• Loading metrics

Mitochondrial dual-coding genes in Trypanosome brucei

  • Laura E. Kirby,

    Roles Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Microbiology and Molecular Genetics, Michigan State University, East Lansing, Michigan, United States of America

  • Donna Koslowsky

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Writing – original draft, Writing – review & editing

    Affiliation Department of Microbiology and Molecular Genetics, Michigan State University, East Lansing, Michigan, United States of America


Mitochondrial dual-coding genes in Trypanosome brucei

  • Laura E. Kirby, 
  • Donna Koslowsky

This is an uncorrected proof.


Trypanosoma brucei is transmitted between mammalian hosts by the tsetse fly. In the mammal, they are exclusively extracellular, continuously replicating within the bloodstream. During this stage, the mitochondrion lacks a functional electron transport chain (ETC). Successful transition to the fly, requires activation of the ETC and ATP synthesis via oxidative phosphorylation. This life cycle leads to a major problem: in the bloodstream, the mitochondrial genes are not under selection and are subject to genetic drift that endangers their integrity. Exacerbating this, T. brucei undergoes repeated population bottlenecks as they evade the host immune system that would create additional forces of genetic drift. These parasites possess several unique genetic features, including RNA editing of mitochondrial transcripts. RNA editing creates open reading frames by the guided insertion and deletion of U-residues within the mRNA. A major question in the field has been why this metabolically expensive system of RNA editing would evolve and persist. Here, we show that many of the edited mRNAs can alter the choice of start codon and the open reading frame by alternative editing of the 5’ end. Analyses of mutational bias indicate that six of the mitochondrial genes may be dual-coding and that RNA editing allows access to both reading frames. We hypothesize that dual-coding genes can protect genetic information by essentially hiding a non-selected gene within one that remains under selection. Thus, the complex RNA editing system found in the mitochondria of trypanosomes provides a unique molecular strategy to combat genetic drift in non-selective conditions.

Author summary

In African trypanosomes, many of the mitochondrial mRNAs require extensive RNA editing before they can be translated. During this process, each edited transcript can undergo hundreds of cleavage/ligation events as U-residues are inserted or deleted to generate a translatable open reading frame. A major paradox has been why this incredibly metabolically expensive process would evolve and persist. In this work, we show that many of the mitochondrial genes in trypanosomes are dual-coding, utilizing different reading frames to potentially produce two very different proteins. Access to both reading frames is made possible by alternative editing of the 5’ end of the transcript. We hypothesize that dual-coding genes may work to protect the mitochondrial genes from mutations during growth in the mammalian host, when many of the mitochondrial genes are not being used. Thus, the complex RNA editing system may be maintained because it provides a unique molecular strategy to combat genetic drift.


Trypanosomes are one of the most successful parasites in existence, inhabiting an incredibly wide range of hosts [1, 2]. The dixenous members cycle between two distinct hosts and can encounter different environments with distinct metabolic constraints. These parasites are unique in that they all possess glycosomes (where glycolysis occurs) as well as mitochondria [3]. The salivarian trypanosomes (e.g. T. brucei, T. vivax) are especially interesting, because they are exclusively extracellular in their mammalian hosts, continuously replicating within the bloodstream over periods of months. During this stage of the life cycle, the mitochondrion is down-regulated, lacking both Krebs cycle enzymes and a functional electron transport chain (ETC) [4]. Successful transition to the fly vector, requires activation of the ETC and ATP synthesis via oxidative phosphorylation. This unique lifecycle leads to a major problem: when the mitochondrial genes are unused, they are not under selection, hence the integrity of these genes are threatened by genetic drift [5, 6]. Exacerbating this, salivarian trypanosomes undergo a severe bottleneck as they transition through the tsetse fly and into the mammalian host, and then within the bloodstream, they undergo multiple bottlenecks at each antigenic switch, as they evade the host immune system [7]. Such bottlenecks create additional forces of genetic drift, where genes can be lost even if their deleterious fitness effect is considerable. These parasites possess several unique genetic features, including RNA editing of the mitochondrial transcripts. RNA editing creates open reading frames in “cryptogenes” by insertion and deletion of uridylate residues at specific sites within the mRNA. The U-insertions/deletions are directed by small guide RNAs (gRNA) and can repair frameshifts, generate start and stop codons and more than double the size of the transcript (for review see [8]). While the mRNA cryptogenes are encoded on maxicircles (25–50 copies per DNA network), the guide RNAs are encoded on thousands of 1 kb minicircles, encoding 3–5 gRNA genes each [9]. This effectively means that the genetic information for the edited mitochondrial mRNAs is dispersed between the mRNA cryptogenes on the maxicircles and the thousands of gRNA coding minicircles. The extensive editing of a single transcript can require more than 40 gRNAs and hundreds of editing events [10]. While the initial gRNA can interact with the 3’ end of the pre-edited transcript, all subsequent gRNAs anchor to edited sequence created by the preceding gRNA. Hence, editing proceeds from the 3’ end to the 5’ end of the mRNA transcript with the terminal gRNA (last one in the cascade) often creating the start codon needed for translation. This sequential dependence means that with even high accuracy rates for each gRNA, the overall fidelity of the process is astonishingly low. A major question in the field has been why this fragile and metabolically expensive system of RNA editing would evolve and persist.

Another level of complexity in the kinetoplastids RNA editing process was the detection of an alternative editing event that leads to the production of a functionally discrete protein isoform. Alternative editing of Cytochrome Oxidase III (COIII) is reported to generate a novel DNA-binding protein, AEP-1, that functions in mitochondrial DNA maintenance [11, 12]. In this transcript, one alternative gRNA generates sequence changes at two sites that links an open reading frame (ORF) found in the pre-edited 5’ end, to the 3’ transmembrane domains found in the COIII edited ORF. This was the first indication, that one cryptogene could contain information for more than one protein. Here, we show that as many as six additional cryptogenes also encode for more than one protein. Analyses of the terminal gRNA populations indicate that gRNA sequence variants exist that can alter the choice of the start codon and the open reading frame by alternative editing of the 5’ end of the mRNA. Mutational bias analyses indicate that six of the mitochondrial genes may be dual-coding, with RNA editing allowing access to both reading frames. Dual-coding genes are defined as a stretch of DNA containing overlapping open reading frames (ORFs) [13, 14]. Of particular interest are dual-coding genes that contain two ORFs read in the same direction: a canonical protein (normally annotated as protein coding in the literature) and an alternative ORF. Maintaining dual-coding genes is costly, as it constrains the flexibility of the amino acid composition of both proteins. Hence, it is thought that dual-coding genes can survive long evolutionary spans only if the overlap is advantageous to the organism [15]. We hypothesize that trypanosomes use dual-coding genes to protect genetic information by essentially hiding a non-selected (ETC) gene within one that remains under selection. Thus, the ability to access overlapping reading frames may be added to a growing list of gene protective strategies made possible by the complex RNA editing process [5, 6, 16].

Materials and methods

Trypanosome growth

T. brucei procyclic clones from IsTAR (EATRO 164), TREU 667 and TREU 927 cell lines were grown in SDM79 at 27°C and harvested at a cell density of 1-3x107. The TREU 667 cell line was originally isolated from a bovine host in 1966 in Uganda [17]. The TREU 927 cell line was originally isolated from Glossina pallidipes in 1970 in Kenya [18]. The EATRO 164 strain was isolated in 1960 from Alcephalus lictensteini and maintained in the lab of Dr. K. Vickerman until being obtained by Dr. Ken Stuart in 1966 [19]. Dr. Stuart derived the procyclic form from the bloodstream form culture in 1979.

Guide RNA isolation, preparation, and sequencing

Mitochondrial mRNAs and gRNAs were isolated as previously described [10]. All RNAs were treated with Promega DNAse RQI. In order to isolate gRNAs from TREU 667 and TREU 927 cells, RNAs were size fractionated on a polyacrylamide gel as previously described [10]. Guide RNAs were then extracted and prepped for sequencing using the Illumina Small RNA protocol [10]. Libraries from TREU 667 and TREU 927 were deep sequenced on the Illumina GAIIx; reads were processed and trimmed as previously described [10].

Messenger RNA preparation and sequencing

In order to isolate target mRNAs, isolated TREU 667 mitochondrial RNAs were reverse transcribed using the Applied Biosystems High Capacity cDNA Reverse Transcription Kit. CR3 cDNAs were amplified via PCR using the following primers (underlined portions are gene specific and non-underlined portions are tag regions used in deep sequencing reaction):



After amplification, the products were purified using the QIAquick PCR Purification Kit, and paired end Illumina deep sequencing was performed on the Illumina Miseq (2x 250 bp paired end run). Low quality results were removed using FaQCs, adapters were removed using Trimmomatic and PEAR was used to merge paired end reads. Finally, Fastx was used to compile identical reads while maintaining the number of redundant reads. CR3 edited transcripts were identified by comparing sequence downstream of the 5' never edited region to the edited CR3 sequence. Guide RNAs were identified by using the mRNA sequences as queries against our existing gRNA databases, as previously described [10].

Mutational frequency and editing conservation analysis

Mitochondrial pan-edited genes were categorized as potentially dual-coding based on identification of extended alternative reading frames and/or presence of identified gRNAs that generate alternative 5’ end sequences. These genes include CR3, CR4, ND3, the 5’ editing domain of ND7, ND9 and RPS12. Nondual-coding pan-edited genes include ATPase 6, COIII, ND8 and the 3’ editing domain of ND7. Partially edited genes include CYb, Murf II and COII. Never edited genes include COI, ND1, ND2, ND4 and ND5. For all analyses, ND7 was considered as two separate coding regions: the 5' editing domain (ND7N) and the 3' editing domain (ND7C) [20]. As we hypothesize that only the 5' editing domain of ND7 is dual-coding, mutation calculations for ND7N was pooled with the dual-coding genes and ND7C was pooled with nondual-coding pan-edited genes. T. brucei and T. vivax mRNA sequences of mitochondrial encoded genes were aligned based on protein sequence using Clustal Omega [21]. Nucleotide sequence mutations were identified and their effects on the amino acid sequence were classified as silent, missense or nonsense mutations. Missense mutations were further divided into three groups based on the PAM 250 matrix where conversions with a value <0 were considered not conserved, conversions with a value 0≤x≤0.5 were considered modestly conserved, and conversions with a value >0.5 were considered strongly conserved [22]. Mutation frequencies were normalized for each gene using nucleotide sequence length. Frequencies were compared using unpaired t-tests.

The extent of editing conservation between T. vivax and T. brucei was calculated by aligning the pan-edited genes based on ACG sequence. For each alignment, each location between an A, C or G nucleotide where a U-residue was inserted or deleted in either sequence was considered an editing site. Editing sites were classified as identical in both sequences, altered in insertion or deletion length, having switched from an insertion site to a deletion site, or only occurring in one of the sequences. Percent editing conservation was based on total number of editing sites within each mRNA. Percentages were compared using unpaired t-tests.

A principal component analysis (PCA) was performed on all three reading frames of the pan-edited genes using the scikit-learn principal component analysis tool [23]. For this analysis, the predicted protein sequences for all three reading frames were aligned using Clustal Omega [21]. Missense, nonsense, and indel mutations were quantified. Missense mutations were further divided into three groups as described above. Each mutation type was quantified and the relative frequency of each mutation calculated based on protein amino acid length. The variables used in the PCA include the protein mutation frequencies and the percentage of identical editing sites in each mRNA. The first reading frame of each gene is defined as the ORF published in the literature.

Data and software availability

CR3 sequence accession number: SAMN06318039. TREU 927 gRNA sequence accession number: SAMN06318154. TREU 667 gRNA sequence accession number: SAMN06318153. NCBI’s Sequence Read Archive.


In T. brucei, analyses of the gRNA transcriptome for the pan-edited transcripts indicate that full editing involves a large number of gRNA populations [10, 24]. In addition, most of the gRNA populations (population defined as guiding the same or near same region of the mRNA) contain multiple sequence classes. The sequence classes most often differ in R to R or Y to Y mutations, hence guide the generation of the same mRNA sequence (A:U and G:U base pairs allowed). During these analyses, we noted that the terminal gRNA population for Cytosine-rich Region 3 (CR3) (putative NADH dehydrogenase subunit 4L [25]), had 3’ sequences that would extend editing beyond the previously identified translation start codon. In addition, this population had several sequence variants that would generate different edited sequences in this region. The most abundant terminal gRNA would introduce a stop codon in-frame with two alternative AUG start codons found near the 5’ end (Fig 1A). Other sequence classes however, would either bring the upstream AUGs into frame, or shift the reading frame. Intriguingly, the alternative +1 reading frame (ARF) did not contain any premature termination codons. In order to determine if these gRNAs were utilized, we used Illumina deep sequencing to identify the most abundant forms of fully edited CR3 transcripts. Surprisingly, we identified multiple forms of the mRNA (Fig 1A, 1B and 1C and S1 Table). The first was the fully edited sequence predicted by the most abundant gRNA identified (Fig 1A). The other transcripts however, had unique editing patterns at the 5’ end (Fig 1B and 1C and S1 Table). Use of these 5’ CR3 sequences allowed us to identify novel gRNAs. Predicted translation of these mRNA sequences indicate that they use the +1 reading frame, and that the protein generated would be the same length as the ORF previously identified. This suggests that CR3 is dual-coding, and that selection of the terminal gRNA determines which reading frame will be used. A re-examination of the terminal gRNAs for the pan-edited genes indicated that at least two other transcripts, NADH dehydrogenase subunit 7 (ND7) and ribosomal protein subunit 12 (RPS12), have identified gRNA sequence variants within the terminal gRNA population that allow access to alternative reading frames (Fig 1D and 1E and S2 and S3 Tables). Interestingly, the alternative gRNA for ND7 generates a +2 frameshift with a 65 amino acid open reading frame. The ND7 transcript is differentially edited in two distinct domains separated by 59 nts that are not edited in the mature transcript (the HR3 region) [20]. Only the 5’ domain is edited in both life cycle stages; full editing of the 3’ domain was only found in bloodstream form (BF) parasites. The stop codon for the +2 frameshift is found within the HR3 region, therefore this alternative protein would be generated by full editing of only the 5’ domain. While the most abundant gRNA in the Eatro BF transcriptome (~50,000 reads) would generate a sequence utilizing the identified ND7 ORF (Fig 1D ORF), the most abundant gRNA (>100,000) in the Eatro 164 procyclic library is the +2 ARF gRNA (Fig 1D ARF and S2 Table). In RPS12, the alternative gRNA deletes an additional U-residue downstream of the existing start codon, shifting the reading frame into the +1 ARF (Fig 1E). Interestingly, in Leishmania tarentolae, a gRNA, gRPS12VIIIa, has been identified that would also shift the frame of the existing start codon into the +1 ARF [26].

Fig 1. Alternative editing of the 5' end of pan-edited genes results in access to different reading frames.

CR3 (A, B, C): Sequenced mRNA variants are aligned with gRNAs and predicted protein sequences. Inserted U-residues are lowercase while deleted U-residues are shown as asterisks. Canonical Watson-Crick base pairs (|); G:U base pairs (:). Previously identified start codons are doubled underlined. Potential upstream AUG start codons are indicated by wave underlines. Alternatively edited nucleotides are shown in red. Common anchor regions are shown in blue. ND7 (D), and RPS12 (E): Predicted mRNA and protein sequences, based on identified gRNAs.

The identification of gRNAs that could alter the reading frame led us to re-analyze the ORFs of the edited transcripts. In addition to CR3, we found extended ORFs in two different frames for Cytosine-rich Region 4 (CR4) and NADH Dehydrogenase (ND) subunit 9, while several others had shorter, but still significant ORFs in alternative frames (Fig 2). We do note that the original sequence publications for both CR4 and RPS12 (CR6) had indicated that the fully edited sequence contained extended ORFs in two different frames [27, 28]. Additionally, NADH Dehydrogenase subunit 3 (ND3) was also considered to be potentially dual coding, based on mutational analysis described below.

Fig 2. Positions of stop codons on all RFs of the edited genes in T. brucei.

For each gene, reading frame 1 (RF1) is designated as the protein ORF previously identified. Hypothetical dual-coding reading frames are shown in red. A6 = ATPase 6; CO = Cytochrome Oxidase; CYb = Cytochrome b; Murf = Maxicircle unidentified reading frame; ND3ND9 = NADH Dehydrogenase subunits.

As we did find potential ARFs in the edited transcripts, we analyzed the predicted ORFs for biases in their mutational pattern. Dual-coding genes often display an atypical codon mutation bias due to constraints imposed by the need to maintain protein function in both genes. In single-coding genes, changes in the third nucleotide of a codon give rise to synonymous amino acids, so this position (N3) is much less constrained. In contrast, in dual-coding genes, the N3 position in one frame is the N1 or N2 position in the alternative frame. Therefore, they have low rates of synonymous mutations [29]. This codon bias has been used to develop algorithms to detect novel overlapping genes [30, 31]. These algorithms however, cannot be used in the analysis of our edited transcripts as the two-component genetic system (mRNAs created by gRNA editing) introduces another layer of mutational constraint [32]. In addition, the edited sequence of the transcripts is known for only a limited number of kinetoplastids, and only the salivarian trypanosomes have the same general life cycle; other kinetoplastids, like Leishmania and T. cruzi, have evolved different infective cycles and are under very different selective pressures [5, 33, 34]. Fully edited sequences are known for T. vivax, the earliest branching salivarian trypanosome [35, 36]. T. vivax differs from T. brucei in that they complete the insect phase of their life cycle entirely within the proboscis of the fly. This parasite has been described as an intermediate stage in the evolutionary pathway from mechanical transmission (ancestral) to full adaptation to the midgut and salivary glands of the tsetse fly [37]. Using the T. vivax sequence, we analyzed mutation patterns in all of the mitochondrially-encoded mRNAs (Fig 3). mRNA sequences were aligned by codons based on their protein alignments (Clustal Omega [21]). Mutated codons were identified and classified as silent, missense and nonsense mutations. Missense mutations were further divided into three groups based on the PAM 250 matrix [22]. These data clearly show that the RNA editing process significantly constrains the types of mutations tolerated within the mitochondrial genome. In comparison to the genes that are not edited (ND1, ND2, ND4, ND5, COI) or have limited editing (CYb, Murf II and COII), a distinct suppression of silent mutations and strongly conserved missense mutations were observed for all of the pan-edited genes, consistent with previous observations (Fig 3) [32]. A suppression of mutations that lead to moderately conserved amino acid replacements was also observed, but these were not as striking due to the low frequency of this type of mutation. No significant difference was observed in the frequency of not conserved missense mutations, though a trend towards a lower frequency of these mutations in the putative dual-coding genes (CR3, CR4, ND3, ND9, 5’ND7 and RPS12) was noted. This was complemented by a significant increase in the frequency of strongly conserved missense mutations in the putative dual-coding genes in comparison to the other pan-edited genes (3’ND7, ND8, A6 and COIII).

Fig 3. Mutational frequencies in mitochondrially encoded genes categorized by effect on amino acid sequence.

T. brucei and T. vivax pan-edited dual-coding (PanEd DC), pan-edited nondual-coding (PanEd NDC), partially edited (PartEd), and never edited (NevEd) mRNA sequences were aligned based on their amino acid alignment (reading frame 1, defined as the reading frame encoding the gene product previously annotated in the literature) [21]. Mutations were categorized as silent (Si), strongly conserved (SC), modestly conserved (MC), not conserved (NC) or nonsense (Non). The amount of conservation was determined using the PAM 250 matrix, where conversions with a value 0<x≤0.5 were considered modestly conserved, and conversions with a value >0.5 were considered strongly conserved, and conversions with a value ≤0 were considered not conserved. Error bars depict standard error. * p<0.05, ** p<0.01 (unpaired t-test).

Surprisingly, while the overall mutational frequency of the fully edited pan-edited genes was similar, a comparison of the conservation of editing patterns did show a significant difference between the putative dual-coding and the other pan-edited genes (Fig 4). The dual-coding genes consistently had a lower conservation of their editing pattern. Upon further examination, we found that most changes in the editing pattern resulted from thymidine insertions and deletions within the maxicircle DNA sequence, which was then corrected by the editing machinery. These types of mutations do not result in a change to the final mRNA sequence once edited. The T. brucei (Tb) dual-coding genes appeared to consistently insert more U-residues, while T. vivax (Tv) had more U-residues encoded within the DNA sequence. Indeed, comparisons of the length of the coding regions of Tb and Tv cryptogenes (unedited sequence) show that the putative dual-coding genes are almost 10% shorter in Tb. In contrast, the non-dual coding cryptogenes are not significantly shorter (~2.5%). Some of the other changes in editing patterns did generate small internal frameshifts as previously described by Landweber and Gilbert [32]. However, the high prevalence of internal frameshifts reported for COIII by Landweber and Gilbert is reflected in our analysis only for COIII and A6.

Fig 4. Percent conservation of editing patterns between T. brucei and T. vivax.

Alignment of the fully edited mRNAs was based on ACG sequence (see S1 Fig). Each editing site was defined as a site on at least one of the two aligned mRNAs where an editing event occurred. Sites were then classified as identical, only identified in one of the two sequences, type switched (one site is an insertion and the other is a deletion), or altered in length. Error bars depict standard error. * p<0.05, ** p< 0.01 (unpaired t-test).

Since differences in the types of amino acid mutations were observed, we performed a principal component analysis on the frequency of mutation types for all three reading frames of the pan-edited genes (Fig 5). In addition, we included the percentage of editing site conservation as a variable. This analysis clearly clustered the putative +1 dual-coding transcripts (reading frame 2). The first component (z-axis,) is strongly based on editing conservation, and separates the dual-coding genes from the other pan-edited genes as expected. While component 2 (x-axis) separated ORF1 and ORF3 from ORF2 of each gene, component 3 clearly separated the dual-coding ORF2s from nondual-coding ORF2s. The ND7N ORF3 was the only exception, and the gRNA data suggests that it is a dual-coding gene using the +2 (ORF3) reading frame. This suggests that an additional layer of mutational constraint beyond that imposed by the RNA editing process can be detected for six of the extensively edited transcripts.

Fig 5. Principal component analysis of frequency of amino acid mutation types and editing conservation between T. brucei and T. vivax pan-edited transcripts.

A. First factorial plan (z-axis: first component, x-axis: second component, y-axis: third component). ND7N = ND7 5’ editing domain, ND7C = ND7 3’ editing domain. ORF2 and ORF3 are defined as the +1 and +2 reading frames, respectively. B. Histogram of eigenvalues for first six components. Eigenvalues represent the amount of the variance accounted for by each component. C. Absolute contribution of each analyzed mutation frequency to components 1, 2, and 3. Amount of conservation was determined using the PAM 250 matrix as described in Fig 3. Mutation type: SC = strongly conserved, MC = moderately conserved, NC = Not conserved, Non = Nonsense, InDel = insertion or deletion. Editing conservation (EdCon) was determined using alignments of edited mRNAs (S1 Fig). Aligned editing sites were characterized as identical or altered.

Because dual-coding genes are often conserved in multiple species, we analyzed the available sequences of other kinetoplastids (Leishmania tarentolae (Lt), Leishmania mexicana amazonensis (Lma), Phytomonas serpens (Ps), Perkinsela CCAP1560/4 (Pk)) to determine if they also contain multiple overlapping reading frames with homology to those found in T. brucei. Interestingly, many of the alternative reading frames did show some homology to the ARFs found in Tb. However, most of these ARFs are punctuated with stop codons (S2 Fig). Extended alternative reading frames are found in CR3, 5’ND7 and RPS12 in Ps. However, the extended ARF in the Ps CR3 is in the +2 reading frame and the ND7 and RPS12 ARFs shows very little homology with the Tb/Tv ARF (S2A, S2D and S2F Fig) [38]. Interestingly, while Perkinsela has lost many of the genes in the mitochondria, RPS12 was retained [39]. The Pk RPS12 ARF possesses a near full open reading frame with one stop codon three codons after an in frame start near the 5' end of the edited transcript. This pattern is reminiscent of the conventionally edited sequences of CR3 and ND7, and could suggest that an alternative edit may remove the stop codon, allowing access to the ARF. The L. tarentolae CR4 orthologue also has two extended ORFs. Interestingly, the published sequence for Lt CR4 appears to switch between the two ORFs (switch appears to occur in a stretch of 13 inserted Us) [40]. This may explain why only the carboxyl half of the published Lt CR4 showed good homology with Tb and Lma [41]. Translation of the Lt ARF does generate a protein with the N-terminus showing high homology to the conventional Tb and Lma CR4, while translation of the published ORF shows some homology to the Tb CR4 ARF (S2B Fig). These data are intriguing enough that these sequences should be re-examined. While most of the other pan-edited transcripts had multiple stop codons in the +1 and +2 reading frames, many did show good homology to the Tb ARF sequences. Particularly intriguing are the ND3, ND8 and ND9 alignments. While internal stop codons are found in Tv, Lt and Lma ND9 ARFs, they show strong homology to the Tb ND9 ARF throughout the protein (S2E Fig). In ND3, the amino ends of the ARFs show strong homology between all four of the Trypanosoma and Leishmania species (S2C Fig). This homology decreases after an internal stop codon found in the same position in 3 of the 4 species. As ND8 is the only other pan-edited gene in Lt, Lam and Ps, we also examined the conservation of the ND8 ORF and ARF, even though our mutational analyses did not tag the ND8 gene as dual coding. While the ND8 ARFs were punctuated by multiple stop codons, they surprisingly also showed areas of strong homology between all 5 species, especially down stream of an internal methionine (S2G Fig). We do note that we cannot rule out the possibility that alternative editing can remove stop codons observed in the ARFs.

Analyses of the ARF predicted proteins suggest that they are all short transmembrane proteins with two or more predicted transmembrane alpha helices (Fig 6) [42, 43]. While functional homologues are often difficult to detect in trypanosomes, searches using the predicted protein sequence of each ARF did identify small molecule transport proteins with limited confidence. Using Phyre2, the ND7 ARF was identified as a homolog of the bacterial sugar transporter SemiSWEET (61.5% confidence) [44, 45]. SemiSWEET, which forms homodimeric structures, is also a distant homolog of the yeast mitochondrial pyruvate carrier 1 (MPC1). This protein has two transmembrane alpha helices and forms a heterodimer with either of the other two pyruvate carrier proteins [46, 47]. While still very speculative, it is intriguing that the small ARF proteins might oligomerize to form small mitochondrial membrane transporters.

Fig 6. Amino acid sequences of ARFs of dual-coding genes.

Predicted transmembrane regions are shaded in gray [42, 43]. *No start codon was identified and the amino acid sequence shown begins at the 5' end or after the first stop codon at the 5' end. Exclamation point indicates premature termination codon.


The work presented here, suggests that as many as six of the extensively edited mRNAs in T. brucei are dual-coding and that it is alternative editing using different terminal gRNAs that allows access to the two different reading frames. Deep sequencing of the 5’ end of CR3 indicates fully edited transcripts that have access to both reading frames are present in the mitochondrial transcriptome and gRNA analyses indicate that three different cell lines contain gRNAs that can alternatively edit the 5’ ends of CR3, RPS12 and ND7. In addition, analyses of the mutational bias in pan-edited genes suggest that an additional layer of mutational constraint is observed in the putative dual-coding genes. While the overall mutational frequency observed for the fully edited mRNAs is similar for all pan-edited genes, the types of amino acid changes that appear to be tolerated are significantly different. This is consistent with these genes having to maintain functional proteins in two different reading frames. Analyses of other trypanosomes, do show that some of the ARFs have intriguing homology to the ARFs identified in T. brucei and T. vivax. However, most of the ARFs are punctuated with stop codons. These data are difficult to interpret because we cannot rule out the possibility that the stop codons are removed by alternative editing events. In addition, the other trypanosome species have evolved very different infective life cycles and are under different selective pressures. For example, P. serpens is a pathogen that infects important crops and is transmitted by sap-feeding bugs. These parasites have glucose readily available in both life cycle stages and are unique in that they lack a fully functional respiratory electron transport chain [48, 49]. For Leishmania, all life cycle stages possess an active Krebs cycle and ETC linked to the generation of ATP [50, 51, 52]. These unique adaptations to different hosts suggest that they may not be under the same evolutionary pressure to maintain dual-coding genes.

Overlapping reading frames are common in viruses, and are thought to persist due to strong genome size constraints [53, 54]. More recently however, over-lapping genes have been identified in mammalian and bacterial genomes [5558]. In these organisms, size is not an issue and the potential advantage of overlapping genes is less clear. For dual-coding genes, the need to maintain both ORFs constrains the ability of each protein to become optimally adapted [15]. As this constraint can be alleviated by gene duplication, it is thought that dual-coding regions can survive long evolutionary spans only if the overlap provides a selective advantage. In mammals, many of the identified dual-coding genes like Gnas1 and XBP1, produce two proteins that bind and regulate each other [5960]. For these proteins, dual-coding may be advantageous for the tight co-expression needed. An alternative model, suggests that under high mutation rates, the overlapping of critical nucleotide residues is advantageous because it may reduce the target size for lethal mutations [61]. This may be particularly important for organisms that have evolved to exist in dual-metabolic environments (two hosts). We hypothesize that the trypanosome mitochondrial ARFs encode small metabolite transporters that provide a distinct growth advantage to bloodstream form parasites. The complete overlap of these small transporter genes with electron transport chain (ETC) genes would protect the integrity of the ETC genes that are required only in the insect host. Thus, in trypanosomes, dual-coding genes may be a mechanism to combat genetic drift during extended periods of growth in non-selective environments. In T. brucei, it is known that a number of bloodstream form essential proteins are functionally linked to Krebs cycle or ETC genes. While not a “classic” dual-coding gene in that production of the alternative protein does not involve overlapping reading frames, the pan-edited COIII gene does contain the information for two distinct proteins, COIII and AEP-1. AEP-1 is important for kinetoplastid DNA maintenance and over-expression of the DNA-binding domain results in a dominant negative phenotype including decreased cell growth and aberrant mitochondrial DNA structure [12]. The nuclear encoded α-ketoglutarate dehydrogenase E2 (α-KDE2) is known to be a dual-function protein, in that it plays important roles in both the Krebs cycle and in mitochondrial DNA inheritance [62]. RNAi knockdowns of this gene in bloodstream form (BF) trypanosomes also show a pronounced reduction in cell growth. Similarly, the Krebs cycle enzyme α-ketoglutarate decarboxylase (α-KDE1) is also a dual-function protein with overlapping targeting signals that allow it to be localized to both the mitochondrion and glycosomes [63]. RNAi knockdowns of α-KDE1 in BF trypanosomes is lethal, suggesting that in addition to its enzymatic role in the Krebs cycle, it plays an essential role in glycosomal function in T. brucei [63]. It has been previously suggested that both alternative editing and dual-function proteins are important mechanisms for expanding the functional diversity of proteins found in trypanosomes [11, 6264]. We hypothesize, that in salivarian trypanosomes, an equally important role for these dual-coding/function genes may be the protection of genetic information.

The “why” of the unique RNA editing process in kinetoplastids has been a long-standing paradox. The complex machinery and the sheer number of gRNAs required to direct the thousands of U-insertion/deletions indicate that this process is metabolically very costly. Initially, it was proposed that U-insertion/deletion editing (kRNA editing) was one of many RNA editing processes that were in fact relics of the RNA world. However, the very different mechanism of the RNA editing systems in existence, and their very limited distribution within specific groups of organisms indicate that they are more likely derived traits that evolved later in evolution [65, 66]. The sheer complexity of the kRNA editing process, with no obvious selective advantage, led to the proposal that insertion/deletion editing arose via a constructive neutral evolution (CNE) pathway [67]. Indeed, RNA editing in trypanosomes is always mentioned in support of CNE as an example of how seemingly non-advantageous, complex processes can arise [68, 69]. More recently however, it has been hypothesized that RNA editing co-evolved with G-quadruplex structures found in the pre-edited mRNAs [16]. These structures are thought to be advantageous in that they can help regulate transcription in order to promote DNA replication and prevent kinetoplast DNA loss. However, they must be removed by the RNA editing system prior to translation [16]. Another prominent hypothesis is that RNA editing is advantageous because it is a mechanism by which an organism can fragment and scatter essential genetic information throughout a genome [6, 70]. Kinetoplast DNA is far less stable than chromosomal DNA, and loss of minicircles due to asymmetric division of the kDNA network have been frequently observed, especially in laboratory cultures of Leishmania [40, 71]. Buhrman et al. [70] suggest that the scattering of essential guide RNA genes throughout the DNA network, would prevent fast growing deletion mutants from outcompeting more metabolically versatile parasites during growth in the mammalian host. Using a mathematical model of gene fragmentation in changing environments (absence of functional selection), they showed a distinct advantage for gene fragmentation. In their model, the number of tolerable generations under periods of relaxed selective pressure was increased by more than 40% before loss of the ability to move to the next life cycle stage. If the dual-coding ARFs give BF trypanosomes a selective growth advantage similar to that observed by the COIII alternative protein AEP1, then the number of “essential” gRNA genes would increase greatly. Currently, only AEP1, A6 and RPS12 mitochondrial genes have been experimentally shown to be essential [12, 7273]. In addition, the presence of alternative editing and dual-coding genes would complement the protection provided by gene fragmentation by also shielding the genes from deleterious point mutations within critical ETC genes. This suggests that the complex RNA editing system found in the mitochondria may therefore provide multiple molecular strategies to increase genetic robustness. Protection of the mitochondrial genome during growth in the mammal would increase the capacity for successful transfer to an insect vector and maximize the parasites long-term survival and spread.

Supporting information

S1 Fig. Alignments of T. brucei and T. vivax pan-edited mRNAs.

ATPase 6 (A), COIII (B), CR3 (C), CR4 (D), ND3 (E), ND7 (F), ND8 (G), ND9 (H), and RPS12 (I). Uppercase letters indicate nucleotides originally encoded in the DNA, lower case u's indicate uridines inserted during editing and asterisks indicate uridines removed during editing.


S2 Fig. Alignments of protein sequences of pan-edited dual-coding genes in L. tarentolae, L. amazonensis, P. serpens, and Perkinsela CCAP1560/4 with T. brucei and T. vivax sequences.

A: CR3, B: CR4, C: ND3, D: ND7 5' Editing Domain E: ND9, F: RPS12, G: ND8 (Nondual-coding). Absent sequences were unavailable. All ORF alignments show published protein sequences. All ARF alignments show +1 or +2 (ND7 only) reading frame translations of the full length mRNA sequences. In the ARF alignments of CR3, ND7 and RPS12, translations were made using the alternative T. brucei mRNA sequences shown in Fig 1. Two alignments of CR3 ARFs are presented to display the P. serpens +2 reading frame which has no stop codons. L. tarentolae CR4 published protein sequence shows limited homology in the C terminus to all other CR4 protein sequences. The edited mRNA has two editing sites where 13 U residues are inserted. If the second of these insertion sites is shortened to 12 U residues, the translation of this mRNA has much better homology to other CR4 proteins. Alignments with translations of the two different sequences (13U and 12U) are both shown, with the location of the altered site highlighted in red. While ND8 does not appear to be dual-coding, this alignment was included as well, for comparison of the conservation of a nondual-coding gene with that of the dual-coding genes. It should be noted that ND8 is the only nondual-coding gene that is pan-edited in L. tarentolae, L. amazonensis, and P. serpens; ND7, A6 and COIII are only partially edited in these species. The internal M observed in the ND8 sequence is highlighted in yellow (see text).! = Termination codon.


S1 Table. Identified CR3 mRNA and gRNA transcripts.

A—C: Major CR3 mRNA and gRNA sequence classes. The CR3 mRNA transcriptome was generated using the TREU667 cell line. Identified sequences were then used to search gRNA transcriptomes from four different cell lines: EATRO 164 Bloodstream (BS), EATRO 164 procyclic (PC), TREU 927 procyclic and TREU 667 procyclic. ORF = previously identified Open Reading Frame (purple protein sequence). ARF = Newly identified Alternative Reading Frame (green protein sequence). Alternatively edited nucleotides are shown in Red. Inserted U-residues are lowercase while deleted U-residues are shown as asterisks. Canonical Watson-Crick base pairs (|); G:U base pairs (:). Previously identified start codons are doubled underlined. Potential upstream AUG start codons are indicated by wave underlines. gRNAs were sorted based on guiding sequence class. Sequence variations observed in the 3’-U-tail were ignored in assigning class. Transcript copy number (Reads), were determined by adding all gRNAs of the same sequence class. Only major sequence classes are shown (defined as containing greater than 100 transcript copies). In the case of rare transcripts, the identified gRNA are shown regardless of copy number. gRNA transcript numbers varied greatly between the different cell lines. Interestingly, the most abundant mRNA (CR3 Form C, 7147 reads) had the fewest identified gRNA reads.


S2 Table. ND7 5’-most gRNA populations and the predicted mRNA sequences generated.

A—J: ND7 gRNA major classes and predicted editing patterns. ND7 terminal (5’ most) gRNA populations and the predicted mRNA sequence generated. Predicted sequences presented are based on the most abundant gRNAs that generate each reading frame found in the four gRNA transcriptome databases. Initial characterization of the ND7 transcript was done using the EATRO 164 cell line and is unusual in that it is edited in two distinct domains [20]. While the 5’ domain was edited in both life cycle stages, complete editing of the 3’ domain was only detected in bloodstream stage parasites. Interestingly, the most abundant EATRO 164 PC (procyclic or insect form) gRNA would generate a sequence that brings the 5’ most AUG into a +2 frame. The ARF is 65 AA long and involves the entire 5’ editing domain. In contrast, the most abundant gRNAs in the EATRO 164 Bloodstream stage library (EATRO 164 BS), would generate sequences that use the originally described ND7 ORF). While gRNA transcript numbers again varied greatly between the different cell lines, all three cells lines had gRNA sequence variants that allowed access to both reading frames.


S3 Table. RPS12 5’-most gRNA populations and the predicted mRNA sequences generated.

A—E: RPS12 gRNA major classes and predicted editing patterns. RPS12 terminal (5’ most) gRNA populations and the predicted mRNA sequence generated. RPS12 differs from both CR3 and ND7 in that the alternative edit that shifts the reading frame occurs just downstream of the previously identified start codon (double-underlined). We do note that the identified alternative gRNAs are rare in all of the gRNA libraries except TREU 667.



We thank the Ken Stuart Lab for trypanosome cell lines and Chris Adami for helpful discussions. We would also like to acknowledge the Dr. Marvis A. Richardson Endowed Fellowship Fund for their recognition of LEK.


  1. 1. Vickerman K. Evolutionary expansion of the trypanosomatid flagellates. Int J Parasitol. 1994;24: 1317–1331. pmid:7729984
  2. 2. Simpson AGB, Stevens JR, Lukes J. The evolution and diversity of kinetoplastid flagellates. Trends Parasitol. 2006;22(4): 168–174. pmid:16504583
  3. 3. Hannaert V, Bringaud F, Opperdoes FR, Michels PAM. Evolution of energy metabolism and its compartmentation in Kinetoplastida. Kinetoplastid Biology and Disease. 2003;2: 11. pmid:14613499
  4. 4. Bringaud F, Riviere L, Coustou V. Energy metabolism of trypanosomatids: Adaptation to available carbon sources. Mol Biochem Parasitol. 2006;149: 1–9. pmid:16682088
  5. 5. Speijer D. Is kinetoplastid pan-editing the result of an evolutionary balancing act? IUBMB Life. 2006;58: 91–96. pmid:16611574
  6. 6. Speijer D. Evolutionary Aspects of RNA Editing. In “RNA Editing”. (Goringer H.U. ed). Nucleic Acids and Molecular Biology. 2008;20: 199–226.
  7. 7. Oberle M, Balmer O, Brun R, Roditi I. Bottlenecks and the Maintenance of Minor Genotypes during the Life Cycle of Trypanosoma brucei. PLoS Pathog. 2010;6: e1001023. pmid:20686656
  8. 8. Read LK, Lukes J, Hashimi H. Trypanosome RNA editing: the complexity of getting U in and taking U out. WIREs RNA. 2016;7: 33–51. pmid:26522170
  9. 9. Hong M, Simpson L. Genomic Organization of Trypanosoma brucei Kinetoplast DNA Minicircles. Protist. 2003;154: 265–279. pmid:13677453
  10. 10. Koslowsky DJ, Sun Y, Hindenach J, Theisen T, Lucas J. The insect-phase gRNA transcriptome in Trypanosoma brucei. Nucleic Acids Res. 2013;42: 1873–1886. pmid:24174546
  11. 11. Ochsenreiter T, Hajduk S. Alternative editing of cytochrome oxidase III mRNA in trypansome mitochondria generates protein diversity. EMBO Reports. 2006;7: 1128–1133. pmid:17008930
  12. 12. Ochsenreiter T, Anderson S, Wood ZA, Hajduk SL. Alternative RNA editing produces a novel protein involved in mitochondrial DNA maintenance in trypanosomes. Molec Cell Biol. 2008;28: 5595–5604. pmid:18606780
  13. 13. Normark S, Bergstrom S, Edlund T, Grundstrom T, Jaurin B, Lindberg FP, et al. Overlapping Genes. Ann Rev Genet. 1983;17: 499–525. pmid:6198955
  14. 14. Liang H. Decoding the dual-coding region: key factors influencing the translational potential of a two-ORF-containing transcript. Cell Research. 2010;20: 508–509. pmid:20436511
  15. 15. Chung W-Y, Wadhawan S, Szklarczyk R, Pond SK, Nekrutenko A. A First Look at ARFome: Dual-Coding Genes in Mammalian Genomes. PLos Computational Biology. 2007;3(5): e91. pmid:17511511
  16. 16. Leeder W-M, Hummel NFC, Göringer HU. Multiple G-quartet structures in pre-edited mRNAs suggest evolutionary driving force for RNA editing in trypanosomes. Sci Rep. 2016;6. pmid:27436151
  17. 17. Hudson KM, Taylor AER, Elce BJ. Antigenic changes in Trypanosoma brucei on transmission by tsetse fly. Parasite Immunol. 1980;2: 57–69.
  18. 18. Gibson W. The origins of the trypanosome genome strains Trypanosoma brucei brucei TREU 927, T. b. gambiense DAL 972, T. vivax Y486 and T. congolense IL3000. Parasit Vectors. 2012;5: 71. pmid:22483376
  19. 19. Agabian N, Thomashow L, Milhausen M, Stuart K. Structural Analysis of Variant and Invariant Genes in Trypanosomes. Am J Trop Med Hyg. 1980;29: 1043–1049. pmid:7435803
  20. 20. Koslowsky DJ, Bhat GJ, Perrollaz AL, Feagin JE, Stuart K. The MURF3 gene of T. brucei contains multiple domains of extensive editing and is homologous to a subunit of NADH dehydrogenase. Cell. 1990;62: 901–911. pmid:2393904
  21. 21. Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, et al. Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega. Mol Syst Biol. 2014;7: 539.
  22. 22. Gonnet GH, Cohen MA, Benner SA. Exhaustive matching of the entire protein sequence database. Science. 1992;256: 1443–1445. pmid:1604319
  23. 23. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;12: 2825–2830.
  24. 24. Kirby LE, Sun Y, Judah D, Nowak S, Koslowsky DJ. Analysis of the Trypanosoma brucei EATRO 164 Bloodstream guide RNA Transcriptome. PLoS Negl Trop Dis. 2016;10: e0004793. pmid:27399202
  25. 25. Duarte M, Tomas AM. The mitochondrial complex I of trypanosomatids–an overview of current knowledge. J Bioenerg Biomembr. 2014;46: 299–311. pmid:24961227
  26. 26. Simpson L, Douglass SM, Lake JA, Pellegrini M, Li F. Comparison of the Mitochondrial Genomes and Steady State Transcriptomes of Two Strains of the Trypanosomatid Parasite, Leishmania tarentolae. PLOS Negl Trop Dis. 2015;9: e0003841. pmid:26204118
  27. 27. Corell RA, Myler P, Stuart K. Trypanosoma brucei mitochondrial CR4 gene encodes an extensively edited mRNA with completely edited sequence only in bloodstream forms. Mol Biochem Parasitol. 1994;64: 65–74. pmid:8078524
  28. 28. Read LK, Myler PJ, Stuart K. Extensive editing of both processed and preprocessed maxicircle CR6 transcripts in Trypanosoma brucei. J Biol Chem. 1992;267: 1123–1128. pmid:1730639
  29. 29. Hanada K, Shin-Han S, Wen-Hsiung L. The Nonsynonymous/Synonymous Substitution Rate Ratio versus the Radical/Conservative Replacement Rate Ratio in the Evolution of Mammalian Genes. Mol Biol Evol. 2007;24: 2235–2241. pmid:17652332
  30. 30. Firth AE, Brown CM. Detecting overlapping coding sequences with pairwise alignments. Bioinformatics. 2005;21: 282–292. pmid:15347574
  31. 31. Firth AE, Brown CM. Detecting overlapping coding sequences in virus genomes. BMC Bioinformatics. 2006;7: 75. pmid:16483358
  32. 32. Landweber LF, Gilbert W. RNA editing as a source of genetic variation. Nature. 1993;363: 179–182. pmid:8387160
  33. 33. Tielens AGM, van Hellemond JJ. Surprising variety in energy metabolism within Trypanosomatidae. Trends in Parasitol. 2009;25(10): 482–490.
  34. 34. Verner Z, Cermakova P, Skodova I, Kovacova B, Lukes J, Horvath A. Comparative analysis of respiratory chain and oxidative phosphorylation in Leishmania tarentolae, Crithidia fasciculata, Phytomonas serpens and procyclic stage of Trypanosoma brucei. Mol Biochem Parasitol. 2014;193L 55–65.
  35. 35. Greif G, Rodriguez M, Reyna-Bello A, Robello C, Alvarez-Valin F. Kinetoplast adaptations in American strains from Trypanosoma vivax. Mutat Res. 2015;773: 69–82. pmid:25847423
  36. 36. Jackson AP, Berry A, Aslett M, Allison HC, Burton P, Vavrova-Anderson J, et al. Antigenic diversity is generated by distinct evolutionary mechanisms in African Trypanosome species. Proc Natl Acad Sci U S A. 2012;109: 3416–3421. pmid:22331916
  37. 37. Morrison LJ, Vezza L, Rowan T, Hope JC. Animal African Trypanosomiasis: Time to Increase Focus on Clinically relevant parasite and Host Species. Trends in Parasitol. 2016;32: 599.
  38. 38. Maslov DA, Hollar L, Haghighat P, Nawathean P. Demonstration of mRNA editing and localization of guide RNA genes in kinetoplast-mitochondria of the plant trypanosomatid Phytomonas serpens. Mol Biochem Parasitol. 1998;93: 225–236. pmid:9662707
  39. 39. David V, Flegontov P, Gerasimov E, Tanifuji G, Hashimi H, Logacheva MD, et al. Gene Loss and Error-Prone RNA Editing in the Mitochondrion of Perkinsela, an Endosymbiotic Kinetoplastid. mBio. 2015;6: e01498–15. pmid:26628723
  40. 40. Thiemann OH, Maslov DA, Simpson L. Disruption of RNA editing in Leishmania tarentolae by the loss of minicircle-encoded guide RNA genes. EMBO J. 1994;13: 5689–5700. pmid:7988566
  41. 41. Maslov DA. Complete set of mitochondrial pan-edited mRNA in Leishmania mexicana amazonensis LV78. Mol Biochem Parasitol. 2010;173: 107–114. pmid:20546801
  42. 42. Käll L, Krogh A, Sonnhammer ELL. A Combined Transmembrane Topology and Signal Peptide Prediction. Method. J Mol Biol. 2004;338:1027–1036. pmid:15111065
  43. 43. Käll L, Krogh A, Sonnhammer ELL. Advantages of combined transmembrane topology and signal peptide prediction—the Phobius web server. Nucleic Acids Res. 2007;35: W429–W432. pmid:17483518
  44. 44. Kelley LA, Mezulis S, Yates CM, Wass MN, Sternberg MJE. The Phyre2 web portal for protein modeling, prediction and analysis. Nat Protoc. 2015;10: 845–858. pmid:25950237
  45. 45. Xu Y, Tao Y, Cheung LS, Fan C, Chen L-Q, Xu S, et al. Structures of SemiSWEET transporters in two distinct conformations. Nature. 2014;515: 448–452. pmid:25186729
  46. 46. Bender T, Pena G, Martinou J-C. Regulation of mitochondrial pyruvate uptake by alternative pyruvate carrier complexes. EMBO J. 2015;34: 911–924. pmid:25672363
  47. 47. Štáfková J, Mach J, Biran M, Verner Z, Bringaud F, Tachezy J. Mitochondrial pyruvate carrier in Trypanosoma brucei. Mol Microbiol. 2016;100: 442–456. pmid:26748989
  48. 48. Porcel BM, Denoeud F, Opperdoes F, Noel B, Madoui M-A, et al. The Streamlined Genome of Phytomonas spp. Relative to Human Pathogenic Kinetoplastids Reveals a Parasite Tailored for Plants. PLoS Genet 2014;10(2): e1004007. pmid:24516393
  49. 49. Nawathean P, Maslov DA.The absence of genes for cytochrome c oxidase and reductase subunits in maxicircle kinetoplast DNA of the repiration-deficient plant trypanosomatid Phytomonas Serpens. Curr Genet. 2000;38(2): 95–103. pmid:10975258
  50. 50. McConville MJ, Naderer T. Metabolic Pathways Required for the Intracellular Survival of Leishmania. Annu Rev Microbiol. 2011;6: 543–561.
  51. 51. Saunders EC, DE Souza DP, Naderer T, Sernee MF, Ralton JE, Doyle MA, Macrae JI Chambers JL, Heng J, Nahid A, Likic VA, McConville MJ. Central carbon metabolism of Leishmania parasites. Parasitology 2010;137: 1303–13. pmid:20158936
  52. 52. Saunders EC, Ng WW, Kloehn J, Chambers JM, Ng M, McConville MJ. Induction of a Stringent Metabolic Response in Intracellular Stages of Leishmania mexicana Leads to Increased Dependence on Mitochondrial Metabolism. PLoS Pathog 2014;10(1): e1003888. pmid:24465208
  53. 53. Belshaw R, Pybus OG, Rambaut A. The evolution of genome compression and genomic novelty in RNA viruses. Genome Research. 2007;17: 1496–1504. pmid:17785537
  54. 54. Brandes N, Linial M. Gene overlapping and size constraints in the viral world. Biology Direct. 2016;11: 26. pmid:27209091
  55. 55. Mouilleron H, Delcourt V, Roucou X. Death of a dogma: eukaryotic mRNAs can code for more than one protein. Nucleic Acids Res. 2016;44: 14–23. pmid:26578573
  56. 56. Liang H, Landweber LF. A genome-wide study of dual coding regions in human alternatively spliced genes. Genome Research. 2006;16: 190–196. pmid:16365380
  57. 57. Ribrioux S, Brungger A, Baumgarten B, Seuwen K, John MR. Bioinformatics prediction of overlapping frameshifted translation products in mammalian transcripts. BMC Genomics. 2008;9: 122. pmid:18325113
  58. 58. Pallejà A, Harrington ED, Bork P. Large gene overlaps in prokaryotic genomes: result of functional constraints or mispredictions? BMC Genomics. 2008;9: 335. pmid:18627618
  59. 59. Klemke M, Kehlenback RH, Huttner WB. Two overlapping reading frames in a single exon encode interacting proteins-A novel way of gene usage. EMBO J. 2001;20: 3849–3860. pmid:11447126
  60. 60. Yoshida H, Oku M, Suzuki M, Mori K. pXBP1(U) encoded in XBP! Pre-mRNA negatively regulates unfolded protein response activator pXBP1(S) in mammalian ER stress response. J Cell Biol. 2006;172: 565–575. pmid:16461360
  61. 61. Peleg O, Kirzhner V, Trifonov E, Bolshoy A. Overlapping Messages and Survivability. J Mol Evol. 2004;59: 520–527. pmid:15638463
  62. 62. Sykes SE and Hajduk SL. Dual functions of α-ketoglutarate dehydrogenase E2 in the Krebs cycle and mitochondrial DNA inheritance in Trypanosoma brucei. Eukaryot Cell. 2013;12: 78–90. pmid:23125353
  63. 63. Sykes SE, Szempruch A, Hajduk SL. The Krebs cycle enzyme-ketoglutarate decarboxylase is an essential glycosomal protein in bloodstream African trypanosomes. Eukaryot Cell. 2015;14: 206–215. pmid:25416237
  64. 64. Ochsenreiter T, Cipriano M, Hajduk SL. Alternative mRNA Editing in Trypanosomes Is Extensive and May Contribute to Mitochondrial Protein Diversity. PLoS ONE. 2008;3(2): e1566. pmid:18270563
  65. 65. Covello PS, Gray MW. On the evolution of RNA editing, Trends Genet. 1993;9: 265–268. pmid:8379005
  66. 66. Gray MW. Evolutionary Origin of RNA Editing. Biochemistry. 2012;51: 5235–5242 pmid:22708551
  67. 67. Gray MW, Lukes J, Archibald JM, Keeling PJ, Doolittle WF. Cell biology. Irremediable complexity? Science. 2010;330: 920–921. pmid:21071654
  68. 68. Stoltzfus A. On the possibility of constructive neutral evolution. J Mol Evol. 1999;49: 169–181. pmid:10441669
  69. 69. Stoltzfus A. Constructive neutral evolution: exploring evolutionary theory’s curious disconnect. Biology. Direct 2012;7: 35. pmid:23062217
  70. 70. Buhrman H, van der Gulik PTS, Severini S, Speijer D. A mathematical model of kinetoplastid mitochondrial gene scrambling advantage. 2013; arXiv:1307.1163v1 [q-bio.PE].
  71. 71. Savill NJ, Higgs PG. A theoretical study of random segregation of minicircles in trypanosomatids. Proc R Soc Lond B BiolSci. 1999;266: 611–20. pmid:10212451
  72. 72. Szempruch A, Choudhury R, Wang Z, Hajduk SL. In vivo analysis of trypanosome mitochondrial RNA function by artificial site-specific RNA endonuclease-mediated knockdown. RNA. 2015;21:1–9.
  73. 73. Aphasizheva I, Maslov DA, Aphasizhev R. Kinetoplast DNA-encoded ribosomal protein S12: a possible functional link between mitochondrial RNA editing and translation in Trypanosoma brucei. RNA Biol. 2013;10 (11): 1679–88. pmid:24270388