Background

Mitochondria of bilaterian animals are commonly a single, closed, circular compact molecule, ranging from about 14 to 16 kb in size, highly conserved in terms of gene content and organization [1]. However, isolated lineages exhibit rapidly evolving and unorthodox mitochondrial architectures; in very rare cases, even mitogenomes fragmented into multiple chromosomes. Mitogenome fragmentation has occurred independently in several bilaterian lineages. For example, most rotifer species for which data are available possess mitogenomes fragmented into two chromosomes [2, 3]. Fragmentation was also reported in parasitoid wasps in the genus Conostigmus (Insecta: Hymenoptera) [4], and nematodes in the genus Globodera (Nematoda: Tylenchida) [5, 6]. Most famously, mitogenome fragmentation is widespread in lice (Insecta: Psocodea), many of which possess mitogenomes fragmented into a varying number of minichromosomes (sometimes as many as 20) [7,8,9,10]. The relationship between fragmentation and gene order rearrangement rate is unclear; there is some evidence that lice with fragmented mitogenomes also exhibit highly elevated gene order rearrangement rates, but no evidence was found that a high rearrangement rate necessarily causes mitogenomic fragmentation [11]. In addition, as gene order rearrangements and size expansions should be selectively constrained [12, 13], mitogenomic size, rearrangement rate, and sequence evolution should all be positively correlated [14].

Crustaceans exhibit strong evolutionary heterogeneity in the rate of architectural rearrangements, with some lineages exhibiting mitogenomic architecture almost perfectly conserved over hundreds of millions of years, and others exhibiting rapidly evolving mitogenomes where no two species exhibit the same gene arrangement, such as Isopoda [15,16,17]. While the vast majority of crustaceans exhibit the standard organization of the mitogenome into a single circular molecule [17], some isopod species exhibit highly destabilized mitogenomic organization, including phenomena such as linearization and dimerization [18, 19], and heteroplasmy (multiple mitotypes existing within a single individual) and fragmented mitogenomic phenotypes have been proposed in Callinectes sapidus (Decapoda) [20]. However, fragmented mitogenomes have not been experimentally confirmed in crustaceans yet.

Potamidae (Brachyura) are a large freshwater crab family comprising more than 500 described species classified into 78 genera [21, 22]. China is the global hotspot of freshwater crab diversity, with multiple new species described in the last two decades [22,23,24,25]. As a result of the destruction of ecological habitats and climate changes brought on by the Anthropocene, freshwater crab populations in China are facing severe challenges [22]. As a part of the major research project aimed at assessing the diversity of freshwater crabs and designing conservation measures necessary for their long-term survival [26], our research team has also sampled one specimen of Bottapotamon fukienense Dai & Lin, 1979, one of the eight recognized species in this genus of freshwater crabs predominantly distributed in China [24]. To contribute molecular data necessary for the understanding of the phylogeny and evolution of freshwater crabs, we sequenced and analysed its mitochondrial genome.

Methods

In this study, we sequenced, assembled, annotated and comparatively analysed the complete mitochondrial genome of B. fukienense.

Specimen collection

Ten freshwater crab specimens were collected from Xibiao Village, Dongping Town, Nanping City, Fujian Province, China in November 2008 (GPS: 118.5716/27.4548). Specimens were preserved in 95% ethanol and stored at 4 °C at the Department of Parasitology of the Medical College of Nanchang University (NCU MCP), Jiangxi, China. The authors compared collected specimens with holotype photos from the Institute of Zoology, Chinese Academy of Sciences, in Beijing, China. On the basis of this analysis, four specimens (2 adult males and 2 adult females) were identified as B. fukienense (Fig. 1, Additional file 1: Figure S1). To confirm that this mitogenome is fragmented across the entire species, we conducted another sampling trip in November 2021. Three specimens (2 adult males and 1 adult female) were collected from Qintang village, Aojiang Town, Lianjiang County, Fujian Province, China (GPS: 119.5085/26.2247). The identity of specimens was again morphologically confirmed against the original holotype specimen of B. fukienense. All experiments were conducted on unprotected invertebrates, so no permits were required for the study.

Fig. 1
figure 1

Photographs of a sampled Bottapotamon fukienense specimen. (A) Dorsal view of the entire specimen. (B) Dorsal view of the cephalothorax. (C) Frontal view of the cephalothorax. Size bars are shown in each panel

Mitogenome sequencing and assembly

Mitogenome sequencing and assembly were conducted roughly following the procedure described before [15]. The sample was rinsed in distilled water, and DNA was isolated from muscle tissue using AidLab DNA extraction kit (AidLab Biotechnologies, Beijing, China). Primers (Table 1) were designed according to the conserved regions of mitochondrial genes from available orthologues. To avoid sequencing artefacts and ensure that mitogenomes are complete and circular, primers were designed to overlap by approximately 100 bp and used to amplify and sequence the entire mitogenome. PCR reaction mixture (50 µl): 5 U µl − 1 TaKaRa LA Taq polymerase (TaKaRa, Japan), 10 × LATaq Buffer II, 2.5 µM dNTP mixture, 0.2–1.0 µM each primer and 60 ng DNA template. The PCR conditions were as follows: denaturation 98 °C/2 min, 40 cycles of 98 °C/10 s, 50 °C/15 s and 68 °C/1 min kb− 1. When the product was not specific enough, PCR conditions were optimized by increasing the annealing temperature and decreasing the number of cycles. PCR products were sequenced using the Sanger method and the same set of primers. After quality proofing via the visual inspection of electropherograms and identity confirmation using BLAST [27], the sequences were assembled using DNASTAR v. 7.1 [28], ensuring that overlapping segments of amplicons are identical.

As results indicated a fragmented mitogenome, to confirm this unusual architecture, we collected a new specimen in 2021, and used the DNA to conduct long-range PCR using a new set of primers (other parameters were the same as above), designed to produce fragments 7 to 8 Kbp in size (Table 1).

Table 1 Primers used for amplification and sequencing of the mitochondrial genome of B. Fukienense. “Fragment” indicates the fragment number, and “Genes” the mitochondrial genes that amplified sequence spanned

Mitogenome annotation

Mitogenome was annotated using inputs of multiple programs: MITOS [29], DNASTAR, and ARWEN (for tRNAs) [30]. Annotations were then manually fine-tuned according to the orthologous sequences using BLAST and BLASTx. PhyloSuite [31] was used to parse and extract the annotation and to generate the file for submission to NCBI’s GenBank.

Comparative and phylogenetic analyses

All 148 available (January 2024) Potamidae mitogenomes were downloaded from GenBank and imported into PhyloSuite. After the removal of all duplicated and unannotated mitogenomes, as well as leaving one mitogenome per species, 42 mitogenomes were left in the dataset. PhyloSuite was used to semi-automatically re-annotate ambiguously annotated tRNA genes with the help of the ARWEN output, uniformise annotation, extract mitogenomic features, translate genes into amino acid sequences, calculate base composition and skews, generate comparative tables, concatenate the alignments, and prepare input files for plug-in programs used for phylogenetic analyses. Two datasets were tested for phylogenetic analyses: concatenated nucleotide and amino acid sequences of 13 mitochondrial protein-coding genes (PCGs). In the former dataset, 17 sequences failed the compositional homogeneity test in IQ-TREE [32], but in the latter dataset, no sequences failed it. As compositional heterogeneity can strongly interfere with phylogenetic reconstruction [33, 34], amino acids were selected to reconstruct phylogenies. Genes were aligned in batch mode using the accurate g-INS-i strategy in MAFFT [35]. The best data partitioning scheme and best-fit evolutionary models for each partition were selected using the Bayesian Information Criterion implemented in ModelFinder [36]: mtVer + F + R4 for the atp6-cox2-cytb-nad3 partition, LG + F + I + G4 for atp8, Q.plant + F + R3 for cox1-cox3, mtVer + F + I + I + R3for nad1-nad4L, mtVer + F + R4 for nad2-nad6, and mtVer + F + R4 for nad4-nad5. Phylogenies were inferred using two algorithms: the Maximum Likelihood (ML) methodology implemented in IQ-TREE, and the Bayesian inference (BI) implemented in MrBayes [37]. We assessed the support for the ML topology using 20,000 UltraFast Bootstrap [38] replicates. The Bayesian inference analysis was conducted as two independent runs (four chains each), and the average standard deviation of split frequencies (SDSF) was used to determine the needed number of generations. SDSF values < 0.01, which indicates convergence, were reached after ≈ 55,000 generations. The SDSF values more or less plateaued after that, so the analysis was allowed to continue till 150,000 generations (SDSF = 0.0085). The first 25% of sampled trees were discarded as burn-in. Phylogenetic studies mostly indicate Gecarcinucidae as the sister family to Potamidae [17, 39, 40], so the only two available mitogenomes (both Esanthelpusa) for this family were used as the outgroup (Potamonautidae, the third Old World freshwater crab clade was not available). TreeSuite function in PhyloSuite 1.2.3 [41] was used to extract branch lengths, and infer the relative composition variability values (RCV), spurious species, and long-branch scores [42]. UGENE was used to infer pairwise differences among genes and generate figures of alignments [43]. For tRNA folding and visualization, we used RNAfold and forna tools of the ViennaRNA Web Services [44]. Tandem repeats were inferred using Tandem Repeats Finder [45]. OGDRAW was used to draw circular mitogenomic maps [46]. RELAX HyPhy tool was used to infer selection pressures; this tool introduces the K parameter, where K > 1 values indicate intensified selection and K < 1 values indicate relaxed selection [47].

Results

Mitogenomic architecture

By relying on the criterion of identical overlapping segments of amplicons, we found that the sequenced mitogenomic fragments of B. fukienense could not be assembled into a single circular mitogenome. Instead, they assembled into two circular chromosomes (Fig. 2; Table 2). More specifically, although the first circular chromosome was assembled completely, it was smaller than expected, was missing genes, and some of the PCR segments could not be incorporated into it. However, the remaining sequenced fragments could be assembled into another circular chromosome. To confirm that the fragmentation is not accidental, and limited to only one specimen or lineage, we sampled a new set of specimens from a different locality (> 100 km distant from the first locality). The DNA extracted from the new specimen was used to conduct long-range PCR using primers designed anew according to the mitogenomic sequence of B. fukienense obtained in previous steps (Table 1). The results produced using this approach confirmed that the mitogenome of this species is fragmented into two circular chromosomes (Fig. 3). Chromosome 1 comprised 15,111 base pairs (bp) and there were 26 genes and one pseudogene (pseudo-nad1) encoded on it (Additional file 1: Figure S2). Chromosome 2 comprised 8,173 bp and there were 12 genes and two pseudogenes (pseudo-trnL2 and pseudo-rrnL) encoded on it (Additional file 1: Figure S3).

Fig. 2
figure 2

The mitogenomic architecture of B. fukienense and the hypothetical scenario of mitogenome fragmentation into chromosomes 1 and 2. The “ancestral_Potamidae” sequence was inferred by [25], and the “ancestral Bottapotamon_fukienense” architecture was inferred by us. Putative tandem duplication events in the ancestral B. fukienense architecture are indicated by red squares, and chromosome 1 is indicated by a black square. Sequence deletions and gene pseudogenisation events are indicated by star symbols. The figure was made using PhyloSuite.

Fig. 3
figure 3

Electrophoresis confirmation of fragmentation into two chromosomes in the mitochondrial genome of B. fukienense. From left to right: 1 kb DNA ladder, with sizes in bases shown; chr1-F1: Chromosome 1, fragment 1 (cytb to atp6, 7697 bases); chr1-F2: Chromosome 1, fragment 2 (atp6 to cytb, 7708 bases); chr2-F1: Chromosome 2, fragment 1 (cox1 to cox1, 7278 bases); cox1 (1255 bases, primers XXCOX1F and R). For primers, see Table 1

Several genes were duplicated when the entire mitogenome was considered (Fig. 2; Table 2). We found two copies of nad1, one on each chromosome: the copy on chromosome 2 (the smaller one) was the functional one, as the copy on chromosome 1 was truncated by approximately 230 bases. Remarkably, the first 703 bases were almost perfectly conserved between the two copies, which indicates a recent duplication. The complete gene was highly conserved compared to other Potamidae orthologues.

A copy of rrnL (or 16S) was present on both chromosomes: a complete gene on chromosome 1 (1326 bp), and a truncated pseudogene on chromosome 2 (752 bp). The copy on chromosome 1 was highly conserved in comparison to most other Potamidae orthologues, whereas the copy on chromosome 2 comprised only the second half of the gene (5’-end-truncated). However, the matching segments of the two copies were also almost perfectly conserved (1 SNP). We checked whether the intergenic space adjacent to pseudo-rrnL corresponds to rrnL, but the two sequences could not be aligned, confirming that this is a deletion and not sequence degradation.

There were also two copies of rrnS (or 12 S): rrnS-1 on chromosome 1 (822 bp) and rrnS-2 on chromosome 2 (823 bp). There were about a dozen SNPs between the two copies (1% divergence). In comparison to the congeneric Bottapotamon lingchuanense, both sequences exhibited 10% divergence. In comparison to other orthologues, the rrnS-1 exhibited a marginally more conserved sequence.

There were two copies of trnL1: one on each chromosome. They exhibited 100% identity, and they were highly conserved in comparison to most other orthologues.

Initial analyses indicated the presence of putative two trnL2 copies on chromosome 2 (Table 2). One copy (annotated as trnL2) was found in the ancestral position, between cox1 and trnS2, whereas the other copy (annotated as pseudo-trnL2) was found between rrnL-2 and rrnS-2. The alignment with orthologues revealed that trnL2 (found in the ancestral position) exhibited high levels of homology, whereas pseudo-trnL2 exhibited almost no homology (Additional file 1: Figure S4). This indicated a possibility of an annotation artefact. Considering the ancestral architecture of this gene (Fig. 2), we assumed that the gene might be a degraded trnV. Indeed, the alignment of pseudo-trnL2 with trnV genes of Potamidae revealed a high level of homology, but also a large deletion, probably rendering it non-functional (Additional file 1: Figure S5). However, the ARWEN algorithm annotated it as trnL2 and successfully folded both sequences into cloverleaf structures (Additional file 1: Figure S6). We further applied the RNAfold and forna tools of the ViennaRNA Web Services, but they failed to produce a standard cloverleaf structure for either of the two genes (Additional file 1: Figure S7). In light of the above analyses, we cannot be certain whether this was a case of tRNA remodelling and the gene is functional, or whether the deletion in the ancestral trnV gene caused a loss of functionality and produced accidental similarity to trnL2.

Table 2 Detailed architecture of the two mitogenomic chromosomes of Bottapotamon fukienense. IGR is the intergenic region, where negative values indicate overlaps

In order to attempt to infer the putative evolutionary history of architectural rearrangements that led to the observed mitogenomic architecture of B. fukienense, we compared it to the ancestral gene order for Potamidae [25] (Fig. 2), exhibited by most species in our comparative dataset (e.g. Huananpotamon lichuanense) (Fig. 4). We also included the gene order of the closest-related available species, B. lingchuanense, as it differed from the ancestral gene order. Notably, this is most likely to be the ancestral order of the clade comprising both the Bottapotamon and Neilupotamon genera, as it was found in most of the Neilupotamon species (Fig. 4). The comparison of the two chromosomes with these two gene orders showed that B. fukienense exhibits a gene order highly similar to B. lingchuanense (Fig. 2). A large section of chromosome 1 perfectly corresponded to the large section of B. lingchuanense mitogenome, spanning pseudo-nad1-L1-rrnL-V-Q-cox2-K-D-atp8-atp6-cox3-G-nad3-A-R-N-S1-E-H-F-nad5-nad4-na4L-T-P-nad6-cytb-rrnS genes. A fragment comprising rrnS-I-M-nad2-W-C-Y-cox1-L2-S2-nad1-L1-rrnL-pseudo-L2 formed chromosome 2. We inferred the ancestral gene order of B. fukienense as being almost identical to the gene order of B. lingchuanense, with only differences in the strand distribution of trnL1, trnL2, and rrnS (Fig. 2). As their strand distribution in B. fukienense corresponds to the ancestral arrangement for Potamidae (Fig. 2), we propose that these three genes underwent strand switches in B. lingchuanense. The additional steps needed to explain the architecture of the two chromosomes comprise tandem duplications of the nad1-L1-rrnL-V and rrnS-I-M sections in the ancestral (non-fragmented) mitogenome (Fig. 2). Following this, double-stranded DNA breaks (//) occurred in two places, between both duplicated sections: nad1-L1-rrnL-V//nad1-L1-rrnL-V and rrnS-I-M//rrnS-I-M. This was followed by the circularisation of the two fragments, and a deletion of the sequence fragment comprising trnI and trnM on chromosome 1 (Fig. 2). To confirm the deletion, we compared the large non-coding region (NCR) between rrnS and trnI on chromosome 2 spanning 1432 bases, and the NCR between rrnS-1 and pseudo-nad1 on chromosome 1 spanning 738 bases. The alignment indicated that the sequence of the smaller NCR was highly conserved in comparison to the larger NCR (≈ a dozen SNPs and deletions). This supports our hypothesis that the downstream fragment of the duplicated rrnS-I-M segment was lost during the fragmentation and recircularization, causing the loss of trnI and trnM genes from chromosome 1. The only extra step required to produce the two chromosomes was the conversion of the ancestral trnV on chromosome 2 into pseudo-trnL2, as discussed above.

We further tested the hypothesis that the existence of tandem repeats may facilitate intramolecular recombination of strand slippage of replication machinery [48]. Tandem repeats were identified in two large NCRs on chromosome 1: 791 bp between trnQ and cox2, and 574 bp between trnE and trnH. In the 791 bp NCR, there were two tandem repeats: 18 × 2 bases and 12 × 9 bases. In the 574 bp NCR, there were seven different tandem repeats: 76 × 2 bases, 19 × 8 bases, 3.4 × 47 bases, 19 × 12 bases, 7.5 × 36 bases, 30 × 12 bases, and 17.4 × 10 bases (Additional file 1: Table S1). These two NCRs do not correspond to regions where putative sequence duplication and mitogenome fragmentation events took place, although the 791 bp NCR is in the proximity of a putative tandem duplication region, comprising nad1-L1-rrnL-V, upstream from trnQ.

Gene order rearrangements and sequence evolution

We merged the two chromosomes of B. fukienense into a single mitogenome and conducted a range of comparative mitogenomic architecture analyses to assess whether it exhibits any other unique evolutionary features, aside from fragmentation and duplication of multiple genes. While gene orders were generally conserved in Potamidae, the two available Bottapotamon species exhibited highly rearranged gene orders (Fig. 4). More precisely, the clade comprising two Bottapotamon and three Neilupotamon species was the only one exhibiting rearrangements of protein-coding genes in comparison to the ancestral arrangement, observed in most Potamidae mitogenomes (Figs. 2 and 4). The sister clade to these two genera, comprising Parapotamon, Sinopotamon, Tenuilapotamon, and Longpotamon genera, exhibited the ancestral PCG order, but the common ancestor of the clade underwent a translocation of two rRNA genes together with a block of tRNA genes. Among the species from the remaining 16 genera, only minor rearrangements in tRNA genes in comparison to the ancestral architecture were observed. Therefore, there are indications of progressively elevated mitochondrial architecture rearrangement rates within the clade comprising these six genera, culminating with the fragmented mitogenome in B. fukienense. The evolutionary scenario of rearrangements is unclear, as all species apart from Neilupotamon papilionaceum exhibit the same PCGs + rRNAs order (notwithstanding gene duplications and fragmentation in B. fukienense), with the large box comprising S2-nad1-L1-rrnL-V-Q genes translocated between cox1-L2 and cox2 genes. However, N. papilionaceum exhibits the ancestral arrangement of the above gene box, but with the M-nad2-W-C-Y gene box translocated within it, between nad1-L1 and rrnL. As we can infer with confidence the existence of annotation and assembly errors in the mitogenome of N. papilionaceum (atp6 and nad6 were not annotated, and at 16,273 bases, the mitogenome was much smaller than mitogenomes of related species, which indicates that it might be incomplete), we also cannot exclude assembly errors. It is necessary to sequence more species belonging to these two genera with fast-evolving mitogenomic evolution.

To test the hypothesis that elevated architectural rearrangement rate and ultimately fragmentation are associated with relaxed purifying selection pressures (i.e. higher overall evolutionary rates), we compared evolutionary rates (branch lengths) among different lineages. Bottapotamon fukienense exhibited an average branch length among the Potamidae (0.393; average = 0.389). As a result of the relatively narrow range of branch lengths in the dataset (0.330 and 0.414), no spurious species were identified in the dataset. We further confirmed this using long-branch scores, which ranged from − 18.8 to 34.6 in Potamidae, and Bottapotamon fukienense (-6.3) exhibited a lower-than-average value within the dataset (average ≈ 0). Similarly, the relative composition variability values (RCV) were low across the dataset (0.017 to 0.049), aside from the major outlier of N. papilionaceum (0.215). As this species exhibited two missing genes, this is the most likely underlying reason for this orders-of-magnitude increase in RCV. Bottapotamon fukienense exhibited an average RCV (0.026) within the Potamidae dataset (average = 0.028; the outlier excluded). Finally, we compared branch lengths between the “fast-evolving” clade comprising the six genera exhibiting relatively rapidly evolving architecture and the remaining 16 “conserved” genera. Again, branch lengths were almost identical between the two clades (0.397 vs. 0.393 respectively; df = 41, t-value = 0.66, p-value = 0.51).

To further test the hypothesis that fragmentation is associated with relaxed purifying selection pressures, we also conducted selection tests. With B. fukienense selected as the test branch, and all other Potamidae as background, the test of differences in selection pressures between the two groups was nonsignificant (K = 0.94, p = 0.68, LR = 0.18). With the two Bottapotamon species selected as the test branches, and all other Potamidae as background, the test of differences in selection pressures between the two groups was nonsignificant (K = 1.02, p = 0.75, LR = 0.10). With the group of six genera with elevated architectural evolutionary rates selected as test branches, the RELAX algorithm found significant (p = 0.001, LR = 12.09) evidence of selection intensification in test branches.

To test the hypothesis that elevated architectural rearrangement rates are associated with increased mitogenome size, we compared mitogenome size among the Potamidae. These analyses can be strongly affected by sequencing and assembly artefacts, so our findings must be interpreted with these limitations in mind. As N. papilionaceum was missing multiple genes and exhibited a small size, we removed it from the dataset. Potamidae exhibited large mitogenomes on average (17.5 Kbp), but when we merged the two P. fukienense chromosomes into a single mitogenome, it exhibited the largest size (23,284 bp) among the Potamidae (Additional file 2: Sheet S1). It was much larger than the second-largest described Potamidae mitogenome: Parapotamon spinescens with 20,027 bp (notably, this mitogenome was incomplete) [25]. Aside from gene duplications, this exceptional size was also a product of this species exhibiting by far the highest ratio of NCRs within the dataset: 34.6%, followed by Parapotamon spinescens at 26.4% (Additional file 2: Sheet S1). Combined, the four species from the two genera with fast-evolving architecture, Bottapotamon and Neilupotamon, exhibited significantly larger mitogenomes than the remaining lineages (19,186 vs. 17564, df = 40, t-value = 2.081, p-value = 0.044). The six genera exhibiting relatively fast architecture evolution also exhibited significantly larger mitogenomes than the remaining lineages (18,670 vs. 17,133, df = 40, t-value = 3.555, p-value = 0.001).

In terms of base composition, the mitogenome (both chromosomes merged) of B. fukienense exhibited a high, but average within the dataset, AT bias of 73.1% (Additional file 2: Sheet S1). The G-base content was the second smallest within the dataset (8.0%) after only Parapotamon spinescens (7.7%). It had a low AT skew (-0.007), but the largest GC skew in the dataset (-0.408). However, GC skew magnitude was not exceptional, as the range in the remainder of the dataset was from − 0.222 to -0. 399. Terrapotamon thungwa (MW697087) exhibited a fully inverted GC skew (0.32), but this was an artefact caused by the authors submitting the minority strand to the GenBank (Fig. 4). There were no other species exhibiting skew inversions in the dataset.

Start and stop codons were highly conserved and standard in PCGs of B. fukienense: ATA, ATG, ATT, and TAG, TAA, T– respectively (Table 1, Additional file 2: Sheet S2). In terms of length and sequence, most PCGs were highly conserved across the entire family. Atp8 was remarkably conserved in both aspects across the Potamidae (159 bp in all available species). This is unusual, as this is commonly the fastest-evolving mitochondrial PCG [49]. Intriguingly, B. fukienense was the only exception in terms of length of atp8, with 162 bases, caused by a single amino acid insertion (S; TCT base triplet) (Additional file 1: Figure S8). Cox1 exhibited an elongation comprising a single base triplet at the 3’-end (TTA), but this was shared by B. lingchuanense and three other closely related species. Nad2 was highly conserved in comparison to other orthologues in B. fukienense, but in congeneric B. lingchuanense, it exhibited a large 15 amino acids insertion and a smaller one of 3 amino acids near the 5’-end. Other PCGs were highly conserved.

Overlaps between genes were small (≤ 2 bases), apart from the nad4-nad4L and atp6-atp8 pairs of genes, which commonly overlap in many animal species [1]. Surprisingly, trnL2 overlapped with cox1 by 5 bases, which is uncommon, as cox1 is a highly conserved gene. However, this overlap is conserved across many Potamidae [25].

In terms of phylogenetic relationships of the remaining lineages, Potamiscus was paraphyletic in both ML and BI topologies, with Potamiscus yongshengensis and Potamiscus motuoensis clustering with Tenuipotamon, Lophopotamon and Aparapotamon, but their exact positions varied between the two topologies (Fig. 4 and Additional file 1: Figure S9). The paraphyly of this genus is a known problem [50]. Longpotamon was also paraphyletic, with Longpotamon parvum comprising a sister lineage to Tenuilapotamon latilum. ML and BI topologies differed in the position of the Indochinamon lineage. The position of the Sinopotamon lineage differed between BI and ML topologies, and its relationship to the L. parvum + T. latilum clade.

Fig. 4
figure 4

The mitogenomic architecture and Maximum Likelihood phylogeny of Potamidae inferred using amino acid sequences of 13 mitogenomic protein-coding genes. Bootstrap support values < 100 are shown at nodes. Esanthelpusa is the outgroup. Two chromosomes of B. fukienense are merged into a single mitogenome

Discussion

Herein, we sequenced the mitochondrial genome of a Chinese freshwater crab species B. fukienense and found that it is fragmented into two chromosomes. At the time of this study (January 2024), there was only congeneric one mitogenome available: B. lingchuanense [51]. It exhibited a standard architecture, but the authors failed to annotate two tRNA genes: tRNA-Ile and tRNA-Met. A recent review of all crustacean mitogenomes failed to observe a single fragmented mitogenome [17], with a minor exception of the mitogenome of the abyssal amphipod Hirondellea gigas, for which the authors submitted two contigs that they failed to assemble into a single mitogenome [52]. This may indicate fragmentation, but it is also possible that they simply failed to sequence noncoding regions completely, preventing the assembly of the two contigs into a circular mitogenome. In addition, heteroplasmy and a minor proportion of fragmented mitogenomic phenotypes have been proposed in Callinectes sapidus (Decapoda) [20], but the authors merely relied on unmapped reads to infer fragmentation and did not confirm this using additional experiments. Similarly, in a recent study of freshwater crab mitogenomes, authors found that mitogenomes of Parapotamon spinescens and Tenuilapotamon latilum could not be circularized due to sequence gaps (two in the former, one in the latter) [25]. Again, we cannot conclude whether this may be indicative of a circularisation or sequencing artefacts. Therefore, to our knowledge, this is the first experimentally confirmed fragmentation of a mitogenome in crustaceans. In addition, as opposed to results in Callinectes sapidus, we did not find indications of the existence of non-fragmented mitogenomic phenotypes in the DNA of B. fukienense. Importantly, we conducted two rounds of sampling and sampled geographically distant populations. This way we confirmed that fragmentation was not limited to a single specimen or population. In addition, the above studies relied on NGS, and there is evidence that in some cases NGS approaches (e.g. Illumina sequencing) can produce incomplete mitogenomic assemblies [53]. Herein, we relied on Sanger sequencing, which is slower and more expensive, but it allowed us to design primers producing overlapping segments and confirm the completeness and fragmentation of this mitogenome. Notably, we also encountered multiple problems in the amplification and sequencing of this mitogenome. Particularly difficult to sequence were nad5, nad2 and trnN genes. This may be caused by the existence of repetitive sequences [54], but also be indicative of heteroplasmy (multiple sequences existing within a single individual), which appears to be common in lineages with fragmented mitogenomes [55]. Regardless, all sequenced fragments assembled into two mitogenomic circles. Finally, a number of factors allowed us to exclude the possibility that these two mitogenomic fragments are numts (nuclear genome-encoded mitogenomic fragments) with high confidence, comprising circularity, large size, the absence of stop codons in PCGs, etc. [56, 57].

We also observed large sequence duplications, producing several duplicated genes (rrnS and trnL1) and pseudogenes (nad1, rrnL and trnL2). While duplications and expansion of noncoding regions are relatively common in crustaceans [17] and other invertebrates, such extensive gene duplications and pseudogenisations are uncommon [58]. They have been reported in several Nematoda lineages [6, 59, 60], and isolated insect [61] and reptile (lizards and salamanders) [62, 63] lineages. Our reconstruction of the ancestral mitogenomic structure for B. fukienense indicated that it probably underwent tandem duplications of two mitogenomic segments, spanning multiple genes. Notably, a similar tandemly duplicated region spanning multiple genes was previously observed in a decapod crustacean, Homarus gammarus [53]. The putative duplication mechanisms comprise the slipped-strand mispairing mechanism [64] or hybridisation of two uneven mitogenomic minicircles [4].

This also brings us to the putative mechanism leading to the observed architecture. The existence of multiple mitotypes (e.g. complete and fragmented) within a single cell and mitogenomic recombination were proposed as key parts of the mechanism explaining the evolutionary occurrence of multipartite mitogenomes in multiple lineages [4, 6, 8, 11, 55]. The generation of long-lived minicircles was also proposed as an intermediate step in explaining the high levels of gene order rearrangements if it is followed by their integration into the full-length mitogenome via homologous recombination [4]. Indeed, it has been shown that mitochondrial architecture of some lice with fragmented mitogenomes evolves by splits and mergers of minichromosomes [65]. Mao et al. [4] proposed that these minicircles should contain the origin of replication in order to be long-lived, but identification of the control region and the origin of replication in crustaceans was proven to be nearly impossible due to high levels of divergence in noncoding regions and the absence of conserved motifs associated with the origin of replication [17]. However, we did identify a highly conserved noncoding segment shared by both mitogenomic chromosomes adjacent to rrnS, which might comprise the control region. In indirect support of this hypothesis, genes flanking the origins of strand replication usually form hotspots of duplications [66,67,68,69], and our reconstruction of the scenario leading to the observed mitogenomic architecture proposed a duplication of the rrnS-I-M gene box (Fig. 2).

We found indications of increased mitogenomic architecture rearrangement rates in the clade comprising sister genera Bottapotamon and Neilupotamon. More precisely, this was the only clade that exhibited rearrangements of PCGs compared to the ancestral arrangement. In comparison to relatively variable tRNA genes, PCGs and rRNAs commonly exhibit much less variable arrangements [58, 70,71,72], so within-family variability in the arrangement of PCGs would be considered unusual in most bilaterian lineages [1, 58, 73]. However, it has been observed in several crustacean families; e.g. Chthamalidae, Allocrangonyctidae, Lysianassidae, Palaemonidae, Nephropidae, and Parastacidae [17, 53, 74]. Intrageneric rearrangements of genes other than tRNAs are even rarer, and a previous study reported that they are completely absent from crustaceans [58]. In this aspect, the genus Bottapotamon, exhibiting two different arrangements of PCGs is unique among crustaceans. This also indirectly supports the observation that fragmentation events are more common in lineages with high architectural evolutionary rates [11]. Notably, the same group of authors stressed that due to the rarity and likely randomness of fragmentation events, there is no direct correlation between the two variables.

Two studies proposed that the fragmentation of mitogenomes is associated with relaxed purifying selection pressures on mitochondrial genes and increased sequence evolutionary rates [55, 75]. Furthermore, elevated architectural evolution rates are correlated with elevated sequence evolution rates in a range of animal lineages [11, 71, 73, 76,77,78]. Contrary to these previous findings, we did not find any evidence of elevated mutational rates or relaxation of purifying selection pressures in B. fukienense, nor in the genera exhibiting rearrangements of PCG and rRNA genes. Several previous studies also reported the existence of exceptions from this presumed correlation between the sequence and architecture evolutionary rates [14, 16, 71, 79].

As a result of sequence duplications described above, B. fukienense also exhibited the largest mitogenome among the Potamidae. This indirectly supports the observation that mitogenome size is often positively correlated with gene order rearrangement rate, observed in some [80], but not all [81] nematodes and flatworms [14]. The mitogenome also exhibited the highest GC skew magnitude within the available Potamidae dataset. However, as GC skew magnitude is driven primarily by the strength of purifying selection pressure [82], NCRs are expected to exhibit elevated GC skews. Therefore, the high skew in this mitogenome is most likely a consequence of the largest ratio of noncoding regions within the dataset (35%), and the associated highly relaxed purifying selection pressures. Indeed, the GC skew value on the coding section of the mitogenome was much lower (-0.313), and only slightly above the average value (-0.284; range from − 0.212 to -0.367) for the entire dataset (Additional file 2: Sheet S1). This indicates that sequence duplications in this mitogenome are not of a very recent evolutionary origin, as a certain evolutionary time is presumably required for GC skew magnitude to increase even in the absence of purifying selection pressures [17, 82, 83]. Contrary to this, relatively low levels of sequence divergence between the duplicated fragments indicate that the duplication did not occur deep in the evolutionary history, so we hypothesise that it might be confined to this species. The presence of pseudogenes further supports this, as they are expected to be removed from the genome by the selection for small size relatively quickly [67]. Further samples from closely related species and additional populations of B. fukienense are needed to place this fragmentation more precisely in evolutionary time.

Conclusions

Fragmentation aside, the mitogenome of B. fukienense exhibited multiple other signs of elevated mitogenomic architecture evolution rates, including the exceptionally large size, duplicated genes, pseudogenisation, and high PCG rearrangement rate, but there is no evidence that this is matched by elevated sequence evolutionary rates or changes in selection pressures. As genomic rearrangements are random evolutionary events, we can speculate with relative confidence that the fragmentation event itself was a nonadaptive evolutionary accident. However, this species clusters within a clade of Potamidae exhibiting progressively elevated architecture rearrangement rates and mitogenomic sizes, culminating with fragmentation in B. fukienense. This overall trend is indicative of elevated architectural dynamics, likely associated with mutations in mitogenomic replication and maintenance machinery encoded in the nuclear genome [71, 84]. The evolutionary future of this fragmentation is unclear. On one hand, small mitochondrial genomes (chromosomes) are believed to have transmission advantages over large genomes/chromosomes due to shorter replication time [7, 55, 85], the fact that the vast majority of animals do not possess fragmented mitogenomes indicates that purifying selection acts against mitogenomic fragmentation. Indeed, given the central role of locomotory capacity in the strength of purifying selection acting on mitochondrial genomes [82, 86, 87], it appears to make sense that a lineage with low locomotory capacity, lice, is the only one with evolutionarily widespread fragmentation of mitochondrial genomes. This may indicate that a B. fukienense phenotype with a complete mitogenome (arising randomly through hybridisation) might have an evolutionary advantage over the fragmented phenotype, and that this species may revert to the ancestral state in its evolutionary future. Continued genetic monitoring of populations of this species might offer novel and important insights into the mitochondrial evolutionary dynamics.