A system where the sequence of genes differs in an unprecedented degree from that of their products is found in a barely known protist group, the diplonemids. Diplonemids are unicellular flagellates thriving predominantly in marine, but also freshwater environments. With only two recognized genera, Rhynchopus and Diplonema, this group is seemingly an insignificant taxon. However, recent environmental explorations uncovered that diplonemids include two additional and enormous clades consisting exclusively of uncultured taxa (DSPDI and DSPDII [5]). The most comprehensive ocean expedition ever conducted (Tara 2009–2013) revealed that diplonemids are among the most abundant and genetically most diverse eukaryotes in the sea [6].

Phylogenetically, diplonemids are very distant to the three best-known eukaryotic groups, animals, fungi and plants (Fig. 2). The sister group of diplonemids are kinetoplastids (including trypanosomes) and the two belong to the super group Excavata, a diverse assemblage of unicellular eukaryotes.

Across eukaryotes, mitochondrial genomes carry genes involved in cellular respiration, ATP synthesis, and protein translation. Mitochondrial DNA (mtDNA) of the diplonemid type species Diplonema papillatum encodes 12 genes of the conventional set (Table 1, atp6 - rns). Transfer RNA genes are absent from this mtDNA; tRNAs are apparently imported from the cytosol as in kinetoplastids and several other eukaryotes.

Most unconventional in diplonemids is their mitochondrial genome architecture and gene structure [7]. While the majority of eukaryotes have a single mitochondrial chromosome [8], that of diplonemids is multi-partite. Specifically, mtDNA of D. papillatum consists of at least 80 distinct circular molecules with a most regular structure (Fig. 3A). Chromosomes are either 6 kbp (class A) or 7 kbp (class B) long. About 9/10 of their length is identical in sequence among members of a given class and mostly contains direct repeats, without coding content. Only a short stretch, termed ‘cassette’, is unique to each chromosome. Cassettes include a single short coding region (up to ∼500 nt) flanked by, on average, 50 nt non-coding sequence on both sides [9].

Cassettes are by far too small to harbor complete, regular mitochondrial genes (except tRNA genes that are absent from D. papillatum mtDNA). Consequently, genes are radically fragmented into up to 11 pieces (modules). The only gene squeezed into a single module is rns, which specifies the mitochondrial small subunit ribosomal RNA (mt-SSU rRNA).

D. papillatum is not the only diplonemid species with such an elaborate genome architecture and gene structure. The same pertains to three other species of the traditional genera Diplonema and Rhynchopus [10]. It will be interesting to examine whether members of the two new clades, DSPDI and DSPDII, adhere to the same gene fragmentation scheme or disclose other unexpected inventions.

After sequencing the pool of mitochondrial chromosomes from D. papillatum, we were able to spot most (or probably all) cassettes. But it was difficult to pinpoint the gene modules within these cassettes and to identify their coding content. The reason is not only that modules are short, but also that gene sequences are highly divergent compared to homologs in other species. Only deep transcriptome sequencing allowed us to locate exact module boundaries within cassettes, and to associate them with standard, full-length genes.

Transcriptome data provided information not just on mature transcripts but also on intermediates. Actually, the set of RNA precursors has allowed to reconstruct all steps involved in the generation of mature RNAs. First, gene modules in Diplonema mitochondria are transcribed separately (Fig. 3B and C). Primary transcripts include adjacent stretches of the chromosomes’ constant regions. Subsequently, all non-coding sequence is removed to end up with bare modules (Fig. 3D). The modules that eventually form the 3′ end of a mature transcript are poly-adenylated (Fig. 3E). Poly(A)-tailed mRNAs are otherwise rare in mitochondria, but common in the nucleus. Interestingly, in Diplonema mitochondria, the stop codon of reading frames is sometimes incomplete in the gene module, but completed post-transcriptionally by poly-adenylation – as also seen in mammalian mitochondria.

Once module transcripts are fully processed at their ends, they are joined into contiguous, full-length RNAs (Fig. 3F). We refer to this process as ‘trans-splicing’, as it involves the linking of coding sequences from separate RNA molecules, in the correct order. As we detail below, the process in diplonemid mitochondria is fundamentally different from intron (trans-)splicing.

Since trans-splicing intermediates in Diplonema mitochondria occur at a detectable steady-state level, we were able to examine whether transcript assembly proceeds in a certain order (e.g. from 3′ to 5′ or vice versa), yet, there is no indication for any directionality. Instead module transcript assembly appears to start simultaneously with any cognate pair and to proceed in parallel until completion [11] (Fig. 3F and G).

Transcriptome analysis also shows an astounding accuracy in the assembly of transcripts, with nearly no module-mix-up from different genes. This raises the question how cognate module partners recognize each other among 82 candidates or more (Table 2, column 2). We tested the most plausible hypothesis, i.e. that neighbor-module recognition and joining is conducted by a conventional intron-splicing machinery. The four known intron types differ in the way how exon/intron boundaries are defined, and by the molecular processes that cleave the precursor RNA at exon/intron boundaries and then join cognate exons (Fig. 1; [reviewed in 4]).

Extensive bioinformatics searches in Diplonema mitochondrial sequences did not detect sequence or secondary structure elements typical for introns nor any other recurrent motifs within or adjacent to modules [11]. As no cis elements appear to exist, there must be some factors acting in trans for guiding accurate joining of module transcripts into full-length mRNAs and rRNAs. The nature of the postulated trans factors (RNA, protein, or DNA?), the underlying biochemical mechanism, and the enzymes involved in Diplonema trans-splicing are yet to be unraveled.

Comprehensive gene/transcript comparisons in D. papillatum uncovered a number of sites in the sequence where transcripts include Us that are absent from the gene sequence. Evidently, these additional Us are inserted post-transcriptionally – reminiscent of U-insertion RNA editing in kinetoplastid mitochondria. Three-quarters of Diplonema mitochondrial transcripts undergo U-insertions, adding more than 220 non-encoded nucleotides (Table 2). In contrast to the sister group, U-deletions seem not to exist in Diplonema mitochondria.

Another difference to kinetoplastids is that in Diplonema, Us are not inserted by transcript cleavage, nucleotide addition and resealing, but rather appended at the 3′ ends of certain module transcripts prior to trans-splicing. For example the module-5 transcript of the gene nad4 (which consists of eight modules, Fig. 4A), is extended by two Us that are retained in the trans-spliced mRNA (Fig. 4B) [7,10].

We recently detected a second, less conspicuous type of post-transcriptional RNA editing in Diplonema mitochondria: nucleotide substitutions [12]. Certain cytidines (Cs) in the gene are replaced by uridines (Us) in the transcript; and further, certain gene-encoded adenosines (As) become inosines (Is) in RNA (inosine is an A that lacks an amino group; Fig. 4C). Both types of substitutions take place after transcription, since transcriptome sequences show pre-edited, partially edited and fully edited modules. We demonstrated that inosines in transcripts originate by in situ nucleotide deamination, i.e. without RNA cleavage. Substitution RNA editing is substantial in D. papillatum mitochondria. Nearly half of the transcripts are edited by C-to-U or A-to-I substitutions, at a total of 114 sites (Table 2).

C-to-U RNA editing in mitochondria has been observed before, particularly in plants [1]. In contrast, A-to-I RNA editing is a hallmark of nuclear transcript processing in animals [13]. Diplonema is the first reported case of A-to-I editing of organellar mRNAs. Moreover, Diplonema's mitochondrial small-subunit ribosomal RNA is the first example of an rRNA containing Is.

Does mitochondrial RNA editing occur only in the type species D. papillatum or in the entire diplonemid group? We found U-appendage RNA editing in all four species examined (D. papillatum, D. ambulator, D. sp.2, and Rhynchopus euleeides), and the sites are precisely conserved [14]. Substitution RNA editing occurs also in all these species; however, the pattern varies considerably between taxa. When comparing the genes and transcripts of nad4 and rns, not a single editing position (out of 74) is in common [12].

RNA editing in diplonemid mitochondria – and in organelles in general – targets exclusively coding regions and is function-critical for the product [1]. This is most obvious in protein-coding genes, where the majority of editing sites coincides with first and second codon positions, thus changing amino acids or eliminating stop codons in precursor RNAs. As a result, the ‘edited’ protein resembles more closely homologs from species that do not feature RNA editing (Fig. 4B and C). In animal nuclear genes, RNA editing plays a diametrically different role. There, editing is typically partial (with various transcript versions coexisting) and affects mostly intronic or terminal, untranslated regions. In fact, RNA editing of animal transcripts plays regulatory roles [13], rather than “fixing” coding sequences.

To decipher aberrant genes, e.g. via trans-splicing and RNA editing as in diplonemids, the cell requires enormously sophisticated machineries. Such a complex system comes with high costs and would seem to be a quite inefficient way of expressing genes. This makes one wonder why non-functional genes have not been corrected at the DNA level, especially over such long evolutionary times. Does ‘doctoring’ at the RNA level provide hidden selective advantages? Not necessarily, because selective advantage is not the only force at work in evolution. Survival just requires that there is no major disadvantage for the cell. Neutral or only marginally harmful features (applied to diplonemids: trans-splicing and RNA editing) can become permanent without selective pressure, e.g. simply by genetic drift [15]. The recent theory of ‘constructive neutral evolution’ provides a compelling framework for the emergence not only of trans-splicing and RNA editing in diplonemids, but also of other ‘fortuitously complex’ processes [16–18].

A decade ago, diplonemids had been a barely known eukaryotic group. Today, they have moved to the spotlight for two reasons. First, they are among the most abundant eukaryotes in global oceans, playing an unsuspected, crucial role in the food web. It will be of great importance to understand the nature of their interactions (mutualistic, parasitic, scavenging, or predatory?) with other ecological entities, and their principal interaction partners (phototrophic eukaryotes – the primary carbon-fixers of the globe – marine bacteria, or viruses?).

The second reason why diplonemids have attracted attention is the novelty of molecular mechanisms involved in mitochondrial gene expression. Dissection of the machinery that performs the unique trans-splicing process will be of great fundamental interest. Research into this system – and any other basic research alike – has the potential of unimagined future applications. Knowing how the diplonemid trans-splicing machinery works might expand our molecular toolbox and find uses in medicine or agriculture. For example, one could imagine RNA level treatment of hereditary diseases in human or rapid RNA-level adaptation of crop plants to fast-changing climatic conditions. Whether these examples are science fiction or vision, only future will tell.

Taken together, studies on diplonemids exemplify that research into neglected eukaryotes continues to open surprising insights in biology, ecology, and biochemistry. It shows that protists are a vastly untapped source for advancing our knowledge about the functioning and evolution of eukaryotic life.

The author declares that she has no competing interest.

This work was supported by an operating grant from the Canadian Institute for Health Research (CIHR, grant MOP-79309) to G.B.