Introduction

Cucumber (Cucumis sativus L.) and melon (C. melo L.) are unique among all plants in that the three genomes show differential transmission: maternal for chloroplast, paternal for mitochondrial, and biparental for the nuclear DNA (Havey 1997; Havey et al. 1998); and they possess the largest mitochondrial genomes among all eukaryotes (Ward et al. 1981). Biochemical and molecular analyses established that the cucumber mitochondrial genome has accumulated a plethora of short repetitive DNAs spread throughout the genome (Ward et al. 1981; Lilly and Havey 2001). Plants of the genus Cucumis represent a unique genetic system among herbaceous dicotyledonous plants because mitochondrially encoded traits can be easily separated from traits conditioned by the chloroplast genome, by taking advantage of differential transmission of the organellar genomes (Havey et al. 1998). Examples of plant phenotypes conditioned by the mitochondrial genome include cytoplasmic male sterility in many species (Kaul 1988), the nonchromosome striped mutants of maize (Coe 1983; Newton and Coe 1986; Newton 1995), the mosaic (MSC) mutants of cucumber (Malepszy et al. 1996; Lilly et al. 2001), and chm-conditioned mitochondrial rearrangements of Arabidopsis thaliana (Martinez-Zapater et al. 1992; Sakamoto et al. 1996).

The highly inbred (>S18) cucumber line B originated from the Polish cultivar Borszczagowski and was selected during inbreeding for uniform regeneration from various cell-culture systems (Burza and Malepszy 1995a). After plants of the wild-type inbred line B are passed through cell cultures, regenerated (R0) plants predominately have a normal phenotype (Burza and Malepszy 1995b; Plader et al. 1998). However, after self-pollination of R0 plants, occasional first (R1) or second (R2) generation selfed progenies from wild-type R0 plants show paternally transmitted, strongly MSC phenotypes expressed in seedlings and adult plants (Malepszy et al. 1996; Ladyzynski et al. 2002). All crosses, backcrosses, and self-pollinations of the wild type by MSC plants showed paternal transmission (Malepszy et al. 1996) and imprinting of paternal nuclear alleles was eliminated (Lilly et al. 2001). We demonstrated that MSC plants possess a deletion in the mitochondrial genome, relative to the parental inbred line B (Lilly et al. 2001), and that testcrosses with MSC as the male parent to wild-type plants produce wild-type seedlings at or below 1% (Malepszy et al. 1996; Lilly et al. 2001). MSC plants are heteroplasmic for both mutant and wild-type mitochondrial genomes, with the former predominating; and wild-type progenies result from sorting of the mitochondrial genome in MSC pollen (Lilly et al. 2001). The passage of cucumber through cell culture may allow mitochondria carrying the deleterious MSC lesion(s) to sort by their reducing negative effects, as previously observed in maize (Gu et al. 1994). Another explanation is that passage through cell culture may induce mutations or be conducive to recombination among direct repeats in the mitochondrial DNA to produce structural rearrangements or deletions affecting the expression of mitochondrial genes. If this latter scenario were true, the passage of cucumber through cell culture may offer a unique opportunity to affect mitochondrial gene expression. In this study, we evaluated for mitochondrial-DNA rearrangements affecting gene expression in MSC and wild-type lines and demonstrated that independently derived MSC lines possess different combinations of mitochondrial polymorphisms.

Experimental procedures

Origin of MSC lines

The origins of wild-type inbred line B, the cultivar Calypso, and MSC lines 11, 16, and 19 have been described by Lilly et al. (2001). Additional MSC and wild-type lines (Fig. 1, Table 1) were produced after self-pollination of plants regenerated from independent cell-culture experiments (Malepszy et al. 1996; Ladyzynski et al. 2002). Wild-type sorters from MSC16 or MSC sorters from wild-type lines were identified at the seedling stage and DNA was isolated using a minipreparation (Lilly et al. 2001).

Fig. 1
figure 1

Examples of phenotypic differences for leaves (A) and fruit (B) among mosaic (MSC) and wild-type (DPR1) cucumber lines regenerated from cell cultures started from the highly inbred line B. The origin of each line is listed in Table 1

Full size image
Table 1 Origins of wild-type and mosaic (MSC) cucumber lines and polymorphisms detected in their mitochondrial genomes. Cell cultures were described by Ladyzynski et al. (2002): LC direct regeneration from leaf callus, Salt regeneration from leaf callus after selection for salt tolerance, Cytokinin cytokinin-dependent embryogenic suspension cultures, Meristem liquid meristematic culture (shoot primordia culture), Protoplast direct protoplast regeneration. Polymorphisms are shown in Figs. 1, 2. Fragments were wild type (WT), MSC16 (MSC), or both (WT/MSC). New Unique fragments not observed in either wild type or MSC16
Full size table

Hybridization of mitochondrial Arabidopsis cosmids and cucumber clones

Total DNA was isolated from cucumber populations using standard CTAB extraction (Havey 1997). The DNAs of MSC3, MSC11, MSC16, MSC19, and wild-type lines B, LCST1, LCST2, and Calypso were digested with 12 different restriction endonucleases (BamHI, DraI, EcoRI, HindIII, SacI, XbaI, AluI, HaeIII, CfoI, MspI, MboI, MseI), subjected to electrophoresis though 0.8% (six-base cutters) or 2% (four-base cutters) agarose gels, and transferred to Nylon membranes (Zetaprobe GT; Bio-Rad, Hercules, Calif.). Fifteen cosmids covering the entire A. thaliana mitochondrial genome (Klein et al. 1994) were individually radiolabeled by nick translation (Invitrogen, Carlsbad, Calif.) and hybridized overnight to blots as described by Havey et al. (1998). Blots were washed once for 5 min in 2× SSC with 0.1% SDS at room temperature and twice for 10 min in 0.1× SSC with 0.1% SDS at 65 °C. Specific coding regions (nad5-exon2, rpl5, rrn18) on cosmids that revealed polymorphisms were amplified by PCR, using primers designed based on A. thaliana (Unseld et al. 1997) and used as probes for hybridizations. Cucumber primers amplifying from the P1 and JLV5 mitochondrial regions (Table 2; Lilly et al. 2001) were P1 (5′-AGG CTC GAC CCC ATT CTA CT-3′) and P2 (5′-TTA GCT GGA CCG ATT TAC GG-3′) and MF1 (5′-GCA GAA GCA ACC TTC ACC TCC GAG AGT TAG-3′), JLV5R (5′-TAA GAC TGC CCA CTA CGC TCC GTG CAA-3′), and rpl5R2 (5′-GCC CTC GGA GAG TGC TTT CTC GT-3′). Primers specific for mitochondrial rrn18 were designed from the mitochondrial region V7 (Neefs et al. 1993). PCR products were gel-purified, cloned by TA-tailing (Invitrogen), sequenced to verify similarities, and used for hybridizations.

Table 2 Genbank accessions for sequenced mitochondrial genomic regions of cucumber
Full size table

Synthesis of mitochondrial DNA libraries

We synthesized mitochondrial genomic libraries of wild-type inbred B and MSC16 from DNA extracted using DNAseI treatments and Percoll-gradient centrifugation (Klein et al. 1994; Lilly and Havey 2001). Cucumber mitochondrial DNA was partially digested with Sau3A, subjected to electrophoresis, and fragments of approximately 10–20 kb were excised from the gel. DNA was purified using the QiaEX gel extraction kit (Qiagen, Valencia, Calif.) and ligated into a dephosphorylated BamHI-digested pUC18 vector, according to the manufacturer′s directions (Amersham, Piscataway, N.J.). Standard electroporation into bacteria and blue-white selection were performed (Sambrook et al. 1989). The average insert size was 15 kb, requiring 1,375 clones for 3× coverage at 99% probability. Two wild-type libraries from Calypso (1,152 clones) and line B (3,072 clones) and one MSC16 library (3,072 clones) were created. Colonies from plates were spotted onto nylon membranes, allowed to grow overnight, and lyzed (Nizetic et al. 1991; Lilly and Havey 2001). Hybridization of the mitochondrial genomic or cDNA clones to these filters revealed genomic clones carrying homologous regions. These clones were aligned by end-sequencing and restriction-enzyme digests. Polymorphic regions were subcloned, sequenced, and compared with the wild-type region in inbred B to establish the genetic basis of rearrangements or deletions in MSC16. Sequencing was completed using random transposon insertions (EZ::Kn2 system; Epicentre, Madison, Wis.). Cycle sequencing reactions were performed (Lilly and Havey 2001) and analyzed on an automated DNA sequencer at the University of Wisconsin Biotechnology Center. Sequence data were edited and trimmed and contigs aligned with Sequencer 3.1.2 (Genecodes, Ann Arbor, Mich.). GeneBank accessions are listed in Table 2.

Mitochondrial rpl5 transcription

Total RNA was isolated from cucumber leaf tissue (Lilly et al. 2001), 20 μg were subjected to electrophoresis on 1.2% formaldehyde-agarose gels (Sambrook et al. 1989), and transferred to Zetaprobe membranes. The membranes were then hybridized overnight at 65 °C in 1 mM EDTA, 0.5 M NaHPO4, and 7% SDS, after which randomly primed, 32P-labeled probes were added. The rpl5 coding sequence and the untranslated 5′ region of rpl5 (corresponding to the bases −3,280 to −2,482 of the wild-type rpl5 genomic region; GenBank accession AY258275) were subcloned as described earlier and hybridized to the RNA-gel blots. After hybridization, membranes were washed twice at 65 °C for 20 min in 1 mM EDTA, 40 mM NaHPO4, 5% SDS, and twice in 1 mM EDTA, 40 mM NaHPO4, 1% SDS, and then subjected to autoradiography.

Mitochondrial rpl5 antibody production and immunoblotting

RT-PCR with primers carrying XhoI restriction sites was used to isolate the cucumber rpl5 cDNA from line B. Primer sequences were 5′-CAT ATG GGT CCG AGC CTA CAA CT-3′ (rpl5-NdeI) and 5′-CTC GAG TTA TTA CAA ATT CAA ATT CCA T-3′ (rpl5-XhoI). RT-PCR products were cloned by TA-tailing and 15 clones were sequenced in both directions. Six editing sites (all C-to-U) were shared among the rpl5 cDNAs from MSC16 and line B. One of these edited clones from line B was chosen and directionally placed, using the XhoI and NdeI sites, into the expression vector pET28 (Novagen, Madison, Wis.). The correct construct was verified by sequencing and was introduced into Escherichia coli expression strain BL21 Codon Plus (Stratagene, La Jolla, Calif.) by heat-shock transformation. Protein expression was induced using isopropyl-β-d-thiogalactopyranoside and, after 4 h induction, the recombinant protein was collected from the insoluble fraction by centrifugation. The pellet was subjected to electrophoresis on SDS-PAGE gels, visualized after staining of adjacent lanes with Coomassie blue, and an adjacent lane of the gel containing the unstained RPL5 was excised, emulsified, and injected into rabbit at the Polyclonal Antibody Service, University of Wisconsin, Madison, Wis. The second bleeding was affinity-purified (Sambrook et al. 1989) and used for Western blots. Mouse monoclonal antibody Anti-PORIN was kindly provided by Dr. Tom Elthon, University of Nebraska, USA. To prepare mitochondrial protein extracts, mitochondria were purified as described above and resuspended in sample buffer (McCabe et al. 1998). For immunoblotting analysis, 30–40 μg of total plant proteins or mitochondrial proteins were loaded on 15% SDS-PAGE gels, subjected to electrophoresis, and blotted onto Hybond C Extra nitrocellulose membrane (Amersham). Membranes were blocked using 5% milk powder and probed with antibodies, according to the membrane manufacturer′s instructions. RPL5 antibody was diluted 1:1,000 and PORIN 1:500. Alkaline phosphatase conjugated anti-rabbit or anti-mouse antibodies (Kirkegaard and Perry Laboratories, Gaithersburg, Md.) were used as secondary antibodies. Both primary and secondary antibodies were diluted in PBS containing 0.1% Tween. Immunoreactive proteins were visualized using the 1-Step NBT/BCIP system (Pierce, Rockford, Ill.).

Results

MSC and wild-type cucumbers regenerated from cell cultures possess rearrangements in the mitochondrial genome

We revealed five structural polymorphisms (Fig. 2) in the mitochondrial genome among wild-type and MSC lines (Fig. 1, Table 1), all tracing back to the highly inbred line B. In addition to the previously described P1 and JLV5 regions (Lilly et al. 2001), new polymorphisms were revealed by rrn18, nad5-exon2, and rpl5 (Fig. 2). All polymorphisms were paternally transmitted (data not shown). We assembled and sequenced mitochondrial contigs across the polymorphic regions. The polymorphisms revealed by rrn18 and nad5-exon2 were due to a rearrangement placing these two coding regions together (Fig. 3, Table 2). For the polymorphism revealed by rpl5, all accessions possessed a 5.2-kb SacI fragment corresponding to the functional rpl5 gene (Fig. 2, Table 2). Wild-type lines B and Calypso possessed a pseudo-rpl5 gene (as defined by premature stop codons) carried on a 15.0-kb SacI fragment (Fig. 2, Table 2). MSC16 and the wild-type lines LCST1 and LCST2 were regenerated from the same leaf-callus experiment (Table 1) and did not possess this pseudo-rpl5 (Fig. 2). MSC16 possessed a unique 6.2-kb SacI fragment (Fig. 2), due to a rearrangement placing a duplicated rpl5 gene next to the forward junction of the JLV5 region (Fig. 3, Table 2). JLV5 is a 15.1-kb region missing in MSC16, present in wild-type line B, and sorting with the MSC phenotype (Lilly et al. 2001). Hybridization of the region (P1) adjacent to the front junction of the JLV5 region revealed polymorphisms between MSC and wild-type lines (Fig. 2). Sequencing an expanded contig of the wild-type JLV5 region (Lilly et al. 2001; Table 2) established that the forward and reverse junctions were located near regions of previously identified short repetitive DNA motifs in cucumber mitochondria (Lilly and Havey 2001). For example, repeat motifs 1, 3, 4, and 5 were located near the front junction of the JLV5 deletion at positions −150 to −110 and near the rear junction at positions −120 to −81. The exact junction points were not homologous to these previously described motifs (Lilly and Havey 2001). The rearrangement that fused the genomic region carrying the rpl5 with the forward junction of the JLV5 deletion occurred at ATTTAGAAGCAA. The rear junction was at or near CTCATTAACGAT. The exact sequence at the rear junction could not be established, because we could not isolate one of the wild-type regions involved in the rearrangement.

Fig. 2
figure 2

Autoradiograms showing RFLPs in the mitochondrial DNA between wild type inbred line B (lane 1), MSC16 (lane 2), wild-type tissue culture lines 1 and 2 (LCST1 in lane 3, LCST2 in lane 4), and wild-type cultivar Calypso (lane 5). Restriction enzymes and probes revealing polymorphisms are shown below the autoradiograms. RFLPs were revealed by cosmids from the Arabidopsis mitochondrial genome carrying the rrn18, rpl5, and nad5-exon2 coding regions (Unseld et al. 1997). The JLV5 deletion was described by Lilly et al. (2001), but was not characterized in lines LCST1 and LCST2. P1 is the genomic region adjacent to the JLV5-DEL rearrangement (Lilly et al. 2001). Approximate fragment sizes (in kilobases) are shown at the right

Full size image
Fig. 3
figure 3

Schematic diagrams of mitochondrial-DNA rearrangements near rrn18-rrn5 with nad5-exon2 (A) and rpl5 with JLV5 (B) in MSC16 relative to the wild-type parental inbred line B. Regions or genes are indicated above the lines

Full size image

We evaluated for the presence of the rearranged rpl5 region among wild-type sorters by intercrossing among MSC16 plants. The frequency of wild-type progenies from MSC16 was 3.3% (24 out of 720). In all cases, DNA-gel blots revealed that the wild-type rpl5 mitochondrial region was recovered with the wild-type phenotype (data not shown). Hybridizations of the P1 region to HindIII-digested DNA of the wild-type line LCST1 revealed both unique (11.0 kb) and MSC16-specific (3.4 kb) fragments (Fig. 2); and intercrossing among LCST1 plants produced 504 wild-type and seven MSC progenies (1.4%). Although line LCST2 possessed no detectable levels of the 3.4-kb MSC-specific fragment in the P1 region (Fig. 2), intercrossing among plants of this line produced 469 wild-type progenies and one MSC progeny (0.2%).

We used primers MF1 and rpl5R2 to amplify a 2076-bp fragment across the front junction of the rpl5-JLV5 rearrangement from MSC16 and wild-type lines B, LCST1, LCST2, and Calypso. Amplicons were observed from MSC16, LCST1, and LCST2; and no fragments were amplified from B and Calypso (Fig. 4). Amplicons from LCST1 and LCST2 were less intense than MSC16, as expected because hybridization of the P1 region revealed weak RFLPs on autoradiograms (Fig. 2). To evaluate for the presence of substoichiometric levels of this rearranged fragment, we loaded onto an agarose gel 45 times the volume of the PCR reaction from B and Calypso (22.5 μl), as compared with MSC16, LCST1, and LCST2 (0.5 μl from each PCR reaction). The autoradiogram from hybridization of the internal rpl5 probe to this gel blot clearly revealed the presence of the 2076-bp amplicon from B, but not Calypso (Fig. 4).

Fig. 4
figure 4

Autoradiogram (top) from the hybridization of rpl5 to PCR-amplified DNA, using primers flanking the rpl5 rearrangement in the MSC16 mitochondrial genome. Lane contents are shown at the top and are listed in Table 1. The approximate size of the hybridizing fragment (in kilobases) is shown at the right. Each lane in the gel (bottom) carries 5 μl of the PCR reaction. The gel blotted to produce the autoradiogram (top) was loaded with 22.5 μl from the PCR reactions of B and Calypso and 0.5 μl from the PCR reactions of MSC16, LCST1, and LCST2

Full size image

Gene expression and protein accumulation in MSC16

We used mitochondrial-RNA-gel blots to evaluate for rpl5 expression in MSC16 and wild-type lines. Transcripts for rpl5 were 4.4 kb and 2.0 kb, both of which were reduced in MSC16 relative to wild-type lines (Fig. 5). We cloned the region upstream from the functional rpl5 from Calypso corresponding to positions −3,280 to −2,482 relative to the rpl5 start codon. Because the recombination event that brought together the P1 and rpl5 regions in MSC16 occurred between positions −1,379 and −1,390, this probe hybridized upstream from the wild-type rpl5, not the rearranged duplicated rpl5 region in MSC16 (Fig. 2), and revealed only the 4.4-kb transcript (Fig. 5). The amount of the 2.0-kb rpl5 transcript was less in MSC16 than in the wild-type lines, but showed a stronger signal intensity than the 4.4-kb transcript (Fig. 5). In spite of reduced transcripts, Western analyses of the RPL5 protein revealed no differences between MSC16 and B (Fig. 6). The nuclear-encoded, cytoplasmically translated, and mitochondrially imported PORIN protein also showed no differences (Fig. 6), indicating that the MSC phenotype does not significantly reduce the transcription or translation of this nuclear-encoded gene.

Fig. 5
figure 5

RNA-gel blot analyses of rpl5 and the untranslated region immediately upstream from rpl5 (5′UTR rpl5). Lane contents are (from the left) wild-type inbred line B, MSC16, and wild-type line LCST1. The approximate sizes of transcripts (in kilobases) are shown at the right. The same blot was hybridized with a mitochondrial rrn18 (V7 fragment) probe and the gel was stained with ethidium bromide to check equal RNA loading

Full size image
Fig. 6
figure 6

Western analyses revealing relative amounts of mitochondrially encoded RPL5 and nuclear-encoded PORIN in mitochondrial protein extracts from inbred B and MSC16. The approximate sizes of polypeptides (in kiloDaltons) are shown at the right

Full size image

Independently generated MSC plants possess different mitochondrial rearrangements

We scored all mitochondrial polymorphisms among MSC and wild-type lines generated from independent cell-culture experiments, all started from the highly inbred line B, and revealed eight unique patterns. MSC16 and MSC19 were both regenerated from the same leaf-callus experiment and shared all polymorphisms (Table 1; Lilly et al. 2001). MSC9 was regenerated after protoplast culture and shared all polymorphisms with MSC16 and MSC19 (Table 1, Fig. 7). MSC12 did not possess the same 6.2-kb rpl5 fragment observed in MSC9, MSC16, and MSC19 (Table 1). However, we were able to amplify this rearranged rpl5 fragment from MSC12, although the intensity of the amplicon was reduced (Fig. 7). MSC3 and four groups of wild-type lines: (1) LC1and CDS1, (2) CDS2, CDS4, and DPR2, (3) CDS3, and (4) LCST1 and LCST2 possessed unique combinations of mitochondrial polymorphisms (Table 1).

Fig. 7
figure 7

Agarose gel showing PCR-amplified DNA fragments using primers MF1, JLV5R, and rpl5R2 (sequences given in the Materials and methods) and (from the left) genomic DNA of MSC16, wild-type line B, mixtures of DNA from B and MSC16, and wild type cucumber cultivars and independently arising MSC lines. The approximate sizes of fragments (in kilobases) are shown at the right

Full size image

Discussion

We revealed structural polymorphisms (Fig. 2) in the mitochondrial genome among wild-type and MSC lines (Table 1) after hybridizing DNA-gel blots with ordered cosmids covering the entire Arabidopsis mitochondrial genome (Klein et al. 1994). Only the polymorphism revealed by rpl5 was unique to MSC16 and not present in the wild-type lines (Fig. 2). This polymorphism resulted from duplication and fusion of the rpl5 genomic region with the previously described JLV5 region (Fig. 3; Lilly et al. 2001). We sequenced across the rearrangement junctions to reveal pockets of repetitive DNAs, consistent with our hypothesis that the MSC phenotype is conditioned by mitochondrial DNA rearrangements due to recombination among short direct repetitive DNA motifs (Lilly et al. 2001). We demonstrated that the rearrangement near rpl5 existed as a sublimon in B, but was not detected in Calypso (Fig. 4). We evaluated for the presence versus absence of the rearranged rpl5 region among wild-type sorters from self-pollination of MSC16, taking advantage of the highly homozygous nuclear genotype to avoid the segregation of nuclear genes affecting the prevalence of sublimons (Hartmann et al. 1992; He et al. 1995; Janska et al. 1998) or conditioning rearrangements (Goff et al. 2002). All 24 wild-type progenies out of 720 selfed progenies possessed the wild-type rpl5 mitochondrial region. The simplest explanation is that wild-type progenies resulted from sorting among wild-type and MSC mitochondria, with the wild type predominating.

We scored the mitochondrial polymorphisms among the wild type and MSC independently regenerated from different cell culture experiments (Table 1) and revealed eight different patterns (Fig. 7, Table 1). These results indicate that the MSC phenotype does not trace back to a single sublimon existing in the B mitochondrial genome and there may be numerous rearranged mitochondrial genomes existing substoichiometrically in inbred B, as reported in Phaseolus (Janska and Mackenzie 1993; Janska et al. 1998; Arrieta-Montiel et al. 2001; Woloszynska et al. 2001). The prevalence of these sublimons may increase during tissue culture (Gu et al. 1994, Kanazawa et al. 1994). This conclusion is supported by the fact that MSC9 and MSC16 shared all mitochondrial polymorphisms, even though these lines were regenerated from independent cell-culture experiments. Another explanation is that passage through cell culture generates de novo rearrangements in the cucumber mitochondrial genome.

RNA-gel blots hybridized with rpl5 revealed two fragments at approximately 2.0 kb and 4.4 kb, both of which showed reduced transcription in MSC16 relative to the wild-type lines (Fig. 5). The longer transcript possessed homology with the genomic region upstream from rpl5. Longer mitochondrial transcripts have been reported in cucumber (Stern and Newton 1985) and other plants (Hoffmann et al. 1999) and may undergo post-transcriptional modifications to reduce their size (de Souza et al. 1991). In spite of the reduced amounts of rpl5 transcripts in MSC16 (Fig. 5), we observed no differences between MSC16 and wild-type lines for the RPL5 protein (Fig. 6). The reduced transcription of rpl5 could be offset by longer protein longevities (Giegé et al. 2000), allowing for the accumulation of similar amounts of protein. Regardless, we were not able to establish that the rearrangement near rpl5 is responsible for the MSC16 phenotype; and its genetic basis remains an enigma. Potential causes of the MSC phenotype not revealed by our analyses could include rearrangements affecting genes present in cucumber but missing from Arabidopsis mitochondrial DNA, such as rps1, rps10, rps13, and sdh3 (Adams et al. 2002), relatively short deletions, insertions, or rearrangements not revealed by DNA-gel blot analyses, or post-transcriptional (Sutton et al. 1993; Maier et al. 1996) or post-translational (Sarria et al. 1998) modifications.

When the highly inbred cucumber line B is passed through cell-culture systems, regenerated (R0) plants are predominately wild type with relatively few somaclonal variants (Burza and Malepszy 1995a; Plader et al. 1998; Ladyzynski et al. 2002). After self-pollination of these wild-type R0 plants, R1 plants are occasionally produced showing the strongly MSC phenotype. Selfed progenies from MSC plants almost exclusively possess the MSC phenotype, with wild-type progenies appearing at about 3% for MSC16. A salient question is why does the MSC phenotype appear after self-pollination of wild-type R0 plants? We demonstrated that the MSC phenotype is paternally transmitted, is not conditioned by paternal imprinting, and is associated with rearrangements in the mitochondrial genome (Malepszy et al. 1996; Lilly et al. 2001; Fig. 2). The passage of cucumber through cell culture may allow for the sorting of previously rearranged sublimons or induce new rearrangements in the cucumber mitochondrial genome. The wild-type R0 plants could possess a mixture of wild-type and rearranged mitochondrial genomes with enough wild-type genomes to condition the wild-type phenotype. The R0 plant producing LCST1 was regenerated from the same leaf-callus experiment that produced MSC11 (Table 1). LCST1 possessed a mixture of wild-type and rearranged mitochondrial genomes (Fig. 2) and produced 0.2% MSC progenies. Abreu et al. (1982) observed relatively few, large mitochondria in cucumber microspores. These few, large mitochondria may represent a bottleneck and contribute to significant random drift among mitochondria incorporated into the sperm cells of cucumber pollen and transferred by the male gametophyte to progenies. The appearance of the MSC phenotype among sexual progenies from wild-type R0 plants would occur when significant numbers of the randomly transferred mitochondria carry the MSC mitochondrial lesion(s).

The MSC16 phenotype is associated with reduced transcription of rpl5 (Fig. 5). Similar DNA rearrangements affecting mitochondrial gene expression have been documented in maize (Feiler and Newton 1987; Lauer et al. 1990). Adjacent mitochondrial coding regions can be co-transcribed, producing relatively long transcripts (Hoffmann et al. 1999). DNA rearrangements upstream from coding regions could affect the cis-acting promoter region, significantly reducing transcription. Even if the rearrangement completely stopped transcription, some wild-type transcript would likely be produced from wild-type sublimons, enabling the plant to survive, depending on the frequency of wild-type mitochondrial genomes and the transcript and protein longevities (Giegé et al. 2000). Presently, there is no efficient way to mutate or alter the expression of specific mitochondrial genes (Jacobs 2001). If the passage of cucumber through cell-culture systems caused de novo rearrangements, or allowed for the sorting of a relatively large number of rearranged substoichiometric mitochondrial DNAs, it could provide a unique opportunity to affect the expression of mitochondrial genes. Relatively large numbers of independently generated MSC lines can be produced after cell culture (Ladyzynski et al. 2002). MSC lines can then be evaluated for rearrangements affecting specific mitochondrial genes to study interactions between mitochondrial and nuclear genes (Moneger et al. 1994; Karpova et al. 2002).