|Home | About | Journals | Submit | Contact Us | Français|
Magnetotactic bacteria are remarkable organisms with the ability to exploit the earth’s magnetic field for navigational purposes. To do this, they build specialized compartments called magnetosomes that consist of a lipid membrane and a crystalline magnetic mineral. These organisms have the potential to serve as models for the study of compartmentalization as well as biomineralization in bacteria. Additionally, they offer the opportunity to design applications that take advantage of the particular properties of magnetosomes. In recent years, a sustained effort to identify the molecular basis of this process has resulted in a clearer understanding of the magnetosome formation and biomineralization. Here, I present an overview of magnetotactic bacteria and explore the possible molecular mechanisms of membrane remodeling, protein sorting, cytoskeletal organization, iron transport and biomineralization that lead to the formation of a functional magnetosome organelle.
Bacteria are remarkable at devising intricate mechanisms to exploit and inhabit a variety of environmental niches. One particularly elegant solution is employed by magnetotactic bacteria, which are capable of navigating along the earth’s magnetic field in order to efficiently find low oxygen environments. This behavior is mediated by a specialized organelle, the magnetosome, consisting of a lipid-bilayer membrane that houses a crystal of the magnetic mineral magnetite (Fe3O4) or greigite (Fe3S4). Individual magnetosomes are held together into one or multiple chains thereby providing the cell with the capability to align in magnetic field. In the nearly forty years since their discovery, these unique microorganisms have captured the imagination of researchers in diverse fields ranging from geology to applied medicine. The last decade has witnessed an explosion in the search for and understanding of the molecular mechanisms of magnetosome formation and biomineralization of magnetic minerals. In this review I present a general overview of this field with a strong emphasis on the recent advances in the discovery and functional analysis of magnetosome formation genes. Given the long history of this field, a single review article cannot be comprehensive in its scope. As such, readers interested in gaining a better understanding of the history and breadth of research on magnetotactic bacteria are encouraged to refer to a number of excellent review articles written over the years (Blakemore, 1982, Bazylinski & Frankel, 2004, Jogler & Schüler, 2009).
Magnetotactic bacteria (MB) were serendipitously discovered almost four decades ago by Richard Blakemore (Blakemore, 1975). While observing sediments collected near Woods Hole, Massachusetts, Blakemore noticed that a conspicuous group of fast-swimming bacteria consistently swam in the same geographic direction, a pattern of behavior that was unrelated to the positioning of the microscope or external stimuli such as light. However, placing a magnet near the microscope slide altered their swimming direction. Using various techniques, Blakemore also showed that these bacteria contained chains of crystal-like inclusions that were rich in iron, thus providing a mechanistic explanation for their ability to respond to magnetic fields (Blakemore, 1975). Interestingly, it has recently come to light that this amazing finding was not the first description of magnetotactic behavior in bacteria (Frankel, 2009). Starting in the late 1950’s and continuing until 1963, Salvatore Bellini, a medical doctor at the University of Pavia in Italy, made a series of observations on the peculiar behavior of bacteria with distinct morphologies in freshwater samples collected from around Pavia (Bellini, 2009, Bellini, 2009). Much like Blakemore, he noticed that certain bacteria accumulated at the edge of a water droplet corresponding to the magnetic North. Using various experimental setups, including simple bar magnets and magnetic coils, he showed definitively that the bacteria of interest were “magnetosensitive.” He further investigated the response of these organisms to oxygen levels and devised methods for their enrichment. These findings were written in Italian and only distributed amongst a few Italian universities. A few years ago, with the help and advocacy of Richard Frankel, these reports were translated to English and published (Bellini, 2009, Bellini, 2009, Frankel, 2009). While Blakemore’s finding was not the original description of MB, it was the first to introduce the worldwide scientific community to these unique microorganisms. By following his intuition and insights, Blakemore launched a field that continues to grow and impact a diverse array of scientific disciplines some four decades later.
The early experiments by Blakemore and others established that MB respond to external magnetic fields through the use of specialized organelles termed magnetosomes (Balkwill, et al., 1980). The magnetosome consists of a magnetic mineral, either the iron oxide magnetite or the iron sulfide greigite, surrounded by a lipid bilayer membrane (Figure 1). Individual magnetosomes, which usually measure 50-70 nm in diameter, are organized into one to several chains within the cell that provide the means for alignment with magnetic fields. In fact, the magnetic properties of the magnetosome chain is sufficient to provide the sensitivity required to stably align with the weak magnetic field of the earth (Frankel & Blakemore, 1980). Surveys into the ecology of MB showed that they are found in most aquatic environments where oxygen and other redox active compounds are horizontally stratified. The opposing gradients of oxygen from the surface of the aquatic environment and sulfide from the bottom of sediments create a region termed the oxic anoxic transition zone (OATZ) where oxygen levels are very low. Most MB surveyed in the environment localize at or close to the OATZ (Simmons, et al., 2004). Another peculiar behavior of MB is that those found in the northern hemisphere, such as the organisms discovered by Blakemore and Bellini, tend to persistently swim northwards while MB from the southern hemisphere tend to swim southwards (Blakemore, et al., 1980). In both cases, this locked-in behavior directs the bacteria to the bottom of the aquatic body they inhabit, a region that is depleted in oxygen. Collectively, these observations have formed the foundations of a model to explain the selective advantage provided by magnetosomes (Figure 2). Under this model, in locales that are sufficiently distant to the equator, the earth’s magnetic field provides a roughly vertical guide through horizontally stratified environments. Thus, by being forced to swim along these magnetic freeways, MB have a simpler route to finding the OATZ as compared to bacteria that rely solely on chemotactic and aerotactic mechanisms. This model is partly supported by laboratory observations showing that MB rapidly move away from advancing oxygen gradients (Frankel, et al., 1997). Additionally, under such experimental setups, the response time of cells is enhanced when a magnetic field is provided as guide along the direction of the gradient (Smith, et al., 2006). As a result, it has been suggested that magneto-aerotaxis is a more adequate term for describing the behavior of these organisms (Frankel, et al., 1997). Magnetosomes provide a passive mode for alignment with the magnetic field and the aerotactic signaling machinery directs the swimming and positioning of the cell in redox gradients. In reality, the behavior of MB in the environment is likely to be more complicated than a simple response to oxygen levels. For instance, several species of MB display phototactic responses, which in one case has been hypothesized to reinforce the magneto-aerotactic behavior by repelling the bacteria from surface waters (Frankel, et al., 1997, Chen, et al., 2011, Shapiro, et al., 2011). In addition, genome sequences of several MB have shown that these organisms have some of the highest numbers of signaling proteins amongst the Bacteria. For instance, some MB contain over 60 predicted methyl-accepting chemotaxis proteins (MCPs), a tremendous number compared to the 5 MCPs of E. coli (Alexandre, et al., 2004). Perhaps, gradients of a multitude of compounds can be coupled to the benefits afforded by alignment and navigation along geomagnetic fields.
Though simple and well supported, several findings over the years have put the magneto-aerotactic model under question. First, some MB produce a large number of magnetosomes, far greater than would be needed to align stably in the earth’s magnetic field (Spring, et al., 1993). Second, MB have been found at or near the equator, a location where magneto-aerotaxis presumably has no selective advantage (Frankel, et al., 1981). Finally, south-seeking MB have been isolated from the Northern hemisphere at a location not too distant from the site of Blakemore’s discovery (Simmons, et al., 2006, Shapiro, et al., 2011). This last finding does not invalidate the magneto-aerotaxis hypothesis, as it may be possible that some bacteria can use their ability to orient in magnetic fields in order to search for compounds that are enriched near the surface rather than at the OATZ. As a result of these confounding observations, other hypotheses have also been proposed for the possible advantages gained by MB through the production of magnetosomes. One idea holds that these organelles can serve as iron storage or detoxification compartments. However, numerous studies have shown that MB can live under iron-poor conditions and, while high iron conditions are detrimental to some nonmagnetic mutants, the elimination of magnetosomes does not significantly harm the bacteria under standard laboratory conditions. (Simmons, et al., 2006, Shapiro, et al., 2011). An alternate model holds that the mechanical forces acting on the magnetosome chain as a cell turns in a magnetic field may somehow be translated into changes in its swimming behavior or perhaps energy generation (Philippe & Wu, 2010). Despite some intriguing molecular connections between magnetosomes and putative chemotaxis proteins (see below) little evidence exists to support this model. Finally, “the magnetosome battery” hypothesis proposes that external redox changes may lead to reversible transitions from magnetite, which contains both ferrous and ferric iron, to maghemite, a magnetic iron oxide that can be formed through the oxidation of magnetite (Kopp & Kirschvink, 2008). The organism can then exploit the change in the redox state of the mineral to generate energy. This model, though not yet supported by experimental data, is intriguing in that it provides an explanation for some of the oddities seen amongst MB. For instance, when viewed from this perspective, organisms with large numbers of magnetosomes may simply be generating more energy from this battery. Additionally, the battery may provide some usefulness to MB at the equator even if they cannot vertically travel through redox gradients. I would also like to propose the somewhat provocative possibility that magnetosomes may provide little to no selective advantage for the organism. Viewed through an evolutionary prism, as long as this process is not significantly disadvantageous to MB there will not be any selective pressure to “lose” the ability to make magnetosomes. In this light, magnetosomes may simply be vestigial, a remnant of a system useful in the past or perhaps the byproduct of the activity of a set of “selfish genes.”
Regardless of the actual function of magnetosomes the study of MB has impacted a number of scientific and applied disciplines. For instance, these organisms likely play an important part in cycling of iron in aquatic environments. While comprehensive environmental surveys of magnetotactic bacteria are rare, some studies suggest that in certain environmental niches they may be present at significant numbers where they constitute up to 5% of the total bacterial population. As a result, in such habitats a significant amount of the dissolved iron, perhaps 1-10%, will be sequestered by MB and will eventually be deposited as magnetite in sediments (Simmons, et al., 2004, Simmons, et al., 2007).
Additionally, the precise control over the composition, size and morphology of magnetite crystals found within magnetosomes has made this system a powerful model for the study of biomineralization (Kirschvink & Hagadorn, 2000). Numerous organisms are capable of transforming inorganic elements, such as calcium and silica, into intricate three-dimensional structures used for a variety of functions (Hildebrand, 2008, Weiner, 2008). However, MB are the most ancient and simplest organisms capable of such activity. In fact, fossilized magnetite chains and crystals have been discovered in numerous terrestrial samples, some dating to a few million to perhaps even billions of years ago (Kirschvink & Hagadorn, 2000, Kopp & Kirschvink, 2008, Schumann, et al., 2008). Additionally, magnetite biomineralization is not limited to bacteria. Numerous organisms, such as honeybees, fish, termites and pigeons are capable of forming magnetite and related iron oxides and most likely use them for direction sensing (Kirschvink, 1981, Kirschvink & Gould, 1981, Mann, et al., 1988, Mikhaylova, et al., 2005, Hsu, et al., 2007). Surprisingly, magnetite has also been found in human brain tissue samples raising the intriguing possibility that these minerals may play some functional role in humans as well (Kirschvink, et al., 1992). Also consistent with these findings is the accumulating evidence that certain neurodegenerative diseases are associated with a build up of precipitated metals in the brain. For instance, patients with a familial disease called neuroferritinopathy can have a build up of magnetite in the brain due to defects in the iron storage protein, Ferritin (Hautot, et al., 2007). Clearly, a deeper understanding of the process of magnetite formation by MB can lead to insights into the mechanisms and evolution of biomineralization. Additionally, understanding this process in a simplified system may provide a route to therapy in cases where magnetite is linked to a pathological condition.
The process of magnetosome formation also requires a sophisticated level of control at the cell biological level. Although underappreciated, many bacteria are capable of forming specialized compartments with features traditionally attributed to eukaryotic organelles (Murat, et al., 2010). The biogenesis of these organelles requires the formation of a lipid bilayer and targeting of a specific set of proteins as well as the localization, maintenance and proper division of the resulting compartment throughout the cell cycle. Given the genetic tractability of MB, magnetosomes are a perfect model for the study of cellular organization and compartmentalization in bacteria.
Finally, the unique properties of these magnetic particles, in particular their uniform size and shape, purity and their production under ambient conditions have made them an attractive choice for potential biotechnological and biomedical applications (Lang, et al., 2007, Yoshino, et al., 2010). For instance, magnetite particles can make great contrast agents for nuclear magnetic resonance imaging (MRI) and they have been proposed as potential therapeutic agents in the hyperthermic treatment of tumors (Ito, et al., 2005, Alphandery, et al., 2011). Magnetic bacteria have also been thought of as potential tools for bioremediation of toxic metals. Under one scenario, MB would sequester toxic compounds within magnetosomes and would then be separated from the water sample with magnets. One barrier to this application is that, in general, the magnetic particles of MB are chemically pure. However, several reports have shown that under the proper conditions exogenous metals such as manganese, cobalt and copper may also be introduced to the magnetosome (Bazylinski, et al., 1993, Staniland, et al., 2008, Keim, et al., 2009). In addition, a recent report shows that MB can precipitate crystals of tellurium, discrete from their magnetite crystals, providing a proof of concept for the use of MB as bioremediation agents (Tanaka, et al., 2010).
For over two decades, the lack of a clear molecular understanding of magnetosomes had proven to be a major block to realizing their potential as models in basic and applied studies. Identification of the genetic elements required for magnetosome formation had stalled for a number of years due to the lack of cultured and genetically tractable model organisms. Magnetospirillum magnetotacticum MS-1, an alpha-proteobacterium originally named Aquaspirillum magnetotacticum MS-1, was the first magnetotactic organism to be isolated in pure culture (Blakemore, et al., 1979). The culturing of MS-1 facilitated numerous studies into the physiology and magnetic properties of MB (Frankel, et al., 1979, Balkwill, et al., 1980). However, its fastidious nature and inability to robustly grow into colonies on plates limited its utility as a workhorse for molecular studies of magnetosome formation. This problem was circumvented by the isolation of two other closely related species of the magnetospirilla, Magnetospirillum grysphiswaldense MSR-1 and Magnetospirillum magneticum AMB-1 (Matsunaga, et al., 1991, Schleifer, et al., 1991). Over the years genetic tools have been developed for AMB-1 and MSR-1 and the two have taken center stage as models for understanding this fascinating process (Matsunaga, et al., 1992, Schültheiss & Schüler, 2003, Komeili, et al., 2004). Two other members of the alpha-proteobacteria, Magnetococcus marinus MC-1 and Magnetovibrio blakemorei MV-1, have also been cultured and have provided crucial insights into the diversity of magnetotactic behavior and genomic organization of magnetosome genes (Meldrum, et al., 1993, Frankel, et al., 1997, Richter, et al., 2007). Finally, the delta-proteobacterium, Desulfovibrio magneticus RS-1, a cultured MB that does not belong to the alpha-proteobacteria, has provided a unique opportunity to explore the broader conservation and evolution of the magnetosome formation process (Sakaguchi, et al., 1993, Nakazawa, et al., 2009).
These cultured species of MB have fueled a dedicated and multifaceted search for the molecular basis of magnetosome formation over the last few years. Three parallel approaches- proteomics, genetic analysis and comparative genomics- have been responsible for this explosion in the knowledge of the molecular factors involved in magnetosome formation. The proteomic analysis of magnetosomes was sparked by the seminal work of Gorby and colleagues on the cellular organization of magnetosomes. By purifying magnetosomes from cell extracts of MS-1 with magnets, they were able to demonstrate that a distinct set of proteins was associated with this cellular fraction (Gorby, et al., 1988). Later work identified the Mam22 protein (also known as MamA or Mms24) as a major constituent of the magnetosome membrane (Okuda, et al., 1996, Okuda & Fukumori, 2001). Similar efforts, focusing on the purification of individual proteins, led to the identification of a handful of other potential magnetosome proteins (Matsunaga, et al., 2000, Okamura, et al., 2001). The big breakthrough, however, came from two major studies by Grünberg et al., who used a careful combination of magnetosome separation techniques, two dimensional gel electrophoresis, N-terminal sequencing and mass spectrometry to provide a comprehensive survey of the magnetosome proteome of MSR-1 (Grünberg, et al., 2001, Grünberg, et al., 2004). Other than compiling an inventory of magnetosome proteins, this work was significant in showing that the genes encoding the vast majority of magnetosome proteins were organized into a few gene clusters conserved in multiple species of magnetotactic bacteria. Proteomic analyses of AMB-1 and RS-1 magnetosomes have also confirmed the presence of some of the proteins found on the MSR-1 magnetosome membrane (Tanaka, et al., 2006, Matsunaga, et al., 2009).
One caveat of proteomic work is that contamination of small amounts of nonmagnetosome proteins can lead to a high level of false positives. In fact, the proteomic analyses of AMB-1 and RS-1 magnetosomes identified many proteins likely to be contaminants from other cellular fractions (Tanaka, et al., 2006, Matsunaga, et al., 2009). Therefore, genetic analyses have provided a useful complement to the proteomics work. Transposon mutagenesis of AMB-1 has been successful in yielding a number of mutants with severe defects in magnetosome formation (Matsunaga, et al., 1992, Nakamura, et al., 1995, Wahyudi, et al., 2001, Komeili, et al., 2004). Interestingly, many of the factors identified in this manner encoded the magnetosome proteins isolated in the large-scale proteomic studies discussed above (Komeili, et al., 2004). Other genetic analyses focused on identifying the genomic lesions responsible for the spontaneous loss of the magnetic phenotype in mutants of MSR-1 and AMB-1 (Schübbe, et al., 2003, Ullrich, et al., 2005, Fukuda, et al., 2006). In both cases, the deletion of a large chromosomal region encoding for known magnetosome membrane proteins was responsible for the loss of magnetic behavior.
Finally, comparative genomic studies have shown that many of the genes implicated by the proteomic and genetic work are conserved across multiple species of MB. Information from the genome sequencing projects of AMB-1, MSR-1, MS-1, MC-1 and MV-1 revealed that four genomic regions, the mamAB, mamGFDC, mms6 and the mamXY gene clusters, appear to be present and mostly conserved amongst these organisms (Richter, et al., 2007). As will be discussed below, some of the genes of the mamAB cluster are also found in the MB that belong to the delta-proteobacteria as well as the Nitrospira.
The genomic segment containing the gene clusters encoding for a majority of magnetosome proteins has many of the hallmarks of genomic islands often associated with pathogenic organisms. For instance, compared to the rest of the genome it has a skewed GC content and is associated with tRNA genes, pseudogenes, transposons and IS elements (Ullrich, et al., 2005, Richter, et al., 2007). As a result it has now been termed the magnetosome gene island or MAI (Ullrich, et al., 2005). In AMB-1, the MAI is demarcated by two ~1.1 kilobase (kb) direct repeats, is approximately 98 kb in length and encodes approximately 100 proteins, representing over 2% of the bacterium’s genetic material (Fukuda, et al., 2006). MSR-1’s MAI does not have clear boundaries but it is estimated to be as large or even larger than that of AMB-1 (Ullrich, et al., 2005). Given its large size, the mere replication of the MAI must represent an energetic burden for the organism. Thus, it may not be surprising to see that the region can be lost or mutated spontaneously through recombination and transposition events (Schübbe, et al., 2003, Ullrich, et al., 2005, Fukuda, et al., 2006). In AMB-1, recombination between the two flanking direct repeats leads to the loss of the entire MAI (Fukuda, et al., 2006). In MSR-1, a much broader array of MAI mutations including its full deletion have been observed (Ullrich, et al., 2005). Interestingly, stress conditions seem to increase the frequency of MAI loss in both organisms. Exposure to oxidative stress or storage in the cold increases the frequency of MAI deletions in MSR-1 to as high as 1 in a 100 cells (Ullrich, et al., 2005). Similarly, the deletion of peroxiredoxin genes in AMB-1 leads to a significant accumulation of MAI deletion mutants under high oxygen conditions (Ge, et al., 2011). These findings imply that a specific mechanism may be in place to delete the MAI upon exposure to certain stresses. However, it is also possible that the frequency of MAI loss is constant and these mutants have a fitness advantage over their wildtype siblings under stress conditions.
The discovery of the MAI has allowed for a broader examination of the evolution and conservation of magnetosome formation amongst distantly related magnetotactic bacteria. While the alpha-proteobacterial species of MB have proven useful as laboratory model systems, some of the more fascinating variations on magnetosome morphology and organization are found outside of this clade. For instance, the delta-proteobacterium Desulfovibrio magneticus RS-1, a member of the sulfate-reducing Desulfovibrio genus, produces unusual bullet-shaped crystals (Sakaguchi, et al., 1993, Pósfai, et al., 2006) (Figure 3A and 3B). The recently completed full genome sequence of RS-1 revealed that it too contains a version of the MAI, with a predominance of genes from the mamAB cluster and a notable absence of genes of the mms6 and mamGFDC operons (Nakazawa, et al., 2009).
Other insights into the ubiquity of the MAI have come from studies of uncultured magnetotactic bacteria. In one effort, screening of metagenomic libraries with probes to magnetosome genes led to the discovery of a putative magnetosome island. This study relied on libraries created from an unknown collection of microorganisms and as such variations in gene content or protein domain organization could not be directly related to specific phenotypes of a known organism (Jogler, et al., 2009). To address these issues, two recent reports have combined environmental enrichment with high throughput sequencing technologies to identify the MAI of two fascinating, yet uncultured MB. Magnetobacterium bavaricum (Mbav) is the most phylogenetically distant of the well-characterized MB. This organism belongs to the Nitrospira and forms hundreds of tooth-shaped crystals arranged in multiple strands within the cell (Spring, et al., 1993, Jogler, et al., 2010) (Figure 3C and 3D). Using large-insert libraries created from enrichments of Mbav, Jogler et al. attempted, without success, to screen for potential homologs of known magnetosome genes (Jogler, et al., 2011). They then amplified and sequenced the genomic DNA extracted from approximately 100 cells of Mbav individually isolated with a micropipette under the microscope. Having found a potential homolog of a magnetosome gene with this method, they reprobed their large-insert libraries and were able to identify a large genomic segment with the distinct characteristics of the mamAB gene cluster found in other MB (Jogler, et al., 2011). In a similar study, Abreu et al. attempted to define the magnetosome genes of the uncultured delta-proteobacterium Candidatus Magnetoglobus multicellularis (Abreu, et al., 2011) (Figure 3E and F). As the name suggests, this organism and its relatives live as multicellular aggregates in which each cell is capable of producing magnetic minerals allowing the cells, as a group, to navigate along magnetic fields (Keim, et al., 2004, Abreu, et al., 2008). The organism is also of great interest since it produces the magnetic mineral greigite and not magnetite in its magnetosomes (Mann, et al., 1990). Although at least one bacterium capable of producing both magnetite and greigite has been described previously (Bazylinski, et al., 1995), there has been a debate about whether the greigite producing MB use a completely different mechanism for biomineralization (Delong, et al., 1993). Abreu et al. enriched for and, using pyrosequencing, generated a partial genome sequence for M. multicellularis. Similar to the Mbav case, a group of genes with homology to the mamAB gene cluster was also found in M. multicellularis (Abreu, et al., 2011). Thus, at least the mamAB cluster, whose genes perform essential functions in magnetosome formation (see below), is found in all species of MB studied to date suggesting a common ancestry for magnetosome formation.
As will be detailed in the next section of this review, the conserved MAI genes are central to various steps of magnetosome formation. Their presence in some of the most diverse species of magnetotactic bacteria suggests that, at least for these organisms, magnetosome formation was invented only once during evolution. The conservation of the mam genes also allows for phylogenetic comparisons to trace the evolutionary history of magnetosome formation. Interestingly, the phylogeny of the mam genes from the alpha-proteobacteria, RS-1, Mbav and M. multicellularis follows the phylogeny of the housekeeping genes from the same organisms (Abreu, et al., 2011, Jogler, et al., 2011). One interpretation of these results has been that magnetosome genes spread to diverse clades of Bacteria through very ancient horizontal gene transfer events thus accounting for the parallel phylogenetic relationship between the MAI and its host genome (Jogler, et al., 2011). An alternate and perhaps more provocative explanation is that the last common ancestor of the Proteobacteria and the Nitrospira could have been a magnetotactic bacterium and over the course time most of the species of these two groups have lost their magnetosome genes (Jogler, et al., 2011). Certainly, the groundbreaking work on Mbav and M. multicellularis should serve as a blueprint for uncovering the magnetosome genes of other, diverse MB.
In the sections below I will discuss how the directed analysis of many of the MAI genes has led to a better understanding of the molecular mechanisms of magnetosome formation. I have attempted to piece together a large number of diverse studies and to facilitate this discussion I have divided the magnetosome formation process into four distinct stages: magnetosome membrane biogenesis, magnetosome protein sorting, magnetosome chain formation and biomineralization. For each of these processes I have provided some context for the importance and the physical nature of the process followed by a discussion of its possible genetic underpinnings.
Much like other cellular organelles, compartmentalization by the magnetosome membrane is thought to provide the proper chemical environment for biomineralization and to protect the cell from potentially toxic byproducts of this activity. Several reports have shown that the magnetosome membrane is an independent structure that serves as the site of mineral formation. The original description of magnetotaxis by Blakemore and a seminal paper by Gorby et al. on the ultrastructure of MB showed that some magnetosomes within a chain are empty and do not contain magnetite crystals (Blakemore, 1975, Gorby, et al., 1988). Additionally, in studies where all explicit sources of iron are eliminated from the growth medium, cells contain long chains of magnetosome membranes devoid of any minerals, even when cultured for multiple generations (Komeili, et al., 2004). Upon addition of iron, small nuclei of magnetite form within these pre-existing membranes and they continue to grow until reaching their final mature size. (Komeili, et al., 2004)
Accumulating evidence over the years has also proven that the magnetosome membrane is a lipid bilayer originating from the inner cell membrane. Early transmission electron micrographs of stained sections of various MB had revealed a “trilaminate” pattern that was reminiscent of the lipid bilayer of the inner cell membrane (Balkwill, et al., 1980). Additionally, purified magnetosomes have a lipid composition with ratios of various phospholipids that resemble that of the inner cell membrane (Gorby, et al., 1988, Grünberg, et al., 2004). Despite these crucial insights, the small size of the magnetosome membrane and its apparent fragility had made it difficult to investigate its precise subcellular arrangement by traditional electron microscopy methods that rely on extensive fixation, staining and sectioning steps. These barriers were overcome with the use of electron cryotomography (ECT) to generate high-resolution views of the ultrastructure and organization of magnetosomes within two species of MB (Komeili, et al., 2006, Scheffel, et al., 2006, Katzmann, et al., 2010). ECT relies on the imaging of rapidly frozen samples at multiple angles relative to an electron beam. A specific algorithm is then used to assemble this image series into a three-dimensional reconstruction of the specimen (Tocheva, et al., 2010). ECT imaging of AMB-1 showed that its magnetosomes are invaginations of the inner cell membrane (Komeili, et al., 2006) (Figure 1B). These invaginations are found across the chain and are either empty or contain magnetite crystals of various sizes, implying that the magnetosome stays as an invagination throughout the biomineralization process. In fact, no magnetosomes that are clearly separated from the inner membrane have been found in this organism. ECT imaging of MSR-1 has also shown that some magnetosomes are adjacent to the inner membrane and can be invaginations of the inner cell membrane (Scheffel, et al., 2006, Katzmann, et al., 2010). However, whether this state applies to all magnetosomes in this organism is not clear at the moment.
The continuity between the magnetosome membrane and the inner cell membrane, and by extension, the open connection between the magnetosome lumen and the periplasmic space, can perhaps provide mechanistic explanations for some of the key aspects of magnetosome formation and magnetotaxis. Since magnetosomes are integrated into the inner cell membrane, it is easy to imagine how changes in the external magnetic field can translate into a realignment of the entire cell. If magnetosomes were free within the cell then a separate structure would be required to provide a connection between the magnetosome chain and the cell body in order to allow for orientation within magnetic fields. Furthermore, an open connection between the magnetosome lumen and the periplasmic space might be important in the transport and transformation of iron during biomineralization. However, this configuration also poses problems that may hamper the formation of magnetic minerals. For instance, a mixing of the content between the two compartments might alter the specific chemical environment that would be needed to form pure crystals of magnetite. Additionally, there is evidence from imaging of other species of MB that are inconsistent with the permanent existence of magnetosomes as inner membrane invaginations. One example can be seen in the recent detailed imaging of Mbav (see below), which clearly shows that some magnetosomes are arranged such that their membranes would almost certainly be separate from the cell membrane (Jogler, et al., 2011). Images of some uncultured magnetic cocci also show a magnetosome chain that runs through the middle of the cell, making it unlikely that the magnetosome membrane is an extension of the inner cell membrane (see figure 3c of (Vali & Kirschvink, 1991)). Perhaps, as seen with other key physical features of magnetosomes, significant variation can exist between species regarding the biogenesis, organization and maintenance of the magnetosome membrane.
These observations suggest that the generation of the magnetosome membrane is a pre-requisite for the biomineralization of magnetite. However, ultrastructural studies on some species of magnetotactic bacteria have led to the surprising suggestion that, in some cases, a recognizable membrane may not surround magnetite crystals. For instance, the imaging of RS-1 by multiple techniques, including ECT, has failed to reveal the presence of a membrane around its magnetite crystals (Byrne, et al., 2010). This is particularly surprising since RS-1 clearly contains a modified version of the MAI (Nakazawa, et al., 2009). Additionally, proteomic analyses have shown that transmembrane domain-containing proteins encoded by the MAI, associate with purified magnetosomes of RS-1, implying that the magnetite crystals are at least transiently associated with a biological membrane during their biomineralization (Matsunaga, et al., 2009). Alternatively, it is possible that magnetosome membranes are so tightly associated with magnetite crystals in RS-1 that they cannot be identified by the electron microcopy-based techniques used in the study. Another organism capable of forming tooth-shaped magnetite crystals is Mbav. This organism is unusual in that it contains hundreds of magnetite crystals per cell arranged into a number of magnetosome chains. Much like RS-1, previous work on Mbav, had also failed to detect magnetosome membranes around its magnetite crystals (Hanzlik, et al., 2002). However, a detailed ultrastructural analysis of Mbav has shown that it likely does contain magnetosome membranes. Imaging of samples prepared by focused ion beam milling or freeze fracture show that the magnetosomes of Mbav have an intricate, rosette-like organization (Jogler, et al., 2011) (Figure 3C and 3D). In positions within the chain expected to contain magnetosomes, empty structures that resemble biological membranes are observed. Based on this and other evidence, the authors reached the conclusion that previous failures to detect magnetosome membranes were most likely due to a tight association of the membrane with the magnetite crystals (Jogler, et al., 2011).
In parallel to the efforts to define the physical nature of the magnetosome membrane, the molecular mechanisms of its biogenesis have also been under intense investigation. The question is of broad significance since processes that require remodeling of cellular membranes are widespread in nature. The ability to bend a lipid-bilayer requires the activity of specific proteins that lower the energetic barriers to exposing the hydrophobic tails of lipid molecules to the aqueous environment of the cell. As a result, eukaryotic cells depend on specific proteins to shape organelles and assist in bending of membranes during vesicle formation, endocytosis and cell division (Graham & Kozlov, 2010). While the components of the cytokinesis machinery of bacteria have been studied in depth, the processes by which bacterial lipid-bounded organelles are created and maintained are generally not understood at a molecular level (Murat, et al., 2010).
Comprehensive genetic analysis of the MAI has provided the most detailed clues into the potential mechanisms of magnetosome membrane biogenesis (Murat, et al., 2010). The mutants that result from the spontaneous deletion of the MAI are defective in the formation of the magnetosome membrane, thus explaining their inability to form magnetite crystals. To identify the genes involved in membrane biogenesis, Murat et al. generated several large deletions of the various putative gene clusters within the MAI of AMB-1 (Murat, et al., 2010). In one deletion, where the entire 18-gene mamAB cluster is deleted, no magnetosome membranes and consequently no magnetic minerals are formed. By deleting each of the genes within this cluster, four genes (mamI, mamL, mamQ and mamB) were identified whose individual deletions result in a loss of the magnetosome membrane from AMB-1 (Murat, et al., 2010). Two of these genes, mamI and mamL, encode small ~70 amino acid proteins with two predicted transmembrane segments. MamI and MamL are only found within MB and do not contain domains with homology to other proteins. However, MamL has a 15 amino acid C-terminal tail that is predicted to form a helix on the cytoplasmic side of the inner membrane. This region is enriched in positively charged residues in a manner that is reminiscent of peptides known to interact with and even cross membranes in a number of different systems (Schmidt, et al., 2010). Thus, in one possible model for magnetosome membrane formation, interaction of MamL’s positively charged residues with the cytoplasmic side of the inner membrane creates an asymmetry that helps to bend and shape the magnetosome membrane. The other two proteins, MamB and MamQ, are part of large families that are conserved beyond MB. MamB is a member of the Cation Diffusion Facilitator (CDF) family of transporters, a group of proteins that participate in the transport of zinc, iron and cadmium in various organisms. Perhaps MamB does not play a direct role in magnetosome membrane biogenesis and is instead acting as a landmark so that magnetosomes are built around a crucial transporter. MamQ is part of an uncharacterized but broadly conserved family of proteins, typified by the LemA protein of Listeria monocytogenes (D’Orazio, et al., 2003). The proteins of this class contain coiled-coil repeat domains that may facilitate interactions to bend the magnetosome membrane. At the moment these are only speculations based on the defects seen when mamI, mamL, mamQ or mamB are deleted. However, these genes may not play a direct role in remodeling of the inner cell membrane. For instance, the possibility that these proteins act in the maintenance, rather than biogenesis, of the magnetosome membrane cannot be ruled out at the moment. In addition, it is likely that mamI, mamL, mamQ and mamB are not the sole players responsible for the biogenesis of the magnetosome membrane. When these four genes are expressed in a strain bearing a deletion of the 18 genes of the mamAB gene cluster no magnetosome membranes are formed suggesting that other factors from this region, which are not necessary on their own, are required in conjunction with mamI, -L, -Q and -B to form a magnetosome membrane (Murat, et al., 2010).
In addition to factors responsible for production of the magnetosome membrane, at least one protein with a potential role in determining the size and shape of this compartment has been isolated in a recent study. A differential purification assay, separating magnetosomes based on the extent of their magnetism, found that certain proteins were enriched in magnetosomes containing smaller crystals. Since these magnetosomes may be in the early stages of magnetosome formation, the authors hypothesized that proteins affecting the biogenesis of the magnetosome membrane may be enriched in this fraction (Tanaka, et al., 2010). This process led to the identification of MamY, a protein encoded by the mamXY gene cluster of the MAI. A deletion of mamY results in the generation of slightly smaller magnetosome membranes and, in vitro, MamY was capable of tubulating purified liposomes. Intriguingly, the authors point to a distant homology between the primary amino acid sequence of MamY and BAR domain proteins, which are involved in membrane remodeling in eukaryotes (Gallop & McMahon, 2005, Tanaka, et al., 2010). However, this weak similarity is mainly due to the existence of coiled-coil domains in MamY and may not be reflective of a conserved functional role between the two classes of proteins. Regardless, the discovery of mamY shows that either directly or indirectly, the proteins within the magnetosome membrane can influence its size and morphology. Perhaps, such proteins are responsible for generating the diversity of magnetosome sizes seen in various species of MB.
As described above, the magnetosome membrane contains a unique set of proteins that differentiate it from other cellular membranes. The combination of biochemical analysis and genetic studies have shown that, depending on the organism and the experimental setup, approximately 20-40 proteins are localized or enriched at the magnetosome membrane (Grünberg, et al., 2001, Grünberg, et al., 2004). These include a number of proteins with one or more predicted transmembrane domains and some that would likely localize to the cytoplasmic side of the magnetosome membrane. While magnetosome proteins constitute a small fraction of the total cellular protein content, recent atomic force microscopy (AFM) imaging of magnetosomes implies that the surface of this membrane is covered in a layer of magnetosome proteins (Yamamoto, et al., 2010). In this breakthrough work, isolated magnetosomes were enveloped in a globular material that could be washed away by an alkaline treatment, a condition known to remove the MamA protein from magnetosomes (Taoka, et al., 2006, Yamamoto, et al., 2010). Furthermore, the globular material re-appeared once recombinant MamA was introduced to the alkaline-treated magnetosomes showing that these structures are likely composed of MamA. This study provides a much-needed complement to electron microscopic work since ECT images do not define any obvious electron-dense structures on the surface of the magnetosome (Komeili, et al., 2006, Scheffel, et al., 2006, Katzmann, et al., 2010). Interestingly, the authors also note that in addition to MamA, other unidentified proteins are also extracted from magnetosomes with alkaline treatment (Yamamoto, et al., 2010). A recent X-ray crystal structure of MamA has also shed light on the mechanisms by which the protein self assembles and potentially provides numerous surfaces for binding of other proteins (Zeytuni, et al., 2011). Thus, one suggestion from this work is that MamA may be bridging the interaction of some magnetosome proteins with the magnetosome membrane.
Since many magnetosome proteins share homology with rather ubiquitous cell membrane proteins, specific mechanisms must be in place to ensure their proper localization to the magnetosome and perhaps their exclusion from other cellular membranes. Thus far, no specific sequence determinants or localization signals have been identified on any magnetosome proteins. However, recent genetic evidence implicates the putative HtrA/DegP family protease, MamE, in the localization of proteins to the magnetosome (Murat, et al., 2010, Quinlan, et al., 2011). Amongst the large number of genes deleted in the Murat et al. study, the individual deletions of mamE, mamO, mamM and mamN have a unique and striking phenotype. These mutants are completely nonmagnetic and produce no electron-dense particles when analyzed by electron microscopy (Murat, et al., 2010). However, they are still capable of forming chains of empty magnetosome membranes (Murat, et al., 2010, Yang, et al., 2010). This phenotype may be due to a direct involvement of the proteins in biomineralization or it may be caused by a general mislocalization of magnetosome proteins. When mamO, mamM or mamN are deleted, a number of magnetosome proteins fused to the Green Fluorescent Protein (GFP) display a wildtype localization pattern. In contrast, the deletion of mamE results in the mislocalization of all of these magnetosome proteins (Murat, et al., 2010, Quinlan, et al., 2011). Interestingly, the putative protease activity of MamE and its presumed heme-binding domains are not required to promote protein localization to the magnetosome (Quinlan, et al., 2011). Given that its predicted enzymatic activity is not required for this process, it is possible that MamE functions as a localization determinant by providing a landmark or platform for recruitment of other magnetosome proteins through direct physical interactions. Such “diffusion and capture” mechanisms are actually quite common in bacteria and have been implicated in numerous processes such as the differential localization of proteins during sporulation in Bacillus subtilis and polar localization of chromosomal origins in Caulobacter crescentus (Rudner, et al., 2002, Bowman, et al., 2008, Ebersbach, et al., 2008, Rudner & Losick, 2010). However, it is important to note that the phenotypes of the mamE deletion may be due to defects in maintenance of proteins at the magnetosome and not reflective of an active sorting mechanism. Furthermore, it is possible that the set of proteins acting at the membrane biogenesis step also have a function in protein localization to the magnetosome membrane. These proteins could act in the recruitment of MamE to the magnetosome or in concert with MamE help to localize specific proteins to the magnetosome. Finally, some magnetosome proteins do not contain any predicted transmembrane domains and they likely find their location in the cell through interactions with other magnetosome membrane proteins.
Most of the experiments used thus far to identify magnetosome proteins and to search for their sorting determinants presume that there is a defined and constant protein content to magnetosomes at all times. However, it is likely that individual magnetosomes within a chain can have a different inventory of proteins at various stages of their development and activity. Some biochemical studies, such as the work with MamY described above, provide support for the idea that there may be variations in the protein content of individual magnetosomes (Tanaka, et al., 2010). Other evidence for this viewpoint has come from live cell imaging studies with GFP fusions to various magnetosome proteins. For example, MamA fused to GFP displays a dynamic localization pattern that is dependent on the growth phase of the bacterial culture. In exponential phase, MamA localizes across the cell as a thin punctate line, whereas in stationary phase it is found as one or two intense foci (Komeili, et al., 2004). Variations in the localization of the biomineralization protein MamC have also been observed when it is fused to GFP. In some cases the protein is found in a line and in others as an intense focus within the cell (Lang & Schüler, 2008). The mechanisms behind and the function of such dynamic changes to localization of magnetosome proteins have yet to be determined. Perhaps, as discussed more thoroughly below, the timing of biomineralization and the number of crystals produced can be modulated through selective localization of certain magnetosome proteins
In order to function as a compass needle to orient the cell in magnetic fields, individual magnetosomes need to be organized into chains. On a cytological level, different species of MB form chains that look quite distinct from each other. For instance, the magnetospirilla, such as AMB-1 and MSR-1, have a single magnetic chain of 15-30 magnetosomes while Mbav contains several hundred crystals organized into multiple chains, each of which is composed of several individual strands of magnetosomes. Even amongst the closely related AMB-1 and MSR-1 species, striking differences exist in the organization and assembly of the magnetosome chain. In AMB-1, empty magnetosome membranes are already organized into chains prior to magnetite biomineralization (Komeili, et al., 2004, Komeili, et al., 2006). In contrast, MSR-1 cells contain an overabundance of disorganized empty magnetosomes that are apparently brought together into a chain through magnetic interactions (Scheffel, et al., 2006). Despite these differences, recent work has shown that a shared mechanism relying on an actin-like cytoskeleton maybe at the core of the magnetosome chain assembly process.
In addition to providing a detailed view of the organization of the magnetosome membrane, the ECT imaging of AMB-1 and MSR-1 also revealed the presence of distinct filaments, with dimensions similar to actin-like filaments, alongside the magnetosome chain (Komeili, et al., 2006, Scheffel, et al., 2006, Katzmann, et al., 2010) (Figure 1C and and4A).4A). Intriguingly, mamK, one of the genes within the MAI, encodes a member of the bacterial actin-like family of proteins (Grünberg, et al., 2001). Bacterial actins were discovered over 20 years ago, where mutations in the mreB gene led to cellular resistance to mecillinam antibiotics and also caused a dramatic change in morphology of E. coli from rod-shaped to round cells. Over the last decade, a series of important advances has shown that these actin-like proteins are nearly ubiquitous in bacteria performing functions ranging from the positioning of cell wall synthesis enzymes to plasmid DNA segregation (Carballido-Lopez, 2006, Shaevitz & Gitai, 2010). While mamK lies within the mamAB gene cluster, a likely functional ortholog of mreB can be found outside the MAI. The obvious homology of MamK to bacterial actins and its association with the key gene cluster for magnetosome formation led to the hypothesis that it might provide a structural scaffold for the organization of the magnetosome chain (Komeili, et al., 2006). Indeed, the deletion of mamK in AMB-1 produced no major defects in biomineralization or magnetosome formation. However, in this mutant, individual magnetosomes were clearly localized in a scattered fashion throughout the cell and did not organize as one coherent chain (Komeili, et al., 2006) (Figure 4A and 4B). Additionally, the magnetosome specific cytoskeleton observed in ECT images was no longer present in this strain (Komeili, et al., 2006). Recently, a deletion of mamK in MSR-1 has confirmed several of the observations seen in the corresponding AMB-1 mutant strain. In the absence of mamK, MSR-1 cells form short chains that are separated by gaps devoid of magnetosomes and the magnetosome cytoskeleton is no longer seen near the magnetosome chain (Katzmann, et al., 2010). Additionally, they appear to form fewer crystals per cell, which may indicate a role in biomineralization or proper segregation of the magnetosome chain (Katzmann, et al., 2010). While these results are fascinating they also leave us with crucial questions regarding the function of MamK filaments in vivo. The resulting disorganization of the magnetosome chain when MamK is absent can be caused by at least two distinct functions of the protein. First, MamK could act early in the chain formation process to bring new magnetosomes to the pre-existing chain. On the other hand, it could act late in the process and maintain the chain structure, perhaps to ensure its proper positioning in the cell during cell division. Such a function has recently been demonstrated for MreB as well as the cytoskeletal protein ParA in organizing the carboxysome organelles of cyanobacteria (Savage, et al., 2010).
While the precise function of this protein remains to be determined there has been a great deal of interest in understanding its physical properties as a member of the bacterial actin family. Bacterial actins are a large and diverse group of proteins with many phylogenetically distinct branches (Derman, et al., 2009). MamK and its homologs from various magnetotactic bacteria form a unique branch within this ever-expanding phylogenetic tree. Interestingly, homologs of MamK are not limited to the MB and can even be found in organisms that are not magnetotactic. For instance, the sulfide-oxidizing arsenate-reducing delta-proteobacterium MLMS-1 (Refseq: NZ_AAQF00000000) has a fully sequenced genome with multiple copies of bacterial actins all of which are part of the MamK clade (see ZP_01287033 as an example). This organism does not contain any of the other major magnetosome genes and does not appear to be capable of forming magnetosomes. Thus, it is likely that MamK has properties that make it useful in situations unrelated to magnetosome organization. Initial biochemical characterization of MamK has shown that the protein can form filaments in an ATP-dependent manner similar to other actin-like proteins and that it can, at some level, associate with magnetosomes (Taoka, et al., 2007). In vivo filament formation by MamK is independent of other magnetosome genes since it can be readily seen in E. coli (Pradel, et al., 2006). However, similar to other actin-like proteins, it is likely that accessory proteins are required to promote the dynamics or function of MamK in MB. Two such interaction partners have been reported in recent studies. MamJ, a protein encoded by the gene directly upstream of mamK, is an acidic protein with an important function in magnetosome chain formation. When this protein is absent, MSR-1 cells fail to form straight magnetosome chains and their magnetosomes are aggregated within the cell (Scheffel, et al., 2006) (Figure 4C). When the mamJ deletion strain is viewed by ECT, magnetosome-associated filaments are not seen attached to magnetosomes leading to the model that MamJ may mediate the interaction between magnetosomes and MamK (Scheffel, et al., 2006, Scheffel & Schüler, 2007). However, it is notable that the phenotype of mamJ and mamK single deletion strains are remarkably different from each other, suggesting that extra functions may be performed by MamJ such that its loss leads to a more severe phenotype than the loss of MamK.
One possible function for MamJ is in promoting the dynamic turnover of MamK filaments in vivo. When assayed by fluorescence recovery after photobleaching (FRAP), MamK filaments fused to GFP display a dynamic behavior that is reminiscent of other actin-like proteins (Draper, et al., 2011). This turnover depends on the putative nucleotide hydrolysis capabilities of MamK and, most importantly, does not occur in the absence of the MAI. One peculiarity of AMB-1 is that it contains two homologs of mamJ, the second of which has been named like-mamJ or limJ. When either mamJ or limJ are deleted, MamK-GFP filaments behave much like they do in wildtype cells. However, when both genes are deleted, MamK-GFP filaments are no longer dynamic (Draper, et al., 2011). The cessation of filament dynamics also results in subtle defects in chain organization such that numerous gaps disrupt the normally continuous row of magnetosomes. Even more striking, the gaps in the chain are filled with bundles of what appear to be MamK filaments. Other factors within the MAI are still thought to be important for filament turnover, since neither MamJ nor LimJ is sufficient to restore MamK dynamics in an MAI deletion strain (Draper, et al., 2011).
MamK has also been shown to interact with Amb0994, a potential methyl accepting chemotaxis protein (MCP) encoded within the MAI, in a bimolecular fluorescence complementation (split GFP) system (Philippe & Wu, 2010). In accordance with its homology to MCPs, when Amb0994 is overexpressed, the motility of the cells in response to changing magnetic fields is impeded. These intriguing observations imply that the role of magnetosomes may extend beyond passively aligning cells in geomagnetic fields. Perhaps, components of the magnetosome can be used to “sense” and translate external magnetic fields into changes in cellular behavior (Philippe & Wu, 2010). However, several outstanding questions need to be resolved before this provocative model is confirmed. For instance, overexpression of Amb0994 may have an effect that is independent of its possible interaction with MamK. Furthermore, the interaction with MamK needs to be confirmed with other, more direct, biochemical methods.
These studies demonstrate that a great deal of potential exists for discovery of interacting partners that may modulate MamK activity and behavior in vivo. Perhaps, through the discovery of these regulators and interaction partners we may be able to get a better view of the specific action of MamK. Also, study of MamK from diverse species may be a key to uncovering how changes in its sequence may have led to changes in the organization and morphology of magnetosome chains from various species. In fact, intriguing variations in MamK sequence are seen amongst the various MAIs examined to date. Most notably, a metagenomic search for magnetosome islands from uncultured microbes revealed a hybrid MamK protein fused to an FtsZ homolog (Jogler, et al., 2009). FtsZ is the bacterial tubulin-like protein most often associated with cell division in bacteria. A hybrid actin-tubulin protein may present interesting opportunities for formation of complex hybrid cytoskeletal structures that could facilitate the organization of unique chain architectures. Additionally, some species of MB are known to contain two homologs of MamK. Recently, Rioux and colleagues have discovered a poorly annotated and cryptic magnetosome island-like genomic locus in AMB-1 (Rioux, et al., 2010). This region, termed the magnetosome islet, contains homologs of several magnetosome genes including mamK, mamE, mamJ, mamQ and mamD. The codon usage within the islet suggests that it may have been acquired via a second independent gene transfer event since it is more similar to the magnetosome island of MC-1 than that of AMB-1. The mamK homolog, termed mamK-like, is capable of forming filaments in vitro as well as when expressed in E. coli (Rioux, et al., 2010). The implications of the existence of such a MamK homolog are unclear at the moment. One unusual feature of MamK-like is that it contains an amino acid substitution at a key residue that is predicted to function in ATP hydrolysis. In other bacterial actins such a change interferes with assembly or dynamics of filaments in a dominant fashion. While it is not known if these two proteins can interact to form filaments, it is likely that a mixture between MamK and MamK-like would similarly change the dynamic behavior of MamK and lead to static filaments. The authors also suggest that the existence of a second MamK protein in AMB-1 may be responsible for the differences seen between the chain organization of this organism and that of other MB (Rioux, et al., 2010). The ubiquity and variations in MamK function and structural organization also need to be examined in various species of MB. The ultrastructural study of Mbav has shown that a relatively thick filament runs along the center of each of the rosette-like magnetosome chains (Jogler, et al., 2011). Surprisingly, no clear MamK homologs were discovered in the MAI of Mbav leading to speculation that this organism may have a novel mode of chain organization (Jogler, et al., 2011). However, the sequencing of the Mbav genome is incomplete leaving open the possibility that a homolog of MamK may exist in this organism. Finally, the AFM studies on purified magnetosomes also suggest a role for MamA in the spacing and organization of magnetosomes. When MamA is washed away with alkaline treatment, the space between neighboring magnetosomes increases by approximately 3-4 nm, a change that is reversed upon the addition of recombinant MamA (Yamamoto, et al., 2010). Further work is needed to show if this effect is also seen in vivo and to decipher the specific mode by which MamA may regulate the structure of the magnetosome chain.
The ultimate goal of building a magnetosome compartment is to provide the environment to safely and efficiently direct the formation of crystalline magnetic minerals with specific shape and size characteristics. Over the years a number of different approaches have been employed to decode the molecular mechanisms by which MB are capable of exerting such precise control over the properties and dimensions of the inorganic crystals formed within their magnetosomes. In addition to identifying a number of important biomineralization factors, they have also led to a better definition of the specific steps involved in the formation of the mineral. Based on this large body of work, I have divided the discussion of biomineralization into four separate topics: iron transport, initiation of crystallization, crystal maturation and control over size and morphology.
The first and perhaps most critical step in magnetite biomineralization is the transport of iron from the extracellular environment into the cell. MB are capable of accumulating significant amounts of intracellular iron that are at least 100 times higher than that found in commonly studied bacteria such as E. coli. For most MB, this process is also regulated in response to environmental oxygen levels. Although many MB are actually microaerophillic such that they can thrive with low levels of oxygen, iron uptake and in turn magnetite formation do not initiate until nearly anaerobic conditions are reached in the surrounding medium (Blakemore, et al., 1985, Schüler & Baeuerlein, 1996, Schüler & Baeuerlein, 1998, Heyen & Schüler, 2003). At the moment it is not known if this regulation is genetically mediated or if it is caused by changes in iron chemistry under anaerobic conditions.
Despite the unusually elevated levels of iron uptake, many MB appear to use fairly generic systems for transport of iron into the cell. Early work with MS-1 showed that this organism was capable of producing hydroxamate-type siderophore molecules, which are often used in the transport of iron to the interior of the cell (Paoletti & Blakemore, 1986). A similar study also confirmed the production of siderophores by AMB-1 (Calugay, et al., 2003). However, siderophores are normally produced under iron deprivation conditions while magnetite biomineralization requires the presence of relatively high iron levels in the growth medium. A more careful analysis of the process in AMB-1 showed that siderophores are not produced during the early phases of magnetite biomineralization (Calugay, et al., 2003).
Instead, when the iron levels in the growth medium are depleted as a result of biomineralization, the cells initiate siderophore production (Calugay, et al., 2003). Thus, siderophores are likely not involved in iron transport during magnetite biomineralization. A broader view into the requirements for iron transport was provided by the analysis of global gene transcription of AMB-1 cells grown with various concentrations of iron (Suzuki, et al., 2006). Interestingly, many of the changes seen, resemble what may be expected in any other bacterium; ferrous iron transporters are upregulated whereas ferric iron transporters are downregulated. However, none of the MAI genes, including many with clear homology to CDF transporters, changed expression in this analysis (Suzuki, et al., 2006). Furthermore, a genetic analysis of the Ferric Uptake Regulator (Fur) transcription factor in MSR-1 suggests that the overall balance of cellular iron must play a key role in the biomineralization process (Uebe, et al., 2010). Deletion of the fur homolog in MSR-1 resulted in the production of fewer magnetite crystals. In addition, the general distribution of iron within the cell was perturbed with a higher portion accumulating bound to a ferritin-type molecule (Uebe, et al., 2010). A similar study also showed that the disruption of another fur-like gene, most likely related to the Irr family of Fur-like proteins, also results in defects in iron accumulation and magnetosome formation in MSR-1 (Yijun, et al., 2007). These results have led to a suggestion that some transport of iron from a cytoplasmic pool into magnetosomes may be required for magnetite synthesis.
Specific iron transporters have also been implicated in the biomineralization of magnetite. The first effort to identify global regulators of magnetosome formation was a transposon mutagenesis screen searching for nonmagnetic mutants of AMB-1. This screen yielded a disruption of magA, a gene encoding for a putative iron transporter (Nakamura, et al., 1995). When expressed in E. coli, MagA allows for ATP-dependent import of iron into purified inverted membrane vesicles (Nakamura, et al., 1995). In AMB-1 cell extracts, the activity of a Luciferase fusion to MagA is associated with the cell membrane fraction as well as the magnetosomes (Nakamura, et al., 1995). However, the proteomic studies on MSR-1 have failed to detect MagA as a magnetosome protein (Grünberg, et al., 2001, Grünberg, et al., 2004). Therefore, at the moment it is unclear if MagA is a direct participant in magnetite biomineralization or if it has a role in general cellular iron uptake, a step that might be a necessary precursor to magnetite biomineralization. While the role of MagA in magnetosome formation needs to be investigated further, recent experiments suggest that it could serve as a candidate for a genetically encoded contrast agent for MRI (Zurkiya, et al., 2008, Goldhawk, et al., 2009). When MagA is expressed in mammalian cells, nanometer-sized magnetic particles precipitate within the cytoplasm. The accumulation of iron is quite significant, allowing for detection with MRI (Zurkiya, et al., 2008, Goldhawk, et al., 2009). This is a clear example of how studies into the process of biomineralization in MB can directly lead to the development of biomedical applications.
Regardless of the mechanism and path of iron transport, the next step in biomineralization is the initiation of crystal formation. This is likely an involved process where the chemical environment of the magnetosome needs to be modified to allow for transformation of concentrated iron into nuclei of magnetite. However, biomineralization of magnetite is not as simple as concentrating iron within a vesicle. In contrast to some other iron oxides, such as the poorly crystalline ferrihydrite, which contain only the ferric form of iron, magnetite is composed of iron in both its ferric (the +3 oxidation) and ferrous (the +2 oxidation) states. How is such a mixed valence state accomplished when in most laboratory experiments the extracellular iron is provided in either the ferric or ferrous form? In the entirety of research on MB, only a handful of experiments have attempted to follow the transformations of iron after its transport into the cell. In one notable experiment, Frankel and colleagues used Mössbauer spectroscopy, a method that can distinguish between various forms of iron oxides, to monitor the progress of biomineralization in MS-1 (Frankel, et al., 1983). Their results indicate that a low-density hydrous ferric oxide (ferrihydrite) mineral and soluble ferrous iron are associated with the cellular fraction while a high-density ferrihydrite and a magnetite signal are found in the magnetosome fraction. These findings form the basis of a model in which soluble iron is first oxidized into a ferrihydrite precursor, transported into the magnetosome and then partially reduced to form magnetite (Frankel, et al., 1983). A similar experiment conducted with MSR-1 reached somewhat differing conclusions (Faivre, et al., 2007). In this case, no ferrihydrite signal is obtained during the biomineralization process whereas one consistent with an iron-containing Ferritin-like protein is detected in the cellular membrane fraction (Faivre, et al., 2007). The authors propose that the iron within this Ferritinlike protein is co-precipitated, along with soluble ferrous iron, to form magnetite crystallites at the cell membrane, which are then matured into magnetite within the magnetosome. While these studies vary significantly in their findings, they both suggest that precursors of magnetite exist in a mineralized state outside of the magnetosome. It is possible that these initial stages of biomineralization occur in the periplasmic space and the large precursor minerals are then transported to the magnetosome through the invaginations seen in ECT images. Alternatively, specific iron transporters may be involved in carrying various forms of iron through the periplasmic space to the cytoplasm and into the magnetosome lumen.
For greigite-containing MB, the transformations of iron to greigite are even less clear since no such organism is currently in culture. However, an intriguing study showed that the magnetosome chains of some greigite formers enriched from various environments contained mackinawite, a nonmagnetic iron sulfide mineral (Pósfai, et al., 1998). Remarkably, when individual particles of mackinawite within a single bacterium deposited on an EM grid were followed over the period of a few weeks they were seen to transform into greigite. The unnatural conditions in which these crystals were monitored have raised questions about the physiological relevance of the mineral transformations observed in this study. However, one possible support for this finding is that even during the earliest observation times, some of the particles within these organisms are a mix of both mackinawite and greigite (Pósfai, et al., 1998).
Even though the specific chemical transformations that lead to the biomineralization of magnetite are not well understood, recent studies have provided some insights into the possible molecular control of these early steps. These findings have mainly come from genetic analyses of mutants that are defective in steps that follow the formation of magnetosome membrane and protein sorting. As described above, deletion of mamM, mamN or mamO in AMB-1 results in completely nonmagnetic cells that are devoid of minerals, yet still produce magnetosome membranes with a full complement of magnetosome proteins (Murat, et al., 2010). Furthermore, transposon insertions within the mamO of MSR-1 lead to similar phenotypes. All three genes are found within the mamAB gene cluster and encode for proteins that are known to be associated with the magnetosome membrane. Like MamB, MamM is predicted to be a member of the CDF family of transporters, which may indicate a potential role in iron uptake. MamN has homology to Na+/H+ antiporters leading to speculation that its activity may result in an increase of the pH within the magnetosome, a condition that is required for in vitro magnetite synthesis. MamO, similar to MamE, is a predicted serine protease of the HtrA family, which may act to activate certain biomineralization enzymes or perhaps degrade inhibitory factors. Given their phenotype, it is likely that at least some of these genes are involved at the earliest steps of magnetite biomineralization including the nucleation of magnetite crystals. Surprisingly, however, no mutants have yet been discovered that “freeze” the biomineralization process such that a precursor mineral is formed instead of magnetite. Perhaps, the factors responsible for such a process have not yet been mutated or their loss is lethal to the cell. It is also possible that transformation of precursors into magnetite occurs sequentially in small batches such that the deletions of factors acting at such a step simply result in an “empty magnetosome” phenotype.
When they are small, magnetite crystals are composed of a single magnetic domain in a superparamagnetic state, meaning that the direction of their dipole moment is not stably maintained. However, once they reach a size greater than 35 nm, they are transformed into stable single domain magnetic crystals. Such a state is central to the magnetosome chain’s ability to reorient cells in magnetic fields. Once biomineralization is initiated the continued maturation of the magnetite crystal takes the mineral through these magnetic transitions that are crucial to its function for the organism.
Interestingly, the phenotypes of mutations in a number of magnetosome genes argue for genetic control of biomineralization during the crystal growth process. For instance, the deletion of mamS, mamT or mamR leads to a reduction in size of the crystal and defects in their morphology (Murat, et al., 2010). The phenotype of the mamS deletion mutant is particularly striking. In this strain many smaller particles appear to be clustered near each other in a manner suggesting the presence of multiple crystals within one magnetosome membrane. This may suggest a misregulation of nucleation or a block in a post-nucleation event, which may be needed for the formation of larger magnetite crystals. Unfortunately, the predicted homologies of these proteins do not allow for an obvious interpretation of their role in magnetite biomineralization. However, MamT and some other magnetosome membrane proteins contain possible heme-binding motifs. If these proteins are indeed heme-modified, they could be directly relevant in electron transfer events to oxidize or reduce mineral intermediates. Alternatively, they may perform a signaling function to ensure the proper progression of biomineralization.
One of the other possible heme-modified magnetosome proteins is MamE and recent work has implicated this protein in the maturation of magnetite crystals as well. As discussed above, when mamE is deleted there is a global defect in magnetosome protein localization, seemingly accounting for this mutant’s lack of biomineralization (Murat, et al., 2010). MamE is a member of the HtrA/DegP family of serine proteases and has a conserved catalytic triad of serine, histidine and aspartate residues. When these residues are mutated to alanines, AMB-1 cells are unable to orient in magnetic fields but can still form small crystals of magnetite (Quinlan, et al., 2011). In a biomineralization time course experiment, this mamE protease mutant is capable of initiating crystal formation at the same rate as the wildtype but growth of the mineral is stalled once an average crystal size of 20-30 nm is reached (Quinlan, et al., 2011). When MamE’s putative heme-binding sites are mutated, an intermediate phenotype is seen where a small number of crystals are able to progress past the 20 nm size and reach maturation (Quinlan, et al., 2011). Surprisingly however, both the MamE protease mutant and the MamE heme-binding mutant can properly localize proteins to magnetosomes. Thus, MamE has at least two separable functions, a protease-independent role in protein sorting and a protease-dependent role in maturation of magnetite crystals (Quinlan, et al., 2011). The targets of this protein’s protease activity could be inhibitors of biomineralization or activators that require proteolysis for their function.
Finally, a surprising candidate for a regulator of crystal maturation has come from the study of a presumed cytoskeletal protein (Ding, et al., 2010). FtsZ is a tubulin-like protein that is nearly ubiquitous in bacteria with an essential role in cytokinesis (Romberg & Levin, 2003). Interestingly, the mamXY gene cluster of the MAI encodes for a second copy of ftsZ termed ftsZ-like. Similar to FtsZ, FtsZ-like is capable of forming filaments in vitro in a GTP-dependent manner (Ding, et al., 2010). However, when ftsz-like is deleted in MSR-1, cell division is unaffected and instead a severe defect in biomineralization is seen (Ding, et al., 2010). In this mutant, crystals are significantly smaller than the wildtype and behave as superparamagnetic particles resulting in cells that cannot align with external magnetic fields (Ding, et al., 2010). The big mystery at the moment is how a cytoskeletal protein that is presumably localized to the cytoplasm would have such a major impact on biomineralization of magnetite. Perhaps, FtsZ-like is involved in the recruitment of some magnetosome proteins to the magnetosome. It may also play a role in the formation or shaping of the magnetosome membrane. If smaller magnetosomes are produced, then the resulting magnetite crystals may also be smaller. It is important to note that a similar deletion of ftsZ-like in AMB-1 does not result in any observable defects in biomineralization, yet again providing an example where the two organisms differ from each other (Murat, et al., 2010).
The sum of these genetic studies indicates that even beyond initiating crystallization, biomineralization depends on several punctuated steps of crystal growth. Given the putative activity of MamE, we have suggested that it may act as a monitoring system to allow for biomineralization to proceed at the appropriate time (Quinlan, et al., 2011). Such a checkpoint may help to balance the availability of iron with the need to build large crystals of magnetite. Accordingly, experiments have shown that environmental conditions can alter the average number of crystals produced by MB. Furthermore, the loss of some magnetosome genes, such as mamA and mamP, changes the number of crystals produced per cell while not affecting the formation of the magnetosome membrane (Komeili, et al., 2004, Murat, et al., 2010). MamP is yet another of the putative heme-containing magnetosome proteins with PDZ domains that may aid in its interaction with other proteins. Its deletion leads to the formation of very few crystals that are somewhat larger than those found in wildtype cells. Perhaps, the heme molecules bound to MamP participate in a signal transduction pathway to determine the number of crystals that should be formed in response to environmental conditions (Murat, et al., 2010). More speculatively, a MamE-dependent checkpoint may allow MB to choose the mode by which they navigate the environment. As stated above, large crystals of magnetite hold their dipole moment stably, which forces the bacterium to orient and navigate along external magnetic fields. In contrast, crystals of the size seen when the putative protease activity of MamE is disabled, are not large enough to force the bacterium into a magneto-aerotactic lifestyle. Perhaps, by controlling biomineralization through MamE, cells can be poised to build a magnetosome chain but still have the option of escaping geomagnetic fields. It is known, in the laboratory, that with constant microaerobic conditions or in the absence of high iron levels, magnetotactic bacteria can thrive without synthesizing magnetite crystals. Thus, it may be possible that in their natural environments, these bacteria can “choose” to exist in a magnetic or nonmagnetic state. To our knowledge, such a scenario has not been directly tested as most surveys of magnetotactic bacteria rely on magnetic separation techniques, which would exclude the potential pool of non-magnetic MB.
The different shapes seen in various species of MB imply that specific proteins, perhaps arranged into a defined matrix, control the final morphology of the mineral. In many biomineralization systems, factors responsible for precipitation and control of mineral size and shape are closely associated with the finished product and the same seems to hold for magnetite formation as well (Kröger, et al., 1999, Kröger, et al., 2000). When magnetosomes are purified from crude cell extracts and treated with a mild detergent solution, the lipid membrane and a majority of magnetosome-associated proteins are removed (Arakaki, et al., 2003). If the resulting “naked” magnetite fraction is then subjected to a harsher detergent and boiling treatment, a few polypeptides, termed the “Mms” proteins, are eluted from the still-intact crystal (Arakaki, et al., 2003). Dissolution of the magnetite crystal with an acid treatment reveals that no proteins are embedded within the mineral. Interestingly, the Mms proteins are encoded by the mms13/mamC, mms7/mamD, mms5/mamG and mms6 genes, which often appear as a coherent gene cluster in MB (Arakaki, et al., 2003). With the exception of MamC, they contain a conspicuous Glycine-Leucine repeat region distantly resembling repeats found in the silk fibroin proteins with a proposed role in the biomineralization of calcium-based minerals. In addition, recombinant Mms6 is capable of binding iron in vitro (Arakaki, et al., 2003). These observations suggested that the Mms proteins could have a direct role in influencing the physical properties of magnetite crystals of MB. To explore this question, recombinant Mms6 was added to an in vitro magnetite synthesis reaction. When produced in this manner, magnetite crystals show a broad heterogeneity in size and shape, a feature that is distinct from the uniform properties of magnetosome magnetite. However, when Mms6 is included in this reaction, a much narrower distribution of crystal shapes and sizes is observed (Arakaki, et al., 2003). Several variations on this initial study have confirmed a role for Mms6 in binding to iron and facilitating the production of magnetite crystals in vitro that more closely resemble those produced in MB (Amemiya, et al., 2007, Prozorov, et al., 2007, Prozorov, et al., 2007, Arakaki, et al., 2010). In addition, the acidic C-terminal portion of Mms6, which is the region that is responsible for iron binding, seems to be sufficient to promote its influence over the properties of magnetite (Amemiya, et al., 2007, Prozorov, et al., 2007, Prozorov, et al., 2007, Arakaki, et al., 2010). Finally, Mms6, or its C-terminal domain, can influence the in vitro properties of cobalt ferrite crystals, a mineral that is not produced by MB or other living systems (Prozorov, et al., 2007).
Combined together, these results clearly show that Mms6 is capable of controlling the morphology of magnetite crystals in vitro and a recent genetic analysis has provided compelling evidence for a similar in vivo function for this protein (Tanaka, et al., 2011). When mms6 is deleted in AMB-1, the resulting magnetite crystals are almost half the size of those formed in wildtype and are distinctly elongated. Furthermore, examination of individual crystals in more detail using high-resolution transmission electron microscopy (HRTEM) shows that certain crystallographic faces, unusual in the wildtype, are more common in this mutant (Tanaka, et al., 2011). Surprisingly, this mms6 deletion strain also has a significant reduction in the level of the “Mms” proteins that are normally tightly associated with the magnetite crystal. The authors attribute this to a potential interaction between the Mms proteins, in which case the presence of Mms6 is presumably required to organize a complex at the crystal surface (Tanaka, et al., 2011). However, it is also possible that Mms6 is required for expression, stability or magnetosome targeting of these Mms proteins.
The other members of the Mms proteins include MamC, MamD and MamG. In MSR-1 these three factors along with MamF are encoded by a presumed operon that is adjacent to the mms6 gene cluster and they account for approximately 35% of the protein content of the magnetosomes (Grünberg, et al., 2001). However, when either mamC or the entire mamGFDC cluster is deleted in MSR-1, only mild effects on biomineralization are observed (Scheffel, et al., 2008). For instance, the average size of the crystals is reduced by 25% and the ability of cells to orient in a magnetic field is reduced. The re-introduction of these genes into the ΔmamGFDC strain results in a dose dependent increase in the size of the crystals. In fact, the ectopic expression of mamC can even enhance the size of the crystals in wildtype cells. These experiments imply that these genes play a redundant role in promoting the biomineralization of magnetite (Scheffel, et al., 2008). Corroboration for such an idea has come from a mutant of AMB-1 where the 8 genes of the mamCDF and mms6 clusters are simultaneously deleted. The phenotype of this mutant strain is noticeably more severe than that of the mamGFDC or mms6 deletions described above (Murat, et al., 2010). This mutant forms small crystals with an elongated habit that are not sufficient to allow the majority of the cells to orient in magnetic fields. It is important to note that this mutant also contains deletions of genes not examined in previous studies. Thus, it is possible that its severe phenotype is due to the activity of one such gene and not the concomitant loss of the characterized mms genes.
These data strongly suggest that the Mms proteins play a central role in regulating the morphology of magnetite crystals. However, this class of proteins has only been found in the MB that belong to the alpha-proteobacteria (Nakazawa, et al., 2009, Jogler, et al., 2010, Abreu, et al., 2011). This is particularly surprising since species such as RS-1 and Mbav create crystals that are elongated and bullet-shaped which indicates the presence of an activity to selectively restrict the growth of certain faces of magnetite during biomineralization. Perhaps, novel factors that fulfill the same role as the Mms proteins are found in these diverse species of MB.
In conclusion, after nearly forty years magnetotactic bacteria and magnetosomes continue to be a rich and fascinating system for a diverse array of research disciplines. In recent years, dramatic advances in defining the molecular basis of magnetosome formation have shown that a core set of magnetosome genes, organized into a distinct genomic island, is conserved in diverse species of magnetotactic bacteria. The preliminary genetic, biochemical and cell biological analysis of these genes also suggests that the organelle is formed through a series of discrete steps (Figure 5). The initial step is the biogenesis of the magnetosome membrane through the action of MamI, MamL, MamQ and MamB. Next is the recruitment of MamE, which in a manner independent of its protease activity can localize a host of proteins to the magnetosome. These proteins can then direct the formation of the magnetosome chain and initiate the biomineralization of magnetite. Finally, the protease activity of MamE promotes the maturation of the crystal to its full size, a process that also requires a number of other factors including the Mms proteins. These insights set the stage for further defining the specific mechanisms by which these factors act every step of the magnetosome formation pathway. Despite the exciting discoveries made recently, a number of important questions remain unresolved in this field. Thus, to end this review I will enumerate some of these open questions with the hope of stimulating discussion and future experimentation:
I would like to thank the Komeili lab for their helpful discussions. In particular, I am grateful to Ertan Ozyamak and Dorotheé Murat for their critical reading of the manuscript. This work was in part supported by the NIH (NIGMS R01GM084122) and the David and Lucille Packard Foundation Fellowship in Science and Engineering.
The latest breakthroughs in understanding the molecular mechanisms of organelle formation and magnetic mineral formation by magnetotactic bacteria are presented in this article.