Experimental analysis of diverse actin-like proteins from various magnetotactic bacteria by functional expression in Magnetospirillum gryphiswaldense

ABSTRACT Magnetotactic bacteria (MTB) produce magnetosomes, which are sensory organelles consisting of nanocrystals of a magnetic iron mineral enclosed by membranes. In the well-characterized Magnetospirillum species of the Alphaproteobacteria, magnetosomes align and form highly ordered chains along filaments that consist of the bacterial actin homolog MamK. The MamK protein is part of a multi-component “magnetoskeleton” that controls the concatenation, positioning, and partitioning of the magnetosome chains (MCs) which serve as cellular compass for efficient navigation in the Earth’s magnetic field. MamK is highly conserved in all MTB; however, it is unknown whether its magnetoskeletal function is preserved, especially in those MTB which exhibit distinct and more complex architectures of MCs and often contain additional putative magnetoskeletal constituents such as the actin-like protein Mad28 with as yet-unknown functions. Here, we studied the ability of magnetosome-associated actins from a wide range of diverse MTB to rescue well-characterized magnetoskeleton mutants of the model Magnetospirillum gryphiswaldense. We found that MamK orthologs from Alpha-, Delta-, Candidatus Etaproteo-, and Nitrospirota-MTB as well as a resurrected MamK LUCA version restored MC assembly to varying degrees and exhibited filamentous localization in M. gryphiswaldense and E. coli. We also identified a novel magnetosome-related protein from the magnetotactic alphaproteobacterium Magnetovibrio blakemorei that substitutes the function of the well-characterized MamJ protein as a molecular adaptor tethering magnetosomes to MamK filaments. Moreover, we demonstrate that Mad28 orthologs from Thermodesulfobacteriota and Nitrospirota are actin-like proteins that can functionally complement mamK mutants of M. gryphiswaldense and which form filamentous structures in vivo and in vitro. IMPORTANCE To efficiently navigate within the geomagnetic field, magnetotactic bacteria (MTB) align their magnetosome organelles into chains, which are organized by the actin-like MamK protein. Although MamK is the most highly conserved magnetosome protein common to all MTB, its analysis has been confined to a small subgroup owing to the inaccessibility of most MTB. Our study takes advantage of a genetically tractable host where expression of diverse MamK orthologs together with a resurrected MamK LUCA and uncharacterized actin-like Mad28 proteins from deep-branching MTB resulted in gradual restoration of magnetosome chains in various mutants. Our results further indicate the existence of species-specific MamK interactors and shed light on the evolutionary relationships of one of the key proteins associated with bacterial magnetotaxis.

from Nitrospirota [former Nitrospirae (37)] and other groups (34,36,38), indicating that mamK is the single most conserved magnetosome gene across MTB (15,31,34,35,(39)(40)(41)(42).Despite this high conservation, it is not yet clear whether the cytoskeletal function of MamK is preserved between various MTB or has diverged over time.For example, some MTB such as AMB-1, Magnetovibrio blakemorei MV-1, and the gammaproteobacteria strain SS-5 and BW-2 have been found to contain two divergent copies of mamK in their magnetosome biosynthesis gene clusters (20,38).In addition, MamJ, which is one of the main MamK interactors in magnetospirilla, seems to be missing outside this group suggesting that MamK dynamics and MC assembly might be controlled differently.A somewhat distinct or more complex mechanism of MC assembly is also suggested by the observation that the number and subcellular arrangement of MCs increasingly diverges in MTB with phylogenetic distance to magnetospirilla.For example, double or multiple MCs are observed in many magnetococci (43,44) and in MTB from the Nitrospirota (40).However, the function of MamK orthologs from these and most other MTB has not yet been experimentally tested due to their inaccessibility.Moreover, MTB belonging to Thermodesulfobacteriota, Nitrospirota, and Candidatus Omnitrophota contain a further putative actin-like protein in addition to MamK, which was termed Mad28 (magneto some associated deep branched).It has been reported that Mad28 in the Nitrospirota Mbav contains a conserved actin-like domain and that Mad28 from the Thermodesulfo bacteriota and Nitrospirota form a clade distinct from other bacterial actin-like proteins (4,38).The mad28 gene is located in close proximity to magnetosome-related genes (34,38) and can be present in multiple divergent copies, which again may be associated with multiplicity of MCs in these MTB.The deltaproteobacterium Desulfamplus magneto vallimortis BW-1, on the other hand, contains only one copy of mad28 besides mamK and forms only a single magnetosome chain of magnetite or greigite crystals (38).Despite its ubiquitous occurrence in Thermodesulfobacteriotaand Nitrospirota-MTB, Mad28 has not been experimentally analyzed so far, again due to a lack of genetically accessible or cultured models from those groups.As a result, its function, polymeric structures, and any potential interaction with MamK have remained elusive.
A powerful method for elucidating the function of foreign proteins from inaccessible donors is to express them in a surrogate host.This approach was recently demonstrated for several conserved magnetosome genes from selected MTB donors, which rescued their orthologous mutants in the tractable model Mgryph (45).Here, we took advantage of the model Mgryph as a heterologous host and its well-characterized ∆mamK and ∆mamKY mutants, where we expressed mamK and mad28 orthologs covering a wide phylogenetic range of MTB and donor strains that exhibit various shapes and arrange ments of crystals and comprising the following representatives: the closely related strain AMB-1 unlike Mgryph forms fragmented MCs (15).The vibriod marine MV-1 from the Alphaproteobacteria produces a single linear MC of about 10 elongated magnetite crystals (46,47).The marine Magnetococcus marinus MC-1 produces a single chain of 10-14 elongated magnetite crystals (47).Solidesulfovibrio magneticus RS-1 is a freshwa ter sulfate-reducing deltaproteobacterium that assembles 12-15 irregular bullet-shaped magnetite crystals in a poorly ordered chain (48,49).Another sulfate-reducing deltap roteobacterium Desulfamplus magnetovallimortis BW-1, however, produces magnetite and/or greigite nanocrystals, aligned in one or more MCs (50).The uncultured Candidatus Magnetobacterium bavaricum Mbav and Ca.Magnetobacterium casensis MYR-1 from the Nitrospirota synthesize between 600 and 1,000 bullet-shaped magnetite crystals that are arranged in several parallel chains, which in turn form rope-like strands or bundles (41,51).Metagenomic mamK from the uncultured Magnetococcus bacterium DCbin4 belonging to Candidatus Etaproteobacteria and HCHbin1 from Nitrospirota, both with unknown morphologies and MC configurations, were also included (34).
We analyzed the localization patterns of the heterologous proteins from different MTB phyla and their potential to complement the mamK and mamKY mutants in Mgryph.We show that MamK orthologs from foreign MTB can complement deletion mutants to different degrees and display similar localization patterns, suggesting that they are functionally equivalent.These analyses were complemented by functional expression of a putative ancestral MamK version, which was resurrected from many extant MamK (Fig. S1; Table S1A).Furthermore, we identified Mg-1g50 as a novel magnetosome-related protein in MV-1 that seems to substitute for MamJ's function in the assembly of MCs.In addition, we provide experimental evidence that the Mad28 protein is a novel magneto some-related bacterial actin-like protein in MTB.

Foreign MamK orthologs restore magnetosome chain formation in ΔmamK Mgryph and ΔmamKY Mgryph magnetoskeleton mutants to different degrees
First, we studied whether the selected MamK orthologs, which share between 71% and 98% sequence similarity and 35% and 95% identity to MamK Mgryph (Table S1B; Fig. 1), can substitute the native MamK function in Mgryph.To this end, we placed these genes under control of the promoter P mamDC45 from Mgryph and transferred them into the magnetoskeleton mutants ΔmamK Mgryph and ΔmamKY Mgryph by Tn5-transposon-medi ated random genome insertion.We then analyzed the degree of complementation by scoring the magnetosome organization and positioning based on transmission electron microscopy (TEM) imaging, as well as the restoration of the cellular magnetic orientation by C mag determination.
To establish the base for our measurements properly, we first re-assessed the ΔmamK Mgryph and ΔmamKY Mgryph deletion mutants.As previously reported by (23), ΔmamK Mgryph exhibited short, off-center, fragmented, and ectopic MCs and a reduced C mag of 0.93 (WT Cmag , 1.43) (Fig. 2A; Fig. S3 and S4C).For ΔmamKY Mgryph , we observed cells with a complete loss of the MC, but magnetosomes were aggregated or localized in magnetic flux-closed rings as reported before (27) (Fig. 2B).However, we also found that few of the ΔmamKY Mgryph cells contained short (<5 particles) and long MC (>10 particles) shifted toward the negative cell curvature, rather resembling the ΔmamY Mgryph single deletion phenotype (Fig. S4D).To address methodological factors that are inherent to our experimental approach, we next re-inserted the native mamK Mgryph gene and determined the degree of complementation.Transfer of the native mamK Mgryph into ΔmamK Mgryph increased the C mag from 0.93 to 1.37 (Fig. S3), and cells predominantly showed mid-cell linear contiguous MCs (~61%) (ΔmamK Mgryph , ~24%) (Fig. 2C; Fig. S4AC), indicating overall functional complementation.Less frequently, we still detected some ectopic (at one or both cell poles) or up to four fragmented MCs (Fig. S4C) resembling the ΔmamK Mgryph mutant phenotype, which might be due to non-native MamK expression levels.Likewise, transfer of mamK Mgryph into ΔmamKY Mgryph restored the ΔmamY Mgryph phenotype, i.e., MCs were again assembled, but shifted to the negative cell curvature, and the C mag increased to 1.12 similar to that described in reference (27) (Fig. 2B; Fig. S3).The frequency of long MC with >10 particles increased from ~4% to 57% in (Fig. 2C; Fig. S4D), and the average C mag , to 0.74 (~290% of ΔmamKY Mgryph ; C mag of ΔmamKY Mgryph = 0.19) (Fig. 2D; Fig. S3).Besides long MCs, a proportion of complemented ΔmamKY Mgryph cells also contained short chains and magnetosome rings but at lower frequency as in the parent ΔmamKY Mgryph (Fig. S3B and  D).
The C mag values of ΔmamK Mgryph and ΔmamKY Mgryph strains, each complemented with the native mamK Mgryph , were averaged and set as reference (100%) to calculate the relative degree of complementation by foreign mamKs, which we transformed next.Trans-complementation of ΔmamK Mgryph with mamK from AMB-1 and DCbin4 resulted in a C mag higher or similar to ΔmamK Mgryph ::mamK Mgryph (Fig. 2D; Fig. S3), and the frequency of non-fragmented mid-cell MC was increased to ~81% and ~67%, respec tively, demonstrating a significant enhancement compared to that of ΔmamK Mgryph (Fig. 2C), indicating a substantial functional substitution of MamK Mgryph .Trans-complementa tion of ΔmamKY Mgryph with mamK AMB1 and mamK DCbin4 increased the frequency of long MCs (>10 particles) to ~76% and ~63% in the populations, respectively, showing a significant enhancement compared to that of ΔmamKY Mgryph (Fig. 2C).Upon expression of mamK AMB1 in ΔmamKY Mgryph , the C mag values were increased by 57% of the C mag relative to the ΔmamKY Mgryph strain trans-complemented with mamK Mgryph , whereas mamK DCbin4 expression in the same genetic background resulted in a 13% higher C mag (Fig. 2D), again suggesting functional equivalence to MamK Mgryph .
MamK from MV-1, MC-1, RS-1, BW-1, Mbav, HCHbin-1, and MYR-1 also increased the frequency of mid-cell non-fragmented long MCs in ΔmamK Mgryph but to a considerably lower degree (~26%-34%) (Fig. 2C).C mag of these trans-complemented strains reached only ~36%-86% of mamK Mgryph (Fig. 2D), with a comparable or even higher frequency of cells with ectopic MCs (Fig. S4C).Likewise, mamK genes in ΔmamKY Mgryph only slightly increased the frequency of long chains (>10 particles, ~5%-14%) compared to Here, the MreB family was used to root the tree, since it forms the most ancestral monophyletic group of Actin like proteins based on Ettema and colleagues (10). of Mgryph strains that were complemented with the native allele was set at 100%, and the C mag of the strains with diverse mamKs, mamK LUCA , mamKJ-like MV-1 , mad28, and parM genes was normalized to the respective value to obtain the relative C mag as indicator for functional complementation.The values of C mag were derived from four independent Tn5 insertion mutants.
To test whether actin-like proteins from non-MTB and not involved in magnetotaxis may have an effect on magnetosome localization, for example, by just restraining their subcellular position sterically, we included ParM from E. coli and transferred this gene into ΔmamK Mgryph and ΔmamKY Mgryph .Whereas transfer of parM E. coli into ΔmamK Mgryph failed for unknown reason, the transfer of parM E. coli into ΔmamKY Mgryph resulted in trans-complemented mutant strain, which, however, did not exhibit a noticeable increase in the occurrence of long mid-cell positioned MCs, and the magnetosome organization largely remained identical to that of the control strain ΔmamKY Mgryph (Fig. 2C; Fig. S4D).Moreover, the trans-complemented mutant displayed a C mag value equivalent to only 2% of that observed in the ΔmamKY Mgryph ::mamK Mgryph strain suggesting the lack of functional restoration (Fig. 2D; Fig. S3).

MamK proteins localize in a linear pattern
To analyze the intracellular localization of the various MamK orthologs, they were N-terminally fused to enhanced green fluorescent protein (EGFP) expressed from low copy number plasmids under control of the tetracycline inducible promoter P tet in E. coli and in Mgryph after chromosomal insertion via Tn5 transposition from the endogenous constitutive promoter P mamDC45 .As reported previously, the control fusion proteins EGFP-MamK Mgryph and EGFP-MamK AMB-1 localized as a continuous line from pole to pole in both E. coli and ΔmamK Mgryph (20,23) (Fig. 3I).EGFP-MamK MV-1 localized from pole to pole as linear structure in E. coli, whereas only a short filamentous signal at mid-cell was observed in ΔmamK Mgryph (Fig. 3III).Localization of EGFP-MamK from Candidatus Etaproteobacteria (MC-1 and DCbin4) in E. coli and ΔmamK Mgryph appeared as straight lines extending along the longitudinal cell axis (Fig. 3iv and v).EGFP-MamK from Thermodesulfobacteriota RS-1 and BW-1 and Nitrospirota Mbav, HCHbin-1, and MYR-1 localized as a long filament in E. coli as well (Fig. 3VI-X).These MamK fusions formed short linear strands or patches in ΔmamK Mgryph (Fig. 3VI-ix), except for EGFP-MamK MYR-1 , which localized as short filament at mid-cell (Fig. 3X).The EGFP-ParM E. coli control localized as filamentous structure in both E. coli and ΔmamK Mgryph (Fig. 3xv).

Complementation by MamK MV-1 depends on a novel MamJ-like protein
While MamK from AMB-1 and DCbin4 restored the C mag similar to MamK Mgryph , MamK from MV-1 and the more distantly related strains RS-1, BW-1, Mbav, HCHbin1, and MYR-1 increased the C mag only slightly.This might be due to reduced interaction efficiency of the divergent foreign proteins with endogenous magnetoskeletal or magnetosomal constituents in Mgryph.One of those is MamJ, which in magnetospirilla acts as a molecular adaptor that tethers magnetosomes to MamK filaments and is necessary for tuning of MamK turnover rates (27).Because of this essential function for MC forma tion, it is likely that less similar or unrelated proteins substitute MamJ in MTB outside magnetospirilla, where MamJ orthologs have not been identified.To test this hypothesis, we first searched for potential as-yet-undetected MamJ-like homologs in the genome of the magnetotactic alphaproteobacterium MV-1.As before, a BLASTP similarity search using the complete protein sequence of MamJ Mgryph as a query failed to generate any significant hit.However, full-length MamJ Mgryph contains the hypervariable non-essential central acidic repetitive (CAR) domain that comprises a direct repetition of an 88 aa motif (residues 81-168 and 169-256) (31), which is poorly conserved even among the closely related magnetospirilla.Therefore, we confined our query to the N-terminal interaction domain (residues 23-81) plus the C-terminus (residues 293-426) of MamJ Mgryph (31).This generated a hit to the gene mg-1g50 (locus tag: BEN30_00020).Similar to mamJ in the magnetospirilla, mg-1g50 is located immediately upstream of mamK-II in the mamDFHKop in MV-1 (39).Despite only low overall sequence similarity to full-length MamJ Mgryph (33%, 15% identity), Mg-1g50 shares 44% similarity (24% identity) and 43% similarity (18% identity) to the N-and C-termini of MamJ Mgryph , respectively, whereas it lacks the non-essential CAR domain and also differs considerably in size (Mg-1g50 MV-1 , 289 aa /30 kDa; MamJ Mgryph , 426 aa/44.3kDa; Fig. S5).Moreover, the three-dimensional structures predicted by Alphafold for MamJ Mgryph (Fig. S5B) and Mg-1g50 MV-1 (Fig. S5C) are distinct.A BLASTP search with full-length Mg-1g50 MV-1 against the non-redundant database failed to reveal orthologues in any magnetic or non-magnetic bacterium.

The resurrected MamK LUCA MTB restores a non-fragmented MC
All MamK orthologs from diverse MTB showed partial to high complementation efficiency in Mgryph, suggesting that the function of the protein is evolutionary preserved.Therefore, we next asked whether a resurrected last common ancestor of MamK from the Pseudomononadota (formerly Protoebacteria), Nitrospirota, Candidatus Omnitrophota, and Planctomycetota phyla might still function as the extant MamK proteins in Mgryph.We used an approach of ancestral sequence reconstruction (ASR) to infer an ancient protein sequence on the basis of few extant ones.We attempted to reconstitute the MamK LUCA sequence from diverse MTB using the maximum likelihood marginal reconstruction algorithm implemented in the Bio++ library (52,53).The generated sequence termed MamK LUCA MTB (MamK LUCA ) shares conserved residues of ATPase activity (Fig. S2), and the amino acid sequence shows 81% similarity and 51% identity to MamK from Mgryph (Table S1B).When the synthesized MamK LUCA was fused to EGFP, it localized as a filament in E. coli like other MamK orthologs.However, unlike extant MamK orthologs, EGFP-MamK LUCA localized homogenously distributed in the cytoplasma and often accumulated in patches in ΔmamK Mgryph (Fig. 3XI), indicating incomplete polymerization.Nevertheless, resurrected mamK LUCA restored 52% of the C mag of MamK Mgryph in ΔmamK Mgryph (Fig. 2D) and also restored mid-cell non-fragmen ted MC to a slightly higher frequency (~29%) than in ΔmamK Mgryph (~24%) (Fig. 2C).MamK LUCA in ΔmamKY Mgryph recovered 24% C mag of the ΔmamKY Mgryph ::mamK Mgryph C mag (Fig. 2D), and the trans-complemented mutant showed a slightly higher frequency (10%) of MC with >10 particles (ΔmamKY Mgryph , ~4%; Fig. 2C).These results suggest that MamK LUCA is functional in Mgryph, however, to a lower degree, possibly due to the lack of a hypothetical cognate ancestral MamJ-like adaptor protein.

Mad28 is an actin-like protein with a MamK-like fold, partially restores MC formation in ΔmamK Mgryph , and forms filamentous structures in vivo
We also studied the putative bacterial actin-like protein Mad28 (Fig. 1), which had been previously discovered in magnetotactic Thermodesulfobacteriota and Nitrospirota (38).To this end, we chose mad28 genes from the thermodesulfobacterium BW-1 and Nitrospirota HCHbin1 and Mbav (with 15%-17% identity and 43%-45% similarity to MamK Mgryph ).Alignment of MamK, Mad28, and MreB sequences revealed that three of the five conserved motifs involved in ATP binding (Fig. S2) are conserved in Mad28.These conserved motifs are a phosphate one binding site (xxxGxx), one adenosine-binding site (xxxxGGxx), and one connecting region 2 (Gx), indicating that Mad28 indeed might have evolved from an ATPase.Mad28 differs from MamK in the phosphate 2 binding site (MamK, xxxDxGxGxx; Mad28, xxxSxGxGxx) and the connecting region 1 (MamK, xEPx; Mad28, xExx), indicating that these motifs evolved differently and that the highly conserved aspartate (D) in phosphate 2 binding site evolved to serine (S).The threedimensional structure of Mad28 from BW-1 modeled by Alphafold can be superimposed to Mad28 from Nitrospirota (HCH1 and Mbav), indicating almost identical structures (Fig. S6A).Remarkably, it also superimposes, if also not perfectly, with the actin-fold of MamK Mgryph (Fig. S6) with a conserved core structure that contains a nucleotide-binding and hydrolysis site (Fig. 4A).

Mad28 interacts with MamK Mgryph and polymerizes in vitro in the presence of ATP-γ-S
formed superimposing filaments (Fig. S7A), suggesting that Mad28 BW-1 and MamK Mgryph may interact with each other.A direct protein interaction between MamK Mgryph and Mad28 BW-1 was probed by a bacterial adenylate cyclase two-hybrid assay (BACTH) (54).BACTH revealed a weak interaction between MamK Mgryph and Mad28 BW-1 (Fig. S7B).However, the interaction was detected in only one of the tested permutations, indicating that the interaction between these proteins is either sterically constrained or limited due to strong competition between the interaction partners (i.e., self-interaction as well as possible putative interactions with other proteins).Mad28 BW-1 showed no interac tion with MamJ and MamY from Mgryph (Fig. S7B).To consolidate the interaction of Mad28 and MamK, we performed an immunoprecipitation assay.For this, we transferred egfp-mad28 BW-1 and only the reporter gene egfp under the control of P mamDC45 into WT Mgryph .Subsequently, we extracted cell lysates from the modified strains and utilized GFP-trap agarose beads to capture both eGFP and eGFP-tagged Mad28 selectively from the cell lysates.Next, we performed Western blotting and immunodetection using antibodies against eGFP and MamK (Fig. S7Ci, Cii).Immunodetection of eGFP indica ted the expression and successful pull down of eGFP and eGFP-Mad28.Interestingly, immunodetection of MamK showed a faint band in the eGFP-Mad28 expressing strain suggesting an interaction between Mad28 and native MamK Mgryph (Fig. S7Cii, right panel).These results suggest that Mad28 may also interact with native MamK orthologs of MTB belonging to Thermodesulfobacteriota and Nitrospirota.
We next asked whether Mad28 can polymerize in vitro in the presence of ATP-γ-S, which is a characteristic of other actin-like proteins.To this end, we expressed and purified Mad28 BW-1 (44.55 kDa) from E. coli cells (Fig. 4B) and performed a polymeri zation assay, similar to that described before (55).TEM analysis of negatively stained Mad28 BW-1 that had been incubated with ATP-γ-S confirmed that Mad28 BW-1 indeed polymerized into filamentous structures (Fig. 4C), which were not detectable in the control in which ATP-γ-S was omitted (Fig. 4D).The well-developed bundles were >2 µm in length and ~29 nm in width, and the smallest bundles were ~7.3 nm in diameter.The dimensions and filamentous appearance of polymerized Mad28 BW-1 are similar to those of the polymerized MamK from Magnetospirillum spp.(length, >2 µm; width, 6-7 nm) (21,55) but distinct from the gently twisted double-helical filaments of ParM (14) and the straight or curved filamentous protofilament bundles or ring-like structures of MreB (56).

DISCUSSION
The role of actin-like proteins in the organization and dynamics of MCs in MTB is still intriguing.MamK is the most highly conserved gene associated with magnetotaxis and present in all known MTB, which suggests that MamK is one of the primordial magnetosome proteins.However, experimental studies have been confined to MamK from the magnetotactic models Mgryph and AMB-1.Therefore, it is unknown whether the numerous insights derived from these related models can be generalized.We used the partial restoration of WT-like MC configuration and, specifically, the frequency of mid-cell non-fragmented MC in the MamK Mgryph mutant as measure to estimate the functionality of introduced foreign actin-like proteins, and the results were normalized to the strains that were complemented with the native mamK Mgryph .Expression of highly similar MamK from AMB-1 and the DCbin4 resulted in a substantially increased frequency of mid-cell non-fragmented MCs, which indicates a high degree of functional ity for these proteins.In contrast, less similar MamK from MV-1, MC-1, RS-1, BW-1, Mbav, HCHbin1, and MYR-1 as well as MamK LUCA also enhanced the frequency of mid-cell non-fragmented MC but to a lesser degree, which indicates limited functionality.These results were confirmed by another complementation experiment that was based on the ∆mamKY Mgryph mutant phenotype, where the frequency of long chains (>10 particles) was used as a measure for the functionality of the bacterial actin-like proteins.Here, we could include ParM from E. coli, a remote actin-like protein, which, however, was not able to complement.The only partial functionality of the mostly distantly related MamK in Mgryph could be due to several factors.While the ATPase domain of diverse MamKs is conserved, regions that mediate interaction with other proteins may have evolved divergently, resulting in limited functionality in Mgryph.Additionally, subopti mal expression levels may also contribute to the limited functionality.Because MamK alone is insufficient to assemble MCs, a further possible explanation for only partial complementation is the absence of cognate heterologous MamK interactors (such as MamJ) in Mgryph.We identified a putative mamJ-like gene (mg-1g50) in MV-1.Strikingly, co-expression of MamK and MamJ-like from MV-1 in ∆mamK Mgryph and ∆mamKY Mgryph significantly enhanced the degree of complementation.This suggests that MamJ-like MamK interactors may well exist in MTB outside the magnetotactic spirilla but simply have been overlooked so far.This indicates that sequence divergence is a determinative factor, likely in concert with the presence of a cognate MamK interaction partner with whom the protein has coevolved.
All tested MamK orthologs fused with EGFP localized as a filament in E. coli, indicating that these proteins share a core characteristic of actin-like proteins.These results show that the linear localization is independent of MTB-endogenous interaction partners and magnetosome membrane proteins, which is consistent with previous results observed with MamK from AMB-1 (15,22).In contrast, these proteins showed different localiza tion in ∆mamK Mgryph .MamK from Alphaproteobacteria (Mgryph, AMB-1, and MV-1) and Candidatus Etaproteobacteria (Magnetococcales MC-1 and Dcbin4) localized as a long filament in ∆mamK Mgryph , whereas those from Thermodesulfobacteriota (RS-1 and BW-1) and Nitrospirota (Mbav and HCHbin1) showed only short linear filaments.This might either be a typical property of these proteins or due to instability or rapid depolyme rization of assembled filaments in Mgryph (i.e., altered subunit turnover or filament dynamics).EGFP-MamK from MYR-1, however, localized as a long filament indicat ing higher stability than other candidates from Nitrospirota.Interestingly, resurrected MamK LUCA localized as a filament in E. coli, which indicated that the protein is able to polymerize.However, the localization in ∆mamK Mgryph was diffuse suggesting that the cellular environment including MamK-depolymerizing factors may play an essential role to form stable MamK filaments.
Mad28 from the thermodesulfobacterium BW-1 and the Nitrospirota HCHbin1 and Mbav also showed partial functionality in ∆mamK Mgryph and ∆mamKY Mgryph strains, as indicated by TEM.Furthermore, these Mad28 were found to localize similar to MamK in E. coli and ∆mamK Mgryph indicating that Mad28 is capable of directly assembling and localizing in a stable polymeric form.The lower degree of MC reconstitution detected by TEM correlated with lower C mag values of complemented strains.The conserved ATPase residues and stable polymerization likely contribute to their partial restoration of MamK function in the deletion mutants, resulting in the reconstitution of magneto some chains at variable degree.Our study provides the first evidence that Mad28 from BW-1 polymerized into stable filaments in vitro.The formation of filamentous structures reflects its specific and distinct intracellular physiological function in Thermodesulfobac teriota and Nitrospirota MTB.We also found that MamK and the putative actin-like Mad28 are able to self-and to cross-interact in vivo (based on bacterial two-hybrid experiments), that Mad28 forms filaments in E. coli, which co-localize with MamK filaments, and that they interact in vitro (based on immunoprecipitation assays).In magnetotactic Thermo desulfobacteriota and Nitrospirota, both genes co-occur, sometimes in several copies.The self-and cross-interacting nature plus interaction with putative MamJ-like proteins might account for the spectacular diversity of MC configurations and, in particular, for the presence of highly ordered multiple MCs in these uncultured bacteria.The Mad28 proteins from BW-1, Mbav, and HCH1 showed structural similarities indicating they may perform similar functions in these organisms.Moreover, even though Mad28 from BW-1 and MamK from Mgryph have low sequence identity/similarity (16%/45%), homology modeling suggests that they share structural similarities.
In conclusion, we provide the first comprehensive and systematic approach to test for conserved functionality of MamK and Mad28 proteins from diverse MTB in Mgryph.The study provides significant insights into the evolutionary relationships and functional conservation of bacterial actin-like proteins associated with magnetosomes.Furthermore, this study indicates the existence of species-specific MamK-interactors and expands the understanding of the molecular mechanisms underlying magnetosome chain organization in diverse MTB.It also establishes Mad28 as a novel bacterial actin likely to play a crucial role in the chain arrangement in MTB.

Bacterial strains and culture conditions
Mgryph strains were cultivated under modified flask standard medium (FSM) (57) at 28°C and 120 rpm agitation in a micro-aerobic environment.Escherichia coli was cultured in lysogeny broth (LB) with shaking at 180 rpm at 37°C.The donor strain E. coli WM3064 (W.Metcalf, unpublished) was grown with 0.1 mM DL-a,Ɛ-diaminopimelic acid.Selection of transconjugants was carried out on agar-solidified media [1.5% (wt/vol)] by adding kanamycin at concentrations of 25 µg/mL (for E. coli) and 5 µg/mL (for Mgryph).Optical densities (OD) were measured photometrically at 565 nm for Mgryph strains and 600 nm for E. coli.The coefficient of magnetically induced differential light scattering (C mag , magnetic response) was determined as previously reported (58).The bacterial strains used in this study are listed in Table S1C.

Molecular and genetic techniques
Oligonucleotides (Table S1D) were purchased from Sigma-Aldrich (Steinheim, Germany).Plasmids were constructed using standard recombinant techniques, as described below.All constructs were sequenced by Macrogen Europe (Amsterdam, Netherlands).The plasmids generated and utilized in this study are listed in Table S1E.The DNA synthe sis of mamKs from MC-1, Dcbin4, Mbav, HCH-1, MYR-1, mamK LUCA , and mad28s from HCH-1 and Mbav was carried out by ATG:biosynthetics GmbH.Sequence-verified DNA fragments were delivered in a pGH standard vector with an ampR (bla) gene for selection on ampicillin.MamKs from MSR-1, AMB-1, MV-1, RS-1, and BW-1 and mad28 from BW-1 were amplified by PCR from the respective genomic DNA.

Phylogenetic tree reconstruction
A phylogenetic tree was built with bacterial actin-like proteins, including MamK, MreB, FtsA, and Mad28 of representative MTB.Sequences were aligned with the MAFFT software (59), then trimmed using the BMGE 1.12 (60) with low stringency parameters (-b 3 -g 0.5 options).The maximum-likelihood tree was built with the IQ-TREE v2.1.3software ( 61) and a substitution model selected using ModelFinder (62) and the BIC.A set of 500 bootstrap replicates were conducted to calculate local support values.MreB was chosen as a root based on the evolutionary history of the actin ATPase protein family reconstructed by Ettema and colleagues (10).

Construction of trans-complementation vectors
To trans-complement the ∆mamK and ∆mamKY strains of Mgryph, we cloned a fragment containing the respective foreign mamK/mad28 gene and parM from E. coli into a Tn5-based insertion vector pBAM-Tn5-P mamDC45 -egfp-HL with the constitutive promoter from Mgryph.To construct Tn5-based constructs containing mamK/mad28/parM gene without a reporter gene egfp, the plasmid pBAM-Tn5-P mamDC45 -egfp-HL was digested with Ndel and EcoRV or Ndel and BamHI and ligated with amplified genes by PCR from respective genomic material digested with same restriction enzymes.To investigate the localization of MamK/Mad28 from distantly related donor strains and ParM from E. coli, we fused the genes with egfp and cloned them into the pBAM-Tn5-P mamDC45egfp-HL.For fusion constructs, the plasmid pBAM-Tn5-P mamDC45 -egfp-HL was digested with EcoRV and dephosphorylated using FastAP thermosensitive Alkaline Phosphatase (Thermo Scientific).The linearized plasmid was ligated with the genes amplified with phosphorylated primers.For the co-expression of mamK and mv-1g50 from MV-1, these genes were placed under the control of P mamDC separated by an optimized ribosome binding site and cloned into pBAM-Tn5-P mamDC45 -egfp-HL digested with Ndel and BamHI by Gibson assembly.The resulting constructs were transferred into Mgryph via conjugation, leading to the random insertion of the expression cassette into the host chromosome while keeping the foreign gene expression levels similar to that of the recipient.

Construction of anhydrotetracycline-inducible expression vectors
For the construction of an anhydrotetracycline-inducible expression vector (pBAM-P tetegfp), the fusion constructs of mamK/mad28/parM based on pBAM-Tn5-P mamDC45 -egfp were digested with Ndel and BamHI or Ndel and XbaI and ligated into pBAM-P tet -egfp digested with same restriction enzymes.

Conjugation and screening of transconjugants
Plasmid transfer by biparental conjugation was performed with donor strain E. coli WM3064 consisting of the verified construct and Mgryph strains as the acceptor strain as described previously (63).The generated trans-conjugant strains are listed in Table S1C.The transconjugants were transferred into 96-well plates with 150 µl of FSM containing the appropriate antibiotic concentration.The mutants were screened for integration of the expression cassette by PCR using primer pairs (Table S1D).

Reconstruction of ancestral MamK
For the ASR of MamK protein, we first selected a set of 29 sequences from representative MTB strains or metagenome assembled genomes belonging to representative Pseudo monadota, Thermodesulfobacteriota, Candidatus Omnitrophota, Planctomycetota, and Nitrospirota phyla (see Fig. S1A for details).Sequences representing the major MTB phyla were selected using the public databases updated in 2019, while keeping a reasonable number of sequences to minimize the number of insertions/deletions. Indeed, actin proteins being highly divergent, their alignment generates many gaps, which is difficult to model in ASR.The multiple-sequence alignment was computed with MAFFT (59) using the default options.The gap positions of this alignment were then manually edited.The phylogenetic tree used for the ASR was computed using IQ-TREE (61).The amino acid substitution model used (LG + I + G 4 ) was selected with the BIC.The tree was rooted using the clade containing the sequences from deep branching phyla including Planctomycetota, Thermodesulfobacteriota, and Nitrospirota phyla (Fig. S1).
For the ASR, we used the marginal maximum-likelihood reconstruction algorithm implemented in the Bio++ software suite (53).Branch lengths were recomputed with Bio++ using the branch and site heterogeneous substitution model COaLA (64).This substitution model uses the amino acid exchangeabilities from the LG model (65), and equilibrium frequencies are computed for the different branches of the tree using a correspondence analysis approach.Site heterogeneity is classically modeled using a gamma correction with four classes.The sequence chosen for experimental resurrection was the one located at the root of the tree.For each position in the sequence, we used the amino acid with maximal posterior probability.

3D structure prediction and superimpose
Prediction of the 3D structure of bacterial actin-like proteins was done using the Chimerax server (https://www.rbvi.ucsf.edu/chimerax).The amino acid sequences of the protein of interest were uploaded onto the Chimerax server, which automatically generated a 3D model through its built-in homology modeling algorithm.To validate the predicted structure, we employed Pymol (version 2.5.0) to superimpose the predicted models of actin-like proteins to compare the models.

Fluorescence microscopy
To study the localization of EGFP fusion proteins, three-dimensional structured illumination fluorescence microscopy was conducted using an Eclipse Ti2-E N-SIM E fluorescence microscope (Nikon) equipped with a CFI SR Apo TIRF AC 100× H NA1.49 oil objective lens, a hardware-based "perfect focus system, " and an Orca Flash4.0LT Plus sCMOS camera (Hamamatsu).Sample preparation, fluorescence excitation with the 488-nm laser line for imaging GFP, and image reconstruction and analysis were carried out as previously described (66).

Protein expression and purification
Recombinant protein expression was carried out in E. coli BL21 (DE3) grown in LB medium supplemented with 50 µg/mL kanamycin.The E. coli cells were transformed with a plasmid (pET28a-Mad28 BW-1 ) encoding Mad28 BW-1 fused to a 6xHis tag.The plas mid (pET28a-Mad28 BW-1 ) was constructed by restriction and ligation where amplified mad28 BW-1 was cloned into pET28a digested with Ndel and BamHI.The culture was inoculated with 1/10 volume of a fully grown overnight culture and cultivated at 37°C until the OD 600 reached 0.6-0.9.The temperature was then reduced to 16°C, and after 30 min, 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) was added to induce gene expression.After 16 hours of induction, the cells were harvested by centrifugation at 10,000 rpm for 10 minutes at 4°C and resuspended in lysis buffer [25 mM HEPES (pH 7.8), 500 mM NaCl, 10 mM imidazole, 10% glycerol, 0.1% Triton-X-100].The cells were incubated on ice for 30 minutes, followed by lysing using a Microfluidizer (microfluidics) by passage at 1,500 bar at 4°C.The crude extract was then centrifuged at 10,000 rpm for 20 minutes at 4°C, and the resulting supernatant was treated as the soluble protein fraction.

Polymerization assay
The purified Mad28 BW-1 was assayed for in vitro polymerization using a protocol adapted from reference (55).A total volume of 30 µL of polymerization buffer, containing 50 mM Tris (pH 7.0), 100 mM NaCl, 14 mM MgCl2, and 30 mM KCl, was used.Purified Mad28 BW-1 protein was added at a final concentration of 10 µM, and the mixture was centrifuged at 15,000 rpm and 4°C for 30 minutes using a Centrifuge 5424R to remove aggregates.Next, a non-hydrolyzable ATP analog, adenosine-5′-(γ-thio) triphosphate (ATP-γ-S), was added at a final concentration of 2 mM, and the samples were incubated for 15 minutes at 30°C.A carbon-coated copper grid was then placed on a drop of the mixture for 6 minutes and washed with ddH 2 O.The grid was negatively stained with 2% (wt/vol) uranyl acetate for 1 minute and observed using a JEM-2100 transmission electron microscope at 80 kV.

Bacterial two-hybrid assay
To conduct protein interaction studies using the adenylate cyclase (CyaA)-based two-hybrid assay (54), we amplified mad28 BW-1 from genomic DNA of D. magnetoval limortis and cloned into pUT18C, pUT18, pKT25, and pKNT25 plasmids using primers (listed in Table S1C) that contained suitable restriction endonuclease sites for clon ing into the two-hybrid variant vectors.The resulting T18-and T25-based plasmids were confirmed by DNA sequencing and co-transformed into the E. coli BTH101 reporter strain.To assess the interaction between the proteins, cells were plated on LB agar supplemented with 40 µg/mL 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-Gal), 0.5 mM IPTG, ampicillin (100 µg/mL), and kanamycin (50 µg/mL) and incubated at 28°C.Colonies were subsequently analyzed for the blue color formation.To visualize the outcome of the experiments, several colonies per plasmid combination were grown overnight at 28°C in LB liquid medium containing IPTG (0.5 mM), ampicillin (100 µg/mL), and kanamycin (50 µg/mL), and 3 µL of culture was spotted onto M63 mineral salts agar supplemented with 0.2% (wt/vol) maltose, X-Gal (40 µg/mL), 0.5 mM IPTG, ampicillin (50 µg/mL), and kanamycin (25 µg/mL).The M63 plates were incubated at 28°C for 2 days and documented with a camera (Panasonic).The blue coloration was regarded as positive.Constructs carrying a leucine zipper fused with the T18-and T25-fragment were used as positive controls.Co-transformants harboring constructs coding for the respective T18-and T25-protein fusions in combination with the corresponding T25and T18-subunit alone served as negative controls.

Immunoprecipitation and immunoblotting
To immunoprecipitate GFP-tagged Mad28 BW-1 , GFP-Trap magnetic agarose beads from Chromotek were used following the manufacturer's protocol.The supernatant of cell lysate of Mgryph strains expressing eGFP-Mad28 BW-1 was incubated with the beads for 2 hours at 4°C with gentle shaking.After incubation, the beads were washed five times with wash buffer, and Mad28 BW-1 was eluted by heating the beads in SDS sample buffer (58 mM Tris-HCl, pH 6.8, 2% SDS, 5% glycerol, 0.1 M DTT, 0.01% bromophenol blue) at 95°C for 10 minutes.The proteins were separated by SDS-PAGE using 12% gels and Tris-glycine running buffer (50 mM Tris-HCl, pH 8.5, 0.19 M glycine, 0.1% SDS), as described in reference (69).
To transfer the proteins onto PVDF membranes, the semi-dry technique was used at 2 mA cm −2 for 2 hours, with Bjerrum-Schafer-Nielsen transfer buffer (48 mM Tris-base, 39 mM glycine, 0.0375% SDS, 20% methanol) (70).The membrane was blocked with 5% non-fat dry milk in Tris-buffered saline containing 0.1% Tween-20 for 1 hour at room temperature and incubated with the primary antibody against the protein of interest overnight at 4°C.Afterward, membranes were washed four times with Tween-Tris-buf fered saline (TTBS) buffer (0.05% Tween 20, 50 mM Tris-HCl pH 7.5, 150 mM NaCl) and incubated with HRP-labeled anti-rabbit secondary antibodies for 1 hour, followed by further washing with TTBS.Immunodetection was performed using the commercial Western Blot Chemiluminescence HRP Substrate by Takara Bio (USA).Finally, images of the gel and blot were captured with a ChemiDoc XRS+ System (Bio-Rad, USA) and processed with ImageLab 6.0.1 software.

Transmission electron microscopy
For TEM analysis, the strains were cultivated anaerobically in FSM at 24°C for 48 hours, fixed in 1.8% formaldehyde, and adsorbed onto carbon-coated copper grids (F200-CU carbon support film, 200 mesh; Electron Microscopy Sciences, Hatfield, UK).The samples were then washed three times with double-distilled water.TEM analysis was performed using a JEM-2100 instrument (JEOL, Ltd., Tokyo, Japan) at 80 kV, and images were captured with a Gatan model 782 ES500W Erlangshen CCD camera (Gatan, Inc., Pleasanton, CA) and Digital Micrograph 1.80.70 software (Gatan, Inc.).ImageJ Fiji V1.50c (71) software was used for data analysis and measurements.The statistical analysis of magnetosome organization in the mutants was performed using GraphPad Prism statistical software.S1 (mBio01649-23-s0008.docx).Protein sequences used for LUCA reconstruction, percentage identities/similarities, strains, primers, and plasmids used in this study.

FIG 1
FIG 1 Maximum-likelihood phylogenetic tree of the actin ATPase protein family showing that Mad28 proteins of magnetotactic Thermodesulfobacteriota and Nitrospirota form a distinct clade compared to other actin-related proteins like MamK, MreB, and FtsA found in magnetotactic bacteria.The tree was drawn to scale, and circles represent statistical support estimated from 500 non-parametric bootstraps.Only significant values above 80 are shown.The sequence database was inspired from that of Monteil and colleagues (4) where accession numbers of crenactin family (Cren) and eukaryotic actin 1 (ACT1) can be found.

FIG 2 (
FIG 2 (A and B) Schematic representation of magnetosome organization in WT Mgryph and the magnetoskeleton mutants ∆mamK Mgryph , ∆mamKY Mgryph , and ∆mamY Mgryph .WT, magnetosomes organized in straight chains that run along the geodetic cell axis; ∆mamK Mgryph , short, fragmented, off-center chains; ∆mamKY Mgryph , agglomerated magnetosomes or magnetic flux-closed rings; ∆mamY Mgryph , magnetosome chain at the negative inner cell curvature.Arrows indicate expected reconstitution of magnetosome chain organization in case of successful trans-complementation with mamK from diverse MTB in ∆mamK Mgryph (A) and in ∆mamKY Mgryph (B).(C) Degree of complementation as inferred from MC organization in ∆mamK Mgryph and ∆mamKY Mgryph .Blue bars, frequency of magnetosome chains positioned at mid-cell in WT Mgryph , ∆mamK Mgryph , and ∆mamK Mgryph that was complemented with the indicated genes.Green bars, frequency of magnetosome chains (>10 particles, resembling ∆mamY Mgryph phenotype) in ∆mamKY Mgryph and ∆mamKY Mgryph that was trans-complemented with the indicated genes.The numbers of analyzed cells (N) correspond to >200.All values are expressed as mean ± standard error of the mean (SEM).Statistical significance is indicated as follows: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, and non-significant (ns) when P > 0.05.Statistical significance was assessed using a non-paired t-test between the control strains (ΔmamK Mgryph in blue and ΔmamKY Mgryph in green) and the corresponding trans-complemented mutants in the control strains.(D) Relative magnetic response (C mag ) of complemented ∆mamK Mgryph and ∆mamKY Mgryph strains.The C mag