A 1988 paper published in Journal of Molecular Evolution presented a fundamental paradox that still dominates the field of plant mitochondrial genome evolution (Palmer and Herbon 1988; https://doi.org/10.1007/BF02143500). These were the early days of molecular biology and genomics; Sanger sequencing was only 11 years old (Sanger et al. 1977), and the first complete flowering plant mitochondrial sequence (Arabidopsis thaliana) was still 9 years away (Unseld et al. 1997). In the same year, 1988, the Office of Human Genome Research was established at NIH with the goal of eventually sequencing the human genome. Mammalian mitochondrial genomes had been sequenced, and were known to be small, invariant in gene order, and to have a high mutation rate (Brown et al. 1979), although it was known that the gene order was different in Drosophila yakuba than in mammals (Clary et al. 1982; Clary and Wolstenholme 1985). However, Arnold Bendich and co-workers had established that the mitochondrial genomes of flowering plants were very large and variable in size (Ward et al. 1981), the other important conundrum that still vexes the field. It was also known that much of the genome was non-protein-coding, and that gene orders are not well-conserved (Makaroff and Palmer 1987). The time was ripe for a thorough study of plant mitochondrial genome organization to understand the evolution of these two different types of changes: point mutations and rearrangements. To solve the problem, species needed to be compared that were different enough to observe a reasonable number of changes, but not so distantly related as to encounter multi-state problems. Jeffrey Palmer, then at the University of Michigan, chose six species in the Brassicaceae: radish, turnip, cabbage, rutabaga, black mustard, and white mustard, (Raphanus sativus, Brassica campestris (now known as Brassica rapa), Brassica oleracea, Brassica napus, Brassica nigra, and Brassica hirta (now known as Sinapis alba), respectively).

In the present day, where samples can be sent off for complete genome sequencing, it is hard to appreciate the significant amount of work that had to be done for these comparisons. PCR was not widely used yet, because publication of the method for amplification using thermostable polymerases occurred that same year (Saiki et al. 1988). Instead, Palmer and Herbon purified mitochondrial DNA from green tissues and characterized the structures by restriction mapping. With these genomes being over 200 kb, this required multiple rounds of restriction digestion, gel electrophoresis, Southern blotting, and hybridization with cloned fragments primarily from B. rapa. The restriction maps were then compared to find their similarities and differences.

Three findings emerged from this work, any one of which would have made it an important paper. First, they were able to compare the synteny of the genomes, and build a phylogenetic tree based on the trait of genome rearrangements, rather than synonymous substitutions. Second, they showed that long, non-tandem repeated sequences are not conserved even in close relatives, and are not completely universal. The one result I look most forward to seeing confirmed or refuted by complete genome sequencing is that Sinapis alba lacks a long, non-tandem repeat (Sinapis arvensis has one). Third, they were able to measure point mutation rates by examining the differences in restriction site cutting, presumably due to single base changes. By comparing the number of point mutations to the number of rearrangements separating related species, they were able to show that rearrangements occur at a higher rate than point mutations, a startling and counter-intuitive result.

Figure 1 illustrates the complex and detailed data provided. These restriction maps were produced by first purifying the mitochondrial DNA, then digesting it with multiple restriction enzymes, and resolving the different size bands by gel electrophoresis. Because most of the fragments of the Brassica rapa genome had been cloned, bands in the other species could be identified by blotting the gel and hybridizing with multiple B. rapa probes (Southern 1975). Each of these five enzymes cuts the genome approximately 30 times, so Palmer and Herbon analyzed 150 bands for each species. They were also able to identify large non-tandem repeats responsible for frequent isomerization of the genomes by recombination (2 kb in B. rapa, napus and oleracea, 9.7 kb in R. sativa, and 6.6 kb in B. nigra). One can only imagine how difficult it was to deal with the background caused by hybridization from the ubiquitous non-tandem repeats of up to about 500 bp found in these genomes (B. nigra, for example, has over 80 such repeats larger than 50 bp (Wynn and Christensen 2019)). Nevertheless, in spite of these technical difficulties, their restriction maps are excellent. Their BglI map of B. rapa compared to accession JF920285 (Chang et al. 2011) shows that the sequence agrees with the restriction map to a remarkable degree, both in sizes and order.

Fig. 1
figure 1

Originally published as Fig. 1 in Palmer and Herbon (1988). Restriction maps of the master chromosome of three Brassica mtDNAs. The circular maps are shown linearized at a BglI site within the 2-kb repeat. Arrows indicate the two copies of this repeat. Numbers between arrows indicate sizes of subgenomic circles. Restriction fragment summations are given at the right of the maps. Filled triangles indicate eight length mutations, which are arbitrarily shown only as insertions. Variable restrictions sites resulting from length mutation and point mutation are indicated with filled circles and stars, respectively. The B. campestris map is modified from (Palmer and Shields 1984)

Full size image

In Fig. 2, they compared the maps of the relatively closely related species B. rapa, B. napus, and B. oleracea. They were able to define multiple rearrangements between these species which defined blocks of synteny and specific inversions that scrambled the genomes. B. napus is an allotetraploid of B. rapa and B. oleracea (Cai et al. 2014; U N 1935). Mitochondrial sequences (Grewe et al. 2014) suggest that B. napus allotetraploids have occurred multiple times with different maternal parents, leading to several different cultivars, and agreeing with the Palmer lab’s previous findings using chloroplast DNA (Palmer et al. 1983). The map of the B. napus cultivar (rutabaga) in Palmer and Herbon (1988) appears to be most similar to the sequenced cultivar B. napus napus, accession number AP006444.1 (Handa 2003) which is very similar to the B. oleracea sequenced genome (Grewe et al. 2014).

Fig. 2
figure 2

Originally published as Fig. 3 in Palmer and Herbon (1988). Relative arrangement of similar sequences in three Brassica mtDNAs. Numbered horizontal arrows above and below the complete SalI maps indicate blocks of sequences whose arrangement has been conserved between genomes. The crossing lines between SalI fragments connect regions of cross-hybridization. Double-headed vertical arrows indicate inversion endpoints. Horizontal arrowheads indicate recombination repeats

Full size image

Palmer and Herbon further compared the arrangements of these genomes with the additional species mentioned above, and were able to characterize the rearrangements that separated the genomes. A parsimony-based analysis using rearrangements as a trait allowed them to construct the phylogenetic tree in Fig. 3. An analysis of this type would be done today using sequence data, with software tools such as GRAPPA (Bader et al. 2001) or GRIMM (Tesler 2002) that only became available in the decade following the Palmer and Herbon (1988) paper. Chromosomal rearrangements had been used as a taxonomic trait previously in Drosophila, using the banding patterns of polytene chromosomes (Dobzhansky and Sturtevant 1938; Sturtevant and Dobzhansky 1936). However, I am not aware of any earlier attempt to use molecular biology data to determine rearrangements for phylogenetic purposes than the work of Palmer and Herbon (1988). It is also not clear to me which method is more tedious. Palmer and Herbon’s findings showed that it could be done, opening the door to subsequent work in metazoans and fungi (Blanchette et al. 1999; Boore and Brown 1998; Sankoff et al. 1992; Smith et al. 1993) and ultimately in flowering plants (Cole et al. 2018; Darracq et al. 2010).

Fig. 3
figure 3

Originally published as Fig. 7 in Palmer and Herbon (1988). Phylogenetic history of mtDNA rearrangements in Brassica. Top Cytoplasmic phylogeny for Brassica based on cp-DNA restriction site mutations (Palmer et al. 1983). This phylogeny is cladistically derived and is not intended to convey divergence times. Numbers of rearrangements are given relative to the reference genome, B. campestris, except for the ten rearrangements that have been shown (Makaroff and Palmer 1988) to distinguish the CMS and fertile mtDNAs of R. sativa. Bottom Sizes of master chromosomes and (where present) subgenomic circles resulting from high frequency recombination at the indicated recombination repeats

Full size image

The final results that make this a landmark paper are mapping Restriction Fragment Length Polymorphisms (RFLPs) in order to measure mutation rates. For this, Palmer and Herbon added an additional 15 restriction enzymes to their analysis, and used Southern blotting against double-digested DNA to find RFLPs. What they found was remarkable. Point mutation rates are very low in plant mitochondria, lower even than in the nucleus, while the rearrangement rates are quite high, essentially scrambling the genomes of even close relatives.

Paradoxes are important in science. Molecular biology was founded by scientists who were inspired by the works of Erwin Schrödinger and Niels Bohr, who had seen physics move forward by analyzing paradoxes (Stent 1968). Nature, they thought, is not paradoxical; it operates according to natural laws. A paradox reveals where our understanding is flawed, and offers an opportunity for important research. Finding paradoxes is important and exciting, because by studying and analyzing the phenomena that appear paradoxical we have the opportunity to learn important new things about nature.

Palmer and Herbon’s work, along with Arnold Bendich’s papers on the large and variable sizes and structures of plant mitochondrial genomes (Bendich 1993, 1996; Ward et al. 1981), established the fundamental paradoxes that continue to drive plant mitochondrial genomics. How can a genome be so full of non-conserved, seemingly useless, junk DNA (Graur et al. 2015)? How can a genome that is so good at repairing nucleotide changes be simultaneously so susceptible to rearrangements? How is it possible that the remarkable variation in junk DNA content and synteny leave mitochondrial function intact? Why is the plant mitochondrial genome so different from the chloroplast and nucleus?

The importance of the Palmer and Herbon (1988) paper is underscored by the citation record. It has been cited every year since its publication except for 1999, emphasizing the central role of the paradox they revealed in plant mitochondrial research. One reason for this is that plant mitochondrial genomes are very different in structure, organization and mutation rates than animal mitochondria or the nucleus, and any paper on plant mitochondrial DNA needs to cite the works of Bendich, Palmer, and others (Wolfe et al. 1987) to remind the reader of these traits. In addition, mitochondrial genome replication, recombination, and maintenance are still not fully understood.

How is it possible that three decades later we do not completely understand why plant mitochondrial genomes behave this way? Although I am convinced that the expansions and variation in non-coding DNA, the low mutation rate in genes, and the high rearrangement rate in non-genes all emerge from the same fundamental process of DNA repair, that process is not understood fully. My preferred hypothesis is that damage, even to a single base in a single strand, is converted to a break and repaired by double-strand break repair (DSBR) (Christensen 2013, 2014). DSBR mechanisms fall into two categories: non-template-directed and template-directed. Non-template-directed repair can be very accurate, but can also lead to rearrangements. Duplications can be produced this way, and probably explain the expansions. Inaccurate repair in genes would be eliminated by purifying selection, and even synonymous substitutions would be reduced in frequency by template-directed repair of even a small region of DNA (Sloan and Taylor 2010; Wynn and Christensen 2015). Rearrangements in non-coding DNA must be neutral, or nearly so, because they accumulate more quickly, as first observed by Palmer and Herbon. Evidence for this hypothesis has been difficult to come by, in part because we are still unable to transform plant mitochondria, and in part because we still do not know all of the proteins involved in repair (Gualberto and Newton 2017). An additional problem is that the repair mechanism is so effective that even an introduced cytidine deaminase in a Uracil-N-glycosylase mutant led to a small number of mutations, none of them fixed, after 10 generations (Wynn et al. 2020). The MSH1 protein, which has novel features in angiosperms possibly allowing it to make double-strand breaks at sites of damage or mismatch has recently been shown to play an important role in DNA repair in plant organelles, providing support for the hypothesis (Wu et al. 2020).

In the years since 1988, plant mitochondrial genomics has not gotten simpler, only more complex. The Palmer lab has shown surprising levels of horizontal gene transfer of mitochondrial DNA between plant species, further complicating the phylogenetics of plants (Bergthorsson et al. 2003; Mower et al. 2010; Rice et al. 2013; Richardson and Palmer 2007). Bendich and co-workers have revealed that the mitochondrial genomes do not exist as circular molecules, even though they can be mapped to circles (Bendich 1993, 1996, 2007), and the DNA in mature tissues is different from in meristematic tissue (Bendich 2013; Kumar et al. 2015; Oldenburg et al. 2013).

It is also interesting that pre-genomics methods are very accurate. Bendich’s lab established the sizes of plant mitochondrial genomes using reassociation kinetics (C0t curves) and the sizes are remarkably close to those determined decades later by sequencing. Palmer’s lab used restriction mapping and Southern blotting, and their results are also remarkably close to current data from sequencing.

Palmer and Herbon (1988) deserves recognition as an important paper in the history of the Journal of Molecular Evolution. The paper showed that plant mitochondrial genomes undergo fundamentally different DNA maintenance processes than the nucleus or animal mitochondria. This 32-year old paper was, and still is, important and relevant, underpinning and informing the field of plant mitochondrial biology.