Bacterial growth conditions
For all experiments Rhodopirellula baltica
cells were grown as chemostat cultures in a mineral medium containing 10 mM glucose as the sole carbon source and 1 mM ammonium chloride as a nitrogen source at 28°C [20
]. Chemostat (Ø 13.5 cm × 25 cm, 1 l, Schott, modified by Ochs, Bovenden) parameters used were: pH 7.4, average dilution rate 0.75 ml/min and pO2
around 100%. The cultures had an OD600 nm
of 0.5 - 0.6 (corresponding to log phase). The cells were harvested after 5 dwell times.
Sample collection, cell lyses, RNA Isolation and cDNA synthesis
After harvesting the R. baltica cultures, an aliquot was collected to serve as the time-zero reference. The culture broth was collected in 500 ml tubes and swirled briefly in an ethanol-dry ice bath to rapidly cool the cultures and prevent shifts in the RNA profile. Subsequently, the broth was centrifuged at 6000 rpm for 20 min at 4°C (Beckman Coulter™ AvantiTM626 J-20XP, JA10 Rotor). The pellets were re-suspended in 0.1 M Tris-HCL and then re-centrifuged. Cell pellets were shock-frozen in liquid nitrogen and stored at -80°C. Total RNA was isolated using the protocol of the TRI Reagent® Kit by Ambion (Austin, USA). The purity and quality of the extracted total RNA was checked with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palto Alto, USA) and gel electrophoresis. cDNA synthesis was performed using the SuperScript direct cDNA labeling kit by Invitrogen (Karlsruhe, Germany) according to the manufacturer's instructions with random hexamers and unlabeled dCTP/dUTP, followed by a three hour reverse transcription incubation step at 46°C. The RT reaction was halted by incubation for 3 min at 95°C. To hydrolyze the RNA, 0.1 M NaOH was added, incubated at 65°C for 15 min and neutralized with 0.1 M HCL. The remaining cDNA was precipitated overnight at -20°C and the pellet washed with 70% Ethanol.
cDNA was directly labeled using the Platinum Bright™ nucleic acid labeling kit based on KREATECH's patented Universal Linkage System (ULS) (Biocat, Heidelberg, Germany) according to the manufacture's protocol.
Concentrations of RNA and cDNA were measured, and incorporation of the dyes Alexa 546 and Alexa 647 were checked using a Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, USA).
Experimental design and sample preparation
In three independent hybridizations conducted for each experiment and time point, the expression profiles of cells that had undergone stress were compared with those of cells at time zero. That is, the array analysis of each Alexa 647 labeled sample was compared with those of Alexa 546 labeled time-zero samples. The data shown are based on the analysis of all three replicates performed for each of the conditions.
Samples for expression profiling and microscopic analysis were collected at 10, 20, 40, 60 and 300 min in all three stress experiments.
Heat shock from 28°C to 37°C
Cells grown continuously at 28°C were collected by centrifugation. An aliquot was removed for RNA extraction and taken as the time zero reference for the heat, cold and salt stress experiments. Aliquots were re-suspended in an equal volume of 37°C medium and returned to 37°C for cultivation.
Cold shock from 28°C to 6°C
Cells grown continuously at 28°C were collected by centrifugation, re-suspended in an equal volume of 6°C medium and returned to 6°C for cultivation.
Salt stress from 17.5‰ to 59.5‰ salinity
Similar to the heat and cold shock experiments, an R. baltica culture was grown in mineral media with 17.5‰ salinity. Cells were harvested and aliquots were transferred to a mineral media with a salinity of 59.5‰.
Whole Genome Array construction, hybridization and image analysis
The whole-genome oligonucleotides for R. baltica SH1T (Pirellula AROS 630 Version 1.0) were purchased from Operon (Cologne, Germany) and diluted to 20 μM concentration in Micro Spotting Solution Plus spotting buffer (Telechem, Sunnyvale, USA). Spotting was done with three replicates per gene, per slide onto GAPS II aminosilane slides (Corning, Schiphol-Rijk, Netherlands) using a SpotArray 24 spotting device (Perkin Elmer, Wellesley, USA) together with 48 Telechem Stealth Pins (Telechem, Sunnyvale, USA). The arrays were subsequently exposed at 245 nm and 360 mJ in the GS Gene Linker (Bio-Rad, München, Germany), followed by incubation at 80°C for at least 3 h. Slides were stored at room temperature in the dark until use.
Blocking, denaturing, hybridization, washing and N2 drying procedures were carried out in an automated hybridization station HS400 (Tecan, Crailsheim, Germany). The spotted arrays were blocked in prehybridization solution containing 250 mM NaCl, 5 mM Tris/HCl at pH 8.0, 50% formamide, 0.5 × SSC, 0.05% BSA, and 1% blocking reagent from Roche Diagnostics, Mannheim, Germany for 45 min at 52°C. For hybridization at least 2 μg of Alexa 546 dye-labeled and 2 μg of Alexa 647 dye-labeled total cDNA were combined and taken up in a final volume of 100 μl DIG Easy Hyb hybridization solution (Roche Diagnostics, Mannheim, Germany). After the blocking step, the sample solution was applied to the arrays, denatured at 95°C for 3 min and hybridized under stringent conditions at 52°C for over 12 hours. After hybridization slides were washed at room temperature in ULTRArray Low Stringency Wash Buffer (Ambion, Austin, USA) and dried by N2.
Signal detection and data analysis
Slides were scanned at a resolution of 5 μm using a ScanArray Express Microarray scanner (Perkin Elmer, Wellesley, USA) with varied laser power and photomultiplier tube (PMT sensitivity) for each slide. The accompanying image analysis software, ScanArray Express Version 4.0, was used for automatic spot detection and signal quantification of both fluorophores. Raw data were automatically processed using the microarray data analysis software tool MADA http://www.megx.net/mada
, developed in-house. Firstly, the spot intensities were corrected for local background (mean spot intensity minus mean spot background intensity). Signals were only assessed as positive if mean spot pixel intensity was higher than the mean local background intensity plus twice the standard deviation of the mean local background pixel intensity. Each gene is spotted in three replicates. Spot replicates with poor quality were removed from the data set according to MADA's outlier test results. This test first computes the standard deviation of all replicates. Secondly, one replicate is omitted and the standard deviation is recalculated; if the deviation differs more than 50% from the previous deviation, the omitted replicate is regarded as an outlier. This procedure is repeated for all replicates
Expression is described through the ratio and intensity, where R is the fluorescence log ratio of the experiment time point relative to the control condition (e.g. R = log2 (result of channel 10 min/result of channel control/reference)) and I is the log mean fluorescence intensity (e.g. I = log10 (result of channel 10 min × result of channel control/reference)).
Each data point represents a regulation factor (ratio) in a logarithmic scale for one gene calculated from the positive replicates for a particular probe coming from two RNA pools (reference and sample). Normalization was carried out by LOWESS fitting on an R-versus-I plot with a smoothing factor of 0.5. Each time point of the time-series experiment was hybridized independently three times. The expression data (ratio) of the three hybridizations were combined to one expression data point (ratio) by averaging and the standard deviation of the average value was calculated. Only ratios with a standard deviation less than 25% were regarded as genes that are regulated. Differentially expressed genes were detected by a fixed threshold cut off method (i.e. a two-fold increase or decrease) based on the results of self-self hybridization. Using the same biological sample, the reference (untreated sample) is labeled twice, once with Alexa 546 and once with Alexa 647, and the variability between the two sets of measurements is calculated to estimate the experimental noise. Ideally, there should not be any variability and all expression points should have a ratio close to zero. In reality, however, this is never the case and thresholds based on the distribution of these data along the y-axis were defined for the further experiments.
Consequently, R. baltica genes detected with intensities resulting in ratios above or below these thresholds can beregarded as up- or down-regulated.
Differentially expressed genes present in the complete time course profile (10, 20, 40, 60 und 300 min) for all three experiments were clustered using the k-means clustering approach (Euclidean distance metric, k = 30 clusters and 49 (max. 500) iterations) [58
] with the software tool Multiexperiment Viewer MeV Version 4.0.2 from the TM4 microarray software suite [59
]. Briefly, the clustering algorithm arranges genes into a given number of clusters, k, according to similarity in their expression profiles across the entire array experiments, such that genes with similar expression patterns are clustered together. The data are displayed in tabular format where each row of colored boxes stores the variation in transcript abundance for each given gene and each column stores the variation in transcript levels of every gene in a given mRNA sample, as detected on one array. The variations in transcript abundance for each gene are depicted by means of a color scale, in which shades of red represents increases and shades of green represent decrease in mRNA levels, relative to the unstressed culture, and the saturation of the color corresponds to the magnitude of the differences. Black coloration indicates no change in transcript level while grey represents missing data.
The genome of Rhodopirellula baltica
was automatically re-annotated based on updated homology searches (June 2005 - MicHanThi [60
]). The updated annotation including all tool results is publicly available at http://gendb.mpi-bremen.de/gendb/BX119912
]. JCoast [62
] was used as a tool for the visualization, interpretation, COG-assignment statistics and comparison of genomic data stored in GenDB V2.2 [63
]. The Venn diagrams were generated by BioVenn [64
Each microarray used in this study contained 7325 known or predicted R. baltica
genes according to Glöckner et al
]. A detailed description of the array can be found at the NCBI's Gene Expression Omnibus (GEO) database under accession number GPL7654. The complete microarray datasets covering the expression of R. baltica
cultures exposed to heat, cold and high salinity, are public available in the GEO repository http://www.ncbi.nlm.nih.gov/geo/
under accession numbers GSE13769, G SE13856 and GSE14075 [65