SEARCH

SEARCH BY CITATION

Keywords:

  • fast sampling;
  • RNA;
  • Saccharomyces cerevisiae;
  • transcriptomics;
  • microarrays;
  • metabolome;
  • fermentors

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References

Mathematical modelling of cellular processes is crucial for the understanding of the cell or organism as a whole. Genome-wide observations, at the levels of the transcriptome, proteome and metabolome, provide a high coverage of the molecular constituents of the system in study. Time-course experiments are important for gaining insight into a system's dynamics and are needed for mathematical modelling. In time-course experiments it is crucial to use efficient and fast sampling techniques. We evaluated several techniques to sample and process yeast cultures for parallel analysis of the transcriptome and metabolome. The evaluation was made by measuring the quality of the RNA obtained with UV-spectroscopy, capillary electrophoresis and microarray hybridization. The protocol developed involves rapid collection by spraying the sample into −40 °C tricine-buffered methanol (as previously described for yeast metabolome analysis), followed by the separation of cells from the culture medium in low-temperature rapid centrifugation. Removal of the residual methanol is carried out by freeze-drying the pellet at −35 °C. RNA and metabolites can then be extracted from the same freeze-dried sample obtained with this procedure. Copyright © 2007 John Wiley & Sons, Ltd.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References

Systems biology studies require integration of large genomics datasets to generate and validate mathematical models of biological systems. Recent development of approaches for global analysis of mRNA (transcriptomics), proteins (proteomics) and metabolites (metabolomics) made possible the quantitative analysis of a large number of cellular components and provided tools for dynamic metabolic modelling. However, global analytical techniques introduce significant challenges for sample collection and processing, particularly in the case of time-series experiments. A sampling technique should provide efficient inactivation of metabolism at all levels and at the same time should be quick and robust. Time delay in collecting samples for various analyses (i.e. RNA, proteins and metabolites) can result in metabolic changes due to sampling techniques skewing the measurements and introducing additional errors.

For parallel analysis of various cellular components (e.g. mRNA, proteins and metabolites) in S. cerevisiae, multiple samples are usually collected from the same culture or from separate cultures grown under the same conditions (Griffin et al., 2002). The time difference in stopping the metabolism between different samples may introduce additional quantitative errors while correlating the levels of transcripts, proteins and metabolites. For example, there is sufficient evidence suggesting that protein levels do not always correlate with the levels of their corresponding mRNA (Griffin et al., 2002; Gygi et al., 1999; Hatzimanikatis and Lee, 1999; Ideker et al., 2001), although this may be partly due to the difference in sampling techniques and collection time required for RNA, proteins and metabolites analyses.

Several sampling methods are currently used to collect yeast cells for RNA extraction and subsequent microarray analysis. The most common method is the sampling of yeast cultures on ice, followed by the separation of cells from the media by centrifugation, and flush freezing cells in liquid nitrogen either directly or after adding cell lysis buffer (Causton et al., 2001; Chen et al., 2003; Gasch et al., 2000; Manna et al., 1996; Rivas et al., 2001). This sampling technique is incompatible with metabolomics studies where rapid quenching of the cellular metabolism is essential, given that many intracellular metabolites have extremely high turnover rates or are chemically and biologically unstable (Theobald et al., 1993).

Over the years several techniques have been developed for rapid inactivation of the metabolism in S. cerevisiae cells (Castrillo et al., 2003; de Koning and van Dam, 1992; Gonzalez et al., 1997). The most efficient technique to rapidly quench metabolism is to flush freeze the culture in liquid nitrogen or liquid carbon dioxide (Schaefer et al., 1999). Although this method provides an almost instantaneous way to stop all the biochemical reactions, it can not be used when it is necessary to separate the cells from the culture medium, and metabolites are diluted before measurement (Gonzalez et al., 1997). Dropping the culture into 60% (v/v) methanol/water solution at −40 °C provides a suitable alternative. This technique was first described by De Koning and Van Dam (1992) and has been improved since then (Castrillo et al., 2003; Gonzalez et al., 1997). Recently it was reported that the method causes leakage of intracellular metabolites, hence precautions should be taken in order to minimize the losses (Villas-Boas et al., 2005). The efficiency of quenching can be further enhanced by spraying the sample as fine droplets into the quenching solution, which increases the sample-quenching liquid surface interface (Chance et al., 1964).

Although there are reports on simultaneous extraction of metabolites, RNA and proteins from a single plant sample (Weckwerth et al., 2004), no such method has been reported for yeast. Developing a sampling technique that allows the parallel analysis of all yeast cellular components from the same sample can significantly minimize the errors introduced by the sampling technique itself.

In the current study we have evaluated several yeast culture sampling techniques and developed a protocol that would allow for parallel analysis of RNA transcripts and metabolites from the same sample. The protocol is based on rapid sampling of the yeast culture in cold buffered methanol, which has been previously used to measure the levels of a large number of yeast metabolites (Castrillo et al., 2003; Gonzalez et al., 1997).

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References

Reagents

Sucrose (ultra-pure) was from ICN Biomedicals, Inc. Uracil, L-leucine and L-histidine were purchased from Fisher Scientific. All other reagents used for yeast media preparation were from Difco. Methanol (Optima grade), tricine (Biograde) and water-saturated phenol were purchased from Acros Organics. Tris–HCl buffer, EDTA solution and sodium dodecyl sulphate (SDS, electrophoresis grade) were purchased from Fisher Scientific. Sodium acetate buffer solution was from Sigma Chemical Co. All other reagents were of analytical grade and were purchased either from Fisher Scientific or from Acros Organics. Double-deionized water was used to prepare all the solutions. Sterile, DEPC-treated RNase-free water (Fisher Scientific) was used for preparing RNA solution. MinElute RNA kit (Qiagen) was used for RNA clean-up.

Yeast strain and growth conditions

The Saccharomyces cerevisiae strain used in this study was BY4743 ([4741/4742] MATa/MATα his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 lys2Δ0/+ met15 Δ0/+ ura3Δ0/ura3Δ0; ATCC No. 201390). Cultures were maintained as glycerol stocks frozen at −80 °C. Work cultures were kept at 4 °C on YPS agar plates (0.1% w/v yeast extract, 0.5% w/v peptone and 2% w/v sucrose). For all experiments cells were grown in controlled batch conditions in 1.3 litre BioFlo110 Modular Benchtop Fermentors (New Brunswick Scientific), maintaining the following growth conditions: 30 °C; agitation speed 500 r.p.m.; pH 6.00 ± 0.02; and dissolved oxygen over 90%. Cells were cultured on 2 × minimal liquid medium with 2% w/v sucrose. The inoculum was prepared by growing cells in minimal liquid medium (6.7 g/l yeast nitrogen base without amino acids, 20 mg/l uracil, 60 mg/l L-leucine and 20 mg/l L-histidine) with 2% w/v sucrose. Cultures were grown in flasks in an orbital shaker at 30 °C and 150 r.p.m. for 24 h. Cultures were diluted to OD600 = 0.3 and were used to inoculate three fermentors, constituting three biological replicates. Samples for RNA extraction were collected from each fermentor at mid-exponential phase at OD600 = 1.5.

Culture sampling

Samples were collected from the fermentors as described by Castrillo et al. (2003). To rapidly quench cell metabolism, cultures (60 ml total volume) were sprayed through the High Impact Narrow Angle Full Cone FullJet nozzle (Spraying Systems Co., Wheaton, IL, USA) into tricine-buffered methanol solution kept at −40 °C in an ethanol/dry ice bath. The final methanol concentration (after sampling) was 60% v/v. The temperature was maintained by addition of dry ice to an ethanol/dry ice bath, and continuously monitored with a mercury thermometer. The buffered methanol solution was continuously stirred with a magnetic stirrer. Samples consisting of 2 ml aliquots were also collected in tubes kept on ice.

Sample preparation

Following the quenching step, samples in buffered methanol solution were divided into 12 ml aliquots, and cells were pelleted by centrifugation for 3 min at 5000 r.p.m. in a refrigerated benchtop centrifuge (Beckman Allegra 6R), using a swing bucket rotor pre-cooled to −10 °C. The temperature of the solution was checked after centrifugation to ensure that it was maintained below −30 °C during the centrifugation step. All the procedures, except for the centrifugation steps, were carried on in a −40 °C ethanol/dry ice bath.

After the centrifugation step, samples were divided into three groups (with three replicates per group) and processed as follows:

  • Group A. Cell pellets were washed with 1 ml cold buffered MeOH and centrifuged at 5000 r.p.m. for 3 min at − 10 °C. After centrifugation the supernatant was discarded and 1 ml cell lysis buffer (10 mM Tris–HCl, pH 7.4, 10 mM EDTA, 0.5% SDS) was added to the cell pellets, mixed by vortexing and samples were immediately frozen in liquid nitrogen and stored at − 80 °C until RNA extraction.

  • Group B. Cell pellets were washed with 1 ml cold buffered MeOH and centrifuged at 5000 r.p.m. for 3 min at − 10 °C. Cells were washed once with 2 ml ice-cold 36 mM tricine buffer, pH 7.4, centrifuged at 5000 r.p.m. for 3 min at 0 °C and resuspended in 1 ml lysis buffer. Samples were frozen in liquid nitrogen and kept at − 80 °C until RNA extraction.

  • Group C. Cell pellets were washed with 1 ml cold buffered MeOH and centrifuged at 5000 r.p.m. for 3 min at −10 °C. Pellets were then freeze-dried at −35 °C for 24 h using the Labconco Freeze Dry System. Freeze-dried cells were resuspended in 1 ml of lysis buffer and frozen in liquid nitrogen. Samples were stored at −80 °C until RNA extraction.

Samples collected on ice were centrifuged at 5000 r.p.m. for 3 min at 0 °C. Cells were resuspended in 1 ml lysis buffer and frozen in liquid nitrogen. Samples (designated as Group D) were then stored at −80 °C until RNA extraction. Samples in groups B and C were exposed to cold buffered MeOH for approximately 10 min during the culture collection, aliquoting and centrifugation steps.

RNA Extraction

Total RNA was extracted using a modified hot phenol protocol (Schmitt et al., 1990). Samples stored at −80 °C in lysis buffer were defrosted on ice. After thawing, 1 ml water-saturated phenol was added to each sample and samples were mixed by inversion. Samples were incubated at 65 °C in a dry heating block for 30 min, with occasional mixing by inversion. The upper phase was transferred to a clean microcentrifuge tube and 1 volume of isopropanol and 1/10 volume of 3 M sodium acetate buffer, pH 5.3, were added to the samples. After mixing by inversion, the samples were centrifuged at 14 000 r.p.m. for 15 min at 4 °C (Eppendorf Centrifuge 5415D). The supernatant was discarded and 500 µl cold 70% (v/v) ethanol was added to the pellet. Samples were centrifuged for 10 min in the same conditions and the RNA pellet was dried for 10 min in a Labconco Centrivap Concentrator at 30 °C. RNA was dissolved in 100 µl sterile DEPC-treated RNase-free water and was additionally cleaned up with Qiagen MinElute Kit, according to the manufacturer's protocol (Qiagen, 2003).

Evaluation of the RNA quality

RNA quality and quantity were evaluated by UV-spectroscopy and by capillary electrophoresis using Agilent 2100 Bioanalyser lab-on-a-chip system. Absorbance spectra in the range 200–400 nm and absorbance levels at 260 nm and 280 nm were measured with a Beckman DU 800 spectrophotometer. RNA concentration was determined based on the absorbance at 260 nm and quality was accessed by calculating the A260 : A280 ratio (Warburg and Christian, 1942).

Probe preparation, microarray hybridization and data acquisition

Microarrays were performed using the Affymetrix GeneChip® system. The system includes: (a) disposable probe arrays containing DNA oligonucleotides on a chip; (b) reagents for RNA amplification and labelling; (c) a hybridization oven for optimizing the binding of the target to the probe arrays; (d) a fluidics station for washing and staining after hybridization; (e) a scanner to read the fluorescent image from the probe arrays; and (f) software to control the instruments and to analyse and manage the resulting genetic information.

RNA samples that passed the quality control check were amplified using the GeneChip® One-Cycle cDNA synthesis Kit, as described in Affymetrix TechNotes (Affymetrix, 2004). Hybridization of labelled targets was performed against the Affymetrix Yeast Genome S98 array following the manufacturer's recommended protocols. Arrays were washed, stained and scanned, and the data obtained was captured and stored within the Core Laboratory Facility Lab Information Management System (CLF LIMS). The Affymetrix standard quality metrics for hybridization, staining and overall chip performance was applied throughout the study (Affymetrix, 2004).

Microarray data analysis

GeneChip Operating Software (GCOS) Version1.1 (Affymetrix) was used for data preprocessing and data normalization. In GCOS, signals are calculated as follows:

  • 1.
    Background subtraction and noise correction are performed to correct probe intensities.
  • 2.
    An ideal mismatch (IM) value is calculated and subtracted from the perfect match (PM) value to minimize the effect of cross-hybridization.
  • 3.
    The adjusted PM intensities are log-transformed to stabilize the variance.
  • 4.
    The biweight estimator is used to provide a robust mean of the probe intensities in each probe set.
  • 5.
    The antilog of the signal calculated from the biweight function is then scaled by using the trimmed mean value (Affymetrix).

Details of the GCOS algorithm are described in Statistical Algorithms Description Document at www.affymetrix.com.

Scaling factor 500 was used to all arrays for global normalization. Normalized data were imported into SAS version 8 (SAS Institute Inc., Cary, NC, USA). An ANOVA model, yjk = µ + Tj + εjk, was constructed for each gene, where yjk represents the intensity of the corresponding gene under treatment j and replicate k, µ represents the overall mean value; Tj represents the effect of treatment j, and εjk represents the error. Contrasts were then performed in SAS to compare the gene expression levels between different samples. For each gene, there were three contrasts: Group B vs. Group C, Group C vs. Group D, and Group B vs. Group D. False discovery rate (FDR; Benjamini and Hochberg, 1995) was used to adjust the p value generated from each contrast. In multiple tests, FDR rate can increase sharply when the number of tests is large. To address this problem, the FDR controlling approach (Benjamini and Hochberg, 1995) was used to adjust the p value generated from each test. The FDR controlling procedure developed by Benjamini and Hochberg is as follows. Consider testing null hypotheses H1, H2, …Hm based on the corresponding p values P1, P2, …, Pm. Let P(1) < = P(2) < = … < = P(m) be the ordered p values, and H(i) be the null hypothesis corresponding to P(i).

Let k be the largest i, for which:

  • equation image(1)

where q* is the defined FDR cut-off (in this study, the FDR cut-off is 0.05), then reject all H(i) if ik. Benjamini and Hochberg (1995) mathematically proved that this procedure can effectively control FDR below or equal to cut-off q*.

Results and discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References

Our initial sampling protocol (Group A) was based on the method developed by De Koning and Van Dam (1992) and later modified by Castrillo et al., 2003). The cell culture was sprayed in buffered methanol kept at −40 °C. Following centrifugation to separate cells from the medium, the cells were washed with fresh buffered methanol solution to remove residual medium. After adding the lysis buffer to the cell pellets, the samples were flush frozen in liquid nitrogen and stored at −80 °C until RNA extraction. Unfortunately, using this sampling protocol we were not able to isolate total RNA from the sample. The presence of residual methanol in the pellet, even at a very low concentration, prevented the separation of the aqueous and organic phases after the phenol addition, making subsequent RNA extraction impossible. Therefore it was necessary to remove residual methanol from the samples prior to the RNA extraction.

To remove traces of methanol from the samples, we have modified the sampling protocol by introducing a sequential wash of the cell pellets with cold tricine buffer (Group B) or freeze-drying (Group C). Both treatments B and C were successful in removing the residual traces of methanol from the sample and did not interfere with the subsequent RNA extraction.

As a reference method of culture sampling and RNA preparation, we employed the widely used ice sampling protocol followed by RNA extraction using the hot phenol method (Group D). This method has been successfully used in numerous microarray studies (Causton et al., 2001; Chen et al., 2003; Gasch et al., 2000).

The purity and integrity of the RNA extracted from samples collected with different sampling protocols were evaluated spectrophotometrically, by capillary electrophoresis and using microarray hybridization.

Both cold buffered methanol sampling protocols (Groups B and C) and the reference protocol (Group D) resulted in isolation of high quality total RNA from S. cerevisiae cells (Table 1, Figure 1). The yield and purity of the RNA was sufficient for microarray analysis and other downstream applications. RNA quality was first evaluated by measuring the absorbance ratio at 280 and 280 nm (A260 : A280). This technique, first described by Warburg and Christian (1942) as a means to measure protein purity in the presence of nucleic acid contamination, is now routinely used to evaluate the purity of nucleic acid samples. The A260 : A280 ratios ranged from 1.8 to 2.0, indicating high purity of the total RNA. At p value cut-off 0.05, Welch's t-test found no significant differences between samples B and D, and samples C and D. However, samples B and C were significantly different. The A260 : A280 ratio for samples that were freeze-dried prior to RNA extraction (Treatment C) was higher than that for samples that were washed with tricine buffer (Treatment B).

thumbnail image

Figure 1. Absorbance spectra (200–400 nm wavelength range) for the RNA samples obtained by different sampling and processing methods of yeast cultures. Details on sample collection and processing are described in Materials and methods

Download figure to PowerPoint

Table 1. A260/A280 ratios obtained for RNA extracted from yeast samples collected and processed by different methods
Sample groupsA260/A280
  • Values are means ± standard deviations of three biological replicates (samples collected from three fermentors). Details on sample collection and processing are described in Materials and methods.

  • *

    ND, not determined; it was not possible to extract RNA from samples A.

AND*
B (tricine-washed)1.87 ± 0.01
C (freeze-dried)1.90 ± 0.01
D (collected on ice)1.85 ± 0.09

Another method that is being increasingly used to evaluate the quality and integrity of the total RNA is based on the electrophoretic fractionation of RNA in microchip using the Agilent 2100 Bioanalyser. This method selects features from signal measurements and uses an algorithm that allows the calculation of an RNA integrity number (RNI), which values range from 10 (intact RNA) to 0 (totally degraded RNA) (Schroeder et al., 2006). Until recently the integrity of RNA samples was estimated from the ratio of ribosomal RNA, 28S : 18S. When the value is about 2, the RNA is considered to be of high quality. However, this method has been shown to be inconsistent, and a ratio 28S : 18S of 2 shows a weak correlation with the RNA integrity (Schroeder et al., 2006). Hence, we used the RIN method for additional evaluation of the RNA quality. The results presented in Table 2 and Figure 2 are also consistent with the high quality of RNA, with RIN values of 9.4–9.8.

thumbnail image

Figure 2. RNA electrophoregrams obtained with the Agilent 2100 Bioanalyser. The two peaks correspond to the 18S and 28S fragments. RNA integrity numbers (RINs) and 28S : 18S ratios presented in Table 2 were calculated from these electrophoregrams. Details on sample collection and processing are described in Materials and methods

Download figure to PowerPoint

Table 2. Bioanalyser results for the RNA samples obtained by different sampling and processing methods of yeast cultures
SamplesSampling technique28S : 18S ratioRIN
  1. All the values obtained for the three biological replicates are shown. The ratios 28S : 18S and the RNA integrity numbers (RINs) were calculated from the electrophoregrams obtained for each sample and presented in Figure 2. More details on sample collection and processing are described in Materials and methods.

B1Collected in cold buffered methanol, washed with tricine buffer, resuspended in lysis buffer and1.29.6
B2frozen in liquid nitrogen1.29.4
B3 1.29.4
C1Collected in cold buffered methanol, freeze-dried, resuspended in lysis1.29.6
C2buffer and frozen in liquid1.49.6
C3nitrogen1.19.8
D1Collected on ice, resuspended in lysis buffer and1.39.5
D2frozen in liquid nitrogen1.39.6
D3 1.19.5

To further evaluate the effect of the sampling technique on the RNA quality and its suitability for large-scale transcriptomics analysis, we performed microarray experiments using the Affymetrix GeneChip® system for the yeast genome (Yeast Genome S98). ANOVA analysis of the microarray results (with a cut-off for FDR corrected p value of 0.05) showed that none of the 9335 probes in the chip was differentially expressed for each treatment pair comparison. Hence, no significant difference in gene expression can be identified among the three sample sets.

Although both treatments B and C resulted in high quality RNA suitable for microarray analysis, treatment B was less reproducible than treatment C and resulted in higher variability of measurements (Figure 3). The coefficients of variation (CV = 100 × standard deviation/average) for each gene in the triplicate samples for each group were calculated. For groups C and D, the coefficients of variation with the highest frequency are in the range of 10 to 20% (Figure 3). However, for group B, the coefficients of variation with the highest frequency are in the range 60–70%. These data suggest that removal of water from the sample by freeze drying improves data reproducibility and can potentially extend the stability and shelf-life of the sample. Therefore, sampling protocol C was identified as the best in our tests.

thumbnail image

Figure 3. Histograms of coefficients of variation for genes in different RNA samples. RNA samples obtained by each sampling method (triplicates) were hybridized to microarrays (Yeast Genome S98 chips, Affymetrix). The coefficients of variation (CV = 100 × standard deviation/average) were calculated for each gene in each triplicate sample. Details on sample collection and processing are described in Materials and methods

Download figure to PowerPoint

In conclusion, we have validated a method for sampling S. cerevisiae cultures that can be used for parallel analysis of the transcript and metabolite levels from the same sample. It involves collecting the cultures in tricine-buffered methanol (Castrillo et al., 2003), separation of the cells from the culture medium by centrifugation, and removal of the residual methanol by freeze-drying of the cell pellet. This method is fast, simple and especially useful in dense time-course experiments, when it is difficult to process the samples immediately. This method can be used in systems biology experiments where the difference in sampling time is critical for further interpretation of the results.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References

This work was supported by NIH-NIGMS Grant R01 GM068947-01. A.M.M. was supported by postdoctoral Grant SFRH/BPD/8033/2002 from the Portuguese Science Foundation (Fundação para a Ciência e Tecnologia). The authors would like to thank Dr. Jacky Snoep (University of Stellenbosch, South Africa) for help with setting up the fermentation facility. The authors also thank Jim Walke and Diogo Camacho for the critical reading of the manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References