Genetic Resistance Testing The invention relates to a method of determining an antibi ^¬ otic resistance profile for a bacterial microorganism and to a method of determining the resistance of a bacterial micro ^¬ organism to an antibiotic drug. Antibiotic resistance is a form of drug resistance whereby a sub-population of a microorganism, e.g. a strain of a bacterial species, can survive and multiply despite exposure to an antibiotic drug. It is a serious and health concern for the individual patient as well as a major public health issue. Timely treatment of a bacterial infection requires the analy ^¬ sis of clinical isolates obtained from patients with regard to antibiotic resistance, in order to select an efficacious therapy . Antibacterial drug resistance (ADR) represents a major health burden. According to the World Health Organization's antimicrobial resistance global report on surveillance, ADR leads to 25,000 deaths per year in Europe and 23,000 deaths per year in the US. In Europe, 2.5 million extra hospital days lead to societal cost of 1.5 billion euro. In the US, the di ^¬ rect cost of 2 million illnesses leads to 20 billion dollar direct cost. The overall cost is estimated to be substantial ^¬ ly higher, reducing the gross domestic product (GDP) by up to 1.6% .

Currently, resistance / susceptibility testing is carried out by obtaining a culture of the suspicious bacteria, subjecting it to different antibiotic drug protocols and determining in which cases bacteria do not grow in the presence of a certain substance. In this case the bacteria are not resistant (i.e. susceptible to the antibiotic drug) and the therapy can be administered to the respective patients. The document

US7335485 describes a method of determining the antibiotic susceptibility of a microorganism, wherein the organism is cultured in the presence of an antibiotic drug to be tested. More recently, sensitive technologies as Mass Spectrometry are applied to determine resistance, but this still requires culturing of the microorganism to be tested in the presence of an antibiotic drug to be tested. Further, in all these techniques, each microorganism to be tested has to be tested against individual antibiotic drugs or drug combinations, re ^¬ quiring extensive, time-consuming and cumbersome tests.

It is known that drug resistance can be associated with ge ^¬ netic polymorphisms. This holds for viruses, where resistance testing is established clinical practice (e.g. HIV genotyp- ing) . More recently, it has been shown that resistance has also genetic causes in bacteria and even higher organisms, such as humans where tumors resistance against certain cyto ^¬ static agents can be linked to genomic mutations.

Wozniak et al . (BMC Genomics 2012, 13 (Suppl 7):S23) disclose genetic determinants of drug resistance in Staphylococcus aureus based on genotype and phenotype data. Stoesser et al . disclose prediction of antimicrobial susceptibilities for Escherichia coli and Klebsiella pneumoniae isolates using whole genomic sequence data (J Antimicrob Chemother 2013; 68: 2234-2244) .

Escherichia coli (E. coli) is a Gram-negative, facultative anaerobic, rod-shaped bacterium generally found e.g. in the lower gastro-intestinal tract of mammals. While many species of the Escherichia genus are harmless, some strains of some species are pathogenic in humans causing urinary tract infec ^¬ tions, gastrointestinal disease, as well as a wide range of other pathologic conditions. E.coli is responsible for the majority of these pathologic conditions.

There remains a need for quick and efficient antibiotic re ^¬ sistance testing. This need is addressed by the present invention.

Summary of the Invention

The inventors performed extensive studies on the genome of E. coli bacteria resistant to antibiotic drugs and found remark ^¬ able differences to wild type E. coli. Based on this infor ^¬ mation, it is now possible to provide a detailed analysis on the resistance pattern of E. coli strains based on individual genes or mutations on a nucleotide level. This analysis in ^¬ volves the identification of a resistance against individual antibiotic drugs as well as clusters of them. This allows not only for the determination of a resistance to a single anti- biotic drug, but also to groups of antibiotics such as lactam or quinolone antibiotics, or even to all relevant antibiotic drugs .

Therefore, the present invention will considerably facilitate the selection of an appropriate antibiotic drug for the treatment of an E. coli infection in a patient and thus will largely improve the quality of diagnosis and treatment.

According to a first aspect, the present invention is di- rected to a method of determining an antibiotic resistance profile for a bacterial microorganism belonging to the spe ^¬ cies E. coli comprising the steps of

a) providing a sample containing or suspected of containing the bacterial microorganism;

b) determining the presence of a mutation in at least one gene of the bacterial microorganism selected from the group of genes listed in Table 4 ;

wherein the presence of a mutation is indicative of a re ^¬ sistance to an antibiotic drug.

Table 4 is depicted in the following: abgB mhpA ybfD yfcO yncG torZ fadA thiE allA mnmC ybfQ yfdF yneK uidB fdx thiM argl mukB ycbF yfdR yphG ycjX fhuB trpC caiC norW ycbS yfdX zraS ydiU fhuC udp csiE ompA yccE yfhM agaD yejA fhuD uxaA cynX ompC yceH ygbN astE yfbL fmt ybiB dadX parC ycgB ygcQ chbG yfiK gudP ybiU elaD pgaA ycgK ygeK eutE ygcR helD ydfI fcl potB yciQ ygeO eutQ ygcU hrpB ydgA fhuA puuC ydbD ygiD fIgL ygfZ ilvA yecA flhA puuE yddV yhaC gcvP ygiF kdpD yehT flu rem ydeK yhal gspO ygjM ldcA yfcN frwC rhsC ydjO yhdP gudD yjju lplA yheN gyrA rhsD yeaU yhgE hemF yj jW menB yhgF hofB Rz yeaX yhiJ kdpE yohG metH yhhQ htrL stfR yeeJ yjbl ldrA ypdB pbpC yhjE hybA tilS yefM yjcF livG yqjA purH yj jG hyfB valS yegE yjfF murB yrfB purK ynfA hyfG xseA yegl yjfZ murP ytfG purL

hyfI yacH yehB yjgL nepl aspS queF

hypF yafE yehl yjgN pphB birA rhaA

ilvY yafT yehM yjhS ptrB cysD rhaB

IsrC yagR yeil yj j J rhaD dapB rplO

IsrF yaiO yfaL ymdC speC dxs srlD

menE ybbB yfaW ynbB tiaE eutA thiC

The presence or absence of a mutation in these genes is test ^¬ ed in relation to the reference strain E. coli K12 substrain DH10B (see also more detailed information in the following and in Example 1) . In a preferred embodiment, step b) com ^¬ prises determining the presence of a mutation in at least two or more genes selected from the group of Table 4, and wherein the presence of a mutation in at least two genes is indica- tive of a resistance to an antibiotic drug. Instead of testing only single genes or mutants, a combina ^¬ tion of several variant positions can improve the prediction accuracy and further reduce false positive findings that are influenced by other factors. Therefore, it is in particular preferred to determine the presence of a mutation in 2, 3, 4, 5, 6, 7, 8 or 9 (or more) genes selected from Table 4.

In a further preferred embodiment, the present method com ^¬ prises in step b) determining the presence of a mutation in at least one gene selected from the group of genes listed in Table 5, and wherein the presence of a mutation in said at least one gene is indicative of a resistance to an antibiotic drug .

The genes according to Table 5 have never been described be ^¬ fore in the context of antibiotic resistance of E. coli bac ^¬ teria. They may be used for the determination of an antibi ^¬ otic drug resistance of E. coli alone or in combination with other genes disclosed herein. abgB yegl ymdC ycjX ldcA yhjE

frwC yehM ynbB ydiU lplA Yj jG

hofB yeil yncG yfbL menB

htrL yfaW yneK yfiK metH

hybA yfcO yphG ygcR pbpC

hyfB yfdF zraS ygcU purH

hyfI yfdR agaD ygfZ purL

IsrF yfdX chbG ygiF queF

potB ygbN eutE ygjM rhaB

puuC ygcQ eutQ yjju rplO

yafT ygeK fIgL yj jW srlD

yagR ygeO gcvP yohG thiC

yaiO ygiD gspO yqjA thiE

ybbB yhaC gudD yrfB thiM

ybfD yhgE hemF ytfG uxaA

ybfQ yhiJ ldrA aspS ybiB

ycbF yjbl livG cysD ybiU ycbS yjcF murP eutA ydfI

ycgB yjfF nepl fadA ydgA

ycgK yjfZ pphB fdx yecA

yciQ yjgL ptrB fhuC yehT

yddV yjgN rhaD gudP yfcN

ydjO yjhS tiaE helD yheN

yeaX yj j J torZ hrpB yhgF

yeeJ ynfA uidB kdpD yhhQ

For E.coli 86 ultra highly significant pairs of genetic posi ^¬ tions and drug resistance (table 2) were identified. The 86 combinations correspond to 35 genetic positions, since the sites are usually significant for more than one single drug. Most importantly, the respective sites are located in 9 genes: hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, yjjJ. These genes thus appear to be critical for antibiotic re- sistance/susceptibility . The identified mutations all lead to amino acid alterations, either to an exchange of amino acid at the respective position or the creation of a new stop- codon. For more detailed information, it is referred to Exam ^¬ ple 1, below.

More generally, the present invention in a further aspect re ^¬ lates to a method of determining the resistance or suscepti ^¬ bility of a bacterial microorganism belonging to the species E. coli to an antibiotic drug comprising:

- providing a sample containing or suspected of containing the bacterial microorganism belonging to the species E. coli ;

- determining from said sample a nucleic acid sequence in- formation of at least one gene selected from the group of hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, and yjjJ; and

- based on the determination of said genetic information determining the resistance or susceptibility to the an- tibiotic drug. In a further embodiment of the invention, the presence of a mutation in at least one gene selected from the group of hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, and yj j J is determined. Thus, the presence of a mutation in at least one or 2, 3, 4, 5, 6, 7, 8 or 9 of these genes can be analyzed.

In a further embodiment, the presence of a mutation in at least one gene selected from the group of the following table 6 is determined. More preferably, the exact amino acid ex- change indicated in Table 6 is determined.

Gene Amino Acid Gene Amino Acid

Name Exchange

Name Exchange

aspS D382E purK N137D

birA Q113H purL D615E

cysD D232N queF K126E

dapB N87K rhaA S406N

dxs A541T rhaB T407A

eutA A210V rplO K39N

fadA V387I srlD M54T

fdx S66T thiC H193R

fhuB G448V thiE A121E

fhuC A122V thiE R43Q

fhuD D76E thiM A122T

fmt V30I trpC L378F

gudP A448V udp I147M

gyrA D87N; D87Y uxaA E236A

helD E671D ybiB G35S

hrpB A413T ybiU M419I

hrpB V240A ydfI A146V

ilvA D401E ydgA F416L

kdpD E376D yecA I195V

ldcA R167Q yehT A106V

lplA A279T yfcN 139V

menB T31A yheN Q49H

metH E1124;E1124D yhgF E737D

mukB S1015N yhhQ R138H parC S80I yhjE I323V

parC S80R yj jc A57V

pbpC H37Q ynfA T84S

purH T366I

Surprisingly, the inventors found that an overlap of muta ^¬ tions in functionally similar proteins of E. coli and K.

pneumoniae exists. Interestingly, when considering the pro- teins that were associated significantly with at least one drug, the inventors found an overlap of 1,746 proteins (same official name and more than 80 percent positives in BLAST in pairwise comparison) that are affected in E. coli as well as in K. pneumoniae. Extending the analysis to the exact AA ex- changes in these proteins, the inventors still detected an overlap of 55 mutated positions that are equal in both organ ^¬ isms. Therefore, the above genes might form a valuable basis for the determination of the antibiotic resistance pattern in both, E. coli and K. pneumonia microorganisms.

According to an optional aspect of the invention, the nucleic acid sequence information can be the determination of the presence of a single nucleotide at a single position in at least one gene.

Thus the invention comprises a method wherein the presence of a single nucleotide polymorphism or mutation at a single nu ^¬ cleotide position is detected. For example, this can be done in at least one gene selected from the group of hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, and yjjJ. Therefore, according to an optional aspect of the invention, the mutation is a mutation which is selected from the group of mutations listed in table 2 (see below in Example 1) . The present invention thus also includes a method of determining an antibiotic resistance profile for a bacte ^¬ rial microorganism belonging to the species E. coli compris ^¬ ing the steps of a) providing a sample containing or suspected of containing the bacterial microorganism;

b) determining the presence of a mutation in at least one position as identified in Table 2 ;

wherein the presence of a mutation is indicative of a re ^¬ sistance to an antibiotic drug.

The determination can be made based on 1, 2, 3, 4, 5, 6, 7, and up to the 35 genetic positions identified in Table 2.

Generally, the method according to the present invention in ^¬ volves determining the resistance of E. coli to one or more antibiotic drugs. These drugs include, but are not restricted to antibiotic drugs selected from the group consisting of am- picillin sulbactam (A.S.), ampicillin (AM), amoxicillin clavulanate (AUG) , aztreonam (AZT) , ceftriaxone (CAX) , ceftazidime (CAZ) , cefotaxime (CFT) , cefepime (CPM) , ciprof ^¬ loxacin (CP) , ertapenem (ETP) , levofloxacin (LVX) , cefuroxime (CRM) , piperazillin tazobactam (P/T) , trimethoprim

sulfamethoxazole (T/S), tobramycin (TO), gentamicin (GM) , cefazolin (CFZ) , cephalotin (CF) , imipenem (IMP), meropenem (MER) and tetracycline (TE) . See also Table 1.

The inventors have surprisingly found that mutations in cer- tain genes are indicative not only for a resistance to one single antibiotic drug, but to groups containing several drugs .

For example, it turned out that in case of the group of lac- tarn antibiotics, the presence of a mutation in the following genes: chbG, eutQ, flgL, gudD, gyrA, ldrA, menE, murB, murP, nepl, parC, pphB, ptrB, rhaD, ydiU, yegE, yegl, yfbL, yfiK, ygcR, ygiF, ygjM, yohG, and/or yrfB should be determined and is indicative for the presence of a resistance against anti- biotics of this group.

The group of lactam antibiotics preferably comprises A.S., AM, AUG, AZT, CFZ, CPE, CFT, CAZ, CAX, CRM, CF, CP, IMP, MER, ETP and/or P/T. The p-value threshold for these identified genes is ≤ 10 ^~45 .

It is within the scope of the present invention that the above determination is done based on a single gene or 2, 3, 4 etc. genes of this group, however, it is preferred to deter ^¬ mine a mutation in all of these genes in relation to the ref ^¬ erence strain K12 substrain DH10B (see also below for further information) .

In a further embodiment, the antibiotic drug is selected from quinolone or aminoglycoside antibiotics and the presence of a mutation in the following genes is determined: agaD, chbG, eutE, eutQ, gcvP, gspO, gyrA, livG, menE, nepl, parC, speC, tiaE, torZ, uidB, yegE, yegl, yejA, ygcU, ygfZ, ygiF, ygjM, yjjU, yjj , ymdC, ypdB, yqjA, and/or ytfG.

The quinolone and aminoglycoside antibiotics preferably are selected from CP, LVX, GM and TO.

Surprisingly, the relevant genes completely overlapped re ^¬ garding a resistance to quinolone and aminoglycoside antibi ^¬ otics; the p-value threshold for these genes is < 10 ^"53 . Also here, it is within the scope of the present invention that the determination is done based on a single gene or in 2, 3, 4 or more genes of this group only, however, it is preferred to determine a mutation in all of these genes in relation to the reference strain K12 substrain DH10B. In a further embodiment, the antibiotic drug is selected from tetracycline and the presence of a mutation in at least one or more of the following genes is determined: astE, chbG, eutQ, flgL, gudD, gyrA, hemF, hypF, kdpE, ldrA, menE, murB, murP, nepl, ompC, parC, pphB, ptrB, and/or rhaD. The p-value threshold is < 10 ^"47 .

In a still further embodiment, the antibiotic drug is select ^¬ ed from trimethoprim sulfmethoxazol and the presence of a mu- tation in at least one or more of the following genes is de ^¬ termined: astE, chbG, eutQ, flgL, gudD, gyrA, ldrA, menE, murB, nepl, parC, ycjX, ydiU, yegE, yfiK, ygcR, ygiF, and/or yrfB. The p-value threshold is ≤ 10 ^~48 .

In a preferred embodiment, the method of the present inven ^¬ tion comprises determining a mutation, wherein the mutation is selected from the group of mutations listed in Table 7. Table 7 is depicted in the following:

Genome Therapy Ref Alt AA Alt AA Gene Exchange Pos

37032 P/T C T V I caiC V270I

206427 AM C A, G T R, K yafE T133R; T133K

319290 P/T A T D E yaiO D36E

1181357 AM C T T M yceH T178M

1368519 P/T G T AA S - A137S

1516808 P/T G A, T AA T, S stfR A114T; A114S

1517573 P/T G C E Q stfR E369Q

1567286 P/T G A AA T ynbB A148T

1615473 AZT, CAX A C, T E V, A yncG E203V;E203A

1684413 AM C T M I ydeK M441I

1974644 A/S T A, C C R, S yeaX C69R;C69S

2052365 ETP A T, C I Stop flhA 1427; I427M codon,

M

2178525 P/T C T G D yefM G74D

2216164 P/T C T, G R K, T fcl R20K;R20T

2233638 ETP, P/T G A, T L F, yegE L447F;L447

Stop

codon

2428172 CP, LVX C A, T D N, Y gyrA D87N; D87Y

2428183 A/S, AM, G A S L gyrA S83L

AZT,

CAX,

CAZ, CFT,

CPE,

CRM, GM,

T/S, TO

2463877 AUG A G V A menE V46A

2565236 ETP G T, A A S, T yfdR A156S; A156T

2725302 P/T G A M I xseA M428I

2755319 P/T T C M T csiE M33T

2924554 P/T A T T S norW T27S

3240296 P/T G A, C F Stop hybA F204;F204L codon,

L

4054212 A/S C A, T E Stop ilvY E184;E184D codon,

D

4525576 AM T C I V yjfZ I78V

4553471 ETP C A, T L I, yjfF L20I;L20F

4575887 P/T T C, G L R, P yjgL L207R;L207P

4636902 AM G A A V - A175V

The inventors found that, apart from the above genes indica ^¬ tive of a resistance against antibiotics, also single nucleo- tide polymorphisms (= SNP's) may have a high significance for the presence of a resistance against defined antibiotic drugs. The analysis of these polymorphisms on a nucleotide level may further improve and accelerate the determination of a drug resistance to antibiotics in E. coli.

For example, a resistance of E. coli against the antibiotic drug AM can be determined by the presence of a single nucleo ^¬ tide polymorphism in at least one, for example 1, 2, 3, 4, 5 or 6 of the following nucleotide positions: 2428183, 4525576, 1684413, 4636902, 1181357, 206427. In an embodiment, the antibiotic drug is A/S and an SNP in at least one, for example 1, 2 or 3 of the following nucleotide positions is detected: 2428183, 4054212, 1974644. In a further embodiment, the antibiotic drug is AUG and a mu ^¬ tation in the following nucleotide position is detected:

2463877.

For a resistance to the antibiotic drug AZT, a mutation in at least one of the following nucleotide positions is detected: 2428183, 1615473.

In a still further preferred embodiment, the antibiotic drug is CAX and a mutation in at least one of the following nucle ^¬ otide positions is detected: 2428183, 1615473.

A resistance to the antibiotic drugs CFT, CP, CPE, CRM, GM, LVX, TO, T/S or CAZ can be detected by a mutation in the nu ^¬ cleotide position 2428183. When the antibiotic drug is ETP, a mutation in at least one, for example 1, 2, 3, or 4 of the following nucleotide posi ^¬ tions is detected: 2052365, 2233638, 4553471, 2565236.

In a further embodiment, the antibiotic drug is P/T and a mu- tation in at least one, for example 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, or 14 of the following nucleotide posi ^¬ tions is detected: 2233638, 2216164, 2725302, 1567286,

2755319, 319290, 3240296, 1517573, 2178525, 2924554, 1516808, 37032, 1368519, 4575887.

Preferably, the resistance to the respective antibiotic drug is tested according to the decision diagrams of Figures 5-20.

A decision diagram or "decision tree" is a tree-like graph for prediction tasks, e.g. classification. Given a data set consisting of a number of samples with feature values (any measurements, here SNPs) and class labels (here resistant/not resistant against a certain drug) , a decision tree models the decision process of inferring the sample class label from its feature values.

To build the model a given data set is used (as described above) : Among all features (SNPs) and their possible values (DNA bases A, C, T, and G) the feature value is selected which achieves the optimal sample separation with respect to the given sample labels. Ideally, that would be a SNP whose value for not resistant samples would be different as for the resistant samples. The selected feature value becomes the root of the tree (the first tree node, often drawn at the top) and the samples are split according to that feature, i.e. samples having that feature value and samples with an ^¬ other value. The resulting subsets of samples form new nodes and the feature selection and splitting process is repeated for each of them separately. This procedure stops if a spe ^¬ cific criterion is fulfilled (e.g. no further improvement or maximal tree size is achieved) . The used graphical representation is defined as follows:

The tree root is always drawn at the top. Each node contains following information:

• Its feature and its value (s) drawn below the node, e.g. SNP 2428183 = G.

· Class label: 0 = not resistant, 1 = resistant.

• Class distribution: The proportion of samples contained in that node belonging to class 0 or 1.

• Proportion of samples contained in that node (w.r.t. to sample number used to build the tree) .

· Color: green = 0, blue = 1, the stronger the color the higher the certainty for the chosen class label.

Generally, the model is built on the so-called training set and its prediction power is tested on the so-called test set (to assess the model performance on unseen data) . Both data sets should be independent and have no intersection. However, if the available data set is not large enough to form a suf ^¬ ficient large training and test data sets, we apply a proce- dure called k-fold cross validation (CV) : We divide our data set into k subsets of equal size, then each of the k subsets is used once as test data and the rest as training data. The final tree is built on the whole data set, so the CV is only used to estimate the performance of the final model.

The classification of a new sample works as follows:

• One starts at the tree root: the value of the root at ^¬ tribute in the sample is checked. If the value is equal to the root value then one goes left to the next node. Other ^¬ wise, one goes right.

• The value of the current node attribute in the sample is checked and it is decided again whether to go left or right. And so on.

· The process stops if one is in a leaf node (terminal node, node without outgoing edges) . The sample gets the same label as that leaf node.

According to an optional aspect of the invention, a detected mutation is a mutation leading to an altered amino acid se ^¬ quence in a polypeptide derived from a respective gene in which the detected mutation is located. According to this as ^¬ pect, the detected mutation thus leads to a truncated or ver ^¬ sion of the polypeptide (wherein a new stop codon is created by the mutation) or a mutated version of the polypeptide hav ^¬ ing an amino acid exchange at the respective position.

According to an optional aspect of the invention, determining the nucleic acid sequence information or the presence of a mutation comprises determining a partial sequence or an en ^¬ tire sequence of the at least one gene.

According to an optional aspect of the invention, determining the nucleic acid sequence information or the presence of a mutation comprises determining a partial or entire sequence of the genome of said bacterial microorganism, wherein said partial or entire sequence of the genome comprises at least a partial sequence of said at least one gene. According to an optional aspect of the invention the sample is a patient sample (clinical isolate) .

According to an optional aspect of the invention determining the nucleic acid sequence information or the presence of a mutation comprises a using a next generation sequencing or high throughput sequencing method. According to a preferred further aspect of this aspect of the invention, a partial or entire genome sequence of the bacterial organism is deter- mined by a using a next generation sequencing or high

throughput sequencing method.

According to an optional aspect of the invention, the method of the invention further comprises determining the resistance to 2, 3, 4, 5, or 6 antibiotic drugs.

In a further aspect, the present invention is directed to a diagnostic method of determining an antibiotic resistant E. coli infection in a patient, comprising the steps of:

a) obtaining or providing a sample containing or suspected of containing E. coli from the patient;

b) determining the presence of at least one mutation in at least one gene as described above, wherein the presence of said at least one mutation is indicative of an antibiotic re- sistant E. coli infection in said patient.

In a still further aspect, the present invention is directed to a method of treating a patient suffering from an antibi ^¬ otic resistant E. coli infection in a patient:

a) obtaining or providing a sample containing or suspected of containing E. coli from the patient;

b) determining the presence of at least one mutation in at least one gene as described above, wherein the presence of said at least one mutation is indicative of a resistance to one or more antibiotic drugs;

c) identifying said at least one or more antibiotic drugs; d) selecting one or more antibiotic drugs different from the ones identified in step c) and being suitable for the treatment of an E. coli infection; and

e) treating the patient with said one or more antibiotic drugs .

According to a preferred embodiment, the patient is a verte ^¬ brate, more preferably a mammal and most preferred a human patient .

Regarding the dosage of the antibiotic drug, it is referred to the established principles of pharmacology in human and veterinary medicine. For example, Forth, Henschler, Rummel "Allgemeine und spezielle Pharmakologie und Toxikologie", 9th edition, 2005 might be used as a guideline. Regarding the formulation of a ready-to-use medicament, reference is made to "Remington, The Science and Practice of Pharmacy", 22 ^nd edition, 2013.

Definitions

Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

The term "nucleic acid molecule" refers to a polynucleotide molecule having a defined sequence. It comprises DNA mole ^¬ cules, RNA molecules, nucleotide analog molecules and combi- nations and derivatives thereof, such as DNA molecules or RNA molecules with incorporated nucleotide analogs or cDNA.

The term "nucleic acid sequence information" relates to an information which can be derived from the sequence of a nu- cleic acid molecule, such as the sequence itself or a varia ^¬ tion in the sequence as compared to a reference sequence. The term "mutation" relates to a variation in the sequence as compared to a reference sequence. Such a reference sequence can be a sequence determined in a predominant wild type or ^¬ ganism or a reference organism, e.g. a defined and known bac- terial strain or substrain. A mutation is for example a deletion of one or multiple nucleotides, an insertion of one or multiple nucleotides, or substitution of one or multiple nu ^¬ cleotides, duplication of one or a sequence of multiple nu ^¬ cleotides, translocation one or a sequence of multiple nucle- otides, and, in particular, a single nucleotide polymorphism (SNP) .

In the context of the present invention a "sample" is a sam ^¬ ple which comprises nucleic acid molecule from a bacterial microorganism. Examples for samples are: cells, tissue, body fluids, biopsy specimens, blood, urine, saliva, sputum, plas ^¬ ma, serum, cell culture supernatant, swab sample and others.

New and highly efficient methods of sequencing nucleic acids referred to as next generation sequencing have opened the possibility of large scale genomic analysis. The term "next generation sequencing" or "high throughput sequencing" refers to high-throughput sequencing technologies that parallelize the sequencing process, producing thousands or millions of sequences at once. Examples include Massively Parallel Signa ^¬ ture Sequencing (MPSS) Polony sequencing, 454 pyrosequencing, Illumina (Solexa) sequencing, SOLiD sequencing, Ion semiconductor sequencing, DNA nanoball sequencing, Helioscope (TM) single molecule sequencing, Single Molecule SMRT (TM) sequenc- ing, Single Molecule real time (RNAP) sequencing, Nanopore DNA sequencing.

Before the invention is described in exemplary detail, it is to be understood that this invention is not limited to the particular component parts of the process steps of the meth ^¬ ods described herein as such methods may vary. It is also to be understood that the terminology used herein is for purpos ^¬ es of describing particular embodiments only, and is not in- tended to be limiting. It must be noted that, as used in the specification and the appended claims, the singular forms "a," "an" and "the" include singular and/or plural referents unless the context clearly dictates otherwise. For example, the term "a" as used herein can be understood as one single entity or in the meaning of "one or more" entities. It is al ^¬ so to be understood that plural forms include singular and/or plural referents unless the context clearly dictates other ^¬ wise. It is moreover to be understood that, in case parameter ranges are given which are delimited by numeric values, the ranges are deemed to include these limitation values.

Figure legends Figures 1-4 are related to Example 2, below.

Figure 1 : An exemplary contingency table for the computation of the Fisher' s exact test and the measures accuracy, sensi ^¬ tivity, specificity, positive predictive value (PPV) , and negative predictive value (NPV) . Numbers are given for amino acid exchange S83L (GyrA) and Ciprofloxacin.

Figure 2 : Overview of mean MIC values for Ciprofloxacin for samples having no mutation in GyrA (S83, D87) and ParC (S80), either one mutation in GyrA and not ParC, both mutations in GyrA and not ParC, or all three mutations.

Figure 3: Panel A: bar chart of genes with highest number of significant sites. Panel B. bar chart detailing the genes with highest number of sites correlated to at least 3 drugs. Panel C. Scatter plot showing for each gene the number of significant sites correlated with at least 3 drugs as func ^¬ tion of total number of significant sites in the gene. Col ^¬ ored genes represent those with highest absolute numbers (yfaL, yehL, YjgL)), with higher frequency of resistance correlated to at least 3 drugs (yjgN) and genes with lower sig ^¬ nificant sites in at least 3 drugs (fhuA, yeeJ) . Panel D. Along gene plot for yigN. The significant sites along the ge- netic sequence are presented as dots, the y-axis shows the number of drug classes significant for the respective site. Below, a so called snake plot of the trans-membrane protein is shown, the affected amino acids are indicated.

Figure 4: Panel A: network diagram showing drugs as rectangles and genes with higher or lower coverage if resistance for the respective drug is shown as circles. mmuP, mmuM, yiel, insN-1 correspond to higher read counts in case of re- sistant isolates while green genes correspond to lower cover ^¬ age. Panel B and C: two example along-chromosome plots. Each sample is represented by a line, black lines correspond to non-resistant and gray lines to resistant isolates. Figure 5: Decision diagram for ampicillin.

Figure 6: Decision diagram for ampicillin sulbactam.

Figure 7: Decision diagram for amoxicillin clavulanate.

Figure 8: Decision diagram for aztreonam.

Figure 9: Decision diagram for ceftriaxone. Figure 10: Decision diagram for ceftazidime.

Figure 11: Decision diagram for cefotaxime.

Figure 12: Decision diagram for ciprofloxacin.

Figure 13: Decision diagram for cefepime.

Figure 14: Decision diagram for cefuroxime. Figure 15: Decision diagram for ertapenem.

Figure 16: Decision diagram for gentamycin. Figure 17: Decision diagram for levofloxacin .

Figure 18: Decision diagram for piperazillin tazobactam. Figure 19: Decision diagram for tobramycin.

Figure 20: Decision diagram for trimethoprim sulfmethoxazole .

EXAMPLES Example 1

Here, a unique collection of genes was identified that allow the determination the resistance of a bacterial microorganism to commonly used antibiotic drugs.

A unique cohort of bacterial samples obtained from 150 clini ^¬ cal isolates was sequenced in order to understand the genetic resistance mechanisms by using High Throughput sequencing. In parallel, classical resistance tests were applied using 21 drugs or combinations of drugs (Table 1)

Table 1 : Antibiotic Drugs

Medication Drugbank ID Abbreviation

Amoxicillin DB00766

Clavulanate DB01060 AUG

Ampici11in DB00415 AM

Ampicillin Sulbactam DB00415 A/S

Aztreonam DB00355 AZT

Cefazolin DB01327 CFZ

Cefepime DB01413 CPE

Cefotaxime DB00493 CFT

Ceftazidime DB00438 CAZ

Ceftriaxone DB01212 CAX

Cefuroxime DB01112 CRM

Cephalotin DB00456 CF

Ciprofloxacin DB00537 CP Gentamicin DB00798 GM

Imipenem DB01598 IMP

Levofloxacin DB01137 LVX

Piperacillin DB00319

Tazobactam DB01606 P/T

Tetracycline DB00759 TE

Tobramycin DB00684 TO

Trimethoprim DB00440

Sulfamethoxaxole DB01015 T/S

Meropenem DB00760 MER

Ertapenem DB00303 ETP

E. coli strains to be tested were seeded on agar plates and incubated under growth conditions for 24 hours. Then, colo ^¬ nies were picked and incubated in growth medium in the pres- ence of a given antibiotic drug in dilution series under growth conditions for 16-20 hours. Bacterial growth was de ^¬ termined by observing turbidity.

Next mutations were searched that are highly correlated with the results of the phenotypic resistance test.

For sequencing, samples were prepared using a Nextera library preparation, followed by multiplexed sequencing using the Illuminat HiSeq 2500 system, paired end sequencing. Data were mapped with BWA (Li H. and Durbin R. (2010) Fast and accurate long-read alignment with Burrows-Wheeler Transform. Bioinfor- matics, Epub . [PMID: 20080505] ) and SNP were called using samtools (Li H.*, Handsaker B.*, Wysoker A., Fennell T., Ruan J., Homer N., Marth G., Abecasis G., Durbin R. and 1000 Ge- nome Project Data Processing Subgroup (2009) The Sequence alignment/map (SAM) format and SAMtools. Bioinformatics , 25, 2078-9. [PMID: 19505943] ) .

The reference sequence was obtained from Escherichia coli str. K-12 substr. DH10B:

LOCUS CP000948 4686137 bp DNA circu- lar BCT 05-JUN-2008 DEFINITION Escherichia coli str. K12 substr. DH10B, complete genome .

ACCESSION CP000948

VERSION CP000948.1 GI:169887498

DBLINK BioProject: PRJNA20079

KEYWORDS

SOURCE Escherichia coli str. K-12 substr. DH10B

ORGANISM Escherichia coli str. K-12 substr. DH10B

Bacteria; Proteobacteria; Gammaproteobacteria; Enterobacteriales ; Enterobacteriaceae ; Escherichia.

REFERENCE 1 (bases 1 to 4686137)

AUTHORS Durfee,T., Nelson, R., Baldwin, S., Plunkett,G. Ill, Burland,V., Mau, B . , Petrosino, J. F. , Qin,X., Muzny, D . M . , Ayele,M., Gibbs,R.A., Csorgo,B., Posfai,G., The inventorsin- stock, G.M. and Blattner, F . R .

TITLE The complete genome sequence of Escherichia coli

DH10B: insights into the biology of a laboratory workhorse JOURNAL J. Bacteriol. 190 (7), 2597-2606 (2008)

PUBMED 18245285

REFERENCE 2 (bases 1 to 4686137)

AUTHORS Plunkett,G. III.

TITLE Direct Submission

JOURNAL Submitted (20-FEB-2008) Department of Genetics and Biotechnology,

University of Wisconsin, 425G Henry Mall, Madi ^¬ son, WI 53706, USA

COMMENT DH10B and DH10B-T1R are available from Invitrogen

Corporation

(http://www.invitrogen.com) .

The mutations were matched to the genes and the amino acid changes were calculated. Using different algorithms (SVM, ho ^¬ mology modeling) mutations leading to amino acid changes with likely pathogenicity / resistance were calculated. Known var- iants from the swissprot database were excluded and all vari ^¬ ants in the respective genes selected. As noted above, for E.coli 86 ultra highly significant pairs of genetic positions and drug resistance (table 2) were iden ^¬ tified. The 86 combinations correspond to 35 genetic posi ^¬ tions, since the sites are usually significant for more than one single drug. Most importantly, the respective sites are located in 9 genes: hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, yjjJ. These genes thus appear to be critical for anti ^¬ biotic resistance/susceptibility. The identified mutations all lead to amino acid alterations, either to an exchange of amino acid at the respective position or the creation of a new stop-codon. Thereby, resistance related variants for the following 6 antiobiotic drugs were detected: CP, LVX, TE,

CFZ, CRM, GM.

Table 2: Identified Mutations

Genome gene Alt

Pos Therapy p-value pos Ref Alt AA AA Gene Exchange

2, 6527E-

90064 CP 22 1140 C T W C hofB W380C

5, 2992E-

90064 LVX 20 1140 C T W C hofB W380C

2, 6527E-

471036 CP 22 31 G A E K allA E11K

5, 2992E-

471036 LVX 20 31 G A E K allA E11K

2, 6527E-

1030161 CP 22 685 C T R C mukB R229C

5, 2992E-

1030161 LVX 20 685 C T R C mukB R229C

4, 7722E-

1161719 CP 13 697 A C N H ymdC N233H

8, 8353E-

1161719 LVX 13 697 A C N H ymdC N233H

2, 6527E-

1161764 CP 22 742 C T R C ymdC R248C

5, 2992E-

1161764 LVX 20 742 C T R C ymdC R248C

3, 4979E-

1238314 CP 18 799 A G L V potB L267V

1238314 LVX 7, 91E-17 799 A G L V potB L267V

2, 5459E-

1238314 TE 07 799 A G L V potB L267V 2, 6527E-

1239076 CP 22 37 C T V F potB V13F

5, 2992E-

1239076 LVX 20 37 C T V F potB V13F

8, 9009E-

1266748 CP 16 189 A G H Q ycgK H63Q

1, 2635E-

1266748 LVX 14 189 A G H Q ycgK H63Q

4, 4331E-

1266748 CFZ 10 189 A G H Q ycgK H63Q

6, 1345E-

1266748 TE 10 189 A G H Q ycgK H63Q

6, 0841E-

1266748 CRM 07 189 A G H Q ycgK H63Q

2, 6527E-

1266829 CP 22 108 G A S R ycgK S36R

5, 2992E-

1266829 LVX 20 108 G A S R ycgK S36R

1, 2048E-

1275217 LVX 15 1489 T C I L ycgB I497L

1, 5912E-

1275217 CP 15 1489 T C I L ycgB I497L

1, 8029E-

1275217 TE 07 1489 T C I L ycgB I497L

6, 7512E-

1275217 CFZ 07 1489 T C I L ycgB I497L

2, 6527E-

1275307 CP 22 1399 G A L M ycgB L467M

5, 2992E-

1275307 LVX 20 1399 G A L M ycgB L467M

3, 7422E-

4582262 CP 11 1407 G A D E valS D469E

7, 4232E-

4582262 LVX 10 1407 G A D E valS D469E

7, 6762E-

4582262 CFZ 07 1407 G A D E valS D469E

1, 1193E-

4582280 CP 12 1389 A G D E valS D463E

1, 2386E-

4582280 LVX 12 1389 A G D E valS D463E

1, 6858E-

4582301 LVX 16 1368 T C K N valS K456N

4, 0168E-

4582301 CP 16 1368 T C K N valS K456N

3, 6594E-

4582313 LVX 16 1356 G A D E valS D452E

2, 3782E-

4582313 CP 15 1356 G A D E valS D452E

4582337 CP 3, 3749E- 1332 G A N K valS N444K 11

1, 2081E-

4582337 LVX 10 1332 G A N K valS N444K

3, 6214E-

4582337 CFZ 08 1332 G A N K valS N444K

4, 0453E-

4582352 CP 16 1317 A G Y * valS Y439*

5, 235E-

4582352 LVX 16 1317 A G Y * valS Y439*

3, 2243E-

4582352 CFZ 07 1317 A G Y * valS Y439*

3, 7422E-

4582382 CP 11 1287 C T L F valS L429F

7, 4232E-

4582382 LVX 10 1287 C T L F valS L429F

1, 4858E-

4582430 CP 11 1239 G A Y * valS Y413*

4, 4757E-

4582430 LVX 11 1239 G A Y * valS Y413*

2, 3492E-

4582430 CFZ 07 1239 G A Y * valS Y413*

1, 1423E-

4582838 LVX 14 831 T C K N valS K277N

1, 2258E-

4582838 CP 14 831 T C K N valS K277N

7, 2547E-

4582838 CFZ 08 831 T C K N valS K277N

6, 011E-

4582937 LVX 18 732 A G D E valS D244E

2, 912E-

4582937 CP 17 732 A G D E valS D244E

7, 6578E-

4582937 GM 07 732 A G D E valS D244E

1, 7589E-

4582943 CP 13 726 G A Y * valS Y242*

1, 3467E-

4582943 LVX 11 726 G A Y * valS Y242*

1, 2817E-

4582943 CFZ 09 726 G A Y * valS Y242*

1, 3866E-

4582943 TE 07 726 G A Y * valS Y242*

1, 7363E-

4582987 CP 22 682 G A L M valS L228M

4582987 LVX 3, 2E-21 682 G A L M valS L228M

3, 6032E-

4583141 CP 18 528 A G D E valS D176E

1, 1539E-

4583141 LVX 17 528 A G D E valS D176E

4583141 CFZ 7, 2389E- 528 A G D E valS D176E 08

5, 4173E-

4666362 CP 19 109 G A D N yj j J D37N

2, 3023E-

4666362 LVX 18 109 G A D N yj j J D37N

5, 4173E-

4666405 CP 19 152 C T A V yj j J A51V

2, 3023E-

4666405 LVX 18 152 C T A V yj j J A51V

9, 4295E-

4666461 CP 08 208 A G T A yj j J T70A

8, 4496E-

4666461 LVX 07 208 A G T A yj j J T70A

9, 4295E-

4666768 CP 08 515 A G H R yj j J H172R

8, 4496E-

4666768 LVX 07 515 A G H R yj j J H172R

1, 6956E-

4666804 CP 22 551 A G H R yj j J H184R

7, 7611E-

4666804 LVX 22 551 A G H R yj j J H184R

4, 5083E-

4666804 CFZ 07 551 A G H R yj j J H184R

9, 4295E-

4666885 CP 08 632 A G Y C yj j J Y211C

8, 4496E-

4666885 LVX 07 632 A G Y C yj j J Y211C

9, 4295E-

4667178 CP 08 925 C G Q E yj j J Q309E

8, 4496E-

4667178 LVX 07 925 C G Q E yj j J Q309E

2, 6527E-

4667191 CP 22 938 G A R H yj j J R313H

5, 2992E-

4667191 LVX 20 938 G A R H yj j J R313H

9, 4295E-

4667359 CP 08 1106 T C V A yj j J V369A

8, 4496E-

4667359 LVX 07 1106 T C V A yj j J V369A

2, 6527E-

4667424 CP 22 1171 G A V I yj j J V391I

5, 2992E-

4667424 LVX 20 1171 G A V I yj j J V391I

1, 2838E-

4667568 CP 17 1315 G A A T yj j J A439T

2, 3051E-

4667568 LVX 17 1315 G A A T yj j J A439T Table 2 the columns are designated as follows: Genome Pos: genomic position of the SNP / variant in the E. coli reference genome (see below) ;

Therapy: the therapy to which the mutation is significantly correlated, multiple therapies are in separate rows (if a SNP is correlated to e.g. 4 therapies this leads to 4 single rows) ;

P-value: significance value calculated using fishers exact test ;

Gene pos: position of the mutation in the gene;

Ref: reference base, A, C, T, G;

Alt: Alternative base associated with resistance;

AA: original Amino acid;

Alt A: changed amino acid;

Gene: affected gene;

Exchange: amino acid exchange in standard nomenclature;

P-value was calculated using the fisher exact test based on contingency table with 4 fields: #samples Resistant / wild type; #samples Resistant / mutant; #samples not Resistant / wild type; #samples not Resistant / mutant

In table 3, the identified genes and gene products are listed and identified by Gene ID of the gene and (NCBI) Accession number of the corresponding protein corresponding to

Table 3: Gene name and Identifier

Accession

Gene name Gene ID No.

hofB 6061494 ACB01286.1

allA 6059827 ACB01630.1

mukB 6060547 ACB02124.1

ymdC 6059214 ACB02240.1

potB 6058608 ACB02318.1

ycgK 6058586 ACB02348.1

valS 6060190 ACB05239.1

yj j J 6058313 ACB05313.1

ycgB 6058539 ACB02358.1 The test is based on the distribution of the samples in the 4 fields. Even distribution indicates no significance, while clustering into two fields indicates significance. Using this approach 35 highly significant, novel genetic po ^¬ sitions or mutations in 9 genes (hofB, allA, mukB, ymdC, potB, ycgK, ycgB, valS, yjjJ) were identified which can be used for and allow the determination of resistance to commonly used antibiotic drugs. All the highly significant muta- tions described herein and listed in table 2 are non- conservative mutations leading to an amino acid exchange or a new stop-codon (designated with a symbol in table 2), and thus to an altered protein. It is thus likely that the iden ^¬ tified 9 genes play a significant role in antibiotic re- sistance and are putative targets for developing new drug candidates .

Example 2

In this example, the inventors evaluate genetic susceptibil ^¬ ity of E. coli to 21 different drugs from five drug classes (see below) . Methods: Antimicrobial susceptibility test (AST) for 1,162 clinical E. coli isolates with varying spectra of resistance to 21 FDA-approved drugs was performed and genomes of all isolates were sequenced. Genetic variants were correlated to the AST data.

Results: The inventors report 25,744 sites in the E. coli ge ^¬ nome significantly correlated to drug resistance. Highest significance was reached for the drugs Ciprofloxacin and Levofloxacin with respect to amino acid (AA) exchange S83L in GyrA ficity and sensitivi ^¬ ty: 9 97%, 98%, 93%), a target for quinolones. The second most significant associa ^¬ tion was observed for ParC, a second target of quinolones (AA exchange S80I, ) . Partic ^¬ ularly many AA exchanges significantly associated with re ^¬ sistance to multiple drugs were discovered in YigN. By ana ^¬ lyzing the sequence coverage on the genome level, the inven- tors identified a gene dose dependency of several genes, in ^¬ cluding mmuP and mmuM, encoding a putative S-methylmethionine transporter and a homocysteine S-methyltransferase . Both loci are associated with resistance against a-lactams and quin- olones .

Conclusion: The inventors here present a high-throughput screening and analysis pipeline to investigate antibiotics resistance in E. coli strains. The results demonstrate the potential of genetics-based tests to predict susceptibility against antimicrobial drugs. In addition, novel correlations of gene dose to resistance are reported.

The inventors carried out a systematic evaluation using E. coli. Specifically, the inventors collected 1,162 E. coli samples over 22 years (1991-2013) across over 60 different institutes. For these isolates, the inventors carried out standard AST for 21 FDA-approved drugs and performed Whole Genome Sequencing (WGS) for the same 1,162 isolates to build a database revealing genetic sites for predicting AST from genetic data.

Methods

Bacterial Strains

The inventors selected 1,162 E. coli strains from the micro ^¬ biology strain collection at Siemens Healthcare Diagnostics (West Sacramento, CA) for susceptibility testing and whole genome sequencing. Antimicrobial Susceptibility Testing Panels

Frozen reference AST panels were prepared following Clinical Laboratory Standards Institute (CLSI) recommendations ¹⁶ . The following antimicrobial agents (with yg/ml concentrations shown in parentheses) were included in the panels: Amoxicil- lin/K Clavulanate (0.5/0.25-64/32), Ampicillin (0.25-128), Ampicillin/Sulbactam (0.5/0.25-64/32), Aztreonam (0.25-64), Cefazolin (0.5-32), Cefepime (0.25-64), Cefotaxime (0.25-

128), Ceftazidime (0.25-64), Ceftriaxone (0.25-128), Cefurox- ime (1-64), Cephalothin (1-64), Ciprofloxacin (0.015-8), Ertepenem (0.12-32), Gentamicin (0.12-32), Imipenem (0.25- 32), Levofloxacin (0.25-16), Meropenem (0.12-32),

Piperacillin/Tazobactam (0.25/4-256/4), Tetracycline (0.5-

64), Tobramycin (0.12-32), and Trimethoprim/Sulfamethoxazole (0.25/4.7-32/608). Prior to use with clinical isolates, AST panels were tested with QC strains. AST panels were consid ^¬ ered acceptable for testing with clinical isolates when the QC results met QC ranges described by CLSI16.

Inoculum Preparation

Isolates were cultured on trypticase soy agar with 5% sheep blood (BBL, Cockeysville, Md.) and incubated in ambient air at 35±1 ^° C for 18-24 h. Isolated colonies (4-5 large colonies or 5-10 small colonies) were transferred to a 3 ml Sterile Inoculum Water (Siemens) and emulsified to a final turbidity of a 0.5 McFarland standard. 2 ml of this suspension was add- ed to 25 ml Inoculum Water with Pluronic-F (Siemens) . Using the Inoculator (Siemens) specific for frozen AST panels, 5 μΐ of the cell suspension was transferred to each well of the AST panel. The inoculated AST panels were incubated in ambi ^¬ ent air at 35±1 ^° C for 16-20 h. Panel results were read visu- ally, and minimal inhibitory concentrations (MIC) were deter ^¬ mined .

DNA extraction

Four streaks of each Gram-negative bacterial isolate cultured on trypticase soy agar containing 5% sheep blood and cell suspensions were made in sterile 1.5 ml collection tubes con ^¬ taining 50 μΐ Nuclease-Free Water (AM9930, Life Technolo ^¬ gies) . Bacterial isolate samples were stored at -20 °C until nucleic acid extraction. The Tissue Preparation System (TPS) (096D0382-02_01_B, Siemens) and the VERSANT® Tissue Prepara ^¬ tion Reagents (TPR) kit (10632404B, Siemens) were used to ex ^¬ tract DNA from these bacterial isolates. Prior to extraction, the bacterial isolates were thawed at room temperature and were pelleted at 2000 G for 5 seconds. The DNA extraction protocol DNAext was used for complete total nucleic acid ex ^¬ traction of 48 isolate samples and eluates, 50 μΐ each, in 4 hours. The total nucleic acid eluates were then transferred into 96-Well qPCR Detection Plates (401341, Agilent Technolo ^¬ gies) for RNase A digestion, DNA quantitation, and plate DNA concentration standardization processes. RNase A (AM2271, Life Technologies) which was diluted in nuclease-free water following manufacturer's instructions was added to 50 μΐ of the total nucleic acid eluate for a final working concentra ^¬ tion of 20 μg/ml. Digestion enzyme and eluate mixture were incubated at 37 °C for 30 minutes using Siemens VERSANT® Am ^¬ plification and Detection instrument. DNA from the RNase digested eluate was quantitated using the Quant-iT™ PicoGreen dsDNA Assay (P11496, Life Technologies) following the assay kit instruction, and fluorescence was determined on the Sie ^¬ mens VERSANT® Amplification and Detection instrument. Data analysis was performed using Microsoft® Excel 2007. 25 μΐ of the quantitated DNA eluates were transferred into a new 96- Well PCR plate for plate DNA concentration standardization prior to library preparation. Elution buffer from the TPR kit was used to adjust DNA concentration. The standardized DNA eluate plate was then stored at -80°C until library prepara ^¬ tion .

Next Generation Sequencing

Prior to library preparation, quality control of isolated bacterial DNA was conducted using a Qubit 2.0 Fluorometer (Qubit dsDNA BR Assay Kit, Life Technologies) and an Agilent 2200 TapeStation (Genomic DNA ScreenTape, Agilent Technolo ^¬ gies) . NGS libraries were prepared in 96 well format using NexteraXT DNA Sample Preparation Kit and NexteraXT Index Kit for 96 Indexes (Illumina) according to the manufacturer's protocol. The resulting sequencing libraries were quantified in a qPCR-based approach using the KAPA SYBR FAST qPCR

MasterMix Kit (Peqlab) on a ViiA 7 real time PCR system (Life Technologies) . 96 samples were pooled per lane for paired-end sequencing (2x lOObp) on Illumina Hiseq2000 or Hiseq2500 se ^¬ quencers using TruSeq PE Cluster v3 and TruSeq SBS v3

sequncing chemistry (Illumina). Basic sequencing quality parameters were determined using the FastQC quality control tool for high throughput sequence data (Babraham Bioinformat- ics Institute) .

Data analysis Raw paired-end sequencing data for the 1,162 E. coli samples were mapped against the E. coli DH10B reference

(NC_010473) (see also above in Example 1) with BWA 0.6.1.20 The resulting SAM files were sorted, converted to BAM files, and PCR duplicates were marked using the Picard tools package 1.104 (http://picard.sourceforge.net/). The Genome Analysis Toolkit 3.1.1 (GATK) 21 was used to call SNPs and indels for blocks of 200 E. coli samples (parameters: -ploidy 1 -glm BOTH -stand_call_conf 30 -stand_emit_conf 10) . VCF files were combined into a single file and quality filtering for SNPs was carried out (QD < 2.0 | | FS > 60.0 | | MQ < 40.0) and indels (QD < 2.0 | | FS > 200.0) . Detected variants were anno ^¬ tated with SnpEff22 to predict coding effects. For each anno ^¬ tated position, genotypes of all E. coli samples were consid ^¬ ered. E. coli samples were split into two groups, low re- sistance group (having lower MIC concentration for the considered drug) , and high resistance group (having higher MIC concentrations) with respect to a certain MIC concentration (breakpoint) . To find the best breakpoint all thresholds were evaluated and p-values were computed with Fisher' s exact test relying on a 2x2 contingency table (number of E. coli samples having the reference or variant genotype vs. number of sam ^¬ ples belonging to the low and high resistance group) . The best computed breakpoint was the threshold yielding the low- est p-value for a certain genomic position and drug. For fur ^¬ ther analyses positions with non-synonymous alterations and p-value < 10-9 were considered. Based on the contingency ta ^¬ ble, the accuracy (ACC) , sensitivity (SENS) , specificity (SPEC), and the positive/negative predictive values (PPV/NPV) were calculated (Figure 1) .

Since a potential reason for drug resistance is gene duplica ^¬ tion, gene dose dependency was evaluated. For each sample the genomic coverage for each position was determined using BED Tools. 23 Gene ranges were extracted from the reference as ^¬ sembly NC_010473. gff and the normalized median coverage per gene was calculated. To compare low- and high-resistance iso ^¬ lates the best area under the curve (AUC) value was computed. Groups of at least 20% of all samples having a median cover ^¬ age larger than zero for that gene and containing more than 15 samples per group were considered in order to exclude ar ^¬ tifacts and cases with AUC > 0.75 were further evaluated. Results

The aim of our study was to demonstrate the feasibility of genetic antimicrobial susceptibility tests (GAST) , to verify our method for known resistance mechanisms, and to discover novel mechanisms. The inventors performed culture-based AST for 1,162 E. coli isolates and 21 antimicrobial drugs belong ^¬ ing to 5 different drug classes: a-lactams, fluoroquinolones, aminoglycosides, tetracyclines, and folate synthesis inhibi ^¬ tors. The complete list of drugs is shown in Table 1. For the same 1,162 E. coli isolates, whole genome sequencing using Illumina' s HiSeq2500 instrument was carried out.

Most significant sites in the E. coli genome In order to calculate genome-wide significance scores, the inventors mapped all 1,162 E. coli genomes to the reference strain DH10B. For each genomic position the inventors determined the base for each sample and discovered 973,226 sites that passed the quality filtering and in which at least one sample had a non-reference base. The respective sites were correlated to the AST data for the 21 drugs using Fisher's exact test. Our analysis revealed 25,744 sites where a genet- ic mutation significantly correlated with at least one drug (p-value<10 ^~9 ) and led to a change in the AA sequence, in ^¬ cluding point mutation and small insertions and deletions. The highest significance was reached for AA exchange S83L in

GyrA and the drug Ciprofloxacin (p = 10 ) . Remarkably, GyrA is one of the targets of Ciprofloxacin. For this position, three AA exchanges, S83L, S83W, S83A, are annotated in

UniProt as conferring resistance to quinolones. For this site, only 5 false positive (0.4%) and 18 false negative sam ^¬ ples (1.6%) were discovered while 1,139 samples were identi- fied correctly, corresponding to accuracy, specificity, and sensitivity of 98.0%, 99.4% and 93.8%, respectively (Figure 1) . Similarly, the second most significant site in GyrA, D87N / D87Y revealed just 12 false positives and 10 false nega ^¬ tives, the respective p-value was 10 ^~206 and the accuracy 98.1%. Again, for this site the D87N exchange is annotated as conferring quinolone resistance in UniProt. For the third and fourth most significant sites, located in the second Ciprof ^¬ loxacin target, ParC, (S80I, E84G) , resistance related vari ^¬ ants have also been described. In Figure 2, the inventors present the means and standard deviations of MICs for Ciprof ^¬ loxacin for samples having no variant in GyrA (S83/D87) and ParC (S80), samples having only one mutation either in GyrA S83 or D87 and not ParC, samples having both mutations in GyrA and not ParC, and samples having all three mutations. Interestingly, the mean MIC values increase from below 1.0 for no or single mutants to above 7.8 for double or triple mutants, which shows that a combination of mutations is nec ^¬ essary to reach a higher level of resistance against Ciprof ^¬ loxacin in this case.

Besides the mutations in type II topoisomerase drug targets (GyrA/ParC) , mutations in genes ygiF (A110T, p=10 ^"67 , acc=86%, spec=89.5%, sens=69.9%) and ygj (A68V, p=10 ^"63 , acc=89.9%, spec=94.4%, sens=67.1%) have also a high significance. Compared to the above-described AA exchanges, these two sites demonstrate a substantially decreased sensitivity and posi ^¬ tive predictive values (PPV) . While the PPV for the four AA exchanges in GyrA and ParC was between 94.8% and 98.2%, the PPV of these two exchanges decreases to 59.0% and 70.8%. This means that the likelihood to be resistant given the exchanged AA is almost as high as the likelihood to be susceptible giv ^¬ en the exchanged AA, limiting the probability that the re- spective AA exchanges are causative.

To discover other AA exchanges that are potentially causative for drug resistance, the inventors filtered the list of all 25,744 sites (at least 150 resistant E. coli isolates carry the AA exchange, NPV>50%, PPV>75%) . This filtering revealed 127 candidate sites (see also Table 4) . Besides the already described exchanges in GyrA and ParC, the inventors discov ^¬ ered AA exchanges in YdjO associated with predicted re ^¬ sistance to different a-lactams (V121E, S120C, V118F, I114V, K111E, and D112N) . Likewise, for lactams the inventors report AA exchanges in YcbS (E848Q, E848*), RhsC (R717Q, W492C), YcbQ (T86I), YagR (S274T) and YeaU (N293K). Finally, the in ^¬ ventors discovered AA exchanges related to quinolones, tetra ^¬ cycline, and lactams in YhaL (altogether 23 different sites) . In addition, the inventors computed the most significant non- synonymous AA exchange for each drug (p-value threshold<l 0 ^{~ 9} ) . Of 21 tested drugs, only two (Imipenem, Meropenem) were not found to be associated with an AA exchange with such a low p-value. Interestingly, the S83L mutation in GyrA is the predominant exchange in 15 drugs. For the drugs Ciprofloxacin and Levofloxacin, of which GyrA is a target, the p-values were however much lower than the p-values for this mutation in association with the remaining 13 drugs (>10 ^~62 vs. <10 ^{~ 209} ) . In addition, the inventors observed again a significant decrease in sensitivity and/or PPV in these cases: either the sensitivity or PPV is below 55% for drugs, of which GyrA is not the target, demonstrating that these measures are effec ^¬ tive for separating mutations in true targets from others. Mutations in known drug targets

In 9 cases, the inventors detected mutations associated with drugs in genes that are also encoding the targets for the re- spective drugs. This includes the mutations associated with Ciprofloxacin and Levofloxacin in GyrA (S83L, D87N, D87Y, D678E, E574D) and ParC (S80I, E84G, E84V, E84A, A192V, Q481H, A471G, T718A, Q198H), mutations associated with Cephalotin in AmpC (K40R, I300V, T335I, A210P, Q196H, A236T, R248C), with Sulfamethoxazole in FolC (A319T, R88C, G217S) , with Cefazolin in MrcB (D839E, QQQP815Q, R556C) and PbpC (L357V, V348A, A15T, A217V, Q495L, V768F, A701E, K766R, K766T, T764S, T764A, R602L, E446G, R669H, A202T) and with Ceftazidime in PbpG (A28V) .

Most affected genes and multi-drug resistant sites

Mutations are not uniformly distributed across E. coli genes: for example, yfaL, fhuA, yehl, yjgL, and yeeJ carry over 120 non-synonymous variants per gene (Figure 3A) ; in yfaL, as many as 182 significant exchanges were discovered. In order to discover sites that are relevant for multi-drug re ^¬ sistance, the inventors calculated the number of AA exchanges significant in association with at least 3 drug classes (Fig- ure 3B) and plotted the respective site counts for each gene in Figure 3C. On average, 35% of all significant sites were associated with at least three drugs. While three genes, yfaL, yehl, and yjgL, had the highest number of AA exchanges, yjgN had a substantially increased number of sites associated with multi-drug resistance (53 of 64 sites, 83%) , while yeeJ (15 of 122 sites, 12%) and fhuA (12 of 166 sites, 7%) carry fewer sites relevant for multiple drug classes than expected. In yjgN, the positions significantly associated with multiple drug classes were concentrated in the terminal regions of the gene (Figure 3D) . Coverage analysis

A potential reason for drug resistance is gene duplication or deletion, which can be observed in our dataset by inspecting the read coverage of different genes in the groups of re ^¬ sistant and susceptible isolates. To estimate the difference in coverage the inventors calculated AUC values for the nor ^¬ malized median coverage per gene in the two groups. Altogeth ^¬ er the inventors discovered 23 cases of abnormal differences in gene coverage between resistant and susceptible bacteria resulting in an AUC > 0.75 (Figure 4A) . The inventors report connections for three a-lactams and two quinolones. Central genes are mmuP and mmuM, encoding for a putative S- methylmethionine transporter and a homocysteine S- methyltransferase, respectively, for which the coverage is substantially higher in bacteria resistant to all 5 drugs. In strains resistant to Levofloxacin and Ciprofloxacin, the inner membrane protein Yiel and InsN-1, a regulator of insertion element, were likewise higher abundant. In contrast, genes encoding glucosyltransferases YaiP, YaiO, outer mem ^¬ brane protein NmpC and DNA-binding transcriptional repressor MngR were less covered in strains resistant to these drugs. Figure 4B and 4C show an example coverage plot for the lower abundant covered yaiP and the higher abundant covered mmuP in strains resistant to Ciprofloxacin. Best diagnostic accuracy was reached for Ciprofloxacin and the gene mmuP, with an AUC value of 0.923, demonstrating that this quantitative infor ^¬ mation allows for accurate separation between resistant and susceptible strains.

Discussion

The considerable and ongoing increase of infections caused by multi-drug resistant pathogens presents a major threat for patients especially in hospital settings. The development of new drugs is a long and expensive venture, and stagnated in the last years despite increasing investments in research and development. The announcement by the FDA in September 2012 to form an internal task force for supporting the development of new antimicrobial drugs emphasizes the importance of this topic. Until these drugs become available, the inventors have to learn how to apply the available ones most efficiently. Abundant prescribing of broad-spectrum antibiotics promotes the development of multi-drug resistance, so a more careful selection of drugs is needed. Thus, the inventors need meth ^¬ ods that can quickly stratify patients and provide them with the optimal therapy. Identifying the genetic loci in the in- fectious agent that are predominantly responsible for an ob ^¬ served resistance or susceptibility is a crucial point for this .

Here, the inventors analyzed 1,162 clinical isolates of E. coli for their susceptibility towards 21 FDA approved drugs and combined this information with whole genome NGS data to identify potential variants that might be causative for the observed resistance patterns. In total, the inventors found 25,744 significant sites (p-value < 10 ^~9 ) . The method cor ^¬ rectly identified already known drug targets in nine

gene/drug combinations: gyrA (Ciprofloxacin, Levofloxacin) , parC (Ciprofloxacin, Levofloxacin) , ampC (Cephalothin) , folC (Trimethoprim Sulfamethoxaxole) , mrcB (Cefazolin) , pbpC (Cefazolin) , and pbpG (Ceftazidime) . To identify other poten ^¬ tial sites that might be secondary drug targets, the inven ^¬ tors applied filtering criteria using the measures NPV/PPV which allowed the inventors to reduce the number of poten ^¬ tially relevant sites from 25,744 to 127 sites. Considering the best drug-target combinations according to the computed p-values, the inventors found the AA exchange S83L in GyrA to be the predominant mutation for 15 drugs. Since only Ciprofloxacin or Levofloxacin are approved drugs for GyrA, the other associations to this protein might be a side-effect of multi-drug resistance. The inventors showed that employing additional measures such as sensitivity, PPV, and NPV facilitates the separation of causative drug targets from other variants as exemplified in this case. Instead of using only single variants, a combination of sev ^¬ eral variant positions can improve the prediction accuracy and further reduce false positive findings that are influ- enced by other factors .

Since gene duplication and/or deletion might also play a role in resistance development mechanisms, the inventors analyzed the gene coverage combined with the resistance data and dis- covered 23 cases of abnormal differences in gene coverage be ^¬ tween resistant and susceptible bacteria. Interestingly, the inventors do not only find an increase of genetic material in resistant bacteria, e.g. for genes mmuP, mmuM, and yiel, but also a decrease in certain genes such as mngB and mngR. While for membrane or transporter proteins both an increase or a decrease of gene dosage can influence drug susceptibility by not allowing a drug to permeate the membranes or to more ef ^¬ ficiently transport it out of the cell, a decrease of the quantity of metabolic enzymes or transcription factors is not as easily interpretable in this context, and might be more or less directly related to the fitness of the isolates.

Another source of information that might improve the accuracy of our analysis are the strain-specific plasmids. Mapping the sequencing data against those plasmids will extend our knowledge about additional resistance mechanisms. In a first approach, the inventors mapped a subset of sequencing data to about 300 E. coli plasmids. Among the genes having the most significant variant sites were e.g. repAl , trbl, psiB, and traG that are directly involved in replication, plasmid transfer, and maintenance and might play an indirect role in resistance development by giving its host the ability to fa ^¬ cilitate spreading of resistance genes. Compared to approaches using MALDI-TOF MS, the present ap ^¬ proach has the advantage that it covers almost the complete genome and thus enables us to identify the potential genomic sites that might be related to resistance. While MALDI-TOF MS can also be used to identify point mutations in bacterial proteins ³³ , this technology only detects a subset of proteins and of these not all are equally well covered. In addition, the identification and differentiation of certain related strains is not always feasible.

The present method allows to compute a best breakpoint for the separation of isolates into resistant and susceptible groups. The inventors designed a flexible software tool that allows to consider besides the best breakpoints also values defined by different guidelines (e.g. European and US guide ^¬ lines) , preparing for an application of the GAST in different countries . Another critical point of this study is that it analysis only included cultured bacteria strains. Several studies used cul ^¬ ture-independent samples from urine, fecal samples, or vagi ^¬ nal swab and applied NGS to identify or characterize the pathogens directly. The advance of the NGS technology, in- eluding the development of new long read sequencers as PacBio and Oxford Nanopore, will further improve and speed up our procedure in the future to develop a culture-independent di ^¬ agnostic test based on NGS data. The inventors demonstrate that the present approach is capa ^¬ ble of identifying mutations in genes that are already known as drug targets, as well as detecting potential new target sites .

References

1. Didelot X, Bowden R, Wilson DJ, Peto TE, Crook DW.

Transforming clinical microbiology with bacterial genome se ^¬ quencing. Nat Rev Genet 2012;13:601-12.

2. Beerenwinkel N, Schmidt B, Walter H, et al . Diversity and complexity of HIV-1 drug resistance: a bioinformatics ap- proach to predicting phenotype from genotype. Proc Natl Acad Sci U S A 2002;99:8271-6.

3. Reuter S, Ellington MJ, Cartwright EJ, et al . Rapid bac ^¬ terial whole-genome sequencing to enhance diagnostic and pub- lie health microbiology. JAMA internal medicine

2013; 173 : 1397-404.

4. Wozniak M, Tiuryn J, Wong L. An approach to identifying drug resistance associated mutations in bacterial strains. BMC Genomics 2012; 13 Suppl 7:S23.

5. Stoesser N, Batty EM, Eyre DW, et al . Predicting antimicrobial susceptibilities for Escherichia coli and Klebsiella pneumoniae isolates using whole genomic sequence data. The Journal of antimicrobial chemotherapy 2013;68:2234-44.

6. Liu YF, Yan JJ, Lei HY, et al . Loss of outer membrane protein C in Escherichia coli contributes to both antibiotic resistance and escaping antibody-dependent bactericidal ac- tivity. Infection and immunity 2012;80:1815-22.

7. Johnson TJ, Siek KE, Johnson SJ, Nolan LK. DNA sequence and comparative genomics of pAPEC-02-R, an avian pathogenic Escherichia coli transmissible R plasmid. Antimicrobial agents and chemotherapy 2005;49:4681-8.