Method for determining the reaction of a microorganism to its exposure to a chemical compound

Abstract

The invention concerns a method for determining the reaction of at least one bacterium of interest to its exposure to an antibiotic implementing a Raman spectroscopy analysis comprising the following steps: Having a biological sample that could contain said bacteria of interest, Preparing at least two fractions of said sample each comprising one or more living bacteria of interest, Capturing, in each fraction, at least one living bacterium of interest by using a binding partner, Exposing at least one of the fractions to at least one concentration of at least one given antibiotic, the other of the fractions being the control fraction, Submitting the bacterium/bacteria of interest contained in the fractions to an incident light and analyzing the resultant light obtained by Raman diffusion by the bacterium/bacteria of interest by Raman spectroscopy in order to obtain as many Raman spectra as bacteria, Treating said spectra in order to obtain a signature of the reaction of the or each bacterium/bacteria of interest to the exposure to said antibiotic and of the control, Comparing the signature obtained accordingly per bacterium of interest to a reference base defined under the same conditions as above, for different bacteria and at least said antibiotic, and Defining a sensitivity clinical profile of said bacterium of interest to said antibiotic.

Claims

1. A method for determining the reaction of at least one bacterium of interest to its exposure to an antibiotic implementing a Raman spectroscopy analysis and comprising the following steps: Having a biological sample that could contain said bacteria of interest, Preparing at least two fractions of said sample each comprising one or more living bacterium/bacteria of interest, Capturing, in each fraction, at least one living bacterium of interest by using a binding partner, Exposing at least one of the fractions to at least one concentration of at least one given antibiotic, the other of the fractions being a control fraction, Submitting the captured bacterium/bacteria of interest contained in the fractions to an incident light and analyzing the resultant light obtained by Raman diffusion by the captured bacterium/bacteria of interest by Raman spectroscopy in order to obtain as many Raman spectra as bacteria, Treating said spectra in order to obtain a signature of the reaction of each bacterium of interest to the exposure of said antibiotic and of the control, Comparing the signature obtained for each bacterium of interest to a reference base defined under the same conditions as above for different bacteria and at least said antibiotic, and Defining a sensitivity clinical profile of said bacterium of interest to said antibiotic.

2. The method according to claim 1, wherein more than two fractions of said sample are prepared, and at least two fractions are exposed respectively to increasing concentrations of said antibiotic.

3. The method according to claim 1, wherein the concentrations of the antibiotic to which each fraction is exposed is within an interval of values including at least one value selected from values characteristic of an antibiotic/species pair comprising the epidemiological cut-off, the clinical breakpoint(s) or concentration panels used in reference methods.

4. The method according to claim 1, wherein the binding partner is directly or indirectly immobilized on a support.

5. The method according to claim 4, wherein the binding partner interacts specifically with the bacterium/bacteria of interest, and wherein the binding partner is selected from proteins, antibodies, antigens, aptamers, phages, and phage proteins.

6. The method according to claim 1, wherein the captured bacteria are marked and sorted.

7. The method according to claim 1, wherein, before submitting the captured bacterium/bacteria of interest to an incident light, bacteria that have not been captured are eliminated.

8. The method according to claim 7, wherein the bacteria that have not been captured are eliminated before or after the step of exposure to the antibiotic.

9. The method according to claim 1, wherein, after the step of exposure of the fractions, said fractions and the control fraction are concentrated and then subjected to the capture step.

10. The method according to claim 1, wherein the antibiotic is in a physiological medium allowing the bacterium/bacteria of interest to remain alive.

11. The method according to claim 1, wherein the exposure to the antibiotic is performed at a temperature of at least 18 C. and at most 40 C.

12. The method according to claim 1, wherein the exposure to the antibiotic is performed for a time of at least 10 minutes and at most 4 hours.

13. The method according to claim 1, wherein, in order to obtain said signature, for each fraction exposed to the antibiotic, the Raman spectrum/spectra of the control is subtracted from the Raman spectra of each fraction.

14. The method according to claim 1, wherein defining the sensitivity clinical profile of said bacterium of interest to said antibiotic comprises determining the Sensitive, Intermediate or Resistant phenotype of the bacterium of interest to said antibiotic.

15. The method according to claim 14, further comprising determining the minimum inhibitory concentration (MIC) of said antibiotic for said bacterium.

16. The method according to claim 14, wherein each fraction comprises at least 2 bacteria of interest in order to obtain at least 2 signals.

17. The method according to claim 15, wherein each fraction comprises at least 2 bacteria of interest in order to obtain at least 2 signals.

18. The method according to claim 2, characterized wherein the concentration of the antibiotic to which each fraction is exposed is within an interval of values including at least one value selected from values characteristic of an antibiotic/species pair comprising the epidemiological cut-off, the clinical breakpoint(s) or concentration panels used in reference methods.

19. The method according to claim 18, wherein the binding partner is directly or indirectly immobilized on a support.

20. The method according to claim 19, wherein the binding partner interacts specifically with the bacterium/bacteria of interest and wherein the binding partner is selected from proteins, antibodies, antigens, aptamers, phages, and phage proteins.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The details and advantages of the invention will emerge from the examples hereinafter, in support to the following figures according to which:

(2) FIG. 1 shows the block diagram of the complete mounting of a system allowing to implement the method of the invention integrating a device of the invention as illustrated in FIG. 2.

(3) FIG. 2 represents the block diagram of the device for conditioning microorganisms belonging to the system illustrated in FIG. 1.

(4) FIG. 3 illustrates placing the samples on a glass slide of the conditioning device illustrated in FIG. 2.

(5) FIG. 4 represents an expression of the reaction of the Escherichia coli strain of the reference ATCC 25922, called EC10, to gentamicin, obtained according to the method of the invention.

(6) FIG. 5 represents an expression of the reaction of the Escherichia coli strain of the reference ATCC 35421, called EC21, to gentamicin, obtained according to the method of the invention, for N=5.

(7) FIG. 6 represents the confusion matrix obtained for the sensitive strain of Escherichia coli of the reference ATCC 25922, called EC10, in the presence of amoxicillin (MIC=6 g/mL) for N=11.

(8) FIG. 7 represents the confusion matrix obtained for the sensitive strain of Escherichia coli of the reference ATCC 35421, called EC21, in the presence of amoxicillin for N=11.

(9) FIG. 8 shows the confusion matrix obtained for the sensitive strain of Escherichia coli of the reference ATCC 35421, called EC21, in the presence of amoxicillin tested with a classifier trained with bacteria exposed to gentamicin for N=11.

(10) FIG. 9 illustrates the signatures obtained for gentamicin concentrations of 0 g/mL, 2 g/mL and 8 g/m.

(11) FIG. 10 represents a simplified block diagram provided in the Example 4.

(12) FIG. 11 represents an expression of the reaction of a bacterial strain exposed to different concentrations of ciprofloxacin, obtained according to the method of the invention.

(13) FIG. 12 represents the confounding matrices obtained for the sensitive strain of Staphylococcus aureus of reference ATCC 25923, called SA44, in the presence of oxacillin (MIC=0.25 g/mL) for N=9 for two experiments, the bacteria being obtained from cultures in a liquid medium.

(14) FIG. 13 represents the confusion matrix obtained for the sensitive strain of Staphylococcus aureus of the reference ATCC 25923, called SA44, in the presence of oxacillin (MIC=0.25 g/mL) for N=9, the bacteria being obtained from cultures in agar medium in Petri dishes.

DETAILED DESCRIPTION

Example 1: Application of the Method of the Invention for Determining the Sensitivity Phenotype of the Escherichia coli Strain of the Reference ATCC 25922 Called EC10 to Gentamicin

(15) The retained conditioning device is constituted of two fluidic chambers and two antibiotic concentrations are tested: c.sub.0: Without antibiotic and Resistance test.

(16) Let c.sub.1 be the concentration of gentamicin: c.sub.1=8 g/mL, which corresponds to the doubling of the concentration 4 g/mL corresponding to the clinical breakpoint as defined by the EUCAST. The objective of this test is to determine whether the bacterium is considered as resistant according to the definitions provided by the EUCAST.

(17) A solution containing the bacteria to be tested is used as a test sample. This solution is obtained by suspending 5.10.sup.7 CFU/mL in order to have a concentration potentially encountered in a clinical sample, for example a urine collection. This bacteria solution is brought into contact with the interface I of the device functionalized by adsorbing polyethylenimine (PEI) (generic capture). After a capture time of 10 minutes allowing the bacteria to come into contact with the functionalization, the interface I is washed with a water solution, this optional step allowing to eliminate the surplus of uncaptured bacteria still in solution. The physiological medium retained in this example is constituted of a poorly enriched mixture of Bouillon TSB-T Trypcase Soy broth (for example of the reference 42100 of bioMrieux) and of PBS 10 (for example obtained from the PBS tablets of the reference A9162, 0100 of the brand AppliChem) in a 1:9 ratio. After dividing this physiological medium into two fractions, an amount of gentamicin (for example of the reference G1397-10ML of Sigma-Aldrich) is added respectively to each of these fractions, allowing to obtain a different concentration of gentamicin c.sub.0 or c.sub.1. The solutions of concentrations c.sub.0 and c.sub.1 produced accordingly are respectively introduced into the chambers C.sub.0 or C.sub.1. The bacteria captured directly from the sample on the surface I are thus exposed, in a suitable medium, to a different antibiotic concentration depending on the chamber in which they are present.

(18) The device is then heated in order to reach a temperature of 37 C. for two hours and then placed in measurement position on the micro-spectrometer. The marking of the captured bacteria is carried out by an automatic procedure based on image analysis, by a conventional procedure for detecting particles, acquired by means of the micro-spectrometer camera and a suitable light source. This marking allows acquiring automatically a series of Raman spectra (S.sub.0 and S.sub.1, respectively) acquired on individual bacteria present in each chamber C.sub.0 and C.sub.1. The number of spectra to be acquired for constituting a dataset depends on the level of requirement on the performances of the tests to be carried out.

(19) As mentioned previously, several methods can be used for treating the data obtained in order to achieve the result. In the present example, a complete method for treating spectra individually comprising a pretreatment step comprising all the phases described above is used in order to maximize the extraction of a signal of interest and the classification.

(20) In this example, a set of at least 2N spectra extracted from the total set M of the acquired spectra is used: N spectra from S.sub.0 and N spectra from S.sub.1. These spectra are drawn without replacement among the available M spectra. The average of N spectra from S.sub.0 is subtracted from each of the N spectra from S.sub.1 and the N spectra from S.sub.0, these two batches of spectra constituting a control test sample and a resistance test sample.

(21) For the classification, a reference database is used in the present example, obtained from similar experiments carried out previously at dates and from different cultures in order to train a classifier obtained using a Support Vector Machine (SVM) with radial kernel. This classifier is trained to recognize two classes, one Without antibiotic effect from spectra from conditions without antibiotic and the other Antibiotic effect from spectra previously acquired under conditions where the concentration is higher than the MIC of the strain(s) used in the reference base. For each test sample constituted of N difference spectra, these difference spectra are tested individually with respect to the classifier and the majority class among the elements of the groups is assigned to the group of N spectra. This majority assignment is based on the good correlation of the results obtained accordingly with the reference methods but can well be modified in order to take into account some other parameters of the tests. For example, a vote to a threshold different from the majority where, as soon as the number of bacteria having no effect exceeds 30%, then a result Without antibiotic effect is conservatively assigned. This threshold may also be adjusted in order to take into account the incubation time: thus for example, if the time of exposure to antibiotics is significantly reduced, or if the tested microorganism has a slower typical doubling time, it is necessary to take into account a lower threshold in order to assign an Antibiotic effect result to this group of spectra. Finally, a more nuanced system where each bacterium is considered in a completely individual way could be adopted. This last embodiment may be advantageous if the method of the present invention is used for research purposes.

(22) In order to illustrate the performances obtained accordingly, the average score obtained for all the results that would be obtained with a combination of 5 spectra per concentration (N=5) out of a total set of total acquired spectra of 294 spectra (M=294) is shown in FIG. 4. This matrix has, in columns, the states Without ATB effect and ATB effect and, in rows, the two concentrations of tested antibiotics. The score indicates the percentage of the test samples that are assigned to a given class by the classifier described above. Thus, it is found that 99% of the samples of N=5 bacteria of the control test sample are classified as Without antibiotic effect and 97.1% of the Sensitivity test samples constituted of bacteria exposed to the test concentration are classified as ATB effect. The strain can therefore be described as sensitive according to the invention with a high reliability on the basis of only a few individual bacteria analyses. The result is consistent with the reference methods [BioMrieux products ETEST (antimicrobial susceptibility testing kit) GM 256 (ref. 412368) and VITEK card N233 (Gram Negative Susceptibility card; ref. 413117)] which give as a result a MIC=1 g/mL, which confirms that the bacterium is not resistant according to the EUCAST, its MIC not being strictly above the threshold defined by this organism.

Example 2: Application of the Method of the Invention for Determining the Sensitivity Phenotype of the Escherichia coli Strain of the Reference ATCC 35421 Called EC21 to Gentamicin

(23) A test identical to that of Example 1 is conducted on another strain of Escherichia coli, the strain ATCC 35421 called EC21 allows confirming the discriminatory nature of the measurement.

(24) The results are shown in FIG. 5.

(25) The reference methods assign a MIC>256 g/mL to this strain which is therefore resistant according to the EUCAST.

(26) The method of the invention confirms this result since in the case where N=5 and M=133, it can be read that 100% of the performed tests do not show a characteristic effect profile of the antibiotic agent.

Example 3: Determination of the Minimum Inhibitory Concentration (MIC) of Two Strains of Escherichia Coli to Amoxicillin

(27) In this example, the aim is to determine the sensitivity phenotype, and to specify a framework of the minimum inhibitory concentration of the Escherichia coli strain of the reference ATCC 25922 called EC10 for an antibiotic, amoxicillin, having a mode of action different from that provided in Examples 1 and 2.

(28) In order to illustrate the discriminating power of this method, the same test performed for a strain of Escherichia coli ATCC 35421 resistant to amoxicillin called EC21 is also shown.

(29) The following concentrations of amoxicillin have been tested.

(30) EC10 sensitive strain (MIC.sub.REF=6 g/mL):

(31) 0 g/mL; 2 g/mL; 4 g/mL; 8 g/mL and 16 g/mL

(32) EC21 resistant strain (MIC.sub.REF=256 g/mL):

(33) 0 g/mL; 4 g/mL and 8 g/mL.

(34) In this example, another embodiment of the provided test corresponding to the described alternative method is illustrated.

(35) A solution containing the bacteria to be tested, EC10 or EC21, is used as test sample. This solution is obtained by suspending 5 10.sup.7 CFU/mL in water in order to have a concentration potentially encountered in a clinical sample, for example a urine collection. This bacteria solution is distributed in 5 filter tubes (for example MICROCON (Centrifugal filters) YM100 from Millipore) at a rate of 150 L per tube. A sufficient amount is added to each of these tubes for the final 250 L of physiological medium allowing the growth constituted in this example of a mixture of PBS of final concentration of 1 (obtained from the PBS tablets of the reference A9162,0100 of the brand AppliChem), of a nutrient medium TSB 0.1 (for example obtained from Bouillon TSB-T Trypcase Soy broth of the reference 42100 of bioMrieux) and of an amount of amoxicillin allowing to achieve respective final concentrations c.sub.0 to c.sub.4 of amoxicillin (for example the reference A8523-10ML of Sigma-Aldrich) as follows: c.sub.0=0 g/mL c.sub.1=2 g/mL c.sub.2=4 g/mL c.sub.3=8 g/mL c.sub.4=16 g/mL

(36) The 5 tubes obtained accordingly are incubated for 2 hours at 37 C. with stirring. Centrifugation at 1200 g for 8 minutes using a centrifuge adapted to the used containers (for example the model 8415C of the brand Eppendorf) then allows retrieving a bacterial pellet on the filter portion of each tube and eliminating the medium. The bacterial pellets are respectively resuspended in water in order to carry out a washing before being again pelletized by centrifugation (1200 g for 10 minutes) always on the filter portion of the tube. These pellets are distributed on a glass slide of the Marienfield type constituting the interface I (not functionalized in this configuration) by means of a swab in corresponding chambers noted C.sub.0 to C.sub.4. In this configuration, the chambers are not necessarily isolated from a physical point of view since no exchange is possible between the different conditions. It is thus possible to use a glass slide whose virtual compartments are clearly identified for each concentration as shown in FIG. 3. The virtual compartments are defined by delimitations materialized in the present example by a labelling previously made on the slide on the opposite side to that where the bacteria are deposited. The glass slide is then deposited on a geneframe constituting the seal J in the conditioning device described above.

(37) The marking of the captured bacteria is performed by a manual procedure in this example based on the visual analysis by the operator of the image acquired by means of the camera of the micro-spectrometer and a light source adapted by an experimenter. This marking allows acquiring a series of at least N Raman spectra acquired on individual bacteria present respectively in each chamber C.sub.0 to C.sub.4. The number of spectra to acquire for constituting a dataset depends on the level of requirement on the performances of tests to be carried out.

(38) The data treatment mode proposed here is identical to that of Examples 1 and 2: a first pretreatment step followed by a step of classifying the acquired spectra using a previously trained classifier. In the examples provided below, the classifier is trained with a reference base containing Without antibiotic effect spectra previously acquired in a condition without amoxicillin antibiotic (0 g/mL) of EC10 bacteria and Antibiotic effect spectra of EC10 bacteria in the presence of 8 g/mL of amoxicillin. The results obtained are shown in the confusion matrix provided in FIG. 6. As previously, this matrix allows demonstrating the robustness of the method by giving the results for a large number of tests.

(39) A transition in the assignment of the spectra groups of the Without ATB effect category to the Antibiotic effect category is observed between the concentrations of 4 g/mL and 8 g/mL. A MIC comprised between 4 g/mL and 8 g/mL can therefore be assigned to this strain according to the assays. This variability to a dilution factor is very frequent in this type of assay, the EUCAST indicates, for example, ranges of MIC variations of [2-8]g/mL for this strain ATCC 25922 during quality controls of the MIC disc diffusion tests, and is therefore consistent with expected results. The result established by the reference methods [BioMrieux products ETEST (antimicrobial susceptibility testing kit) AM 256 (ref. 412253) and VITEK card N233 (Gram Negative Susceptibility card; ref. 413117)] is 6 g/mL for this strain, which is also consistent with this result.

(40) The same type of experiment carried out on the strain EC21 resistant to amoxicillin gives the results shown in FIG. 7. No transition is observed and the vast majority of the measured groups is assigned to the without ATB effect category. A MIC>8 g/mL can be assigned to this strain with this test, which is also consistent with the results obtained by the reference method.

(41) As shown in FIG. 8, identical results are obtained by training the classifier on a reference base containing again spectra of bacteria not exposed to antibiotics in order to recognize the Without ATB effect class and Antibiotic effect spectra of bacteria exposed to a concentration greater than the MIC of another antibiotic molecule belonging to a different family, for example the gentamicin of the previous example.

(42) Similar results to those exposed previously are found. This example proves that it is possible to perform the search for the antibiotic effect of an unknown substance on the bacterial strain tested in this manner and could therefore be applied to the molecule screening.

Example 4: Determination of the Effect of an Unknown Substance on a Bacterial Strain

(43) In this example, the aim is to determine the sensitivity phenotype, and to precise a framework for the minimum inhibitory concentration, of a bacterial strain, for example, Escherichia coli strains of the reference ATCC 25922 called EC10, to a substance considered as unknown.

(44) For the needs of the test, a known antibiotic molecule but not belonging to the previously used antibiotic families, is used: ciprofloxacin from the fluoroquinolones family.

(45) The embodiment of the previous example is used for this example.

(46) Only the results obtained for the first 4 concentrations will be explicitly illustrated because an antibiotic effect is rapidly detected for this molecule. A series of N Raman spectra is acquired in each of the chambers from C.sub.0 to C.sub.3. The used concentrations c.sub.0 to c.sub.3 are as follows: c.sub.0=0 g/mL c.sub.1=0.005 g/mL c.sub.2=0.015 g/mL c.sub.3=0.064 g/mL

(47) As previously, the steps carried out for performing the pretreatment of the spectra are as follows: the removal of saturated spectra the removal of cosmic rays the realignment the extraction of the specific bacterial signal the removal of deviants the region of interest and the signal normalization

(48) In order to perform a test, an average of N spectra acquired for each tested concentration is made and an average of N spectra of the concentration c.sub.0 is subtracted from the result. For the concentration c.sub.0, N spectra different from the N spectra used for the subtraction of the reference state are selected. This operation aims at overcoming all variations that are not correlated with the exposure to the antibiotic, under the measurement conditions. A series of 4 test spectra representative of each concentration is thus obtained.

(49) In this example, an unsupervised classification method is used based on the research and the use of at least one spectral signature characteristic of an antibiotic effect. In order to identify this signature, a previously acquired dataset is used for gentamicin and EC10 strain, whose MIC (1 g/ml) is known for this antibiotic. The dataset is used constituted by the N pretreated spectra acquired on bacteria exposed to one of the concentrations (or more) higher than the MIC in order to extract a characteristic effect. An average of the set of N (or n*N) spectra of the dataset is made and the average of N spectra acquired in the absence of antibiotic is subtracted from the result. This result is described as reference signature. It is this reference signature that is retained to describe the data acquired by exposing bacteria to ciprofloxacin, which is considered as unknown in this case.

(50) The signature set constructed accordingly is shown in FIG. 9. The two signatures obtained for the concentrations c.sub.1 and c.sub.3 are similar: the same peaks are modified but the intensity is here correlated to the concentration. It should be noted that this difference of intensity could be used to quantify the impact of a given concentration but only in some strain/antibiotic configurations.

(51) The signature extracted from the concentration of 8 g/mL of gentamicin will be used to analyze the 3 sets of test spectra. In order to do this, the proximity of each test spectrum to the selected signature is evaluated. The evaluation of the distance between the tested spectrum and the signature will be made in this example using a simple Euclidean distance in the space of the spectra but several other distances allow evaluating this proximity (Mahalanobis, L1 . . . ). A threshold is defined empirically with respect to other identical reference experiments carried out previously, the choice of this threshold can be optimized by conventional methods (ROC . . . ) according to the required level of result which can significantly differ between the applications (IVD diagnosis, pharmaceutical screening . . . ). The obtained distance measurement is then compared with this threshold value in order to define the proximity between the retained signature and the difference spectra acquired in the presence of the different concentrations of the molecule considered as unknown. If the distance is below the threshold, the test is positive and an effect of the antibiotic molecule has been detected. If the distance is above this threshold then no effect is detected.

(52) In an advantageous embodiment, it is possible to establish several increasingly strict detection thresholds. In this example, two thresholds are used according to this principle: the first, less strict, allows detecting a significant variation of the tested spectra while the second, stricter, allows describing a great proximity of the modification to the signature. A test of the distance of the tested spectrum to the signature is therefore performed relative to the first threshold. If the test is not passed, then no effect is detected. If the test is passed, the measured distance is below the first threshold, then an effect is detected. A second test is then performed by using the second threshold and if this test is passed, the Antibiotic effect result is assigned to the complete test. If this second test is not passed then the Other effect result is assigned to the complete test. This configuration allows detecting easily spectral modifications reaching a given concentration but not having sufficient similarity with the reference signature. This configuration therefore simply allows overcoming a portion of the possible hazards occurring during a test (strong inhomogeneity of the capture surface, presence of parasitic particles . . . ). A simplified diagram is shown in FIG. 10.

(53) Another way of performing an equivalent test would be to perform a test by using directly the Euclidean standard or other standard of the tested spectrum and to compare it with a significance threshold chosen threshold in order not to take into account the conventional variations related to the measurement mode (sensor noise level, biological variability . . . ) and then to perform the test at a single strict threshold. If the standard exceeds a certain threshold then the spectrum is significantly different from a near-zero difference spectrum if there were no changes in the concentration and then can be compared strictly to the reference signature, for example by measuring its distance to the signature according to the used standard.

(54) The results obtained are shown in FIG. 11. The detection of an effect for concentrations greater than 0.005 g/mL is thus observed. An antibiotic effect of ciprofloxacin can therefore be assigned to the strain tested for concentrations greater than 0.005 g/mL. This test would therefore define a concentration of 0.005 g/mL as a MIC. A new test could possibly be performed by adding the lowest concentrations between 0 and 0.005 g/mL if necessary. The EUCAST data for this strain, indicate a known MIC of 0.008 g/mL and an acceptable variation range of 0.004 g/mL to 0.016 g/mL. A test performed with a suitable etest (bioMrieux) allows measuring a MIC of 0.008 g/mL which confirms that the obtained result is in accordance with the reference methods [BioMrieux products ETEST CI 32 (antimicrobial susceptibility testing kit; ref. 412311) and VITEK card N233 (Gram Negative Susceptibility card; ref. 413117)].

Example 5: Application of the Method of the Invention for Determining the Sensitivity Phenotype of the Strain Staphylococcus Aureus of the Reference ATCC 25923 Called SA44 to Oxacillin

(55) The retained conditioning device is constituted of two fluidic chambers and two antibiotic concentrations are tested: c.sub.0: Without antibiotic and Resistance test.

(56) Let c.sub.1 be the concentration of oxacillin: c.sub.1=8 g/mL which corresponds to multiplying by a factor of 32 the concentration of 0.25 g/mL which corresponds to the clinical breakpoint as defined by the EUCAST. The objective of this test is to determine whether the bacterial strain is considered as resistant according to the definitions provided by the EUCAST.

(57) A solution containing the bacteria to be tested is used as a test sample. This solution is obtained by suspending 5.10.sup.7 CFU/mL in order to have a concentration potentially encountered in a clinical sample, for example a urine collection. The bacteria can come from a culture in a liquid medium or a culture in agar medium in Petri dish. This bacteria solution is brought into contact with the interface I of the device functionalized by adsorbing polyethylenimine (PEI) (generic capture). After a capture time of 10 minutes allowing the bacteria to come into contact with the functionalization, the interface I is washed with a water solution, this optional step allowing to eliminate the surplus of uncaptured bacteria still in solution. The physiological medium retained in this example is constituted of a poor mixture of BHI Brain Heart Infusion Bouillon (for example of the reference 42081 of bioMrieux) and of PBS 10 (in a 1:9 ratio). After dividing this physiological medium into two fractions, an amount of oxacillin (for example of the reference 00353 of TCI Europe) is added respectively to each of these fractions, allowing to obtain a different concentration of gentamicin c.sub.0 or c.sub.1. The solutions of concentrations c.sub.0 and c.sub.1 produced accordingly are respectively introduced into the chambers C.sub.0 or C.sub.1. The Bacteria captured directly from the sample on the surface I are thus exposed, in a suitable medium, to a different antibiotic concentration depending on the chamber in which they are present.

(58) The device is then heated in order to reach a temperature of 37 C. for two hours and then placed in measurement position on the micro-spectrometer. The marking of the captured bacteria is carried out by an automatic procedure based on image analysis, by a conventional procedure for detecting particles, acquired by means of the micro-spectrometer camera and a suitable light source. This marking allows acquiring automatically a series of Raman spectra (S.sub.0 and S.sub.1, respectively) acquired on individual bacteria present in each chamber C.sub.0 and C.sub.1. The number of spectra to be acquired for constituting a dataset depends on the level of requirement on the performances of the tests to be carried out.

(59) As mentioned previously, several methods can be used for treating the obtained data in order to achieve the result. In the present example, a complete method for treating spectra at the individual level is used, comprising a pretreatment step comprising all the phases described above in order to maximize the extraction of a signal of interest and the classification.

(60) In this example, a set of at least 2N spectra extracted from the total set M of the acquired spectra is used: N spectra from S.sub.0 and N spectra from S.sub.1. These spectra are drawn without replacement among the available M spectra. The average of the N spectra from S.sub.0 is subtracted from each of the N spectra from S.sub.1 and N spectra from S.sub.0, these two batches of spectra constituting a control test sample and a resistance test sample.

(61) For the classification, a reference database is used in the present example, obtained from similar experiments performed previously at dates and from different cultures in order to train a classifier obtained using a Support Vector Machine (SVM) with radial kernel. This classifier is trained to recognize two classes, one without antibiotic effect from spectra from conditions without antibiotic and the other Antibiotic effect from spectra previously acquired under conditions where the concentration is higher than the MIC of the strain(s) used in the reference base. For each test sample constituted of N difference spectra, these difference spectra are tested individually with respect to the classifier and the majority class for each of the elements of the groups is assigned to the group. This majority assignment is based on the good correlation of the results obtained accordingly with the reference methods but can well be modified in order to take into account some other parameters of the tests. For example, a vote to a threshold different from the majority where as soon as the number of bacteria having no effect exceeds 30%, then a result Without antibiotic effect is assigned conservatively. This threshold may also be adjusted in order to take into account the incubation time: thus for example, if the time of exposure to antibiotics is significantly reduced, or if the tested bacterium has a slower typical doubling time, it is necessary to take into account a lower threshold in order to assign an Antibiotic effect result to this group of spectra. Finally, a more nuanced system might be adopted where each bacterium is considered in a completely individual way. This last embodiment may be advantageous if the method of the present invention is used for research purposes.

(62) In order to illustrate the performances obtained accordingly, the average score obtained for all the results that would be obtained with a combination of 9 spectra per concentration (N=9) out of a total set of total acquired spectra, 337 spectra (M=337), is shown in FIGS. 12 and 13. These matrices show in columns the states Without ATB effect and ATB effect and in rows the two tested antibiotic concentrations. The score indicates the percentage of the test samples that are assigned to a class given by the classifier described above. Thus, in FIG. 12, it is found that 100% of the samples of N=9 bacteria of the control test sample are classified as Without antibiotic effect and that a number greater than 97% of the sensitivity test samples consisting of bacteria exposed to the test concentration are classified as ATB effect. The strain can therefore be described as sensitive according to the invention with a high reliability on the basis of only a few individual bacteria analyses. The results observed in these figures are consistent with the reference methods [BioMrieux products ETEST OX 256 (antimicrobial susceptibility testing kit; ref. 412432) or VITEK card P631 (Gram Positive Susceptibility card; ref. 414961)] which give as a result a MIC=0.25 g/mL, which confirms that the bacterial strain is not resistant according to the EUCAST, its MIC not being strictly above the threshold defined by this organism.