The heavy strand dilemma of vertebrate mitochondria on genome sequencing age: number of encoded genes or G + T content?

Differential ultracentrifugation is a classic technique of cell biology and microbiology used to separate cell components, such as nuclei, mitochondria, chloroplasts, lysosomes, peroxisomes, vesicles, ribosomes, and cytoplasm (Garber & Yoder 1983). Regarding the study of the mitochondria, a cesium chloride (CsCl) solution has been used in a technique named Buoyant density ultracentrifugation to separate nuclear DNA from mitochondrial DNA (Welter et al. 1988; Zimmerman et al. 1988). Therefore, the two distinct bands obtained in the CsCl ultracentrifugation were used to define a heavy (H) and a light (L) DNA strand for a given mitochondrion. These studies have been classically conducted for many clades (Corneo et al. 1966; Sinclair et al. 1967; Brack et al. 1972; Beridze & Tabidze 1976; Garber & Yoder 1983) and they are still done in a smaller pace to isolate cell components for further analyses (Zhao et al. 2005). However, with the development of PCR and Sanger sequencing techniques, researchers did not need to separate the mitochondria from other cellular components to perform the sequencing of their DNA. Nowadays, more than 10 years after the development of next-generation sequencing (NGS) techniques, mitochondrial genomes can be obtained as a byproduct of whole genome sequencing. Although the amount of mitochondrial reads can differ because of the type of tissue, animal group and/or sample extraction protocols, we have found that one mitochondrial read can be found for each 200–1000 reads of the nuclear genome (Uliano-Silva et al. 2015; Perini Vda et al. 2016; Prosdocimi et al. 2016). In some cases the amount of reads from mitochondrial DNA is so high that assemblers masked them as repeats while performing whole genome assemblies. It was just after assembling and annotating many complete mitochondrial genomes of various organisms (Gomes de S a et al. 2015; Uliano-Silva et al. 2015; Perini Vda et al. 2016; Prosdocimi et al. 2016; Souto et al. 2016; Uliano-Silva et al. 2016) that we decided to take a closer look into the H and L strand assignments. We were astonish to realize that many vertebrate mitochondrial genomes published to date seem to be inaccurately annotated in terms of light and heavy strands (Arnason et al. 2006, 2007; Zhang et al. 2012; Yoon et al. 2013; Ren et al. 2014; Pan et al. 2015; Zhang et al. 2015; Zou et al. 2015; Meng et al. 2016; Sun et al. 2016). Classic works showed that mitochondrial L-strand could be defined as the one that has the lower content of guanines and thymines (Taanman 1999; Munn 1975). According to Vinograd et al. (1963), the titration of N–H protons of thymines and guanines would increase the buoyant density of denatured DNA, showing that Gþ T content is responsible for the buoyant density of mitochondrial strands. When sequencing and analyzing the mitogenome of the blue-fronted amazon parrot, Amazon aestiva (Lima et al. manuscript in preparation), we were confronted with observation that the majority of the genes in this molecule was encoded by the L-strand, i.e. the strand presenting the lower Gþ T content (in this case 38.06%). The first work describing a complete mitochondrial genome from a species from the Amazona genus suggested that most genes were present in the H-strand (Urantowka et al. 2013). We were also amazed to observe that the most significant avian mitochondrial genome classically described for the chicken (Desjardins & Morais 1990), have provided contradictory information about heavy and light strand assignments. While saying that the heavy strand presented most of genes in the first paragraph of their results, Desjardins and Morais (1990) have also shown that this strand presents the lower amount of Gþ T content (37.3%) and showed most of their genes being encoded on this strand, according to the legend of their Figure 2. On the other hand, in one of the most classical and cited work of mitochondrial genomes dated from 1981, Anderson and collaborators described for the first time the whole human mitochondrial genome. They state that most genes of the human mitochondrion were encoded in the light-strand (Anderson et al. 1981). However, contrarily to Anderson’s work (Anderson et al. 1981), Taanman (1999) states that the human mitochondrion H-strand is the one with higher amount of genes encoded. In order to bring light to this question, we downloaded all available vertebrate mitochondria from RefSeq (November 2016) and calculated their Gþ T content, associating this information with the number of genes described on each strand. We found out that nearly all vertebrate mitogenomes (4200 out of 4205) have their light strand presenting most genes and the lowest Gþ T (Supplementary Table 1). The single exceptions are from 5 fishes from different clades, with 2 from the same genus Johnius. Those exceptions present the majority of genes encoded in the heavey strand (Accession numbers: NC_005800, NC_013879, NC_021130, NC_022464 and NC_008222). We suspect that the statement proposed by Taanman (1999) that most vertebrate mitochondrial genes were on the H-strand was replicated downstream until most