Strains used

All the strains used in this study are isogenic to CW252, an intronless mitochondrial genome with a nucleus isogenic to that of W303 [33] with the exception of the BCS1 Puf3p binding site mutation studies. In this latter experiments a BY4742, bcs1::kanMX4 (Euroscarf) was used for FISH and Q-PCR analyses. The puf3-Delta strain was constructed by homologous recombination, using a PCR-amplified DNA fragment from the puf3 (YLL013c) locus of the strain BY4741, YLL013c::kanMX4 (Euroscarf).

Extraction of mitochondrion-associated RNA

Mitochondrion-associated RNA and total cellular RNA were extracted as previously described [19]. Cells were grown at 30°C in galactose medium (1% bactopeptone, 1% yeast extract, 2% galactose, 0.1% KH2PO4 and 0.12% (NH4)2SO4). Cells were incubated for 10 minutes and fractionated, in the presence or absence of 200 µg/ml cycloheximide. Cells were treated with puromycin to release mitochondrion-bound cytoplamic ribosomes. The cells were fractionated in the presence of 2.1 mM puromycin and 1 M KCl as previously described [10]. RNA was extracted and purified from the various fractions using the RNEasy Mini Kit (Qiagen). Three independent RNA preparations (mitochondrial + cellular RNA) were obtained in the absence of drugs and two independent preparations were obtained for each treatment with cycloheximide or puromycin and in the absence of Puf3p. The quality of the extraction was checked by quantitative PCR analysis as previously described [19].

Array hybridization and normalization

The microarray data and the related protocols are available at the GEO web site (www.ncbi.nlm.nih.gov/geo/), accession number: GSE9393. Briefly, the cellular and mitochondrion-associated RNAs of each independent preparation were reverse-transcribed and labeled with Cy3 or Cy5 dye using the indirect labeling procedure. We then hybridized 1.5 µg of cellular and mitochondrial labeled cDNA with ab 8 x 15K S. cerevisiae DNA chip manufactured by Agilent. Each array hybridization was replicated using dye-swap. Arrays were read using a GenePix 4000B scanner (Molecular Devices) and the signal segmentation was done using the GenePix Pro 6.1 software. Data pretreatment was applied on each result file to discard GenePix flag and saturating spots. The data were normalized without background substraction by the global Lowess method performed with the Goulphar software [34].

Percentage of mRNA asymmetric localization and SAM analysis

In RT-PCR analyses, the percentage of each mRNA localized in mitochondrial fraction was calculated in a manner similar to that used in a previous study [19]. For the calculation of the mitochondrial localization rate from microarray data, we used a method adapted from [19]. A detailed protocol is presented in Doc S1. Briefly, we estimated the mitochondrial purification yield from the median log2 values of the mitochondrial fraction/total cell extract ratios of 33 RNAs transcribed from the mitochondrial genome (8 mRNAs, 2 rRNAs and 23 tRNAs). This median value was set at 100% and served as the normalization factor used to transform all the log2 values into mitochondrial localization rates. These raw rates were corrected to take into account the biochemical contamination due to the non-mitochondrial fraction. This task is difficult because the mitochondrial fraction is always contaminated by membrane fractions [23]. However, as negative control, we considered 416 mRNAs not connected with mitochondria (according to YPD, see supplemental table S4) and calculated a median contamination value, which was substracted from the raw localization rates as described in [19] to obtain the final corrected and normalized mitochondria localization rates (MLR values) for each mRNA probed on the array. To generate a list of mRNAs which were statistically enriched in the mitochondrial fraction in our experiments, the microarray data were analyzed with the SAM (significance analyses of microarrays) method implemented in the TMEV software (www.tm4.org), with default parameters and a false discovery rate of 5% [35]. This list was used to set a MLR value cutoff with a reasonable rate of false positive. We observed that a cutoff of 8% led to 3% of mislocalized mRNA according to the list generated by SAM (Figure S1), with a total false positive rate of 8% (5% of SAM FDR plus 3% of discrepancies between SAM and the MLR value cutoff). Therefore, among the 489 MLR-mRNAs identified in this study, only 40 may be false positives.

FISH analysis and in situ mRNA localization

WT cells were grown on galactose medium (1% bactopeptone, 1% yeast extract, 2% galactose) to mid-exponential growth phase and were treated as described in a previous study [12]. The combined hybridization of three to five antisense oligonucleotides probes was used to enhance the visualization of each mRNA. Typically, each probe was designed such that it contained 50 to 55 nucleotides, with five aminoallyl thymidines and a coherent Tm value. After synthesis by Eurogentec (San Diego, CA), the probes were directly labeled with CY3 or CY5 fluorochromes and hybridized to fixed cells as previously described [19].

Image analyses

Fluorescent imaging and three-dimensional sampling by multiple-stage microscope acquisition allows collecting data within the whole cell but conclusions driven in the absence of 3D reconstitution are highly speculative. To assess relative RNA localizations we developed a software dedicated to distance measurement between two fluorescent probes and allowing analysis of a large cell population. We could thus quantify the distance between mitochondria, revealed with mitochondrial rRNA probes, and a specific mRNA. This analysis is based on 1) 3D segmentation and extraction of object features (coordinates and intensity). 2) Distance measurement of each mRNA particles to the mitochondrial surface. 3) Statistical analysis of data generated: distances are intensity-weighted and for each cell the distribution of mRNA- mitochondria distance is analysed to extract the percentage of particles in the close vicinity of mitochondria (a threshold of 100nm was chosen to be consistent with microscope resolution). In these conditions if more than 35% of the studied mRNA is localized to mitochondrial vicinity there are 95% chances that the corresponding cell exhibits mRNA asymmetric localisation (see results for ATP16 mRNA). Data represent fraction of cell population with localized mRNA with confidence intervals calculated using binomial test. Such analyses could be routinely carried out on more than 300 cells. Details characteristics and availability of the distance-FISH software, called CORSEN, are in preparation (Jourdren et al).