We constructed two ferritin chimeras by connecting the C terminus of one ferritin subunit to the N terminus of the other using a linker consisting of a FLAG epitope flanked by a short flexible region. A schematic of the two subunit configurations is shown in . The construct with H-linker-L orientation is designated H*L, and the opposite orientation (L-linker-H) is denoted L*H. Additionally, as benchmark positive controls, we produced individual H and L subunit transgenes that were modified to have a strong Kozak initiation sequence for improved translational efficiency, comparable with L*H and H*L chimeras. The transgene design contained a transcript devoid of an iron-responsive region. The recombinant ferritins and benchmark controls were expressed using AdV5-based gene delivery vectors in human osteosarcoma (U2OS) cells.
Fig. 1 Schematic of chimeric ferritin constructs. The diagrams show the configurations of the H*L (top) and L*H (bottom) gene sequences, including the promoter (CMV), the Kozak sequence, the first two codons of cDNA for reference to the reading frame, the subunit-joining (more ...)
To demonstrate the relative molecular size of the fused subunits, we used denaturing SDS-PAGE immunoblotting, as shown in . The L*H and H*L subunits have an apparent molecular weight of about 40 kDa, as expected, which is approximately double the weight of the L subunit used as a control. We next confirmed the proper polymer assembly of the L*H and H*L chimeric ferritins by comparing their migration rate to ferritin L and H homopolymers in native gels (). The comparable band sizes show that all of the ferritins were expressed at similar levels. The native western blot also shows supramolecular ferritin species, i.e., multimeric or aggregate forms of the assembled chimeric shells. The slower migrating, multimeric species of the L*H and H*L appear most similar to those of L (). In contrast, the native H shells resolve predominantly as a single native species ().
Fig. 2 Relative apparent sizes of native and denatured ferritins. The ferritin and control (lacZ) proteins were expressed in U2-OS cells. a Denaturing SDS-PAGE western blot stained for L subunit. The length of the L*H and H*L polypeptide chain is approximately (more ...)
We purified the chimeric ferritins expressed in U2OS cells and imaged the metal–amino acid complexes using TEM. As shown in , both L*H () and H*L () proteins assembled into spherical shells [21
], similar to the control ferritins (). shows the results of the TEM-based size estimates represented by the mean diameter of 30 randomly selected ferritin shells. The mean diameters were significantly different with p
< 0.05 using one-way ANOVA. Tukey's method for pair-wise difference showed that the L*H chimera was significantly larger than the control H and H*L with a confidence interval of 0.95.
Fig. 3 Morphology of chimeric ferritin shells visualized by TEM. In these micrographs, a is H ferritin, b is horse spleen ferritin, c is L*H, and d is H*L. The chimeric ferritin shells are similar in phenotype to the control. Micrographs of negatively stained (more ...)
Fig. 4 Molecular size of chimeric ferritins and controls. a Apparent molecular size estimated from the TEM images (average of n = 30 shells). Overall, L*H is slightly larger than the other ferritins. b Hydrodynamic size measured via DLS, where the range of the (more ...)
Additionally, we used DLS to measure the hydrodynamic sizes of the purified ferritin samples, as shown in . Overall, the DLS results are consistent with the engineered ferritins forming intact shells. There is a perceived upward size shift of the L*H curve on ; however, statistical analysis of the DLS data did not reveal significant size differences among the ferritins measured.
Since ferritin compartmentalization can affect the MRI contrast [22
] and possibly cytotoxicity, we next investigated the subcellular localization of the different ferritins in U2OS cells using confocal microscopy. As shown in , the chimeric proteins display a similar distribution to that of L, which is mostly cytoplasmic with low nuclear presence. In contrast, H displays a high degree of nuclear localization, in addition to presence in the cytoplasm (, white arrows). All controls, as well as cells expressing the chimeras stained separately for the L and H subunits, are included in the “Electronic supplementary material
” (Fig. S1
). In order to quantify the observed difference in subcellular distribution, we computed Manders colocalization coefficients [20
] for all ferritins. As expected, the H homopolymer had the highest nuclear localization, significantly higher than L and the engineered chimeras (). Overall, the Manders colocalization coefficients did not show any statistically significant difference between H*L and L or L*H ().
Fig. 5 Confocal microscopy of subcellular distributions of recombinant ferritins. a Distributions of the L subunit, H subunit, L*H and H*L. The nuclear localization of H ferritin is marked with white arrows. All other ferritins are cytoplasmic. U2OS cells were (more ...)
displays the in vitro cellular iron loading. We used one-way ANOVA followed by a Tukey's test to look into the group variations. We found a modest increase in iron loading in cells expressing H and both the separate H and L subunits. Interestingly, the L*H construct loads significantly more iron (p < 0.01) compared to cells co-expressing the H and L subunits. All constructs featured the same Kozak initiation sequence and had the same expression level as seen on the western blots. We used the same multiplicity of infection for all groups. In the case of co-transduction of H and L, there should be the same level of gene transcripts of the H and L subunits; however, we had no control over the final ratio at the stage of shell assembly.
Fig. 6 a Iron content of recombinant ferritin expressing cells. Cells (HEK293) were transduced with the transgenes LacZ, H, L, L*H and H*L. Following transduction, the cultures were incubated in low-serum media supplemented with 2 mg/ml purified holotransferrin. (more ...)
Also intriguing is the finding that the iron loading of the H*L construct appears to be no different than those of the control or single subunit expressing cells. Since colorimetric measurement of cellular iron could also include the labile iron pool (LIP) in addition to the ferritin shell content, we also looked into the amount of IRP2 expressed in cells. Iron-dependent proteasomal degradation of IRP2 makes this protein a very good estimator of the cytosolic pool of iron [23
]. shows the relative levels of IRP2 in control cells and all ferritin constructs. The L*H-expressing cells have the highest levels of IRP2, which suggests that the measured iron content () is not part of the LIP. As expected, the nontransduced cells and cells expressing control vector LacZ have very low levels of IRP2 (). Overall, these data suggest that the L*H has the greatest tendency to store iron among the ferritin constructs studied.
The NMR transverse relaxation rate (R2) of U2OS cell pellets expressing the different ferritin constructs are shown on . Overall, we found that L*H induces a significantly higher R2 relaxation rate in the cell pellets compared to the other constructs (p < 0.01). In addition, the relaxation rates of H and H*L are higher than that of the control (p < 0.05). We also measured the longitudinal relaxation rates (R1) of the same samples and saw no statistically significant differences (data not shown). We note that the iron-loading data () does not exactly follow the trend seen in the pellet R2 measurement. displays the bulk iron content of the cell, whereas NMR assays the indirect effect of the iron on the water relaxation rate, which may depend on where the iron is situated within the cell and its effective paramagnetic moment at that site (e.g., aggregated within ferritin cores vs. associated with other proteins or in LIP).
Fig. 7 NMR relaxation rates and MRI of transgene-expressing cell pellets. U2OS cells were transduced with the different ferritin constructs and controls, including LacZ, H, L, L*H and H*L. a Measurements of R2 of cell pellets at 500 MHz; results are the average (more ...)
The results of T2-weighted MRI of the above ferritin-transduced cell pellets are shown in ; as expected, the LacZ-negative control is the brightest and the L*H pellet is the darkest (i.e., maximum reporter contrast). It is worth noting that the NMR data () display the means of triplicate samples, while the MRI data show a single plane across one representative sample (we note that due to the inherent variation of the cell pellets, the pellet intensities observed in the single T2-weighted MRI slice shown do not exactly match the trend seen in the bulk, averaged NMR data for certain samples). Overall, there is an agreement among the iron-loading, NMR and MRI data in that the L*H appears to be the most efficacious reporter.
To ensure that the expression of the ferritin chimeras is not overtly cytotoxic to cells, we performed MTT proliferation assays [25
] on 293T cells expressing the different ferritins. Overall, we found no statistically significant difference in the proliferation levels among control LacZ-expressing cells and those expressing the L*H and H*L constructs at 48 h post-transduction (Fig. S2