Protein bands 1, 2a and 3 were in-gel digested as described previously (Shevchenko et al., 1996) with minor modifications. Briefly, the gel spots were distained with 200 mM NH4HCO3 in 50% acetonitrile, washed 4 times with water, shrunk with 100% acetonitrile and dried in a Speed-Vac (Jouan Model 1022, St-Herblain, France). Trypsin solution was prepared in 50 mM NH4HCO3 and 5 mM CaCl2 in 10% acetonitrile to 20 ng of trypsin per microlitre. An adequate volume of trypsin solution was added to cover the gel spot, incubated on ice for 1 h then incubated at 37°C for 13 h. After digestion, the gel spots were sonicated 15 min with 20 mM NH4HCO3 once, followed by 5% formic acid in 50% acetonitrile three times. The combined extracts were concentrated by Speed-Vac, diluted to about 10% acetonitrile and concentrated again to about 15 µl, followed by C18 Zip-Tip (Millipore, Billerica, MA, USA) purification according to the manufacturer's protocol. The peptide solutions, after Zip-Tip cleaning, were concentrated by Speed-Vac, diluted to about 10% acetonitrile and concentrated again to about 20 µl. The resultant sample (3 µl) was loaded pneumatically into a 75 µm i.d. × 360 µm o.d. capillary HPLC column packed with 10.2 cm of 5 µm AQUA C18 (Phenomenex, Torrance, CA, USA). LC/MS analysis was performed using an LTQ mass spectrometer (Thermo Electron Corp. San Jose, CA, USA), Magic 2002 HPLC (Michrom BioResouces, Auburn, CA, USA) and an in-house built micro-ESI interface. Solvent A was 0.5% acetic acid in water, B, 0.5% acetic acid in acetonitrile. The gradient started at 5% B, held 5 min, increased to 10% B in 1 min and held for 4 min, to 40% B in 45 min, to 80% B in 1 min and held for 9 min, then down to 5% B in 5 min and held for 10 min. The flow rate was about 0.3 µl min−1. The MS1 scan range was 400–2000 m/z units, each MS1 scan was followed by 10 MS2 scans for the 10 most intense ions in the MS1 scan. Default parameters under the Xcalibur 1.4 data acquisition software (Thermo) were used, with the exception of an isolation width of 3.0 m/z units and normalized collision energy of 40%. Raw mass spectral data were searched by SEQUEST (Eng et al., 1994). The ORF database included normal and reversed sequences from microbial genomes from NCBI (Concise Microbial Protein Database, state of curation as it existed on 20 January 2009, 3.5 Gigabytes total size for forward and reversed sequences) or Lake Washington composite metagenome (LWM, 96 Mbytes) (DOE Joint Genome Institute, http://img.jgi.doe.gov/cgi-bin/m/main.cgi) and the human subset of the nrdb (NCBI). The SEQUEST search results were filtered by DTASelect Ver. 1.9 (Tabb et al., 2002), then input into a FileMaker script developed in-house for further processing. The parameters were selected to yield a false-positive rate (Peng et al., 2003) of about 4% based on hits to the reversed LWM sequences (fully tryptic peptides only, Sequest Xcorr thresholds of 1.9, 2.2 and 3.3 for singly, doubly and triply charged peptides respectively). A minimum of two unique peptides was required to identify each protein, following the definition of uniqueness used by DTASelect in which each charge state of a given precursor peptide is considered separately (Tabb et al., 2002). Proteins shown in Table 4 with a single unique peptide, using a more restrictive definition of uniqueness (Xia et al., 2006), were identified with multiple charge states associated with the same amino acid sequence. Only peptides with many repeated observations (spectral counts) typically seen for the primary components of gel spots were retained in the final data set.