Samples were collected from 4 different classrooms where children ages 0–4 years were taught daily. Environmental samples were taken with dual tip sterile cotton swabs (BBL CultureSwab™, catalog # 220135, Becton Dickinson, Sparks, MD) and these were stored in sterile-labeled tubes for immediate transport back to the lab. Toys and surfaces were sampled on a fixed surface area of approximately 13 cm2. During the course of the study the same furniture surfaces were sampled repeatedly, while the toys varied between samplings.
Bacterial Culturing Methods
Immediately upon return to the lab, one tip of the dual-tip swab samples was placed into 7 mL of nutrient broth (Difco™, Becton Dickinson, Sparks, MD) and allowed to incubate overnight at 37°C. The second tip of each dual-tip swabs was labeled and placed at -80°C for later analysis. Overnight cultures of nutrient broth (Difco™) were used to streak 5% blood (Blood Agar Contact Plate, Hardy Diagnostics, Santa Maria, CA) and nutrient agar plates (Difco™, Becton Dickinson, Sparks, MD), which were also incubated overnight at 37°C. To minimize outside contamination, all culturing was performed in a biological hood using sterile instruments. The following day, plate growth and colony morphology were evaluated and recorded.
DNA Extraction and PCR Amplification of Colonies and Swabs
DNA was extracted using a lysozyme-extraction protocol  directly from the bacterial colonies picked off plates using a sterile toothpick. One colony was selected from each observed morphology type. We used the same protocol to isolate DNA directly from the swabs (the environmental samples) for culture-independent analysis.
For the environmental extractions, cotton from the swab samples was removed using a sterile razor blade and placed into the lysozyme reaction mixture. The reaction mixture had a total volume of 200 μl and included the following final concentration: 20 M Tris, 2 mM EDTA (pH 8.0), 1.2% P40 detergent, 20 mg ml-1 lysozyme, and 0.2 μm filtered sterile water (Sigma Chemical Co., St. Louis, MO). We used the same reaction buffer and extraction method for isolating DNA from the cultured organisms (Table 1). For the cultured bacteria, we used a sterile toothpick to pick a single colony from the agar plates, which was then swirled into the reaction mixture. Samples were incubated in a 37°C water bath for thirty minutes. Next, Proteinase K (DNeasy Tissue Kit, Qiagen Corporation, Valencia, CA) and AL Buffer (DNeasy Tissue Kit, Qiagen Corporation, Valencia, CA) were added to the tubes and gently mixed. Samples were incubated in a 70°C water bath for 10 min. All samples were subjected to purification using the DNeasy Tissue Kit.
Following extraction, the DNA was quantified using a NanoDrop ND-1000 Spectrophomtometer (NanoDrop Technologies, Willmington, DE). We created PCR-based clone libraries from nine of the swabs collected over the course of the study. With one exception, swabs we chose came from the same surface type (the toy shelf). Six of the swabs were selected to represent each month from October 2005 through April 2006, while the other three were selected as duplicates for three of the sampling days to determine the consistency of contamination across surfaces. This strategy allowed us to find the most consistently abundant types of bacteria and to detect any significant changes in diversity that might occur during the "cold and flu" season.
The PCR reactions utilized published bacterium specific primers primers 8F (5'-AGAGTTTGATCCTGGCTCAG-3') and 805R (5'-GACTACCAGGGTATCTAATCC-3') to amplify the 16S rRNA gene. The ~800 bp PCR products from amplification using these primers includes a portion of the 16S rRNA gene that has been shown to be particularly useful for database analysis and identification of bacterial sequences . PCR was carried out in a total reaction volume of 50 μl including 1 μl (approx. 10 ng μl-1) of sample DNA as template, each deoxynucleoside triphosphate at 200 μM, 1.5 mM MgCl2 in 10× buffer (10× concentration: 500 mM 1 M KCl, 100 mM 1 M Tris HCl pH 8.4, 1% Triton-X, 15 mM MgCl2), each primer at 0.4 μM, 4 μl of bovine serum albumin (10 mg ml-1), and 0.5 μl of REDTAQ™ DNA polymerase (1 unit μl-1; Sigma-Aldrich Inc., St. Louis, MO). Between twenty-five and thirty cycles of PCR amplification were performed for the environmental swab samples and the bacterial colony samples. We used the lowest numbers of cycles that yielded a visible band on an agarose gel in order to minimize over-amplification of rare sequences and production of chimeric sequences. All PCR cycles included an initial denaturation step at 94°C for 1 min, an annealing step at 55°C for 45 sec and an extension step at 72°C for 1.5 min. The amplification cycles were preceded by a one-time denaturing step at 94°C for 5 min prior to the first cycle and included a final 72°C extension for 10 min to ensure complete extension for efficient cloning. Products were cleaned using Qiagen's QIAquick PCR Purification Kit.
Cloning and Sequencing
The cleaned PCR products were cloned using the TOPO TA Cloning Kit for Sequencing (Invitrogen™, Carlsbad, CA) according to the manufacturer's instructions. Transformed One Shot chemically competent E.coli cells were plated on LB-agar plates containing 50 μg ml-1 ampicillin and top plated with X-gal and IPTG. Next, colonies with inserts were randomly selected with a sterile toothpick and grown overnight at 37°C in 150 μl of LB broth (Fisher Biotech, Fair Lawn, NJ) containing 6% glycerol, and 1 μM ampicillin in a 96-well plate. Subsequent to cloning, a PCR amplification was performed on each of the 96 wells. The universal bacterial primers M13F (5'TTATGTAAAACGACGGCCAGT) and M13R (5'GGAAACAGCTATGACCATG) were used. Sequencing of PCR products was completed by the San Diego State MicroChemical Core Facility using an ABI 377 DNA sequencer.
Database and Phylogenetic Analyses
The sequence chromatogram files were imported and analyzed using XplorSeq 2.0, a program written by Dr. Dan Frank at the University of Colorado (unpublished). XplorSeq imports chromatograms and determines the quality of the sequence using automatic base calling software . The program also processes a batch BLAST through the NCBI database and outputs files that can easily be transferred to Microsoft Excel or the sequences exported as text files in the Fasta format.
We used the Fastgroup II program , to trim the 3' end of all the cleaned sequences, count replicate sequences for determining abundance of clones in libraries, and to identify a single representative for further alignment and phylogenetic analysis. The count data also allowed us to estimate the sequence coverage for each library. Coverage (C) was calculated using the following equation,
C = 1 - n/N
, where n is the number of unique OTU sequences observed and N is the total number of OTUs (i.e., sum of unique OTUs plus OTUs observed more than once) .
After Fastgroup analysis, the sequences were aligned using the NAST alignment software . This program aligns rRNA gene sequences to a diverse set of full-length rRNA gene sequences that have been rigorously aligned using the RNA secondary structure. From here the aligned sequences were imported into the ARB application . Bacterial colony species identifications were completed using a combination of BLAST results and phylogenetic analysis using the ARB program. Clone library sequences and other sequences identified in GenBank were aligned using ARB and exported as Nexus files for phylogenetic analysis using PAUP*  and MrBayes version 3.1.2 .
Phylogenetic analyses were performed with two different data sets (see Fig. 1 and Fig. 2). For each of the data sets, trees were constructed using three different methods: Bayesian, Maximum Parsimony (MP) and Maximum Likelihood (ML). The MODELTEST program  was used to choose the DNA substitution model that best fit our particular dataset. Bayesian analyses were performed using the General Time Reversible model  with a gamma-distributed among-site substitution rate heterogeneity and a fraction of sites constrained to be invariable (GTR+I+G).
All Bayesian analyses were done with four independent Markov chains run for 3,000,000 MCMC generations. Trees were sampled every 200 generations with a burn-in of 2000 trees. The best Maximum Parsimony (MP) tree, or set of trees, was found through a random addition sequence heuristic search strategy with 100 replicates. The maximum number of trees kept during each search was capped at 1000. For the MP bootstrap analyses, we performed MP searches on 100 bootstrap replicated datasets using the same heuristic search strategy except with 10, rather than 100, search replicates. We also performed a Maximum Likelihood (ML) analysis using the GTR+I+G model of evolution and a random addition sequence heuristic search strategy with 10 replicates to find the highest likelihood tree.