Evaluation of the large-scale deletion method
We first assessed two different techniques with respect to their usability and efficiency to introduce large genomic deletions: A Cre-lox based method, which had been used for excision of larger fragments before [28, 29, 35], and an allelic replacement method based on two consecutive double-crossovers counterselected by lethal GalK [34]. These were tested on two different regions (∆M01/~ 16 kb, and ∆M04/~ 66 kb) of the MAI (Fig. 1).
By using the Cre-lox based method, plenty of clones containing the desired ∆M01 and ∆M04 deletions could be isolated (typically around 20–30 clones with excised target regions per 96 screened clones). Using allelic replacement, between 15–30 clones with the desired double-crossover were typically obtained from 96 screened clones after the final counterselection step. As expected, we found that the use of longer homologous regions of about 1.5–2.5 kb is favorable to yield high numbers of positive clones for larger (>ca. 20 kb) deletions, whereas fragments larger than 2.5 kb were difficult to clone by overlap PCR. Excluding time for cloning, Cre-lox based deletions in our hands typically required about 6 weeks because of the need of three consecutive cycles of laborious conjugation, plate growth, clonal selection and screening, which are particularly cumbersome and time-consuming in the rather slow-growing M. gryphiswaldense. In contrast, after some streamlining of the workflow, by GalK selection and double-crossovers a clean unmarked deletion mutant was typically obtained and PCR-verified in only about 3 weeks. During this study, this method later also proved to be highly efficient for deletions of up to about 100 kb. While in most cases proper excisions by double-crossovers could be confirmed by sequencing of PCR products spanning over the excision site, occasionally we identified clones which yielded amplicons of expected size, but did not have lost their insensitivity against kanamycin, indicating the Kmr marker harbored on the suicide vector to be still residing in the genome. This issue is exemplified by a clone in which we had attempted a ~ 68 kb deletion spanning from feoAB1op to mamABop (region M08, see below and Fig. 2). Despite their kanamycin insensitivity, all cells had apparently lost the ability to form magnetosomes as expected. However, genome resequencing revealed a large part (~ 44 kb) of the deletion target to be still residing in the chromosome, and a large part (~ 11.4 of ~ 11.7 kb) of the suicide vector was inserted next to it. Conspicuously, the orientation of the homologous downstream region had become inversed, and a ~ 2 kb fragment of mamABop (comprising mamH, mamI and a part of mamE) was dislodged from its native position, while the rest of this operon (including several essential magnetosome genes) was absent (Fig. 2), thereby explaining the loss of the magnetic phenotype. Likely, deletion of ~ 47 kb exceeding targeted M08 region had occurred by homologous recombination between two nearly identical ~ 750 bp stretches of two integrase genes residing in R4 and R6, respectively.
Notably, in this clone we also found the suicide gene galK (encoding the lethal galactokinase) to be inactivated by insertion of IS elements, thereby prohibiting proper counterselection in the presence of galactose, but favoring the occurrence of spontaneous homologous and non-homologous rearrangements instead. Similarly, during the further course of our mutagenesis approach, false positive clones instead of the intended ‘clean’ deletions were frequently obtained, in particular for difficult or essential targets. Resequencing of all such suspicious clones revealed that this was always accompanied by galK inactivation due to IS insertions (Fig. 2). Nonetheless, considering the benefits of the GalK-based method, it was chosen for all subsequent deletions in this work.
Deletion and replacement of the MAI and adjacent regions
We next generated a library of strains in which we aimed to delete all key magnetosome biosynthesis genes plus as much as possible of the interspacing and flanking gene content from the ~ 100 kb MAI [3]. This region is known to be particularly rich in genetic junk and comprises 39 putative mobile genetic elements [26,27,28,29] (Fig. 1, blue arrows). We genetically dissected the MAI and its neighboring region for testing their relevance regarding survival, cell growth and magnetosome biosynthesis. By excluding genes assumed to be relevant or essential for cell growth (e.g. tRNAs and rRNAs), we predicted a region of ~ 134 kb comprising all known key magnetosome clusters and genes potentially irrelevant to the magnetosome formation (Fig. 1), including region R2 that seemed to be successfully deleted in Ullrich et al. (2010), while it appeared to be non-deletable in Lohße et al. (2011). The whole ~ 134 kb region was divided into eight separate regions (R1–8) representing putative deletion targets, which comprised known magnetosome biosynthesis operons (R1, R3, R5, R7), intervening regions (R2, R4, R6) and a flanking region adjacent to the MAI (R8). Since regions R2 and R8 are spanning large chromosomal areas containing many hypothetical genes with unknown function, they were further divided into smaller parts for deletion. In summary, all regions were covered by 17 partially overlapping deletion targets spanning from ~ 2.5 kb (feoAB1op) up to ~ 100 kb (∆M13) (Fig. 1 and Table S2).
Despite of repeated attempts, we failed to enforce proper deletions of ∆M06–M09 (Fig. 1, dashed bars), which all include the region R2, thereby supporting the assumption by Lohße et al. (2011) of a non-deletable part in this region. By deletions ∆M14 and ∆M15 this non-deletable part was narrowed down to a region of 15.2 kb including msr1_02770–msr1_03000 (Fig. 1), which in addition to several hypothetical genes encodes a putative toxin-antitoxin system (msr1_02860–msr1_02870) that might prevent its simultaneous deletion.
For all other targets, mutants could be readily generated as intended, yielding strains ∆M01–∆M05 and ∆M10–∆M17 with defined single deletions ranging from ~ 2.5 kb (feoAB1op, deleted in a later step) up to ~ 100 kb (∆M13) (Fig. 1, grey bars). The ∆A13 mutant from Lohße et al. (2011) (not to be confused with ∆M13, this study), already lacking mms6op, mamGFDCop and mamXYop (Fig. 1), was used as parental strain for the additional deletion of mamABop and feoAB1op to generate ∆M01 and ∆M03 mutants, respectively. To generate ∆M02, strain ∆A13∆mms5/mmxF lacking mms6op, mamGFDCop, mamXYop and mms5/mmxF (R. Uebe, unpublished) was used to delete the mamABop. Further deletion of regions R4 and R6 in the ∆M02 background then yielded ∆M05 (Fig. 1). All other deletions were introduced into the WT parent. ∆M01–∆M05 showed WT-like cell size, shape and morphology, but displayed slightly impaired swimming motility as their parent strains ([29], R. Uebe, unpublished).
As expected, all deletions comprising the known magnetosome clusters were impaired in magnetosome biosynthesis to different degrees. Mutants ∆M01–∆M05 and ∆M12–∆M13 lacking the mamABop were entirely devoid of magnetosomes, whereas ∆M11 (deletion of R7 with mamXYop, but all other mam/mms/feo clusters still present) essentially phenocopied the known intermediate magnetic phenotype typically caused by mutation of the mamXYop (Figs. 3 and S1) [36]. This phenotype is characterized by a reduced (40–80% of the WT) Cmag (a light-scattering based proxy for the average magnetic orientation of bacterial cells in liquid media [37]), with WT-like magnetite crystals flanked within the magnetosome chain by poorly crystalline flake-like particles. By contrast, elimination of regions outside the mam/mms/feo clusters (∆M10, ∆M14–∆M17 in R2 and R8) resulted in a WT-like magnetosome phenotype (Fig. S1). These mutants ∆M10 and ∆M14–∆M17, covering 15 putative mobile genetic elements, phage-related genes and several hypothetical genes, also displayed a WT-like cell growth at 28 °C under aerobic conditions (data not shown).
However, all non-magnetic mutant strains in which deletions covered the mamABop (∆M01–∆M04) displayed a growth advantage over the WT by reaching higher cell densities (ca. 10–35%) under aerobic conditions or moderate heat stress at 33 °C (Fig. 4). An exception was strain ∆M05, which showed the same mild growth deficiency (lower cell yields) as its parent, probably due to an unidentified spontaneous second site mutation. Growth of non-magnetic ∆M01–∆M04 and ∆M13 mutants under anaerobic conditions was indistinguishable from the WT. However, in the presence of oxidative stress generated by H2O2, ∆M01–∆M04 grew to higher, and ∆M13 to lower densities than the WT, respectively (Fig. 4). Deleted genes in ∆M13 include a putative aerotaxis-related gene and several hypothetical genes, the loss of which might have caused the decreased sensitivity to oxidative stress.
Next, we tested whether the magnetic phenotypes could be restored by a compact version of all key magnetosome biosynthesis operons. To this end, a transposable cassette comprising feoAB1op, mms6op, mamGFDCop, mamABop, and mamXYop without intervening gene content was utilized. This cassette was harbored on pTpsMAG1 comprising the MycoMar (tps) transposase gene [38]. Reinsertion of the cassette at several random chromosomal locations in ∆M01–∆M04 and ∆M13 restored magnetosome biosynthesis to WT-levels (Figs. 3 and S1). This again confirmed that deleted genes apart from the mam/mms gene clusters are dispensable for magnetosome biosynthesis in M. gryphiswaldense. The presence of an extra copy of the endogenous feoAB1op seems to have no effect on magnetosome biomineralization, but it should be removed in future engineering steps to avoid unintended recombination events. After ‘re-magnetization’, growth rates of ∆M01::pTpsMAG1–∆M04::pTpsMAG1 and ∆M13::pTpsMAG1 were reduced to WT-levels under aerobic conditions and moderate heat stress. These findings indicate that magnetosome biosynthesis represents a significant burden that prevents cells from reaching higher cell yields observed in non-magnetic mutants. Under anaerobic conditions, complemented ∆M01::pTpsMAG1–∆M04::pTpsMAG1 and ∆M13::pTpsMAG1 strains showed WT-like cell yields. Under oxidative stress, complemented ∆M04::pTpsMAG1 revealed slight growth deficiencies (reduction by ~ 12% of WT OD), while the complemented ∆M13::pTpsMAG1 exhibited significantly reduced growth compared to the WT (reduction by ~ 70% of WT-level; Fig. 4).
Of note, in some of the non-magnetic mutants (∆M01–∆M05 and ∆M13) (Fig. 3) TEM revealed the presence of numerous (ca. 90 per cell) irregularly shaped conspicuous electron dense particles ranging 10–125 nm in size (in the following referred to as ‘EDP’), scattered over the entire cell. Analysis of strains ∆M03 and ∆M05 by high-resolution electron microscopy revealed that EDPs were amorphous. In addition, energy-dispersive X-ray spectroscopy (XEDS) showed that the inorganic inclusions were rich in potassium, phosphorus and oxygen, while no significant amounts of iron could be detected (Fig. 5). Variation of culture conditions such as growth in low-iron medium [25] supplemented with 10 μM 2,2′-dipyridyl as non-metabolizable iron chelator, or in medium oversaturated with 250 μM Fe (III)-citrate did not affect the number, size or appearance of EDPs (data not shown), confirming their independence from iron. Formation of EDPs was neither affected by variation of the phosphate concentration in the medium (0–3 mM), suggesting that low residual phosphate was still saturating for EDP formation. Furthermore, EDPs remained present in cells even after restoration of magnetosome biosynthesis by pTpsMAG1 complementation (Figs. 3 and S1). This indicates that the formation of EDPs is independent of magnetosome biosynthesis, but somehow linked to the deleted genes outside the five key magnetosome biosynthetic clusters. Because of their apparent irrelevance for magnetosome biosynthesis and growth, the identity and formation of EDPs was not explored further in this study.
Overall, the strain with the largest deletion that exhibited WT-like magnetosome biosynthesis upon complementation was ∆M13. In this mutant, a contiguous stretch of ~ 100 kb including all mam and mms6 operons (~ 27 kb) but feoAB1op, interspaced or flanked by ~ 73 kb of irrelevant or problematic gene content was deleted and substituted by a contiguous, yet functional version of magnetosome biosynthetic gene clusters (Fig. 1).
Deletion of putative determinants for magnetosome biosynthesis outside the MAI
Next, we assessed the role of candidate genes with putative roles during magnetosome biosynthesis located outside the MAI. One group of these candidates was recently retrieved by genome-wide transposon mutagenesis, in which a colony appearance deviant from the dark-brown color of the WT served as a proxy for impaired magnetosome biomineralization [31]. Another category was comprised of candidate genes, whose gene products were found to be genuinely associated with magnetosome particles purified from disrupted M. gryphiswaldense cells [32]. Most interesting targets for mutagenesis were further selected based on their conservation in other magnetospirilla and/or a conspicuous genomic neighborhood. This resulted in the following list of deletion targets (Fig. 6; Table S3):
Candidates identified by Tn5-mutagenesis [31]
– A clone with a reduced Cmag was linked to a hit in msr1_17870, which is part of a putative operon comprising eleven genes (msr1_17870–17940) that is conserved in two other magnetospirilla (Table S3). It has predicted functions related to the TonB-system, which is known to form energized, gated pores that bind and internalize iron chelates in Gram-negative bacteria [39]. Here, we deleted the entire 9.5 kb operon region.
– Several Tn-insertants within a huge (31 kb) monocistronic gene (msr1_20490) became suspicious because of their slightly altered colony appearance [31]. The gene encodes a single giant putative surface protein with a predicted mass of 1147 kDa and a repetitive structure, which belongs to the FecR/concanavalin A-like lectin/glucanase superfamily [31]. It is also conserved in several other magnetic and non-magnetic magnetospirilla (Table S3). In our study, we deleted the entire open reading frame of msr1_20490.
– Conspicuously, msr1_24180 was also hit by several independent Tn5-insertions [31] and is conserved in most magnetospirilla (Table S3). It contains a lysylphosphatidylglycerol synthase transmembrane region with putative function in cell wall modification [31]. We deleted msr1_24180 (~ 1 kb) in this study.
– The first four genes (msr1_30910–30940) of a six-gene operon were hit several times independently [31] and are conserved in several magnetospirilla (Table S3). The predicted functions (e.g., a glycosyl transferase gene, a dTDP-sugar isomerase, a methyltransferase and epimerase/dehydratase (NAD) gene) may play an important role in cell wall biogenesis or modification reported by Silva et al. (2020). msr1_30910–30940 (~ 3.5 kb) were deleted in this study.
– msr1_33570 and msr1_33770 are hypothetical genes which were also retrieved by the Tn-screen. They are conserved in many magnetospirilla (Table S3). Both genes were deleted (∆msr1_33570, 1.2 kb, ∆msr1_33770, 0.3 kb).
Candidates identified by magnetosome membrane proteomics [32]
-
MSR1_13180 (10 kDa), MSR1_16710 (9 kDa) and MSR1_19470 (11 kDa) are transmembrane proteins with unknown functions, but orthologs in many magnetospirilla. All three respective genes were deleted individually (0.27 kb, 0.249 kb, 0.33 kb, respectively).
-
MSR1_30840 is a transmembrane protein (33 kDa, four TMH) predicted as a putative peptidase, encoded next to potential LPS core biosynthesis genes, which is also conserved in two other magnetospirilla (Table S3). In addition to its detection in the magnetosome membrane [32], msr1_30840 is within close genomic neighborhood (7.4 kb) to msr1_30910–30940, all having received several Tn5-hits [31]. ∆msr1_30840 was generated in this study (0.951 kb).
Deletion mutants of all targeted genes could be obtained in a straightforward manner. Some of the null mutants (∆msr1_20490, ∆msr1_30910–30940, ∆msr1_30840) displayed a slightly reduced Cmag (< 1), compared to WT-levels of 1–2, and the cell shape of ∆msr1_20490 seemed to be more spiralized. However, TEM analysis revealed the presence of magnetosomes apparently indistinguishable from the WT with respect to number, size, shape and alignment in all mutants (Fig. S2). Hence, contrary to the previous hypotheses, these genes play no obvious and strong role in magnetosome biosynthesis under the tested conditions.