In this work, we provide evidence for the existence of a nuclear factor involved in the splicing of a specific mitochondrial intron in higher plants. In the Nicotiana sylvestris nuclear NMS1 mutant, defective in both vegetative and reproductive development, the first intron of the nad4 transcript encoding the complex I NAD4 subunit is not removed, whatever the tissue analysed. Transcript patterns of other standard mitochondrial genes are not affected in NMS1. However, numerous polypeptides are missing in two-dimensional in organello mitochondrial protein synthesis patterns and several nuclear and mitochondrial complex I subunits are present in trace amounts. This indicates that translational or post-translational steps in the synthesis of other mitochondrial proteins are affected. All of these defects co-segregated with the abnormal phenotype in the offspring of a NMS1 × wild-type cross, showing that they are controlled by the same nuclear gene (MS1) or tightly linked loci. Such a complex situation has been described in chloroplasts and mitochondria of fungi, but never in higher plant mitochondria.
Mitochondria are semi-autonomous organelles, the genome of which encodes only a small part of mitochondrial (mt) proteins ( Tzagoloff & Myers 1986). The majority of mitochondrial proteins are nuclear-encoded and imported after processing of the N-terminal signal peptide. In all eukaryotes, biogenesis and functioning of mitochondria are thus largely under the control of the nuclear genome, from the transcriptional to the post-translational steps. In yeast and fungi, several cases of precise control of mtRNA processing by the nuclear genome have been described, namely in intron splicing (reviewed in Grivell 1995). In higher plants, in which the genome is larger than in other eukaryotes ( Schuster & Brennicke 1994), nuclear control of mitochondrial gene expression is less well documented ( Binder et al. 1996 ). Most data concern the post-transcriptional action of the Rf genes restoring fertility in cytoplasmic male sterile mutants (CMS) (reviewed in Schnable & Wise 1998). Although mtDNA contains numerous type II introns, essentially located in genes encoding complex I subunits (nad genes) that are either cis- (nad4, nad7) or trans-spliced (nad1, nad2, nad5), in none of the plant species studied so far has an example of nuclear control of plant mtRNA splicing been documented.
We previously reported that the ms1 recessive nuclear mutation conferring morphological abnormalities in vegetative and reproductive organs in the NMS1 (formerly MST1) Nicotiana sylvestris mutant co-segregated with the absence of a 40 kDa mitochondrial-encoded polypeptide, suggesting that the wild-type MS1 allele controls mtDNA expression ( De Paepe et al. 1990 ). NMS1 plants presented similar developmental abnormalities, although more severe, than those of cytoplasmic CMSI and CMSII male-sterile N. sylvestris plants ( Chétrit et al. 1992 ; Li et al. 1988 ). These mutants were found to be totally (CMSII) or partially (CMSI) deleted for the mitochondrial nad7 gene, which encodes a subunit of the respiratory chain complex I ( Gutierres et al. 1997a ; Pla et al. 1995 ). As a result of the mtDNA deletion, CMS mitochondria are defective for NAD7. In plants, as in other eukaryotes, complex I or NADH:ubiquinone oxidoreductase, EC 188.8.131.52, is composed of more than 30 subunits, either mitochondrial- (nad genes) or nuclear-encoded ( Rasmusson et al. 1998 ). All NAD subunits are part of the membrane arm of the complex, except NAD7 and NAD9, which belong to the peripheral arm protruding into the matrix. In both CMSI and CMSII, in addition to NAD7, at least two other peripheral complex I subunits, NAD9 and a 38 kDa nuclear-encoded subunit are lacking or present in highly reduced amounts ( Gutierres et al. 1997a ).
NMS1 plants grow more slowly than CMS, display more severe abnormalities in vegetative and reproductive organs, and only occasionally set a few seeds under high sodium lighting. However, CMS plants produce about 50% of normal pollen amounts under these conditions. The ms1/MS1 hybrids resulting from reciprocal crosses between NMS1(ms1/ms1) and wild-type T(MS1/MS1) plants, display normal morphology and seed set ( De Paepe et al. 1990 ).
In this paper, we investigated transcriptional and post-translational aspects of mitochondrial gene expression associated with the NMS1 phenotype in the offspring of ms1/MS1 hybrids. We found that, among the standard mitochondrial genes analysed, only nad4 transcripts were affected, in that the first intron remained unspliced. However, in organello mitochondrial synthesis patterns presented numerous differences compared to the wild-type T, and immunodetection studies showed that NMS1 mitochondria are defective for NAD7, NAD9, and the 23 kDa and 38 kDa nuclear-encoded complex I subunits. Such mitochondrial expression defects were not age- or tissue-specific, as they were found in young plants, adult leaves and pollen.
Absence of splicing of the nad4 first intron in NMS1 mitochondria
In order to determine whether mitochondrial transcript patterns were affected in NMS1 plants, we conducted RNA blot experiments using probes for the standard respiratory genes nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9, cob, cox1, cox2, cox3, atp1, atp6 and atp9 (details in Experimental procedures). nad4 and nad7 are cis-spliced in N. sylvestris as in the other plant species investigated so far, whereas nad1, nad2 and nad5 are trans-spliced, and each individual transcription unit was tested ( Gutierres et al. 1997b , 1999; Pla et al. 1995 , and unpublished results). The only genes coding for ribosomal proteins identified in N. sylvestris, i.e. rpl2 and rps13 ( Vitart et al. 1992 and unpublished results), were also tested. As seen in Fig. 1, most gene probes gave complex transcript patterns, due either to the presence of processing intermediates or co-transcription with neighbouring mitochondrial sequences, or both. These particularities of plant mitochondrial gene expression render it difficult, if not impossible, to identify the mature transcript of most mitochondrial genes (see for example the paper of Gutierres et al. 1999 concerning the nad1 gene). In spite of this complexity, no clear-cut differences were found between T and NMS1 patterns, except for nad4 ( Fig. 1, lanes 13 and 14): the 1.8 kb transcript corresponding in size to the mature nad4 transcript of wheat ( Lamattina & Grienenberger 1991) was barely detectable in NMS1, whereas a major 3.2 kb transcript was visible.
In order to determine the nature of this abnormal nad4 transcript, we first investigated the structure of the nad4 gene in N. sylvestris. In all plant species analysed so far, nad4 is interrupted by introns, from one to three, depending on the species ( Gass et al. 1992 ). Using the wheat nad4 exon3 sequence as a probe, we found that the N. sylvestris nad4 sequence was located in a 6.2 kb SacI fragment from an mtDNA cosmid library ( Chétrit et al. 1992 ; Lelandais et al. 1998 ). PCR experiments were carried out, using either this fragment or total N. sylvestris T, CMSII or NMS1 DNAs as templates, and oligonucleotides designed from the conserved regions of plant nad4 exon sequences ( Fig. 2). Results showed that the N. sylvestris nad4 gene was similarly composed of four cis-spliced exons in all three genotypes, having approximately the same size as in B. campestris ( Gass et al. 1992 ) and wheat ( Lamattina & Grienenberger 1991): about 400, 500, 450 and 80 bp in size, respectively ( Fig. 2a, lanes 1–14). The intron lengths, about 1.4, 3.5 and 1.9 kb in size, were also well conserved ( Fig. 2a, lanes 15–38). Exon- and intron-specific PCR fragments ( Fig. 2b) were hybridized to leaf total RNA extracted from 2-month-old T, CMSII and NMS1 plants. Probes corresponding to exon 1 ( Fig. 2c, lanes 1–3), exon 2 ( Fig. 2c, lanes 7–9) and exon 3 ( Fig. 2c, lanes 10–12) all gave a main 1.8 kb signal in T (lanes 1, 7, 10) and CMSII (lanes 2, 8, 11; same results for CMSI, not shown). This indicates that this signal corresponds to the mature nad4 transcript in N. sylvestris as in wheat ( Lamattina & Grienenberger 1991). In contrast, all exon probes gave a 3.2 kb signal in NMS1 ( Fig. 2c, lanes 3, 9, 12). The first intron probe, about 1.4 kb in size, gave a very slight signal when hybridized to T ( Fig. 2c, lane 4) and CMSII ( Fig. 2c, lane 5) mtRNA, whereas it hybridized strongly to the 3.2 kb RNA species in NMS1 ( Fig. 2c, lane 6). The third intron probe, about 1.9 kb in size, gave weak signals around 2–2.5 kb in all genotypes ( Fig. 2c, lanes 13–15), most likely corresponding to unprocessed RNA species. Taken together, these results show that the first nad4 intron is essentially not spliced in NMS1 whereas other introns are normally processed; the second intron was not tested, as it is about 3.5 kb in size ( Fig. 2a,b) and thus cannot contribute to the 3.2 kb signal. Similar results were obtained in NMS1 plants whatever their generation and age. For example, an adult NMS1(ms1/ms1) plant ( Fig. 2d, lane 3) and a plantlet of its selfed progeny ( Fig. 2d, lane 4) showed identical patterns, although transcripts were more abundant at the young stage. In contrast, MS1/MS1 F1 plantlets, resulting from the cross of the same ms1/ms1 plant with a wild-type MS1/MS1 plant, displayed a normal nad4 northern pattern ( Fig. 2d, lane 2). The defective splicing was not tissue-specific, as NMS1 nad4 transcription patterns were identical in leaves and pollen-containing anthers ( Fig. 2e). In all tissues, similar very low traces of the normal 1.8 kb transcript could be detected.
It could not be determined whether an NAD4-like polypeptide was synthesized in NMS1 mitochondria, since no antiserum is presently available. However, it seems very likely that the long 1.4 kb nad4 first intron would contain stop codons in N. sylvestris as is the case for other nad4 introns in which no open reading frame larger than 75 codons was found ( Gass et al. 1992 ). Thus, only a truncated NAD4 polypeptide of about 15 kDa, roughly corresponding to the first exon size, is expected in NMS1, at the best.
Decreased amounts of several in organello synthesized mitochondrial polypeptides in NMS1
Previous mono-dimensional SDS–PAGE of in organello synthesized mitochondrial polypeptides revealed the absence of a 40 kDa polypeptide in NMS1 ( De Paepe et al. 1990 ). We have screened for other changes by two-dimensional SDS–PAGE. Among the 50 or so autoradiographic spots visible in T patterns, about 10 were absent or present in trace amounts in NMS1 (arrowed in Fig. 3a,b). Major differences concerned a polypeptide of about 55 kDa, pI 7.2, two spots in the 40 kDa range, pI 5.8 and 6, and two spots in the 25–28 kDa range, pI 6.5. The 55 kDa spot is a candidate for NAD4, as its molecular weight calculated from gene sequence is 56 kDa ( Lamattina & Grienenberger 1991). However, it is well known that highly hydrophobic polypeptides, such as membrane NAD subunits, may migrate faster in SDS–PAGE than would be expected from their molecular weight ( Walker 1992). In wheat, it has been proposed recently that NAD4 has an apparent migration corresponding to 44 kDa ( Combettes & Grienenberger 1999); it is thus possible that NAD4 could be one of the spots missing in the 40 kDa range.
NMS1 leaf mitochondria are defective for mitochondrial and nuclear-encoded complex I subunits
It was previously shown in both CMSI and CMSII plants that the absence of NAD7 resulting from the deletion of the nad7 gene ( Pla et al. 1995 ) was associated with the loss of both NAD9 and the nuclear-encoded 38 kDa subunit ( Gutierres et al. 1997a ). We checked whether similar defects occurred in the NMS1 complex I following the defective processing of nad4. We first compared T and NMS1 leaf mitochondria by Western experiments using antisera directed against the Neurospora crassa 49 kDa subunit ( Weiss et al. 1991 ) homologous to plant NAD7, and the wheat NAD9 polypeptide ( Lamattina et al. 1993 ) ( Fig. 4). Leaf mitochondrial proteins of MS1/ms1 fertile hybrids and CMSI plants were used for comparison. The 42 kDa signal corresponding to NAD7 in the T line ( Fig. 4, lane 1) was not detectable in NMS1(ms1/ms1) as in CMSI ( Fig. 4, lanes 3 and 4). Similarly, the 28 kDa protein corresponding to NAD9 in T ( Fig. 4, lane 5) showed only traces in both mutants ( Fig. 4, lanes 7 and 8). Other complex I nuclear-coded subunits were screened using bovine antisera. Among the eight antisera directed, respectively, against the 75, 51, 31, 24, 23, 19, 14.5 and 10 kDa bovine subunits, only the 23 kDa reacted differently on T and mutant blots: the 25 kDa band seen in the T panel was missing in both NMS1 and CMSI panels ( Fig. 4, lanes 9–12). Other antibodies either did not specifically react with N. sylvestris mitochondrial proteins or did not reveal differences among genotypes (not shown). All signals absent in NMS1 were present in normal amounts in the MS1/ms1 hybrid ( Fig. 4, lanes 2, 6 and 10). In control experiments, similar amounts of the formate dehydrogenase enzyme (FDH), 40 kDa in size in potato mitochondria ( Colas des Francs-Small et al. 1993 ), were found in all genotypes studied ( Fig. 4, lanes 13–16).
Finally, to screen for the nuclear-coded 38 kDa complex I subunit previously shown to be absent in the CMS mutants ( Gutierres et al. 1997a ), two-dimensional SDS–PAGE patterns of silver-stained leaf mitochondrial proteins were compared among T, CMSII, NMS1 and MS1/ms1 hybrid plants ( Fig. 5a–d). The 38 kDa subunit was absent in NMS1 ( Fig. 5c) but present in the fertile MS1/ms1 hybrid ( Fig. 5d).
Lack of NAD7 and NAD9 in NMS1 pollen mitochondria
We determined whether the NMS1 complex I defects were specific to leaf, or whether they were also displayed in pollen. We previously reported that, although NMS1 plants were completely sterile under low light, they occasionally shed low amounts of pollen under high light ( De Paepe et al. 1990 ). This was confirmed by the Alexander test which stained some NMS1 pollen grains purple ( Alexander 1969), suggesting they had not lost viability ( Fig. 6). Interestingly, all pollen from the MS1/ms1 hybrid stained purple, indicating a sporophytic control of sterility. In T pollen, electron micrographs revealed the presence of numerous mitochondria, with well-differentiated cristae ( Fig. 7a). In CMSI ( Fig. 7b,c) and NMS1 ( Fig. 7d,e) non-aborted pollen, mitochondria were also clearly identifiable, although in some grains they were swollen without well-defined cristae ( Fig. 7c,e), whereas in other grains they displayed a normal structure ( Fig. 7b,d).
In wild-type mature pollen, NAD7 and NAD9 could easily be detected by Western blot analysis ( Fig. 4, lane 17), but it was not possible to purify mitochondrial proteins from NMS1 pollen in sufficient amounts for comparison. Therefore, we undertook pollen comparisons using in situ immunocytochemistry. Anther cross-sections were treated with NAD7 and NAD9 antisera using a silver-enhanced immunogold technique. FDH antiserum was used as a control. All antisera gave positive signals for T pollen grains, showing a homogeneous silver–green stain under epipolarization optics ( Fig. 8a,d,g). Pollen of both CMSI and NMS1 gave positive signals using the FDH antiserum, indicative of a persistent metabolic activity ( Fig. 8h,i). In contrast, using anti-NAD7 or anti-NAD9 antisera, only a very weak heterogeneous labelling signal was obtained in both mutants ( Fig. 8b,c,e,f). These results suggest that only very low amounts of NAD7 and NAD9 were present in NMS1 and in CMSI pollen mitochondria (same results with CMSII, not shown). Control experiments consistently gave negative results for all genotypes ( Fig. 8j–l), indicating that the positive signals obtained with specific antisera were not artefactual.
Absence of the 38 kDa subunit in NMS1 pollen and its presence in the fertile hybrid pollen were demonstrated by two-dimensional SDS–PAGE followed by silver staining (not shown).
Presence of low amounts of NAD7 and NAD9-containing complex I in NMS1 leaf mitochondria
We wished to determine whether NMS1 mitochondria were completely devoid of NAD7 and NAD9, or whether low amounts of complex I containing these subunits were still assembled. Complex I was purified using an immunoaffinity method developed to purify complex I in wheat ( Combettes & Grienenberger 1999) and N. sylvestris ( Gutierres et al. 1999 ), based on the use of anti- NAD9–protein A columns coupled to agarose beads. Unexpectedly, in view of the apparent lack of NAD9 in CMS and NMS1, a protein fraction was retained by the columns for all genotypes; however, on a leaf weight basis, only 10–20% of the wild-type amount was obtained for both mutants. Silver-stained SDS–PAGE electrophoretic patterns of the retained protein fractions consisted of more than 30 polypeptides and differed only slightly among genotypes, suggesting they corresponded to a near completely assembled complex I ( Fig. 9a). These results could be explained by low amounts of NAD9-containing complex I (called NAD9+ complex I) in mutant mitochondria. This was confirmed by Western analyses which showed the presence of NAD9 in both CMS and NMS1 retained fractions ( Fig. 9b). Also, in NMS1, the NAD9+ complex contained normal amounts of NAD7, while no traces of NAD7 could be detected in the CMS complex, as expected since the unique copy of the nad7 gene was deleted ( Gutierres et al. 1997a ; Pla et al. 1995 ).
The molecular analyses presented here show that the NMS1 (ms1/ms1) mutant is affected in mitochondrial post-transcriptional steps, as well as in respiratory complex I composition: (1) the first intron of the mitochondrial nad4 transcript essentially remains unspliced; (2) the relative abundance of about ten in organello synthesized mitochondrial polypeptides is strongly decreased; and (3) two mitochondrial-, NAD7 and NAD9, and two nuclear- encoded subunits of complex I, the 38 and the 23 kDa proteins, all belonging to the peripheral arm of the complex, are absent or present in reduced amounts.
Thus, with respect to the composition of their complex I peripheral arm, NMS1 mutants resemble the previously characterized CMSI and CMSII mtDNA deletion mutants. However, the two types of mutations differ in the mode of inheritance: the CMS-associated traits are maternally inherited, while those of NMS1 are nuclear-inherited. Indeed, in all aspects, phenotypical or molecular, reciprocal (NMS1 × T) hybrid plants were identical to wild-type. Co-segregation of molecular defects with the abnormal phenotype in sexual offspring of these hybrids indicated that they are controlled either by the same recessive nuclear mutation or by mutations in tightly located loci. The fact that no changes could be detected in NMS1 mtDNA, whatever the generation tested, also strongly suggests that the wild-type Ms1 locus controls mtDNA expression rather than structure, contrary to the CHM locus of Arabidopsis ( Martinez-Zapater et al. 1992 ) or the bean Fr restorer gene ( He et al. 1995 ).
MS1 is a specific mitochondrial group II intron splicing factor
In the mitochondrial nad4 gene, the wild-type MS1 allele specifically controls the splicing of the first intron, as other introns are normally removed. The reason for such specificity has not been elucidated, but it has already been noted in B. campestris ( Gass et al. 1992 ) and lettuce ( Geiss et al. 1994 ) that splicing of nad4 intron 1 is delayed as compared to that of other introns. However, in N. sylvestris, nad4 intron 1 did not significantly hybridize to wild-type mtRNA from seedlings, leaf, pollen or anthers (not shown), indicating that this intron is rapidly processed and degraded in all tissues. This suggests either the presence of a more efficient form of the MS1 factor in N. sylvestris than in other species, or that intron/exon boundaries are different and more easily processed.
Genetic evidence for intron II splicing factors in plant mitochondria has never been demonstrated, in contrast to the situation found in yeast or Neurospora. In fungi, both group I and group II introns can self-splice in vivo and in vitro ( Podar et al. 1998 ), depending on the presence of an intron-coded maturase activity, but in vivo splicing is under the control of many nuclear factors (reviewed in Grivell 1995). In higher plants, nuclear genes involved in chloroplast RNA maturation have also been described ( Barkan et al. 1994 ; Hess et al. 1994 ). Such factors may or may not be intron-specific, showing that intron splicing requirements can differ ( Hübschmann et al. 1996 ; Jenkins et al. 1997 ).
The NMS1 splicing defect seems to be gene-specific, as Northern analysis of other standard intron-containing mitochondrial genes, namely nad genes and cox2, did not reveal qualitative differences as compared to T. Also the transcript pattern of the rpl2 gene, found to contain an intron in N. sylvestris ( Vitart et al. 1992 and unpublished results) and rice ( Kubo et al. 1996 ) was not affected. Taken together, these results strongly suggest that the action of MS1 at the splicing level is essentially, if not totally, gene- and intron-specific. However, it cannot be ruled out that any of the transcripts encoded by the still unidentified open reading frames present in higher plant mtDNA ( Unseld et al. 1997 ) may be incorrectly processed in the mutant. Whatever the case, our results provide the first evidence for the presence in higher plants of a nuclear factor, MS1, directly or indirectly involved in the splicing of a specific mitochondrial intron.
Lack of mitochondrial and nuclear-encoded complex I subunits in NMS1: translational or post-translational control?
Although the transcription of the two mitochondrial-encoded NAD7 and NAD9 complex I subunits does not seem defective, these subunits are present only at low levels, as they can be detected immunologically in blots carrying the purified complex only. Two explanations for their low abundance can be proposed.
1The ms1 mutation is associated, directly or indirectly, with changes in mitochondrial protein synthesis or degradation rates. A dual function in mitochondrial transcription and translation has been reported for several yeast genes, for example PET54, involved in both the translation of cox3 and the processing of cox1 pre-mRNA ( Brown et al. 1994 ). Two mitochondrial tRNA synthetases also participate in intron splicing ( Lambowitz & Perlman 1990). In maize chloroplasts, the crp1 gene was shown to be necessary for both the translation of petA and petD and for the processing of petD from a polycistronic precursor ( Barkan et al. 1994 ). CRP1 processing and translation functions were suggested to be independent ( Fisk et al. 1999 ). Also, barley mutants lacking chloroplast ribosomes are impaired in the splicing of the rps12 and rpl2 transcripts ( Hübschmann et al. 1996 ). A general translation function for MS1 must be discarded, as if it were the case, the whole pattern of in organello mitochondrial protein synthesis would be altered. Therefore, a more specific function should be considered, for example restricted to the translation of NAD subunits. Some of the polypeptides missing in the two-dimensional in organello synthesis patterns are candidates for NAD7 (in the 40 kDa range) and NAD9 (in the 25–28 kDa range). However, immunodetection studies failed to establish a clear-cut relation between autoradiographic spots and immuno signals (not shown). Moreover, a role for MS1 in the control of mitochondrial gene translation cannot account for the absence of the 23 and 38 kDa nuclear-encoded subunits.
2Reduced amounts of either nuclear- or mitochondrial-encoded complex I subunits may result from an assembly defect, due to absence of NAD4. In chloroplasts, it is well known that unintegrated subunits of photosynthetic complexes are rapidly degraded (reviews by Rochaix 1996; Stern et al. 1997 ). In CMS mitochondrial mutants, it was previously proposed by Gutierres et al. (1997a) that the lack of NAD9 and of the 38 kDa subunit could result from mis-assembly of the peripheral arm due to loss of NAD7. Such a proposition was supported by the fact that in beef and Neurospora, all these subunits belong to the peripheral IP fraction of complex I ( Walker 1992; Weiss et al. 1991 ). At first view, this explanation is less relevant in the case of NMS1, as NAD4 is thought to belong to the membrane arm. However, in NCS2 maize, a similar absence of several peripheral subunits was observed following the deletion of the nad4 mitochondrial gene ( Karpova & Newton 1999). NAD4 and the other missing subunits could belong to the ‘connexion fraction’ linking the membrane and the peripheral arms, similar to that found in the minimal complex I of bacteria ( Friedrich et al. 1995 ).
As in other examples of mutations affecting multi-subunit complexes of plant organelles, it is difficult to distinguish direct effects from pleiotropic effects. It is, however, clear that the defective splicing of nad4 does not result from a structural complex I defect, as processing is normal in CMS mutants ( Fig. 2c) although their complex I presents similar features to NMS1.
Why is the phenotype more severe in NMS1 than in CMS?
The altered structure of complex I, a key coupling site of the respiratory chain, is likely to be directly responsible for the phenotypic abnormalities of N. sylvestris mutants. In contrast to the CMS-related polypeptides in other species, for example Petunia ( Conley & Hanson 1994; Young & Hanson 1987), sunflower ( Monéger et al. 1994 ) and common bean ( Abad et al. 1995 ), RNA and protein changes are observed in both sporophytic and gametophytic tissues. This may explain why the whole plant development is affected and not only male gametogenesis, as is the case in CMS of other species. This also seems to be the situation for the maize NCS2 complex I mutant, which is even more affected than N. sylvestris mutants as it presents aborted reproductive organs ( Yamato & Newton 1999). In NMS1 and in CMS, plant development could be due, at least in part, to the presence in low amount, about 10–20% relative to wild-type, of a NAD9-containing complex, called NAD9+ ( Fig. 9). In NMS1, this complex also contains NAD7 but whether or not it contains NAD4 could not be ascertained since no corresponding antiserum is available. Trace amounts of NAD4 could be synthesized from the normal 1.8 kb transcripts present at the sub-stoichiometric level ( Fig. 2c). Thus, whether the NAD9+ complex I from NMS1 is identical to wild-type complex I is not known. However, since silver-stained SDS–PAGE patterns of the fractions retained by the NAD9–protein A column present differences among genotypes, it is likely that the NAD9+ complex differs from wild-type in both NMS1 and CMS.
Determination of the exact composition of the NMS1 complex I, if present, requires further investigations using immuno-affinity columns coupled to other anti-complex I subunits, or blue native electrophoresis, as used in the case of the maize NCS2 mutant ( Karpova & Newton 1999). However, whatever their complex I amount and subunit composition, respiration measurements performed in vitro using purified mitochondria showed a near complete collapse of complex I activity in both mutants, which probably survive due to the enhanced activity of non-proton-pumping NAD(P)H dehydrogenases ( Gutierres et al. 1997a ; Sabar et al. 1998 ). Several non-exclusive possibilities may explain why NMS1 plants are much more severely affected than CMS as regards growth and fertility ( De Paepe et al. 1990 ).
1The two mutants probably differ in their membrane arm composition, NMS1 being essentially devoid of NAD4, but not CMS. In fact, whether CMS contains NAD4 is currently unknown, but this seems very likely, as the nad4 gene is not located in the region deleted in CMS mtDNA ( Lelandais et al. 1998 ) and its transcription pattern is normal ( Fig. 2c). In contrast, CMSII plants were recently found to be defective for NAD1 ( Gutierres et al. 1999 ). Thus, it appears that NAD4 absence could be more deleterious to plant cells than that of NAD1. At first view, this appears surprising as NAD1 is thought to be the ubiquinone-binding site ( Earley et al. 1987 ). In fact, NAD4 absence could result in the loss of the membrane subunits in plants as it does in human mutant cells carrying a point mutation in the nad4 gene ( Hofhaus & Attardi 1993). This is also the case for the NAD4-deficient NCS2 mutant of maize ( Karpova & Newton 1999), which has a most abnormal phenotype ( Newton & Coe 1986; Yamato & Newton 1999).
2In spite of the very similar absence in complex I activity presented by both mutants, external NAD(P)H dehydogenases are much less active in NMS1 than in CMS ( Sabar et al. 1998 ). In this respect also, NMS1 resembles CMS less than it does maize NCS2 plants, which do not present any activation of external NAD(P)H dehydrogenase activity ( Marienfeld & Newton 1994). This strongly suggests that activation of these pathways may directly or indirectly depend on the complex I composition.
3If some aberrant forms of NAD4, either corresponding to the first exon or to the unspliced 3.2 kb transcript, are integrated into complex I or directly into the mitochondrial membrane, they could affect mitochondrial functioning, as the URF13 polypeptide does in maize Texas CMS ( Levings 1993).
4Phenotypic abnormalities could be associated with the absence of some of the other in organello synthesized mitochondrial proteins.
Further molecular, biochemical and physiological studies of N. sylvestris nuclear and cytoplasmic complex I mutants would contribute to a better understanding of both mtDNA expression and complex I assembly and functioning.
The N. sylvestris parental line (T) is a fertile botanical line provided by the Institut des Tabacs (SEITA, Bergerac, France). NMS1, CMSI and CMSII mutants were obtained by protoplast culture as previously described ( Prat 1983). Plants were grown in greenhouses under a 16 h 24°C day and 8 h 17°C night regime. To induce pollen fertility, NMS1 plants were illuminated with SON-T AGRO (Philips) sodium lamps.
Leaf and anther mtRNAs were purified and separated by electrophoresis in 1.2% agarose, 6% formaldehyde as described in Pla et al. (1995) . All probes are homologous N. sylvestris sequences, except nad4 exon 3 (wheat), cob (maize) and rps13 (Petunia) as described in Lelandais et al. (1998) . The nad7 probe was N. sylvestris exon 3 ( Pla et al. 1995 ). N. sylvestris exons/introns nad4 probes were synthesized by PCR using oligonucleotides designed from the corresponding sequences of the B. campestris nad4 gene ( Gass et al. 1992 ):
Whole anthers collected 1 day before anthesis were fixed in 4% paraformaldehyde and 0.2% glutaraldehyde in 0.1 m sodium phosphate/0.05 m sucrose, for 24 h at 4°C, dehydrated in a graded ethanol series and embedded in LR white resin (London Resin Co, Woking, Surrey, UK). For immunocytochemistry, 1–2 mm sections, adhered to Biobond-treated slides, were pre-incubated in a blocking solution of 1% BSA in TBST buffer (Tris-buffered saline plus Tween), containing 0.1 m Tris–HCl pH 8, 0.9% NaCl, 1% Tween-20 for 30 min, and then incubated overnight with the following primary antibodies: Neurospora crassa complex I 49 kDa subunit (homologous to plant NAD7), wheat NAD9 subunit (both used at a 1:20 dilution) or potato formate dehydrogenase (FDH) used at a 1: 0 dilution (gifts from Professor H. Weiss, Dr J.M. Grienenberger and Dr R. Rémy, respectively). The antibody–antigen complexes were revealed using a goat anti-rabbit IgG (British Biocell International) as the secondary antibody, labelled with gold particles of 1 nm. Amplification of the gold labelling was achieved with the silver enhancing kit (British Biocell International) according to the manufacturer's protocol. Sections were counter-stained with toluidine blue and fuschin red in 50% aqueous ethanol, and photographed with bright field/epipolarization optics on a Zeiss Axiophot microscope. Three independent experiments were performed.
Anthers were fixed with 2% glutaraldehyde in 0.1 m phosphate buffer (pH 6.8) for 3.5 h at 4°C, washed extensively with the same buffer, post-fixed in 1% osmium tetroxide in phosphate buffer for 3 h and washed with distilled water. After a graded series of ethanol/propylene oxide dehydration steps, anthers were embedded in araldite resin for 60 h at 48°C. Ultra-thin sections were obtained with a LKB III ultramicrotome, stained in 9% uranyl acetate in methanol, post-stained in lead citrate (Reynolds) and observed with a Philips EM 208 (80 kV).
Complex I purification
Following isolation from fully expanded leaves, mitochondria were submitted to three cycles of freeze/thaw, and were incubated on a rotating wheel in resuspension buffer (25 m m Tris–HCl pH 7.8, 0.2 m NaCl, 1 m m EDTA, 0.1 m m EGTA, 3% Triton X-100, 0.5 m m phenylmethanesulphonide fluoride (PMSF)) for 90 min at room temperature (20–22°C) followed by 3 h at 4°C. PMSF was added every 90 min at the same concentration as initially. The extract was then centrifuged at 50 000 g for 30 min to remove membrane debris. For immunoaffinity chromatography, mitochondrial proteins including membrane complexes which remained in the supernatant were loaded on an anti-NAD9 column, prepared as described in Gutierres et al. (1999) . Analysis of the eluted polypeptides was performed by SDS–PAGE.
Purification of leaf and pollen mitochondria, SDS–PAGE electrophoresis and Western blotting of mitochondrial proteins
Leaf and pollen mitochondrial proteins were purified, electrophoresed and transferred for Western blotting as described by De Paepe et al. (1993) . High-resolution two-dimensional electrophoresis was performed and gels silver-stained as described by Colas des Francs-Small et al. (1992) ; at least four different mitochondrial protein preparations were analysed for each genotype.
We thank Professor H. Weiss (Institut für Biochemie der Universität, Dusseldorf), Dr R. Rémy (Institut de Biotechnologie des Plantes, Orsay) and Dr J.M. Grienenberger (Institut de Biologie Moléculaire des Plantes, Strasbourg) for the gift of antibodies. We are grateful to M. Hodges for critical reading of the manuscript, to O. Roche for cytological work and to D. Froger and R. Boyer for photographic art. This work was supported by the CNRS.