Identifying Antimicrobial Resistance Genes with DNA Microarrays

Abstract

We developed and tested a glass-based microarray suitable for detecting multiple tetracycline (tet) resistance genes. Microarray probes for 17 tet genes, the β-lactamase bla_TEM-1 gene, and a 16S ribosomal DNA gene (Escherichia coli) were generated from known controls by PCR. The resulting products (ca. 550 bp) were applied as spots onto epoxy-silane-derivatized, Teflon-masked slides by using a robotic spotter. DNA was extracted from test strains, biotinylated, hybridized overnight to individual microarrays at 65°C, and detected with Tyramide Signal Amplification, Alexa Fluor 546, and a microarray scanner. Using a detection threshold of 3× the standard deviation, we correctly identified tet genes carried by 39 test strains. Nine additional strains were not known to harbor any genes represented on the microarray, and these strains were negative for all 17 tet probes as expected. We verified that R741a, which was originally thought to carry a novel tet gene, tet(I), actually harbored a tet(G) gene. Microarray technology has the potential for screening a large number of different antibiotic resistance genes by the relatively low-cost methods outlined in this paper.

Antimicrobial resistance testing is an invaluable method for characterizing clinical and environmental bacteria, although a resistance phenotype provides only limited information about the identity of the actual resistance genes harbored by the microbes (15). In contrast, when the specific antibiotic resistance genes are identified, this information can be used to investigate the diversity and the spatial and temporal distribution of resistance genes within and between different host reservoirs, populations, and environments (1, 14). When identifying specific resistance genes, researchers have used a variety of methods, including DNA-DNA hybridization on membranes and gene-specific PCR assays (10, 14). PCR assays are relatively rapid and can be used with a variety of sample types, such as individual bacterial colonies or environmental samples (4, 5, 9). Multiplex PCR assays have been devised for detection of a large number of genes (12), and this approach could be used for different classes of antibiotic resistance genes. For example, it should be possible to devise multiplex PCRs to detect the 35 genetically diverse tetracycline resistance genes (8).

DNA microarrays offer an alternative method for screening for the presence of a wide diversity of genes. In this format, probes specific to each gene are deposited onto a solid substrate (usually glass) in a lattice pattern. DNA is then labeled and hybridized to the array, and specific target-probe duplexes are detected with a reporter molecule (21). In this study, we investigated the feasibility of using DNA microarrays as a tool to screen bacterial isolates for tetracycline resistance genes. Our initial efforts involved short oligonucleotide probes (25-mer), but we found this array design to be insensitive to low-copy-number genes (data not shown). Herein we demonstrate a more conventional format in which microarray probes consist of ca. 550-bp PCR products. This microarray functions with a high degree of specificity and is suitable for detection of multiple tetracycline resistance genes from multiple genera of bacteria.

MATERIALS AND METHODS

Controls and test isolates.

Seventeen cloned tet genes were used as PCR templates and hybridization controls for the microarray (Table 1). A diverse group of test isolates (n = 48), including 12 different genera, were assembled from previous studies (10, 14) and from bovine sources (Table 2). We included Escherichia coli transconjugants that harbored tet genes from other sources, including isolate 33, which harbors plasmid R714a from a Providencia species (13). All isolates either harbored known tet genes or required MICs that were consistent with resistance based on National Committee for Clinical Laboratory Standards breakpoints for E. coli (17, 18).

TABLE 1.

Primers and predicted amplification products used as microarray probes

Gene	Primersa	GenBank accession no.b	Locationc	Size (bp)d
tet(A)	TTGGCATTCTGCATTCACTC	X75761	1822	494
	GTATAGCTTGCCGGAAGTCG		2316
tet(B)	CAGTGCTGTTGTTGTCATTAA	V00611	275	571
	GCTTGGAATACTGAGTGTAA		846
tet(C)	ATATCGTCCATTCCGACAGC	Y19114	92	502
	CTGACTGGGTTGAAGGCTCT		594
tet(D)	GCAAACCATTACGGCATTCT	X65876	1638	546
	GATAAGCTGCGCGGTAAAAA		2184
tet(E)	TATTAACGGGCTGGCATTTC	L06940	518	544
	AGCTGTCAGGTGGGTCAAAC		1062
tet(G)	GCTCGGTGGTATCTCTGCTC	AF133140	1209	550
	CAAAGCCCCTTGCTTGTTAC		1759
tet(H)	TCAACCACACTTTGGATGCT	U00792	986	550
	CCCATTTTTGTGCCAATTTC		1536
tet(L)	ACTGGGTGAACACAGCCTTT	X60828	322	548
	CAGGAATGACAGCACGCTAA		870
tet(M)	ACACGCCAGGACATATGGAT	M85225	740	536
	ATTTCCGCAAAGTTCAGACG		1276
tet(O)	GCGGTAATTATGGGAAACGA	M18896	735	550
	TTTCCCGCTGTTCAGATTTC		1285
tet(S)	CGCTACATTTGCGAGACTCA	L09756	1267	555
	GAATGCCACTACCCAAAGGA		1822
tet(W)	GGGAAATTGTTCGGACAGAC	AJ222769	1090	549
	AACGGATACCATCCCTGACA		1639
tet(Z)	GATGAGATGGGGAAGGTTCA	AF121000	15666	544
	CTTGTTGGTAACCCGGAAGA		16210
tet(30)	CCGTCATGCAATTTGTGTTC	AF090987	1281	550
	TAGAGCACCCAGATCGTTCC		1831
tetA(P)	TTGGGGGAGTTTTAACAGGA	L20800	1292	564
	TTCCAAAAATTCCAAACCAA		1856
tetB(P)	TGGGGCAAATTTCAACAAAG	L20800	2710	537
	TCAACAACTCCCCCATTTTC		3247
otr(B)	CAAGAAACTGGCGATCGTG	AF061335	969	545
	GGAACCAGGTCATGACGAAC		1514
blaTEM-1	CCAATGCTTAATCAGTGAGG	L09137	1629	858
	ATGAGTATTCAACATTTCCG		2487
16S rDNA	AGAGTTTGATCMTGGCTCAG	AE000452	8	528
	ATTACCGCGGCTGCTGG		536

^aPrimers designed for this study. Primers for the bla_TEM-1 and 16S rDNA genes were adopted from references 2 and 16, respectively.

^bGenBank accession number for the reference sequence used to design the microarray probes.

^cPrimer location within the reference sequence.

^dPredicted size of the microarray probe.

TABLE 2.

Field isolates and other test strains used in this study

Isolate no.a	Species	Geneb	Intensityc	Thresholdd
1	Acinetobacter sp.	None		11,130
2	Acinetobacter sp.	None		1,113
3	Acinetobacter woffii	None		259
4	Escherichia coli	None		4,549
5	Escherichia coli	None		15,589
6	Pantoea sp.	None		17,606
7	Providencia rettgeri	None		12,467
9	Pseudomonas putida	None		23,694
10	Ralstonia pickettii	None		15,160
11	Aeromonas hydrophila	tet(A)	4,895	4,035
12	Aeromonas hydrophila	tet(A)	33,819	26,808
13	Citrobacter freundii	tet(A)	65,535	51,741
14	Citrobacter freundii	tet(A)	62,044	48,931
EC3e	Escherichia coli	tet(A),(B)	65,535	73,099
EC4	Escherichia coli	tet(A)	26,433	21,129
EC5	Escherichia coli	tet(A)	65,535	52,592
EC6	Escherichia coli	tet(A)	26,002	20,713
EC8	Escherichia coli	tet(A)	49,000	38,695
EC9	Escherichia coli	tet(A)	14,199	11,304
15	Escherichia coli	tet(A)	61,251	55,416
SSuTA1	Escherichia coli	tet(A)	48,778	38,454
SSuTA2	Escherichia coli	tet(A)	38,426	30,303
SSuTCH1	Escherichia coli	tet(A)	33,717	26,769
SSuTCH2	Escherichia coli	tet(A)	36,390	28,889
SSuTCH3	Escherichia coli	tet(A)	9,964	8,073
17	Pseudomonas fluorescens	tet(A)	48,640	39,580
18	Pseudomonas fluorescens	tet(A)	61,988	51,568
19	Pseudomonas fluorescens	tet(A)	29,881	24,043
20	Pseudomonas fluorescens	tet(A)	51,838	41,179
22	Pseudomonas sp.	tet(A)	59,082	46,700
24	Brevundimonas vesicularis	tet(B)	60,669	51,384
25	Enterobacter sakazakii	tet(B)	59,673	50,585
SSuT1	Escherichia coli	tet(B)	24,672	20,814
SSuT2	Escherichia coli	tet(B)	34,982	29,552
SSuT3	Escherichia coli	tet(B)	52,593	44,385
SSuTA3	Escherichia coli	tet(B)	61,319	48,373
26	Serratia liquifaciens	tet(B)	64,873	57,964
27	Citrobacter freundii	tet(D)	12,974	10,353
28	Citrobacter freundii	tet(D)	8,499	6,872
29	Aeromonas hydrophila	tet(E)	65,535	60,134
30	Aeromonas hydrophila	tet(E)	23,357	18,635
31	Aeromonas hydrophila	tet(E)	63,840	50,459
32	Aeromonas hydrophila	tet(E)	39,391	31,303
33f	Escherichia coli	tet(G)	32,777	25,887
34	Acinetobacter radioresistens	tet(H)	42,763	37,004
EC2	Escherichia coli	tet(H)	65,535	51,935
35	Moraxella sp.	tet(H)	60,039	53,763
EC1	Escherichia coli	tet(Z)	65,535	52,207

^aIsolate reference number. Both EC2 and EC3 harbored cloned genes and do not represent field isolates.

^bGene detected by the DNA microarray. Gene identity was confirmed by previous work (14) or by PCR with gene-specific primers (strains designated with an SSuT prefix).

^cMedian pixel intensity of the microarray probe.

^dDetection threshold based on the average median pixel intensity for tet probes plus 3 × SD.

^eEC3 was clearly positive for both tet(A) and tet(B).

^fIsolate 33 harbored plasmid R714a from a strain of Providencia (13).

Microarray probes.

We used PCR to generate probes for the microarray. Cloned DNA from 17 of 35 recognized tet genes was extracted by using a DNeasy kit (Qiagen, Valencia, Calif.) (Table 1) (http://faculty.washington.edu/marilynr/). Primers for each gene were identified with Primer3 software (20) with the goal of producing products ca. 550 bp in length (Table 1). This probe size was selected based on previous success by similar procedures for microarray hybridizations (6). Additional PCR products were used as controls, including the 16S ribosomal DNA (rDNA) gene (rrn from E. coli) and the bla_TEM-1 gene (from pBR322). All predicted PCR products were compared by using Vector NTI software (InforMax Inc., Bethesda, Md.) to verify that probe sequences were unique (<85% sequence identity between predicted probe sequences). Oligonucleotides were synthesized by Invitrogen Corp. (Carlsbad, Calif.) with no modifications and with basic desalting after synthesis. PCR mixtures (100 μl) were composed of 1× reaction buffer (Fisher Scientific, Pittsburgh, Pa.), 200 nM each deoxynucleoside triphosphate (dNTP), 2 mM MgCl₂, 4 U of Taq polymerase, 400 nM each primer, and ca. 40 ng of template DNA. The thermal cycler conditions included 2 min of denaturation at 95°C followed by 30 cycles of 96°C for 30 s, 60°C for 30 s, and 72°C for 30 s and a final 10-min extension time at 72°C. Each PCR product was verified by gel electrophoresis and ethanol precipitated. Products were resuspended in sterile water and quantified by spectrophotometry.

Slide preparation.

Multiple microarrays were printed on glass slides so that independent microarrays were contained within each of 12 wells defined by a Teflon-masked surface (Erie Scientific, Portsmouth, N.H.); the hydrophobic nature of the masking permitted independent samples to be hybridized within each well. Slides were derivatized with an epoxy-silane monolayer as described previously (7). Briefly, slides were sonicated for 2 min in a detergent solution (Contrad 70 detergent; Fisher Scientific), rinsed with deionized water, and dried by compressed air. Slides were then soaked for 1 h in 3 N HCl, rinsed, dried, and then incubated in a 2% solution of 3-glycidoxypropyltrimethoxysilane (Sigma Aldrich, St. Louis, Mo.) in high-performance liquid chromatography HPLC-grade methanol. After 15 min, slides were rinsed in 100% methanol and dried by compressed air.

Microarray construction.

PCR products were diluted to 75 ng/μl in print buffer (100 mM Na₂HPO₄, 200 mM NaCl, 0.01% sodium dodecyl sulfate [SDS]; pH ∼11), heat denatured (95°C, 5 min) with a thermal cycler, and allowed to cool to room temperature. Each probe was then deposited at a fixed location within each masked well by a robotic spotter (BioRobotics Microgrid II; Woburn, Mass.). Each probe was printed as four replicate spots within each array, and every array included an arbitrary oligonucleotide probe (25-mer) conjugated with biotin. These biotin pseudoprobes served as positive controls for the detection chemistry and served as orientation points for image processing. After being spotted, slides were baked for 1 h at 130°C under vacuum (22 Hg) and stored at room temperature while protected from light.

Hybridization and detection.

Target DNA was prepared by growing a test isolate in Luria-Bertani (LB) broth or on an LB plate. Cells were pelleted, and DNA was extracted with a DNeasy kit (Qiagen) followed by quantification by spectrophotometry. Target DNA (1 μg) was nick translated according to the manufacturer's instructions (biotin-dATP; BioNick kit; Invitrogen Corp.), except reaction mixtures were incubated for 2 h. The labeled DNA was used directly with hybridization buffer for the 48 test strains, whereas 1:1,000 dilutions were prepared from pure plasmid extractions containing cloned positive control genes (Table 1). To more closely simulate test samples, positive control clones were also prepared with a DNeasy kit extraction so that both genomic DNA and plasmid DNA were present in the labeling reaction mixture, and a 1:5 dilution was used in the subsequent hybridizations. Labeled DNA was ethanol precipitated and resuspended in 75 μl of hybridization buffer, composed of 4× SSC (0.6 M NaCl plus 0.6 M Na citrate) and 5× Denhardt's solution (0.001% Ficoll, 0.001% polyvinylpyrrolidone, 0.001% bovine serum albumin). In preparation for hybridization, slides were incubated with TNB (100 mM Tris-HCl [pH 7.5], 0.15 mM NaCl) with 0.5% blocking reagent (biotin Tyramide Signal Amplification [TSA] kit; Perkin-Elmer Life Sciences, Boston, Mass.) for 30 min at room temperature. Labeled DNA was heat denatured (95°C for 3 min), briefly centrifuged, and held on ice until applied to the slide. Blocking buffer was aspirated and immediately replaced with 35 μl of denatured target per well. Slides were then placed into a humidified, conical tube (50 ml) and submerged into a water bath overnight at 65°C.

After incubation, the labeled target was aspirated and could be stored at −20°C for replicate hybridizations. Slides were washed with a mixture of 1× SSC and 0.2% SDS at hybridization temperature (4 min) followed by 0.1× SSC-0.2% SDS at room temperature (4 min) and 0.1× SSC at room temperature (4 min) (11). Stringent washes were followed by three 1-min washes in TNT buffer (0.1 M Tris-HCl [pH 7.5], 0.15 M NaCl, 0.05% Tween 20). A streptavidin-horseradish peroxidase conjugate (TSA kit) was diluted 1:100 in TNB and incubated in each well for 30 min, followed by three 1-min washes in TNT. Ten percent fetal equine serum in 2× SSC was incubated in each well for 30 min to provide a protein surface for tyramide binding followed by three 1-min washes in TNT buffer. Biotinyl tyramide (diluted 1:50 in amplification diluent) was incubated in each well for 10 min. After biotinyl tyramide had been washed from the slide, 2-mg/ml streptavidin conjugated to Alexa Fluor 546 (Molecular Probes, Eugene, Oreg.) was added in 1× SSC-5× Denhardt's solution, and the slides were incubated for 1 h before being given three 1-min washes in TNT followed by drying by high-speed centrifugation. Images were captured with an arrayWoRx^e scanner (Applied Precision, Issaquah, Wash.), and image quantification was accomplished with softWoRx software (Applied Precision). Median pixel values are reported as signal intensity (averaged for replicate probes). Positive detection was confirmed by either PCR (primers from Table 1) or by DNA-DNA hybridization (10, 14). Resistant strains that failed to hybridize to any tet probes were hybridized a second time to confirm the finding.

RESULTS AND DISCUSSION

Microarray validation.

In all cases, there was a one-to-one correspondence between a control gene and its respective probe, and many hybridizations with cloned genes were positive for the bla_TEM-1 probe (Fig. 1). The latter result reflected the presence of the ampicillin resistance gene harbored by many of the cloning vectors. The tetA(P) and tetB(P) genes were contained on the same plasmid, so both probes were expected to be visible for this hybridization. We subsequently nick translated individual PCR products and demonstrated that both probes were specific for their respective targets (data not shown).

FIG. 1.

Hybridization results illustrating specificity of tet gene probes. Probes were printed in quadruplicate in two columns. From top to bottom of each array, the left column included tet(A), tet(B), tet(C), tet(D), tet(E), tet(G), tet(H), tet(L), tet(M), and tet(O). The right column included tetP(A), tetP(B), tet(S), tet(W), tet(Z), otr(B), tet(30), 16S rDNA, bla_TEM-1, and a biotin marker.

Threshold for detection.

Our positive control hybridizations yielded some minor cross-hybridization with other probes—most commonly involving tet(A) (Fig. 1). Therefore, we selected an arbitrary threshold for positive detection based on three standard deviations (3× SD) above the average signal intensity for all tet probes on the array (excluding 16S rDNA, bla_TEM-1, and biotin spots). For example, the results for 3× SD for the tet(A) and tet(S) hybridizations (Fig. 1) were 54,532 and 55,003, which yielded very clear detection thresholds for these hybridizations (Fig. 2). We adopted this procedure for scoring subsequent hybridizations. If more than one tet probe had a high signal intensity, then a 3× SD detection threshold may not be appropriate. For example, the threshold for the tetA(P) and tetB(P) hybridization (Fig. 1) was 73,754, which exceeds the maximum possible signal intensity, 65,535. Therefore, each hybridization needs to be examined to avoid loss of information where more than one tet gene may be present.

FIG. 2.

Example of quantified signal intensity for positive control hybridizations tet(A) (upper panel) and tet(S) (lower panel). The horizontal line represents a detection threshold equivalent to the average median intensity plus 3× SD for all tet probes.

Sample hybridizations.

We tested 48 isolates on the microarray: 38 isolates were identified as having a single tet gene, and 1 isolate (EC3) harbored two tet genes, tet(A) and tet(B). The nine remaining strains did not hybridize to any of the tet genes present on the array (Table 2). In the latter case, all nine field isolates were positive for the 16S rDNA probe, indicating that failure to detect a gene was not due to failed labeling and hybridization procedures, but presumably was due to the absence of a corresponding probe. Both positive and negative hybridization results for 38 isolates were described previously by using DNA-DNA hybridization for the 17 tet genes (10, 14). Positive detections for the remaining nine isolates (Table 2) were confirmed with gene-specific primers (this study). The E. coli isolate harboring plasmid R741a (isolate 33) was shown to have tet(G), and this was confirmed by sequence analysis (data not shown).

Assay design.

For the proposed technique to have applicability in a high-throughput setting, it would clearly require shorter incubation times and automation. In this study, we used overnight incubations before proceeding with TSA detection simply as a matter of convenience. We also tested the feasibility of reducing the hybridization incubation to 3 or 6 h by using tet(B) as a test target. Probe signal was not saturated at 3 h (intensity, 54,266), but saturation was evident by 6 h (intensity, 65,535), indicating that a truncated hybridization procedure is feasible.

Our prototype microarray demonstrates the feasibility of developing relatively high-throughput, planar microarrays for identification of resistance genes. These probe fragments can be cloned into a common vector to make the production process more efficient, and because PCR is not required to generate targets, the array is limited only by the physical constraints of the slide. For example, we have designed a new masked formatting that accommodates over 400 probes per well. If no masking is used, then it is not unusual to incorporate over 10,000 probes on a single slide. Nevertheless, microarrays designed from PCR products have the disadvantage of requiring positive control constructs for the PCR template. These long probes are less sensitive to minor point mutations, such as those responsible for protection against extended-spectrum β-lactams (3). Longer probes, such as the PCR products employed here, are clearly more sensitive for target detection (22), but a compromise may be achieved by using long (ca., 70- to 100-mer) oligonucleotide probes or peptide nucleic acids (19).