Silent gene clusters encode magnetic organelle biosynthesis in a non-magnetotactic phototrophic bacterium

The phototrophic species Rhodovastum atsumiense G2-11 acquired MGCs from an unknown alphaproteobacterial MTB by recent HGT

In a systematic database search for novel MGCs, we identified several orthologs of known magnetosome genes in the recently released draft genome sequence of the culturable anoxygenic phototroph Rhodovastum atsumiense G2-11 [25]. This finding was unexpected as, after isolation of G2-11 from a paddy field more than 20 years ago, no magnetosome formation has been reported [26]. Furthermore, no MTB has been identified so far among phototrophs or within the Acetobacteraceae family to which G2-11 belongs [26] (Fig. 1a).

Since the available genome version (NZ_VWPK00000000) was highly fragmented (226 contigs), and the magnetosome genes were distributed over several contigs with gaps, we first re-sequenced and assembled the complete genome using long reads generated by Illumina and Nanopore technologies. The resulting genome consists of one chromosome (6.48 Mb) and eight plasmids ranging from 10,690 bp to 220,129 bp in size (Fig. 1b, Supplementary Fig. S1). The putative magnetosome genes localize within a single region (27.5 kb) on the chromosome, compactly organized in four operon-like clusters comprising the following genes: mag123, (mms6-like1)(mmsF-like1)mamH1IEKLMOH2, (mms6-like2)(mmsF-like2), and feoAmBm-mamPAQRBST (Fig. 1c). They include all genes thought to be essential for magnetosome biosynthesis (mamIELMOQB) [27] and appear to be intact, as no obvious frameshifts or nonsense mutations could be detected. Protein alignments using BLASTP suggested that the closest orthologues of the magnetosome genes from G2-11 are found among alphaproteobacterial MTB (Supplementary Table S1). The phylogenetic analysis of concatenated amino acid sequences of the magnetosome proteins MamKMOPAQBST demonstrated that the sequences from G2-11 reliably cluster with those of a recently discovered uncultivated calcium carbonate producing MTB CCP-1 [28] (Fig. 1d). The low completeness of the CCP-1 genome does not allow us to reliably infer the relationship between G2-11 and CCP-1 using phylogenomic markers, e.g., ribosomal proteins, as we conducted for the Acetobacteraceae family (Fig. 1a). Nonetheless, CCP-1 has been shown to belong to Azospirillaceae based on the 16 S rRNA tree [28], which occupies an outgroup position related to Rhodospirillaceae and Acetobacteraceae, according to the latest Alphaproteobacteria phylogeny [29]. This incongruence in the phylogenetic positions of these two species and their magnetosome genes is best explained by the HGT of the MGCs in G2-11 from an unknown alphaproteobacterial MTB probably related to Azospirillaceae. The additional phylogenetic analyses of the two parts of the MGCs individually (MamKMO and MamPAQBST) did not support different evolutionary histories of these parts, suggesting a single transfer of these clusters from the same organism as the most likely scenario (Supplementary Fig. S2).

Although the magnetosome genes from G2-11 and CCP-1 have a close phylogenetic relationship, the comparative analysis of their MGCs revealed considerable differences in their organization. First, G2-11 lacks several accessory magnetosome genes (mamX, mamZ, and mamD), which were previously shown to be universally present in alphaproteobacterial MTB and, although being not essential, are important for proper magnetic crystal formation in MSR-1 and M. magneticum AMB-1 [30, 31]. Their absence in G2-11 could be explained by functional differences in the magnetosome biosynthesis pathways, incomplete horizontal transfer of the MGCs, or a secondary loss of these genes in G2-11. Furthermore, the MGCs of CCP-1 are interspersed by >20 genes with no homology to known magnetosome genes (Fig. 1c). In contrast, the compact MGCs in G2-11 include only a few genes that could not be associated with magnetosome biosynthesis.

Tetranucleotide usage patterns are frequently employed as a complementary tool to group organisms since they bear a reliable phylogenetic signal [32]. Likewise, deviations of tetranucleotide usage in a certain fragment from the flanking genome regions can indicate HGT [21]. Comparison of the z-normalized tetranucleotide frequencies of the MGCs (27.5 kb) with the flanking upstream (117.7 kb) and downstream (79.5 kb) fragments showed a considerably lower correlation between them (Pearson’s r = 0.88 with both flanking fragments) than between the flanking fragments themselves (Pearson’s r = 0.97, Fig. 1e). This indicates a significant difference in the tetranucleotide composition of the MGCs compared to the flanking genomic regions and supports a foreign origin of the magnetosome genes in G2-11 suggested by the phylogenetic analysis. Besides, the presence of a mobile element (transposase) and position of the MGCs directly downstream of a tRNA gene, a common hotspot for integration of genomic islands [33,34,35], suggests that the MGCs of G2-11 are indeed located on a genomic island, i.e., represent MAI, like in many other MTB [20, 21]. Unfortunately, the lack of other representatives of the genus Rhodovastum makes it impossible to infer whether the MAI was transferred directly to G2-11 or the last common ancestor of the genus. Nonetheless, its compact organization and conspicuous tetranucleotide usage suggest a relatively recent HGT event.

G2-11 does not form magnetosomes under laboratory conditions

Although magnetosome genes discovered in G2-11 comply with the minimal set required for magnetosome biomineralization in MSR-1 [36], no magnetosomes have been detected in this organism. It might have several explanations: (i) the strain might switch to the magnetotactic lifestyle only under very specific, yet not tested, conditions; (ii) it once was able to synthesize magnetosomes in its natural environment but lost this ability upon subcultivation due to mutations before its characterization; (iii) the strain might naturally not exploit magnetotaxis as its genes might be non-functional or not actively expressed. To clarify which of these explanations is most likely, we first tested whether G2-11 can form magnetosomes under different laboratory conditions. To this end, the strain was cultivated photoheterotrophically, anoxic or microoxic, in a complex medium with potassium lactate and soybean peptone, as commonly used for MSR-1 (FSM) [37], as well as in minimal media with different C-sources previously shown to support growth in G2-11 (glucose, pyruvate, L-glutamine, and ethanol) [26]. All media were supplied with 50 μM ferric citrate to provide sufficient iron for magnetite biomineralization. Since magnetosome biosynthesis is possible only under low oxygen tension, aerobic chemoheterotrophic growth of G2-11 was not tested. The best growth was observed in the complex FSM medium and a minimal medium with glucose or pyruvate, whereas L-glutamine and ethanol supported only weak growth (Supplementary Fig. S3). Irrespective of the growth stage, none of the tested cultures demonstrated magnetic response as measured by a magnetically induced differential light scattering assay (Cmag) [38]. Consistently, micrographs of cells collected from stationary phase cultures did not show any magnetosome-like particles (Supplementary Fig. S3). This confirmed that G2-11 indeed cannot biosynthesize magnetosomes, at least under the conditions available for the laboratory tests. During cultivation, we also noticed that G2-11 cells did not move at any growth stage despite the initial description of this organism as motile using a single polar flagellum [26], and containing several flagellum synthesis operons and other motility-related genes. Moreover, the cells tended to adhere to glass surfaces under all tested conditions and formed a dense clumpy biofilm immersed in a thick extracellular matrix (Supplementary Fig. S3a-ii).

Considering that G2-11 generally lacks magnetosomes and appears to have a stationary lifestyle, which is not consistent with magnetotaxis, we assessed whether the maintenance of MGCs comes at fitness costs for the organism. To this end, we deleted the entire region containing the magnetosome genes (in the following, referred to as the MAI region) using the genetic tools we established for G2-11 in this work (Supplementary Fig. S4a, see Materials and Methods for details). After PCR screening, replica plating test, and genome re-sequencing, two of G2-11 ΔMAI mutants were selected for further analysis (Supplementary Fig. S5). These mutants showed no significant differences in the growth behavior compared to the wildtype (WT) when incubated in minimal media supplied with acetate or pyruvate as a sole carbon source (Supplementary Fig. S4b). This finding suggests that the presence of the magnetosome genes neither provides benefits nor poses any substantial metabolic burden for G2-11, at least under the given experimental conditions.

RNAseq reveals poor expression levels and antisense transcription in the MGCs of G2-11

We set on to determine whether the magnetosome genes are transcribed in G2-11. To this end, we analyzed its whole transcriptome for the photoheterotrophic conditions, under which the best growth was observed, in two biological replicates. The expression levels of all the encoded genes calculated as TPM (transcripts per million) demonstrated a high correlation between the two replicates (Pearson’s r = 0.98). Most genes of the (mms6-like1)(mmsF-like1)mamH1IEKLMOH2 cluster were only poorly or not transcribed at all (Fig. 2a, Supplementary dataset). Transcription of mms6-like1, mamF-like1, mamL, mamH1, mamI, and mamK, for example, did not pass the noise background threshold (TPM ≤ 2) in both replicates and were unlikely to be expressed, whereas mamE, mamM, mamH2, feoAm, and feoBm slightly exceeded the threshold in at least one replicate and might be weakly transcribed (Fig. 2a). Although the TPM of mamO (TPM = 5.67–6.10, Supplementary dataset) exceeded the background threshold, the coverage plot reveals that the number of mapped reads sharply rises at its 3’-end, whereas the 5’-end has low read coverage (Fig. 2b). This indicates the presence of an internal transcription start site (TSS) and its associated promoter within the coding sequence of mamO instead of the full transcription of the gene. Localization of an active promoter within mamO was recently described in MSR-1, suggesting that the transcriptional organization of MGCs may be more broadly conserved across MTB than assumed previously [39].

Transcription of genes within the mag123, (mms6-like2)(mmsF-like2), and mamAPQRBST clusters significantly exceeded the threshold, with the expression levels of mag1, mamT, and mamS being above the overall median. At the same time, antisense transcription was detected in the mamAPQRBST region, with the coverage considerably exceeding the sense transcription (Fig. 2b). This antisense RNA (asRNA) likely originated from a promoter controlling the tRNA gene positioned on the negative strand downstream of mamT. Such long asRNAs have the potential to interfere with sense transcripts, thereby significantly decreasing the expression of genes encoded on the opposite strand [40].

In summary, the RNAseq data revealed extremely low or lack of transcription of several genes that are known to be essential for magnetosome biosynthesis (mamL, mamI, mamM, mamE, and mamO) [27, 41]. Additionally, the detected antisense transcription can potentially attenuate expression of the mamAPQRBST cluster that also comprises essential genes, i.e., mamQ and mamB. Although other factors, like the absence of several accessory genes mentioned above and the potential accumulation of point mutations, might also be involved, the lack or insufficient transcription of the essential magnetosome genes appears to be the primary reason for the absence of magnetosome biosynthesis in G2-11.

Magnetosome proteins from G2-11 are functional in a model magnetotactic bacterium

Although visual inspection of the G2-11 magnetosome genes did not reveal any frameshifts or other apparent mutations, accumulation of non-obvious functionally deleterious point substitutions in the essential genes could not be excluded. Therefore, we next tested whether at least some of the magnetosome genes from G2-11 still encode functional proteins that can complement isogenic mutants of the model magnetotactic bacterium MSR-1. In addition, we analyzed the intracellular localization of their products in both MSR-1 and G2-11 by fluorescent labeling.

One of the key proteins for magnetosome biosynthesis in MSR-1 is MamB, as its deletion mutant is severely impaired in magnetosome vesicle formation and is entirely devoid of magnetite crystals [42, 43]. Here, we observed that expression of MamB_[G2-11] partially restored magnetosome chain formation in MSR-1 ΔmamB (Fig. 3a, b-i, b-ii). Consistently, MamB_[G2-11] tagged with mNeonGreen (MamB_[G2-11]-mNG) was predominantly localized to magnetosome chains in MSR-1, suggesting that the magnetosome vesicle formation was likely restored to the WT levels (Fig. 3b-iii).

Another essential protein MamQ is also involved in magnetosome vesicle formation, and its deletion eliminates magnetosomes in MSR-1 and other magnetospirilla [27, 41]. Expression of MamQ_[G2-11] in MSR-1 ΔmamQ initiated the biosynthesis of very tiny and scarce magnetosomes (Fig. 3c-i, c-ii). mNG-MamQ_[G2-11] was localized in several intracellular patches, which distribution resembled that of the particles observed in the TEM micrographs (Fig. 3c-iii).

MamK is an actin-like filamentous protein, which is an essential structural component of the intracellular “magnetoskeleton” that aligns magnetosomes into linear chains [44]. Deletion of mamK in MSR-1 leads to the formation of disrupted short magnetosome chains instead of a continuous long chain typical for the WT (Fig. 3d-i). Expression of MamK_[G2-11] in MSR-1 ΔmamK resulted in the restoration of a normal magnetosome chain in most of the observed cells (Fig. 3d-ii). mNG-MamK_[G2-11] demonstrated linear signal indicating the filament formation [45] (Fig. 3d-iii). Since distinguishing ΔmamK from WT or a complemented phenotype can be difficult in shorter cells, we additionally transferred mamK_[G2-11] into the MSR-1 ΔmamKY mutant [46]. In MSR-1 ΔmamKY, both magnetosome chains and their positioning are disrupted leading to the formation of magnetosome clusters or very short linear and ring-shaped chains (Fig. 3e-i, e-ii), which represent a more unambiguous phenotype than ΔmamK. Complementation of this mutant by a functional mamK should result in restoration of long chains, which would be positioned to the outer cellular curvature instead of the geodesic line of the helical cell since mamY is absent [46]. Indeed, expression of MamK_[G2-11] in MSR-1 ΔmamKY resulted in a population that included a considerable number of cells having long (≥10 particles) magnetosome chains, that were absent from MSR-1 ΔmamKY (Fig. 3e-iii). Evaluation of >50 cells for each of two randomly selected insertion mutants MSR-1 ΔmamKY::mamK_[G2-11] revealed that the long magnetosome chains were restored in 35-40% of the population (Fig. 3e-iv). Of note, mNG-MamK_[G2-11] formed slightly shorter filaments in MSR-1 ΔmamKY than in ΔmamK, which were also characteristically displaced to the outer cell curvature due to the lack of mamY [46] (Fig. 3e-v).

MamJ attaches magnetosomes to the MamK filament in MSR-1, mediating their chain-like arrangement. Elimination of mamJ disrupts this linkage, causing magnetosomes to aggregate owing to magnetic interactions [47] (Fig. 3f-i). In MSR-1, MamJ is encoded within the mamAB operon, between mamE and mamK. Within the (mms6-like1)(mmsF-like1)mamH1IEKLMOH2 cluster of G2-11, there is an open reading frame (ORF) encoding a hypothetical protein that is located in a syntenic locus (Fig. 1c). Although the hypothetical protein from G2-11 and MamJ from MSR-1 differ considerably in length (563 vs. 426 aa), share only a low overall sequence similarity (31%), and are not identified as orthologues by reciprocal blast analyses, multiple sequence alignments revealed a few conserved amino acids at their N- and C-termini (Supplementary Fig. S6). Moreover, in both proteins, these conserved residues are separated by a large region rich in acidic residues (pI 3.3 and 3.2) suggesting that the G2-11 protein might be a distant MamJ homolog. To test if it implements the same function as MamJ, we transferred this gene to MSR-1 ΔmamJ. Interestingly, it indeed restored chain-like magnetosome arrangement, which, however, often appeared as closed rings rather than linear chains (Fig. 3f-ii). Despite this difference, it indicated the ability of the hypothetical protein (hereafter referred to as MamJ-like_[G2-11]) to attach magnetosomes to MamK, suggesting that in the native context, it can have a function identical to MamJ. Consistently, its fluorescently labeled version was often observed in ring-like structures within the cytoplasm of MSR-1 ΔmamJ, suggesting that it is indeed localized to magnetosomes (Fig. 3f-iii).

In magnetospirilla, magnetosome proteins MmsF, MamF, and MmxF share an extensive similarity. Their individual and collective elimination gradually reduces the magnetite crystal size and disrupts the chain formation in MSR-1 (Fig. 3g-i; Paulus, manuscript in preparation). The MAI of G2-11 includes two genes, whose products have high similarity to these proteins, designated here as MmsF-like1_[G2-11] and MmsF-like2_[G2-11]. Expression of each of them in the MSR-1 ΔmmsFΔmamFΔmmxF triple mutant (ΔF3) partially restored the magnetosome size and led to the formation of short magnetosome chains in MSR ΔF3::mmsF-like1_[G2-11] (Fig. 3g-ii) or clusters in MSR-1 ΔF3::mmsF-like2_[G2-11] (Fig. 3g-iii, iv). Consistently, fluorescently tagged mNG-MmsF-like1_[G2-11] and mNG-MmsF-like2_[G2-11] localized to magnetosomes in the pattern resembling that in the TEM micrographs of the complemented corresponding mutants (Fig. 3g-v, vii), or were perfectly targeted to the magnetosome chains in MSR-1 WT (Fig. 3g-vi, viii).

In G2-11, MamB_[G2-11]-mNG, mNG-MamQ_[G2-11], MamJ-like_[G2-11]-GFP, mNG-MmsF-like1_[G2-11], and mNG-MmsF-like2_[G2-11] were patchy-like or evenly distributed in the inner and intracellular membranes (Supplementary Fig. S7). No linear structures that would indicate the formation of aligned magnetosome vesicles were observed in these mutants. As expected, mNG-MamK_[G2-11] formed filaments in G2-11 (Supplementary Fig. S7c).

Expression of MamM, MamO, MamE, and MamL failed to complement the corresponding deletion mutants of MSR-1 (not shown). Although detrimental mutations in the genes cannot be excluded, this result can be attributed to the lack of their native, cognate interaction partners, likely due to the large phylogenetic distances between the respective orthologues.

Transfer of MGCs from MSR-1 endows G2-11 with magnetosome biosynthesis that is rapidly lost upon subcultivation

Having demonstrated the functionality of several G2-11 magnetosome genes in the MSR-1 background, we wondered whether, conversely, the G2-11 background is permissive for magnetosome biosynthesis. To this end, we transferred the well-studied MGCs from MSR-1 into G2-11, thereby mimicking an HGT event under laboratory conditions. The magnetosome genes from MSR-1 were previously cloned on a single vector pTpsMAG1 to enable the one-step transfer and random insertion into the genomes of foreign organisms [23]. Three G2-11 mutants with different positions of the integrated magnetosome cassette were incubated under anoxic phototrophic conditions with iron concentrations (50 μM) sufficient for biomineralization in the donor organism MSR-1. The obtained transgenic strains indeed demonstrated a detectable magnetic response (Cmag = 0.38 ± 0.11) [38], and TEM confirmed the presence of numerous electron-dense particles within the cells (Fig. 4), which, however, were significantly smaller than magnetosome crystals of MSR-1 (ranging 18.5 ± 4.3 nm to 19.9 ± 5.0 nm in three G2-11 MAG insertion mutants vs 35.4 ± 11.5 nm in MSR-1 WT, Fig. 4b) and formed only short chains or were scattered throughout the cells (Fig. 4a, c-i). Mapping of the particle elemental compositions with energy-dispersive X-ray spectroscopy (EDS) in STEM mode revealed iron- and oxygen-dominated compositions, suggesting they were iron oxides. High-resolution TEM (HRTEM) images and their FFT (Fast Fourier Transform) patterns were consistent with the structure of magnetite (Fig. 4c). Thus, G2-11 was capable of genuine magnetosome formation after acquisition of the MGCs from MSR-1.

In at least three independent transfer experiments, we noticed that the ability to synthesize heterologous magnetosomes was highly unstable in G2-11 upon subcultivation. The Cmag of the transgenic cultures started to decline soon after the transfer, and the magnetic response became eventually undetectable in all of them after 10–15 daily culture passages. Concurrently, the mutant cells in this non-magnetic state were devoid of magnetosomes. To understand the mechanism of the trait loss, we sequenced the genomes of three randomly selected newly magnetized mutants (hereafter, C1-3) immediately after the genetic transfer, and again after the magnetic response had been lost from the cultures. All three mutants demonstrated a rapid decline in Cmag after the 8th passage (Fig. 5a), whereas the rate at which the cultures transitioned to a completely non-magnetic state (nonMAG) varied among the clones. As expected, TEM observations confirmed the loss of magnetosomes (Fig. 5b). Genome analysis showed that in two out of three insertion mutants (C2 and C3), the entire integrated magnetosome cassette was deleted in their nonMAG descendants (Fig. 5c). Visual inspection of the reads mapped to the insertion locus and the sequences flanking the integrated cassette revealed that a large fraction of the reads (87.4% and 96.9% in C2 and C3, respectively) was mapped to a restored wildtype sequence (except leaving a single nucleotide insertion in place of the deleted cassette), indicating a complete excision of the integrated cassette in most cells (Fig. 5d). Since on pTpsMAG1 the magnetosome cassette is flanked by inverted repeats recognized by the mariner transposase for mobilization and insertion, we believe that these repeats could be recognized and re-used for the excision of the cassette in G2-11, mediated either by intrinsic recombinases or one of many transposases encoded in its genome.

In contrast to C2 and C3, no mutations could be detected in the nonMAG state of C1. This suggests that, in addition to the cassette deletion, other mechanisms to suppress the expression of foreign magnetosome genes, e.g., transcriptional silencing, are likely involved. Besides, the native MGCs present in G2-11 were not affected in either of the mutants. Overall, this experiment demonstrated that although G2-11 can synthesize magnetosomes upon acquisition of the foreign magnetosome genes, their expression imposes a significant negative selection pressure on G2-11, causing the gene deletion (C2 and C3) or potential suppression of expression (in C1).

Source: Ecology - nature.com