Engineering intracellular biomineralization and biosensing by a magnetic protein

Abstract

Remote measurement and manipulation of biological systems can be achieved using magnetic techniques, but a missing link is the availability of highly magnetic handles on cellular or molecular function. Here we address this need by using high-throughput genetic screening in yeast to select variants of the iron storage ferritin (Ft) that display enhanced iron accumulation under physiological conditions. Expression of Ft mutants selected from a library of 10(7) variants induces threefold greater cellular iron loading than mammalian heavy chain Ft, over fivefold higher contrast in magnetic resonance imaging, and robust retention on magnetic separation columns. Mechanistic studies of mutant Ft proteins indicate that improved magnetism arises in part from increased iron oxide nucleation efficiency. Molecular-level iron loading in engineered Ft enables detection of individual particles inside cells and facilitates creation of Ft-based intracellular magnetic devices. We demonstrate construction of a magnetic sensor actuated by gene expression in yeast.

Figure 1. Fluorescent reporter system used to probe intracellular iron mineralization by Ft.

(a) Schematic diagram of yeast cells containing an iron-responsive reporting system. Sequestration of cytosolic iron (red dots) into Ft (grey) triggers translocation of an iron-responsive transcription factor, AFT1 (orange), into the nucleus, where it induces transcription of an FTR1–GFP fusion protein (blue/green). Iron accumulation by effective Ft variants therefore results in a green signal (right). (b) Yeast cells transformed with empty vector (Vec), SPFt and the SPFt mutant E94G/K142R, which lacks iron mineralization capability, were incubated in minimum media overnight. Differences in FTR1–GFP expression are visible in the fluorescence micrographs at right, with SPFt but not E94G/K142R effective at upregulating the reporter; corresponding phase contrast images are shown at left. Scale bar, ∼10 μm. (c) FACS histograms showing the distribution of GFP-associated fluorescence observed in yeast cell populations transformed with vector, SPFt and E94G/K142R.

Figure 2. Selection of SPFt mutants by high-throughput genetic screening.

(a) Summary of the fluorescence-activated cell sorting (FACS)-based yeast genetic screening procedure. Control yeast cells lacking the FTR1–GFP reporter (neg) or positive cells harbouring the reporter and a SPFt gene library (Lib) were grown in minimum media. The yeast populations were presorted to remove debris and aggregated cells, and then used to establish a criterion (green outline) designed to reject cells lacking a functional reporter construct. From among Lib cells that passed this criterion, roughly 5% of cells which displayed the highest GFP fluorescence intensities (black label) were selected during each FACS run. Multiple rounds of selection and regrowth were performed (arrows) to enrich library mutants which induced the highest levels of fluorescent reporter expression. (b) A histogram showing the distribution of GFP fluorescence intensity in the yeast cell population transformed with the initial library (Lib, red), and following one to four successive rounds of enrichment (S1–S4). (c) Flow cytometry distributions of GFP fluorescence intensity of yeast cells transformed with SPFt (red) and three mutants identified through the screen, L55P (green), F57S (cyan) and F123S (magenta) incubated in minimal media overnight. Cytosolic iron content of intact yeast (d) and molecular-level iron loading by purified SPFt variants (e) was measured for each of the selected mutants using a bathophenanthrolinedisulfonate-binding assay following 16 h incubation of the corresponding cells in iron-rich medium. Error bars denote s.e.m. of three or more independent measurements. (f) Native gel analysis of purified SPFt and mutant nanoparticles stained with Coomassie blue for protein content (top) and Prussian blue for iron content (bottom), showing substantially increased iron content of the selected SPFt mutants.

Figure 3. Structural analysis of SPFt variants.

(a) X-ray crystal structure of PFt displaying the internal cavity of the protein in which one of the subunits is highlighted in yellow (left panel). Enlarged image of the highlighted subunit (right) shows the relative positions of sidechains mutated in the selected biomineralization mutants (L55P, F57S and F123S) with respect to the ferrooxidase residues (yellow) and the known iron binding sites (grey balls). (b) Cryo-EM images of purified SPFt, L55P, F57S, and F123S, showing formation of 12 nm spherically shaped nanoparticles in each case. SPFt samples also display differences in electron dense iron core formation, as indicated by the variable frequency of ‘empty' particles in the images (for example, yellow arrowheads). Scale bar, 50 nm. (c) The percentage of particles containing electron dense cores was computed by analysing 400 particles in cryo-EM images of each SPFt variant. All selected mutants displayed higher frequencies of core formation than the starting clone (t-test, P≤0.04), with L55P showing the greatest effect. Error bars denote s.e.m. of measurements from two independent samples.

Figure 4. Engineered SPFt mutants are effective hypermagnetic probes in yeast.

(a) Yeast cells transformed with empty vector (Vec), human heavy chain Ft (HFt), SPFt, L55P, F57S and F123S were pelleted and imaged in a 7 T MRI scanner. Relaxation rates (1/T₂) were computed from the MRI signal amplitudes. Inset, corresponding T₂-weighted spin echo MRI image of yeast cell pellets in microtiter wells (echo time=24 ms, repetition time=2,000 ms). (b) Isolation of yeast cells transformed with vector (blue), SPFt (red) and L55P (green) following application to a magnetic column. Cells were recovered during flow-through (FT), wash and elution phases of a magnetic cell separation protocol. Inset shows the percentage of cells retained until the elution phase, with L55P performing ∼4 times better than SPFt. Error bars denote s.e.m. of three independent measurements. a.u., arbitrary unit.

Figure 5. Detection of intracellular SPFt particles in ultrastructural cell images.

(a) Representative TEM image of a yeast cell following transfection with SPFt L55P and growth in overnight in medium containing 1 mM ferric citrate before sample preparation (left, scale bar, 500 nm). A closeup of the region identified by the dashed box at left region shows electron dense puncta such as those indicated by arrowheads (top right, scale bar, 100 nm). Similar puncta are not apparent in a comparable region from control cells (bottom right). (b) Automated analysis of TEM images from SPFt-expressing (n=10) or control cells (n=4) enables quantification of puncta that correlate with coefficient ≥0.9 to a Gaussian spot with full width at half height of 7 nm, comparable to the expected SPFt mineral core size. The difference in the density of puncta in SPFt L55P-expressing versus control cells is significant with t-test P=0.05.

Figure 6. Construction of a SPFt-based intracellular magnetic biosensor.

(a) Design of a genetically encoded magnetic biosensor for dynamic gene reporting using engineered SPFt. The sensor consists of two components: constitutively expressed SPFt L55P nanoparticles (grey) and streptavidin (SA) tetramers (orange) expressed from a galactose-inducible gene shown in the nucleus. When SA is upregulated, it crosslinks SPFt oligomers via their Strep-tag II moieties (blue), forming clusters that enhance T₂ relaxation rates and provide a means for magnetic detection of galactose-induced SA expression. (b) Representative cryo-EM images showing aggregate formation by purified SPFt L55P in the presence (below) but not the absence (above) of SA. Scale bar, 50 nm. (c) SA-dependent changes in the relaxation rate (1/T₂) displayed by solutions of 0.2 μM SPFt L55P holomers loaded with 520 μM Fe in the presence of increasing tetramer concentrations of an SA variant optimized for stability and Strep-tag II binding. Error bars show s.e.m. of three independent titrations, and the inset displays representative T₂-weighted MRI images corresponding to conditions shown in the graph. (d) Magnetic detection of SA-mediated intracellular clustering of SPFt. SPFt L55P and galactose-inducible SA were coexpressed in yeast as schematized in a. After 16 h in galactose-free medium (10 mM Fe), cells were transferred to 2% raffinose media without iron for 2 h. Then galactose was added to the culture at a final concentration of 2% to induce SA expression and cells were harvested at 0, 2 or 4.5 h for analysis. T₂ relaxation rates measured by MRI were normalized by cell pellet iron content and were compared with control experiments performed with PFt L55P (grey bars), which accumulates iron but is not crosslinked by SA. Normalized 1/T₂ values at 2 h and 4.5 h were significantly higher in the SPFt-expressing cells (Student's t-test P=0.02–0.04, n=3), consistent with the mechanism in a and the results of c. Error bars indicate s.e.m. of three independent experiments and corresponding protein levels revealed by Western blotting against SA and the Strep-tag II moiety of SPFt are presented above.