A microarray was developed to simultaneously detect Cryptosporidium parvum, Cryptosporidium hominis, Enterococcus faecium, Bacillus anthracis and Francisella tularensis in water.
A microarray was developed to simultaneously detect Cryptosporidium parvum, Cryptosporidium hominis, Enterococcus faecium, Bacillus anthracis and Francisella tularensis in water.
A DNA microarray was designed to contain probes that specifically detected C. parvum, C. hominis, Ent. faecium, B. anthracis and F. tularensis. The microarray was then evaluated with samples containing target and nontarget DNA from near-neighbour micro-organisms, and tap water spiked with multiple organisms. Results demonstrated that the microarray consistently detected Ent. faecium, B. anthracis, F. tularensis and C. parvum when present in samples. Cryptosporidium hominis was only consistently detected through the use of shared probes between C. hominis and C. parvum.
This study successfully developed and tested a microarray-based assay that can specifically detect faecal indicator bacteria and human pathogens in tap water.
The use of indicator organisms has become a practical solution for monitoring for water quality. However, they do not always correlate well with the presence of many microbial pathogens, thus necessitating direct monitoring for most pathogens. This microarray can be used to simultaneously detect multiple organisms in a single sample. More importantly, it can provide occurrence information that may be used in assessing potential exposure risks to waterborne pathogens.
Microbial contamination of water supplies represents a persistent health concern (Craun et al. 2010). The large number of pathogenic micro-organisms that could potentially contaminate a water supply makes it very difficult to rapidly and effectively monitor a system in a cost-effective manner. Traditionally, indicators of faecal pollution have been used to predict human health risks associated with pathogen contamination of the water supply. However, the correlation between the presence of indicators and waterborne pathogenic microbes has been variable (Field and Samadpour 2007). Until a more reliable surrogate system is available, specifically monitoring for human pathogens is currently the most accurate approach to assess human health risks associated with contamination of our water systems.
Polymerase chain reaction (PCR) has become increasingly useful for detecting and genotyping waterborne pathogens in environmental samples because it is rapid, specific, sensitive and relatively inexpensive. It has been successfully used to study the occurrences and/or levels of bacteria (Bej et al. 1991), viruses (Wolf et al. 2010) and parasites (Yang et al. 2008, 2009). One common theme among those studies listed is that all used PCR targeting only a single locus (e.g. 16S or 18S rRNA in the case of bacteria and Cryptosporidium, respectively) to detect and type each pathogen. To conduct multi-locus typing or detect multiple pathogens in the same sample, several PCR reactions are required, separately or in a single tube. This multi-assay approach can be problematic as it requires the sample to be divided into several separate PCR reactions, which can increase the chances of false negatives due to an organism of interest ending up in the ‘wrong’ aliquot. Alternatively, a multiplex reaction could be carried out in a single tube, but the sensitivity can be compromised as all the reactions will compete for the same PCR reagents. Primer mispairing and cross-reactivity can also occur resulting in the amplification of nontarget regions.
DNA microarrays offer the advantage of interrogating a sample for thousands of loci, in a single sample, making it amenable to identifying pathogens and characterizing the microbial community structure in various matrices (Avarre et al. 2007). Initially, microarrays like the ‘PhyloChip’ were used to understand bacterial community structures in the environment (DeSantis et al. 2007). This PhyloChip microarray was developed to contain DNA probes designed to hybridize to 16S rRNA prokaryotic sequences. Although it proved useful for elucidating the overall community structure and diversity of prokaryotic organisms in various natural environments, it was not designed to have the resolution to specifically identify microbial pathogenic species.
In other studies, microarrays were designed to detect sequences specific to pathogens of interest. For example, Lee et al. described the use of DNA microarrays to detect bacteria (Lee et al. 2008) and protozoa (Lee et al. 2010) in water. Both studies describe a two-step PCR-DNA microarray assay that first amplifies the 16S or 18S rRNA gene, respectively, followed by hybridization of these products onto a low-density DNA microarray. These results demonstrated that the two-step PCR + microarray worked well in detecting Aeromonas hydrophila, Bacillus cereus, Clostridium perfringens, Enterococcus faecalis, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa and Salmonella spp., for the bacterial microarray, and Cryptosporidium parvum, Acanthamoeba castellanii and Giardia duodenalis for the protozoan microarray. Similarly, a low-density microarray capable of genotyping Cryptosporidium spp. using the hsp70 gene was developed (Straub et al. 2002). This approach also required PCR amplification of the hsp70 gene prior to hybridization onto the microarray. Nevertheless, single-nucleotide polymorphism (SNP) analysis using this Cryptosporidium-specific microarray was able to differentiate C. parvum and C. hominis species in water matrices; however, it remains to be determined if it can identify other Cryptosporidium species and genotypes. In another study, Brinkman and Fout (2009) also demonstrated that noroviruses could be successfully genotyped in raw surface and finished drinking water samples using microarrays. More recently, Leski et al. (2009) developed a high-density resequencing microarray that has the capability of detecting 84 different types of pathogens ranging from bacteria, protozoa, viruses, including biothreat agents like B. anthracis, Francisella tularensis and Ebola virus. The approach adopted, however, employed PCR amplification of target DNA prior to hybridization to the microarray.
A common theme in the cited studies is the requisite PCR amplification prior to microarray analysis. Although this potentially increased the detection sensitivity of the microarray, it may not be practical in situations where a limited amount of sample is available, as it would require splitting the samples into several PCR reactions to amplify and detect multiple gene targets. In addition, PCR can introduce amplification biases or inefficient target amplification resulting in reduced assay sensitivity or false negative signals (Wang et al. 2004; Leski et al. 2009). Thus, a one-step direct DNA microarray detection design is a more practical approach for multi-locus genotyping and for simultaneously identifying several unrelated pathogens (e.g. bacteria, viruses and protozoa), especially in situations where the concentrations of micro-organisms are likely to be high. Alternatively, other approaches that increase nucleic acid levels without introducing amplification biases, such as whole genome amplification, might be useful when the concentration of microbes is expected to be low (Wu et al. 2006). In either case, the first step in the development of an effective microarray assay is to design a chip that is able to specifically identify the organisms of interest.
In this study, we describe a direct (i.e. non-PCR based) microarray assay designed to detect waterborne pathogens encompassing a bacterial indicator of water quality, two pathogenic protozoa and two bacterial select agents. This microarray was evaluated using samples containing DNA from targeted pathogens, nontargeted near-neighbour micro-organisms and a mixture of both groups spiked into a complex concentrated tap water matrix. Our results showed that this type of approach is more effective at simultaneously detecting multiple pathogens in pure sample as well as in a complex environmental matrix than conventional single or multiplex PCR detection technology.
Purified C. hominis oocysts (TU502) (Akiyoshi et al. 2002) were obtained from Tufts University. The Iowa isolate of C. parvum (Harley Moon; Waterborne, Inc., New Orleans, LA, USA) and C. muris (RN66), obtained from Dr. U. Iseki, Osaka University, Medical School, Osaka, Japan, were both propagated in-house using CF-1 mice (Jackson Laboratories, Bar Harbor, ME, USA) as previously described (Miller et al. 2007; Ware and Villegas 2010). Briefly, C. parvum oocysts were purified from the faeces by sieving, step sucrose gradients and caesium chloride purification, while C. muris oocysts were purified using a continuous Percoll gradient (Arrowood and Donaldson 1996). Purified oocysts were then resuspended in reagent grade water containing 100 U ml−1 penicillin and 100 μg ml−1 streptomycin (Gibco, Gaithersburg, MD, USA) and stored at 4°C. Oocysts were used within 3 months of isolation. Giardia muris (kindly provided by Drs Erik Hewlett and John Andrews, Case Western Reserve School of Medicine, Cleveland, OH, USA), G. duodenalis (H3; Assemblage B), Enterocytozoon hellem (CDC:0291:V213), Enterocytozoon intestinalis (50502, ATCC) and Enterocytozoon cuniculi (50502, ATCC) were isolated from cysts/spores propagated at the US EPA animal facilities as previously described (Schaefer et al. 1991; Hester et al. 2002; Hayes et al. 2003). Purified organisms were stored in sterile 1× phosphate-buffered saline (1× PBS; Sigma, St Louis, MO, USA). Toxoplasma gondii (VEG strain) was kindly provided by J.P. Dubey (USDA-ARS, Beltsville, MD, USA).
Seed stocks of B. anthracis (Sterne; Battelle Lot No.: BAS0306) were initially streaked onto tryptic soy agar (TSA) to confirm purity based on colony morphology. An aliquot of the stock was then inoculated in tryptic soy broth (TSB) and incubated, with shaking, for approximately 24 h at 35–37°C. Similarly, F. tularensis (LVS strain; Battelle Lot No. 011006) was streaked onto TSA containing 5% sheep blood to confirm purity based on colony morphology. An aliquot of the F. tularensis stock was then inoculated into cysteine heart broth (CHB) and incubated, with shaking, for approximately 24 h. Enterococcus faecium (ATCC 19434) was grown in 50 ml of brain heart infusion broth (BHIB) and incubated, with shaking, overnight at 37°C. One ml of the overnight culture was seeded into 100 ml of BHIB and incubated with shaking for 7 h at 37°C. The culture was then centrifuged to pellet bacterial cells. Bacterial cell pellets were stored at −70°C.
Genomic DNA from at least 1 × 107 fresh C. hominis, C. parvum, C. muris, G. duodenalis, G. muris, E. cuniculi, E. hellem and E. intestinalis (oo)cysts or spores was isolated using the MasterPure DNA purifications kit as per manufacturer's protocol (Epicentre Biotechnologies, Madison, WI, USA). DNA concentration was determined using the ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA).
Genomic DNA from B. anthracis and F. tularensis was extracted using a Gentra PureGene Genomic Prep kit (Qiagen, Valencia, CA, USA) in accordance with the manufacturer's protocol. The purity of the genomic DNA was analysed by gel electrophoresis and by its 260/280 nm ratio. DNA concentration was measured using PicoGreen dsDNA kit (Invitrogen, Carlsbad, CA, USA). Once quantified, the genomic DNA samples were stored at −70°C. Genomic DNA from Ent. faecium was extracted using a MasterPure Complete DNA and RNA kit (Epicentre Biotechnologies) as per the manufacturer's instructions. Genomic DNA from Bacillus thuringiensis (35646-D), B. cereus (10987-D), Shigella flexneri (29903-D) and E. coli K-12 (10798-D) was purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). Genomic DNA was stored at −70°C.
An aliquot of a concentrated tap water sample described in Brinkman and Fout (2009) was used to provide a matrix background for assessment of method performance. Briefly, 1000 l of dechlorinated tap water was filtered through the Virosorb 1MDS electropositive cartridge filter (Cuno, Meriden, CT, USA). Trapped particles were eluted twice with beef extract, and then further concentrated using celite (Fluka, St Louis, MO, USA). To remove potential inhibitors of molecular detection techniques, the sample was ultracentrifuged through a sucrose pad, then purified by solvent extraction and further concentrated with a microcon YM-100 unit (Millipore, Billerica, MA, USA). The final sample was concentrated in 1× PBS; 0·2% bovine serum albumin (Sigma).
The microarray platform used in this study was the CombiMatrix CustomArray microarray (CombiMatrix Corp., Mukilteo, WA, USA). The CombiMatrix 12 K and 4 × 2 K CustomArray were generated by randomly placing replicates of each identified capture probe among the array. A positive control using a prokaryotic ribosomal sequence was included in each microarray chip to determine successful labelling and hybridization. Genomic DNA was labelled with the Cy3-labelled dCTP (GE Healthcare, Piscataway, NJ, USA) using the BioPrime DNA Labeling System according to the manufacturer's protocol (Invitrogen). Labelled DNA probes were purified using a size exclusion purification column and quantified using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific). Hybridization of samples to the microarrays used the following procedure. First, each hybridization chamber was filled with nuclease-free water and foil tape was used to seal the chamber portals. The array was then incubated in a rotisserie hybridization oven at 65°C for 10 min, then removed and cooled to room temperature. The nuclease-free water was removed from each chamber and replaced with prehybridization solution [6× saline sodium phosphate EDTA solution (SSPE): 0·05% Tween-20, 20 mol ml−1 ethylenediaminetetraacetic acid (EDTA), 5× Denhardt's solution, 100 ng μl−1, denatured salmon sperm, 0·05% sodium dodecyl sulphate (SDS)] and then incubated at 47°C for 30 min. Hybridization solution (6× SSPE, 0·05% Tween-20, 20 mol ml−1 EDTA, 25% diformamide, 100 ng μl−1 salmon sperm DNA, 0·04% SDS) was added to approximately 2 μg of Cy3-labelled DNA, incubated at 95°C for 3 min and then cooled on ice for 3 min. The array was removed from the hybridization oven and the prehybridization solution was discarded and replaced with the hybridization mixture containing the Cy3-labelled target. The portals were again sealed using foil tape and the array was rotated in the rotisserie oven at 47°C overnight. Prior to washing the array, Wash 1 solution (6× SSPE, 0·05% Tween-20) was warmed to the incubation temperature of 47°C. The array was taken from the hybridization oven and the hybridization mixture was removed from the chamber. The preheated Wash 1 solution was added to the chamber and gently rinsed by pipetting the solution up and down. Wash 1 solution was removed and replaced with fresh preheated Wash 1 solution. The array was returned to the hybridization oven and incubated for 5 min at 47°C. The array was taken from the hybridization oven, Wash 1 was removed from the chamber and the chamber was rinsed using Wash 2 solution (3× SSPE, 0·05% Tween-20). The array was incubated at room temperature with fresh Wash 2 solution for 5 min, and the same wash procedure was used with Wash 3 (0·5× SSPE, 0·05% Tween-20), Wash 4 (2× PBS, 0·1% Tween-20) and final Wash (2× PBS) solutions. The final wash was performed twice and the final wash solution was retained in the chamber until the array was imaged. The arrays were scanned using an Axon Instruments Genepix 4000B and 4200A microarray scanners (Molecular Devices, Sunnyvale, CA, USA). Hybridization intensity values for each of the oligonucleotide capture probes on the arrays were analysed using Microsoft Excel (Microsoft, Redmond, WA, USA).
The hybridization intensity values were used to evaluate the results of the hybridization experiments. For each capture probe on the array, the mean hybridization intensity value from all the replicates (3–4 identical probe sequences) present on the array was determined. Analysis proceeded with a threshold criterion of 10 000 fluorescence intensity units. In this way, any capture probe that measured a hybridization intensity value of 10 000 or greater was considered a positive hybridization result. A sample was called positive if 33% of the total capture probes in the probe set designated to identify a target micro-organism were positive. These criteria were derived using positive and negative control samples during the initial validation of the microarray probes (data not shown) as well as those presented in Fig. 2 where the fluorescence intensity values detected fall below or above 10 000 units, respectively, for individual probes in the probe set.
Using Figur software (GenArraytion, Rockville, MD, USA), approximately 100 candidate probe sequences were initially identified from each target micro-organism's genomic sequence (Supporting Information, Table S1). These sequences were selected based on four criteria that were designed to reduce the risk of false-negative or false-positive results, including cross-reactivity, while meeting length requirements for the microarray. The criteria called for sequences that (i) were perfect alignment matches with the organism of interest, (ii) had alignments with very low expectation values (e-values; lower e-values indicate higher percentages of matching alignments), (iii) matched to the organism of interest, or other related organisms, or (iv) had some alignment divergence, specifically when alignment lengths fall at or below 50% of the prospective probe length (e.g. at or below 20 of the total 40 bases). This last criterion can be problematic because a 100% identity score can be obtained, even though the alignment length may only span 20 bases of the possible 40. At this point, the match is not likely to be the organism of interest, although it may represent an unidentified variant. These sequences were considered to maximize the candidate pools although it increased the chances of false positives. At times, no sequences were identified by the Figur software, other than the organisms of interest. Even beyond this point, any matches from similar organisms to that of the sequence origin were also kept. All other matches outside these criteria were removed.
The identified candidate probes were then evaluated based on their thermodynamic properties such as melting temperature, formation of secondary structures and self-self dimerization (data not shown). Based on these results, the top 15–25 candidate probes for each organism were selected and spotted onto Version 1 of the multi-pathogen microarray. These probes were further evaluated for specificity on select genomic DNA from various organisms (Fig. 1). Francisella tularensis, B. anthracis, C. parvum and C. hominis genomic DNA were specifically hybridized to probes designed for each organism. In addition, probes designed to detect both C. parvum and C. hominis probes easily detected genomic DNA from either parasite (Fig. 1e).
Probes for faecal indicator bacteria, Ent. faecium, were also designed and tested for specificity. In Fig. 1f, all 23 probes selected specifically detected Ent. faecium genomic DNA with signals ranging from 20 000 to over 60 000 fluorescence intensity units. Based on these initial performance characteristics of the hybridization probes against genomic DNA from irrelevant and closely related organisms, Version 2 of the multi-pathogen microarray was developed consisting of 21, 21, 15, 20, 23 and 13 oligonucleotide probes for C. hominis, C. parvum, shared C. parvum/C. hominis, B. anthracis, Ent. faecium and F. tularensis, respectively (Table 1). This new microarray design was then retested for cross-reactivity.
|Target micro-organism||Oligonucleotide probe number|
|Version 1||Version 2||Version 3|
|C. parvum + C. hominis||15||15||7|
The genomic material of near-neighbour micro-organisms to the target micro-organisms was harvested as described in the materials and methods, mixed and hybridized to the microarray to test the capture probes for cross-reactivity (Fig. 2). These near-neighbours were C. muris, G. duodenalis, G. muris, E. coli, B. thuringiensis and Sh. flexneri; Ent. faecium was included as a positive control. Unexpectedly, cross-reactivity of the probes to near-neighbour micro-organisms occurred. Those capture probes that shows cross-reactivity (hybridization intensity values >10 000) were further eliminated from the probe set. Table 1 shows the number of noncross-reactive capture probe set for each target micro-organism on the final Version 3 of the multi-pathogen microarray design.
To further evaluate the cross-reactivity of the Version 3 microarray design, mixtures of nontarget micro-organism from C. muris, G. muris, E. coli (K-12) and T. gondii were prepared and hybridized to the chip (Fig. 3a sample 1). Results revealed very little cross-reactivity of the selected probes to these samples. Only one probe from the F. tularensis and Ent. faecium probe sets was positive (>10 000 fluorescence intensity units) illustrating that 86% and 95% of the probe sets, respectively, were specific. In a separate experiment, genomic DNA from Ent. faecium, E. coli (K-12), B. thuringiensis, Sh. flexneri, C. muris, G. duodenalis, G. muris, E. cuniculi, E. intestinalis and E. hellem was hybridized to the same chip. Based on the microarray results, only the Ent. faecium probe set was positive (100% of probes in the final set), thus correctly identifying the presence of Ent. faecium DNA in a sample containing irrelevant DNA (Sample 2). In sample 3 in Fig. 3a, genomic DNA from C. hominis, C. parvum and Ent. faecium was mixed together and hybridized to the chip. Results revealed that 100, 83, 100, 28, 0 and 100% of capture probes in each probe set were positive in the shared C. parvum and C. hominis, C. parvum, C. hominis, F. tularensis, B. anthracis and Ent. faecium probe sets, respectively. In sample 4, 500 ng of genomic DNA from C. hominis, C. parvum and Ent. faecium and 250 ng of genomic DNA of each of F. tularensis and B. anthracis were mixed and hybridized to the microarray. The hybridization results from the multi-pathogen microarray revealed that 42, 16, 7·6, 100, 100 and 90% of capture probes were positive in the shared C. parvum and C. hominis, C. parvum, C. hominis, F.tularensis, B. anthracis and Ent. faecium probe sets, respectively. In the last sample, 250 ng of genomic DNA from C. hominis, C. parvum and Ent. faecium and 100 ng of genomic DNA from F. tularensis and B. anthracis were mixed and hybridized to the microarray. Results were 71, 66, 7·6, 100, 100 and 100% of capture probes were positive in the shared C. parvum and C. hominis, C. parvum, C. hominis, F. tularensis, B. anthracis and Ent. faecium probe sets, respectively.
Genomic DNA from C. hominis, C. parvum and Ent. faecium, at least 100 ng each, was spiked into a concentrated tap water sample (equivalent to 33 l of tap water). The hybridization results revealed that 85, 83, 15, 0, 0 and 100% of capture probes were positive in the shared C. parvum and C. hominis, C. parvum, C. hominis, F.tularensis, B. anthracis and Ent. faecium probe sets, respectively.
To decipher the hybridization results of Version 3 of the multi-pathogen microarray, a sample was determined to be positive for the presence of a particular micro-organism if at least 33% of the capture probes in the probe set were positive (hybridization intensity values were >10 000). Using this approach, the presence/absence of Ent. faecium was correctly identified in all samples. Similarly, the F. tularensis and B. anthracis probes accurately detected the presence or absence of these organisms in the samples, although there was a low level of cross-reactivity with some of the F. tularensis probes (samples 1 and 3). However, the percentage of these cross-reactive probes still fell below the minimum amount of positive probes (33%) that has been established to determine if that particular organism is present in the sample.
The evaluation of the C. parvum and C. hominis probe sets was more complex. The probe sets correctly showed that these organisms were not present in samples 1 and 2 and genomic DNA from closely related species, like C. muris did not cross-react with any of the probe sets. Similarly, in sample 3, the shared C. parvum/C. hominis, C. parvum and C. hominis probes were positive, correctly identifying the presence of genomic DNA from those micro-organisms. In sample 4, the C. hominis or C. parvum probe sets failed to correctly detect the presence of DNA from these organisms despite large amounts of spiked DNA in the sample. However, the shared C. parvum/C. hominis probes indicated the presence of at least one of these species in the sample. Similar to sample 4, C. hominis was not detected by the Cryptosporidium species-specific probes in sample 5. However, C. parvum was detected, as 66% of the probes were positive.
Finally, in Fig. 3b, version 3 of the multi-pathogen microarray was tested by spiking genomic DNA in a true environmental matrix. In this experiment, tap water concentrates were spiked with C. hominis, C. parvum and Ent. faecium genomic DNA prior to being hybridized to the microarray. Results revealed that the Version 3 of the microarray correctly identified the contents of the water sample. F. tularensis and B. anthracis were not detected, while, with the exception of C. hominis, C. parvum and Ent. faecium were easily detected amidst the tap water matrix.
The goal of this work was to be able to simultaneously detect genomic targets of multiple micro-organisms in a single sample. Using our microarray, we consistently detected the presence of F. tularensis, B. anthracis, Ent. faecium, C. hominis and/or C. parvum from preparations of pure genomic DNA and from genomic DNA seeded into a concentrated tap water matrix. Although the individual probe sequences for the targeted Cryptosporidium species showed inconsistent detection, the probes from genomic sequences used for the shared C. parvum and C. hominis worked effectively. In this way, the positive signal resulting from sample hybridization to the shared C. parvum and C. hominis probe set would indicate that either or both of the two human infectious species are present. Overall, the use of microarrays as a detection tool has several advantages, especially when compared to PCR as a detection tool. For example, a microarray can facilitate the detection of several loci from several microbial targets that go beyond the range of multiplex PCR assays. This keeps a particular sample from having to be split for several PCR reactions, which reduces the overall sensitivity of detection. In addition, the use of multiple genomic targets increases the confidence in certifying a detection event. Furthermore, detection of genetically modified or recombinant microbes present in a sample is possible through the use of probes to multiple loci.
A major disadvantage to using a microarray detection approach is the potential decrease in sensitivity as compared with PCR. However, studies by Vora and others suggested that PCR preamplification is not necessarily important for increasing microarray sensitivity (Vora et al. 2004). Additional studies performed using decreasing serial dilutions of F. tularensis genomic DNA indicated that this microarray assay was able to detect down to at least 20 genomic copies without a PCR preamplification step (data not shown). These results further suggest that the PCR step may not be necessary and instead could compromise the accuracy of the microarray detection assay as PCR can introduce biases in a mixed template reaction. Thus, this single-step microarray detection assay is a more ‘unbiased’ approach with high-throughput detection capabilities. Alternatively, other amplification techniques, such as whole genome amplification, hold promise as a way of increasing the microarray signal without biasing results (Wu et al. 2006). Regardless of the approach used, the first step is the development of an accurate microarray assay.
The design of Version 3 of the multi-pathogen microarray consists of 58 probes, each detecting unique loci, spotted in triplicate. This reduction from the initial 100 candidate probes was due to stringent in silico analysis of each probe's thermodynamic properties. Moreover, those passing the initial analysis were further tested empirically for specificity. Among those selected, 4, 13, 6, 7, 21 and 7 probes specifically detect B. anthracis, C. hominis, C. parvum, shared C. parvum/hominis, Ent. faecium and F.tularensis, respectively. That is equivalent to at least 174 PCR reactions, if carried out conventionally.
Designing the probes to use in custom arrays is a challenge. While unique, thermodynamically optimal probes can be deduced using specialized software and algorithms in silico, the experimental performance of such probes is not always as predicted. As we found in this study, those probes that were selected in silico for Version 2 of the microarray, some performed as predicted but there were other probes that did not (Fig. 2). It is anticipated that in silico tools will continue to improve to more accurately predict actual probe sequence performances minimizing the amount of time and resources invested in optimizing the microarray designs. Additional efforts focusing on improving target preparation and microarray hybridization steps in the procedure are warranted if additional target organisms are included on the current list.
The custom microarray designed in this study has several potential applications. It can be used as suggested here, for detecting the presence of microbes in water samples collected from various sources. In addition to tap water, as evaluated in this study, this tool could prove to be beneficial for detecting the selected pathogens in surface, ground, marine, waste and reclaimed waters. Another potential application could be in monitoring microbial contamination in the food industry (Rasooly and Herold 2008). This microarray could also be used in the surveillance of two biothreat agents, F. tularensis or B. anthracis. As the results of this work are promising, additional studies are ongoing to improve upon the reported microarray. Future work includes the redesign of the probes for C. hominis, inclusion of new probes to detect additional microbial pathogens (e.g. viruses) and field testing of the microarray to determine the presence of microbes in various aquatic environments. New designs should also consider including a viability determination component to discriminate between viable and nonviable organisms, such as propidium monoazide (PMA) (Brescia et al. 2009; Nocker et al. 2009).
As molecular-based technologies continue to improve and become easier to use, it is anticipated that the capital equipment needed and the cost of conducting a microarray-based detection assay will become more affordable, user friendly and easier to adopt in various laboratories. The microarrays developed in this study now provide a useful triage type tool that enables the research community, public utilities and homeland security to simultaneously monitor water supplies for select agents (B. anthracis and F. tularensis), regulated pathogens (C. parvum and C. hominis) and faecal indicator bacteria (Ent. faecium) that have been accidentally or intentionally contaminated. This could potentially improve response time to mitigate this problem.
The authors would like to thank David Lattier and John Martinson for critical review of this manuscript. US Environmental Protection Agency, through its Office of Research and Development, funded and managed the research described here. It has been subjected to the Agency review and approved for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation by EPA for use.