Paternally Transmitted Mitochondria Express a New Gene of Potential Viral Origin

Mitochondrial ORFans (open reading frames having no detectable homology and with unknown function) were discovered in bivalve molluscs with doubly uniparental inheritance (DUI) of mitochondria. In these animals, two mitochondrial lineages are present, one transmitted through eggs (F-type), the other through sperm (M-type), each showing a specific ORFan. In this study, we used in situ hybridization and immunocytochemistry to provide evidence for the expression of Ruditapes philippinarum male-specific ORFan (orf21): both the transcript and the protein (RPHM21) were localized in spermatogenic cells and mature spermatozoa; the protein was localized in sperm mitochondria and nuclei, and in early embryos. Also, in silico analyses of orf21 flanking region and RPHM21 structure supported its derivation from viral sequence endogenization. We propose that RPHM21 prevents the recognition of M-type mitochondria by the degradation machinery, allowing their survival in the zygote. The process might involve a mechanism similar to that of Modulators of Immune Recognition, viral proteins involved in the immune recognition pathway, to which RPHM21 showed structural similarities. A viral origin of RPHM21 may also support a developmental role, because some integrated viral elements are involved in development and sperm differentiation of their host. Mitochondrial ORFans could be responsible for or participate in the DUI mechanism and their viral origin could explain the acquired capability of M-type mitochondria to avoid degradation and invade the germ line, that is what viruses do best: to elude host immune system and proliferate.


Introduction
One intriguing observation that emerged from sequencing of whole mtDNAs is the occurrence of numerous ORFans (open reading frames having no detectable homology and with unknown function) showing homologies in closely related species only. Interestingly, some of these genes were shown to be involved in key biological functions (Monchois et al. 2001;Khalturin et al. 2008Khalturin et al. , 2009. For example, novel mtDNA-encoded proteins are known to act on gamete formation in angiosperm plants exhibiting cytoplasmic male sterility (CMS) (Chase 2007). Novel lineage-specific mitochondrial ORFs (Breton et al. 2009(Breton et al. , 2011a(Breton et al. , 2011bGhiselli et al. 2013) were also discovered in animals with doubly uniparental inheritance (DUI) of mitochondria (Skibinski et al. 1994a(Skibinski et al. , 1994bZouros et al. 1994aZouros et al. , 1994b. In metazoans, mitochondria are commonly inherited by strictly maternal inheritance (SMI) (Birky 2001), but, in some bivalve molluscs, two mitochondrial lineages are present, one transmitted through eggs (F-type) and the other through sperm (M-type).
All the analyzed DUI species show the following two interesting features: 1) The existence of sex-biased progeny was observed in species of the Mytilus edulis complex (Saavedra et al. 1997;Kenchington et al. 2002;Cogswell et al. 2006), and recently also in Ruditapes philippinarum . The sex bias appears to be controlled by the nuclear genome of the females which can be accordingly subdivided in three classes: those producing predominantly males, those producing predominantly females, and those producing females and males at intermediate frequencies.
2) In embryos of DUI species the distribution of sperm mitochondria follows two different patterns: aggregated or dispersed (Cao et al. 2004;Obata and Komaru 2005;Cogswell et al. 2006;Milani et al. 2011Milani et al. , 2012. In Mytilus, it was possible to correlate the pattern type to the sex of the embryo: male-biased offsprings showed a prevalence of aggregated patterns, while female-biased offsprings a prevalence of dispersed patterns. It was then proposed that in developing males the aggregate of sperm mitochondria enters the primordial germ cells (PGCs), providing an early mechanism through which sperm mitochondria are segregated into the germ line of male embryos; after that, the M mitochondrial genome becomes dominant in male germ line, as indicated by the exclusive presence of M-type mtDNA in spermatozoa (Venetis et al. 2006;Ghiselli et al. 2011). In DUI female embryos, sperm mitochondria are dispersed, diluted or possibly degraded, and the mitochondrial inheritance would be as in SMI species.
The peculiar segregation pattern of spermatozoon mitochondria, correlated with the existence of sex-biased lineages, led to the hypothesis that M-type mitochondria may have an active role in gonad masculinization, achieved through a series of specific signals between nucleus and mitochondria (Kenchington et al. 2002). This is not unthinkable, because the involvement of mitochondria in germ line development is well established. For example, in Drosophila, mitochondrial ribosomes are found extramitochondrially and are required to produce proteins that are necessary for germ cell formation (Amikura et al. 2001 and references therein). Actually, in early embryos, a mitochondrial-type translation is required for germ cell formation, which is disrupted by the injection of prokaryotic translation inhibitors (Amikura et al. 2005). Moreover, translation of nuclear-encoded proteins by mitochondrialtype ribosomes was detected in mammalian sperm (Gur and Breitbart 2006).
The molecular mechanisms underlying DUI are still unknown; however, many structural and functional features of M and F mtDNAs were proposed as candidates for a role in mitochondrial inheritance and germ line establishment/differentiation Milani, Ghiselli, Guerra, et al. 2013;Zouros 2013). Among these candidates, novel lineagespecific ORFs found in DUI species belonging to the families Unionidae (Breton et al. 2009(Breton et al. , 2011a, Mytilidae (Breton et al. 2011b), and Veneridae  were proposed. So far, the existence of the translation product was verified in the unionid Venustaconcha ellipsiformis only (Breton et al. 2009(Breton et al. , 2011a. Milani, Ghiselli, Guerra, et al. (2013) used multiple in silico approaches to perform a comparative analysis of DUI mitochondrial ORFans and proposed their origin through viral endogenization.
In the DUI species R. philippinarum, the novel lineage-specific ORFs are localized in different position in the two mitochondrial genomes. The F-specific ORF is located in the largest unassigned region of the F-type mtDNA and is not present in the M-mtDNA, while the M-specific ORF is located in the Unassigned Region 21 (UR21) of the M-type mtDNA, and is not present in the F-mtDNA. Transcriptomic data  showed that, while the F-specific ORF is poorly transcribed compared with the mitochondrially encoded electron transport chain (ETC) genes, the M-specific ORF is transcribed at the same level of ETC genes. Moreover, while the F-specific ORF shows a few nonsynonymous substitutions , the amino acid sequence of M-specific ORF is 100% conserved in all analyzed males. These findings suggest M-specific ORF functionality. This study is focused on the M-specific ORF, which will be here referred to as orf21. We performed an in-depth analysis on orf21 transcript and on its product, RPHM21 (from R. philippinarum Male mitochondrial protein 21): we localized orf21 transcripts by in situ hybridization (ISH), and synthesized a specific antibody to verify the existence, by Western blotting and confocal microscopy, of RPHM21 in adults and early embryos. Moreover, we performed additional analyses of orf21 flanking region and RPHM21 structure to further support its possible viral derivation and to find similarity with known proteins. We discussed the putative function of RPHM21, its role in DUI and embryonic development, its origin and the possibility that such novel gene was co-opted acquiring important functions in R. philippinarum development.

Materials and Methods
orf21 Transcript Localization with ISH ISH analysis was performed on R. philippinarum gonads and surrounding tissues of five specimens, three males and two females. Overall, we processed and checked 60 sections. An antisense probe against the strictly conserved novel mitochondrial ORF in the Male Unassigned Region 21 (MUR21) of M-type mtDNA was produced. The probe, containing the entire orf21 sequence of 519 bp, was generated from the M-type mitochondrial whole genome sequence present in GenBank (AB065374). The primers used were orf21-forward 3 0 -GGTAGCACAAGGTTTCCAGAGTTTATGTGT-5 0 , and orf21reverse 3 0 -ACTTGTAAACCAGGGGTAAGAGGTCACA-5 0 . The T7 RNA polymerase-binding sequence was added 5 0 to the antisense DNA sequence. RNA extraction was performed by TRIzol RNA Isolation Reagents (Life Technologies). The target sequence was amplified by PCR from cDNAs obtained by SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen -Life Technologies). A Digoxigenin (DIG)-labeled riboprobe was obtained using the Roche in vitro transcription labeling protocol (Roche DIG RNA labeling kit). ISH was performed on 10-20 mm cryosectioned sections, following the method in (Milani, Ghiselli, Nuzhdin, et al. 2013). Samples were analyzed with a Nikon Eclipse 80i microscope and images were captured using NIS-Elements D3.2 software.

Spawning Induction and Fertilization
Ruditapes philippinarum clams used for mating were collected during the reproductive season (July/August 2012). The spawning was induced with 30-min cycles of cold (22 C) and warm (29 C) artificial sea water (i.e., reverse osmosis water added with RedSea Coral Pro aquariology sea salt). Two matings were performed: two females were fertilized with sperm from the same male. After fertilization, embryo development was visually checked with an optical microscope. Embryo samples were collected at 40-45 min, 60-65 min, and 2 h, timing that corresponds to 2, 4-8, and 16-32 blastomere stages (see Milani et al. 2012).

SDS-PAGE and Western Blotting
Pieces of male and female gonads were frozen in liquid nitrogen, stored at À80 C, and then processed by SDS-PAGE (sodium dodecyl sulphate-polyacrylamide gel electrophoresis) and Western blotting. Samples were homogenized in RIPA buffer (Tris-HCl 50 mM, pH 7.4, added with 150 mM NaCl, 1% NP-40, 0.25% Na-deoxycholate, 1 mM EDTA) in the presence of protease inhibitor cocktail tablets (Complete Mini, Roche) and 1 mM PMSF, utilizing an Ultra Turrax T25 Janke & Kunkel IKA-labortechnik. Next, samples were centrifuged 10 min at 4 C at 10,000 rpm. The supernatant was treated with 0.25-0.6 U/ml of DNase I (from bovine pancreas; Sigma-Aldrich) and 0.004-0.009 U/ml of RNase A (from bovine pancreas; Sigma-Aldrich) in a 50 mM MgCl 2 solution, followed by 30 min incubation on ice, to remove nucleic acids that prevented a correct running during electrophoresis.

Gonads
Pieces of gonads were fixed in 3.7% paraformaldehyde and 0.1% glutaraldehyde solution in K-PIPES buffer (pH 7) for 3 h and embedded in 7% agar. Sections of about 100-mm thickness were made using a Lancer Vibratome Series 1000 and post-fixed with increasing concentrations of methanol. Sections were rehydrated and then treated 1 h and 15 min with sodium borohydride 70 mM in TBS (pH 7.4) at RT. After rinsing 1 h and 15 min in TBS-Triton 0.1%, sections were treated with 0.01% Pronase E, as above, but for 18 min at RT. After, samples were permeabilized adding TBS-Triton 1% and left over night at 4 C.
Controls were performed using samples in which the first or the second antibody was omitted, or samples treated only with normal serum. Moreover, in additional control samples, the synthetic peptides used in antibody production were added to the primary antibody solution at a ten-fold concentration, finally accessing the decreasing staining intensity and confirming the reaction specificity (the abcam protocol was utilized as method to block with immunizing peptide). Imaging was recorded by a confocal laser scanning microscope (Leica confocal SP2 microscope), using Leica software.

Structure and Function Analyses
The orf21 translation (putative protein RPHM21; from the M-type complete mitochondrial genome; GenBank: AB065374) was analyzed with several programs. Given the ORFan status of RPHM21, we needed to find signatures of short sequences and remote homology, so we used tools that were specifically developed for these tasks. ProTeus (http://www.proteus.cs.huji.ac.il/, last accessed February 10, 2014; Bahir and Linial 2005) detects previously known and potentially overlooked signatures in protein termini. It focuses on signatures that are often undetected by most search methods due to an inadequate statistical significance score that fails to recognize very short signatures (Bahir and Linial 2005). HMMER (http://hmmer.janelia.org/search/phmmer, last accessed February 10, 2014; Finn et al. 2011) searches sequence databases for protein homologs using profile hidden Markov models. Compared with other search tools based on the classic scoring methodology, it is more accurate and more able to detect remote homologs (Finn et al. 2011 To predict protein secondary structure, we used Quick2D (http://toolkit.tuebingen.mpg.de/quick2_d, last accessed February 10, 2014): it finds secondary structure features like alpha helices, extended beta sheets, coiled coils, transmembrane helices (TMHs), and disorder regions. The advantage of this tool is that it combines the results of up to 11 different structure prediction algorithms (see website for details). Quick2D was used to predict protein secondary structure of both RPHM21 and MK3, an E3 ubiquitin ligase (also named MIR1 of [Murid herpesvirus 4], NCBI Reference Sequence: NP_044852.1) to which many function similarities were found with RPHM21.
Since structural predictions revealed the putative presence of transmembrane domains in both RPHM21 and MK3, we used TM-Coffee (http://tcoffee.crg.cat/apps/tcoffee/do:tm coffee, last accessed February 10, 2014; Chang et al. 2012) to produce their alignment. To obtain a model for protein transmembrane topology, we used PredictProtein (http://www.predictprotein.org/, last accessed February 10, 2014; Rost et al. 2004), which utilizes 30 different methods (see Rost et al. 2004 for details) of structural prediction. To infer the subcellular localization, we also used LocTree2 (https://www.rostlab.org/services/loctree2/, last accessed February 10, 2014; Goldberg et al. 2012) that is particularly efficient in de novo prediction. In general, given the absence of RPHM21 detectable homologs in the databases, we decided to use tools able to detect remote homology and to predict protein features de novo or by structural similarity. Moreover, we favored multialgorithm approaches (and combinations of them), because we feel that getting converging results from several different methods makes the derived assumptions stronger, strengthening the results.
TOPO2 was used to obtain a protein structure representation (http://www.sacs.ucsf.edu/TOPO2/, last accessed February 10, 2014). An analysis of the whole MUR21 nucleotide sequence was performed with BlastN. All databases were accessed by August 2013.

Results
Localization of orf21 Transcripts and RPHM21 Protein ISH with orf21 antisense probe in R. philippinarum male gonad showed a specific transcription in both the acinus (gonadic unit containing germinal cells) and its lumen (the cavity of the acinus that is full of gametes at maturity), indicating a positive reaction with both spermatogenic cells ( fig. 1A and B) and mature spermatozoa that are free in the lumen ( fig. 1C-E). The staining was stronger along the acinus wall ( fig. 1). The probe did not stain the eggs ( fig. 1F), in agreement with the absence of M-type mtDNA from all the egg samples tested so far .
Immunoblotting of gonadic extracts was performed using an antibody (anti-RPHM21) generated against peptides synthesized from the predicted amino acid sequence in the C-terminus of the protein (two adjacent peptides of 17 and 16 aa). In male gonadic extracts, the Western blot showed a band of~20 kDa ( fig. 2). Actually, 19.45 kDa is the molecular weight of RPHM21 estimated from the amino acid sequences. No band of the corresponding molecular weight was detected by Western blot of female gonadic extracts ( fig. 2).
Only male gonadic tissue showed a staining with anti-RPHM21. The staining was localized in spermatogenic cells, with a deeper labeling in mature spermatozoa, both in mitochondria (in the four mitochondria of the spermatozoon midpiece) and in the nucleus ( fig. 3). No staining was visible in surrounding tissues (connective and intestine) or in eggs ( fig. 3G).
Some of the embryos of the two matings showed a spotted staining, while others did not show any staining. The spot number and dimension appeared to increase from the twoblastomere stage to the 32-cell embryo. Four-and eight-cell embryos had a deep staining localized around the animalvegetal axis ( fig. 4). In 32-blastomere embryos, the staining was localized in micromeres. No staining was present in control embryos ( fig. 4A).

Function and Structure of RPHM21
The analysis of the whole MUR21 nucleotide sequence with BlastN revealed in orf21 one significant similarity with Bacteriophage phi-C31 complete genome (table 1). On the whole, 44 hits of phages and viruses were detected in MUR21, mainly in orf21 sequence and in the repeat region upstream orf21 (table 1 and fig. 5).
RPHM21 analysis with HMMER showed one significant hit (E-value ¼ 0.05), that is Cydia pomonella granulosis virus (Baculoviridae, dsDNA virus) ORF140 similar to XcGV ORF178 (containing a domain similar to Fibroblast growth factor). The other three hits found were as follows: a Zinc-binding protein (Drosophila melanogaster) (Evalue ¼ 0.33); Autophagy protein Apg17 (Dictyostelium discoideum) (E-value¼ 0.59); ABC-2 family transporter protein (Thermofilum pendens Hrk 5) (E-value ¼ 0.9). Despite the nonsignificant E-values (HMMER E-values of 0.1 or less are significant; see HMMER User Guide Version 2.2, page 9), all these three hits are supported by other in silico analyses (Milani, Ghiselli, Guerra, et al. 2013). We also would like to point out that the E-value (which measures the number of hits that are expected to be found by chance when searching a database) depends strongly on two parameters: the database size and the sequence length. The larger the database and the , where a lumen will form after sperm maturation, is not stained; (B) mature spermatozoa begin to be stored in the acinus center during the lumen formation (asterisk). orf21 transcript appears to be more present (higher staining) along the acinus wall (arrow in A and B); (C-E) mature acini with the lumen full of stained mature spermatozoa (s) (well evident in D). (A-E scale bars ¼ 100 mm); (F) no staining is present in R. philippinarum eggs (n ¼ egg nucleus; c ¼ egg cytoplasm) (F scale bar ¼ 25 mm). Positive signal with antisense probe hybridization stained dark blue.
shorter the query sequence, the greater the chance to find a match by chance. As we are dealing with short sequences searched in quite large databases, it is possible that nonsignificant but borderline results as those reported are actually reliable.
FUGUE found two hits with high confidence: the first is ankyrin repeat domain of trpv1 (transient receptor potential cation channel) (Z score ! 4.0, 95% confidence), and the other hit is sprpn10 vwa domain (26s proteasome regulatory subunit rpn10) (Z score ! 3.5, 90% confidence), a ubiquitin interacting motif. ProTeus found 1 group with signature FIL: electron transport GO biological process, supporting some kind of membrane association. PredictProtein predicted the presence of two TMHs (20-37 and 43-61 residues). Mitochondria were indicated as possible subcellular localization (LOCtree implemented in PredictProtein: reliability index ¼ 6 over 9; values !4 denote good predictions), and also the nucleus, even if with a low prediction score (LOCtree2 prediction: score ¼ 17 over 100). The proposed RPHM21 position/orientation is reported in figure 6. According to the protein secondary structures predicted with Quick2D, both RPHM21 and the ubiquitin ligase MK3 (protein resulted similar to RPHM21) of Murid herpesvirus 4 showed two TMHs and a cytoplasmic C-terminus (supplementary figs. S1 and S2, Supplementary Material online).

Discussion
Localization of orf21 Transcript and Its Protein RPHM21 The Manila clam R. philippinarum is strictly gonochoric and its gonad forms every year at the beginning of the spawning season, after which it is degraded (Devauchelle 1990). The gonad includes strongly branching acini made up of germinative epithelium supported by connective tissue (fig. 1F). Gonia localize at the periphery of the acinus, while mature gametes are free in the acinus lumen (Devauchelle 1990). ISH performed with orf21 antisense riboprobe showed a specific transcription in the acinus and its lumen, indicating a positive reaction with both spermatogenic cells and mature spermatozoa ( fig. 1): the staining was present in spermatogenic cells when the acinus lumen was empty, while it was restricted to spermatozoa when they filled the lumen. orf21 showed an high staining also along the acinus wall, the acinus border formed of undifferentiated germinal cells ( fig. 1A and B). The observation that the transcript was present in the acinus wall points to an early transcription, since the first stage of sexual cell differentiation. When spermatogenesis arrests, orf21 transcription in immature germinal cells stops, but orf21 transcripts are still visible in the acinus lumen, meaning that they are stored in mature sperm. As expected, female tissues were not stained, being orf21 specific of the M-mtDNA, which is commonly absent from them .
The existence of an orf21 product (RPHM21 protein) in male gonads was verified by the presence in the Western blot of a band of about 20 kDa ( fig. 2), consistent with the estimated molecular weight of RPHM21 (19.45 kDa). RPHM21 was expressed in male germ cells, while it was absent in female germ cells, since no RPHM21 band was detected. Moreover, RPHM21 protein immunolocalization showed that male germ cells were specifically stained with anti-RPHM21 ( fig. 3), on the contrary, no staining was detected in eggs ( fig. 3G). The staining was localized in spermatogenetic cells, with a deeper labeling in mature spermatozoa, both in mitochondria and nucleus ( fig. 3F). Therefore, orf21 is highly transcribed since the first stage of male sexual cell differentiation (acinus wall cells, i.e., spermatogonia) ( fig. 1), but only in the following stages (moving toward the acinus lumen) the transcripts are translated producing RPHM21 ( fig. 3). In mature sperm, both transcript and the protein were detectable ( figs. 1 and 3). This is the first time a lineage-specific mitochondrial protein is immunodetected in DUI early embryos. Overall, the spot number and dimension in the stained embryos increased from the two-blastomere stage to the 32-cell embryo ( fig. 4). It has to be considered that RPHM21 visualized in developing embryos can be the protein amount already present in the spermatozoon ( fig. 3) and imported with fertilization, plus the product of translation of orf21 transcript stored in spermatozoa ( fig. 1). Moreover, in embryos, newly transcribed orf21 and its product may contribute to the observed staining, thus explaining the increasing of spot number and dimension from the first embryonic stages to the 32-cell embryo.
Four-and eight-cell embryos showed a deep staining around the animal-vegetal axis ( fig. 4). We already showed that germ line determinants are localized in the same embryonic area , so the positioning of RPHM21 in the same region and its increasing amount during embryo growth supports a role in development. In 32-blastomere embryos, the staining was localized in micromeres ( fig. 4). The presence of RPHM21 in all analyzed embryos may appear to be in contrast with its male-specific activity, as supported by our observations in adult tissues. Nevertheless, two considerations have to be done. First, different sex-specific post translational regulation of RPHM21 might be involved in early developmental stages: as discussed in Milani, Ghiselli, Nuzhdin, et al. (2013) main differences between males and females are likely to be quantitative rather than qualitative, since generally there are thresholds to be exceeded to have effects or changes. Second, the presence of RPHM21 in male and female embryos could also have different outcomes because of different epistatic interactions with sex-specific products. In the present work, we had no chance to know the sex of analyzed embryos, because genetic markers of sex are not available. Nonetheless, a sex ratio ranging from 100% females to 90% males was observed in DUI progenies (see Zouros 2013 for a review). The production of sex-biased families for R. philippinarum will allow us to predict the sex of the offspring and check for differential RPHM21 transcription and expression during embryo development and eventually find at what age the differentiation takes place.

Viral Origin of orf21
Recently, in silico analyses of mitochondrial ORFans suggested their possible origin through viral sequence endogenization (Milani, Ghiselli, Guerra, et al. 2013). One clue pointing in that direction was the identification of CRISPR-associated sequences (Cas) in two of them: R. philippinarum MORF and Paphia euglipta ORF (Milani, Ghiselli, Guerra, et al. 2013). Clustered regularly interspaced short palindromic repeats (CRISPR) are a family of DNA direct repeats separated by nonrepetitive spacers of regular size. CRISPR are found in most bacterial and archaeal genomes and seem to provide resistance against mobile genetic elements (Jansen et al. 2002;Bolotin et al. 2004Bolotin et al. , 2005Mojica et al. 2005;Makarova et al. 2006;Barrangou et al. 2007): after viral infection, hosts integrate CRISPR in their genome, enhancing the resistance to phage infection probably via RNA interference (Bolotin et al. 2005). Cas proteins show similarity to helicases and repair proteins (Pourcel et al. 2005) and are thought to be involved in the propagation and functioning of CRISPR (Bolotin et al. 2005). Interestingly, in R. philippinarum, the region upstream orf21 is characterized by repeats that, as CRISPRs, are separated by spacers (motifs e and b, of 28 and 26 bp, respectively; see fig. 5 and Ghiselli et al. 2013). Although other motifs found in R. philippinarum mtDNA intergenic regions showed homology with motifs related to mtDNA replication and transcription, all the performed analyses failed to identify any similarity of e and b with known motifs .
BlastN analysis showed many hits from phages and viruses in MUR21, mainly in the orf21 and in the repeat region (table 1 and fig. 5). Another clue in favor of the hypothesis of a viral origin of RPHM21 comes from the analysis performed with HMMER that retrieved a significant hit with a mitogen (fibroblast growth factor, FGF; i.e., a factor that triggers mitosis) of a baculovirus. FGFs are fundamental for the processes of proliferation and differentiation in a wide variety of cells and tissues, and are involved in embryonic development, cell growth and differentiation, and morphogenesis. Overall, the finding of multiple virus-related hits and the presence of some features that could be ascribable to a CRISPR/ Cas system further support a viral origin of orf21.

Putative Function of RPHM21
Given the absence of related proteins in databases, the function of RPHM21 has to be inferred from the presence of structural motifs showing similarities with known domains. Considering previous data (Milani, Ghiselli, Guerra, et al. 2013) and the data presented here, we can summarize RPHM21 main putative features: 1) the presence of two TMHs; 2) a possible viral origin; 3) the presence of domains involved in immune system and cell cycle regulation; 4) a binding site for ubiquitin, and 5) the presence of domains involved in cytoskeleton interactions. Given these premises, we found a striking similarity between RPHM21 and MK3, a modulator of immune recognition (MIR), which is a viral protein involved in the immune recognition pathway (Coscoy and Ganem 2003). Many viruses have evolved strategies for evading T-cell-mediated host immunity by preventing the display of surface molecules that are recognized by the immune system (Coscoy and Ganem 2003). Many viral regulators cause the ubiquitination of these surface molecules, controlling the trafficking and/or the direct degradation of their targets. MIRs and homolog proteins are involved in such mechanisms functioning as ubiquitin ligases. The proteins of the ubiquitination machinery comprise several ubiquitin-like proteins and hundreds of ubiquitin-conjugating enzymes. Ubiquitin is mainly known for its role as a tag that induces protein degradation either by the proteasome or through targeting to lysosomes, but it is well established that ubiquitination is also a component of most cellular signaling pathways for the control of cell adhesion,  Control sample: eight-blastomere embryo in which the first antibody (anti-RPHM21) was omitted; no staining is detected. Nuclei in green (TO-PRO3); (B) two-blastomere embryo showing a spotted staining of RPHM21 (in red) more concentrated in the region flanking the cleavage between the two blastomere (the animal-vegetal axis is indicated by a white circle) and around nuclei; (C) four-blastomere embryo showing a spotted staining around the animal-vegetal axis (indicated by a white circle); (D) RPHM21 localization in a fourblastomere embryo in which the biggest blastomere (the D one) is shown shortly before the formation of the following segmentation furrow that will form micromere 1d and macromere 1D of the eight-blastomere stage. The future position of micromere 1d (its nucleus is indicated by an arrow) will be in proximity to the animal-vegetal axis where also the nucleus of the polar body is visible (arrowhead). It is evident that micromere 1d will take up most of the cytoplasmic region containing the labeled spots; (E) in 32-blastomere embryos bigger immunostained spots are localized only in micromeres (M ¼ macromeres, m ¼ micromeres); (F) a detail of the 32-blastomere embryo showing some RPHM21 spots surrounding a micromere nucleus (arrowhead). RPHM21 antibody staining in red; TO-PRO3 nuclear dye in green. All the embryos are visualized (at confocal microscope) along Z axis, that coincides to the animal-vegetal axis direction. polarity and directional migration (Schaefer et al. 2012). Most of the in silico predicted functions of RPHM21 such as reorganization of cytoskeleton, cell migration, cell cycle control, chromatin remodeling, and transcriptional control are shared with MK3 (Ronkina et al. 2008;Milani, Ghiselli, Guerra, et al. 2013). Moreover, the two proteins show structure similarities ( fig. 6; supplementary figs. S1 and S2, Supplementary Material online): as predicted for RPHM21, MIRs present two putative transmembrane domains and N-and C-terminus both arranged on the cytosolic side of the membrane they are associated with (Sanchez et al. 2002;Coscoy and Ganem 2003). Nonetheless, the fact that RPHM21 is localized in both mitochondria and nucleus of spermatozoa agrees with the wide distribution that characterizes many of the ubiquitination proteins cited, as well as MIRs (Kurz et al. 2002;Ronkina et al. 2007Ronkina et al. , 2008. All that considered, RPHM21 is similar to ubiquitinationrelated proteins that participate in cytoskeletal organization, transcription, cell fate determination, differentiation, proliferation and migration, and to viral proteins involved in the immune recognition pathway acting as ubiquitin ligases.

Role of RPHM21 in Mitochondrial Inheritance and Sex Determination
As shown in previous studies (Cao et al. 2004;Obata and Komaru 2005;Cogswell et al. 2006;Milani et al. 2011Milani et al. , 2012, during the development of DUI male embryos, spermatozoon mitochondria show a peculiar distribution, the aggregated pattern, which is absent in female embryos and in species lacking DUI. This pattern is most likely the result of specific interactions between sperm mitochondria and cytoskeleton by which they are actively transferred to the PGCs, where they become dominant during male germ line formation. It was proposed that the midbody derived from the central spindle of the first embryonic division is involved in positioning the aggregate of spermatozoon mitochondria in the middle of the first cleavage furrow together with germ line determinants . Given RPHM21 predicted functions and its localization in sperm mitochondria and around the animal-vegetal axis of embryos (figs. 3 and 4), and given the capability of viral proteins to interfere with mitochondrial positioning altering cytoskeleton conformation (Doorbar et al. 1991;Galluzzi et al. 2008;Ohta and Nishiyama 2011), we suggest an involvement of this novel mitochondrial protein in the interactions with specific motor proteins that would carry sperm mitochondria to the central spindle, where the cleavage furrow of the two-cell stage forms. We propose that RPHM21 prevents the recognition of male mitochondria by the degradation machinery, allowing their survival in the zygote. The process might involve a mechanism similar to that of MIRs, including the detach of specific tags associated with spermatozoon mitochondria, that normally lead to their degradation by the ubiquitination machinery. Moreover, RPHM21 cytoskeleton-binding domains support a role in the distribution pattern of spermatozoon mitochondria observed in early DUI embryos. The outcome of the interactions between RPHM21 and the cytoskeleton would depend on the abundance, in the zygote, of nuclearencoded factors which would determine the fate of sperm mitochondria (Milani, Ghiselli, Nuzhdin, et al. 2013).
The tight association between sex and type of mitochondrial inheritance in DUI organisms raised the question about the causality of this relationship: Are M-type mitochondria responsible for inducing spermatogenesis (therefore inducing maleness)? There is no clear-cut evidence that can answer this question, actually different studies reached opposite conclusions. On the one hand, the presence of DUI (and specifically the presence of a functional genome-specific mitochondrial ORFan) appears to be strictly related to gonochorism in freshwater mussels (Bivalvia Unionidae). In fact, Breton et al. (2011a) found DUI in all the analyzed gonochoric species, while it was missing in all the hermaphrodites, in which the ORFs showed heterogeneous macromutations. On the contrary, Kenchington et al. (2009) proposed that M-type mtDNA inheritance and maleness are not causally linked in marine mussels (Bivalvia Mytilidae). The matter still remains unsolved, but if we want to confirm or rule out the possibility that M-type mitochondria have a role in gonad differentiation, we need to keep investigating the functional and structural differences between M and F-type mtDNAs, among which their specific ORFans. Several DUI models based on breeding experiments in Mytilus (Diz et al. 2013;Zouros 2013) proposed that DUI phenotypes in the progeny (i.e., sex, distribution patterns of sperm mitochondria and presence/absence of M-type mtDNA) are controlled by the maternal genotype. A reanalysis of sex-determination models in Mytilus (Yusa et al. 2013) confirmed the maternal effects hypothesis, but also predicted the existence of secondary factors inherited from both parents: such minor sex-determining factors could be nuclear or mitochondrial. Accordingly, transcriptomic differences found in gonads of R. philippinarum males belonging to families with opposite sex-bias lead to the hypothesis that also some factors carried by the spermatozoon could be involved in the determination of DUI phenotypes . Some of these factors could be nuclear-encoded (Milani, Ghiselli, Nuzhdin, et al. 2013), but the best candidate for a possible mitochondrial factor is RPHM21. Supposing that it was a masculinizing factor, sperm from different males might carry different amount of transcript and/or protein, determining the different quantity of protein in the embryo thus shifting its development toward maleness.

Viral Protein Co-Option
The inferred functional features of RPHM21 and its increasing amount in R. philippinarum early embryos ( fig. 4) are consistent with a possible role for this protein during embryo development. Indeed, RPHM21 showed similarities with proteins involved in transcription activation and chromatin modification related to cell cycle, cytoskeleton and centrosome organization, spindle orientation, cell proliferation, migration and differentiation, pointing also to a role in intercellular signaling during development (Milani, Ghiselli, Guerra, et al. 2013). Also, a viral origin of this paternally transmitted mitochondrial gene could support the hypothesis of a developmental role: some integrated viral elements are involved in the early development of their host, as an endogenous retrovirus recently reported to be the earliest expressed gene detected following fertilization in mouse (Kigami et al. 2003;Kageyama et al. 2004). Another example of retroviral gene co-opted by the host with a role in early development is syncytin, a retroviral envelope proteins present in mammals (Mi et al. 2000). This viral gene is indispensable for maternal-fetal exchange, tissue remodeling during placental development and protecting the developing fetus from the maternal immune response (Cross et al. 1996;Lala and Hamilton 1996;Munn et al. 1998). Other retroviral-related genes are expressed in germ cells: some proteins of viral origin expressed predominantly in the testis, and localized in discrete dots in the nucleoplasm, seem to have a role during sperm differentiation (Inoue et al. 2000), and an infection by a vertically transmissible virus is the cause of masculinization (conversion of secondary sexual characters) of female isopods (Oniscoidea, Crustacea) (Juchault et al. 1991). Several hits regarding M-type mtDNA-specific ORFs point to a role in spermatogenesis and reproduction (Milani, Ghiselli, Guerra, et al. 2013): these hits might indicate a cooption of the novel proteins in sex-specific functions. In this concern, it was proposed that marine viruses manipulate life cycle and reproduction of their metazoan hosts, as, for example, in the sea slug Elysia chlorotica (Gastropoda, Plakobranchidae), in which endogenous and vertically transmitted retroviruses were suggested to be partly responsible for the slug reproductive cycle (Pierce et al. 1999).

Conclusions
It is quite clear that in DUI organisms sperm mitochondria evolved the ability to escape the degradation mechanism of the zygote and invade the germ line. Although such ability is well documented, whether M-type mitochondria play a role in gonad development and/or spermatogenesis is still unknown. The mitochondrial ORFans could be responsible for or participate in the DUI mechanism and a viral origin could well explain the acquired capability of M-type mitochondria to avoid the degradation, enter the germ line and proliferate, that is what viruses do best: to elude host immune system and proliferate. However, this does not explain the association with male sex. A mitochondrion that acquires the ability to escape degradation would be strongly advantaged and would be preferentially transmitted to the progeny, through both males and females, thus quickly increasing its genotype frequency in the population and eventually reaching fixation. In this case, we would not observe DUI. One possible explanation is that M-type mtDNA germ line invasion can succeed only in FIG. 6.-A proposed localization for RPHM21. The N-terminus is localized in the cytoplasm, followed by the two TMHs (in blue) inside FIG. 6.-Continued the mitochondrial outer membrane, and by the long C-terminus that is also localized in the cytoplasmic side of the membrane. The synthesized antigenic peptides used to generate antibodies are in orange. males, maybe because of epistatic interactions between the factors causing DUI and maleness factors (e.g., genes with male-biased expression). Another explanation is that the endogenization of the viral protein conferred the mitochondrion also a sex-distortion capability. In this case, a nuclear restorer gene would have evolved to avoid a shift of the sex-ratio toward males (see Passamonti and Ghiselli 2009 for a detailed discussion).
The most studied DUI taxa (Mytilidae, Unionidae, and Veneridae) are evolutionary distant, so it is also plausible that DUI mechanism of action and its interactions (if any) with sex-determining genes are different (especially if DUI evolved multiple times; Milani, Ghiselli, Guerra, et al. 2013). Most likely, several maternal and nuclear factors contribute to sex determination forming a complex network, and DUI factors could intervene at different levels of the sex determination cascade, maybe leading to different effects in different taxa.