Accurately quantifying low-abundant targets amid similar sequences by revealing hidden correlations in oligonucleotide microarray data

Marcelino et al. 10.1073/pnas.0601476103.

Supporting Text

File in this Data Supplement:

Supporting Text




Supporting Text

Experimental procedures

Bacterial strains, environmental samples, and RNA extraction

Bacterial cell cultures were obtained from American Type Culture Collection (ATCC) and grown according to the specified protocol [Escherichia coli K12 ATCC 10798, Vibrio cholerae N16961 El Tor ATCC 39315, Vibrio vulnificus ATCC 27562, Vibrio alginolyticus ATCC 19108, Vibrio anguillarum (ATCC 14181), and Vibrio splendidus (ATCC 33125), Vibrio fischeri (strain MJ1)]. Total RNA was isolated from 100 ml of actively growing bacterial cultures (»5 ´ 107 cells/ml) using the RNAeasy extraction kit (Qiagen).

To create artificial RNA communities, varying amounts of total RNA of different bacteria were mixed with total RNA of Saccharomyces cerevisiae (Sigma), tRNA fractions of bovine liver (Sigma), and tRNA fractions of wheat germ (Sigma). In the particular case of V. cholera, we added total RNA of V. cholera at different proportions (0.05%, 0.1%, and 1%) to various amounts of total RNA of an artificial RNA community (0.5, 2.5, and 12.5 mg of an artificial RNA community).

For experiments using natural communities, water samples for RNA extraction were collected monthly during March through August 2004 at high tide from the marine end of the Plum Island Sound estuary (Northeastern Massachusetts) (1, 2). Total bacteria ranged between 5.83 ´105 (March 2004) and 2.53 ´ 106 (July 2004) cells per milliliter of seawater as measured by microscopic count of 4,6-diamidino-2-phenylindole (DAPI) stained cells. Two-liter samples were prefiltered through a 60-mm nylon mesh screen (Spectrum) and a 2.7-mm filter (GD/F, Millipore), followed by concentration onto 0.22-mm Sterivex filters (Millipore). Filters were stored at -20°C until total RNA was extracted by bead-beating, as described in ref. 3. Final RNA purification was by application of the RNAeasy extraction kit (Qiagen). Quantification of total RNA in each sample was performed by using Ribogreen RNA Quantitation Kit R-11490 (Molecular Probes) on a fluorescent plate reader (Synergy HT).

Template labeling with Cy3 dye

Three types of samples were labeled: (i) artificial RNA communities where total RNA of known bacteria was added to a variable background of nonbacterial RNA at different proportions, (ii) total RNA extracts from coastal seawater spiked with a known amount of total RNA of V. cholerae, and (iii) total RNA extracts from unamended coastal seawater samples obtained from different months to identify naturally occurring bacterial taxa.

Each sample was spiked with 10 different mRNAs from the plant Arabidopsis thaliana (Stratagene) at concentrations ranging from 0.004 to 4 ng. These served as positive controls for normalization of all microarray experiments to allow comparison across experiments.

Each mixture was performed in duplicate or triplicate, labeled, and hybridized to individual arrays.

RNA mixtures were labeled with 1.25 mM FluoroLink Cy3-dCTP (Amersham Pharmacia) in the presence of 100 mM of random 9-mer primer (MWGbiotech), 100 mM each 789R bacterial 16S rRNA primer and 636R bacterial 23S rRNA primer (www.psb.ugent.be/rRNA/index.html), 2.5 mM of dATP, dGTP, and dTTP, and 1.25 mM dCTP, and 200 units of Superscript II Reverse Transcriptase (Life Technology, Rockville, MD) in a final volume of 30 ml, for every 2.5 mg of total RNA. Each mixture was incubated at 42°C for 1.5 h to synthesize first-strand cDNA. Reactions were terminated by adding 3 ml of 2.5 M NaOH and incubated for 15 min at 37°C followed by addition of 15 ml of 2 M Hepes-free acid (pH 5.5) to neutralize the reaction. Each reaction was purified by CyScribe GFX-purification columns (Amersham Pharmacia), and the labeled first-strand cDNA was eluted in 120 ml of elution buffer. Each purified and labeled total RNA mixture was then dried on a speed vacuum and resuspended in 12 ml of Pronto! hybridization buffer (Corning), deposited onto one array of 18 ´ 18 mm covered with a coverslip and mounted onto an hybridization chamber (Corning) according to Corning manufacturer.

Microarray construction

Probe design. The oligonucleotide-based array contained 139 probes targeting 23S and/or 16S rRNA sequences of 50 different bacteria obtained from GenBank (www.ncbi.nlm.nih.gov) and RDP databases (http://rdp.cme.msu.edu), as well as from sequences of Vibrio environmental isolates from the same site (2) and sequences from a 16S clone library constructed from the same sample site (1). To identify potential probe regions, sequences were aligned using the Genetics Computer Group Wisconsin package software (Accelrys) for sequence alignment and determination of distance matrices for all probes. Oligonucleotide properties calculator (www.basic.northwestern.edu/biotools/oligocalc.html) was also used to check for self-complementarity and melting temperature. The selected oligonucleotide sequences were compared against the entire GenBank database using BLAST, and unique sequences were used for array printing. Probes targeting human, mouse, and rat genes were used as negative controls and designed as described above, to determine background hybridization signal for the methodology. For a list of probe details, see National Center for Biotechnology Information (NCBI) Probe Database PUIDs 6103259-6103399.

Microarray design and fabrication. All oligonucleotide probes (Operon Technologies) were adjusted to a concentration of 50 mM in 3 SSC (3 M NaCl and 0.3 M NaB3-citrate ´ 2H2O) and 1.5 M betain (Sigma) and spotted in five replicate spots on CMT-GAPS amino silane-coated glass slides (Corning, Inc.), using a MicroGrid (BioRobotics) arrayer. After printing, the slides were exposed to 625 mJ of UV irradiation and stored at room temperature.

Microarray hybridization. All arrays were prehybridized and hybridized with Pronto! Microarray Hybridization Kit (Corning) according to instructions, at 42°C. After 18 h of hybridization, the arrays were washed using Pronto! Microarray washing solutions at 60°C. The arrays were scanned using ArrayWoRx scanner (Applied Precision), and raw fluorescence data were acquired using Image Analysis software from MolecularWare (Digital Genome). The mean fluorescence of each spot (average of 70 pixel measurements) for the Cy3 channel, together with the mean fluorescence of the local background surrounding each spot (average of 200 pixel measurements) was used for further processing.

Microarray data processing before analysis by the methodology.

1. Filtering out spots with high local background. Spots with background higher than two standard deviations above the average background for all spots in the entire array were removed from the data set. This ensured that spots containing fluorescence due to inconsistencies such as random dye spots, scratches, and "comet effects" were not included in the data set.

2. Normalizing hybridization intensities to compare fluorescence across experiments. Variability in signal intensity within and between replicate arrays was normalized to compare hybridization results across microarray experiments. Variability in signal intensity can result from differences in probe printing, labeling efficiency, target quality and quantity, scanner settings, and hybridization conditions, among other sources. The raw hybridization signal of each spot was normalized by the average of the five most intense spots for control genes of the plant A. thaliana (Stratagene).

3. Testing consistency among replicate experiments. A regression analysis was run in a pairwise manner among replicate experiments, and the linear regression coefficients were then used to transform the data sets. We applied a combination of two tests, the t test and variance of the means (VOM), to determine whether replicate experiments can be considered consistent and therefore used for further processing. Each test is run for each combination probe j and bacteria k (»5-20 replicate spots per data set). Data sets were considered consistent if for each combination jk, P value ³0.01 and VOM = 0.2 (

, where M(1) and M(2) are the means for experiments 1 and 2, respectively).

4. Outlier Removal. Signal intensities outside three standard deviations of the mean for a given probe j and bacteria k pair were considered outliers and removed from the data set.

RNA abundances measured with RT-quantitative PCR (QPCR)

The RNA abundance of Vibrio splendidus in environmental samples was validated by reverse transcriptase QPCR by synthesizing the first-strand cDNA as described above without the addition of cye-dye. QPCR was carried out with Vibrio group-specific16S rRNA primers using a competitive internal standard and separation and quantification of the resulting amplicons by constant denaturing capillary electrophoresis (CDCE) as described in ref. 3.

1. Acinas, S. G., Klepac-Ceraj, V., Hunt, D. E., Pharino, C., Ceraj, I., Distel, D. L. & Polz, M. F. (2004) Nature 430, 551-554.

2. Thompson, J. R., Pacocha, S., Pharino, C., Klepac-Ceraj, V., Hunt, D. E., Benoit, J., Sarma-Rupavtarm, R., Distel, D. L. & Polz, M. F. (2005) Science 307, 1311-1313.

3. Thompson, J. R., Randa, M. A., Marcelino, L. A., Tomita-Mitchell, A., Lim, E. & Polz, M. F. (2004) Appl. Environ. Microbiol. 70, 4103-4110.

This Article

  1. Published online before print September 1, 2006, doi: 10.1073/pnas.0601476103 PNAS September 12, 2006 vol. 103 no. 37 13629-13634
  1. OA Abstract
  2. Figures Only
  3. OA Full Text
  4. Full Text (PDF)
  5. » Supporting Text
About Our New Site Design >>
Link: Advanced Search

This Week's Issue

  1. July 1, 2008, 105 (26)
  1. Current Issue
  1. Alert me to new issues of PNAS

Most