figure 1 microarray analysis to study the subcellular localization of mRNAs involved in mitochondrial biogenesis.
After cycloheximide blocking of translation polysomes associated with mitochondria (MP) and free cytoplasmic polysomes (FP) were purified and the corresponding mRNA species were purified and treated to be analyzed with high density oligonucleotide microarrays. The total mRNA population
(TOT) was extracted in parallel with the two other fractions and used as a reference. After normalization and signal treatment a synthetic localization value called MFI (Membrane versus Free Index) was calculated as indicated. 97% probe sets, corresponding to 96% genes and 94% mitochondrial genes have MFI values. figure 2
Comparison of the high-density oligonucleotide analysis (MFI) with the previous cDNA microarray analysis (MLR).
A Spearman rank correlation coefficient between the MFI values obtained in this work and the previous MLR values from previous experiments was established as a function of the number of MLR experiments. The MFI experiments which are more accurate provides results for more than 95% of yeast genes. Number of genes for which the MLR was calculated: 1 experiment: 1267; 2 experiments: 666; 3 experiments: 442; 4 experiments:
272; 5 experiments: 119; 6 experiments: 101. figure 3
Whole-genome MFI distribution.
Individual results (including protein localization and mfi class) for most of the yeast genes are available from our database. As examples, genes which are known to code for mitochondrially localized products and belonging to the two extreme 5% MFI values are listed, thus the mRNAs of genes named on the right are translated to the vicinity of mitochondria. figure 4
Cumulative distribution of MFIs and protein localization.
Empirical cumulative distribution functions of MFIs for mRNAs encoding proteins localized in three kinds of subcellular compartments. The left
(green) and right (red) curves correspond respectively to 443 cytosolic and 136 plasma membrane/endoplasmic reticulum localized proteins
(corresponding mRNAs are known to be translated respectively on free cytosolic and membrane-bound polysomes). The 329 mitochondrial proteins are in the curve in the middle (blue), their mRNA localization being intermediate between that of the two reference populations mentioned above. Empirical cumulative distribution functions of MFIs for mRNAs encoding proteins localized in three kinds of subcellular compartments.
The left (green) and right (red) curves correspond respectively to cytosolic and plasma membrane/endoplasmic reticulum localized proteins
(corresponding mRNAs are known to be translated respectively on free cytosolic and membrane-bound polysomes). Mitochondrial proteins are in the curve in the middle (blue), their mRNA localization being

intermediate between that of the two reference populations mentioned above. figure 5
Phylogenetic categories of genes coding for mitochondrial proteins and
MFI values.
Mitochondrial proteins were previously grouped into four classes corresponding respectively to: SC=genes specific to Saccharomyces cerevisisae, E=genes with homologues in Eucaryotes exclusively, BE=genes with homologues in Bacteria and Eucaryotes, ABE=genes with homologues in
Archaebacteria, Bacteria and Eucaryotes. Bard are Tukey-Cramer 95% confidence intervals. This graph hence shows a very significant effect of the Prokaryotic/Eucaryotic origin of genes on their mRNAs MFIs. figure 6
Mitochondrial mRNA localization and mRNA length.
The MFI values of 305 genes known to have mitochondrialy localized products are plotted against the corresponding mRNA length (semi-log scale). The Pearson correlation coefficient between MFIs and log mRNA length is 0.6. The red dots correspond to genes of prokaryotic origin
(ABE in Figure 5). The higher number of nascent polypeptide-chain associated complex (NAC) found on longer mRNA could attract corresponding polysomes to mitochondrial vicinity (see text for details). figure A
Effect of quantile normalization on MvA plots and boxplots of logintensities.
Data for Perfect Match signal of chips MP1 and MP2. One can see that the intensity dependance of the ratio is corrected, leading to an horizontal
MvA plot and aligned boxplots. figure B
Log-intensities for spike mRNAs.
Rows : experiments MP1, MP2, FP1, FP2, TOT1, TOT2; columns : experiments
MP1, MP2, FP1, FP2, TOT1, TOT2. On the diagonal are log-intensity histograms for each experiments. While some experiment-dependant effects can be guessed from these graph, they seem clearly intensity-dependant
(non-linears). Since these effect remain quite smalls and since the number of control mRNA spots is limited we did not apply any further correction, which would have been hazardous. figure C
MFI for replicate 1 (chips MP1, FP1, TOT1) versus MFI for replicate 2
(chips MP2, FP2, TOT2).
This graph shows a good correlation between the two independant subcellular fractionation and sets of chips.

figure D
ANOVA results.
A significant effect of evolutionary groups on mRNA localization is demonstrated here (alpha = 5% ; see text for multicomparison analysis). figure E
Comparison of the length of genes coding for mitochondrial proteins and cytoplasmic proteins in yeast.
The left (blue) and right (green) curves correspond respectively to 360 genes coding for mitochondrial proteins and to 540 genes coding for cytoplasmic proteins. Both curves are kernel density estimates with gaussian kernel and appropriate bandwidth. figure F
Analysis of the quality of RNA purified from the two polysomal preparations examined.
A. Two micrograms of RNA purified from spheroplasts (S), mitochondrionbound polysomes (M) and free-cytoplasmic polysomes (P) were subjected to electrophoresis using 1 % TAE agarose gel with ethidium bromide. The 25S and 18S rRNAs were visualized (rRNAs), the integrity of the preparation was then confirmed. The two preparations used for the two oligonucleotides array analysis were shown (S1, M1, P1 and S2, M2, P2).
B. 200 ng of RNA for each preparation was used for reverse transcription. The products obtained were subjected to twenty-five cycles of PCR with the following specific oligonucleotides inside each open reading frame :
ATP2: 5' tagaagaaataaagcttaaaccaagggg 3' and 5' catgtccagtgggaaagcgaggcaaga 3'
ATM1: 5'tcatagttcttgctggtcttttag 3' and 5' caatgacactatctgggaaaatgt 3'
OXA1: 5' ttccgcttctacttcggaccttatcgc 3' and 5' attggaatttgtagcatgggtgcggcc 3'
COX3: 5' atgacacatttagaaagaagtagacatcaa 3' and 5' tattacctgcgattaaggcatgatgactat 3'
COX6: 5' tatacaacaatgttatcaagggccata 3' and 5' ttcatcatgtgcgtcagaatacttcct 3'
ACT1: 5' gaggttgctgctttggttatt 3' and 5' gtggtgaacgatagatggacc 3'
VTC4: 5' atgaagtttggtgagcacttgagcaag 3' and 5' gcgacctgaagttctggcaatatcata 3'
SSH1: 5'atgtctggttttcgtctaattgatatc 3' and 5' ctttgtttacatagcacctggaacacc
3'
10% of the amplified products were subjected to electrophoresis; the size of each amplified product is shown at the bottom. The figures shows+the result for RNAs purified from Spheroplasts (S1), mitochondrion-bound polysomes (M1) and free-cytoplasmic polysomes (P1).