Team:Kyoto/Project/Magnetosome Formation


count down




Studies of synthetic biology produced various E. coli whose functions are multiply expanded. To utilize such E. coli in the real world, it is necessary to place them in the proper spot and have them stay there (Fig. 1). However, we have very limited technique to realize this. Solving this problem and utilizing transformed E. coli, we tried to create E. coli which have magnets inside their cells. Therefore, we focused on magnetotactic bacteria and introduced its gene to E. coli.

Fig. 1 Imagine if we can move E. coli as we want.

About magnetotactic bacteria

In 1975, a microbiologist Richard P. Blakemore first discovered a kind of magnetotactic bacteria, which is now known as Magnetospirillum magnetotacticum (MS-1).

He observed this bacterium under his microscope and realized that they move along the magnetic field lines of the Earth's magnetic field, and named this microorganism magnetotactic bacteria (or MTB)[1].

After this, many other species of MTB were discovered, such as Magnetospirillum magneticum (AMB-1), Magnetospirillum gryphiswaldense (MSR-1). And most of these MTB are able to thrive only in an oxic-anoxic transition zone [2].

The most impressive feature of them is the magnetosomes (Fig. 2).The magnetosomes are intracellular structures that consist of magnetic, iron-mineral crystals enveloped by a membrane vesicle known as the magnetosome membrane. The magnetosome crystals typically size from 35 nm to 120 nm. Inside the MTB, the magnetosomes are organized in one or more straight chains, parallel to the long axis of the cell, which function as the compass aid MTB to reach regions of optimal oxygen concentration [3].

Fig. 2 MTB has magnets inside their cell

In M. magneticum or M. gryphiswaldense, the hypothetical formation process of magnetosome can be roughly divided into three steps (Fig. 3). The first step is vesicle formation. The inner membrane of MTB swells out and makes a vesicle. The second step is iron uptake. Transporters in the magnetosome membrane pump in Fe2+/Fe3+, creating high iron ion concentration in the vesicle. The third step is bio-mineralization. Providing high iron ion concentration, magnetosome proteins crystalize iron ion, making magnetite crystal (Fe3O4) [4].

Fig. 3 How a magnetosome is formed in MTB

Gene cluster involved in magnetosome formation

A gene cluster called Magnetosome Island (MAI) is a group of genes related to magnetosome formation [5].

MAI is highly conserved among MTB and contains 4 operons — mamAB operon, mamGFDC operon, mamXY operon, mms6 operon. In a previous research published early this year, Kolinko et al. introduced these 4 operons from M. gryphiswaldense into photosynthetic bacteria Rhodospirillum rubrum, which is phylogenetically close to Magnetospirillum sp., yet does not make magnetosomes. They observed small vesicles formed in R. rubrum after these genes were introduced [6] (Fig. 4). They also demonstrated that the R. rubrum strain carrying these vesicles can be collected by a permanent magnet, showing that the small vesicles indeed function as magnetosomes.

Fig. 4 MAI plays a critical role in magnetosome formation

To unravel the function of genes in MAI, researches had been conducted by Dirk Schüler et al. through knocking out each one of them. In MTB, magnetosome membrane cannot be formed if either one of mamL, mamQ or mamB, which belongs to mamAB operon, is knocked out. Proteins of the three genes, though their function mechanisms are not so clear, are considered to be integrated with the inner membrane and triggers vesicle formation [4].

Magnetosome vesicle formation in E. coli

OUC-China 2013 also tried to magnetosome membrane formation in E. coli. They introduced mamL, mamQ, mamB, mamI and mamK to E. coli and indicated that introducing these genes affected the localization of MamC: GFP fusion protein in the cells. However, their constructs lack RBS in proper position, that is, they did not insert RBS 6 bases upstream of start codon. In addition, they did not have a mean to detect vesicles at high resolution. These difficulties hinder their experiments from the observation of the vesicle formation. Moreover, in 2014, new paper was published and mamL, mamQ and mamB were considered to be essential, and mamI and mamK were suggested to be inessential for the first step of magnetosome membrane formation.

Generally, E. coli do not have any internal structures enclosed by lipid membranes. Therefore, it is impossible for E. coli to make magnetosomes without forming lipid vesicles first. As we described above, in magnetosome formation of MTB, three proteins – MamL, MamQ and MamB, are considered to be integrated with the inner membrane and triggers vesicle formation. Here we started from magnetosome vesicle formation in E. coli by expressing these three proteins (Fig. 5). In the result below, we show successful reconstitution of the magnetosome-like vesicles in E. coli. The characterization of the induced vesicles and roles for MamL, MamQ and MamB are also discussed.

Fig. 5 mamL, Q and B are considered to relate to vesicle formation.

Experiments & Results

Cloning and Expression of mamL, mamQ and mamB.

We purchased M. magneticum AMB-1 genome DNA was purchased from ATCC. Using the genome, mamL, mamQ, and mamB genes were amplified and cloned. During the construction of the plasmids, the RBS sequence BB_B0034 RBS was inserted to the 6 nucleotide upstream of the start codon of each gene, following the instruction of iGEM standard parts. As shown in Fig. 6, histidine tag was fused to the C-terrminus of mamL and mamQ, respectively, for the purpose of the examination of the expression. These genes were fused and placed under the constitutive promoter BBa_J23100.

Fig. 6 Construction workflow
mamL, Q, B were cloned from genomic DNA of AMB-1 by PCR. After adding constitutive promoters and RBSs to amplified CDSs, 2 plasmid pLQB and pLQ were constructed.

Next, we checked whether our plasmid can produce M. magneticum proteins from the synthetic operon. Fig. 7 shows the detection of mamQ expression by immunoblotting. In this experiment, E. coli cells grown in rich media to their stationary phase were lysed and the total protein was separated by SDS-PAGE. Proteins were transferred to PVDF membrane and examined by HRP conjugated anti-His5 antibody. When mamQ is included in the synthetic operon, a 30 kDa band is clearly observed. This band is not visible when the plasmid lacks mamQ. These results indicate that at least one gene from the synthetic operon was produced in E. coli, demonstrating that both transcription and ribosome binding work normally on this construct.

Fig. 7 the detection of mamQ expression by immunoblotting
mamQ was detected between 25 kDa and 35 kDa.

We failed to detect mamL, a small hydrophobic protein with 85 amino acids. This might be attributed to some technical reasons, i.e., inefficiency of small protein capture on the membrane, or the difficulty of the solubility in the loading buffer. In any case, since mamQ protein was detected, we reasoned that the other two genes (mamL and mamB) are expressed as well.

Detection of vesicle formation by transmission electron microscope (TEM)

To check whether mamL, mamQ and mamB induce vesicle formation in E. coli, the plasmid pLQB (Fig. 6) or empty vector (pSB1C3 inserted only T7 promoter and RBS) introduced strains were grown to their stationary phase, and examined by the transmission electron microscope (TEM). In Fig. 8, 8 photos selected from each strain are presented.

Fig. 8 Magnetosome-like vesicle formation of pLQB and pLQ transformants
Plasmid pSB1C3 (inserted only T7promoter and RBS) was transformed into E. coli the negative control group (upper) and plasmid pLQB was transformed into E. coli the pLQB group (middle). Plasmid pLQ was transformed into E. coli the pLQ group(lower) also made magnetosome-like vesicles. Protein and lipid membrane was stained.

As shown clearly, the outer membrane of strains with pLQB plasmids tend to show winding structure, suggesting the effect of ectopic expression of membrane proteins on the membrane integrity. More interestingly, we frequently observed magnetosome-like lipid vesicles with a radius of about 100 nm in a pLQB-dependent manner. We counted the number of lipid vesicles per sections of E. coli. The observed vesicle number in pLQB transformant is significantly higher than negative control. The results are shown in Fig. 9. These results are consistent with the model that the expression of M. magneticum mamL, mamQ and mamB can induce the first step of magnetosome formation; the vesicle formation.

Fig. 9 Significant difference between control and the two transformants
The percentage of sections that contain magnetosome-like lipid vesicles calculated from the pictures from TEM.

While the pLQB transformant could make vesicles, it is unclear whether the three proteins are all necessary for vesicle formation in E. coli or not. Interestingly, E. coli has a protein homologous to mamB, fieF (ferrous iron effulux protein F, 22% identical, 40% similar by delta-blast). We hypothesized that mamB can be eliminated from the synthetic operon without changing the efficiency of the vesicle formation. For this purpose, we constructed plasmid pLQ (Fig. 6) and introduced it to E. coli. As expected, the pLQ transformant also induced lipid vesicles (Fig. 8).

Finally we examined the difference between negative control and pLQB or pLQ (Fig. 9). These results suggest pLQ as well as pLQB would induce vesicle formation in E. coli.

Hypothetical process of magnetosome-like vesicle formation in E. coli

Out of 2000 cells inspected, we found some of the candidates of immature vesicles in the pLQ transformant. Analyses of these samples led us to a model of the magnetosome-like vesicle formation process in E. coli. In our model, the inner membrane is first curved. Then gradually the space between the inner membrane and the outer membrane become larger. Finally the vesicle is formed (Fig. 10).

Fig. 10 process of magnetosome-like vesicle formation
The processes of magnetosome-like lipid vesicle formation in the pLQ transformant.
Fig. 11 Not all section had vesicles
Magnetosome-like vesicles were surely detected but not all section had them.

Efficiency of the vesicle formation by pLQ or pLQB was evaluated

Not all sections we observed had vesicles (Fig. 11). We then evaluated our parts.

We need to know the number of E. coli individuals which made magnetosome vesicles after transformation to estimate the efficiency of our parts. However, because we used TEM to observe E. coli, samples of E. coli were fixed by resin and sliced into 100 nm thick sections from different angles. We could not determine the percentage of vesicle-formed E. coli directly from the picture of slices. Therefore, we analyzed the pictures from TEM through reasonable calculation in order to determine the percentage.

Here we defined one value, vesicle-observation-rate (VOR). VOR is the average possibility that magnetosome vesicles were reflected in one section of E. coli (not one individual). Theoretical value of VOR can be calculated from the size of E. coli and the vesicles; we assumed 500 nm size magnetosome at maximum and the size of each E. coli transformant was calculated from the TEM image. We calculated parts efficiencies using theoretical and measured VOR. Specific formula and concept are described in "Material and Method" page.

We noticed that the transformants with pLQB or pLQ are larger than the control strain(Fig. 16). Their average size were 1.44 (pLQ) and 2.46 (pLQB) times larger than negative control (Fig. 13). Using these values, we calculated the theoretical VORs for pLQ (21.7%) and for pLQB (13.1%). These are the expected probability of vesicle apperance in a single section of E. coli when all cells harbor one vesicle.

When we measured actual VORs from the experimental data, the values were 9.1% (pLQ) and 10.0% (pLQB). Thus, the parts efficiencies, calculated by dividing measured VOR by that theoretical one, were 41.9% (pLQ) and 76.3% (pLQB), respectively. These values were obtained based on a model that assumes 500 nm magnetosome at maximum size of the vesicles, to avoid over-estimation of the parts efficiency (see Material and Method section for more detail). We next performed more precise estimation. When we assume the vesicle size at 120 nm (magnetosome of magnetotactic bacteria is 35 to 120 nm), the probabilities of the model are 9.9% (pLQ) and 6.0% (pLQB), so the efficiencies are 91.9% and 166.7%, respectively. These results together indicate that the efficiency of vesicle formation by pLQ and pLQB introduction is reasonably high. Most, if not all, of the transformants with these plasmids harbor at least one vesicle per cell.

Fig. 12 Our model of E. coli
We used this model to calculate theoretical VOR (vesicle observation rate).
Fig. 13 Measurements of the height and radius of transformants for the parts efficiency estimation.
We measured sections to make graph 1. and 2. To determine height, we measured their mean and extracted a part of samples which are larger than the average. Then, we took an average of the extracted samples. However, we determined radius from average of data.

E. coli growth is retarded when pLQ or pLQB are introduced

The transformants with pLQ or pLQB frequently show elongated form in TEM images (Fig. 16). We reasoned that this is due to the inefficient cytokinesis of these strains. To address this issue, we investigated the growth curve of these strains (Fig. 14). While the growth rate of pL (Fig .15) transformant was same as the control, that of pLQ and pLQB transformants were much lower than the control strain. These results are in good consistence with the model that pLQ or pLQB expression is partially harmful for the cell division (presumably at the cytokinesis).

Fig. 14 E. coli growth is retarded when pLQ or pLQB are introduced
The growth rate of pLQ and pLQB transformants were lower than pL transformants and control. Culture at 37 °C in LB medium.
Fig. 15 Construction of mamL
After adding constitutive promoters and RBSs to amplified CDS, plasmid pL were constructed.


Vesicle formation

We observed vesicle formation induced in pLQ and pLQB transformants. Only few vesicle-like structures were detected in negative control cells, suggesting that mamL and mamQ are the positive determinant that induce the magnetosome-like vesicles in E. coli.

In some previous studies, it was shown that E. coli can produce vesicle-like structures without specific machinery.[8] These vesicle-like structures are called "inclusion bodies". When misfolded or too abundant proteins are produced in E. coli, E. coli make inclusion bodies to depredate the harmful aggregation of the proteins. We carefully compared the inclusion bodies and our vesicles. Inclusion bodies are protein aggregates so they are found as black spots (200nm~ 500nm) when examined by TEM. These are distinct from what we observed in pLQB or pLQ transformants. All of the induced vesicles contain no electron-dense materials and we could not find any black spots whose size are 200nm~500nm.

So far, we have no further evidence than the TEM images for the induction of the magnetosome-like vesicle formation. Clearly, we need more detailed characterization of these vesicles, such as the subcellular fractionation and biochemical analyses of the vesicles. The localization of mamL, mamQ and E. coli membrane proteins by immuno-electron microscope technology will also reveal the detailed structure of the vesicles.

mamB is not essential for the magnetosome-like lipid vesicle formation in E. coli

In previous studies, mamB was considered to be an essential gene for the vesicle formation of magnetosome, since magnetosome was not formed when mamB was knocked out from magnetotactic bacteria. However, in our experiments presented here, only two factors, mamL and mamQ, were sufficient for the formation of the magntetosome-like vesicles in E. coli. As mentioned above, E. coli has a mamB homolog, fieF. This protein is a metal binding factor of the cation diffusion facilitators (CDF) involved in the ferrous iron efflux. It is reasonable to assume that this protein plays an equivalent role for mamB complexed with mamL and mamQ. No clear homologs for mamL and mamQ are found in the genome of E. coli.

It is still unclear whether mamL is really required for the vesicle formation in E. coli. As we showed in Fig. 7, we could not detect any expression of mamL by the immunoblotting experiments. While we believe that our synthetic operon produce all of the genes introduced, it is possible that mamL is not expressed from the plasmid. We have no clear result for the significance of mamL in our system. To clarify this point, we need another plasmid that expresses only mamQ.

Huge E. coli emerged in pLQ and pLQB transformants

As described earlier, we noticed most of the pLQ and pLQB transoformants are larger than the control strain(Fig. 16). They also show the winding outer membrane structures. We presume that these phenotypes can be attributed to the partial inhibition of cytokinesis. In the case of a mutant of FtsZ[9], a ring structure forming protein essential for cytokinesis, elongated cells are observed as well when the mutant strain is placed in the non-permissive temperature, showing that the proper alignment of FtsZ protein on the middle of the cell surface is indispensable for E. coli cytokinesis. Based on these information, we assume that the disruption of healthy membrane structure by ectopic expression of mamL and mamQ interferes with the alignment of FtsZ ring.

Fig. 16 The cells of pLQ and pLQB transformants were larger than control
pLQ and pLQB transformants' average size were 1.44 (pLQ) and 2.46 (pLQB) times larger than negative control.

The growth of the tansformants of pLQ and pLQB are lower than that of pL transformant and negative control. These results clearly show that the induction of vesicle formation is harmful for the cells. In the present study, we used a constitutive promoter and a high copy plasmid for the expression of the synthetic operon. We need to find the best condition for the vesicle formation by reducing the production rate of mamL and mamQ.


Through our experiments, we revealed that the ectopic expression of M. magneticum mamL and mamQ is sufficient to induce magnetosome-like vesicle formation in E. coli. We observed vesicles by TEM and showed that the efficiency of vesicle formation is nearly 100%.

Future work

In the present study, we successfully induced the magnetosome-like vesicles in E. coli. We will next try to fulfill these empty containers by magnetites. mamM gene is the next candidate to be introduced, as it works as iron transporter with mamB[10]. Addition of mamM to our synthetic operon will enable us to observe bioremediation of iron ions into the vesicles.

Introduction of the full of 4 operons (mms6, mamGFDC, mamAB, mamXY, 26-kb in total) is a technically challenging but intriguing experiment. Since our study revealed that the first step, vesicle formation, can be achieved by the expression of the subset of these genes, we believe that the total formation of functional magnetosome in E. coli is also feasible.

The induced vesicles shown in this study can be applied for other systems, too. Expression of other types of transporters will expand the use of these vesicles. If the specificity of the iron transporter is modified by mutation(s) properly, one might collect precious metals, rare metals, or toxic compounds in the vesicles. Using the magnetosome formation system in parallel, these technologies will provide us a novel tool to concentrate various compounds more efficiently from nature.


  • [1] Blakemore, Richard. "Magnetotactic bacteria." Science 190.4212 (1975): 377-379.
  • [2] Bazylinski, Dennis A. "Controlled biomineralization of magnetic minerals by magnetotactic bacteria." Chemical Geology 132.1 (1996): 191-198.
  • [3] Richard B. Frankel and Dennis A. Bazylinski Magnetosome Mysteries, ASM news (2004)
  • [4] Lohße A, Borg S, Raschdorf O, et al. Genetic dissection of the mamAB and mms6 operons reveals a gene set essential for magnetosome biogenesis in Magnetospirillum gryphiswaldense[J]. Journal of bacteriology, (2014): JB. 01716-14.
  • [5] Grünberg K, Wawer C, Tebo B M, et al. A large gene cluster encoding several magnetosome proteins is conserved in different species of magnetotactic bacteria[J]. Applied and environmental microbiology, (2001), 67(10): 4573-4582.
  • [6] Müller R, Zhang Y, Schüler D. Biosynthesis of magnetic nanostructures in a foreign organism by transfer of bacterial magnetosome gene clusters[J]. (2014).
  • [7] iGEM OUC-China 2013
  • [8]Richard R. Burgess, Refoluding solubilized Inclusion Body Proteins, Methods in Enzymology, Volume 463, (2009), p260
  • [9]Joe Lutkenhaus FtsZ ring in bacterial cytokinesis Molecular Microbiology Volume 9 Issue 3 August (1993) pages 403–409
  • [10]Dietrich H. Nies How iron is transported into magnetosomes Molecular Microbiology (2011) 82(4), 792–796