically from other disorders manifested by similar symptoms. For example, Yersinia, Campylobacter and Salmonella are pathological bacterial species that may infect the ileocecal area and cause infectious ileocecitis, which (together with other infective enteritis conditions) is (are) a common mimicker of appendicitis. In such cases, pain in the right lower quadrant is the predominant symptom, which may lead to a surgical operation, a so-called laparotomy, for suspected appendicitis. However, at surgery a normal appendix is removed unnecessarily, while there is edematous thickening of ileum and cecum, and enlarged mesenteric lymph nodes, a condition which is principally self-limiting and innocuous^6,7,8.

Antibiotics have significantly reduced the threat posed by infectious diseases during their use in the last century. Over the years, antimicrobials have saved the lives of millions of people and the drugs have increased the life span of the human population. This achievement is now threatened by the development of bacteria that show resistance to antibacterial agents. Today, antibiotic resistance is a huge and serious concern constituting an immense threat to public health and modern medicine⁹. In the majority of cases of resistance to β-lactam antibiotics (all having the β-lactam ring as a common structural element, e.g. penicillins, cephalosporins, cephamycins,

carbapenems and monobactams), the resistant bacteria produce an enzyme called β-lactamase that degrade penicillins by opening up the ring of the molecule thus generating an inactive compound. During the last 50 years there have been an enormous spread of different β-lactamases and the spread of expanded-spectrum β-lactamases (ESBL) now constitutes a huge clinical problem worldwide.

Global metabolic profiling, also called metabolomics or metabonomics, has been used to detect changes in biomarkers as a means of monitoring patient health status¹⁰. Due to the large chemical diversity of the metabolites found in samples from different body fluids and tissues, it is necessary to employ numerous analytical methods to separate and detect as many of these compounds as possible. Most of these methods are based on high- throughput separation and mass spectrometry techniques combined with powerful computer-based multivariate data analysis¹¹. It has been shown that unbiased metabolic profiling can be used to distinguish between benign prostate, clinically localised prostate cancer and metastatic disease¹². Thus, metabolites that vary under these conditions can be used as biomarkers to create metabolic signatures that can serve as a diagnostic tool for monitoring a particular disease (French G. L. (1990), Reviews in Medical Microbiology 1 :168-174).

Summary of the invention

On account of the above-mentioned problems related to infections, it is of great importance to identify or diagnose an infection in an individual at an early stage, in order to provide a relevant and efficient treatment of the infection. It may also be crucial to monitor the development of an infection in an individual and for example study the effects of a treatment given to the individual.

Further, due to the rapid spread of antibiotic resistance, there is an increased clinical need to enable rapid diagnosis not only of the infectious agent itself but also of its metabolite profile of resistance to antibiotics. In order to maximize the use and effectiveness of antibiotic treatment, novel strategies have to be developed to rapidly assess the metabolite profile of antibiotic resistance of bacterial infections.

In one aspect, the present invention relates to a method for monitoring, identifying and/or diagnosing an infection in an individual, comprising:

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(b) analysing a metabolite composition of said body fluid sample and/or said tissue sample to determine a test metabolite profile of said body fluid sample and/or said tissue sample; and

(c) comparing said test metabolite profile with a control metabolite profile.

More specifically, the invention pertains to a method for monitoring, identifying and/or diagnosing an infection in an individual, comprising:

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(b) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said body fluid sample and/or said tissue sample; and

(c) comparing said test metabolite profile with a control metabolite profile.

In an embodiment of the invention, (b) comprises applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to analyse the metabolite composition of said body fluid sample and/or said tissue sample.

In another embodiment, (b) comprises processing the data of the metabolite composition using hierarchical multivariate curve resolution (H- MCR).

In a preferred embodiment, (b) comprises

(b1 ) applying gas chromatography/time of flight-mass spectrometry

(GC/TOF-MS) to said body fluid sample and/or said tissue sample; and

(b2) processing the data of the metabolite composition using

hierarchical multivariate curve resolution (H-MCR). In an embodiment of the invention, c) comprises performing a multivariate statistical calculation of a separation between said test metabolite profile and said control metabolite profile.

In a preferred embodiment, said multivariate statistical calculation is an orthogonal partial least squares discriminant analysis (OPLS-DA).

In another embodiment, (c) comprises comparing the test metabolite profile obtained in (b) with control metabolite profiles of healthy and infected individuals, respectively.

In particular, the metabolite profiles compared in (c) comprise (i) a simple profile of a small number of identified and/or unidentified metabolites unique to an infected individual and/or (ii) a complex profile of a large number of mostly unidentified metabolites. The metabolite profiles compared in (c) may comprise a combination of (i) and (ii).

The present invention further relates to a method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic, comprising the method as described above.

In one embodiment, said antibiotic has been administered to said individual prior to obtaining a body fluid sample and/or a tissue sample from said individual according to (a), and a control metabolite profile originates from a body fluid sample and/or a tissue sample obtained from said individual before said antibiotic has been administered to said individual.

In an alternative embodiment, the method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic comprises

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(a') dividing said body fluid sample and/or said tissue sample into aliquot parts and distributing said aliquot parts into separate reaction chambers, wherein each reaction chamber comprises an antibiotic or a negative control;

(a") incubating said aliquot parts in said reaction chambers;

(b) analysing a metabolite composition contained in said reaction chambers, including processing data of the metabolite composition, to determine a test metabolite profile of the aliquot part in each reaction chamber; and

(c) individually comparing said test metabolite profiles with a control metabolite profile from bacteria resistant to or susceptible to an antibiotic;

(d) identifying the infection as being caused by bacteria resistant to or susceptible to an antibiotic.

In said method, the test metabolite profile(s) may comprise metabolites being degradation products of said antibiotic and/or metabolites from lysed bacteria.

The invention also relates to a method as described above, wherein said individual is further affected by a chronic disease.

The invention further relates to a method as described above wherein said infection is monitored over time and said individual receives treatment for said infection.

The present invention also relates to a method for assessing the efficacy of a treatment of an infection, comprising the steps

(a) identifying an individual as having said infection;

(b) obtaining a body fluid sample and/or a tissue sample from said individual;

(c) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said sample;

(d) comparing said test metabolite profile with a control metabolite profile;

(e) treating said individual for said infection with said treatment;

(f) repeating steps b)-d) any number of times while performing said treatment.

The invention also relates to a method as described above for clinical studies of pharmaceutical candidates for treatment of infection.

The invention further relates to apparatus for monitoring, identifying and/or diagnosing an infection in an individual, said apparatus being configured to perform the method according to any one of the embodiments described above. The invention also encompasses a computer-usable medium

comprising a computer program code configured to cause a processor to execute the method according to any one of the embodiments described above.

Definitions

All terms used in the present specification are intended to have the meaning usually given to them in the art. For the sake of clarity, some terms are also defined below.

The terms "metabolite profile", "metabolite pattern", "marker profile" and "marker pattern" shall be construed as meaning a profile or pattern of one or more identified or unidentified metabolites. Such a metabolite profile may consist of a simple profile comprising a small number of identified or unidentified metabolites unique to an infected individual and/or to an infectious organism or a complex profile comprising a large number of mostly unidentified metabolites.

The term "small number of metabolites" shall be construed as meaning at least three metabolites.

The term "large number of metabolites" shall be construed as meaning about one hundred metabolites, or more.

In the context of monitoring an infection, the term "over time" shall be construed as meaning over a period starting on day 1 and ending within one year, such as within 3-6 months, such as within 7-21 days.

The term "antibiotic" shall be construed as including compounds useful for treatment of bacterial, viral, fungal or insecticidal infections.

Brief description of the drawings

Figure 1 . Yersinia infection followed by in vivo imaging of FVB/N mice infected with 5.3 x 10⁷ CFU Y. pseudotuberculosis Xen4. a) Mice exhibiting a bioluminescent signal on the indicated days after onset of infection, b) Health status of FVB/N mice monitored by measuring levels of glucose in blood, weight, and haemoglobin (Hb) in blood. Errors bars represent SEM for five mice analysed at each time point (control mice N = 2). Figure 2. Yersinia infection followed by metabolonnics. Scatter plot of OPLS-DA scores (first predictive component, t[1 ], vs. first orthogonal component, to[1 ]) based on plasma metabolomics data for Yersinia- infected FVB/N mice, a) Separation of data on one mouse prior to infection (black dots) and another after onset of infection (white dots). The calculated OPLS- DA model could predict 66.1 % of the between-class variation according to cross-validation (p = 0.005, CV-ANOVA). b) Separation of data on control (black boxes) and infected (white boxes) mice at corresponding time points after onset of infection. The calculated OPLS-DA model could predict 70.6% of the between-class variation according to cross-validation (p = 0.025, CV- ANOVA). For visualization purposes the x-axis in the figure is reversed, c) Separation of data on 3 mice with cleared infection (black stars to the left) and mice with persistent infection (black stars to the right) 52 days after onset of infection. The calculated OPLS-DA model could predict 99.3% of the between-class variation according to cross-validation (p < 0.001 , CV- ANOVA).

Figure 3. Prediction of Yersinia infection clearance in FVB/N mice by metabolomics. a) Scatter plot of OPLS-DA scores (first predictive component, tPS[1 ], versus first orthogonal component, toPS[1 ]) showing the difference between control (black boxes) and Yersinia- infected (white boxes) mice at corresponding time points after onset of infection. The model is based on plasma metabolites found to be significant for modelling the difference between mice with, respectively, cleared and persistent infection, and it reveals a distinct difference between control and infected mice (predicting 62.8% of the between-class variation according to cross-validation, p = 0.005 by CV-ANOVA). Applying the model data on samples taken 52 days after onset of infection from one control (star in the lower left quadrant), three mice with persistent infection (stars in the lower right quadrant), and three with resolved infection (stars in the upper right quadrant) clearly illustrates that the mice with cleared infection are predicted closer to the control group than to the group with persistent infection (as pointed out by an arrow), and the control mice are predicted to be in the control group, b) Bar plot of OPLS-DA scores (first predictive component, tPS[1 ]) showing the difference between control (black bars) and infected (white bars) mice at corresponding time points after the onset of infection. The model is based on three plasma metabolites (UN19, UN50, and UN76) highlighted as potential markers for monitoring and predicting infection, and it reveals a distinct difference between control and infected mice (predicting 67.3% between-class variation according to cross-validation, p = 0.0001 by CV-ANOVA). Applying the model to data on samples collected 52 days after onset of infection from one control (black bar), three mice with persistent infection (white bars), and three with resolved infection (grey bars) clearly shows that the animals with cleared infection are predicted closer to the control group than to the group with persistent infection, and the control mice are predicted to be in the control group.

Figure 4. Metabolites predicting Yersinia infection clearance. Bar plots of the areas under the resolved chromatographic peaks in the plasma metabolomics data. Mice with persistent infection (white bars), mice with cleared infection (grey bars), and controls (black bar) at 52 days post-onset, a) Bar plot of data on the potential marker UN19 (p = 0.016 for the difference between animals with persistent and cleared infection), b) Bar plot of data on the potential marker UN50 (p = 0.023 for the difference between animals with persistent and cleared infection), c) Bar plot for the potential marker UN76 (p = 0.007 for the difference between persistent and cleared infection).

Figure 5. Francisella, Pseudomonas, and Borrelia infections followed by metabolomics. Scatter plot of OPLS-DA scores based on plasma metabolomics data for control mice and mice infected with Francisella, Pseudomonas and Borrelia. a) Separation of data between control mice (Control; black dots), control mice for Pseudomonas aeruginosa (Control P; black circles) and mice infected with Francisella tularensis (F; black

diamonds), Pseudomonas aeruginosa (P; black squares), Borrelia burgdorferi (BB; black triangles) and Borrelia duttonii (BD; black stars). The calculated OPLS-DA model could predict 57.6% of the between-class variation according to cross-validation (p < 0.0001 , CV-ANOVA). b) Separation of data on mice infected with Borrelia burgdorferi (black triangles) and Borrelia duttonii (black stars). The calculated OPLS-DA model could predict 67.3% of the between-class variation according to cross-validation (p < 0.0001 , CV- ANOVA). c) Separation of data between control mice (black dots) and mice infected with Francisella tularensis (black diamonds). The calculated OPLS- DA model could predict 88.6% of the between-class variation according to cross-validation (p < 0.0001 , CV-ANOVA). For visualization purposes the x- axis is plotted in reversed order, d) Separation of data between control mice for Pseudomonas aeruginosa (black circles) and mice infected with

Pseudomonas aeruginosa (black squares). The calculated OPLS-DA model could predict 90.1 % of the between-class variation according to cross- validation (p < 0.002, CV-ANOVA). For visualization purposes the x-axis is plotted in reversed order, e) Separation of data between control mice (black dots) and mice infected with Borrelia burgdorferi (black triangles). The calculated OPLS-DA model could predict 89.6% of the between-class variation according to cross-validation (p < 0.0001 , CV-ANOVA). f) Separation of data between control mice (black dots) and mice infected with Borrelia duttonii (black stars). The calculated OPLS-DA model could predict 89.6% of the between-class variation according to cross-validation (p < 0.0001 , CV- ANOVA). For visualization purposes the x-axis is plotted in reversed order.

Figure 6. Prediction of ampicillin resistance followed by metabolomics. Scatter plot of OPLSA-DA scores based on plasma metabolomics data for control mice and mice infected with strains of Yersinia pseudotuberculosis. Small metabolites are released to serum after ampicillin treatment of an ampicillin susceptible strain of Y. pseudotuberculosis (black triangles to the left). On the contrary, no effect is seen as a result of ampicillin treatment of the resistant strain expressing a TEM-1 beta-lactamase (black diamonds in the upper right quadrant) when compared to a non-infected control (white circles in the lower right quadrant). Instead, the difference between the control and the resistant strain (in the orthogonal direction) is likely to be explained by the reaction between TEM-1 beta-lactamase and ampicillin.

Figure 7. Plot showing the metabolic pattern in blood samples 1 , 4 and

8 hours after antibiotic treatment in a mouse model of bacterial infection. The metabolic pattern in untreated, uninfected controls is indicated by one grey box for each individual. Efficient treatment, meaning mice treated with cloxacillin against MSSA and vancomycin against MRSA, are shown as black boxes. Mice infected with cloxacill in-resistant MRSA and treated with cloxacillin, shown as unfilled boxes, are separated from the other two groups indicating that the animals do not survive the infection.

Figure 8. Plot showing the metabolic pattern in blood samples 1 , 4 and

8 hours after treatment with antibiotics in a mouse model of bacterial infection. The metabolic pattern in untreated, uninfected controls is indicated by a grey box for each individual. Successful treatment, that is vancomycin against MRSA, shown as black boxes, is clearly separated from control animals. Animals infected with MRSA (resistant to cloxacillin) and treated with cloxacillin are shown as unfilled boxes. This latter group is not separated from the control group, indicating that the bacteria are not affected by the treatment.

Figure 9. Plots showing examples of two metabolites that exhibit a change in concentration upon antibiotic treatment as compared to control. The Y axis denotes normalized difference in concentration between metabolites. Black boxes = controls; grey boxes = MRSA treated with vancomycin; and unfilled boxes = MRSA treated with cloxacillin. A) Metabolite with an increased concentration in mice infected with MRSA and treated with vancomycin compared to control animals. However, no difference could be seen when comparing the concentration of the metabolite in mice infected with MRSA and treated with cloxacillin and in control animals. B) Metabolite with a decreased concentration in mice infected with MRSA and treated with vancomycin compared to control animals. However, no difference could be seen when comparing the concentration of the metabolite in mice infected with MRSA and treated with cloxacillin and in control animals.

Detailed description of the invention

The present invention comprises monitoring, identifying or diagnosing an infection in an individual by performing a method of metabolic profiling. The method comprises obtaining a metabolite profile and comparing said metabolite profile with a control metabolite profile. Said metabolite profiles may consist of a simple profile comprising a small number of identified or unidentified metabolites unique to an infected individual or a complex profile comprising a large number of mostly unidentified metabolites, i.e. a

metabolite pattern. Preferably, the method comprises classifying a

combination of both kinds of said profiles.

The inventors have found that individual metabolites that are unique to an infected individual and/or to an infectious organism may be detected by the method as disclosed herein. This is a surprising result since the infectious organisms producing said unique metabolites exist in very low concentrations (10⁶) in comparison to the very much higher concentrations (10¹²) of non- pathogenic bacteria that are normally present in animals and humans.

In one aspect, the present invention relates to a method for monitoring, identifying and/or diagnosing an infection in an individual, comprising:

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(b) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said sample; and

(c) comparing said test metabolite profile with a control metabolite profile.

In an embodiment of the invention, (b) comprises applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS)¹³ to analyse the metabolite composition of said body fluid sample and/or said tissue sample. Further, (b) may comprise processing the data using hierarchical multivariate curve resolution (H-MCR)¹⁴.

Alternatively, liquid chromatography/mass spectrometry (LC/MS)¹⁵ may be used for analysing the metabolite composition of the sample and the thus acquired data may be processed by use of XCMS¹⁶. The analysis of the metabolite composition of the sample may also be performed by using ultra pressure liquid chromatography/mass spectrometry (UPLS/MS)¹⁵, nuclear magnetic resonance (NMR)¹⁷ or capillary electrophoresis/mass spectrometry (CE/MS)¹⁸.

In a preferred embodiment, (b) comprises (b1 ) applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to said body fluid sample and/or said tissue sample; and

(b2) processing the data using hierarchical multivariate curve resolution (H-MCR).

In an embodiment of the invention, c) comprises performing a multivariate statistical calculation of a separation between said test metabolite profile and said control metabolite profile.

In a preferred embodiment, said multivariate statistical calculation is an orthogonal partial least squares discriminant analysis (OPLS-DA)¹⁹.

Alternatively, said multivariate statistical calculation is made by PLS-

DA or a combination of orthogonal signal correction (OSC)²⁰ and PLS-DA²¹. Other calculation tools that may be used are Support Vector Machines²², neural networks²³ or various clustering methods²⁴.

It is further shown that it is possible to use metabolic profiling to rapidly assess if the infectious bacterial agent is resistant towards β-lactam

antibiotics having an active β-lactamase enzyme. The underlying idea is based on the argument that resistant bacteria should not be affected by exposure to β-lactam antibiotics. In contrast, sensitive bacteria should lyse after exposure to the antibiotic, and as a consequence, small intracellular metabolites should be released from the bacteria that could be detected using metabolomics.

The method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic is performed by administering an antibiotic to an individual and obtaining a body fluid sample or a tissue sample from said individual, followed by analysing the metabolite profile of said individual and comparing it to a control metabolite profile. If resistant bacteria are present in the sample, the antibiotic AB will be degraded or modified to other metabolites ABX. On the other hand, if bacteria susceptible to the antibiotic are present in the sample, no ABX will be formed. Instead, the bacteria will be lysed by the antibiotic, thus releasing small metabolites that can be detected by identification of a metabolite profile. A resistant bacterium will not lyse and therefore will not generate such small metabolites. In an alternative embodiment, the method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic comprises obtaining a body fluid sample or a tissue sample from an individual, dividing said sample into aliquot parts and distributing said aliquot parts into separate reaction chambers, wherein each reaction chamber comprises an antibiotic or a negative control, and further incubating said aliquot parts in said reaction chambers, analysing the metabolite composition contained in said reaction chambers, including processing data of the metabolite composition, to determine a test metabolite profile from each reaction chamber, and individually comparing said test metabolite profiles with a control metabolite profile from bacteria resistant to or susceptible to an antibiotic, which makes it possible to identify the infection as being caused by bacteria resistant to or susceptible to an antibiotic.

The invention also relates to a kit-of-parts for performing the above described method, comprising a plurality of reaction chambers wherein an antibiotic is pre-deposited in a portion of the reaction chambers and a portion of the reaction chambers do not have pre-deposited antibiotic in order to function as a negative control. Several kinds and/or classes of antibiotics may be deposited in the respective reaction chambers in order to facilitate simultaneous testing of resistance to a number of antibiotics. The plurality of reaction chambers may be provided physically integrated with each other, such as on one or more microtiter plates, or physically separated from each other, such as individual test tubes. The kit may further comprise buffers and media suitable for performing the method and/or instructions for use.

Thus, resistance towards different β-lactam derivatives and other antibiotics causing bacterial cell lysis accompanied by release of metabolites can be rapidly assessed by a simple blood test. This will be noticed as a change in metabolic profile thus generating a specific signature that can be evaluated and deciding for the selection of the antibiotic of choice. In addition, this application can be combined with the application of diagnosing infectious agents. Thus, a combination of these two applications of the present method of metabolic profiling can give the correct diagnosis as well as the antibiotic resistance metabolite profile of the bacterial infection. The present method may also be used for monitoring, identifying or diagnosing infectious disease caused by a viral, fungal or insecticidal species. Non-limiting examples of genera of viruses that may be monitored, identified or diagnosed by the method according to the present invention are the hepatitis C virus, hepatitis B virus, HIV, cytomegaloviruses (CMV),

hantaviruses, influenzaviruses, papillomaviruses, dengue viruses, viruses causing encephalitis and viruses causing so-called "emerging diseases", e.g. SARS, H5 avian influenzavirus and Nipah virus.

The invention also relates to an apparatus for performing the method according to the invention. Said apparatus is configured to receive a body fluid sample and/or a tissue sample obtained from an individual, and the apparatus comprises an analyser unit configured to analyse a metabolite composition of said body fluid sample and/or said tissue sample; a processor unit to process data of the metabolite composition to determine a test metabolite profile of said sample; and a comparator unit configured to compare said test metabolite profile with a control metabolite profile. Said different units may be connected to a central processing unit.

More specifically, the apparatus may comprise:

- a GC/TOF-MS unit to analyse the metabolite composition of said body fluid sample and/or said tissue sample;

- a processor unit to process data of the metabolite composition

using H-MCR;

- a calculator unit configured to perform a multivariate statistical

calculation of a separation between said test metabolite profile and said control metabolite profile, said calculator unit optionally being configured to perform an OPLS-DA.

The present invention and its embodiments can be realised in many ways. For example, one embodiment of the present invention includes a computer-usable medium comprising instructions stored thereon that are executable by a computer system located in one or several units of the above-described apparatus for performing the method of the invention. The instructions executable by the computing system and stored on the computer- usable medium perform the method steps of the present invention as described herein. The invention may be performed by use of hardware and/or software.

The invention is further described in the following examples. The examples are illustrative and shall not be considered as limiting the scope of the invention.

Example 1

With the aim to explore the possibility of using metabolomics to monitor disease progression after infection with Yersinia pseudotuberculosis, FVB/N mice were orally infected with a sub-lethal dose of the pathogen. The infection was followed for 1 15 days, and blood samples for metabolomic analyses were collected at different time points. The infection was also monitored using the in vivo imaging system IVIS®Spectrum, which allows real-time imaging of an infection in living mice^25,26,27,28. The animals were infected with a wild-type strain of Y. pseudotuberculosis, which emits a bioluminescent signal due to the luxCDABE operon integrated on the virulence plasmid. Such signals were detected in the intestinal tract from day 1 in 14 out of 15 mice (data not shown).

Two phases of infection were monitored: an acute phase starting from the onset of the infection peaking on days 5-10 and the subsequent phase during which the infection declined in the majority of the infected mice. A parallel group of five infected mice were sacrificed on day 7, and organs associated with a gut infection were dissected out and analysed. The results revealed the presence of Y. pseudotuberculosis in the intestinal tract (cecum and Peyer's patches), which agrees with earlier findings^29,30. During the acute phase (days 5-10), the mice showed external signs of infection such as low activity and a tendency towards dishevelled fur.

During the second phase, which lasted five to seven days, the infection was either cleared (two of 14 mice), or developed to a systemic infection and mortality (three of 14) (Fig. 1 a). In the latter case, the bioluminescent signal also appeared in the area of the liver and spleen, and the mice were sacrificed before death due to infection. However, the infection remained in two mice after 52 days. Importantly, the animals with persistent infection did not show any symptoms such as changes in blood status, glucose, or weight (Fig. 1 b), indicating that the infection was mild and involved only a limited number of bacteria. The two chronically infected mice also had yersiniae in the faeces (3.1 x 10⁶ and 0.21 x 10⁶ CFU/g faeces respectively). In sharp contrast, the mice that did not emit any light signal had no viable yersiniae in faeces, which indicated that infection was cleared. After 1 15 days of infection all mice were sacrificed, and organs were removed and analyzed. No signal could be detected in any organs from mice that had been considered to be cleared based on the MS analysis in the anaesthetised mice. However, it was evident that the light-emitting yersiniae in mice with persistent infection were restricted to the intestinal/caecal area. Notably, no signals were evident in Peyer's patches or mesenteric lymph nodes at this stage of infection. The pathogen titres of the cecum samples from mice with persistent infection were found to be around 1 x 10⁶ yersiniae per cecum. Therefore, it is concluded that the MS® Imaging System can detect even a small number of the bacteria and that it can accurately monitor a persistent symptomless infection, which will make it possible to discriminate between persistence and

clearance.

Blood was drawn from the mice at discrete time points before and during the entire infection period, and sampling was also continued after the animals appeared to have cleared the infection. Initially, samples from groups of five mice were pooled to ascertain the average status of the mouse population after infection. At the end of the experiment, individual blood samples were drawn from mice with persistent infection, mice with cleared infection, and non-infected mice (controls). The serum metabolic profile was determined by gas chromatography/time of flight-mass spectrometry

(GC/TOF-MS) after extraction of the low-molecular-weight components to obtain a screening of the metabolome³¹. The acquired data were

preprocessed using a hierarchical multivariate curve resolution strategy (HMCR)¹⁴ followed by multivariate statistical analysis entailing orthogonal partial least squares discriminant analysis (OPLS-DA)¹⁹. Use of this approach to achieve predictive metabolomics has proven successful in a variety of metabolomics applications^32,33. Multivariate analysis was performed to compare the metabolic profiles of plasma represented by 277 putative metabolites from mice (i) at pre- infection and post-onset of infection and (ii) at time points post-onset with the profiles of corresponding plasma from control mice. The results revealed a significant separation in the first model directions (predictive OPLS

components; p = 0.005 and p = 0.025 respectively, based on ANOVA of cross-validated scores [CV-ANOVA]³⁴ indicating that metabolite profiles were present in the metabolomics data (Fig. 2a, b). Metabolic changes associated with infection, that is, significant changes in metabolites that occurred in both models and accordingly were not confounded by time of sampling or inter- individual variation, included the following: elevated levels of UN19, UN23, UN47, UN50, glucose-6-phosphate, and homoserine; depleted levels of UN01 and UN76, where UN denotes unidentified metabolite.

Multivariate comparison also revealed a significant metabolic difference between mice with persistent infection at 52 days post-onset and mice with no evident signs of infection at 52 days post-onset (p < 0.001 based on CV-ANOVA) (Fig. 2c). This highlights the possibility of detecting plasma metabolite profiles to monitor the fate of infection on an individual basis.

Interestingly, an equally strong model of the above-mentioned separation (persistent vs. cleared at day 52; p < 0.001 based on CV-ANOVA) was obtained when using only the metabolites that had the most significant impact on the separation of control and infected mice (i.e. 10 metabolites with p<0.05 in both infection models, i.e. pre infection vs. post infection and control vs. infected), which indicates coherence of metabolic markers for the two conditions studied. Furthermore, only the metabolites that had the greatest impact on separation of persistently infected and cleared mice were used to model the difference between control and infected mice. A significant difference was found (p = 0.005, CV-ANOVA), which was similar to the difference noted in the corresponding model based on all metabolites.

Furthermore, samples taken 52 days after the onset of infection (from one control, three mice with persistent infection, and three with clearance) were compared with the model generated by separating control and infected mice. A clear metabolite profile emerged in which mice with persistent infection were actually predicted to be in the infected group, and the control mice were predicted to be in the control group. Notably, the animals that had cleared the infection at 52 days post-onset were all predicted to be closer to the control group compared to the mice that were still infected (Fig. 3a). Use of OPLS model loadings and univariate p-values to investigate the metabolites that contributed to detection of infection and prediction of clearance or persistence singled out three metabolites in particular (UN19, UN50, and UN76) as a potential metabolite profile for diagnosing and monitoring the fate of infection. An OPLS-DA model comparing control and infected mice and based solely on the three markers revealed a clear separation associated with infection in one OPLS component (p = 0.0001 by CV-ANOVA). Prediction using the samples from day 52 yet again proved the capacity of the method to distinguish between persistent and cleared infection (Fig. 3b). Plotting the area under the resolved chromatographic peaks for the three highlighted metabolites was done to verify that each one actually had the ability to serve as a marker of infection clearance (Fig. 4a-c).

Example 2

To further study the general applicability of metabolic profiling as a novel diagnostic tool we infected mice with four different bacterial pathogens, Pseudomonas aeruginosa, Francisella tularensis, Borrelia duttonii and Borrelia burgdorferi^{35,36,37,38,39} , which were shown to cause different disease metabolite profiles. Blood was drawn from the infected mice at day 1 to day 10 after the onset of infection (depending on the infectious agent) and the metabolite profiles of plasma samples including age matched non-infected controls were identified (Fig. 5a). We could in accordance with the earlier results show that there was a significant change in the metabolic profile of these animals detectable already 24 hours after onset. Importantly, each infection generated a unique metabolic profile that was stably maintained throughout the whole experiment. Notably, the two different Borrelia infections (Borrelia duttonii and Borrelia burgdorferi) could be clearly distinguished based on their metabolite profiles in a separate OPLS model (p < 0.0001 by CV-ANOVA; Fig. 5b, c). This was also the case for all included infections compared to control (Pseudomonas aeruginosa, p = 0.002; Francisella tularensis, p < 0.0001 ; Borrelia duttonii, p = 0.005; Borrelia burgdorferi, p < 0.0001 . p-values by CV-ANOVA; Fig. 5d-e). In total, 138 metabolites (57 identified and 81 unidentified) were significantly altered between control mice and the individual infections (p < 0.05; Table 1 , below). Out of these, 1 1 metabolites were common for all four infection types (e.g. increased levels of p-hydroxybenzoic acid, and depleted levels of thryptophan, L-cysteine and serine). Interestingly, the same changes could also be detected in acute Yersinia infection for p-hydroxybenzoic acid, thryptophan and serine.

Furthermore, 85 metabolites were common for at least two infection types and 53 were unique for a specific infection. This suggests that our

metabolomic approach can be employed to detect a general infection status as well as to provide diagnostics of individual infection types by means of unique markers, or more likely by predictive metabolite profiles.

Table 1 (below) shows significant metabolite changes (p < 0.05) in blood from mice with acute infection by Borrelia burgdorferi (BB), Borrelia duttonii (BD), Pseudomonas aeruginosa (P) and Francisella tularensis (F) and includes a comparison to acute Yersinia (Y) infection. The arrows are indicating direction of change compared to control (†: increase, J,: decrease). Identity is given for metabolites with a verified identity in database search. Metabolic changes due to acute Yersinia infection in the identified significant metabolites for the other four infections (BB, BD, P and F) are given with corresponding p-values for the change between control and Yersinia infected mice. The Yersinia samples were analysed separately, and therefore no complete comparison was made between the Yersinia samples and the other samples. Table 1

Significant metabolite changes during acute infection by differen pathogens.

Metabolite ID BB BD P F Y

1 -Monooctadecanoylglycerol Ϊ †

1 -Monooleoylglycerol Ϊ Ϊ Ϊ Ϊ

2,5-diaminovalerolactam Ϊ Ϊ

2-Monooleoylglycerol Ϊ Ϊ Ϊ Ϊ

3-Hydroxyproline Ϊ Ϊ

Arabitol Ϊ Ϊ

Arachidonic acid † Ϊ

Arginine Ϊ Ϊ

Aspartic acid Ϊ Ϊ † p < 0.05

Campesterol Ϊ Ϊ †

Cholesterol Ϊ Ϊ †

Citric acid † † p = 0.09

Creatinine Ϊ

Dehydroascorbic acid dimer †

Docosahexaenoic acid Ϊ

Fumaric acid Ϊ Ϊ

Galactose Ϊ Ϊ

Galactosyl glycerol Ϊ

Glycerol Ϊ

Glycerol-2-phosphate Ϊ

Glycine Ϊ

Heptadecanoic acid † † p = 0.07

Hexadecanoic acid † Ϊ

Inositol Ϊ

Itaconic acid †

L-Cystine Ϊ Ϊ Ϊ Ϊ

L-Glutamic acid Ϊ

Linoleic acid †

Malic acid Ϊ Ϊ Ϊ p < 0.05

Maltose Ϊ †

Methionine Ϊ Ϊ

Myo-inositol-1 -phosphate Ϊ Ϊ Ϊ Ϊ

Myristic acid † † p = 0.05

Oleic acid †

Palmitoleic acid † Ϊ

Panthotenic acid

Pentose Ϊ

Phenylalanine †

Phosphoric acid Ϊ

p-Hydroxybenzoic acid † † † † † p = 0.06

Hydroxybenzoic acid methyl ester Ϊ

Proline Ϊ

Pyroglutamic acid † Ϊ p < 0.05

Valine Ϊ

Example 3

With the object to explore the possibility of using metabolic profiling to rapidly assess the antibiotic resistance metabolite profile of bacterial 5 infections, BALB/c mice were orally infected with strains of Yersinia

pseudotuberculosis. One group of mice was infected with a virulent wild-type strain not expressing a TEM-1 β-lactamase, thus being susceptible to β- lactams, while a second group of mice was infected with the isogenic counterpart of the virulent wild-type strain harbouring the plasmid pDSRed

10 expressing the TEM-1 β-lactamase, thus conferring resistance to ampicillin. In addition, a third non-infected group of mice was used as control. Ampicillin at a concentration of 0.5 mg per mouse was intravenously injected (i.v.) 96 hours after onset of infection and blood was drawn from the mice 1 , 2, 4 and 8 hours thereafter. The plasma metabolomic profiles were determined by

15 means of gas chromatography/time of flight-mass spectrometry (GC/TOF- MS) following extraction of the low molecular weight components to obtain a screening of the metabolome³¹. Acquired data were preprocessed using a hierarchical multivariate curve resolution strategy (HMCR)¹⁴ followed by multivariate statistical analysis by applying orthogonal partial least squares

20 discriminant analysis (OPLS-DA)¹⁹. The results from this analysis revealed that there was a clear difference in the metabolite profile between the three groups of mice (fig. 6) and it was evident that the ampicillin sensitive bacteria were affected by the antibiotic treatment and released a number of novel metabolites while the resistant bacteria showed a minor response to the addition of the drug comparable to the non-infected control group.

When mice are subjected to an oral infection of Y. pseudotuberculosis, two phases of infection can be monitored; one acute phase peaking at days 5-7, and thereafter the infection declines. Examination of mice being in the acute phase shows that the bacteria are present in the intestinal tract, preferentially in cecum and Peyer's patches and in some cases the bacteria reach the mesenteric lymph nodes. The pathogen occasionally breaks through this barrier causing a systemic infection. The above infected mice were examined for spread of the pathogen and in no case the bacteria had become systemic (data not shown). This result shows that the above method can be employed successfully not only for systemic infections but also for infections being contained within different compartments. Example 4

In this experiment, strains of methicillin-resistant Staphylococcus aureus (MRSA) (resistant to doxacillin but sensitive to vancomycin) and methicil I in-sensitive Staphylococcus aureus (MSSA) (sensitive to both doxacillin and vancomycin) were employed. Both strains were clinical isolates from Umea University, Sweden. It is expected that any clinical isolates with the same resistance pattern will give the same results. MRSA or MSSA were used to infect BALB/c mice via the i.v. route. 24 hours after infection, blood was drawn as negative control, after which doxacillin (20-40 mg/kg; Axellia Pharmaceuticals ApS, Denmark) or vancomycin (1 10 mg/kg; Meda AB, Sweden) was given i.v. to both groups of mice. Blood was drawn from the mice 1 , 4 and 8 hours thereafter. The serum metabolic profiles were determined by means of gas chromatography/time of flight-mass

spectrometry (GC/TOF-MS) after extraction of the low molecular weight components, to obtain a screening of the metabolome. Acquired data were preprocessed using a hierarchical multivariate curve resolution strategy (HMCR) followed by multivariate statistical analysis by applying orthogonal partial least squares discriminant analysis (OPLS-DA). It was established that administration of vancomycin resulted in the release of small metabolites in both groups of mice. In contrast, cloxacillin lysed the MSSA strain, releasing metabolites, but did not lyse the MRSA strain. This analysis indicated what drug could be used to treat the induced acute septicemia. Results from the experiments with MRSA and MSSA are presented in Fig 7-9.

Example 5

This experiment investigated whether it was possible to determine whether a strain of Y. pseudotuberculosis expressed ESBL. Therefore, a plasmid expressing ESBL was introduced into the wild-type

Y. pseudotuberculosis-pXenA (Caliper Life Science, USA). The resulting strain and the wild-type were used to infect BALB/c-mice orally. The ESBL plasmid had been isolated from a clinical isolate from the Swedish Institute of Infectious Control, Sweden, and transformed into Y. pseudotuberculosis- pXen4. It is expected that any ESBL plasmid from a clinical isolate will give the same result. Three days later, when yersinosis had been established as determined by in vivo imaging, either of the β-lactam antibiotics imipenem and cefotaxim (both from Sigma-Aldrich, USA) was injected i.v. after the taking of control blood samples. Doses of antibiotics were determined by extrapolation from suggested doses for humans and adjusted according to literature on animal studies. Blood was drawn from the two groups of mice 1 , 4 or 8 hours after administration of antibiotic. The serum metabolic profiles were determined by GC/TOF-MS and data processing as described in previous Examples. It was established that the wild-type strain released small metabolites into the blood after treatment with either cefalosporin or imipenem, indicating bacterial cell lysis. Conversely, the strain expressing ESBL was found to be resistant towards cefalosporin but sensitive towards treatment with imipenem, demonstrating that ESBL was expressed and possessed β-lactamase activity that could hydrolyse cefalosporin. Thus, it was possible to detect ESBL expressed from bacteria confined in the intestine by analysis of the metabolic profile of the blood. This is truly a remarkable finding given the limited number of Yersinia bacteria present in the intestine, in comparison with the number of bacteria present in the normal flora and, in addition, given the fact that the infection was not systemic. Materials and methods

Chemicals

All compounds were of analytical grade except where stated otherwise. lsoFlo®Vet (Orion Pharma Animal Health, Sweden), heparin (LEO Pharma, Denmark), ampicillin (Sigma), buprenorphine (Temgesic®, Schering-Plough), ketamine (Ketalar®, Pfizer), medetomidine (Domitor®, vet. Orion Pharma). The stable isotope-labelled internal standard compounds (ISs) [¹³C₅]-proline, [²H₄]-succinic acid, [¹³C₅,¹⁵N]-glutamic acid, [1 ,2,3-¹³C₃]-myristic acid, [²H₇]- cholesterol, and [¹³C₄]-disodium a-ketoglutarate were purchased from

Cambridge Isotope Laboratories (Andover, MA, USA); [¹³Ci2]-sucrose, [¹³C₄]- palmitic acid, and [²H ]-butanediamine-2HCI were from Campro (Veenendaal, Netherlands); [¹³Ce]-glucose was from Aldrich (Steinheim, Germany); [²H₆]- salicylic acid was from Icon (Summit, NJ, USA). N-Methyl-N- trimethylsilyltrifluoroacetamide (MSTFA) supplemented with 1 %

trimethylchlorosilane (TMCS) and pyridine (silylation grade) was purchased from Pierce Chemical Co. Stock solutions of the reference compounds and ISs (Internal Standards) were prepared at the same concentration (0.5 g μΓ¹) in Milli-Q water or methanol.

Mice

10 weeks female FVB/N mice (The Jackson Laboratories, Maine, USA) and 8 weeks female BALB/c mice (Taconic, Denmark) were given normal mouse chow and water ad libitum, and were housed under standard conditions. The experiment was according to the guidelines of the Umea University Animal Ethics Committee.

Mouse infections

The bioluminescent Y. pseudotuberculosis strain Xen 04 (Caliper Life Sciences, Hopkinton, MA, USA) was subcultured on Luria Broth (LB) agar plates supplemented with kanamycin (50 pg/ml). Bacteria to be used for infection were grown overnight in 100 ml of broth at 26 °C. Cultures were collected and resuspended to 1 .3 x 10⁷ CFU/ml in 50 ml of sterilized tap water supplemented with 150 mM NaCI. FVB/N mice were deprived of water and food for 18 h and then given water containing the bacteria. Each cage contained five mice, which had free access to bacteria-containing water for 6 h, and each animal consumed approximately 4 ml of the water. The infection dose was determined by viable count and drinking volume

measurements. For bacterial counts, organs were dissected out and homogenized in PBS, and serial dilutions of the homogenates were spread on agar plates supplemented with kanamycin.

Francisella tularensis strain LVS was grown on modified Thayer-Martin agar plates⁴² at 37 °C in 5 % CO2 atmosphere. For mouse infection, LVS was grown for 16 h on plates, diluted in PBS (pH 7.4) and 100 μΙ was injected subcutaneously in the groin area of BALB/c mice³⁵. The infection dose was determined by viable count and found to 7 x 10⁵ CFU.

BALB/C mice were infected subcutaneously with 1 x 10⁵ infectious Borrelia burgdorferi N40 (obtained from Alan G Barbour, UC Irvine)⁴⁰. Positive infection was confirmed by cultivation of ear tissue, from mice sacrificed at day 10 post infection.

BALB/C mice were infected subcutaneously with 1 x 10⁵ Borrelia duttonii 1 120 (Guy Barandon, Institute Pasteur, Paris, France)⁴¹. To confirm infection, spirochetemia development was monitored in blood by microscopic examination, and serum was withdrawn at the day of peak spirochetemia, day 6 post infection.

P. aeruginosa was used to infect BALB/c mice by a combination of previously described methods^38,39. In short, a preheated metal bar was held against the shaved back of the mice for 60 s. Subcutaneous administration of buprenorphine 100 pg/kg every 12 h was used as analgesia and ketamine 75 mg/kg + medetomidine 1 mg/kg (intraperitoneal) as anaesthesia. 1 x 10⁵ CFU P. aeruginosa ATCC 19660 (strain 180) was injected into the site of the burn; PBS was used for the control animals.

For the ampicillin resistance experiment, BALB/c mice were infected with bacteria in the drinking water as described above. One group of mice was infected with 1 .6 x 10⁸ CFU/ml wild-type strain Xen 04 not expressing a TEM-1 β-lactamase, a second group was infected with with 1 .9 x 10⁸ CFU/ml of Xen 04 harbouring pDSRed expressing the TEM-1 β-lactamase. Day 4 post infection, mice received 0.5 mg ampicillin in 100 μΙ PBS intravenously in the mouse tail vein. Blood samples were collected at 1 , 2, 4, and 8 h post injection of ampicillin, and plasma prepared as described above. Control mice were not infected but received ampicillin or PBS 96 h post infection.

β-lactamase resistance by S. aureus was analyzed in 8 week old, female BALB/c mice. Clinical isolates of MRSA or MSSA were used to infect mice in groups of 6, and 2 x 10⁶ CFU in 150 μΙ PBS were injected

intravenously. 24 h post infection, 3 mice in each group received 0.8 mg cloxacillin and 3 mice recieved 2.2 mg vancomycin in 150 μΙ PBS i.v. Blood samples were collected before onset of infection, 24 h after infection, and 1 , 4, and 8 h post injection of antibiotics. Control mice were not infected but received cloxacillin 24 h post infection. Plasma was prepared as described above.

Imaging of bioluminescence

In vivo images were acquired every two to four days using an MS® Spectrum (Caliper Life Sciences) according to the manufacturer's instructions. Acquisition and analysis of total photon emission were performed with Living Image software (version 3.1 , Caliper Life Sciences). Prior to image

acquisition, mice were anaesthetised with a constant flow of 2.5% isoflurane (IsoFlo® Vet, Orion Pharma, Animal Health) mixed with oxygen supplied from an XGI-8 gas system (Caliper Life Sciences). During the imaging process, the animals were kept anaesthetised. Images were acquired for 1-120 s with a binning of 8. All other photographic parameters were held constant.

Health status measurements

Haemoglobin (Hb) levels in blood were measured using a HemoCue reader and cuvettes (HemoCue AB, Angelholm, Sweden), and glucose by the GlucoSurePlus system (HaeMedic AB, Sweden). Analysis was performed prior to anaesthesia. Preparation of plasma samples

Tail blood samples were drawn prior to anaesthesia at indicated times the experiment, and added to heparin to achieve an end heparin

concentration of 50 U/ml. Thereafter, the samples were centrifuged at 19,600 g for 10 min at 4 °C, and separated plasma was placed in a new tube and stored at -80 °C.

Sample preparation for GC-MS analysis

Each plasma sample was prepared for GC-MS analysis as follows. A frozen sample was allowed to thaw at 37 °C for 15 min, and immediately thereafter 450 μΙ of extraction solution (methanol/water 8:1 with 1 1 ISs [each 7 ng μΓ¹]) was added to a 50-μΙ aliquot of the plasma, and the extraction sample was vortexed for approximately 10 s. For a plasma sample of insufficient volume, a 40- or 30-μΙ aliquot was extracted with a proportionally lower amount of extraction solution. After vortexing, the extraction sample was shaken vigorously at 30 Hz for 1 min in a MM301 Vibration Mill (Retsch GmbH & Co. KG, Haan, Germany), then placed on ice for 120 min, and subsequently centrifuged at 19,600 g for 10 min at 4 °C. A 100-μΙ aliquot of the supernatant was transferred to a GC vial and evaporated to dryness. Methoxymation with 15 μΙ of methoxyamine solution in pyridine (15 μg μΓ¹) was carried out at room temperature for 16 h. Finally, the sample was trimethylsilylated with 15 μΙ of MSTFA at room temperature for 1 h, after which 15 μΙ of heptane (containing 0.5 μg of methyl stearate as injection IS) was added.

GC-TOFMS analysis

Using an Agilent 7683 series autosampler (Agilent, Atlanta, GA), a 1 -μΙ aliquot of derivatised sample was injected (splitless mode) into an Agilent 6980 GC equipped with a 10 m x 0.18 mm inner diameter fused-silica capillary column chemically bonded with 0.18 μιτι DB5-MS stationary phase (J&W Scientific, Folsom, CA). The injector temperature was set at 270 °C, and helium was used as carrier gas at a constant flow rate of 1 ml min^"1. In each analysis, the purge time was set to 60 s at a purge flow rate of 20 ml mirf ¹ and an equilibration time of 1 min. The column temperature was initially kept at 70 °C for 2 min and then increased by 30 °C min^"1 to 320 °C, and held at that temperature for 2 min. The column effluent was introduced into the ion source of a Pegasus III TOFMS instrument (Leco Corp., St Joseph, Ml). The transfer line temperature was set at 250 °C for the transfer line and at 200 °C for the ion source. Ions were generated by a 70-eV electron beam at a current of 2.0 mA. Masses were acquired from m/z 50 to 800 at a rate of 30 spectra s^"1, and the acceleration voltage was turned on after a solvent delay of 165 s. Files of acquired samples were exported to MATLAB 7.3 (R2006b)

(Mathworks, Natick, MA) in NetCDF format for further data processing and analysis.

Hierarchical multivariate curve resolution (HMCR)

All data pre-treatment procedures, such as baseline correction, chromatogram alignment, time-window setting, and hierarchical multivariate curve resolution (H-MCR)¹⁴ were performed in MATLAB using custom scripts. All 90 samples (including repeated extractions) were subjected to H-MCR. Alignment and smoothing using a moving average was carried out before dividing the chromatograms into 69 time windows, from which a total of 277 chromatographic profiles (peaks representing putative derivatised

metabolites) with corresponding mass spectra were resolved by a cross- validation method. Before multivariate modelling, all peak areas were normalized to peak areas of 12 internal standards eluting over the entire chromatographic time range. Mass spectra of all detected compounds were compared with spectra in the NIST library 2.0 (as of 31 January 2001 ), using the in-house mass spectra library database established by Umea Plant Science Center (UPSC) or the mass spectra library maintained by the Max Planck Institute in Golm (http://csbdb.mpimp- golm.mpg.de/csbdb/gmd/gmd.html).

Data analysis

The H-MCR-processed GC-TOFMS metabolomics data were mean- centred and scaled to unit variance prior to any multivariate data analysis. Orthogonal partial least squares discriminant analysis (OPLS-DA) was performed to model the systematic variation in the metabolomics data related and orthogonal to pre-defined sample classes among the mouse plasma samples, here considering pre- vs. post-onset infection, control vs. post-onset infection, persistent vs. resolved infection and inter infection comparisons. Cross validation was used to determine the predictive ability of the models, and ANOVA based on the cross-validated OPLS-DA score vectors (CV- ANOVA) was employed to calculate p-values for the differences between the pre-defined sample classes in the respective models. The OPLS-DA model loadings combined with univariate p-values (Student's t-test) were used to highlight significant metabolites associated with diagnosis and/or monitoring of infection. Verification of detected metabolite profiles was done by use of OPLS-DA model predictions of samples with known fate and by calculation of univariate p-values (Student's t-test) for the pre-defined differences of interest.

Discussion

Based on the examples described above, we conclude that metabolic profiling represents a novel and interesting method for monitoring, identifying and/or diagnosing infectious diseases. The disclosed method is remarkably sensitive given the high number of bacteria present in the normal bacterial flora of the infected animal. Although metabolic profiling has already been applied to predict the outcome of diabetes, cirrhosis, and cancer in preclinical and clinical studies^12,43,44, this remarkable sensitivity was unexpected.

Moreover, many metabolites are species-specific and might therefore be exploited as biomarkers for clinical applications. Indeed, it is particularly interesting to apply metabolic profiling to bacteria, viruses, and parasites in different body fluids or tissue samples, because, as we have shown here, it is likely that during infection such pathogens generate species-specific biomarkers that can be used for precise diagnosis and as tools to monitor progression of the disease. However, being able to be identified is not an absolute prerequisite for specific biomarkers associated with an infecting organism, since, as we also have demonstrated; it is highly probable that changes in metabolic profiles will be sufficient to leave a fingerprint that can be identified by pattern recognition and coupled to a specific infectious agent.

The present invention is not limited to the above-described preferred embodiments. Various alternatives, modifications and equivalents may be used. Therefore, the above embodiments should not be taken as limiting the scope of the invention.

References

1 . Ternhag, A. et al., Emerg Infect Dis 14, 143-148 (2008).

2. Lamps, L. W. et al., Am J Surg Pathol 27, 220-227 (2003).

3. Saebo, A. et al., Eur J Intern Med 16, 176-182 (2005).

4. Schiellerup, P. et al., J Rheumatol 35, 480-487 (2008).

5. Grayston, J. T., J Infect Dis 181 SuppI 3, S402-410 (2000).

6. Puylaert, J. B. et al., Eur Radiol 7, 3-9 (1997).

7. Sakellaris, G. et al., Pediatr Surg Int 20, 155-157 (2004).

8. Antonopoulos, P. et al., Emerg Radiol 15, 123-126 (2008).

9. Hawkey, P. M. and Jones, A. M., J Antimicrob Chemother 64 SuppI 1 , i3-10 (2009).

10. Schnackenberg, L. K. and Beger, R. D., Pharmacogenomics 7, 1077- 1086 (2006).

1 1 . Jonsson, P. et al., J Proteome Res 5, 1407-1414 (2006).

12. Sreekumar, A. et al., Nature 457, 910-914 (2009).

13. Weckwerth, W. et al., Proc Natl Acad Sci USA 101 , 7809-7814 (2004). 14. Jonsson, P. et al., Anal Chem 77, 5635-5642 (2005).

15. Wilson, I. D. et al., J Chromatogr B Analyt Technol Biomed Life Sci 817, 67-76 (2005).

16. Smith, C. A. et al., Anal Chem 78, 779-787 (2006).

17. Nicholson, J. K. and Wilson, I. D., Nat Rev Drug Discov 2, 668-676 (2003).

18. Soga, T. et al., Anal Chem 74, 2233-2239 (2002).

19. Bylesjo, M. et al., Journal of Chemometrics 20, 341 -351 (2006). 20. Wold, S. et al., Chemometrics and Intelligent Laboratory Systems 44, 175-185 (1998).

21 . Sjostrom, M. et al., Pattern Recognition in Practice II. (Elsevier,

Amsterdam, 1986).

22. Burges, C. J. C, A Tutorial on Support Vector Machines for Pattern Recognition. (Kluwer Academic Publishers, Boston, 1998).

23. Steeb, W-H., The Nonlinear Workbook: Chaos, Fractals, Neural Networks, Genetic Algorithms, Gene Expression Programming, Support Vector Machine, Wavelets, Hidden Markov Models, Fuzzy Logic with C++, Java and SymbolicC++ Programs., 4th edition, ed. (World Scientific

Publishing, 2008).

24. Jain, A. K. et al., ACM Computing Surveys (CSUR) 31 , 264-323 (1999).

25. Isaksson, E. L. et al., Infect Immun (2009).

26. Heuts, F. et al., Microbes Infect (2009).

27. Hutchens, M. and Luker, G. D., Cell Microbiol 9, 2315-2322 (2007).

28. Vuong, C. et al., J Infect Dis 198, 258-261 (2008).

29. Mecsas, J. et al., Infect Immun 69, 2779-2787 (2001 ).

30. Logsdon, L. K. and Mecsas, J., Infect Immun 71 , 4595-4607 (2003).

31 . Trygg, A. J. et al., Anal Chem 77, 8086-8094 (2005).

32. Chorell, E. et al., J Proteome Res 8, 2966-2977 (2009).

33. Wuolikainen, A. et al., Amyotroph Lateral Scler 10, 229-236 (2009).

34. Eriksson, L. et al., Journal of Chemometrics 22, 594-600 (2008).

35. Forslund, A. L. et al., Mol Microbiol 59, 1818-1830 (2006).

36. Larsson, C. et al, Curr Opin Infect Dis 22, 443-449 (2009).

37. Barthold, S. W. et al., J Infect Dis 162, 133-138 (1990).

38. Dale, R. M. et al, Antimicrob Agents Chemother 48, 2918-2923 (2004).

39. Kobayashi, M. et al., J Leukoc Biol 83, 1354-1362 (2008).

40. Fikrig, E. et al., Proc Natl Acad Sci U S A 89, 5418-5421 (1992).

41 . Larsson, C. et al., Microbes Infect 8, 2213-2219 (2006).

42. Sandstrom, G. et al., Infect Immun 45, 101 -106 (1984).

43. Bain, J. R. et al., Diabetes 58, 2429-2443 (2009).

44. Xue, R. et al., Clin Chem Lab Med 47, 305-310 (2009). ITEMIZED LIST OF EMBODIMENTS

1 . A method for monitoring, identifying and/or diagnosing an infection in an individual, comprising:

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(b) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said body fluid sample and/or said tissue sample; and

(c) comparing said test metabolite profile with a control metabolite profile.

2. The method according to item 1 , wherein (b) comprises applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to analyse the metabolite composition of said body fluid sample and/or said tissue sample.

3. The method according to any one of the preceding items, wherein

(b) comprises processing the data of the metabolite composition using hierarchical multivariate curve resolution (H-MCR).

4. The method according to any one of the preceding items, wherein (b) comprises

(b1 ) applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to said body fluid sample and/or said tissue sample; and

(b2) processing the data of the metabolite composition using

hierarchical multivariate curve resolution (H-MCR).

5. The method according to any one of the preceding items, wherein c) comprises performing a multivariate statistical calculation of a separation between said test metabolite profile and said control metabolite profile.

6. The method according to item 5, wherein said multivariate statistical calculation is an orthogonal partial least squares discriminant analysis (OPLS- DA).

7. The method according to any one of the preceding items, wherein

(c) comprises comparing the test metabolite profile obtained in (b) with control metabolite profiles of healthy and infected individuals, respectively. 8. The method according to any one of the preceding items, wherein the metabolite profiles compared in (c) comprise (i) a simple profile of a small number of identified and/or unidentified metabolites unique to an infected individual and/or (ii) a complex profile of a large number of mostly unidentified metabolites.

9. The method according to item 8, wherein the metabolite profiles compared in (c) comprise a combination of (i) and (ii).

10. A method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic, comprising the method according to any one of the preceding items.

1 1 . The method according to item 10, wherein said antibiotic has been administered to said individual prior to (a), and a control metabolite profile originates from a body fluid sample and/or a tissue sample obtained from said individual before said antibiotic has been administered to said individual.

12. The method according to item 10, comprising

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(a') dividing said body fluid sample and/or said tissue sample into aliquot parts and distributing said aliquot parts into separate reaction chambers, wherein each reaction chamber comprises an antibiotic or a negative control;

(a") incubating said aliquot parts in said reaction chambers;

(b) analysing a metabolite composition contained in said reaction chambers, including processing data of the metabolite composition, to determine a test metabolite profile of the aliquot part in each reaction chamber; and

(c) individually comparing said test metabolite profiles with a control metabolite profile from bacteria resistant to or susceptible to an antibiotic;

(d) identifying the infection as being caused by bacteria resistant to or susceptible to an antibiotic.

13. The method according to item 1 1 or 12, wherein the test metabolite profile(s) comprises metabolites being degradation products of said antibiotic and/or metabolites from lysed bacteria. 14. The method according to any one of the preceding items, wherein said individual is further affected by a chronic disease.

15. The method according to any one of the preceding items, wherein said infection is monitored over time and said individual receives treatment for said infection.

16. A method for assessing the efficacy of a treatment of an infection, comprising the steps

(a) identifying an individual as having said infection;

(b) obtaining a body fluid sample and/or a tissue sample from said individual;

(c) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said sample;

(d) comparing said test metabolite profile with a control metabolite profile;

(e) treating said individual for said infection with said treatment;

(f) repeating steps b)-d) any number of times while performing said treatment.

17. The method according to any one of the preceding items for clinical studies of pharmaceutical candidates for treatment of infection.

18. An apparatus for monitoring, identifying and/or diagnosing an infection in an individual, said apparatus being configured to perform the method according to any one of items 1 -17.

19. A computer-usable medium comprising a computer program code configured to cause a processor to execute the method according to any one of items 1 -17.

Claims

1 . A method for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic, comprising:

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(b) analysing a metabolite composition of said body fluid sample and/or said tissue sample, including processing data of the metabolite composition, to determine a test metabolite profile of said body fluid sample and/or said tissue sample; and

(c) comparing said test metabolite profile with a control metabolite profile.

2. The method according to claim 1 , wherein said antibiotic has been administered to said individual prior to (a), and a control metabolite profile originates from a body fluid sample and/or a tissue sample obtained from said individual before said antibiotic has been administered to said individual.

3. The method according to claim 1 , comprising

(a) obtaining a body fluid sample and/or a tissue sample from said individual;

(a') dividing said body fluid sample and/or said tissue sample into aliquot parts and distributing said aliquot parts into separate reaction chambers, wherein each reaction chamber comprises an antibiotic or a negative control;

(a") incubating said aliquot parts in said reaction chambers;

(b) analysing a metabolite composition contained in said reaction chambers, including processing data of the metabolite composition, to determine a test metabolite profile of the aliquot part in each reaction chamber;

(c) individually comparing said test metabolite profiles with a control metabolite profile from bacteria resistant to or susceptible to an antibiotic; and (d) identifying the infection as being caused by bacteria resistant to or susceptible to an antibiotic.

4. The method according to any preceding claim, wherein the test metabolite profile(s) comprises metabolites being degradation products of said antibiotic and/or metabolites from lysed bacteria.

5. The method according to any preceding claim, wherein (b) comprises applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to analyse the metabolite composition of said body fluid sample and/or said tissue sample.

6. The method according to any preceding claim, wherein (b) comprises processing the data of the metabolite composition using hierarchical multivariate curve resolution (H-MCR).

7. The method according to any preceding claim, wherein (b) comprises

(b1 ) applying gas chromatography/time of flight-mass spectrometry (GC/TOF-MS) to said body fluid sample and/or said tissue sample; and

(b2) processing the data of the metabolite composition using hierarchical multivariate curve resolution (H-MCR).

8. The method according to any preceding claim, wherein (c) comprises performing a multivariate statistical calculation of a separation between said test metabolite profile and said control metabolite profile.

9. The method according to claim 8, wherein said multivariate statistical calculation is an orthogonal partial least squares discriminant analysis (OPLS-DA).

10. The method according to any preceding claim, wherein (c) comprises comparing the test metabolite profile obtained in (b) with control metabolite profiles of healthy and infected individuals, respectively.

1 1 . The method according to any preceding claim, wherein the metabolite profiles compared in (c) comprise (i) a simple profile of a small number of identified and/or unidentified metabolites unique to an infected individual and/or (ii) a complex profile of a large number of mostly unidentified metabolites.

12. The method according to claim 1 1 , wherein the metabolite profiles compared in (c) comprise a combination of (i) and (ii).

13. An apparatus for identifying an infection as being caused by bacteria resistant to or susceptible to an antibiotic, said apparatus being configured to perform the method according to any one of claims 1 -12.

14. A computer-usable medium comprising a computer program code configured to cause a processor to execute the method according to any one of claims 1 -12.