Identification of the meiotic toolkit in diatoms and exploration of meiosis-specific SPO11 and RAD51 homologs in the sexual species Pseudo-nitzschia multistriata and Seminavis robusta

Genes exclusive to meiosis

Compared to other studies, our list of meiosis-specific genes (Tables 1 and 2) contained three more ZIP (Zipping up meiotic chromosomes protein) genes, ZIP2, ZIP3 and ZIP4 [64], all reported to be required for the synaptonemal complex in budding yeast [47, 64, 65]. Moreover, we list as meiosis-specific both SPO11-1 and SPO11-2, therefore the total number of meiosis-specific genes increases from 11 [9] to 15 (Table 1). Of these 15 genes, five were detected in all the diatom genomes surveyed (Table 2). In the Seminavis robusta dataset, these five genes were all clearly upregulated during mating (Fig. 4), as was the one tested in Pseudo-nitzschia multistriata (SPO11-2, Fig. 5). The role of sex in the ecology and biogeochemical function of natural diatom populations has been challenging to investigate with classical methods as diatom sexual stages are difficult to recognize and preserve poorly [15]. The genes identified in the present study might prove useful as molecular markers to detect events of diatom sexual reproduction in nature.

Ten meiosis-specific genes could not be identified in any of the diatom genomes. The absence of some of these genes has also been reported for other species (Additional file 5: Table S1) and has been correlated with the presence of alternative structures and pathways required during meiotic recombination [66–69].

While duplications of some of the meiotic toolkit genes have been observed in protists and other organisms (Additional file 5: Table S1), the diatom genes were all present in single copies, except for the notable expansion observed in the RAD51 family.

Does Rad21 function as a component of the cohesin complex both during mitosis and meiosis in diatoms and other Stramenopiles?

Although the SMC1, SMC2, SMC3, SMC4, SMC5 and SMC6 genes, whose products are functional components of the cohesin and condensin complexes (required for chromatin organization during cell division), were identified in diatoms, REC8, an important component of cohesin complex and meiosis-specific homolog of RAD21 [50, 70], seemed to be absent in the diatoms investigated (Additional file 5: Table S1). Gene expression analysis of the RAD21 homolog during sexual reproduction in P. multistriata and S. robusta supports the hypothesis that REC8 might be substituted by RAD21, which would function as a component of the cohesin complex both during mitosis and meiosis (Figs. 4 and 5). Indeed, during mammalian meiosis, the mitotic RAD21 cohesin has been shown to perform the role of REC8 [30, 71]. On the contrary, in the ciliate Tetrahymena thermophila that is lacking RAD21, it has been shown that REC8, the meiotic homolog of RAD21, replaces the function of RAD21 during mitosis [72]. The loss of REC8 appears to be a character shared by all Stramenopiles (Additional file 5: Table S1).

SPO11-2 is the meiosis-specific SPO11 paralog

SPO11, encoding a conserved protein involved in DNA DSBs formation and thus in initiation of homologous recombination, was found to have two paralogs in diatoms, SPO11-2 and SPO11-3/TOP VIA (Fig. 2). In many plants, including Arabidopsis thaliana, three paralogs of SPO11 have been reported, of which SPO11-1 and SPO11-2 are meiosis-specific, while SPO11-3/TOP VIA has a topoisomerase function as it interacts with topoisomerase VIB (Top VIB) and is required during vegetative growth [27]. However, most animals, insects and yeasts possess the meiosis-specific SPO11-1 homolog [73] (Additional file 5: Table S1) and lack the other counterpart of topoisomerase assembly, the TOP VIB homolog [73]. In diatoms, red algae and prasinophytes, the SPO11-1 gene seems to be lost, although the SPO11-2 and SPO11-3/TOP VIA homologs are conserved [74, 75]. Although functional differentiation between SPO11-1 and SPO11-2 is not fully resolved in plants, in A. thaliana SPO11-2 mutants the male and female meiosis is severely disrupted, while the mutation does not affect vegetative growth [27]. This suggests that SPO11-2 is involved in meiotic recombination. Similar to many plants, it had been hypothesized that the SPO11-2 homolog in diatoms is involved in meiosis, while the product of SPO11-3/TOP VIA homolog may interact with the Top VIB subunit and might be involved in vegetative growth [27, 73, 74, 76]. The present study demonstrated that SPO11-2 mRNA levels were significantly upregulated during sexual reproduction in P. multistriata and S. robusta, providing the first experimental evidence for the functional distinction between SPO11 paralogs in diatoms. Gene expression studies in the centric diatom Thalassiosira weissflogii also revealed the meiosis-specific role of SPO11-2 (Additional file 6). SPO11-1 is absent in all members of the Stramenopiles analyzed to date, and might have been lost early in the divergence of the alveolate and stramenopile lineages.

Homologous recombination and strand exchange in the absence of DMC1

Archaeal Rad-A homologs Rad51 and Dmc1 work collectively in homology search and strand exchange processes during meiotic recombination, although it is unclear how they cooperate [54, 77]. However, in mitotic cells only Rad51 is induced and carries out recombination [77]. Tsubouchi et al. [78] have proposed two different pathways of homology searching during meiosis. In the first pathway, Dmc1 and Rad51 act together with Hop2 and other accessory proteins to accomplish efficient homology searching. In budding yeast, mutation in the HOP2 gene results in inappropriate homology searching, leading to extensive synaptonemal complex formation between non-homologous chromatids [41, 79]. In this Dmc1-dependent pathway, Hop2 interacts with Mnd1 downstream of Dmc1 and Rad51 homology searching, with Hop2 being a major DNA binding protein and Mnd1 the foremost protein interacting with Rad51 [42, 80]. In the second pathway, only Rad51 is involved in homology searching. The same study [78] also reported that overexpression of RAD51 suppresses defects in DMC1 mutants, indicating Rad51 can carry out effective homology searching independently. Crismani et al. [81] recently reported that in Arabidopsis Rad51 can work together with Mcm8 to repair meiotic double strand breaks when the Dmc1-dependent major repair pathway fails. The Mcm family of DNA helicases consists of nine homologs of which six (Mcm2-7) are conserved across the eukaryotic kingdom and function as heterohexamer helicase in DNA replication [28], whereas the other three (Mcm8-10) are less conserved, with Mcm8-Mcm9 being proved to work during meiotic recombination [82, 83]. Mcm8 and Mcm9 form a stable complex and promote recruitment of Rad51 to the DNA damage sites [83, 84]. Although diatoms lack DMC1 and HOP2 genes, they possess five to six homologs of RAD51 (either one or two homologs of RAD51-A, and one homolog each of RAD51-B, RAD51-C, XRCC3 and REC-A) and MCM8 and MCM9. In A. thaliana, RAD51-C and XRCC3 have been shown to be involved in meiotic recombination [85, 86]. In P. multistriata, we observed that RAD51-A1 and RAD51-C were upregulated during sexual reproduction, and in S. robusta RAD51-A, RAD51-B and RAD51-C appeared to be upregulated during sexual reproduction. Based on the RNA-seq data produced for S. robusta, MCM8 and MCM9 also appeared upregulated during meiosis (Fig. 4). Intriguingly, a homolog of the HOP2 gene, whose protein product forms a heterodimer with Mnd1, could not be found, while MND1 was found in diatoms. Either Hop2 is highly diverged in diatoms and is currently beyond the detection by homology searches or another mechanism involving Mnd1 exists, as the S. robusta homolog is indeed up-regulated during meiosis. Therefore, diatoms may have evolved an alternative meiotic double strand break repair pathway that does not involve Dmc1. Absence of DMC1, HOP2 and MND1 homologs has been reported for certain sexually reproducing organisms such as Caenorhabditis elegans, Drosophila melanogaster and Neurospora crassa (Additional file 5: Table S1) and absence of HOP2 has been reported in Gallus gallus, Phytophthora species, and some fungi [6]. As DMC1 and HOP2 orthologs were identified in the Ectocarpus siliculosus genome but not in sexual diatoms, the losses of a DMC1 dependent DSB repair pathway and of HOP2 might be specific to certain lineages of Stramenopiles and do not correlate with obligate asexuality.

Does cross-over occur without canonical synaptonemal complex (SC) in diatoms?

The synaptonemal complex (SC), a proteinaceous structure, is developed during early prophase I of meiosis and is thought to juxtapose homologous chromatids to enhance crossing over during meiotic recombination [45]. Almost all animals, plants and fungi capable of meiosis possess the SC [87]. The ZMM (Zip, Msh, Mer) group of proteins includes seven functionally similar yet structurally diverse proteins that coordinate recombination events and SC formation during meiosis [45]. Functionally, ZMM proteins can be categorized into three subgroups. Subgroup I includes Mer3 and Msh4-Msh5, subgroup II includes Zip2, Zip3 and Zip4 while subgroup III includes the Zip1 protein. The Zip1 protein produces a stable connection between two homologous chromosomes [47, 88], Zip2, Zip3 and Zip4 facilitate protein-protein interactions [64, 65, 89] and Mer3, Msh4 and Msh5 promote DNA recombination [45, 90]. However, homologs of Zip1 among fungi (Zip1, [47]), animals (Sycp1, [91]) and plants (Zyp1, [88, 92]) are highly divergent. Similarly, the other Zip proteins have functional analogues in fungi (Zip2, Zip3 and Zip4 [64, 65]), animals (Zhp-3/Hei10 [93], Zip4H/Tex11 [94]) and plants (Zip3/Hei10 [95], Zip4 [96]) albeit with very low or no sequence similarity among different taxa. The diatom genomes contained neither identifiable homologs of ZIP genes (Table 2), nor of HOP1 and RED1, the components of lateral elements in SCs (Table 2 and Additional file 5: Table S1) and thus, diatoms may lack canonical SCs. Such a possibility was proposed for ciliates. In ciliate genomes, none of the genes related to canonical SCs were detected [9] and the corresponding absence of a canonical SC in the ciliate Tetrahymena thermophila was also supported by microscopic observations [67, 97]. The presence of a rigid siliceous frustule and dense ring of chromatin around the central spindle at metaphase make it difficult to perform routine cytological studies in diatoms [16, 98], however, in some species SC-like structures have been reported [99, 100], so SC-like structures might involve unidentified proteins that have replaced the Zip and Hop1 protein functions. Alternatively, the homologous proteins are present in diatoms but have diverged so much that they are unrecognizable.

DNA Mismatch repair genes (MSH and MLH gene family)

The MSH gene family comprises homologs of bacterial MutS genes that are important for DNA mismatch recognition and repair (MMR). MSH genes have been reported in all eukaryotes and are fundamentally involved in the initial recognition of nucleotide mismatch during repair [101, 102]. Although seven MSH homologs (MSH1-7) have been identified among eukaryotes, MSH1 and MSH7 are less conserved [102]. Msh proteins form heterodimers, Msh2-Msh6 is principally involved in MMR during mitosis whereas Msh4-Msh5 functions during meiosis [32, 44], stabilizing single strand invasion intermediates formed during early stages of meiotic recombination [90, 103]. Further, it directs Holliday Junction resolution towards crossover formation following an interference sensitive pathway [103, 104]. T-DNA insertional mutation of MSH4 of Arabidopsis exhibits reduced fertility with no effects on normal vegetative growth [103]. MSH2, MSH4, MSH5 and MSH6 were identified in all diatom genomes investigated and they were upregulated during meiosis in S. robusta indicating the presence of a complete and active MMR machinery.

Prokaryotic MutL homologs (Mlh) of DNA MMR proteins are another important group of conserved meiotic genes that work in coordination with Msh homologs. Multiple copies of MLH homologs (MLH1-3 and PMS1-2) are present in eukaryotes [105]. Mlh1 and Pms1/2 form heterodimers and interact with Msh2-Msh4 or Msh4-Msh5 heterodimers to remove DNA mismatches during replication [101, 105, 106]. Msh heterodimers initiate DNA MMR by recognizing and binding to unpaired and impaired bases. In addition, they activate the Mlh complex endonuclease that further incises DNA mismatches [107]. Mlh1-Pms1 is the major heterodimer and in some eukaryotes the Mlh family contains multiple homologs (all of which form heterodimers with Mlh1) [107]. Diatoms contain MLH1 and PMS1 whereas MLH2 and MLH3 were not detected, suggesting that the Mlh1-Pms1 complex plays a major role in MMR in diatoms. The latter hypothesis is supported by the upregulation of both genes during mating in S. robusta. Although MLH2 was not detected in other SAR supergroup members examined, as is the case for diatoms, an MLH3 homolog was detected in E. siliculosus, suggesting that MMR varies among Stramenopiles.

The meiotic toolkit in P. tricornutum and T. pseudonana

Centric and pennate diatoms differ in many aspects of their life cycles, and meiosis also differs in many ways, with a different number of gametes produced in pennates (generally two isogamous gametes) with respect to centrics (one large sized female gamete and many small sized male gametes) [108, 109]. Nevertheless, our results suggest that the molecular machinery employed in meiotic recombination may be shared by all diatom species studied. This includes P. tricornutum and T. pseudonana for which a sexual phase has never been reported. Differences could only be found for the RAD51 family, where P. multistriata, P. multiseries and F. cylindrus appeared to have a duplicated RAD51-A gene, a canonical version of XRCC3 appeared to be present only in P. tricornutum, while T. pseudonana lacked MUS81.

This would suggest that the apparent lack of a sexual phase for P. tricornutum and T. pseudonana is unlikely due to major losses in the meiotic toolkit genes. In spite of the rapid evolution of the genomes of these two diatoms and the presence of a significant amount of transposable elements in their genomes, the meiotic genes are conserved. Since meiotic genes have been reported in the genome of asexual organisms [33, 110], it cannot be excluded that these diatom species are truly asexual, with meiosis-related genes having undergone neo-functionalization and becoming employed in non-meiotic processes such as DNA repair. Moreover, detailed analyses should include an assessment of the integrity of the meiosis-related genes identified, to rule out a recent accumulation of mutations rendering the genes non-functional (i.e., species-specific or even strain-specific loss of sex after isolation in culture). For example, the meiotic SPO11-2 homolog in the T. pseudonana genome appeared to be missing the N-terminal portion of the gene found in other diatoms (Additional file 4). The presence of meiosis related genes in the genome of T. pseudonana, even if some genes may be subject to recent loss-of-function mutations (in the CCMP1335 genome), suggests that some members of the species may have retained this capacity, as seen recently in the coccolithophore Emiliania huxleyi [36]. However, at least for P. tricornutum, which is a pennate diatom (pennate diatoms are generally heterothallic), few strains have been used in laboratories across the world and it could be that the right mating partner has never been used in crossing experiments. This species is both important for fundamental research and promising in biotechnology, and the ability to conduct laboratory breeding would greatly enhance this potential. The isolation of additional wild type P. tricornutum strains should be pursued in order to verify if sexual reproduction can be induced in the laboratory.