Transcript
Provided for non-commercial research and educational use only. Not for reproduction, distribution or commercial use. This chapter was originally published in the book Methods in Enzymology, Vol. 496 published by Elsevier, and the attached copy is provided by Elsevier for the author's benefit and for the benefit of the author's institution, for non-commercial research and educational use including without limitation use in instruction at your institution, sending it to specific colleagues who know you, and providing a copy to your institution’s administrator.
All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing copies or access, or posting on open internet sites, your personal or institution’s website or repository, are prohibited. For exceptions, permission may be sought for such use through Elsevier's permissions site at: http://www.elsevier.com/locate/permissionusematerial From: B. B. Ward and N. J. Bouskill, The Utility of Functional Gene Arrays for Assessing Community Composition, Relative Abundance, and Distribution of Ammonia-Oxidizing Bacteria and Archaea. In Martin G. Klotz and Lisa Y. Stein, editors: Methods in Enzymology, Vol. 496, Burlington: Academic Press, 2011, pp. 373-396. ISBN: 978-0-12-386489-5 © Copyright 2011 Elsevier Inc. Academic Press
Author's personal copy
C H A P T E R
F I F T E E N
The Utility of Functional Gene Arrays for Assessing Community Composition, Relative Abundance, and Distribution of AmmoniaOxidizing Bacteria and Archaea B. B. Ward1 and N. J. Bouskill Contents 374 375 375 379 380 381 384 389
1. 2. 3. 4. 5. 6. 7. 8. 9.
Introduction DNA Microarrays: Introduction to Microarrays Probe Selection Target Preparation Array Printing, Hybridization, and Scanning Factors That Influence Hybridization Results Array Applications Possibilities and Limitations Detailed Protocol for Functional Gene Microarrays Using Oligonucleotide Probes References
390 394
Abstract Ammonia-oxidizing bacteria (AOB) and archaea (AOA) transform ammonium to nitrite, an essential step in the complete mineralization of organic matter, leading to the accumulation of nitrate in oxic environments. The diversity and community composition of both groups have been extensively explored by sequence analysis of both 16S rRNA and amoA (encoding the critical enzyme, ammonia monooxygenase subunit A) genes. In this chapter, the power of the amoA gene as a phylogenetic marker for both AOB and AOA is extended to the development and application of DNA microarrays. Functional gene microarrays provide high throughput, relatively high resolution data on community
Department of Geosciences, Princeton University, Princeton, New Jersey, USA Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California, USA
1
Methods in Enzymology, Volume 496 ISSN 0076-6879, DOI: 10.1016/B978-0-12-386489-5.00015-4
#
2011 Elsevier Inc. All rights reserved.
373
Author's personal copy 374
B. B. Ward and N. J. Bouskill
composition and relative abundance, which is especially useful for comparisons among environments, and between samples in time and space, targeting the microbial group that is responsible for a biogeochemical transformation of interest, such as nitrification. In this chapter, the basic approaches to the design of probes to represent the target groups AOB and AOA are described, and the protocols for preparing hybridization targets from environmental samples are provided. Factors that influence the hybridization results and determine the sensitivity and specificity of the assays are discussed. A few examples of recent applications of amoA microarrays to explore temporal and spatial patterns in AOB and AOA community composition in estuaries and the ocean are presented. Array data are lower resolution than sequencing, but much higher throughput, thus allowing robust statistics and reproducibility that are not possible with large clone libraries. For specific functional groups, arrays provide more direct information in a more economical format than is possible with next generation sequencing.
1. Introduction Both ammonia-oxidizing bacteria (AOB) and the more recently discovered ammonia-oxidizing archaea (AOA) perform the critical step of oxidation of ammonia to nitrite in the nitrogen cycle of aquatic and terrestrial environments. Although many AOB are available in culture, and much has been learned from their genomes, extensive sequencing of both 16S rRNA and ammonia monooxygenase subunit a (amoA) genes from environmental samples demonstrates that the most abundant AOB in the environment are not represented in the culture collection. Similarly, many more operational taxonomic units (OTUs) corresponding to sequences of amoA genes from AOA are known than are represented in the very limited culture collection of AOA. Therefore, molecular methods are crucial for investigations of the distribution and ecology of ammoniaoxidizing microorganisms. The very existence of AOA was first discovered on the basis of amoA gene sequence information (Treusch et al., 2005; Venter et al., 2004) and initial clone library studies confirmed that AOA are widespread in the environment (Francis et al., 2005). Clone libraries targeting the functional genes of nitrifying bacteria continue to be the most common tool for these investigations. In this chapter, we describe a methodology building on sequence information obtained from clone libraries that provides a high-throughput approach for investigating the distribution and community composition of AOB and AOA assemblages without additional sequencing.
Author's personal copy Functional Gene Microarrays for AOB and AOA
375
2. DNA Microarrays: Introduction to Microarrays Functional gene microarrays provide a method for rapid, high resolution, high-throughput analysis of microbial community composition. They are applicable to DNA or RNA samples from any environment and for any target organism or gene for which a suitable sequence database exists or can be obtained. Hybridization data provide relative quantification of abundance and distribution of different archetypes or OTUs, such that robust comparisons can be made between samples, even though absolute quantification (i.e., number of organisms possessing each gene type) is not yet possible. Here we describe the development and application of DNA microarrays for the study of AOB and AOA. The rationale for focusing on these organisms is their essential role in the nitrogen cycle of aquatic and terrestrial environments, including wastewater treatment and agricultural systems. AOB and AOA are both relatively constrained taxonomically, as far as is presently known, so these groups can be comprehensively targeted and studied using the functional gene approach. A microarrays is a small solid support, usually a glass slide, plastic, or silicon thin film, on which a large number of discrete samples of biological materials (e.g., DNA) are fixed in an orderly arrangement. Those described here are based on glass microscope slide substrates. The slides are commercially coated (e.g., Corning, Agilent) with materials that allow DNA to be precisely bound to the substrate, but do not produce innate fluorescence or allow nonspecific binding of DNA, RNA, or other biological materials. The probes (DNA fragments representing the genes to be detected, see below) are printed robotically (DeRisi et al., 1997) and bound to the slide in a precise pattern. The targets (complementary DNA derived from DNA or RNA from the unknown sample, see below) are prepared by coupling a fluorescent tag to the sample DNA. The fluorescent target is then hybridized to the array, and the target molecules bind to the specific probes with complementary sequences. The amount of fluorescence associated with each probe is then quantified by laser scanning. The overall process is represented in the flow chart in Fig. 15.1 for reference throughout the steps described in the following sections.
3. Probe Selection Both AOB and AOA arrays use the amoA (ammonia monooxygenase subunit A) gene as the basis of detection of the two microbial groups. At present, the public databases contain thousands of sequences for both AOA
Author's personal copy 376
B. B. Ward and N. J. Bouskill
Align potential probe sequences
ATCGGTTAACTCGGGG......... ATCGCTTAACTTGGGG.......... TTCGCTTAACTCGGGG.......... TTCGCTTAACTTGGGG..........
Collect sample, extract DNA/RNA ATCGGTTAACTCGGGG.......... ATCGCTTAACTTGGGG.......... TTCGCTTAACTCGGGG.......... TTCGCTTAACTTGGGG..........
Make cDNA
Design/select probes, verify with in silico and experimental methods
Synthesize specific oligos, with 20-mer internal standard
Digest genomic DNA
Print probes onto array Amplify DNA/cDNA linearly using Klenow polymerase, incorporate dUaa 13463577
+
+ Scan and analyze Conjugate Cy3 to dUaa labeled fragment
Hybridize target and 20-mer standard to array
Figure 15.1 Flow chart of main steps in target and probe preparation and array analysis. The steps in the protocol for target preparation and hybridization are described more fully in the following detailed protocol.
and AOB amoA genes. Depending on the hybridization conditions and array format, the amount of sequence diversity between probe and target that allows hybridization can vary, and is not always predictable from
Author's personal copy Functional Gene Microarrays for AOB and AOA
377
sequence or thermodynamic data alone (Pozhitkov et al., 2006). Perfect match (PM) (i.e., 100% sequence identity between probe and target sequence) probe–target pairs usually hybridize well, and the stringency of hybridization conditions constrains the binding of mismatch (MM) targets. The degree of mismatch that still allows binding depends on the length of the probe and target fragments, the GC content and melting temperature, the location of mismatch basepairs (the stretch of matching or mismatched regions), and in some cases, the secondary structure (e.g., presence of hairpin structures). The empirical testing of probe design criteria and the behavior of target–probe interactions is well described by He et al. (2005). For oligonucleotide probes, the intensity of hybridization signal varies with the degree of probe–target sequence identity (He et al., 2005; Taroncher-Oldenburg et al., 2003). That is, under normal hybridization conditions with targets prepared from complex environmental samples, most probes are not 100% specific. A small amount of PM target can yield as strong a signal as a much larger amount of MM target. Thus, the challenge is to devise a probe set with appropriate specificity to distinguish between ecologically relevant sequence types. In most applications, 50 and 70 bp oligonucleotide probes bind well with targets within 85% sequence identity (He et al., 2005; Taroncher-Oldenburg et al., 2003). Thus for the AOA and AOB arrays, we have designated the probes as archetype probes, implying that the signal represents the relative abundance of an archetype, or group of sequences with 85–87% identity. This level of resolution is appropriate for AOB, for which the culture collection can inform the question of ecologically relevant distinctions. For cultivated AOB, 16S rRNA sequence divergence of 2–8% defines the genus level (Koops et al., 2003). amoA sequence varies by less than 80% within species, and by much more between species in the same genus (Purkhold et al., 2000). Using 85% as the cutoff (15% divergence), we have developed an algorithm (Bulow et al., 2008) that identifies an optimal probe set from the set of all homologous sequences. The optimal probe set contains the minimal number of probes required to allow all known sequences to hybridize with at least one (and preferably only one) of the probes. The algorithm performs essentially the same function as OTU definition according to DOTUR (Schloss and Handlesman, 2005) and other groups have developed similar criteria for probe selection (He et al., 2005, 2007). The archetype probe suite that should allow detection of all known AOB amoA sequences contains about 30 probes (Bouskill et al., 2010), and a similar number suffices to cover the entire known sequence database for AOA amoA sequences. The oligonucleotide probes are synthesized commercially. Even with robotic printing and high quality substrates, the intensity of hybridization can vary across the surface of a single array. Therefore, some
Author's personal copy 378
B. B. Ward and N. J. Bouskill
form of internal standardization is desirable. The standard approach for expression arrays in disease diagnostics is to employ a two-color competitive hybridization method. Two competing targets are prepared with different fluorescent tags (e.g., red and green), typically one representing the “control” condition and one prepared from “treatment” conditions. The two targets are then competitively hybridized to the same array, and the relative abundance for each target is determined from the fluorescence ratio (FR) between the two colors. The most commonly used fluorescent markers are Cy3 and Cy5, which are reactive water-soluble fluorescent dyes of the cyanine dye family: Cy3 (green: 550 nm excitation, 570 nm emission) and Cy5 (red: 650 nm excitation, 670 nm emission). In environmental samples, it is not obvious how to find or construct a “control” or “normal” sample. In theory, one could produce a mixture of every known sequence represented on the array and use that as the control. We have decided that this approach is not useful because the array hybridizes to all sequences within the 15% limit, not just the PM targets, and it is impossible to manufacture targets of every possible sequence variant for every probe. This has led some investigators to forego the two-color ratio approach altogether and to rely on single color, scaling the fluorescence intensity of each feature to that of a PM control (e.g., Bodrossy and Sessitsch, 2004). If hybridization is not perfectly even across the array, then the single color method may be prone to experimental artifacts, because it would be difficult to include sufficient PM probes to allow calibration of every feature. We have devised an alternative approach that includes controls and an internal standard two-color method to allow standardization at multiple levels. Firstly, the probe oligonucleotide (70 bp) is synthesized as a 90-mer with the additional 20 bp representing a nonsense sequence that is identical for every probe. The experimental target is labeled with one fluorescent tag, typically Cy3. In each hybridization experiment, a Cy5-labeled reference oligonucleotide, complementary to the nonsense 20-mer, is included in the hybridization mixture, and this 20mer binds to the complementary part of the 90-mer. Thus, each feature produces a Cy5 signal, regardless of the presence or absence of specific targets. When the specific target binds, each feature then yields a Cy3/Cy5 ratio. Since each probe includes both 70-mer and 20-mer, the ratio represents the ratio of unknown to standard binding and this ratio is not dependent upon the absolute amount of probe or target. Secondly, every block of features on the array includes a probe called “Mixall,” which is a mixture of all probes on the array. Any variation of hybridization quality across the array will be reflected in the intensity of Mixall signal, allowing for normalization of feature signal within each block. In practice, we usually find that data quality is excellent using the internal standard ratio without Mixall normalization, but the Mixalls provide a quality control check.
Author's personal copy Functional Gene Microarrays for AOB and AOA
379
4. Target Preparation The crucial step in target preparation is the incorporation of a fluorescent molecule into the DNA or cDNA to be hybridized to the array (see protocol; Fig. 15.1). The cyanine dyes are usually synthesized with Nhydroxysuccinimidyl (NHS) esters on the nitrogen side chains. These reactive groups bind to aliphatic amine groups. Therefore, in order to bind to DNA, the DNA must first be modified to incorporate aminomodified nucleotides, which can then be chemically linked to the esterified dye. The DNA modification is obtained by incorporating a modified nucleotide into the target DNA, and this can be done in the following three ways. (a) PCR targets: Use of specific PCR products as hybridization targets has the advantage of specificity and sensitivity. Because all the DNA in the target preparation comprises the specific gene fragment of interest, small mass of target is sufficient for a strong reaction, and nonspecific hybridization is minimized. Depending on the array format, 50–200 ng of PCR target suffices for a single array hybridization experiment. (i) One-step direct labeling: Incorporate the Cy3-modified nucleotides into the target molecule during the PCR step. Instead of an equimolar mixture of the four dNTPs, some of the dCTP is replaced with Cy3- or Cy5-modified dCTP (in a 1:1 or 2:1 ratio) (Taroncher-Oldenburg et al., 2003). Perform three replicate PCRs and combine the products before hybridization. (ii) Two-step labeling: Instead of an equimolar mixture of the four dNTPs, part of the dTTP is replaced with amino-allyl modified dUTP (dUaa; in a ratio of up to 1:10 dTTP:dUaa). After PCR, the purified fragment is chemically linked to Cy3- or Cy5-NHS ester. (iii) Three-step indirect labeling of PCR products: Perform the PCR as usual with an equimolar mixture of dNTPs. Then perform a nonlinear amplification of the PCR product using random primers and the Klenow polymerase. During the Klenow reaction, replace the equimolar dNTP mixture with the mixture containing dUaa (previous protocol). After the Klenow reaction, the purified product is chemically linked to Cy3- or Cy5-NHS ester. The indirect method involves the most steps but is often preferable because substitution of modified nucleotides often reduces the efficiency and yield of the PCR reaction in the one- and two-step labeling protocols. The Klenow reaction can amplify 25 ng of PCR product into 1000 ng of randomly labeled product and is therefore very useful for preparing targets from difficult PCR products. This 40fold amplification is the level cited in the guidelines for a typical
Author's personal copy 380
B. B. Ward and N. J. Bouskill
Klenow reaction use in labeling kits. We find that the main variable in determining the yield is the dUaa concentration. Under the conditions prescribed here, a yield even better than 40-fold is often obtained. (b) Whole DNA (WDNA) targets: Although specific and sensitive, PCR targets have all the problems of bias and selectivity of standard PCR. Therefore, it may be preferable to avoid PCR in target preparation. Instead of labeling a PCR product, as in (iii) above, simply perform the Klenow reaction with 25 ng of total DNA or total cDNA extract. Total genomic extracts are usually first digested with a restriction enzyme to allow better access to the DNA during the Klenow reaction. If low sample DNA concentration is a problem, the DNA or cDNA can first be amplified using whole genome amplification (Wu et al., 2006), in which case as little as 1 ng of original DNA or cDNA extract is sufficient. For WDNA targets, it is necessary to use 500–1000 ng or more of target DNA in each hybridization because most of the target preparation is not specific for any of the probes on the array. An advantage of WDNA targets, however, is that many different unrelated genes can be spotted on the same slide and all detected with the same target preparation.
5. Array Printing, Hybridization, and Scanning The probe oligonucleotides are printed onto Corning Ultragaps or Erie Scientific Epoxy Silane slides using a DeRisi style arrayer (DiRisi et al., 1997) and pins from Parallel Synthesis. The printing step can be performed by a local array facility or a number of commercial companies to your specifications. Dried slides are stored in a desiccator until ready to use. Slides are cross-linked at 70 mJ using a Stratagene Stratalinker just prior to hybridization, or baked at 80 C for 2 h prior to storage. If it is necessary to minimize excessive background hybridization, slides can be prehybridized using Blocking Agent. Prepare the hybridization mixture immediately before use. The mixture typically contains the labeled target, a control target (if using the two-color competitive approach) or control oligonucleotide (for the internal standard approach), and a hybridization buffer containing blocking agents. Quickly add the mixture to the slide or coverslip and seal. Incubate with gentle mixing at the desired temperature for 16 h. After hybridization, the slides are washed to remove excess target, dried, and scanned. The scanning procedures are platform specific and cannot be addressed generally here. After scanning, the arrays are analyzed using commercial software to quantify fluorescence of each dye in each feature.
Author's personal copy Functional Gene Microarrays for AOB and AOA
381
The fluorescence data are subjected to various filters to identify significant signals using the following steps: (1) Calculate the signal intensities for each feature by subtracting the background fluorescence for each channel (i.e., the wavelengths 532 nm (Cy3) and 635 nm (Cy5)). (2) Calculate the mean background fluorescence across both channels and for all features (i.e., amoA archetypes, controls) and identify significant signals as those having a signal intensity two standard deviations above the mean background fluorescence. (3) Calculate the ratio of Cy3:Cy5 fluorescence for each feature. For methods using a ratio of two dyes, remove from further analysis any features that do not have significant signal for both wavelengths. For single color analysis, features with no significant signal above background are considered zero. Compute the average of replicate features for each probe for the subsequent analysis. To subject the array data to subsequent statistical analysis, the absolute signal strength, the FR, or a relative fluorescence ratio (RFR) can be used, depending on the approach. For the internal standard approach used for the AOA and AOB arrays described here, RFR is the most robust measure. RFR is calculated as the contribution of each probe to the sum of FR for all probes in the set (e.g., all AOA amoA probes on the array).
6. Factors That Influence Hybridization Results For arrays containing hundreds or thousands of probes, it is not realistic to perform empirical validation tests for each probe. Such testing of some of the earlier applications with simpler arrays has been described (Bodrossy et al., 2003; He et al., 2005; Taroncher-Oldenburg et al., 2003). For large arrays, the predicted specificity is usually evaluated by computing various physical parameters for the probes: free energy of binding (http:// frontend.bioinfo.rpi.edu/applications/mfold/cgi-bin/dna-form1.cgi), percent identity with high identity MM targets, etc. (e.g., He et al., 2005, 2010). The principles derived from these characterizations are generally applied and a posteriori tests rely on reproducibility and internal correlations for validation of results. It is useful to illustrate a few general principles of array behavior, however, so that protocols can be optimized within the constraints of the array format. (a) Hybridization conditions: As with any hybridization based assay, the stringency of the hybridization and wash conditions are critical to the sensitivity and specificity of the assay. When using a small array containing a small number of relatively similar probes (e.g., all representing the same gene fragment), the GC ratio of all the probes is similar and one hybridization temperature can be optimal for all the probes. In complex
Author's personal copy 382
B. B. Ward and N. J. Bouskill
arrays, in which many different genes are represented, a single temperature may not be optimal for all probes because some genes have higher GC content than others, and thus their hybrids will be stable at higher temperatures. In selecting the probe region for multiple genes (see above for probe selection criteria), melting temperature is one of several features to be optimized. Similarly, the stringency of wash conditions after hybridization can influence specificity and cross reactivity. Longer low temperature washes tend to produce the best results, but stringency and signal strength must be balanced to obtain the best sensitivity for targets of interest. This will always require optimization with the targets and probe sets for each study. (b) Cross reactions: The characteristics of oligonucleotide probes were discussed above. Using optimized probe selection and hybridization conditions can yield highly specific and reproducible results, but some degree of nonspecific hybridization, that is, cross reactions, is probably unavoidable and cannot always be detected experimentally. A good way to address this problem is to include multiple probes for the same target organism. Multiple positive reactions are thus robust proof of target presence, while a mixture of positive and negative reactions would imply nonspecific reactions for some of the probes. This approach dramatically increases the number of probes required and the bioinformatics requirements, but is currently possible through commercial applications such as Nimblegen (http://www.nimblegen.com). (c) Length of incubation: Hybridizations are conveniently carried out “overnight,” a time period that ranges from 8 to 16 h. Target binding kinetics follow a Langmuir function (Dai et al., 2002), so that PM targets require longer to reach equilibrium than do MM targets. Thus, longer hybridization times should yield higher specificity of binding. (d) Concentration of target: The intensity of hybridization signal should increase with increasing target concentration. For an AOB amoA array, which was spotted with 25 pmol of probe in each feature, the relationship between concentration of PM target and fluorescence is shown in Fig. 15.2. It follows the generally expected linear increase in fluorescence at lower target concentrations, but appears to saturate, as all probe hybridization sites are filled (Held et al., 2003). Thus, it would be desirable for unknown targets to be present in the linear range of the response curve: if all targets are present at saturating concentrations, then no differences in signal will be discerned. In practice of course, it is not possible to optimize the concentration of every unknown target. The least amount of target that yields a strong signal is perhaps the best approach. The unknown sample, however, contains an undefined mixture of target and nontarget molecules, such that targets of similar total DNA concentration may contain quite variable amounts and mixtures of potential target molecules. There is essentially no way to
Author's personal copy 383
Functional Gene Microarrays for AOB and AOA
7 104 6 104 5 104
FR
4 104 3 104 2 104 1 104 0
0
10
20
30
40
50
60
70
DNA conc. (ng)
Figure 15.2 Fluorescence ratio (Cy3/Cy5 ¼ archetype fluorescence/internal standard oligo fluorescence) as a function of amount of target in hybridization mixture. The target was prepared using the three-step PCR protocol (amoA gene amplified with amoA-1F and amoA-2R primers (Rotthauwe et al., 1997)) from a culture of Nitrosospira multiformis, which is a member of AOB archetype A3.
account for this variability, which is a good reason to use relative abundance (percent of total signal) data in comparing microarray results for different samples. (e) Sensitivity: It would be useful to know the detection limit for individual target sequences in unknown samples, but this is not simple to determine. Figure 15.2 shows that 1 ng of labeled PM target produces a statistically significant signal. How much sediment must be extracted, or water must be filtered, in order to end up with a target that contains 1 ng of an individual archetype target? For mixed unknown samples, a large volume of water, often 4 L or more, is usually filtered and extracted. From this DNA extract a minimum of 25 ng of total DNA can generally produce 1000 ng of labeled target, the standard amount of mixed WDNA target recommended for the AOA array. Most of the 1000 ng in the target does not represent specific target molecules, however, and it’s not possible to determine how much of the mixed target represents a single sequence. If one could design quantitative PCR primers with the exact specificity of the archetype probes, this comparison could be made. But it is not generally possible to obtain
Author's personal copy 384
B. B. Ward and N. J. Bouskill
such specificity with qPCR. All of the archetype probe sequences can be obtained with one or two sets of amoA primers, and those are the primers most often used for qPCR of total AOA (Francis et al., 2005; Wuchter et al., 2006). As mentioned above in the protocols for WDNA target preparation, 1000 ng of target can be obtained from as little as 1 ng of total DNA extract using whole genome amplification, so sensitivity is not a great limitation to the array technology. (f) Probe/target capacity: Although all probe/target pairs follow the general relationship of increasing hybridization with increasing target concentration, the ratio of hybridization strength to target concentration (i.e., the slope of the linear portion of the curve in Fig. 15.2) varies among probe/target pairs. This is referred to as the probe capacity and cannot always be predicted from the thermodynamic properties of the PM or MM molecules (Held et al., 2003; Pozhitkov et al., 2006). For an AOB array, we showed that even among PM probe/target pairs, this capacity varies greatly (Ward et al., 2007). This means that the absolute hybridization signal cannot be interpreted in terms of absolute target concentration. In addition, the capacity varies between PM and MM probe/ target pairs for the same probe. In an unknown sample mixture, it is not known what portion of the signal is due to PM versus MM hybridizations. In an attempt to optimize specificity for highly similar probes, Marcelino et al. (2006) developed a mathematical approach for analysis of microarray data. Although not generally applicable to all array formats, this paper (Marcelino et al., 2006) is a useful resource for discussion of the hybridization behavior of oligonucleotide probes and targets. In practice, protocols are usually optimized and then performed consistently in order to remove variability from factors such as hybridization time and total DNA concentration.
7. Array Applications To illustrate the application of the functional gene microarray approach to study of AOB and AOA, we present a few recent examples. Both AOA and AOB are present in most environments, and their relative abundance often varies in correlation with environmental variables (e.g., Santoro et al., 2008), particularly in estuarine environments. AOB are more abundant in the upper freshwater reaches of Chesapeake Bay, where they exhibit an annually repeating pattern of community assembly (Bouskill et al., 2010). Using an AOB amoA array containing 26 betaproteobacterialAOB archetype probes representing all published sequences, 4 years of sampling showed that the assemblage varied consistently with season
Author's personal copy 385
Functional Gene Microarrays for AOB and AOA
(Bouskill et al., 2010). The water column was sampled three to four times each year over a 4-year period and the relative abundance of all known AOB lineages examined simultaneously. The AOB assemblages varied predictably with season. While assemblages were often dominated by estuarine archetypes (typical of estuaries that experience wide salinity ranges), seasonally reoccurring dynamics were also observed for “rare” archetypes (i.e., those that were rarely a major component of the RFR). An example of AOB assemblage variation over time is shown for a station in the upper Chesapeake Bay at 1 m above the bottom (Fig. 15.3,
TP1_D 25 RFR
20 15 10 5 0 2001
2002
2003
2004
TP2_D
RFR
20 15 10 5 0 2001
2002
2003
2004
TP3_D 40 RFR
30 20 10 0 2001
2002
2003
2004
August
2001
May
October October August
2002
2003
2004
Figure 15.3 Temporal classification of covarying archetypes based on K-means discrimination analysis at 1 m above the bottom in upper Chesapeake Bay. The left panel represents the RFR of individual archetypes in each temporal pattern (TP) over time. The identification of the archetype is not relevant for the overall analysis, which is to determine the seasonal behavior of the archetype as a whole. The right panel shows a centroid compiled from the cumulative behavior of all the archetypes in each cluster over time. This centroid is without magnitude and incorporates the trend of the RFR signals in each cluster through time.
Author's personal copy 386
B. B. Ward and N. J. Bouskill
left panel). Applying a discrimination analysis to the array dataset yielded a time series analysis in which different archetypes were classified by their temporal RFR patterns. Here the 26 AOB amoA archetypes grouped into three significant clusters, with three distinctly different temporal patterns (TPs), regardless of the absolute magnitude of the hybridization signal. The three TPs identified at the deep depth showed strong seasonal reoccurrence, as illustrated in the centroid for each group (Fig. 15.3, right panel). The profile of TP1-D appears to be an October signal (except for Oct. 2002), TP2-D was a repeating August pattern and TP3-D appeared as a spring pattern with peaks in all four April samples. The array was also sensitive enough to demonstrate community resilience following a significant perturbation event to the surface water (Bouskill et al., 2010). The assemblage changed dramatically in response to flooding caused by a hurricane, but the “normal” assemblage returned within a few months. Following the hurricane flooding, the community was dramatically altered; the estuarine archetypes disappeared and the assemblage was streamlined down to two archetypes. Physiological characteristics for the two archetypes could explain their success under the flood conditions caused by the hurricane. Resilience was demonstrated by the return to a “normal” assemblage following removal of the perturbation. For reasons of cost and labor, it would have been impractical to attempt to detect the patterns found in this dataset using clone libraries or either Sanger or next generation sequencing. The current AOA array (Bouskill et al., submitted) contains 31 archetype probes, and was developed from the public database of amoA sequences in 2009. amoA sequences continue to accumulate in the public database rapidly and most of them are the result of clone library studies using the same primer sets. Recent checks of newly published sequences shows that the current probe set should cover the vast majority if not all of them. The AOA array has been used to characterize community composition in samples from diverse environments (estuaries to oceans, sediments, and water column) around the world (Chesapeake Bay, Pacific, Atlantic, and Indian Oceans) (Bouskill et al., submitted). For two samples, the community composition derived from targets prepared by PCR and WDNA are compared in Fig. 15.4. In both cases, the RFR values derived from both target methods are very similar; the RFRs estimated from the two targets are highly correlated. The same dominant archetypes were identified by both methods and greater variability is associated with detection of archetypes that are responsible for less than 5% of the total signal. Nevertheless, both targets contained some signal from most of the archetypes, and the array analysis yields greater diversity estimates than obtained from a typical clone library. This suggests that use of PCR to obtain amoA sequences for AOA does not dramatically bias the estimates of community composition for targets derived from the same DNA extraction. This may not be true for every probe set and probably depends on the PCR primers and the divergence of the target genes.
Author's personal copy 387
Functional Gene Microarrays for AOB and AOA
A
ETSP_St.10_80 m 0.25 WDNA
0.015
WDNA-generated targets
0.2
0.15
R2 = 0.686
0.01 0.005
R2 = 0.964
0 0
0.002
0.004
0.006
0.008
0.01
PCR
0.1
0.05
0
0
0.05
0.1 0.15 PCR-generated targets
0.2
0.25
B Sargasso Sea_100 m 0.04 2
WDNA
WDNA-generated targets
R = 0.8765
0.03
0.3
0.02 0.01 0
0.2
R2 = 0.9854 0
0.01
0.02
0.03
0.04
PCR
0.1
0
0
0.05
0.1
0.15 0.2 PCR-generated targets
0.25
0.3
0.35
Figure 15.4 Comparison of RFR signals from targets prepared from the same sample by the three-step protocol for amoA PCR fragments and whole DNA (WDNA). (A) Eastern Tropical South Pacific Station 10 80 m and (B) Sargasso Sea 100 m, April 2004.
Using WDNA targets, we compared the relative abundance of AOA archetypes (RFR) among three samples (Fig. 15.5). Each colored bar represents the percent of total fluorescence (RFR) attributed to a particular archetype as the average of duplicate arrays prepared from the same target preparation. In the deep water (1 m above the sediment) of the seaward end of Chesapeake Bay (CB_300D) and 100 m depth of the open ocean of the
Author's personal copy 388
B. B. Ward and N. J. Bouskill
AOA32_CN8C_20_EF382433 0. AOA31_EF500_19O12_EF106947 AOA29_DS2_1_EF382468 AOA28_DS2_6_EF382473 0. AOA27_HB_29_EU022786
Relative fluorescence ratio (RFR)
100 90
AOA26_AOAC-S_SA09_EU339380 AOA25_DS2_2_EF382469 AOA24_DS4_20_EF382456 0. AOA23_HF770_22G04_EF106902 AOA22_AJ41-4_EU553368 AOA21_HF770_36M12_EF106908 0. AOA20_CN8C_17_EF382430
80 70 60 50 40 30 20 10 0 CB300_D_2003 (Coastal)
SS_M (Open Ocean)
ETSP_St.23_20 m (Offshore)
AOA19_MG85-37_EU553389 AOA18_TOB_61_DQ501021 AOA17_JCS82-4_EU553403 0. AOA16_HB_13_EU022770 AOA15_S33_A_12_EU025184 AOA14_CB3_14 0. AOA13_DS2_16_EF382483 AOA12_R60-70_278_DQ534884 AOA11_AOAC-U_SB06_EU339389 AOA10_AOAB_SH04_EU339454 0. AOA9_GOC-C-450-2_EU340536 AOA8_BS2-130MD3_EF414247 AOA7_AOAC-S_SA10_EU339381 AOA6_QY-A38_EF2072146 0. AOA5_TOB_159_DQ501119 AOA4_TOB_44_DQ501003 AOA3_GEO_OT2_AM260489 0. AOA2_BS80E_D4_EF414277 AOA1_OA-SA10-64_AB373281
Figure 15.5 Stacked bar plots comparing AOA community composition in three samples using targets prepared by the three-step protocol using WDNA.
Sargasso Sea (SS_M), archetype AOA12 is a major component. The clone library sequences that define this archetype were all derived from soils or estuarine sediments. The relative dominance of this archetype in the Sargasso Sea is not consistent with the robust biogeography of clone library sequences (Francis et al., 2005; Prosser and Nicol, 2008), which would suggest we should expect quite different assemblages in Chesapeake Bay and the Sargasso Sea. The Chesapeake Bay and Sargasso Sea samples were also similar in the relative contribution of archetypes AOA4 (sequences derived from various sediment environments, and including the hot spring enrichment culture, Nitrososphaera gargensis) and AOA26 (representing soil phylotypes). The appearance of sediment clades in the Sargasso Sea sample is unexpected, but unlikely due to artifacts of the method, such as cross reactions between probes. The level of identity between these probes representing sediment clades and those representing oceanic clades (<79%) is well below the threshold for cross reactivity in this format and there are no internal regions of very high identity within the 70 bp probe sequence. The array may be detecting sequences that are not amplified with the primers used to build the clone libraries upon which the biogeography of AOA has been described. The community composition of the Chesapeake Bay and Sargasso Sea samples was, however, quite different from the community composition found at 20 m depth at a 4000 m deep station in the Eastern Tropical South Pacific (ETSP) (Fig. 15.5). The dominant archetype in the ETSP sample was AOA9, which represents sequences obtained from deep ocean water with low oxygen content. AOA2, which represents AOA sequences from shallower depths in the water column, was also important only in the
Author's personal copy Functional Gene Microarrays for AOB and AOA
389
ETSP sample. AOA21, representing sequences derived from deep ocean water, and AOA31, representing sequences derived from the water column of Monterey Bay, CA, were both much more important in the ETSP than either of the other two samples. Interestingly, archetype AOA1, which includes the cultivated ammonia-oxidizing archaeon, Nitrosopumilus maritimus, and a large number of published AOA amoA sequences from clone libraries, did not make a significant contribution to the array signal from any of these samples, suggesting once again that even this cultivated type with oligotrophic characteristics (Martens-Habbena et al., 2009) is not representative of many natural assemblages.
8. Possibilities and Limitations Microarrays have been developed and applied for many different additional functional groups, including nitrogen fixers (Moisander et al., 2007; Zhang et al., 2007), methane oxidizers (Bodrossy et al., 2006), denitrifiers (Bulow et al., 2008), sulfate reducers (Loy et al., 2002, based on 16S rDNA genes, rather than functional genes), among others. The most ambitious functional gene microarray developed to date is the GeoChip, now in its third generation (GeoChip 3.0; He et al., 2010) with 27,812 50bp oligonucleotide probes representing 292 genes or enzymes, including ammonia oxidation. The GeoChip and the many other smaller more focused arrays mentioned here all share some similar attributes, advantages, and limitations. Because the sequence must be known to be included on the array, the array cannot be used to discover new functional gene sequences. The probes may hybridize with sequences that are similar to but different from known sequences, but those new sequences are not retrieved for further analysis, and totally unknown genes that might fulfill the same function cannot be discovered with the array. Therefore, it is not an exploratory tool. Arrays do, however, make it possible to detect and describe the diversity of known genes with greater breadth and less expense than clone libraries (DeSantis et al., 2007). For example, a typical small clone library might contain 20–100 clones, usually 20–30 (Francis et al., 2005). In a typical environment, where one or a few phylotypes are dominant, many of those clones would be redundant. The total number of different OTUs typically reported for a clone library study of AOA with an OTU definition of 5% is on the order of 4–12 (Francis et al., 2005). Even with a much coarser OTU definition (15% for the array), many more OTUs, up to the total number of probes on the array, might be detected in each sample analyzed on the array (Fig. 15.4). If natural assemblages are often structured by a few abundant and many rare types, then a typical clone library would not include most of them, especially the rare types, while the array can detect them, in both PCR and WDNA preparations.
Author's personal copy 390
B. B. Ward and N. J. Bouskill
Neither clone libraries nor array hybridization yields reliable quantitative information, but in combination with qPCR, improvement is likely possible. Clone libraries offer greater resolution because exact sequences are obtained, but are constrained by time and expense to much less coverage than is possible with arrays. Because most sequences have been derived by PCR, the array approach cannot be free of PCR bias, but this can be minimized by the use of WDNA targets to avoid PCR bias in sample analysis. Clone libraries are very rarely replicated, while replicate sample analysis is rapid and simple for microarrays. Next generation sequencing may eventually replace microarray analysis, but this may not be soon. For focused analysis of specific functional groups, microarrays provide highthroughput analysis for which every datum is relevant, avoiding the necessity of extensive bioinformatics to sift through millions of reads to find the few sequences of importance for a particular functional gene.
9. Detailed Protocol for Functional Gene Microarrays Using Oligonucleotide Probes I. Target preparation starting from PCR products or genomic DNA (or cDNA made from RNA) extracted from an environmental sample. This protocol contains the steps required for dUaa labeling during a Klenow random priming reaction followed by conjugation with Cy dyes. 1. Begin with clean target DNA. When using PCR products, prepare at least three standard PCR reactions of the target sequence (or more if necessary to obtain enough for a few reactions at 25 ng each at the Klenow stage). Combine all the products and clean them using Qiaquick columns (standard Qiaquick protocol, Qiagen), so that you elute the pooled products in 2 30 mL of water or EB. Reduce this volume down to 20 mL or less (using a speed vac) if necessary to concentrate prior to the random priming reaction (see below). Measure the DNA concentration (using Pico Green, Invitrogen) after this step in order to select the appropriate volume for the next step. When using total genomic DNA (WDNA), it is necessary to cut the DNA into smaller fragments for labeling. Use any four cutter restriction enzyme. Clean up the digested DNA or cDNA with Qiaquick colums (standard protocol) or with an EtOH precipitation. Measure the DNA concentration of the genomic DNA. Use 50 ng in the digest: DNA HinF1 Buffer Water
x mL to contain 50 ng 0.1 mL (or 1 unit) 5 mL 50 (sum of above volumes) ¼ mL
Author's personal copy 391
Functional Gene Microarrays for AOB and AOA
Incubate 2 h 37 C. Add 5 mL 3 M Na-acetate and 100 mL 100% EtOH, mix well, ppt 15 min on ice. Spin 15 min in microfuge, remove supernatant with a pipet. Add 50–100 mL 70% EtOH, mix well, spin 5 min, remove supernatant with pipet. Dry 10 min at 40 C in the speed vac. Measure the DNA concentration after this step in order to select the appropriate volume for the next step. If using cDNA, it’s probably already in fragments, so proceed directly to the labeling reaction, assuming your cDNA is clean (if in doubt, Qiaquick). 2. Use the cleaned product in a random priming labeling reaction using the BioPrime kit (Invitrogen). Only the random primers (octomers), the buffer and the Klenow, not the dNTP mix from the kit, are used in the reaction. Make your own dNTP mix (below) containing dUaa Reaction mixture: 2.5 random primer mix DNA (25 ng) plus water to total volume 1.2 mM dNTP mix with dUaa (see below) Klenow enzyme
20 mL 24 mL 5 mL 1 mL
Mix the random primers and diluted DNA and heat at 95 C for 5 min, then put on ice. Add the dNTP mix and the enzyme, mix gently, and incubate at 37 C for 2–4 h. Add 5 mL stop buffer and proceed. Klenow dNTP mix (10 reactions worth): 10 mM dACG mix 10 mM dTTP 10 mM dUaa dH2O
19.8 mL 2.2 mL 24.2 mL 3.8 mL
3. Clean up the random priming reaction following the standard Qiaquick protocol except that modified EB and PE buffers (see below), not the Qiagen buffers, must be used. Use phosphate buffer (see below) for washes (2–3 is enough) and elute in 2 30 mL water or phosphate EB (see recipes below) (NOT Qiaquick’s EB!). Store product frozen at this stage or dry down to pellet and store pellet frozen. In general, use the PO4 buffers for cleanup when cleaning dUaa preps—the Qiaquick buffers contain Tris and the Cy3 can couple to the amino groups.
Author's personal copy 392
B. B. Ward and N. J. Bouskill
EtOH precipitation works very well here too (50 mL reaction, add 5 mL. 5 N NaOAc þ 100 mL 100% EtOH, wash with 70% EtOH). Resuspend the pellet in 50 mL of water. Measure the DNA concentration using PicoGreen. Aliquot the volume containing up to 1000 ng of dUaa DNA into a clean tube and reduce it to dryness using a speed vac. PO4 buffers 1. 1 M K2HPO4 ¼ 17.4 g/100 mL 1 M KH2PO4 ¼ 13.6 g/100 mL 2. 9.5 mL 1 M K2HPPO4 þ 0.5 mL 1 M KH2PO4 ¼ 1 M KPO4, pH 8.5 3. 0.5 mL 1 M KPO4 þ 15.25 mL dH2O þ 84.5 mL EtOH ¼ PO4 wash buffer (use this instead of Qiaquick PE) 4. 10 mL 1 M KPO4 þ 2.49 mL dH2O ¼ PO4 EB (use instead of Qiaquick EB)
4. Couple the Cy3 dye to the dUaa labeled DNA as follows. First, dissolve one dye pellet (40,000 pmol each, Monofunctional NHSester Cy3, GE Lifesciences) in 45 mL of DMSO. Keep it frozen and use in aliquots of 4.5 mL each (see below). The dye is not very stable in water or DMSO so do not make up more than you will use. Hydrate the dry dUaa DNA pellet (upto 1000 ng DNA) from step 3 in 4.5 mL of 100 mM NaCO3 buffer (pH 9) for 15 min. Then add 4.5 mL of the dissolved dye pellet, mix with the pipet without bubbling, and incubate in the dark for 3 h to overnight. Add 4.5 mL of 4 M hydroxylamine to quench the reaction and incubate a further 15 min in the dark (this last step is not necessary if you move immediately to the clean up step). Clean up the labeled product with the Qiaquick kit but follow these directions. Add 25 mL dH2O to the reaction before adding 200 mL PB buffer Add 3 mL of 3 M NaOAc, then mix the pink solution and add it to the column and spin through. Do 5 750-mL washes with PE or the alternate wash buffer (see above) Elute in 2 30-mL washes of EB or water Save 1 mL for PG and Cy3 quantification and dry down the rest of the eluted target, store frozen. Do two conjugation reactions of 1000 ng each and combine them. Measure the DNA concentration and perform the hybridization with up to 1000 ng per array. For PCR products, 200 ng suffices, but for WDNA targets, 500–1000 ng is preferable.
Author's personal copy Functional Gene Microarrays for AOB and AOA
393
II. Hybridization protocol. This protocol is designed for use with Agilent gasket slides (http://www.agilent.com). It can be modified by scaling the volumes (and Cy5 amount) appropriately for 2-, 4-, and 8-well slides. The volumes given here are for 4-well slides. Total volume per well: 2-well gaskets: 200 mL 4-well gaskets: 100 mL 8-well gaskets: 80 mL 1. Hybridization mixtures Make hybridization mixture on ice: 50 – (x þ 1) mL H2O 50 mL 2 Agilent buffer x mL Cy3-target (e.g., 1000 ng worth for mixed environmental sample WDNA target) 1 mL Cy5-ref oligo (i.e., 0.5 pmol for 2-well slides, less for 8-well) 2. Working quickly, or preferably in an ozone free room, mix the hybridization mixture briefly (do not vortex) and heat at 95 C 5 min in wet block covered with foil. Cool on ice 2 min. Set up the gasket: Place the gasket coverslip in an Agilent hybridization chamber with the gasket side up. Apply the entire 100 mL hybridization mixture to one well inside gasket. Add mixtures to all gaskets (the volume will be about half of the gasket volume). Carefully place the array slide on top of coverslip, print side down. Close and secure hyb chamber Knock the chamber sharply on the counter to make all the bubbles coalesce into one large one in each gasket. It is important that the bubbles do not stick to the edges, make sure the entire solution flows when you turn the chamber. Place the hybridization chamber in the prewarmed hybridization oven and be sure to balance each chamber with another in exactly the opposite position in the rack. 3. Incubate at 60 or 65 C (depending on probe set) in the oven rotating overnight (16 h or longer). 4. Prepare the wash solutions (below).
Author's personal copy 394
B. B. Ward and N. J. Bouskill
Open the hybridization chamber, place slide and gasket (still stuck together) in a container of Wash #1. Use forceps to pry the coverslip off the slide while submerged. Be careful not to touch the slide with your fingers except on the edge, and try not to break the slide. 5. Transfer slide quickly to 2nd batch of Wash #1. Wash in a closed plasticware container (or for a single slide use a 50-mL tube) on shaker 20 min at 100 rpm. Cover so that the slide is not exposed to light. Transfer the slide to a container of Wash #2 and shake in the dark for 20 min. Transfer the slide to a container of Wash #3 and shake in the dark for 20 min. Remove the slide from the wash and dry it with a short spin in slide minicentrifuge or place the slide in a 50-mL tube supported by tissue paper in the bottom (do not let it touch the slide) and spin at 1000 rpm for 1 min in a table top centrifuge. Wash recipes: Wash #1 (1 L per use) Wash #2 (0.5 L per use) Wash #2 (0.5 L per use)
50 mL 20 SSC 10 mL 10% SDS up to 1 L in dH2O 50 mL 20 SSC up to 1 L in dH2O 5 mL 20 SSC up to 1 L in dH2O
After washing, store the slide in an ozone free room or desiccator in the dark until scanning. Slides are relatively stable at this point and can be stored or rescanned if necessary.
REFERENCES Bodrossy, L., and Sessitsch, A. (2004). Oligonucleotide microarrays in microbial diagnostics. Curr. Opin. Microbiol. 7, 245–254. Bodrossy, L., Stralis-Pavese, N., Murrell, J. C., Radajewski, S., Weilharter, A., and Sessitsch, A. (2003). Development and validation of a diagnostic microbial microarray for methanotrophs. Environ. Microbiol. 5, 566–582. Bodrossy, L., Stralis-Pavese, N., Donrad-Koszler, M., Weilharter, A., Reichenauer, T. G., Schofer, D., and Sessitsch, A. (2006). mRNA-based parallel-detection of active methanotroph populations by use of a diagnostic microarray. Appl. Environ. Microbiol. 72, 1672–1676.
Author's personal copy Functional Gene Microarrays for AOB and AOA
395
Bouskill, N. J., Eveillard, D., O’Mullan, G. D., Jackson, G. A., and Ward, B. B. (2010). Seasonal and annual reoccurrence in ammonia-oxidizing bacterial population structure. Environ. Microbiol. Doi: 10.1111/j.1462-2920.2010.02362.x. Bouskill, N. J., Eveillard, D., Chien, D. M., Jayakumar, A., and Ward, B. B. (submitted). Distribution and abundance of ammonia-oxidizing organisms across environmental gradients. Bulow, S. E., Francis, C. A., Jackson, G. A., and Ward, B. B. (2008). Sediment denitrifier community composition and nirS gene expression investigated with functional gene microarrays. Environ. Microbiol. 10, 3057–3069. Dai, H. Y., Meyer, M., Stepaniants, S., Ziman, M., and Stoughton, R. (2002). Use of hybridization kinetics for differentiating specific from non-specific binding to oligonucleotide microarrays. Nucleic Acids Res. 30, e86. DeRisi, J. L., Iyer, V. R., and Brown, P. O. (1997). Exploring the metabolic and genetic control of gene expression on a genomic scale. Science 278, 680–686. DeSantis, T. Z., Brodie, E. L., Moberg, J. P., Zubieta, I. X., Piceno, Y. M., and Andersen, G. L. (2007). High-density universal 16S rRNA microarray analysis reveals broader diversity than typical clone library when sampling the environment. Microb. Ecol. 53, 371–383. DiRisi, J. L., Iyer, V. R., and Brown, P. O. (1997). Exploring the metabolic and genetic control of gene expression on a genomic scale. Science 278, 680–686. Francis, C., Roberts, K., Beman, J., Santoro, A., and Oakley, B. (2005). Ubiquity and diversity of ammonia-oxidizing archaea in water columns and sediments of the ocean. Proc. Natl. Acad. Sci. USA 102, 14683–14688. He, Z., Wu, L., Li, X., Fields, M., and Zhou, J. (2005). Empirical establishment of oligonucleotide probe design criteria. Appl. Environ. Microb. 71, 3753–3760. He, Z., Gentry, T. J., Schadt, C. W., Wu, L., Liebich, J., Chong, S. C., Huang, Z., Wu, W., Gu, B., Jardine, P., Criddle, C., and Zhou, J. (2007). GeoChip: A comprehensive microarray for investigating biogeochemical, ecological and environmental processes. ISME J. 1, 67–77. He, Z., Deng, Y., Van Nostrand, J. D., Tu, Q., Xu, M., Hemme, C. L., Li, X., Wu, L., Gentry, T. J., Yin, Y., Liebich, J., Hazen, T. C., et al. (2010). GeoCHip 3.0 as a highthroughput tool for analyzing microbial comunity composition, structure and functional activity. ISME J. 4, 1167–1179. Held, G. A., Grinstein, G., and Tu, Y. (2003). Modeling of DNA microarray data by using physical properties of hybridization. PNAS 100, 7575–7580. Koops, H.-P., Purkhold, U., Pommerening-Roser, A., Timmermann, G., and Wagner, M. (2003). The lithoautotrophic ammonia-oxidizing bacteria. In “The Prokaryotes: An Evolving Electronic Resource for the Microbiological Community,” (M. Dworkin, ed.), 3rd edn. Springer-Verlag, New York. Loy, A., Lehner, A., Lee, N., Adamczyk, J., Meier, H., Ernst, J., Schleifer, K. H., and Wagner, M. (2002). Oligonucleotide microarray for 16S rRNA gene-based detection of all recognized lineages of sulfate-reducing prokaryotes in the environment. Appl. Environ. Microbiol. 68, 5064–5081. Marcelino, L. A., Backman, V., Donaldson, A., Steadman, C., Thompson, R. J., Preheim, S. P., Lien, C., Lim, E., Veneziano, D., and Polz, M. F. (2006). Accurately quantifying low-abundant targets amid similar sequences by revealing hidden correlations in oligonucleotide microarray data. PNAS 103, 13629–13634. Martens-Habbena, W., Berube, P. M., Urakawa, H., de la Torre, J. R., and Stahl, D. A. (2009). Ammonia oxidation kinetics determine niche separation of nitrifying Archaea and Bacteria. Nature 461, 976–982.
Author's personal copy 396
B. B. Ward and N. J. Bouskill
Moisander, P. H., Morrison, A. E., Ward, B. B., Jenkins, B. D., and Zehr, J. P. (2007). Spatial-temporal variability in diazotroph assemblages in Chesapeake Bay using an oligonucleotide nifH microarray. Environ. Microbiol. 9, 1823–1835. Pozhitkov, A., Noble, P. A., Domazet-Loso, T., Nolte, A. W., Sonnenberg, R., Stehler, P., Geier, M., and Tautz, D. (2006). Tests of rRNA hybridization to microarrays suggest that hybridization characteristics of oligonucleotide probes for species discrimination cannot be predicted. Nucleic Acids Res. 34, e66. Prosser, J. I., and Nicol, G. W. (2008). Relative contributions of archaea and bacteria to aerobic ammonia oxidation in the environment. Environ. Microbiol. 10, 2931–2941. Purkhold, U., Pommerening-Roser, A., Juretschko, S., Schmid, M. C., Koops, H.-P., and Wagner, M. (2000). Phylogeny of all recognized species of ammonia oxidizers based on comparative 16S and amoA sequence analysis: Implications for molecular diversity surveys. Appl. Environ. Microbiol. 66, 5368–5382. Rotthauwe, J.-H., Witzel, K.-P., and Liesack, W. (1997). The ammonia monooxygenase structural gene amoA as a functional marker: Molecular fine-scale analysis of natural ammonia-oxidizing populations. Appl. Environ. Microbiol. 63, 4704–4712. Santoro, A. E., Francis, C. A., de Sieyes, N. R., and Boehm, A. B. (2008). Shifts in the relative abundance of ammonia-oxidizing bacteria and archaea across physicochemical gradients in a subterranean estuary. Environ. Microbiol. 10, 1068–1079. Schloss, P. D., and Handlesman, J. (2005). Introducing DOTUR, a computer program for defining operational taxonomic units and estimating species richness. Appl. Environ. Microbiol. 71, 1501–1506. Taroncher-Oldenburg, G., Griner, E., Francis, C. A., and Ward, B. B. (2003). Oligonucleotide microarray for the study of functional gene diversity of the nitrogen cycle in the environment. Appl. Environ. Microbiol. 69, 1159–1171. Treusch, A. H., Leininger, S., Kletzin, A., Schuster, S. C., Klenk, H.-P., and Schleper, C. (2005). Novel genes for nitrite reductase and Amo-related proteins indicate a role of uncultivated mesophilic crenarchaeota in nitrogen cycling. Environ. Microbiol. 7, 1985–1995. Venter, J., Remington, K., Heidelberg, J., Halpern, A., Rusch, D., Eisen, J., Wu, D., Paulsen, I., Nelson, K., Nelson, W., Fouts, D., Levy, S., et al. (2004). Environmental genome shotgun sequencing of the Sargasso Sea. Science 304, 66–74. Ward, B. B., Eveillard, D., Kirshtein, J. D., Nelson, J. D., Voytek, M. A., and Jackson, G. A. (2007). Ammonia-oxidizing bacterial community composition in estuarine and oceanic environments assessed using a functional gene microarray. Environ. Microbiol. 9, 2522–2538. Wu, L. Y., Liu, X., Schadt, C. W., and Zhou, J. Z. (2006). Microarray-based analysis of subnanogram quantities of microbial community DNAs by using whole-community genome amplification. Appl. Environ. Microbiol. 72, 4931–4941. Wuchter, C., Abbas, B., Coolen, M. J. L., Herfort, L., van Bleijswijk, J., Timmers, P., Strous, M., Teira, E., Herndl, G. J., Middelburg, J. J., Schouten, S., and Damste, J. S. S. (2006). Archaeal nitrification in the ocean. PNAS 103, 12317–12322. Zhang, L., Hurek, T., and Reinhold-Hurek, B. (2007). A nifH-based oligonucleotide microarray for functional diagnostics of nitrogen-fixing microorganisms. Microb. Ecol. 53, 456–470.
