WO2000053798A1

WO2000053798A1 - Detection of microorganisms responsible for urinary tract infections

Info

Publication number: WO2000053798A1
Application number: PCT/GB2000/000792
Authority: WO
Inventors: Selwayan Saini; Jan Leiferkus
Original assignee: Cranfield University
Priority date: 1999-03-06
Filing date: 2000-03-06
Publication date: 2000-09-14
Also published as: GB9905102D0; AU2927400A

Abstract

A sample (e.g. urine from a patient possibly having a urinary tract infection) is tested for microorganisms by: 1) enabling growth of the micoorganisms; 2) addition of substrates convertible by microorganisms of interest into detectable products; and 3) detection of the products. Different microorganisms of interest may have different patterns of utilization of the substrates and thus lead to characteristic products mixtures. Thus analysis of the product mixture can be used to identify the microorganisms.

Description

DETECTION OF MICROORGANISMSRESPONSIBLEFORURINARY TRACT INFECTIONS

This invention describes a novel method that permits the identification of microorganisms responsible for Urinary Tract Infections (U Is) in 5 humans and some animals. In the preferred embodiment, the method significantly improves on current UTI diagnostic tests as it is significantly faster, can be performed at lower cost, and can be automated or designed to be very simple, removing the requirement of a skilled operator. In addition, the method permits parallel antibiotic 0 sensitivity testing which normally requires several days before a result is known.

In the preferred methodology, a urine sample is firstly collected from a patient presenting symptoms of UTI. A sterile, non-selective growth 5 medium is then added to the urine to encourage growth of endogenous infective UTI microorganisms. The next stage involves the addition of certain c einical factors to urine that enable UTI microorganisms to produce certain indicator or marker products. These products are then detected using a variety of methods . Detection can be achieved within 0 the body of the liquid or in the headspace above the urine sample or both depending on the detection technique employed.

Different UTI microorganisms in urine produce different indicator products depending on the chemical nature of the added factors . Examples 5 of chemical factors, indicator products and detection methods that can be used for the detection of the two most common UTI bacteria, E. coli and P. mirabilis in urine , are described .

Urinary Tract Infections The urinary tract in humans is lined with a sheet of epithelium that is continuous with that of the skin. Therefore, the epithelial surface is a potential pathway for entry of microorganisms from the outside world. Most urinary tract infections (UTIs) arise by the ascent of bacteria following colonisation of the periurethral area by faecal organisms (especially Escherichia coli ) . Hematogenous infection of the kidney is much rarer. The main defences against UTIs are the flow of urine and the shedding of epithelial cells to which bacteria may be attached. Immune defences (humoral or cellular) play little role here. In view of the ready access of bacteria to the urinary tract, it is not surprising that UTIs are second in incidence only to infections of the respiratory tract. They rank first among the bacterial diseases of adults that come to the attention of physicians. The majority of patients are women, probably because the female urethra is much shorter than the male urethra. An antibacterial effect of prostatic secretions may also offer some protection to the male. Thus, bacteriuria (the presence of bacteria in the urine) , whether symptomatic or asymptomatic, is generally more common in women than in men at all ages and recurrent episodes of UTIs afflict about one in ten women at some time in their life. As many as 20% of women in the US are known to have an episode of urinary tract infection by the age of 30 years. A number of bacteria are known to cause UTIs (in order of prevalence) : Escherichia, Proteus mirabilis, other enterobacteria Cilebsiella, E-ifcerojbacler, Serratia, other members of Proteus sp.), Chlamidiae, Staphylococcus saprophyticus, Streptococci (enterococci and group B streptococci) , Pseudomonas aeruginosa (in complicated (nosocomial) infections) , Neisseria goπorrheae, Mycoplasma species ( . hominis, M. geni talium, Ureaplasma urealyticum) . Further species types include Trychomonas vaginalis and different species of Amoebae which are of the protozoa genera.

State-of-the-art Diagnostic Methods for UTIs

The most common tests for urinary tract infections include:

1. .Nitrite dipstick test (eg Biotron) . Undertaken by GP - the test gives a non-specific indication of bacterial presence, no measure of bacterial numbers is made or potential to replicate. Gives results in minutes but gives no information of the identity of the bacteria present.

2. Microscopy. Undertaken in a clinical microbiology department. Involves the visual identification of bacterial presence and' again no measure is made of potential of microorganisms to replicate. Requires transport of urine sample to the laboratory, minimal sample preparation. Once analysed, gives result in minutes but gives limited information of the identity of the bacteria present.

3. Automated analysis. Undertaken in a clinical microbiology department. Involves the automated assessment of cell and bacterial numbers in urine. Several commercial instruments available (for example, the Questor™ system) . The test requires transport of urine sample to the specialist laboratory. Requires limited sample preparation, however once processed, gives result in minutes but gives no information of the identity of the bacteria present . 4 . Urine cul ture . Undertaken in a clinical microbiology department. This is the definitive and most informative assay currently available. Essential bacteria within the urine sample are cultured on specialist selective agar plates. Several protocols exist and information obtained ranges from initial classification of bacteria, through to characterisation of antibiotic sensitivity. Requires transport of urine sample to the laboratory and some sample preparation. Greatest limitation is that assay requires from 1 to 5 days to enable adequate bacterial growth.

Recently, more elaborate and highly specific test procedures such as gene probes, antigenic tests (antibody reactions, counter- immunoelectrophoresis, E ISA, latex agglutination, immunofluorescence, etc.) have also been proposed for detecting bacteria in UTIs. However all these techniques require obligatory and extensive sample preparation steps, are expensive and require a high level of user competence, significantly raising the overall cost of the test.

The Invention

The invention is embodied in a specific method that provides information on the type of UTI microorganisms present in urine samples and when used with certain detection methods, the number of UTI organisms present. The test can be configured to produce a result in a few hours depending on the mode of detection.

The method is based on the identification of UTI microorganisms in urine, or in the headspace above urine, as opposed to a culture medium. Urine may be diluted with additional components if necessary (generally ≥50% v/v urine) .

The method generally consists of three component stages - (1) (if necessary) growth of organisms in urine to increase numbers and provide biochemical activity; (2) addition of chemical factors known to be converted into marker products; (3) Detection of marker products using one or more techniques .

Importantly, the method does not require extensive sample preparation such as prior plating or bacteria isolation as used in many existing "rapid" methods such as UTI commercial test kits (such as bioMerieux'ε Rapidec UR, which requires previous plating and incubation) . Instead the invention permits the operator to analyse untreated urine samples and obtain results within a matter of hours following sample collection.

In a preferred embodiment, following collection of a urine sample from a patient suspected of having an infection of the urinary tract, the following step are undertaken as part of the method of detection of UTI organisms: 1. Addition of a non-specific growth medium to urine to encourage growth of endogenous microorganisms . Such media are well known to those skilled in the art and include brain heart infusion media, nutrient broth etc. 2. Addition of specific chemical factors to the urine which are utilised by endogenous UTI microorganisms to produce certain products that may be used as indicators or markers for that organism when a suitable detection method is employed. The invention is based on the fact that different UTI organisms utilise specific added chemical factors producing different indicator products. As long as the detection method (or combination of detection methods) used is sufficiently discriminatory for different indicator products, it is possible to identify UTI infecting organisms present in urine. It should be noted that these chemical factors need not be natural growth substrates, metabolic substrates, or biological precursors.

Any chemical factor which is converted, by the target organism, into an analyte which indicates, or provides part of an indication, of the presence of that organism is also included in the term chemical factor. 3. The detection method can be one that analyses either the urine itself or the headspace above the urine or both.

The invention described above is demonstrated in the following experimental detection systems for UTI organisms in urine (with the exception of detection method 1) . Brief Description of Drawings Figure 1

Typical sensor array response {E. coli samples are shown) . The maximum response was used for the further data processing.

Figure 2

Figure showing the maximum average sensor responses for samples containing the different bacteria used with liquid medium A. It is clearly seen that sensors 3, 6, 7 and 8 provide the most information while sensors 4 and 5 are redundant.

Figure 3

Plot of the first two principal components produced by PCA using liquid media A. Good discrimination between different types of samples have been achieved. Dots show the datapoints which were included into the analysis (base data) , ^® symbols show points not included in the analysis (interpolations for test samples) . It is clearly seen that interpolated points are falling within their corresponding base data clusters. Clusters are marked as follows: 1- Blank; 2 - E. coli ; 3 - P. mirabilis; 4 - Mixture of 2 and 3; 5 - E. cloacae; 6 - K. aerogenes .

Figure 4

Shows a plot of the first two principal components produced by PCA using urine. Some discrimination is seen although there is significant overlap between the mixed culture of E. coil and Proteus and E. coli alone . Figure 5

Shows the plot obtained using the first and second functions in multiple discriminant analysis using urine test media. In this case, clear improvement over Figure 4 is seen in the clustering of the different samples as indicated in the diagram. Unseen samples (indicated by ^®) fall into their corresponding base data clusters .

Figure 6

Graph showing DPSV voltammograms obtained when different volumes of headspace sample obtained from E. coli infected urine were injected into the electrolyte solution. Results of injections from urine with no infection (blank) , an air sample and lOm of liquid urine are also shown.

Figure 7

Graph showing DPSV voltammograms obtained when different volumes of headspace sample obtained from Proteus -niraJbi--is infected urine were injected into the electrolyte solution. A clear voltammetric peak is seen in this graph at -0.1V which increases in magnitude with the volume of head space sample injected into the electrolyte. Results of injections from urine with no infection (blank) , an air sample and lOmL of liquid urine sample are also shown which appear to have a negligible effect on the measurement.

Description of Embodiments 1. In the first example, urine is replaced by a standard laboratory growth medium in order to illustrate the basic features of the method. In further examples, the test is used with simulated infected urine samples. The organisms, E. coli , Proteus mirabilis, K. aerogenes and E. cloacae were detected using a non-selective liquid growth media (peptone water) supplemented with the chemical factors -Methionine and Lactose. The indicator products were detected in the headspace above the medium using an array of metal oxide sensors in 18 hours. 2 . E. coli and Proteus mirabilis, the two most frequently infecting UTI organisms, in simulated infected urine samples using non-selective growth media (peptone water) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected in the headspace above the urine using an array of metal oxide sensors in 18 hours.

3. E. coli and Proteus mirabilis, the two most frequently infecting UTI organisms, in simulated infected urine samples using non-selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected in the headspace above the urine using an array of metal oxide sensors in 6 hours . 4. Detection of E. coli and Proteus mirabilis in simulated infected urine samples using non-selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator product, ethanol, was detected in the body of the urine using an enzymatic detection method in 18 hours. 5. E. coli and Proteus mirabilis , the two most frequently infecting UTI organisms, in simulated infected urine samples using non-selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected using a scanning electrochemical detection method in 18 hours .

Experimental description

General Methods

Preparation of Bacterial cultures

The strains of E. coli , P. mirabilis, K. aerogenes and E. cloacae were obtained from clinical specimens (blood and urine samples) .

Indicator odour detection using a commercial ' electronic nose' device (used in detection 1,2,3)

A number of commercial electronic nose devices have appeared in recent years . These vary in the sensor material and number of sensors employed but are all generally used for odour analysis. The device employed in this work was the eNOSE 5000 produced by EEV Ltd, UK.

Detection 1 - Determination of E. coli , Proteus mirabilis , K. aerogenes and E. cloacae using standard laboratory growth medium

Summary - A non-selective liquid growth media (peptone water) was supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected using an array of metal oxide sensors housed within a commercial electronic nose instrument in 18 hours.

Specific Method Liquid medium was prepared from Peptone Water (Oxoid, UK) according to the manufacturer's instructions, supplemented with chemical factors 0.1M L-methionine and 1% (wt/vol) of lactose and sterilised by autoclaving at 120° ° C for 15 mins. The final medium was distributed with sterile precautions into 240 ml screw-cap jars, with septa in the caps, in 30 ml aliquots. After inoculation they were incubated at 37° ° C for 18 hours (unless otherwise stated) , after which containers were allowed to cool to room temperature (20° ° C ±±2°° C) .

A glass syringe was used to remove 5 cm³ headspace samples though the jar's septum and injected into a gas sampling system of a commercially available electronic nose containing an array of eight metal oxide semiconductor electrodes (EEV Ltd) .

Results and Discussion Figure 1 shows the response obtained at each of the eight MOS elements- in the sensor array on exposure to the headspace sample from an E. coli culture. The general shape of the graph in this figure is representative of the other cultures that were analysed (individual cultures and mixtures - see later) . The responses provided by each sensor in the array were reduced to the maximum response for each sensor, giving eight variables for each of the ten different headspace samples obtained from culture flasks . A number of different methods are used for graphical representation of data to help with pre-processing evaluation; a bar graph was chosen for the current work (Figure 2) . It is noticeable that individual sensors respond differently to volatiles of the headspace from each sample. Specific profiles are obtained for each bacterium that were later used for differentiation between the cultures. It is also apparent that negative responding sensors (sensors four and five) give very similar responses for all the samples and can therefore be ignored.

There are numerous computational techniques available that can aid the classification of responses acquired from a sensor array. These include multivariate data reduction techniques such as principal components analysis (PCA) , multiple discriminant analysis (MDA) and artificial neural networks (ANNs) . PCA was chosen here because it is relatively straightforward to apply and its results are readily visualised and explained. PCA is capable of reducing a large number of variables to a much smaller number of principal components which capture the vast majority of variance in the data (Jackson, 1980) , by removing cross- correlation and redundancy. This reduces the dimensionality of data considerably, enabling effective visualisation, regression and classification of multivariate data (Massart et al . , 1988) . By reducing instrumental responses comprising a large number of variables to a small number, n, of principal components, each response can be represented as a point in -.-dimensional space (if n is less than four, these points can be conveniently visualised) . In a successful discrimination, responses appear in distinct clusters in space according to their classification and the space can be divided into regions associated with the different classifications. Unseen responses can therefore be identified by the region of n-dimensional space in which they fall .

A total of ten response patterns were obtained for each bacterial culture (E. coli , P. mirabilis, E. cloacae, K. aeroges and the clinically relevant mixture of P. mirabilis and E. coli) . Seven of these were used to generate regression models and the remaining three were used to test the models generated. Figure 3 shows the results of applying PCA to the sensor array data. Using two principal components, the sensor responses clearly form distinct clusters according to the bacterial content of test medium A. All unseen samples fall into the appropriate cluster, indicating 100% classification accuracy. Blank medium, E. coli, P. mirabilis, E. cloacae, K.aeroges and the clinically relevant mixture of P. mirabilis and E. coli are all clearly distinguishable.

2. Detection 2 - Determination of E. coli and Proteus mirabilis in simulated infected urine samples

Summary - This was achieved using non-selective growth media (peptone water) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected using an array of metal oxide sensors housed within a commercial electronic nose instrument within 18 hours. Specific Methods

Urine samples were collected from one of the authors and sterilised by filtration through a 0.2 mm cellulose nitrate membrane. The urine test medium consisted of 0.2M potassium di-hydrogen phosphate/ disodium hydrogen phosphate buffer at approximately pH 7.2. The buffer was sterilised in an autoclave for 20 minutes at 120°C. Following this, the chemical factors, 0.4M L-methionine and 4% (w/v) lactose were dissolved in phosphate buffer while hot (60 °C) and sterilised by filtration through cellulose nitrate membrane filters, 0.2 mm pore size. Double strength peptone water (Oxoid, UK) was autoclaved as before. The final urine test medium was prepared by mixing 1 part buffer, 1 part double strength peptone water with two parts of urine.

Clinically significant bacteriuria (bacteria in urine) corresponds to at least 5.10^s bacterial cells/ml. In order fcr the experiment to correspond to the clinical situation and be compatible with the electronic nose assay, the bacterial inoculum was chosen to be 5- 10" cells /ml of liquid medium. This imitates a one in two dilution of infected urine containing at least 10⁵cells/ml. Cells counts revealed actual concentrations of cells were in the correct order of magnitude: E. coli - 4.8-10⁴ cells/ml, P. mirabilis - 5.1- 10" cells/ml (per ml of culture medium) . A glass syringe was used to remove 5 cm³ headspace samples though the jar's septum and injected into a gas sampling system of a commercially available electronic nose containing an array of eight metal oxide semi-conductor electrodes (EEV Ltd) . These samples were then incubated for 18 hours at 37 °C after which time they were subjected to odour analysis using the electronic nose. A glass syringe was used to remove 5 cm³ headspace samples though the jar's septum and injected into a gas sampling system of a commercially available electronic nose containing an array of eight metal oxide semi-conductor electrodes (EEV Ltd) .

Results and Discussion

E. coli and P. Mirabilis are responsible for up to 80% of uncomplicated UTIs and thus their detection and differentiation in urine is of clear benefit to medical diagnosis. Similar response curves were obtained with urine test medium inoculated with E. coli and P. Mirabilis as previously seen in Figure 1 above. The number of samples tested and the treatment of data was identical to experiment 1. With PCA, using two principal components , the sensor responses formed clusters as seen in Figure 5 according to the bacterial content of the sample. Seven of these were used to generate regression models and the remaining three were used to test the models generated. All unseen samples fall into the appropriate cluster, indicating 100% classification accuracy with the exception of E. coli and mixed clusters, which show some degree of overlap. In order to improve the accuracy of classification, a different chemometric operation was used based on multiple discriminant analysis (MDA) . Using MDA, clearly resolved clusters are obtained and unseen samples are seen to fall into the appropriate clusters indicating 100% classification accuracy. 3. Detection 3 - Determination of E. coli and Proteus mirabilis in simulated urine in 6 hours

Summary - detection in simulated infected urine samples using non- selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected in the headspace above the urine using an array of metal oxide sensors in 6 hours .

Detection 2 showed the test method is compatible with urine and capable of determining clinically significant numbers of UTI bacteria. In this detection, we aim to reduce the time of diagnosis by using an alternative test medium (brain heart infusion) with the chemical factors, methionine and lactose. In addition, a further method of data analysis was employed (artificial neural networks) to analyse the signals generated by the array based sensors of the electronic nose.

Specific Methods

The non-specific growth medium comprised: sterile 0.2 M KH₂P0₄. Na₂HP0₄ buffer at pH 7.2. Double strength brain heart infusion. 0.4 M methionine, 4% lactose dissolved on hot buffer (60%) which was subsequently sterilised by filtration prior to use. The final composition was one part buffer + 1 part BHI plus 2 parts urine plus 0.4 M methionine, 4% lactose. After six hours incubation in 20ml containers, a glass syringe was used to remove 5 cm³ headspace samples though the container' s septum and injected into a gas sampling system of a commercially available electronic nose containing an array of eight metal oxide semi-conductor electrodes (EEV Ltd) .

Results and Discussion

The data from the eight sensors, which comprise the electronic nose, were interpreted using an artificial neural network (ANN) , with the aim of classifying the sample from which the data were collected. The network used is of the feed-forward type. This comprised:

1. A linear input layer, to which data from the electronic nose is presented. 2. A log-sigmoidal hidden layer.

3. A log-sigmoidal output layer, consisting of one neuron for each class of odour.

Every neuron in each layer is attached to every neuron in the next layer by a number of connections with adjustable weights. Before the ANN can be used it must be suitably trained. This involves presenting the responses for a number of known samples to the ANN and adjusting the connections between network layers in order to achieve the desired mapping between input (data from the nose) and output (identity of the sample) . The output layer is defined such that a neuron outputs a value of 1 if the sample belongs to the class associated with that neuron, and outputs 0 otherwise. The training algorithm used is backpropagation of errors, with dynamic learning rate and momentum. Once the network had been trained, interpretation of the data acquired from a sample involved several steps :

1. Pre-processing. The raw data from each sensor takes the form of a tailing peak of resistance over time, typically several minutes in length. This response is reduced to a single number by using only the peak height. The data for all the sensors are- then mean centred and rangescaled.

2. ANN processing. The pre-processed data are presented at the input layer of the ANN and the values of the output neurons determined.

3. Interpretation of ANN result. The class of the sample is identified by determining the output neuron with the highest value. If several outputs are close to 1 then the result is deemed to be inconclusive.

Results of neural network analysis

Neural networks as described above were simulated using software developed in-house. Networks were trained to differentiate between blank (no infection), Proteus mirabilis and E.coli. Using a leave-one- out method of cross-validation it was found that networks with five hidden neurons were able to correctly classify the responses of every sample from a set of 13 (5 Proteus, 3 E.coli and 5 blank) . Nature of chemical indicator products generated by E. coli and Proteus mirabilis

Work using gas chromatography shows that Proteus in non-specific media supplemented with 0.1M L-methionine, forms large amounts of the products dimethyl disulphide and methyl mercaptan.

In contrast, it has been shown that E. coli forms very small amounts of sulphur products from methionine additives and preferentially utilises lactose (or arabinose) to produce large amounts of ethanol. Ethanol is the principal product of lactose fermentation in E. coli and therefore can be used as a indicator compound in the detection of this organism.

Based on knowledge gleaned from gas chromatography on the different types of principal products produced by E. coli and Proteus from added chemical factors, it becomes feasible to design alternative detection systems to the array based electronic nose detector described in examples 1-3. The following two examples (4 and 5) take advantage of the different, products produced by E. coli and Proteus mirabilis using two different detection systems.

4. Detection 4 - Determination of E. coli and Proteus mirabilis in urine on the basis of ethanol detection Summary - Detection in simulated infected urine samples using non- selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator product, ethanol, was detected in the body of the urine using an alcohol detection method in 18 hours.

Specific Methods The non-specific growth medium previously used in detection method 3 was employed. This comprised: 0.2 M KH₂P0₄.Na₂HP0₄ sterile buffer at pH 7.2. Double strength brain heart infusion. 0.4 M methionine, 4% lactose dissolved on hot buffer (60%) which was subsequently filtered prior to use. The final composition was one part buffer + 1 part BHI plus 2 parts urine plus .4 M methionine, 4% lactose.

Principle of detection:

Ethanol produced by E. coli (or other UTI organisms) in urine is oxidised to acetaldehyde with the simultaneous reduction nicotinamide adenine dinucleotide (NAD) . This reaction is catalysed by the enzyme, alcohol dehydrogenase (ADH) . The reduced NAD (NADH) absorbs light at 340nm. Thus, an increase in absorbance at 340nm is due to production of NADH from NAD and thus directly proportional to the amount of ethanol oxidised. This detection system is available in a commercial Sigma Diagnostics" ethanol test kit (single assay vials Cat. No. 332 -A each containing 1.8 mmol NAD, 150 Units ADH (yeast) and buffer salts. Glycine buffer reagent 0.5 mol/L, pH 9.0, with trapping agent). The ethanol test kit method was used to determine the ethanol content of urine inoculated with either E. coli and Proteus mirabilis or blank. A volume of 10 mL of urine that had been incubated for 18 hours was added to the 3ml Sigmal test kit vial and was then incubated at 23°C for 10 minutes. The absorption was then determined at 340nm and converted to the ethanol concentration from a standard curve.

Results

Ethanol mg/ml after 18 hours Urine Blank 3.8

Proteus mirabilis 18.8

E. coli 190

Clearly a significantly larger amount of the indicator product, ethanol, is produced from E. coli compared to Proteus mirabilis over the 18 hour time frame. This is a useful result for the detection of E. coli since measuring ethanol at an early stage in the incubation period should provide an ethanol signal due to E. coli since Proteus mirabilis would have produced negligible amounts of ethanol. Moreover, 90-95% of non- hospital infections are due to E. coli alone (mixed infections are rare). Thus the enzymatic method for ethanol from E. coli infections may be used as a rapid method when used in conjunction with the invention. It is further conceivable that the concentration of alcohol measured could be related to the number of UTI bacterial cells. It should also be possible to design alternative alcohol sensing methods such as amperometric enzyme electrodes utilising alcohol oxidase or alcohol dehyrogenase .

5. Detection 5 - Determination of E. coli and Proteus mirabilis in urine using a scanning electrochemical technique

Summary - Detection of E. coli and Proteus mirabilis in simulated infected urine samples using non-selective growth media (brain heart infusion) supplemented with the chemical factors L-Methionine and Lactose. The indicator products were detected using a scanning electrochemical detection method in 18 hours.

Specific Methods

Separate Proteus, E. coli and blank urine samples were incubated for 18 hours in 20ml containers after which 5ml, 10 or 15ml headspaces were removed using a syringe and injected into a electrochemical cell containing electrolyte. The cell contained 10 ml of 0.1 M NaOH as electrolyte. Electrochemical measurements were performed using an Autolab Pstat 10 (Eco-Chemi) . Electrochemical measurements were based on dual pulse staircase voltammetry and were carried out in static electrochemical cell. A magnetic stirrer (Rank Bros) was used to return the cell to homogeneity between scans by stirring at 500 rpm for 10 s. A 1.6 mm diameter platinum disk working electrode (BAS) was used with a platinum wire counter (BAS) and a Ag/AgCl reference electrode (BAS) .

Dual Pulse Staircase Voltammetry. Each DPSV scan consisted of a 3 s, 0.7 V pulse, to remove adsorbed fouling agents and form platinum oxide on the electrode surface, and a 2 s, -0.9 V pulse, to regenerate the surface by removing the oxide layer, followed by a scan from -1.0 V to +1.0V in steps of 10 mV at a rate of 0.5 V s"¹. The current was recorded at the end of each potential step during the scan.

Results and Discussion

Figures 6 and 7 show the DPSV data obtained from urine samples containing E. coli and Proteus mirabilis respectively. The greatest feature observed in infected urine samples compared to non-infective samples (blanks) was the major peak centred at -0.1V vs. Ag/Ag Cl reference electrode. Interestingly., when similar volumes of headspace are used for each bacterial sample, there is a significantly greater response seen with Proteus mirabilis compared to E..col (which shows a very weak signal) . Since the principal products from methionine utilisation by Proteus mirabilis are dimethyl disulphide and methyl mercaptan, it follows that the principal voltammetric peak may correspond to electrochemical detection of these compounds in the experiment .

If detection metnod 4 (which detects the principal product (alcohol) from E. coli) and detection method 5 (which appears to detect sulphurous products from Proteus mirabilis) were used together, and in conjunction with commercial nitrite dipsticks, it is conceivable that a combined UTI test for E. coli and P. mirabilis could be developed.

An alternative method of detecting Proteus mirabilis alone could be based on a commercial electrochemical gas sensor for organic sulphur compounds used in conjunction with the method described in this patent.

The invention can also be used to provide antibiotic sensitivity testing, possibly in parallel with identification and/or quantification of organisms . Thus two samples of the material under investigation (e.g. urine) are taken, and one is tested with one or more antibiotics. Both samples are then subjected to the detection method provided by the invention. Comparison of the results shows the effect of the antibiotics. For example if the antibiotic-treated sample apparently contained microorganisms A and B whereas the other sample contained A, B, C and D, it could be concluded that the original material contained A, B, C and D, of which A and B were resistant to the antibiotic used. Relevant Literature

Cowan & Steel, Manual for Identification of Medical Bacteria, Cambridge

University Press, UK Hayward, N. J. , Jeavons, T. H. , Nicholson, A. J. C, Thornton, A. G.,

Development of specific tests for rapid detection of Escherichia coli and all species of Proteus in urine, J. Clinical Microbiology, 195,

Sept. 1977.

Mitruka, B. M. , Jonas, A. M. Rapid identification of bacteria in cultures, blood, and other body fluids of animals by gas chromatography,

Bacteriol . Proc. , 72, 87, 1972.

Kundargi, R. S. and Mitruka, B. M. , Rapid detection of bacterial infections by gas chromatography, Proc. Int . Congr. Bacteriology,

Jerusalem, 1973. Mitruka, B. M. , Detection and identification of microorganisms by gas chromatography, Symp. Proc. Int . Congr. Food Microbiology and Hygiene,

Kiel, Germany, 1974.

Carlsson, J., Simplified gas chromatographic procedure for identification of bacterial metabolic products, Appl . Microbiol . , 25, 287, 1973.

Mitruka, B. M., Ed., Gas Chromatographic Applications in Microbiology and Medicine, John Wiley & Sons, USA, 1975.

Moore, W. E. C, Relationships of metabolic products to taxonomy of anaerobic bacteria, Int . J. Syst . Bacteriol . , 20, 535, 1970. Mitruka, B. M. , Rapid and automated identification of microorganisms in clinical samples by gas chromatography, Symp. Applied Microbiology,

John Wiley & Sons, New York NY, USA, 1973.

Brooks, J. B., Identification of disease and disease causing agents by analysis of spent culture media and body fluids with electron capture gas liquid chromatography, Microbiology - 1375, Schlessinger, D. , Ed.,

American Society for Microbiology, Washington, DC, USA, 1975.

Mitruka, B. M., Jonas, A. M. , Alexander, M. , Kundargi, R. S., Rapid differentiation of certain bacteria in mixed populations by gas chromatography, Yale J. Biol . Med. , 46, 104, 1973.

Gardner, J. w. , Barlett, P., A brief history of electronic noses,

Sensors And Actuators B, 18-19, 211, 1994.

Bartlett, P., Elliott, J., Gardner, J. W. , Application of, and developments in, machine olfaction, Annali di Chimica (Societa Chimica Italiana) , 87, 33, 1997.

Craven, M. A., Gradner, J. W. , Bartlett, P. N. , Electronic noses - developments and future prospects, Trends in Analytical Chemistry, 15,

9, 486, 1996.

G. H., Bartlett, P. N. , Gardner, J. w. , Ed., Sensors and sensory systems for an electronic nose, NATO ASI Series E: Applied Sciences, Vol. 212,

Kluwer, Dordrecht, Belgium, 1992.

Kress, E., Ed., Handbook of Biosensors and Electronic Noses: Medicine,

Food and Environment, CRC Press, Boca Raton, 1996.

Chatfield, c, Collins, A. J. , Introduction to Multivariate Analysis, Chapman and Hall , London, 1980. Moy, . , Collins, M. , Electronic noses and artificial neural networks,

American Laboratory, Vol. 28, No. 3, 22, 1996.

Byfield, M. P., May, I. P., Olfactory sensor array systems - the electronic nose, GEC Journal of Research, 13, 1, 17-27, 1996. Pearce, T. C, Computational parrallels between the biological olfactory pathway and its analogue 'the electronic nose' , Biosystems, 41,

43-90, 1997.

Ping, W., Jun, X., A novel recognition method for electronic nose using artificial neural network and fuzzy recognition, Sensors and Actuators, 37, 3 , 169 -174 , 1996

Gibson, T. D., Posser, 0., et al. Detection and simultaneous identification of microorganisms from headspace samples using an electronic nose, Sensors and Actuators B, 44 , 413 -422 , 1997 Dickinson, T. A., White, J., Kauer, J. S., Walt, D. R., Current trends in 'artificial-nose' technology, Trends in biotechnology, 16, 250-258,

1998

Haugen, J. E., Kvaal, K., Electronic Nose and artificial neural network,

Meat Science, 49, No . Suppl 1 , 273 -286, 1998

Roussel, S., Forsberg, G., et al, Optimisation of electronic nose measurements. Part I: methodology of output feature selection, Journal of Food Engeneering 37, 207-222 , 1998.

Jackson, J. E. (1980). Principal components and factor analysis: Part I

- Principal components. J. Qual. Tech, 12 , 201 -213. Massart, D. L., Vandeginste, B. G. M. , De ing, S. N. , Michotte, Y. & Kaufman, L. (1988) . Chemometrics : A Textbook. Amsterdam: Elsevier Science Publishers .

Claims

1. A method of detecting microorganisms in a sample comprising:

(a) if necessary, treating said sample to enable growth of the microorganisms therein to increase their numbers and/or biochemical activity; (b) adding one or more chemical factors known to be convertible by one or more candidate microorganisms possibly present in said sample into detectable marker products; and

(c) analysing for the presence of said marker product or products.

2. A method according to Claim 1 wherein the analysis of step (c) is carried out on a liquid portion of the sample.

3. A method according to Claim 1 or Claim 2 wherein the analysis of step (c) is carried out on a gas sample withdrawn from headspace over the sample.

4. A method according to any preceding claim wherein step (a) involves the

addition of a nonspecific growth medium.

5. A method according to any preceding claim wherein the sample is a urine sample.

6. A method according any preceding claim wherein the candidate microorganisms are microorganisms known to cause urinary tract infections.

7. A method according to Claim 6 wherein the candidate microorganisms comprise one or more of E. coli, Proteus mirabilis, K. aerogenes and K cloacae.

8. A method according to Claim 6 wherein the candidate microorganisms comprise one or more of E. coli and Proteus mirabilis.

9. A method according to any preceding claim wherein a said chemical factor is a substrate convertible by E.coli into ethanol, which is a marker product in step (c).

10. A method according to Claim 9 wherein said substrate is lactose or arabinose.

11. A method according to any preceding claim wherein a said chemical factor is a substrate convertible by Proteus mirabilis into one or more sulphur compounds which are sufficiently volatile to be detected in the headspace over the sample.

12. A method according to Claim 11 wherein said substrate is methionine.

13. A method according to any preceding claim wherein step (c) employs an array

of sensors having different relative sensitivities to a plurality of marker compounds, and the outputs of the sensors are analysed to determine response patterns

characteristic of particular candidate microorganisms.

14. A method according to Claim 13 wherein said analysis of sensor outputs employs a multivariate data reduction technique.

15. A method according to Claim 13 wherein said analysis of sensor outputs employs principal component analysis, multiple discriminant analysis or artificial neural network analysis.

16. A method according to any preceding claim wherein a marker product is ethanol and this is detected quantitatively using an enzymatic assay.

17. A method according to any preceding claim wherein a marker product is a sulphur compound and this is detected electrochemically.

18. A method according to any preceding claim including parallel antibiotic sensitivity testing, wherein first and second like samples are provided; one or more

antibiotics are added to one of said samples; both of said samples are then submitted to the detection method; and the results are compared, thereby providing an indication of the efficacy of the antibiotic(s).