Engineering Genetically-Encodable MRI Contrast Agents for in vivo Imaging
By Yuri Matsumoto
B.S. Chemistry Massachusetts Institute of Technology, 2006
SUBMITTED TO THE DEPARTMENT OF BIOLOGICAL ENGINEERING IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN BIOLOGICAL ENGINEERING AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY
FEBRUARY 2014 MASSACHUSETTS 1ThJN 0 Yuri Matsumoto, 2014. All rights reserved. OF TECHNOLOGY
The author hereby grants MIT permission to reproduce JUN 3 0 201 and to distribute publicly paper and electronic copies of this thesis document in whole or in part LIBRARIES i any medium now known or hereafter created.
Signature of Author: Signature redacted Department of Biological Engineering January 1, 2014
Certified by:_Signature redacted Alan P. Jasanoff Associate Professor of Biological Engineering Nuclear Science and Engineering, and Brain and Cognitive Sciences Thesis Supervisor
redacted CertifiedSignature Forest White Associate Professor of Biological Engineering Chairman, Graduate Program Committee
1 2 Thesis Committee
A ccepted by...... K . D ane W ittrup Professor of Chemical & Biological Engineering Chairman of Thesis Committee
A ccepted by...... A lan P. Jasanoff Associate Professor of Biological Engineering, Nuclear Science and Engineering, and Brain and Cognitive Sciences Thesis Supervisor
A ccepted by...... Angela B elcher W.M. Keck Professor of Energy in Materials Science and Engineering and Biological Engineering Thesis Committee Member
3 Abstract
Magnetic resonance imaging (MRI) is gaining recognition as a powerful tool in biological research, offering non-invasive access to anatomy and activity at high spatial and temporal resolution. However, the range of biological phenomena accessible to measurement by MRI is limited, due to a current lack of molecular-level methods for detecting physiological processes in living organisms. One way to overcome this limitation is to develop contrast agents that report physiological events at a molecular level. Traditionally MRI contrast agents have been based on small molecules that chelate paramagnetic ions such as Gd (III), but synthesis and delivery of such exogenously applied agents are complicated. Genetically-encodable MRI sensors may overcome some of these issues. In this thesis, we describe new class of MRI contrast agents which will be broadly applicable as genetically-controlled tools for in vivo imaging. The major goal of my thesis research was to improve the sensitivity of the existing protein-based MRI contrast agent, ferritin (Ft) by inducing it to accumulate larger number of iron atoms per particle in a physiological environment. Using a high throughput genetic screening process, we obtained
Ft mutants that show threefold greater cellular iron accumulation than mammalian heavy chain
Ft. In another project, we used the engineered Ft to develop a dynamic gene reporter that responds to changes in gene expression levels in vivo via aggregation-dependent MRI contrast changes. Successful creation of genetically-encodable MRI contrast agents that are robust and sensitive enough to be applied in vivo will enable neuroscientists and biologists to study molecular processes of living subjects.
Thesis Supervisor: Alan P. Jasanoff
Title: Associate Professor of Biological Engineering, Brain and Cognitive Sciences and Nuclear
Science and Engineering
4 Acknowledgements
I would like to thank my advisor, Alan Jasanoff for his continuous support while I pursued my doctoral degree. He has generously supported me financially and intellectually whenever I needed help with my projects. He respected my decisions and allowed me to explore many different ideas which helped me gain valuable experience as a scientist.
I would also like to thank my committee members, Dane Wittrup and Angela Belcher for their helpful insights and advice with my thesis work.
I am also grateful to my former and current colleagues in Jasanoff lab. In particular, I would like to thank Tatjana Atanasijevic and Gil Westmeyer for teaching me laboratory techniques when I first started working in the lab. I would like to thank Mariya Barch and Victor Lelyveld for helpful discussions with my projects.
I would like to thank various facilities at MIT (CSBI biosil cluster, MIT Center for Materials Science and Engineering, nanotechnology materials core facility, and flow cytometry core facility) which helped me execute experiments for my publications.
I am grateful to my funding sources, a Friends of the McGovern Institute Fellowship and a Siebel Scholar Fellowship for providing financial support in my 3 rd and 5th year of my Ph.D.
I am also extremely grateful to my parents in Japan who have raised me with patience while I explored the world with overwhelming curiosity. My mother in particular has been my best friend and mentor for as long as I can remember.
Finally, I would like to express the deepest appreciation to my husband, Nicholas Tham for being kind, patient and resourceful while I struggled through my thesis work. Besides being a responsible husband, he has been my technical support, chef, therapist and workout buddy. I would not have been able to complete my study without his support and for that I am eternally indebted to him.
5 Table of Contents
A b stract...... 4
A cknow ledgem ents...... 5
T able of C ontents...... 6
1. Introduction
1.1. Goals of molecular imaging ...... 9
1.2. Magnetic resonance imaging (MRI) as a molecular imaging tool ...... 9
1.3. Theoretical basis of MRI contrast agents: TI and T2 contrast agents...... 11
1.4. Advantages of protein-based MRI contrast agent...... 13
1.5. R elaxivity of Ft...... 14
1.6. Ft-based MRI gene reporter: Initial studies...... 15
1.7. Advantages and challenges of Ft-based MRI contrast agent...... 16
2. Engineering intracellular biomineralization to produce hypermagnetic genetically- encoded nanoparticles
2.1. A bstract ...... 18
2.2. Introduction, results and discussions ...... 19
2.3. M aterials and m ethods ...... 25
2.4. Acknowledgements ...... 31
2.5. Figure captions ...... 32
2 .6. F igures ...... 35
2.7. Supplem entary m aterial ...... 39
3. Clusters of genetically engineered hypermagnetic nanoparticles report dynamic changes with MRI
3.1. Abstract ...... 45
6 3.2. Introduction, results and discussions ...... 46
3.3. Materials and methods ...... 50
3.4. Acknowledgements ...... 55
3.5. F igure caption s ...... 55
3 .6 . F ig u res ...... 5 8
3.7. Supplementary Material ...... 61
4. T2 relaxation induced by clusters of superparamagnetic nanoparticles: Monte Carlo simulations.
4 .1. Ab stract ...... 64
4.2. Introduction ...... 65
4 .3. M ethods ...... 67
4.4. Results and discussion ...... 69
4.5. C onclusions ...... 74
4.6. A cknow ledgem ents ...... 75
5. Conclusions and future directions
5.1. SPFt-L55P nanoparticles and the genetic screen...... 76
5.2. Ft-based dynamic gene reporter ...... 75
Appendix A: Metalloprotein-based MRI probes (review article) ...... 79
Appendix B: Towards finding mutant ferritins with higher magnetic moment by a high gradient m agnetic cell sorting screen...... 106
Appendix C: A novel protein-based kinase activity sensor for MRI...... 114
R eferen ces ...... 12 9
7 8 1. Introduction
1.1 Goals of molecular imaging
Historically molecular imaging has focused on identifying tissue abnormalities in patients by locating radiolabeled tracers whose distributions in the living subjects are perturbed by non- specific macroscopic physical, physiological or metabolic changes'. More recently, the goal of molecular imaging has shifted to imaging specific proteins, genes or biochemical molecules within intact living subjects for a system-level understanding of biology, earlier detection of diseases, and evaluation of therapeutic approaches The ultimate goal is to idcntify where specific molecules are present in the body, the level at which they are present and how the distribution and level change over time or following an intervention. This shift in emphasis was made possible by advances in development of specific molecular imaging agents tailored to specific imaging modalities such as positron emission tomography (PET), optical imaging, and magnetic resonance imaging (MRI). Molecular imaging agents can be made specific and functional by linking the molecular probe that binds to the target molecules to the contrast generating moiety whose contrast is triggered only after the molecular interactions of the agent and the target take place. These agents provide us a quantitative understanding of cellular and molecular processes such as gene expression, protein expression, trafficking, and protein-protein interactions in the physiologically relevant context 1. Moreover, due to their ability to non- invasively track molecular events in a living subject over time, they are particularly useful in developing drugs and assessing disease progression.
1.2 MRI: a molecular imaging tool
9 In proton MRI2, the form of MRI most commonly applied in laboratories and clinics, populations of nuclear spins arising from hydrogen nuclei primarily in water are perturbed and monitored to generate images. At thermal equilibrium in a strong magnetic field (Bo), the water proton spins align weakly with the applied field and give rise to a net "longitudinal" magnetization parallel to BO. This magnetization is unobservable, but can be detected following application of radiofrequency energy pulses which tilt the magnetization vector off of the Bo axis and give rise to a nonzero "transverse" magnetization component. After excitation, the transverse magnetization component decays away with a time constant T2 (the transverse relaxation time) and the overall magnetization returns to thermal equilibrium with a time constant T (the longitudinal relaxation time). The shorter T is, the more frequently an MRI signal can be repeatedly measured per unit time; areas of a specimen with short T therefore give rise to a larger average MRI signal. Conversely, the shorter T2 is, the more rapidly the observable component of magnetization disappears, and the lower the MRI signal becomes.
MRI has distinct advantages over other in vivo molecular imaging techniques such as PET and optical imaging. The spatial resolution of MRI can be as good as 25ptmI, whereas that of PET and noninvasive forms of optical imaging is of the order of mm. Moreover, unlike most optical imaging techniques, MRI has excellent tissue penetration, which makes it suitable for imaging intact living animals. Typically the temporal resolution of functional MRI is on the order of seconds, which is slightly lower than optical imaging (~milliseconds) 3. The major limitation of
MRI, however, is its low sensitivity due to its inherently low signal to noise ratio (SNR), several orders of magnitude lower than PET and optical imaging. The sensitivity of PET is in the range of 10~" to 10-2 M, whereas optical imaging achieves 10-5 to 10-17 M (bioluminescence
10 imaging) and 10-9 to 10-12 M (fluorescence imaging)1 . One way to improve the sensitivity of
MRI is to develop contrast agents with high sensitivity, which is the main theme of the thesis.
1.3 Theoretical basis of MRI contrast agents: Ti and T2 contrast agents
Paramagnetic species that influence contrast in MRI by reducing the T1 and T 2 relaxation times are called contrast agents4' 5. Accelerated relaxation arises from coupling between the magnetic dipole of the contrast agent and the nuclear spins of water molecules that interact with the agent through bonds ("inner sphere" interactions) or through space ("outer sphere").
Relaxation rates R 1 (= 1/TI) and R 2 (= 1/IT2) are generally linear with contrast agent concentration. The slopes of these relationships are referred to as the T, and T 2 relaxivities, r,
and r 2, respectively, which measure the strength of the contrast agent and are expressed in units
of mM- s-1. Greater relaxivity is beneficial to an MRI contrast agent, because the agent can then
be applied at lower doses or to greater effect at any given concentration. Both r1 and r 2 vary
strongly with the magnetic field strength, B 0, and depend on physical parameters including the
electron spin number (S) or magnetic moment of the contrast agent, the number of coordinated
inner sphere water molecules (q), the time constant for inner sphere water exchange (rM), and the
rotational correlation time of the agent (TR). Inner sphere contributions to relaxivity are described
by the theory of Solomon, Bloembergen, and Morgan 6-8 , and apply to metalloproteins with
adjustments to account for slow rotation in macromolecules9 . Outer sphere contributions are
described by related theories applicable to mononuclear 0 and particulate" 13 metal complexes.
Determinants are thoroughly discussed in a number of secondary references14' 15. Relaxivity
determinants are important not only because they explain how contrast agents may be optimized,
but also because they provide potential mechanisms for designing MRI-detectable sensors.
11 Most MRI contrast agents or sensors tend to affect TI-weighted MRI scans more than T2-
weighted scans, or vice versa, and are correspondingly referred to as T1 or T2 agents. T1 agents
generally have an rj/r2 ratio of about 1:2 and contain one or a small number of paramagnetic
ions. Classical T1 agents are exemplified by complexes of Gd3+ with small chelators like
diethylenetriaminepenaacetic acid (DTPA) 6 ' 17 and are the most commonly applied agents in
clinical MRI; analogous proteins can include porphyrin prosthesis or directly bound metal ions.
At typical field strengths for clinical or preclinical MRI (> 1 T), most T, agents have values from
1 to 10 mM's-1. In biological samples with T, values typically near 1 s, T, agents need to be applied at concentrations near 100 pM to induce substantial contrast effects.
T2 agents have r2/r] ratio greater than ~10 (a necessary condition because T2 values are much shorter than Ti values in vivo) and are best exemplified by superparamagnetic nanoparticles
1-20 (SPNs)- , contrast agents that incorporate discrete crystalline domains that exhibit highly cooperative magnetic behavior. A biosynthetic analog to SPNs is ferritin (Ft) 13, an iron storage protein that accumulates minerals in a 12 nm shell like structure formed from 24 polypeptide chains. SPN T2 agents produce high r2/rj ratio because they become magnetized by the BO field and create microscopic magnetic perturbations experienced by diffusing water molecules in solution. These perturbations affect T2 relaxation more than T1, particularly for particles with highly magnetic mineral cores over -3 nm in diameter and at high Bo strengths (>1 T), where
SPN magnetizations tend toward an asymptotic "saturation" point 21 . The relaxivity of T2 agents strongly depend on R2/D, where R is the mineral core radius and D is the solvent self-diffusion constant; SPNs with larger size shorten T2 more effectively, up to a so-called static dephasing limit 2 near ~50nm for typical SPNs. Most SPN contrast agents incorporate iron oxide; synthetic iron-containing SPNs usually have r2 values of 50-500 mM Fe~1 s-1. Given typical background T2
12 values near 100 ms in tissue, synthetic SPNs applied at concentrations of 1-10 pM Fe can produce substantial effects under optimal T2-weighted imaging conditions.
1.4 Advantages of protein-based MRI contrast agents
Although the majority of MRI contrast agents have been based on small organic
1 molecules such as Gd3+ chelates 23,24 and synthetic nanoparticls (SPNs) 8,20, metalloprotein MRI contrast agents present a variety of advantages. Thanks to the advancement of molecular biology techniques, proteins are much easier to synthesize and modify than organic molecules; numerous protein engineering techniques may be applied to tune metalloprotein properties for MRI
(reviewed in 25). Due to their larger size, protein-based contrast agents are retained in the blood pool for a longer time than small molecule agents, allowing for longer imaging times in some types of experiments26,27 . Another advantage of protein contrast agents with respect to small molecules is in the area of biomolecular target detection, the objective of molecular imaging. The large surface area and numerous functional groups presented by protein interfaces render these macromolecules especially suited to binding potential ligands with high affinity and specificity.
Although naturally occurring proteins may not contain both desirable paramagnetic ions and target specificity, their amenability to engineering approaches makes them an attractive choice
for developing specific molecular imaging agents. Furthermore, most cytosolic or secreted metalloproteins are evolved to remain in solution and interact with well-defined ligands; this
limits the potential for biological environments to substantially degrade relaxivity or analyte
sensing capabilities by adversely affecting water proton interaction parameters. Lastly, in some
cases, protein contrast agents can indeed be targeted and expressed in vivo using gene delivery
methods, a la GFP. Such reporters generate relatively static contrast, but can be used to non-
13 invasively monitor gene expression level and allow particular types of cells to be tracked in vivo over time. The advantages and limitations for such reporter gene will be discussed in details in the following sections.
1.5 Relaxivity of Ft
Ft is an iron-storage protein ubiquitously found in all living organisms except for fungi28
Canonincal Ft homologs consist of 24 peptides that are self-assemble into a spherical shell and the iron mineral core. Mammalian Ft has been reported to bind up to -4500 Fe atoms per 24- mer 29, providing a stoichiometric advantage of over a hundred, compared with heme proteins, in terms of sheer iron accumulation. Due to its similarity to SPNs, Ft is an obvious platform for engineering genetically-encodable T2 MRI contrast agents. While chemically synthesized SPNs are made up of a strongly magnetic iron oxide called maghemite, Ft contains hydrated iron oxide crystals called ferrihydrite. Attempts to model the relaxation of ferritin with the transverse relaxation model of outer-sphere mechanism, which successfully modeled maghemite nanoparticle relaxation, failed to describe the linearity of ferritin relaxivity with respect to the
13 applied magnetic field , 14. A proton exchange dephasing model (PEDM) was proposed by
Gossuin et al. to address the linear dependence of ferritin relaxivity to the external magnetic
3 field, B 0 . The PEDM relaxation is primarily due to the protons jumping between the adsorption sites on the surface of ferrihydrite core while exchanging with the bulk water. Protons adsorbed on different sites of the particle surface experience slightly different magnetic fields, which are characterized by a Lorentzian distribution. The model assumes that the diffusion correlation time of protons in the solvent, D, is small compared to the time it spends on adsorption sites, M.
14 Under such conditions, the transverse relaxivity of ferritin is expressed as the product of the number of exchangeable protons, q, a Lorentzian constant, k, and the external magnetic field, Bo.
q r 2 = k BO
It is important to note that since the transverse relaxivity of ferritin is proportional to the applied magnetic field, imaging at higher magnetic field significantly improves the sensitivity of the technique. In contrast, SPNs, commonly used synthetic T2 contrast agents have magnetization that saturates at around 1 T, thus imaging at higher field than 1 T does not provide an added advantage. Linear increase of r2 with the magnetic field for a ferritin in aqueous solution has been experimentally verified with field up to 11.7 T3 1. Furthermore, the relaxivity of ferritin per unit concentration of iron is primarily dependent on the absolute concentration of iron in the ferritin solution and not on the size of ferrihydrite core if the number of iron atoms per ferritin is above 502.
1.6 Ft-based MRI gene reporter: Initial studies
Visualizing gene expression level at a cellular resolution in vivo requires a robust gene reporter which translates the target gene expression level into a signal detectable by MRI. In the case of optical imaging, a well-characterized gene reporter such as green fluorescent protein has been widely available, but this is not the case for MRI. An early effort to express myoglobin in transgenic mice did not produce substantial MRI contrast changes33 , but some contrast changes have been achieved by overexpressing Ft in cells and animals. The effect of overexpressing Ft was first demonstrated in C6 glinoma cells transfected with murine heavy chain Ft (one of two isoforms, heavy and light, expressed in mammals) 34. In the same year, Genove et al. showed for the first time that ectopic adenovirus-driven Ft overexpression leads to detectable T2 contrast in
15 rodent brains, allowing them to non-invasively monitor the gene expression with MR135 . The use of Ft as a contrast agent has also been explored by Deans et al., who showed enhanced T2 contrast in mouse neural stem cell line by coexpressing human H-chain Ft and human transferrin receptor 36. Later papers reported T2 effects of Ft overexpression in transgenic mice37 and transplanted cells in vivo38 4 .
1.7 Advantages and challenges of Ft-based MRI contrast agent
Ft is a great platform for engineering MRI contrast agents because of its similarity to
SPNs and has several advantages over the synthetic SPN contrast agents. Practical advantages of
Ft compared with synthetic SPNs include its regular structure and predictable size, both of which can aid in characterization or engineering the relaxivity of Ft-based contrast agents. Another advantage of Ft over SPN contrast agents is that Ft is fully genetically-encodable, and therefore can be delivered to intracellular locations using preexisting gene delivery techniques whereas
SPN contrast agents have to rely on non-specific transport across the plasma membrane.
Intracellularly localized contrast agents may be able to report gene expression levels and sense the levels of important intracellular molecular messengers such as calcium ions. Due to its genetically-encodable nature, Ft can be produced by cells locally and the Ft transgenes can be inherited by the daughter cells whereas exogenously applied agents would be diluted by spontaneous degradation as well as cells divisions. For this reason, the contrast due to Ft persists for a longer period than exogenously applied agents and therefore Ft is more suitable for longitudinal studies where, for example, implanted cells are tracked for over many months.
Finally, by using cell-type specific promoters, Ft based contrast agents allow us to image specific types of cells in vivo, which is virtually impossible with SPNs.
16 One of the major limitations of Ft as an MRI contrast agent for in vivo imaging is its low relaxivity, resulting in modest contrast changes at a typical magnetic field strength. Saturation magnetization of Ft is only 0.9-1.2 emu/g 41, which is significantly lower than that of SPN contrast agents (60-100 emu/g )42 . Because the expected T2 effects of iron oxide cores are proportional to their magnetization, the relaxivity of Ft at saturating B0 is in principle only about
1% that of synthetic iron oxides for equivalent core sizes. This means that, we need 100 times more Ft particles to produce the same amount of T2 contrast changes generated by similarly sized
SPNs. Another limitation is that the endogenous iron level has to be sufficiently high so that the cells can take up enough iron to load iron into the overexpressed Ft as well as to sustain other cellular activities involving iron such as respiration. Finally, the timescale of Ft expression and iron loading may be too slow to capture the rapidly changing gene expression levels. The major focus of my thesis is to improve the sensitivity of Ft-based MRI contrast agents (Chapter 2).
Another aspect of my thesis addresses the kinetics of the contrast agents by creating a MRI gene reporter which responds to dynamic changes in gene expression levels via aggregation mechanism (Chapter 3). Lastly, a theoretical study was conducted to systematically predict the
T2 effect of aggregated superparamagentic nanoparticles as a function of cluster size, nanoparticle size, interparticle distance and cluster geometry (Chapter 4).
17 2. Engineering intracellular biomineralization to produce hypermagnetic genetically-encoded nanoparticles
2.1. Abstract
Noninvasive 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 engineering "hypermagnetic" forms of the iron storage protein ferritin (Ft), which can act as genetically encoded iron oxide nanoparticles suitable for production inside cells. We developed a high throughput genetic screening approach in yeast to substantially improve the magnetic properties of spontaneously biomineralized Ft nanoparticles under physiological conditions. The screen was applied to a library of 107 variants of a thermostable Ft from Pyrococcusfuriosus,fused to an affinity tag that facilitates detection, purification, and characterization of the mutant proteins. Engineered Ft nanoparticles show threefold greater cellular iron accumulation than mammalian heavy chain Ft, as well as over fivefold higher contrast in magnetic resonance imaging and robust retention on magnetic separation columns. Mechanistic studies indicate that improved Ft magnetism arises in part from increased iron oxide nucleation efficiency, which suggests strategies for further engineering of intracellular protein nanoparticle biomineralization in diverse contexts.
18 2.2. Introduction, results and discussions
Magnetic approaches to biological experimentation are particularly promising because they interact minimally with biological processes, do not involve radiation, and have already led to powerful imaging and manipulation techniques. Existing magnetic biotechnologies are of limited value for studying molecular and cellular level phenomena, however. The best known magnetic measurement technique, magnetic resonance imaging (MRI), detects a complex mixture of tissue properties which relate only indirectly to underlying molecular and cellular phenomena.
Molecular MRI measurements can be made using contrast agents that combine magnetic properties with other functionalities 43-46, but these agents need to be delivered exogenously.
Techniques for magnetic modulation of biological systems have been demonstrated at cellular level47-49, but also tend to depend on exogenous nanoparticles that are difficult to apply in vivo.
Although manipulation of cellular magnetism and magnetic image signals has also been
34 demonstrated using genetic techniques , 35, 50-53, the effects tend to be weaker or less specific than approaches based on synthetic magnetic nanoparticles.
A strongly magnetic protein could provide a basis for robust modulation or detection of well-defined molecular-level phenomena. A promising starting point for generation of such a
54 molecule is ferritin (Ft), an iron storage protein found in most animal, plant, and bacterial cells .
Ft proteins consist of a spherical shell of 24 identical or closely homologous polypeptide chains, in which a reservoir of hydrated iron oxide accumulates and can be rapidly mobilized according to physiological needs. Ft variants have been used as magnetic gene reporters and components of
magnetically-responsive genetic devices3 4' 35' 55, but Ft is far less potent than synthetic nanoparticles of similar volume and often contains far fewer iron atoms than its core structure
could in principle accommodate28 . In vitro manipulation of Ft mineralization has enabled the
19 generation of highly magnetic species5 6 , but the resulting protein complexes cannot be applied in conjunction with genetic techniques and suffer similar limitations to those of synthetic nanoparticles. We therefore designed a system that would allow us to specifically enhance the magnetic properties of intracellularly expressed Ft in a systematic and high-throughput fashion.
Our protein engineering approach was based on the hypothesis that mutant Ft molecules that sequester iron compounds most effectively would display optimal magnetic properties-a view motivated by the fact that both greater Ft iron content5 7 and denser, unhydrated iron oxide mineralization56 can result in higher per-particle magnetic moments. Iron accumulation by Ft variants is expected to reduce cytosolic iron concentration by mass action principles, so we established a reporting system in yeast whereby expression of Ft mutants could be evaluated for induction of a low cytosolic iron phenotype. In Saccharomyces cerevisiae, intracellular iron level is regulated by the iron responsive transcriptional activator Afti, which under low iron conditions translocates into the nucleus and regulates genes involved in iron uptake58. One of the genes upregulated by Aftl encodes the cell surface high-affinity iron transporter, FTR159; by monitoring expression of an FTR1-green fluorescent protein (GFP) fusion reporter 60, we could therefore identify individual cells that display low cytosolic iron concentrations (Fig. la). This system was intended as a tool for selecting mutant Ft variants that robustly sequester cellular iron, and that would therefore induce greater FTR1-GFP expression and fluorescence than Ft variants with less potent iron binding capacity.
As a template for random mutagenesis and screening, we choose to work with a Ft from the thermophilic bacterium Pyrococcus furiosus (PFt). PFt has the advantage that it is highly thermostable (Tm > 120 C) 6 1, and therefore likely to be more tolerant to mutations introduced to alter biomineralization than human heavy chain Ft (HFt) (Tm = -77 oC) 62, which has been used
20 for the majority of biotechnological applications of Ft in the past. In addition, PFt forms homooligomeric nanoparticles which require only a single polypeptide, in contrast to mammalian
Fts that incorporate two chains, making PFt structure and chemistry simpler and more predictable. To facilitate isolation and analysis of PFt variants, we fused an affinity tag (Strep-tag
II) to the N-terminus of PFt to form a construct abbreviated SPFt (Supplementary Fig. S Ia). The tag had minimal effect on protein folding and iron loading functions of the protein
(Supplementary Figs. Slb-e). SPFt was expressed in yeast cells bearing the FTR1-GFP reporter and induced elevated fluorescence, compared with control cells bearing no SPFt or harboring a compromised SPFt with E94G and K142R substitutions that eliminate ferroxidase activity of the protein (Fig. lb). Results of fluorescence microscopy were further validated by fluorescence- activated cell sorting (FACS) analysis (Fig. 1c), and were consistent with the explanation that
SPFt expression sequesters cytosolic iron and boosts FTR1-GFP reporter expression.
In order to isolate mutants that preferentially accumulate more iron in vivo, we subjected the entire PFt coding sequence in SPFt to polymerase chain reaction (PCR)-based random mutagenesis. After transfection, this resulted in a library of 10 million yeast clones expressing randomly mutated SPFt variants with an average mutation rate of 1 nucleotide changes per gene
(Supplementary Fig. S2). This relatively low mutation rate was chosen in order to avoid accumulation of deleterious mutations which could obscure beneficial but rare mutations. The yeast library was incubated in a minimum media and sorted by FACS to obtain cells exhibiting highest levels of FTRl-GFP fluorescence. Cells in the top 5% were propagated for a subsequent round of sorting (Fig. 2a), and the procedure was repeated. After four rounds (Fig. 2b), we
sequenced the sorted population and identified three mutations that were enriched among the
selected yeast cells: L55P, F57S, and F123S.
21 To confirm the Ft dependence of iron reporter expression in the selected clones, plasmids
for SPFt L55P, F57S, and F123S were isolated and retransformed for reanalysis by FACS;
fluorescence histograms were consistent with the screening results (Fig. 2c). As an additional test
of the iron accumulation phenotype, we incubated the three selected clones in iron-supplemented
media and measured the total cellular iron content (Fig. 2d) and iron content of purified SPFt proteins (Fig. 2e and f). The most effective of the SPFt mutants, L55P, induced 1.6 + 0.2 times
greater cellular iron accumulation than wild-type SPFt and 2.6 ± 0.3 times greater accumulation than HFt. Compared with SPFt, the L55P mutant also exhibited almost double the number of
iron atoms per Ft 24-mer, indicating that the cellular iron loading phenotype originates largely
from an increase in iron sequestration by Ft at the molecular level. For both L55P and F57S mutants, significant enhancement of cellular iron accumulation (Student's t-test, p = 0.002, n = 6
for L55P and p = 0.003, n = 6 for F57S) and molecular-level Ft iron loading (p = 0.00003, n = 6
for L55P and p = 0.02, n = 4 for F57S) were observed. These results prove for the first time that intracellular Ft biomineralization processes can be engineered to produce substantial gains in iron accumulation by individual protein macromolecules.
In an attempt to understand the mechanism by which primary sequence mutations in SPFt lead to enhanced iron accumulation in Ft holomers, we performed a series of characterization experiments. By inspecting the crystal structure of PFt63 , we saw that all three mutant residues point toward the inside of the iron storage cavity and lie on the B and D helices close to a site thought to be involved in oxidation of Fe2 ions that enter the PFt core (Fig. 3a). We speculated that the mutations might therefore affect either the enzymatic functionality of PFt or the structure of the iron oxide core itself. To test these ideas, we began by measuring the iron loading and release kinetics of the SPFt variants; no significant differences in uptake or release rates were
22 found (Supplementary Table S1). To examine potential structural effects of the mutations, we characterized the purified protein nanoparticles by high-resolution cryo-electron microscopy
(cryo-EM), a powerful technique that allows imaging of proteins in the near-native environment.
Micrographs confirmed that SPFt and the variants all form 12 nm cage-like structures as expected (Fig. 3b). There was however a striking variation in the prevalence of electron dense cores discernible among the four SPFt variants. Only 68.3 1.3 % of wild-type SPFt nanoparticles contained dark core structures, whereas 96.1 ± 0.1 %, 87.0 ± 0.3 %, and 78.3 ± 1.5
% of the L55P, F57S, and F123S mutants, respectively, appeared electron dense (Fig. 3c).
Increased core formation in each mutant was significant with respect to SPFt (p = 0.03 for L55P, p = 0.04 for F57S, p = 0.04 for F 123S; n = 2 samples with 400 particles/sample), suggesting that an increased ability of the mutant proteins to nucleate mineral core formation might largely account for their ability to accumulate a larger number of iron atoms per protein molecule.
Our strategy for engineering hypermagnetic SPFt variants was predicated on the notion that iron sequestration by SPFt mutants would accompany enhanced magnetic properties. To demonstrate this, we explored the utility of hypermagnetic SPFt variants in imaging and high gradient magnetic cell separation (HGMS) applications. For MRI experiments, the same yeast samples used for the iron assays in Fig. 2d were pelleted and imaged in a 7 T magnet using a spin-echo acquisition sequence. The transverse relaxation rate (lIT 2) of cells transformed with the most iron-rich Ft mutant, L55P, was significantly higher than that of cells expressing wild type SPFt (58.2 ± 3.7 s 1 vs. 30.0 ± 2.5 s-, p = 0.001, n = 4) or human HFt (21.9 ± 0.9 s-, p =
0.001, n = 3), indicating that the hypermagnetic mutant L55P indeed shows higher sensitivity as an intracellularly expressed MRI contrast agent (Fig. 4a). The ability of SPFt L55P to enhance magnetic capture in HGMS was assessed by comparing the mutant protein to wild-type SPFt and
23 Ft-free control cells. Yeast cells expressing L55P were retained with four times greater efficacy than cells transformed with SPFt (Fig. 4b), demonstrating that the increased cellular magnetization due to expression of hypermagnetic mutant protein nanoparticles significantly improved the sensitivity of magnetic cell sorting process (p = 0.007, n = 3).
In this report, we have shown that a high-throughput protein selection strategy can be applied to enhance intracellular molecular-level biomineralization within Ft variants, resulting in proteins with substantially improved ability to induce magnetic phenotypes under physiological conditions. The ability to engineer spontaneous biomineralization processes that occur within intracellular proteins could be useful for a broad array of biotechnological applications. Because the enhanced mineral accumulation and magnetism generated here is explicitly associated with
Ft nanoparticles, as opposed to cellular mineral content more generically5 1 , 52, 64, the resulting hypermagnetic proteins can provide means for detection or manipulation of processes that occur on a molecular level, such as magnetic biosensing6 5 and control of signaling55 . Mechanistic analysis of the SPFt mutants identified here indicated that single amino acid substitutions significantly enhanced the uniformity of mineral formation within SPFt expressed in yeast. This result could not have been predicted from the PFt structure alone and shows that screening for
66 68 iron sequestration phenotypes can complement traditional site-directed mutagenesis studies - to expand knowledge about the mechanisms of iron mineralization by Ft. Altering mineral nucleation could also prove to be a general and versatile route for tuning intracellular biomineralization, particularly if unnatural mineral species are desired69 . In the future, high throughput screening approaches like the one introduced here could also be used to engineer additional metalloproteins, and could further alter magnetic properties or other physical parameters of genetically expressed nanocomplexes.
24 2.3. Materials and Methods
Yeast strain, growth conditions, and genetic methods
We used the haploid yeast (Saccharomyces cerevisiae) strain BY4742/FTRI-GFP (MAT a
FTR1-GFP:.:HISMXhis3Al leu2AO lys2AO ura3AO)60 (gift from Dr. Christopher Burd) as a host for expression of all Ft variants. We grew yeast cells in a dropout medium without histidine (SD-
HIS) made with a dry culture medium (Teknova, Hollister, CA) or in a YPAD medium: 10 g/L yeast extract (BD Biosciences, San Jose, CA), 20 g/L of Bacto Peptone (BD Biosciences, San
Jose, CA), 20 mg/L of adenine hemisulfate, and 20 g/L glucose. We transformed yeast cells with expression plasmids using the Frozen-EZ Yeast Transformation 1I kit (Zymo Research, Irvine,
CA).
Construction of Strep-tag II/ferritin fusion proteins
We used Escherichiacoli NEB 10 cells (New England Biolabs, Ipswich, MA) for plasmid construction. In order to create an expression plasmid with a dominant selectable marker, we used the polymerase chain reaction (PCR) to amplify a geneticin resistant cassette, KanMX4 from a plasmid pFA6-kanMX4 70 kindly provided by Dr. Peter Philippsen. We subcloned the
PCR product containing KanMX4 fragment into the pHVX2 yeast expression plasmid generously supplied by Dr. Hennie Van Vuuren 71. We then made a point deletion to destroy a superfluous EcoRI site by the QuikChange Lightning Kit (Agilent Technologies, Santa Clara,
CA) to yield the host plasmid, pHVX2G, used for subsequent expression of Ft constructs in our experiments. We amplified ferritin gene of Pyrococcusfuriosus (PFt) from the genomic DNA of the bacteria (ATCC, Manassas, VA). A Strep-tag II sequence (WSHPQFEK), spacer (GTSS),
25 and restriction sites were genetically fused at the 5' end of the PFt gene and the PCR product was subcloned into pHVX2G to yield plasmid pHVX2G-SPFt (Supplementary Table 2).
SPFt expression and affinity purification
For expression of SPFt, we inoculated yeast cells with expression plasmids in 1 mL of
YPAD media with 200 pg/mL Geneticin (Life Technologies, Carlsbad, CA) and incubated overnight at 30 'C. We then diluted the cultures into a fresh medium at OD 60 0 ~0.04 and incubated them for 16 hrs at 30 'C before harvesting. We washed the freshly harvested yeast with 30 mL of phosphate buffered saline (PBS) + 10 mM ethylenediaminetetraacetic acid
(EDTA) twice and finally resuspended in PBS. We lysed yeast cell pellet with Y-PER Plus
(Thermo Scientific, Waltham, MA), benzonase nuclease (EMD Millipore, Billerica, MA) and protease inhibitors according to the manufacturer's instructions. We then centrifuged the lysate at 3,000g for 20 min at 4 'C. SPFt protein was purified by applying the cleared lysate into the
Strep-Tactin sepharose column (IBA, Goettingen, Germany) according to the manufacturer's instructions, except EDTA was omitted from the wash and the elution buffers. We buffer exchanged and concentrated the purified protein into the wash buffer using a spin filter with 100 kDa cutoff membrane (EMD Millipore, Billerica, MA). We measured the protein concentrations by the Pierce 660 nm Protein Assay (Thermo Scientific, Waltham, MA), with bovine serum albumin (BSA) as a standard.
Transmission electron microscopy of purified SPFt
For conventional transmission electron microscopy (TEM), we applied 1-3 PI of 0.05 mg/mL SPFt sample onto a carbon/copper coated grid (Electron Microscopy Sciences, Hatfield,
26 PA), removed excess solution with a filter paper, and let it dry for 30 seconds. We then applied
15 iL of 1% phosphotungstic acid (pH 7.0) over the sample for about 10 seconds and removed the excess stain with a filter paper. The grid was dried at room temperature for at least 1 h before imaging with a JEOL 2010 HRTEM instrument (JEOL, Tokyo, Japan).
For cryo-electron microscopy, we applied 5 pl of the protein and buffer solution on a lacey copper grid coated with a continuous carbon film and removed excess sample without damaging the carbon layer using a Gatan Cryo Plunge III (Gatan, Pleasanton, PA). We mounted the grid on a Gatan 626 cryo-holder equipped in the TEM column and kept it under liquid nitrogen throughout the transfer into the microscope and the subsequent imaging session. We imaged the
SPFt samples on a JEOL 2100 FEG microscope (JEOL, Tokyo, Japan) using a minimum-dose method that was essential to avoid sample damage under the electron beam. We imaged at 200 kV with a magnification setting of 60,OO0x for assessing particle size and distribution and recorded the images on a Gatan 2k x 2k UltraScan CCD camera (Gatan, Pleasanton, PA).
In order to calculate the percentage of filled cores, we counted 400 particles per sample and divided the number of filled particles by 400. For each SPFt variant, we obtained cryo-EM images of the protein samples from two different batches in order to calculate mean, s.e.m., and statistical parameters.
Library construction
We carried out library construction using error-prone PCR, using parameters described previously 72 . The entire SPFt gene except for the Strep-tag II sequence was subjected to mutagenesis over 30 error-prone amplification cycles, which yielded on average one amino acid mutation per SPFt gene. The linearized vector was prepared by digesting pHVX2G with ApaI
27 and XhoI followed by gel purification. We transformed yeast with the SPFt library according to the method developed by Benatuil et a173 with a few modifications. We mixed 1.5 pg of digested plasmid and 0.5 tg of error-prone PCR product with 100 pL of electrocompetent cells (-1.6 x
109 cells/mL) in a disposable electroporation cuvette with 0.2 cm gap (Bio-Rad, Hercules, CA) on ice for 5 min. We electroporated the cells at 3 kV using MicroPulser electroporator (Bio-Rad,
Hercules, CA), resulting in time constants ranging from 4.8 to 5.3 ms. After electroporation, we immediately transferred the cells to 1:1 mix of 1 M sorbitol: YPAD medium and incubated in 30
'C for 3 h. We then harvested cells by centrifugation and resuspended in SD-HIS with 200 pg/mL of Geneticin and incubated for 2 days before freezing them for long-term storage at -80
C. Typical transformation efficiency was 0.5-1 x 107 transformants per pg of plasmid DNA.
The library diversity was tested by sequencing randomly picked 24 colonies.
Measurements of iron content in yeast cells and purified SPFt
We used a colorimetric assay based on the protocol of Tamarit et al.74 to quantify the iron content of yeast cells and the purified protein. This method relies on the Fe2+-dependent optical absorbance of bathophenanthrolinedisulfonic acid (BPS) at 535 nm at pH 5.4. As standards, we dissolved known amounts of ferrous ammonium sulfate in 3% nitric acid.
For measuring the iron content of yeast cells, we digested 4.2 x 108 cells by boiling in 200 tL of 3% nitric acid for 2 h, and centrifuged at 10,000g for 5 min. In order to measure the concentration of iron in SPFt, a 1:1 ratio of purified protein and 3 % nitric acid solution were mixed and boiled for 15 min followed by centrifugation at 10,000 g for 5 min. In both cases, the iron quantification assay was applied to the supernatant of the resulting samples. Iron loading stoichiometries of the protein samples were computed by dividing the iron concentrations by the
28 protein concentrations, as measured by the 660 nm Protein Assay (Thermo Scientific, Waltham,
MA).
High-throughput screening of yeast cells with SPFt library using FACS
We inoculated 1 x 108 cells in a 20 mL SD-HIS medium containing 200 ptg/mL of
Geneticin at 30 C overnight (about 16-20 h). We harvested the cells in a culture tube and resuspended in a sterile PBS such that the cell density was about 5 x 107 cells/mL. We filtered the cells with a sterile membrane with 40 pm pores immediately before sorting. Similarly, we prepared negative control samples using the BY4742 background strain without the FTR1-GFP reporter. We set up a flow cytometry protocol using the control yeast samples. First, the yeast population was gated with forward and side scattering channels to remove debris and aggregated cells. We then collected cells displaying green fluorescence in the top -5%, indicating high
FTR1-GFP expression. We propagated these cells overnight in 4 mL of SD-HIS medium supplemented with 200 tg/mL of Geneticin.
Measurement of iron oxidation and release kinetics
We monitored the kinetics of iron oxidation by SPFt variants by an optical assay75 . We prepared SPFt samples with 100 Fe/24-mer in 100 mM MOPS, pH 7.0. We added ferrous
ammonium sulfate solution (1 mM), made in degassed distilled water to the protein solution
(final concentration of 0.1 pM) at a 500-fold molar excess of iron(II). Following a mixing dead
time (~5 s), we recorded the optical absorbance of the mixture at 315 nm every 2 s for 5 min. We
used a disposable cuvette with a 1 cm path length and recorded the spectra with SpectraMax M2
Microplate reader (Molecular Devices, Sunnyvale, CA). We calculated the specific activity,
29 defined as the micromoles of iron(III) formed per minute per milligram of 24-mer SPFt by
dividing the change in absorbance of the reaction mixture over the first 30 s by the extinction
coefficients of SPFt variants and the amount of protein in the reaction. Extinction coefficients for wild type SPFt, L55P, F57S, and F123S were 2.6 ± 0.1, 2.6 ± 0.1, 2.7 ± 0.1, and 2.8 ± 0.1 mM-
Icm', respectively.
We measured the kinetics of iron release from preloaded SPFt variants by monitoring time
dependent formation of the BPS complex with Fe2 released from iron-loaded Ft variants. We used purified SPFt samples that were loaded aerobically with 1,000 Fe atoms per molecule.
These samples were diluted to a final concentration of 0.1 pM SPFt in an iron mobilization assay buffer that included MOPS (0.1 M, pH 7.0), sodium acetate (20 mM), and BPS (1 mM). We measured the absorbance values at 535 nm every 30 s for 3 hours using SpectraMax M2
Microplate reader. We took the first 3.5 min of the data and computed the initial rate of iron release using the standard curve constructed using freshly made ferrous ammonium sulfate
solutions.
Yeast cell pellet MRI
We prepared the yeast samples as described in SPFt expression and purification section.
After we washed the cells twice with PBS supplemented with 10 mM EDTA, the supernatant was decanted and 100 pL of the cell suspension was loaded into the wells of a microtiter plate.
Unused wells were filled with PBS. We centrifuged the plate at 1,500g for 3 min and placed it in
a 12 cm outer diameter birdcage transceiver for imaging in a 20-cm-bore Bruker 7 T Avance III
MRI scanner. We imaged a 2 mm slice through the cell pellet samples with the field of view of 5
x 5 cm and the data matrices were 256 x 256 points. We used T2-weighted spin echo pulse
30 sequence with multiecho acquisition; repetition time (TR) was 2 s, and echo time (TE) ranged from 5 ms to 150 ms in 5 ms intervals. We used custom routines written in Matlab (Mathworks,
Natick, MA) to reconstruct the images and computed relaxation time constants (T2) by fitting image intensity data to exponential decay curves.
Magnetic cell sorting
High gradient magnetic separations of yeast cells were performed using magnetic columns
(Miltenyi Biotec, Bergisch Gladbach, Germany) inserted into a Frantz Canister Separator, Model
L-1CN (S. G. Frantz Company Inc., Tullytown, PA). Briefly, we suspended yeast cells at the density of 2 x 108 cells/mL in a sorting buffer consisting of PBS supplemented with 2 mM
EDTA and 0.5% BSA. After equilibrating the column with the sorting buffer, we applied the yeast cells on the column in the presence of an externally applied magnetic field of 0.6 T followed by a wash with the sorting buffer. We then switched off the magnetic field and eluted the cells from the column with the sorting buffer. We collected the flow through, the wash and the elution fractions from the column into a 96-well microtiter plate. We carried out optical density measurements at 600 nm to estimate the cell densities of each fraction and computed the percentages of cells retained on the columns.
2.4. Acknowledgements
We thank D. S. Yun from the nanotechnology materials core facility at the Koch Institute for technical support with electron microscopy. We also thank staff of the flow cytometry core facility at Koch Institute for assistance. This research was supported by NIH grants DP2-
31 OD002114, RO1-NS076462, and RO1-MH103160 to APJ. YM was supported by a Siebel
Scholar Fellowship and a Friends of the McGovern Institute Fellowship.
2.5. Figure captions
Figure 11 Fluorescent reporter system used to probe intracellular iron accumulation by Ft a, Schematic diagram of yeast cells containing an iron-responsive reporting system.
Sequestration of cytosolic iron (red dots) into Ft (gray) 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 storing ability, 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. c, FACS histograms showing the distribution of GFP-associated fluorescence observed in yeast cell populations transformed with vector, SPFt, and E94G/K142R.
Figure 2 1 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 harboring 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
32 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 (right panel) 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 minimum 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 1Structural 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)24 . Enlarged image of the highlighted subunit (right) shows the relative positions of sidechains mutated in the selected iron loading mutants (L55P,
F57S, and F123S) with respect to the ferrooxidase residues (yellow) and the known iron binding sites (gray balls). b, CryoEM images of purified SPFt, L55P, F57S, and F123S, showing formation of 12 nm spherically shaped nanoparticles in each case. SPFt samples also display
33 differences in electron dense iron core formation, as indicated by the variable frequency of
"empty" particles in the images (e.g. yellow arrowheads). Scale bar = 50 nm. c, The percentage of particles containing electron dense cores was computed by analyzing 400 particles in cryoEM 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 1Engineered SPFt mutants are highly effective hypermagnetic probes 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/T 2) 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 = 2000 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.
34 2.6. Figures
Figure 11 Fluorescent reporter system used to probe intracellular iron accumulation by Ft
a Low Ft iron High Ft iron
100 bur K 3 VecO 80 SPR S60' 40ec SPFtE 20
E94G/ 0 1 2 3 4 K142R Log(GFP fluorescence)
35 Figure 2 1Selection of SPFt mutants by high-throughput genetic screening
a4 C 100. 2 -AS3 a F57S F123S
2 Lib 60 SPFR L55P 23 40 V01 S4 40 0 2D 0 1 2 3 4 LoWGFP fluorescenc) o 1 2 3 4 0 1 2 3 4 Log(GFP fuorescenc.) Log(GFP uo Snce)
d 1s0 0 1200 f I m100
4W0 Prussan bue Qw) 50 i
36 Figure 3 1Structural analysis of SPFt variants
a C100
L55P F7 80
0*F123S
60 - *4
b
37 Figure 4 Engineered SPFt mutants are highly effective hypermagnetic probes
a wo b 80.6
201 0.2 SPFt .II A0$ 04 d 0- f -)P441O Wash
38 2.7. Supplementary material
Supplementary Table S1. Kinetics of iron oxidation and release by SPFt variants*
Ferritin Iron oxidation Iron release variant specific activity (U/mg) initial rate (pM/min.) SPFt 0.22 ± 0.03 0.44 ± 0.04 L55P 0.26 ± 0.01 0.39 ± 0.00 F57S 0.31 ± 0.01 0.54 ± 0.02 F123S 0.28 ± 0.03 0.43 ± 0.01 *Reported values reflect mean and standard error of 3 independent measurements.
Supplementary Table S2. DNA sequence of plasmid pHVX2G-SPFt
GCGCCCAATACGCAAACCGCCTCTCCCCGCGCGTTGGCCGATTCATTAATGCAGCTGGCACGACAGGTTTCCCGACTGGAAAG CGGGCAGTGAGCGCAACGCAATTAATGTGAGTTAGCTCACTCATTAGGCACCCCAGGCTTTACACTTTATGCTTCCGGCTCGTA TGTTGTGTGGAATTGTGAGCGGATAACAATTTCACACAGGAAACAGCTATGACCATGATTACGCCAAGCTTTCTAACTGATCTA TCCAAAACTGAAAATTACATTCTTGATTAGGTTTATCACAGGCAAATGTAATTTGTGGTATTTTGCCGTTCAAAATCTGTAGAAT TTTCTCATTGGTCACATTACAACCTGAAAATACTTTATCTACAATCATACCATTCTTATAACATGTCCCCTTAATACTAGGATCAG GCATGAACGCATCACAGACAAAATCTTCTTGACAAACGTCACAATTGATCCCTCCCCATCCGTTATCACAATGACAGGTGTCATT TTGTGCTCTTATGGGACGATCCTTATTACCGCTTTCATCCGGTGATAGACCGCCACAGAGGGGCAGAGAGCAATCATCACCTGC AAACCCTTCTATACACTCACATCTACCAGTGTACGAATTGCATTCAGAAAACTGTTTGCATTCAAAAATAGGTAGCATACAATTA AAACATGGCGGGCACGTATCATTGCCCTTATCTTGTGCAGT-TAGACGCGAAT1-TTCGAAGAAGTACCTTCAAAGAATGGGGT CTCATCTTGTTTTGCAAGTACCACTGAGCAGGATAATAATAGAAATGATAATATACTATAGTAGAGATAACGTCGATGACTTCC CATACTGTAATTGCTTTTAGTTGTGTATTT1-TAGTGTGCAAGTTTCTGTAAATCGATTAA1TnTFUTTT CTTTCCTCT 1FATTAAC CTTAATTUTTATTTTAGATTCCTGACTTCAACTCAAGACGCACAGATATTATAACATCTGCACAATAGGCATTTGCAAGAATTACT CGTGAGTAAGGAAAGAGTGAGGAACTATCGCATACCTGCATTTAAAGATGCCGATTTGGGCGCGAATCCTTTATTTTGGCTTC ACCCTCATACTATTATCAGGGCCAGAAAAAGGAAGTGTTTCCCTCCTTCTTGAATTGATGTTACCCTCATAAAGCACGTGGCCTC TTATCGAGAAAGAAATTACCGTCGCTCGTGATTTGTTTGCAAAAAGAACAAAACTGAAAAAACCCAGACACGCTCGACTTCCTG TCTTCCTATTGATTGCAGCTTCCAATTTCGTCACACAACAAGGTCCTAGCGACGGCTCACAGGTTTTGTAACAAGCAATCGAAG GTTCTGGAATGGCGGGAAAGGGTTTAGTACCACATGCTATGATGCCCACTGTGATCTCCAGAGCAAAGTTCGTTCGATCGTAC TGTTACTCTCTCTCTTTCAAACAGAATTGTCCGAATCGTGTGACAACAACAGCCTGTTCTCACACACTCTTTTCTTCTAACCAAGG GGGTGGTTTAGTTTAGTAGAACCTCGTGAAACTTACATTTACATATATATAAACTTGCATAAATTGGTCAATGCAAGAAATACA TATTTGGTCTTTTCTAATTCGTAG1TrTTCAAGTTCTTAGATGCTTTCT--I-TTCTC1TTTTACAGATCATCAAGGAAGTAATTATCT AC1TTTACAACAAATATAAAACAAGATCGGAATTCTAGAAATGTCTTGGTCTCACCCACAATTCGAAAAGGGGCCCGGTACTA GTAGTTTGAGCGAAAGAATGCTCAAGGC TTAAATGACCAGCTAAACAGGGAGCTTTATTCTGCATATCTATACTTTGCCATGG CTGCCTACTTTGAAGATCTTGGCCTTGAAGGTTTCGCCAACTGGATGAAGGCTCAGGCTGAAGAAGAGATTGGGCATGCACTG AGGTTCTACACTACATCTACGATCGCATGGTAGGGTGAGCTTGATGAAATTCCAAAGCCTCCAAAGGAGTGGGAGAGCCC ATTAAAAGCTTTTGAAGCTGCTTACGAGCATGAGAAATTCATAAGCAAGTCCATATATGAATTGGCAGC1TTAGCAGAGGAGG AAAAAGATTACTCGACGAGGGATC-AGAGTGTUATcACGAGCAGGTTGAGGAAGAGGCCAGCGTAAAGAAAATACT GGACAAGTTAAAGTTTGCTAAGGACAGTCCTCAAATATTGTTCATGCTTGATAAGGAGTTGAGTGCGAGAGCTCCAAAGCTCC CAGGGCTCTTAATGCAGGGAGGAGAGTAACTCGAGGGATCTGCGATAGATCAATT11TTTCTTTTCTCTTTCCCCATCCTTTACG CTAAAATAATAGTTTATTTTAIIIIIIGAATATTTTTTATATATACGTATATATAGACTATTATTTATCTTTTAATGATTATTAAG ATTTrTATTAAAAAAAAATTCGCTCCTCTTAATGCCTTTATGCAGTTTT1TITFCCCATTCGATATTTCTATGTTCGGGTTCAGC GTATTTTAAGTTTAATAACTCGAAAATTCTGCGTTCGTTAAAGCTTGCATGCCTGCAGGTCGACTCTAGAGGATCCCCGGGTAC
39 CGAGCTCGAATATTCACTGGCCGTCGTTTTACAACGTCGTGACTGGGAAAACCCTGGCGTTACCCAACTTAATCGCCTTGCAGC ACATCCCCCTTTCGCCAGCTGGCGTAATAGCGAAGAGGCCCGCACCGATCGCCCTTCCCAACAGTTGCGCAGCCTGAATGGCG AATGGCGCCTGATGCGGTATTTTCTCCTTACGCATCTGTGCGGTATTTCACACCGCATATATCGGATCGTACTTGTTACCCATCA TTGAATTTTGAACATCCGAACCTGGGAGTTTTCCCTGAAACAGATAGTATATTTGAACCTGTATAATAATATATAGTCTAGCGCT TTACGGAAGACAATGTATGTATTTCGGTTCCTGGAGAAACTATTGCATCTATTGCATAGGTAATCTTGCACGTCGCATCCCCGG TTCATTTTCTGCGTTTCCATCTTGCACTTCAATAGCATATCTTTGTTAACGAAGCATCTGTGCTTCATTTTGTAGAACAAAAATGC AACGCGAGAGCGCTAATTTTTCAAACAAAGAATCTGAGCTGCATTTTTACAGAACAGAAATGCAACGCGAAAGCGCTATTTTA CCAACGAAGAATCTGTGCTTCA]TTTTGTAAAACAAAAATGCAACGCGAGAGCGCTAAT1TrTCAAACAAAGAATCTGAGCTGC ATTTTTACAGAACAGAAATGCAACGCGAGAGCGCTATTTTACCAACAAAGAATCTATACTTC TTTGTTCTACAAAAATGCAT CCCGAGAGCGCTAT1T[TCTAACAAAGCATCTTAGATTACTTTTTTTCTCCTTTGTGCGCTCTATAATGCAGTCTCTTGATAACTTT TTGCACTGTAGGTCCGTTAAGGTTAGAAGAAGGCTACTTTGGTGTCTATTTTCTCTTCCATAAAAAAAGCCTGACTCCACTTCCC GCGTTTACTGATTACTAGCGAAGCTGCGGGTGCATTTTTTCAAGATAAAGGCATCCCCGATTATATTCTATACCGATGTGGATT GCGCATACTTTGTGAACAGAAAGTGATAGCGTTGATGATTCTTCATTGGTCAGAAAATTATGAACGGTTTCTTCTATTTTGTCTC TATATACTACGTATAGGAAATGTTTACATTTTCGTATTGTTTTCGATTCACTCTATGAATAGTTCTTACTACAATTTTTTTGTCTAA AGAGTAATACTAGAGATAAACATAAAAAATGTAGAGGTCGAGTTTAGATGCAAGTTCAAGGAGCGAAAGGTGGATGGGTAG GTTATATAGGGATATAGCACAGAGATATATAGCAAAGAGATACTTTTGAGCAATGTTTGTGGAAGCGGTATTCGCAATATTTT AGTAGCTCGTTACAGTCCGGTGCGTTTTTGGTTTTTTGAAAGTGCGTCTTCAGAGCGCTTTTGGTTTTCAAAAGCGCTCTGAAGT TCCTATACTTTCTAGCTAGAGAATAGGAACTTCGGAATAGGAACTTCAAAGCGTTTCCGAAAACGAGCGCTTCCGAAAATGCA ACGCGAGCTGCGCACATACAGCTCACTGTTCACGTCGCACCTATATCTGCGTGTTGCCTGTATATATATATACATGAGAAGAAC GGCATAGTGCGTGTTTATGCTTAAATGCGTACTTATATGCGTCTATTTATGTAGGATGAAAGGTAGTCTAGTACCTCCTGTGAT ATTATCCCATTCCATGCGGGGTATCGTATGCTTCCTTCAGCACTACCCTTTAGCTGTTCTATATGCTGCCACTCCTCAATTGGATT AGTCTCATCCTTCAATGCTATCATTTCCTTTGATATTGGATCGATCCGATGATAAGCTGTCAAACATGAGAATTAATTCTACCCTA TGAACATATTCCATTTTGTAATTTCGTGTCGTTTCTATTATGAATTTCATTTATAAAGTTTATGTACACGTACGCTGCAGGTCGAC CGTACGCTGCAGGTCGACGGATCCCCGGGTTAATTAAGGCGCGCCAGATCTGTTTAGCTTGCCTCGTCCCCGCCGGGTCACCC GGCCAGCGACATGGAGGCCCAGAATACCCTCCTTGACAGTCTTGACGTGCGCAGCTCAGGGGCATGATGTGACTGTCGCCCGT ACATTTAGCCCATACATCCCCATGTATAATCATTTGCATCCATACATTTTGATGGCCGCACGGCGCGAAGCAAAAATTACGGCT CCTCGCTGCAGACCTGCGAGCAGGGAAACGCTCCCCTCACAGACGCGTTGAATTGTCCCCACGCCGCGCCCCTGTAGAGAAAT ATAAAAGGTTAGGATTTGCCACTGAGGTTCTTCTTTCATATACTTCCTTTTAAAATCTTGCTAGGATACAGTTCTCACATCACATC CGAACATAAACAACCATGGGTAAGGAAAAGACTCACGTTTCGAGGCCGCGATTAAATTCCAACATGGATGCTGATTTATATGG GTATAAATGGGCTCGCGATAATGTCGGGCAATCAGGTGCGACAATCTATCGATTGTATGGGAAGCCCGATGCGCCAGAGTTG TTTCTGAAACATGGCAAAGGTAGCGTTGCCAATGATGTTACAGATGAGATGGTCAGACTAAACTGGCTGACGGAATTTATGCC TCTTCCGACCATCAAGCATTTTATCCGTACTCCTGATGATGCATGGTTACTCACCACTGCGATCCCCGGCAAAACAGCATTCCAG GTATTAGAAGAATATCCTGATTCAGGTGAAAATATTGTTGATGCGCTGGCAGTGTTCCTGCGCCGGTTGCATTCGATTCCTGTT TGTAATTGTCCTTTTAACAGCGATCGCGTATTTCGTCTCGCTCAGGCGCAATCACGAATGAATAACGGTTTGGTTGATGCGAGT GATTTTGATGACGAGCGTAATGGCTGGCCTGTTGAACAAGTCTGGAAAGAAATGCATAAGCTTTTGCCATTCTCACCGGATTCA GTCGTCACTCATGGTGATTTCTCACTTGATAACCTTA 1TVTGACGAGGGGAAATTAATAGGTTGTATTGATGTTGGACGAGTC GGAATCGCAGACCGATACCAGGATCTTGCCATCCTATGGAACTGCCTCGGTGAGTTTTCTCCTTCATTACAGAAACGGCJTTITC AAAAATATGGTATTGATAATCCTGATATGAATAAATTGCAGTTTCATTTGATGCTCGATGAGTTTTTCTAATCAGTACTGACAAT AAAAAGATTCTTGTTTTCAAGAACTTGTCATTTGTATAGT]T1TTTATATTGTAGTTGTTCTATTTTAATCAAATGTTAGCGTGATT TATATTTTTTTTCGCCTCGACATCATCTGCCCAGATGCGAAGTTAAGTGCGCAGAAAGTAATATCATGCGTCAATCGTATGTGA ATGCTGGTCGCTATACTGCTGTCGATTCGATACTAACGCCGCCATCCAGTGTCGAAAACGAGCTCGATTCATCGATGACGTCAG GTGGCACTTTTCGGGGAAATGTGCGCGGAACCCCTATTTGTTTATTTTTCTAAATACATTCAAATATGTATCCGCTCATGAGACA ATAACCCTGATAAATGCTTCAATAATATTGAAAAAGGAAGAGTATGAGTATTCAACATTTCCGTGTCGCCCTTATTCCCTTTTTT GCGGCATTTTGCCTTCCTGTTTTTGCTCACCCAGAAACGCTGGTGAAAGTAAAAGATGCTGAAGATCAGTTGGGTGCACGAGT GGGTTACATCGAACTGGATCTCAACAGCGGTAAGATCCTTGAGAGTTTTCGCCCCGAAGAACGTTTTCCAATGATGAGCACTTT TAAAGTTCTGCTATGTGGCGCGGTATTATCCCGTATTGACGCCGGGCAAGAGCAACTCGGTCGCCGCATACACTATTCTCAGAA TGACTTGGTTGAGTACTCACCAGTCACAGAAAAGCATCTTACGGATGGCATGACAGTAAGAGAATTATGCAGTGCTGCCATAA CCATGAGTGATAACACTGCGGCCAACTTACTTCTGACAACGATCGGAGGACCGAAGGAGCTAACCG CTTTTTTG CACAACATG GGGGATCATGTAACTCGCCTTGATCGTTGGGAACCGGAGCTGAATGAAGCCATACCAAACGACGAGCGTGACACCACGATGC CTGTAGCAATGGCAACAACGTTGCGCAAACTATTAACTGGCGAACTACTTACTCTAGCTTCCCGGCAACAATTAATAGACTGGA TGGAGGCGGATAAAGTTGCAGGACCACTTCTGCGCTCGGCCCTTCCGGCTGGCTGGTTTATTGCTGATAAATCTGGAGCCGGT GAGCGTGGGTCTCGCGGTATCATTGCAGCACTGGGGCCAGATGGTAAGCCCTCCCGTATCGTAGTTATCTACACGACGGGGA GTCAGGCAACTATGGATGAACGAAATAGACAGATCGCTGAGATAGGTGCCTCACTGATTAAGCATTGGTAACTGTCAGACCAA
40 GTTTACTCATATATACTTTAGATTGATTTAAAACTTCATTTTTAATTTAAAAGGATCTAGGTGAAGATCCTTTGATAATCTCAT GACCAAAATCCCTTAACGTGAGTTTTCGTTCCACTGAGCGTCAGACCCCGTAGAAAAGATCAAAGGATCTTCTTGAGATCCTTT TTTTCTGCGCGTAATCTGCTGCTTGCAAACAAAAAAACCACCGCTACCAGCGGTGGTTTGTTTGCCGGATCAAGAGCTACCAAC TCTTTTTCCGAAGGTAACTGGCTTCAGCAGAGCGCAGATACCAAATACTGTCCTTCTAGTGTAGCCGTAGTTAGGCCACCACTT CAAGAACTCTGTAGCACCGCCTACATACCTCGCTCTGCTAATCCTGTTACCAGTGGCTGCTGCCAGTGGCGATAAGTCGTGTCT TACCGGGTTGGACTCAAGACGATAGTTACCGGATAAGGCGCAGCGGTCGGGCTGAACGGGGGGTTCGTGCACACAGCCCAG CTTGGAGCGAACGACCTACACCGAACTGAGATACCTACAGCGTGAGCTATGAGAAAGCGCCACGCTTCCCGAAGGGAGAAAG GCGGACAGGTATCCGGTAAGCGGCAGGGTCGGAACAGGAGAGCGCACGAGGGAGCTTCCAGGGGGAAACGCCTGGTATCTT TATAGTCCTGTCGGGTTTCGCCACCTCTGACTTGAGCGTCGATTTTTGTGATGCTCGTCAGGGGGGCGGAGCCTATGGAAAAA CGCCAGCAACGCGGCCTTTTTACGGTTCCTGGCCTTTTGCTGGCCTTTTGCTCACATGTTCTTTCCTGCGTTATCCCCTGATTCTG TGGATAACCGTATTACCGCCTTTGAGTGAGCTGATACCGCTCGCCGCAGCCGAACGACCGAGCGCAGCGAGTCAGTGAGCGA GGAAGCGGAAGA
Supplementary figures captions
Supplementary Figure 1 Design and characterization of affinity-tagged PFt a, Schematic of the DNA construct leading to self-assembled SPFt. Each polypeptide chain contains an N-terminal Strep-tag II (blue), a GTSS linker (white), and the Ft gene from
Pyrococcusfuriosus (gray); these proteins form homooligomers of 24 subunits each. b, Native gel analysis of horse spleen ferritin (HSF), containing approximately 3,000 Fe atoms per 24-mer, and purified wild type SPFt, aerobically loaded with 1,500 Fe/24-mer, together stained with
Coomassie blue (left) and Prussian blue (right). The Prussian blue stain indicates iron content semiquantitatively, and shows that SPFt is capable of loading iron in vitro. c, Coomassie-stained sodium dodecylsulfate polyacrylamide gel showing that affinity purification yields highly pure
SPFt, with a single band near the expected molecular weight of 22 kD indicated by a black arrowhead. Molecular weight standards (std) are shown at left. d, Transmission electron micrograph of SPFt with negative staining showing iron mineral cores surrounded by protein shells of about 12 nm in diameter, consistent with the crystal structure of this PFt. Scale bar = 20 nm. e, Dynamic light scattering size histogram of SPFt nanoparticles, showing an average hydrodynamic diameter of about 12 nm.
41 Supplementary Figure 2 1Mutation rates in the initial SPFt library
Distribution of the number of nucleotide (blue) and amino acid (red) mutations per gene in the
SPFt library used as the starting point for screening. The average number of DNA-level mutations is 1.0 and the average number of coding changes is 0.6.
42 Supplementary figures
Supplementary Figure 1 Design and characterization of affinity-tagged PFt
a b protein iron
6 ,ww v
40 30 40
20 20L I 15La. 10 0Log(diameter)
43 Supplementary Figure 2 Mutation rates in the initial SPFt library
12 Enuosodes 10 *an*1oacf,
I6
2
0 0 1 2 . 3 Muwkms gnb
44 3. Clusters of genetically engineered hypermagnetic nanoparticles report dynamic changes in gene expression with MRI
3.1. Abstract
Magnetic resonance imaging (MRI) offers high spatial resolution and deep tissue penetration, making it an ideal tool for non-invasively monitoring gene expression levels in vivo.
Only a handful of MRI gene reporters have been developed 34,35,44,45,50,53, however, and they are not robust enough to be widely used in vivo. One of the most promising MRI gene reporters is based on a cytosolic iron-storage protein, ferritin (Ft), but its usage is limited to monitoring long term static gene expression because there is a delay between the gene expression and the contrast produced by Ft depending on the iron availability in the cell76 . Here we introduce a new design of Ft-based MRI probe that can report dynamic changes in gene expression in a cell. Our sensor is based on a chimeric ferritin nanoparticle made up of 24 identical polypeptides; each containing iron mineralization and sensing domains. As illustration, we showed that our MRI probe can detect dynamic changes in gene expression levels in yeast cells within 2 hours after induction, with contrast effects that are stable for several hours.
45 3.2. Introduction, results and discussions
Non-invasively monitoring gene expression levels in living organisms is challenging. For optical imaging, well-characterized gene reporters such as green fluorescent protein have been widely available but its usage is limited to cell culture and optically accessible small organisms.
MRI allows imaging of opaque tissues at cellular resolution, but lacks robust gene reporters which translate the target gene expression level into a signal detectable by MRI. Previously reported MRI gene reporters often require synthetic chemical moieties such as superparamagnetic nanoparticles (SPNs) 44 and gadolinium chelates 45, making it difficult to synthesize and deliver to intracellular locations.
Use of genetically-coded MRI contrast agents such as Ft could solve the issues with synthesis and delivery. Ft is a self-assembling protein nanoparticle with a 12 nm spherical shell made up of 24 identical or similar polypeptides containing an iron mineral core, which is weakly magnetic and generates contrast in MRI. Several groups have used Ft as a gene reporter for
MRI 34-4 0, 64, but its low sensitivity and static contrast do not allow sensing of dynamic changes in gene expression. After Ft gene is introduced into a cell, there is a delay between the onset of expression and contrast generation by Ft because expression and iron loading process take time 7. In vitro experiments have shown that aggregation of Ft due to chemical cross-linking77 or protein-protein interactions65 increased their T2 relaxation rates in MRI. Because aggregation process is relatively fast compared to protein expression and iron loading, we proposed to develop an aggregation-based Ft gene reporter that responds to dynamic changes in gene expression levels.
Our sensor is made up of two components: Ft nanoparticles, which act as a source of
MRI contrast, and crosslinking proteins, which promote aggregation of the nanoparticles and
46 increases their T2 relaxation rate (Figure 1 a). Each Ft monomer contains an N-terminal sensing
(Strep-tag II) domain and the mineralization domain derived from thermostable Pyrococcus furiosus Ft (PFt); the fusion construct is referred to as SPFt and described in Chapter 2. Twenty
four sensing domains are displayed outside of each nanoparticle to serve a dual purpose, first, to
facilitate SPFt purification and second, to interact with crosslinking proteins to form aggregates.
We used a mutant version (L55P) of the mineralization domain that has been optimized for iron
loading capability through a high-throughput screening (Chapter 2). SPFt-L55P nanoparticles
were expressed in yeast, purified with a Strep-Tactin column and iron loaded with 2530 ± 117
iron atoms per nanoparticle on average. We used streptavidin homo-tetramers (SA) as a
crosslinking protein, which binds tightly to Strep-tag 1I domains of the nanoparticles and induces
particle aggregation (Figure la and Ib). We used a commercially available recombinant SA for
the following in vitro experiments.
As an initial demonstration of nanoparticle aggregation, we prepared two tubes with
purified SPFt-L55P solutions and added SA to one and buffer to the other, followed by a quick
centrifugation (~5 sec) to sediment aggregated species. A pellet is visible only in the tube with
SA, indicating that substantial aggregation of nanoparticles takes place in the presence but not
the absence of SA (Figure ic). SA-dependent aggregation of SPFt-L55P was also visible under
cryo-electron microscopy (Figure 1 d), a technique that allows imaging of biological samples at
near native environment. The radius of the control nanoparticles appeared about 12 nm as
predicted whereas that of the SPFt-L55P nanoparticles mixed with SA appeared significantly
larger than the size of a single particle. We also measured the size of the aggregates by dynamic
light scattering (DLS). Figure le shows that the hydrodynamic radius of the aggregates is more
than 100 fold larger than that of unaggregated control nanoparticles (8.5 ± 0.6 nm vs 1107 ±
47 106.0 nm). The aggregation process is partially reversible. When excess biotin is added to the aggregated nanoparticles, the hydrodynamic radius of the aggregates reduced to ~50-100 nm, but never went down to the original size of unaggregated nanoparticles.
We further characterized the aggregation behavior of the Ft nanoparticles and demonstrated that the aggregation of nanoparticles is dependent on the concentration of the crosslinker by measuring the DLS signals as a function of SA concentration (Figure 2a). We also showed that the sensitivity of the sensor can be modulated by introducing mutations in the vicinity of Strep-tag II binding site of SA. We engineered a SA variant that incorporated four mutations (E44V, S45T, V47R and W120A) and an N-terminal T7 tag to facilitate its folding and purification 78, and we named this variant STm (Supplementary Figure 1). The first three mutations increased its binding affinity to Strep-tag II9 and the last mutation decreased its binding affinity to biotin80 , which facilitated bacterial cell growth and improved its yield. The midpoint of the binding curve (EC50) dropped to 0.16 pM with STm compared to 12.3 pM with
SA, showing that these mutations improved the sensitivity of the aggregation sensor by almost two orders of magnitude. Increase in sensitivity is important especially when the sensor is applied in a cell where high levels of crosslinker expression may not be feasible. The corresponding MRI measurements of the SPFt and STm mixtures were made (Figure 2b). As expected, we observed an increase in the relaxation rate (R2) as the concentration of crosslinker was increased. The relaxation rates of the aggregated samples were more than two-fold higher than that of unaggregated samples.
One of the key objectives of this study was to develop a sensor which responds to dynamic changes in gene expression levels. Previously developed aggregation-based sensors worked on the time scale of minutes to hours in vitro 3' 65. We analyzed the kinetics of
48 aggregation formation by taking time course measurements with DLS. Figure 2c shows the average time course of 7 independent experiments where STm was added to the solution of
SPFt-L55P and briefly mixed by pipetting. We observed a rapid increase in cluster size after addition of the crosslinker. Within the first 15 sec, aggregation size increased about 10 fold; the size kept increasing for the next minute, indicating that this sensor is appropriate for measuring changes in gene expression levels which generally take minutes to hours.
In order to demonstrate that the aggregation of nanoparticles can induce MRI contrast in a cellular environment, we cotransformed two episomal plasmids - one containing SPFt-L55P and the other containing STm into yeast cells. We first incubated yeast in an iron supplemented rich medium overnight to load iron into constitutively expressed SPFt-L55P nanoparticles. Then these cells were transferred to a galactose medium to induce expression of STm, and the cells were harvested at a few time points for Western blot and MRI measurements. As a control, we used Ft variant PFt-L55P, which is identical to SPFt-L55P but without the Strep-tag II, such that it cannot interact with the crosslinker. Compared to the control cells, the normalized relaxation rates (R2/total cellular iron concentration) of the experimental cell pellets were significantly higher at both 2 h and 4.5 h time points (p = 0.02 and 0.04, respectively) (Figure 3 bottom panel). As soon as 2 h after induction, the cells with aggregated nanoparticles showed about 14% increase in the normalized relaxation rate, and the difference increased to 20% at the 4.5 h time point. From the Western blot images, it is apparent that STm is expressed only after induction at
2 h and 4.5 h time points in both control (STm + PFt) and experimental (STm + SPFt) samples, whereas SPFt is expressed only in experimental samples and not with control samples, as expected, at all time points (Figure 3 top panel). We conducted the same experiment with a non- magnetic mutant version of SPFt, SPFt -E94G/K142R (Chapter 2) and did not observe
49 significant difference in normalized relaxation rates between control and experimental samples
(Supplementary Figure 2). This indicates that the increase in relaxation rate is due to the magnetic effect of the aggregated SPFt-L55P nanoparticles and not to the large protein aggregates interacting with cells or influencing endogenous contrast in yeast.
In this report, we introduced an aggregation-based dynamic MRI gene reporter which features simple design and an iron-loading optimized Ft variant. The fully genetically-encoded nanoparticles were easily made in a cell and displayed fixed number of sensing domain per particle, leading to predictable aggregation behavior. Previously reported systems required multiple species of nanoparticles that are made up of several different types of polypeptides, making it difficult to apply in vivo, but our system contains only one type of nanoparticle made up of single chain of Ft. Together with the crosslinking protein, the system allowed non-invasive detection of dynamic changes in gene expression by MRI in a cell. We demonstrated for the first time that the cellular MRI contrast can be manipulated by the aggregation status of nanoparticles.
Our design of aggregation-based gene reporter is simple enough to be applied for in vivo experiments and can be generalized for variety of purposes. By replacing the promoter fragment which controls the expression of crosslinker protein, we could use this system to monitor the expression levels or functional status of proteins involved in a variety of biological processes, potentially including embryogenesis, disease progression, and gene therapy treatment.
3.3. Materials and methods
Yeast strain and genetic methods
We used the haploid yeast (Saccharomyces cerevisiae) strains, YPH499 (MATa ura3-52 lys2-801_amber ade2-101_ochre trplA63 his3A200 leu2Al) (ATCC, Manassas, VA) for
50 expression of SPFt. For coexpression study, we used another haploid yeast strain,
BY4742/FTR]-GFP (MAT a FTR1-GFP::HISMXhis3Al leu2AO lys2AO ura3AO) 60 (gift from
Dr. Christopher Burd). Yeast cells were transformed using Frozen-EZ Yeast Transformation II kit (Zymo Research, Irvine, CA) according to the manufacturers' instructions.
Plasmid construction was carried out in Escherichiacoli NEB 10p cells (New England Biolabs,
Ipswich, MA) and protein expression was induced in E. coli BL2 1 DE3plysS (Life technologies,
Carlsbad, CA). All reagents and purified streptavidin (SA) were purchased from Sigma unless otherwise noted.
SPFt-L55P expression and affinity purification
For expression of SPFt, we transformed YPH499 cells with a plasmid, pHVX2G-SPFt-
L55P, previously made (Chapter 2) and inoculated in 1 ml of YPAD medium: 10 g/L yeast extract (BD Biosciences, San Jose, CA), 20 g/L of Bacto Peptone (BD Biosciences, San Jose,
CA), 20 mg/L of adenine hemisulfate, and 20 g/L glucose with 200 tg/ml Geneticin (Life technologies, Carlsbad, CA) for an overnight incubation at 30 C. We then diluted the cultures into a fresh medium at OD 600 ~0.02 and incubated them for 18 to 24 hrs at 30 C. We harvested yeast cells and lysed as described previously (Chapter 2). After the affinity purification with a
Strep-Tactin sepharose column (IBA, Goettingen, Germany), we buffer exchanged the eluted protein and concentrated into the standard assay buffer containing 100 mM 3-(N- morpholino)propanesulfonic acid (MOPS) at pH 7.0, using a spin filter with IOOkDa cut off membrane (EMD Millipore, Billerica, MA). We measured protein concentrations by 660 nm
Protein Assay (Thermo Scientific, Waltham, MA) with BSA as a standard.
Gene construction of T7-tagged streptavidin variant, STm
51 We used the polymerase chain reaction (PCR) with High-Fidelity Phusion master mix
(New England Biolabs, Ipswich, MA) to construct the gene of SA variant that contains N- terminal T7 tag and four mutations (E44V, S45T, V47R, and W120A). We used a plasmid, pSAI
T7-SA W120A (a gift from Dr. Blake Peterson)81 as a template for an inverse PCR to introduce the following three mutations (E44V, S45T, V47R) to obtain a new plasmid, pSAI STm. We amplified the gene encoding STm from pSA 1 STm and subcloned it into NdeI/EcoRI sites of an
E.coli expression plasmid, pT7-7 plasmid (a gift from Dr. Nicholas Reiter), resulting in the pT7-
7 STm.
Expression and affinity purification of STm
To express STm, we transformed E. coli with the plasmid, pT7-7 STm and grown in M9 minimum medium supplemented with 100 pg/ml ampicillin at 37 'C. Once the culture reached
OD 600 -0.8, we induced the recombinant protein expression with 0.4 mM isopropyl P-D-1- thiogalactopyranoside (IPTG) for 4 h at 30 'C. We harvested and lysed cells with BugBuster reagent (EMD Millipore, Billerica, MA) supplemented with protease inhibitor cocktail III (EMD
Millipore, Billerica, MA) and Lysonase TM Bioprocessing Reagent (EMD Millipore, Billerica,
MA) for 30 min at room temperature. Insoluble fractions were removed by centrifugation at
10,000 g for 40 min. The soluble fraction of lysate was used for the affinity purification of STm using T7-Tag Affinity Purification Kit (EMD Millipore, Billerica, MA) according to the manufacturer's instructions. We then buffer exchanged the purified protein and concentrated into the assay buffer. We measured protein concentrations by 660 nm Protein Assay (Thermo
Scientific) with BSA as a standard.
Cloning of yeast expression plasmid containing STm
52 Yeast expression plasmid containing STm was constructed with Zeocin as a selection marker such that it can be used to co-express with SPFt in yeast. We amplified the 1.2 kb fragment containing Zeocin resistance cassette from pPICZA (Life technologies, Carlsbad, CA), digested with BstBI and AatIl, and cloned into pSA 1 T7SA to replace TRP 1 marker, thereby producing the pSAZ T7SA plasmid. A gene encoding STm was amplified from pSAI STm, digested with NheI and XhoI and cloned into pSAZ T7SA to replace T7SA with STm to yield the pSAZ STm plasmid.
Coexpression of SPFt-L55P and STm in yeast
We transformed yeast cells with two expression plasmids, pHVX2G-SPFt-L55P and
pSAZ STm. We first incubated the yeast cells in a rich medium with 2 % glucose and 10 mM
ferric citrate over night to allow SPFt-L55P expression and iron loading. We then transferred the
yeast cells into 2 % raffinose/ 0.1 % glucose medium and incubated for 2 h. We then induced
expression of STm by adding galactose at a final concentration of 2 % and harvested cells at 0 h,
2 h and 4.5 h time points to measure expression levels and make MRI measurements with the
cell pellet.
Dynamic Light Scattering (DLS) measurements
We performed DLS measurements on a DynaPro DLS system (Wyatt Technology, Santa
Barbara, CA), at 30 'C with averaging over 72 acquisitions each with a 2-s integration time. The
laser power was set to 25%. We used 16 pl of protein sample for each measurement and each
sample was measured in triplicates.
Cryo-electron microscopy of purified SPFt-L55P with and without SA
Sample preparation and imaging methods were described in detail in Chapter 2.
Western blot experiments
53 For western blotting experiments, the whole cell lysate samples were prepared from yeast cells freshly harvested after overnight incubation according to the method8 2 developed by von der Haar with a few modifications. Equivalent numbers of yeast cells (2. 1x 108) were resuspended in 100 pil of the lysis buffer and boiled for 10 min. The cell suspensions were neutralized by the addition of 2.5 pl of 4 M acetic acid, vortexed for a minute, and boiled for another 10 min. We then added 25 pL of the loading buffer to the samples and centrifuged them at 10,000 g for 5 min before loading onto a 12% Mini-Protein TGX Precast gel (Bio-Rad,
Hercules, CA). We ran the protein gels at 160 V for 30 min and transferred the separated proteins onto PVDF membranes (Bio-Rad, Hercules, CA) at 100 V for 40 min at 4 C. The membranes were blocked with 5 % fat-free milk in Tris-buffered saline (AMRESCO, Solon,
OH) containing 0.1 % Tween 20 (TBST) for 30 min at 4 C. We washed the membranes once with TBST for 5 min and incubate it with Strep-Tactin-HRP (IBA, Goettingen, Germany) at
1:4000 dilution in TBST for 1 h at room temperature. For imaging expression of STm, we used anti-Streptavidin antibody conjugated with -HRP (Abcam, Cambridge, Eng) at 1:10000 dilution in TBST. After washing the membranes three times with TBST, we visualized the SPFt-L55P and STm bands with a chromogenic substrate Opti 4CN (Bio-Rad, Hercules, CA) according to the manufacturer's instructions. Images of the membranes were taken and processed with ImageJ software for quantitative analysis.
Non-invasive measurements of gene expression levels with MRI
We prepared protein samples and yeast cells in microtiter plates for MRI measurements.
For yeast samples, the cells were washed twice in PBS with 1 mM EDTA and once in PBS. After the last wash, decant the supernatant and load 100 pl of the cell suspension into the wells of a microtiter plate, in which unused wells were filled with PBS. The plate was centrifuged at 1500 g
54 for 3 min and placed in a 20-cm-bore Bruker 7 T MRI scanner. We imaged a 2 mm slice through the cell pellet samples and the field of view was 5 x 5 cm and the data matrices were 256 x 256 points. We used T2-weighted spin echo pulse sequence with multiecho acquisition; repetition
time (TR) was 2 s, and echo time (TE) ranged from 5 ms to 150 ms with 5 ms intervals. Images
were reconstructed and analyzed using custom routines written in Matlab, and the relaxation time
constants (T2) were computed by exponential fitting of the image intensity data.
Measurements of iron content in yeast cells
We used a colorimetric assay"3 developed by Tamarit et al with a few modifications to
quantify the iron content of yeast cells and the purified protein. Details of the method were
described in chapter 2.
3.4. Acknowledgements
We thank D. S. Yun from the nanotechnology materials core facility at the Koch Institute
for technical support with electron microscopy. This research was supported by NIH grants DP2-
OD002114, RO1-NS076462, and RO1-MH103160 to APJ. YM was supported by a Siebel
Scholar Fellowship and a Friends of the McGovern Institute Fellowship.
3.5. Figure captions
Figure 1. Ft-based sensor changes cluster size in the presence of a crosslinker (a) Schematic
diagram of general strategy of aggregation-based dynamic gene reporter consists of two
components: Ft nanoparticles (gray) with Strep-tag II (cyan) and a streptavidin tetramer (SA) as
a crosslinking protein. When SPFt nanoparticles and SA are both expressed in a cell, large
clusters of nanoparticles are formed due to the interactions between Strep-tag II and tetrameric
55 SA. (b) Crystal structures of SA (red) interacting with Strep-tag II (cyan). (c) Formation of Ft nanoparticles aggregates occurs rapidly and visibly. Tubes containing SPFt nanoparticles were mixed with buffer and SA solution. We observed the formation of large clusters of nanoparticles only after they are exposed to SA for a few seconds. (d) Cryo electron micrographs of control and aggregated SPFt-L55P. Scale bar is 50 nm. (e) Dynamic light scattering (DLS) experiments of Ft nanoparticles with and without SA. The hydrodynamic radius of SPFt-L55P nanoparticles increased about 100 compared to the size of single nanoparticle.
Figure 2. Sensitivity range of the Ft-based MRI gene reporter and its time course of aggregation formation (a) Sensitivity range of Ft-based sensor and improvement of sensitivity by mutagenesis. DLS measurements of the hydrodynamic radii of nanoparticles clusters displayed as the normalized values to the maximum data point with SA (empty circle) and STm
(solid circle) as crosslinkers. Fitted curves for each sample are also shown for SA (dashed line) and STm (solid line). STm has higher affinity to Strep-tag II sequence and therefore shifted the titration curve to the lower concentration ranges of STm and results in a more sensitive gene reporter. (b) STm-dependent aggregation of Ft nanoparticles induces changes in R2 relaxation rates (s-). The inset panel shows MRI images of the SPFt-L55P nanoparticles. For (A) and (b), error bars show s.e.m. of 3 independent trials. (c) Time course of SPFt-L55P sensor aggregation measured by DLS. Data points were collected every second and STm was added roughly 10 after the start of the experiment. Error bars show s.e.m. of 7 independent trials.
56 Figure 3. Non-invasive measurements of gene eypression with Ft-based MRI gene repnorter
Ft sensors respond to changes in gene expression by modulating T2 relaxation rates in MRI measurements of yeast cell pellets. The SPFt-L55P nanoparticles were constitutively expressed in yeast cells while STm expression was induced by galactose as shown in western blot images in the top panel. After STm expressed is induced, T2relaxation rate normalized by iron concentration of yeast cells (R2/Fe) expressing SPFt-L55P nanoparticles was significantly higher than that of yeast cells expressing control nanoparticles, PFt-L55P without the Strep-tag II. Error bars show s.e.m. of 3 independent experiments.
57 3.6. Figures
Figure 1. Ft-based sensor changes cluster size in the presence of a crosslinker
a) b)
SA
c) d) e) 1 - SA + SA - SA + SA 100 dspn 10 Ip down 5 10
1 -SA +SA
58 Figure 2. Sensitivity range of the Ft-based MRI gene reporter and its time course of aggregation formation
a) b) _ c) 5 -1 ST added
0.8 0 10 ji!0.6~ 10 0.4 5 f 0.2 0 8 1 0 0.01 0.1 1 10 100 0.1 1 10 0 25 50 75 100 SA(STm)/Ft ratio STmn/Ft ratio Time (sac)
59 Figure 3. Non-invasive measurements of gene expression with Ft-based MRI gene reporter anti-Streptavidin anti-Strep-tagil
n ST+ PFt oc"0 CST + SPFt
2 0 0 h 2 h 4.S h
60 3.7- Supplementary material
DNA sequence of STm: ATGGCTAGCATGACTGGTGGACAGCAAATGGGTCGCGACCAGGAGGCCGGCATCACCGGCACCTGGTA CAACCAGCTCGGCTCGACCTTCATCGTGACCGCGGGCGCCGACGGCGCCCTGACCGGAACCTACGTGAC CGCTCGCGGCAACGCCGAGAGCCGCTACGTCCTGACCGGTCGTTACGACAGCGCCCCGGCCACCGACG GCAGCGGCACCGCCCTCGGTTGGACGGTGGCCTGGAAGAATAACTACCGCAACGCCCACTCCGCGACC ACGTGGAGCGGCCAGTACGTCGGCGGCGCCGAGGCGAGGATCAACACCCAGTGGCTGCTGACCTCCGG CACCACCGAGGCCAACGCCGCGAAGTCCACGCTGGTCGGCCACGACACCTTCACCAAGGTGAAGCCGTC CGCCGCCTCCATCGACGCGGCGAAGAAGGCCGGCGTCAACAACGGCAACCCGCTCGACGCCGTTCAGC AGTAA
Supplementary figure captions
Supplementary Figure 1. (a) Schematic of the DNA construct of T7-tagged streptavidin variant,
STm. T7-tag (yellow) is introduced in the N-terminus of the mutant streptavidin (red) with a three amino acid linker, RDQ. (b) Coomassie stained SDS-PAGE gel showing that affinity- purified recombinant STm protein appears at around 17 kDa, as predicted.
Supplementary Figure 2. Control experiments were conducted with a non-magnetic mutant version of SPFt, SPFt -E94G/K142R. The mutant nanoparticles were constitutively expressed in yeast while STm expression was induced by galactose, as shown in Western blot images in the top panel. There was no significant difference in normalized relaxation rates between control (ST
+ PFt E94G/K142R) and experimental samples (ST + SPFt E94G/K142) at all time points.
61 Supplementary Figure 1.
E44V W120A a) b) SMSTM 37 kDa 0
25
20
15
10
62 Supplementary Figure 2.
anti-streptavidin
anti-Strep-tagli
50 ST + PFR E94G/K142R DST + SPR E94GK142R f~40
N 30
E A 20 0 z 10 0 0 hr 2 hr 4.5 hr
63 4. T2 Relaxation Induced by Clusters of Superparamagnetic Nanoparticles: Monte Carlo Simulations
4.1. Abstract
MRI-detectable sensors for a variety of analytes have been formed by coupling presence or activity of the analyte to the aggregation of superparamagnetic nanoparticles (SPIOs). In each case, the analyte induces or disrupts crosslinks between the nanoparticles; subsequent changes in the spatial distribution of particles in turn induce changes in T2 relaxation rates, which may be measured by imaging. Several studies have shown that aggregation of "ultrasmall" SPIOs often referred to as mononuclear iron oxide contrast agents (MIONs) dramatically shortens T2.
Another study showed that aggregation of larger, -50 nm diameter SPIOs leads to increases in
T2. The relationship between particle size and a single particle relaxivity has been studied
experimentally and theoretically by Gillis, Brooks, Muller, and colleagues. However, T2 relaxation rate induced by a cluster of particles have not been well characterized in terms of its dependence on the cluster geometry and the particle diameter. In an effort to understand the relationship between SPIO clustering and relaxivity in this context, we used Monte Carlo
methods to simulate proton T2 relaxation in the presence of SPIO aggregates of different geometries created by particles with a range of diameters.
64 4.2. Introduction
MRI-detectable sensors for a variety of analytes have been formed by coupling presence or activity of the analyte to the aggregation of superparamagnetic nanoparticles (SPIOs) 84 . In each case, the analyte induces or disrupts crosslinks between the nanoparticles; subsequent changes in the spatial distribution of particles in turn induce changes in T2 relaxation rates, which may be measured by imaging. Several studies have shown that aggregation of "ultrasmall" SPIOs often
85 referred to as mononuclear iron oxide contrast agents (MIONs) dramatically shortens T2 .
Another study showed that aggregation of larger, ~50 nm diameter SPIOs leads to increases in
T243. An approximately linear relationship was observed, with negative slope between T2 relaxivity and cluster diameter over a threefold range.
Differences in the aggregation-dependent T2 changes evoked by nanoparticles of different sizes may be reconciled with theoretical and experimental work of Gillis, Brooks, and colleagues22,8"-8. These authors showed that for small particles, dynamic sampling of the microscopic fields induced by SPIOs prevents the development of NMR phase dispersion among spins diffusing around the particles; this context is referred to as the motional averaging regime
(MAR), and is governed by classic relationships from outer sphere theory 87' 89. For large particles, phase dispersion forms, but may be partially refocused using spin echo techniques; this is referred to as the echo-limited or slow motion regime (SMR)90'9'. For particles of radius rp in a medium characterized by self-diffusion constant D, the magnitude of a diffusional correlation time D = r ID determines the transitions between relaxation regimes. If the echo time (TE
2TCP) is not too short, a relaxation maximum referred to as the static dephasing limit92 is reached between the MAR and SMR, when:
65 TD - (1) where A(Or is the rms angular frequency shift at the particle surface, a material-specific quantity equal to roughly 3 x 107 rad/s for magnetite. In aqueous medium with D = 2.5 x 10-5 cm 2 /s, Eq.
(1) is satisfied for magnetite particles of ~30 nm diameter (rD = 10-4 ms). R2 is maximized for nanoparticles of approximately this size, provided that echo spacings rcep rD are used; for small particles R2 grows in proportion to rD, and for larger particles R2 declines with increasing radius.
Although it is tempting to generalize these principles of single particle relaxivity to clusters of particles, there are potential complications to account for. First, the density, and hence the apparent magnetization, of particle clusters may differ significantly from that of single particles of equivalent size. Second, microstructural differences (shape and surface features) between similarly sized single particles and clusters may distinguish criteria for MAR and SMR relaxation for different particle geometries. Aggregation of superparamagnetic components into
SPIO contrast agents has been explored by Roch et al. , who used analytical approaches to
explain the lower T, relaxivity and higher T2 relaxivity of polynuclear SPIOs, with respect to
MIONs. These authors did not discuss aggregation phenomena specifically relevant to aggregation-based MRI sensors, however. In an effort to understand the relationship between
SPIO clustering and relaxivity in this context, we therefore used Monte Carlo methods to simulate proton T2 relaxation in the presence of SPIO aggregates of different geometries.
66 4.3. Methods
Computations were performed using methods adapted from earlier studies.8'9496. Random distributions of magnetic nanoparticles were created for each choice of aggregate geometry. In
one set of simulations, "isotropic" aggregates of a given number of particles per cluster were
assembled iteratively from single particles, assuming a homogenous probability of attachment of
any new particle to sites on the surface of the growing aggregate. In another set of simulations,
aggregates were assumed to be linear chains. In both cases, overlap between particles was not
allowed, and a fixed interparticle (center to center) distance was used to space each particle from
the particle to which it was attached. Clusters were randomly rotated and placed into a cubic
simulation volume with edges 100 times the diameter of the particles being modeled. The
number of particles in the simulation volume was chosen such that the partial volume of
magnetic material was 3.14 x 10-6. To prevent artifacts due to the finite size of the simulation
volume, the volume was treated as a repeating unit (unit cell) in an unbounded periodic space
with P1 symmetry.
The diffusion of single zero-radius spins through each particle distribution was initially
simulated over intervals of 20-120 ms, in steps of 10-1000 ps. The simulation time of was chosen
such that the total duration was longer than T2 for each set of conditions. At every time step, the
magnetic field and concomitant phase shift experienced by the diffusing spin was calculated by
adding contributions from all particles present in the simulation volume. Each particle was
modeled as an impenetrable sphere giving rise to a longitudinal (z) magnetic field:
B,(d,0)= 4 Yr3 3cos2 0 ) (2)
67 where r is the particle radius, Aor is the rms angular frequency shift at the particle surface, y is the gyromagnetic ratio, d is the distance from the particle center, and 0 is the angle between the z axis and the position where Bz is evaluated. To achieve consistency with earlier studies 22, 86, we used Aor = 2.36 x 107 rad/s and D = 2.5 x 10-5 cm 2/s.
For each set of simulation parameters, phase evolution was computed for 80-100 independent spin trajectories passing through 60-200 independent particle distributions (i.e.
4,800-20,000 trajectories per particle distribution). For simulations of Carr-Purcell spin echo signal, the accumulated phase of each spin was inverted at times t = (1 + 2n) 'Cp, for integer n >
0 and echo spacing 2Tcp. Data presented in the figures were generated using Tcp = 0.5 ms. Phase dispersions present at each echo time were converted to normalized NMR signal intensity
[1(t)/Io] using the formula:
I(t)/Io = (cos [$i (t) (3) where ( ) denotes the average over spin trajectories indexed by i, and where yPi(t) is the phase associated with the ith spin trajectory at time t. R2 values were obtained by linear fitting to the negative logarithm of the normalized signal as a function of time. In all cases, the goodness of fit parameter from curve fitting was greater than 0.98.
Simulations were implemented using C++ routines running on a Linux-based processing cluster at MIT. Further data processing and display was performed using Matlab (Mathworks,
Natick, MA). Error bars are shown for all data points in the figures, and correspond to the standard deviations for multiple relaxation simulations performed in each case (n = 3, unless otherwise noted).
68 4.4. Results and discussion
Compact clusters induce relaxation equivalent to similarly sized single particles
T2 relaxation rates (R2 values) induced by single particles with TD ranging from 1 X 10- to
2.5 x 10-2 ms were predicted from simulations and are plotted in Figure IA. The inverted "V"
shaped curve closely follows findings of earlier studies22,86 and shows asymptotic behavior
corresponding to theoretical predictions of relaxation behavior in the MAR and SMR. Using
identical approaches, the relaxation induced by clusters of two, three, six, and fifteen particles
each was predicted for three particle sizes. For these calculations, clusters were modeled as
compact ensembles with interparticle (center to center) distances equal to the particle diameter.
SPIO particle diameters were chosen from size ranges characteristic of the MAR, the boundary
between the MAR and SMR [TD ~ 104 ms, cf Eq. (1)], and the SMR; particles had TD x 1-, 54
1.6 x 10-4 , and 4 x 10-3 ms, respectively, corresponding to small, medium, and large diameters of
20, 40, and 200 nm, given the diffusion constant we used (2.5 x 10-5 cm 2/s).
Figure IB summarizes R2 values computed for clusters of the sizes we examined. For
comparison with single particle relaxivities, an effective T D was assigned to each cluster
configuration by approximating the cluster radius as
r ~ N/3)
where rp is the radius of a single particle, N is the number of particles per cluster, and the
effective TD rc2 ID. The data confirm some expected effects of particle clustering: aggregation
of particles in the MAR increases R2, aggregation of particles in the SMR decreases R 2, and
aggregation of intermediately sized particles has marginal effect. If the data of Figure 1 A are
69 linearly interpolated to the TD values following from Eq. (4), the corresponding interpolated and simulated R2 values are closely matched, with a correlation coefficient (CC) of 0.99, and a root mean squared deviation (RMSD) of 11%. Slightly worse correspondence (CC 0.98, RMSD
30%) was obtained using an alternative estimation of effective TD for clusters, based on a formula for radius of gyration, applied to the actual cluster distributions used in our simulations:
r ~ i (5)
where i indexes the particles within a cluster and di is the distance from each particle center to the center of mass of the cluster. Estimating cluster radii using a fractal dimension9 7 [exponent in Eq. (4) between 1/3 and 1/2] did not improve the correlation, whether or not cluster R2 values were rescaled to account for the average magnetization and volume fractions of aggregates less dense than single particles [by factors of (re/N"'rP)3 or (NI/ 3 rp/rc)2 in the MAR and SMR, respectively; cf ref. 7]. We expect that modeling clusters as fractal aggregates would be more effective for predicting relaxation by clusters of significantly more than fifteen particles each.
Our analysis here, however, indicates that T2 relaxation by clusters of 2-15 superparamagnetic nanoparticles is well approximated in numerical simulations by the relaxation induced by single particles of equivalent volume and dipole moment, despite the more complex geometrical characteristics of these clusters.
Dependence of relaxation rate on interparticle distance differs between clusters of small and large particles
70 We next used the simulation approach to determine how T2 relaxation induced by aggregated particles is dependent on interparticle distances (dpp). Figure 2 shows calculated relaxation rates for clusters of three or fifteen nanoparticles as a function of dp,, and indicates a significant difference between the behavior of particles in the MAR (panel A) and SMR (panel
B) size ranges. The small and large particles had TD of 4 x 10-5 and 4 x 10- 3 ms, corresponding respectively to diameters of 20 nm and 200 nm, again assuming D = 2.5 x 10~5 cm 2 /s. For clusters of small particles, the relaxation rates declined sharply with interparticle distance, such
that for dpp = 8rp the R2 had decreased approximately halfway back to the single particle relaxation limit. The dissociated single particle limit is eventually reached for dpp > 40rp. For
clusters of large particles, where increases in dp result in higher R2, there are only negligible
changes in relaxivity for dp < 20rp. The midpoint between the relaxation induced by compact
clusters and dissociated single particles is reached for dpp > 50rp, and single particle relaxivity
behavior is achieved for dpp 1 00rp.
These results suggest that the thickness of coatings and the size of macromolecular
conjugates involved in actuating SPIO-based sensors may have a significant effect on the
performance of sensors formed from small nanoparticles (ID << 10- ins). For instance, nucleic
acid segments incorporated into some MRI sensors98' 99 have lengths of roughly 10 nm per 30
duplex base pairs; according to the data of Figure 2A, DNA spacers of 30 bp would decrease the
dynamic range of relaxation changes induced by aggregation of small particles (e.g. MIONs) by
roughly 10%. Conjugated macromolecules on the order of 10 nm are unlikely to affect the
relaxivity changes induced by large SPIO aggregation sensors, however.
The results of Figure 2 also imply significant particle size-dependent differences in the time
scales by which disaggregation of SPIOs might influence relaxation changes. For 15-particle
71 clusters under our choice of D, the midpoints for R2 changes caused by disaggregation of densely clustered 20 nm and 200 nm diameter particles occurred at d,, = 73 nm and 6.9 Pm, respectively
(values obtained by interpolation of data in Figure 2). Diffusion constants (at 20 'C in water) for the small and large particles are estimated by the Stokes-Einstein relation to be 2.1 x 10-7 and 2.1
X 10-8 cm 2 /s, respectively, and can be used to approximate half-times for relaxation changes caused by diffusion limited dissociation of particles from compact clusters:
(6) ri/ 2 (d1/2 -4r )/2D,
where rp is the particle radius, Dp is the particle diffusion constant, and dI/2 is the value of dp associated with half-maximal R2 changes. Eq. (6) and the d112 and Dp values computed above imply that TI/ 2 values for dissociation of 15-particle clusters would be roughly 0.1 ms for small particles and 10 s for large particles, a discrepancy of five orders of magnitude. It should be noted that the actual "off times" for aggregation-based T2 sensors may depart significantly from these diffusion-based estimates, depending on the mechanism by which disaggregation proceeds.
Half times for relaxation changes due to SPIO agglomeration (the reverse of what we consider here) may be estimated using a different approach, and are discussed extensively elsewhere 7.
Anisotropic cluster geometry reduces relaxation by particles in the MAR
We finally applied the Monte Carlo simulation approach to determine to what extent the geometrical arrangement of magnetic particles in a cluster affects relaxation. We focused specifically on a comparison of two cases, both with dp = 2rp: particles packed compactly, with isotropic distribution of attachment points (as in Figure 1 B), and particles associated in "rods,"
72 ciioh thait the nrtirle centers are collinear We considered ensemhes of six narticles ner chuster, again using three particle sizes (TD 4 x 1 1.6 x 10-4, and 4 x 10-3 ms). We found that relaxation induced by isotropic aggregates and rods did not differ significantly when the clusters were composed of large or medium-sized particles, but that linear chains composed of six small particles gave rise to 18 ± 6% lower R2 values than isotropic six-particle clusters. R2 values observed for linear chains of three particles were roughly equivalent to those observed for isotropic clusters; larger clusters were not examined because of computational requirements.
The lower R2 values observed for linear particle chains in the MAR cannot be explained with reference to the radius of gyration of the chains [see Eq. (5)], which is almost two-fold larger for linear chains of six particles, and would erroneously predict that linear ensembles produce higher
R2. Rather, the finding of reduced relaxation by chains of small particles suggests that motional averaging is more efficient around chains than around isotropically configured clusters, perhaps because of the short characteristic distance for magnetic field changes planes perpendicular to the axes of linear chains.
The comparison between relaxation produced by isotropic clusters and linear chains has two implications for MRI sensor design: First, the fact that R2 values were similar for both
particle geometries implies that sensors could be constructed based on agglomeration
mechanisms that result in incorporation of SPIOs into non-compact structures, including fibrils
or extended polymers (e.g. amyloid plaques and microtubules), without significant loss of
sensitivity. Second, the relatively modest difference in R2 between linear and compact chains of
small particles suggests that SPIO-based sensors could be based on reorganization of particles
between these structures (e.g. a linear chain that collapses into a compact ensemble in the
73 presence of a molecular target), but that such sensors would be unlikely to produce large changes in relaxation.
4.5. Conclusions
Monte Carlo simulations modeled relaxation changes induced by aggregation of magnetic nanoparticles, and allowed us to identify properties relevant to the design of SPIO-based sensors for MRI. According to the calculations, compact clusters with dpp = 2rp create T2 relaxation effects similar to single particles containing equivalent amounts of magnetized material.
Increasing the interparticle distance for clusters of small particles in the MAR produces sharp decreases in R2, while increasing dpp for clusters of large particles in the SMR results in only
gradual increases in R2, followed by eventual return to the relaxation rate associated with dissociated particles. The shape of particle aggregates (compact vs. linear) has a modest effect on relaxivity of clusters in the MAR, but negligible effect on clusters in the SMR. We have not attempted to explain the simulation results in terms of relaxation theory. We do note, however, that the relative independence of R2 on the configuration of clusters of relatively large particles
(Figures 2B and 3) is consistent with the idea that interactions within an "inner region" around these particles do not strongly affect observed relaxation rates in the SMR22, 86. For the large particles we considered in our simulations, the inner region is bounded by a radius of roughly twenty times the particle radius [cf Eq. (14) in ref.2 2 ], within which changes in dp or cluster shape are predicted theoretically to make little difference; this prediction is supported by results presented here. R2 values associated with particle aggregates in the MAR are more sensitive to changes in cluster configuration. Our modeling suggests that aggregation-based sensors composed of small particles may be optimized by reducing SPIO coating thickness, and that
74 MRI contrast changes could he generated by manipulating geometric parameters within individual clusters.
4.6. Acknowledgement
This work was generously supported by grants from the Raymond and Beverly Sackler
Foundation and the NEC Corporation Research Fund. We gratefully acknowledge Pierre Gillis for detailed comments on the manuscript, and David Cory for helpful discussions.
75 5. Conclusions and future directions
5.1. SPFt-L55P nanoparticles and the genetic screen
In this thesis, we developed a Ft-based MRI sensor with significantly improved sensitivity through a high throughput genetic screening approach. We demonstrated for the first time that the intracellular mineralization can be controlled at a molecular level. By taking advantage of iron regulatory system of yeast, we were able to select for mutant Ft nanoparticles that efficiently accumulate iron in a cellular environment. In our study, we only executed one round of mutagenesis and selection because we did not observe improvements in the 2nd round, however, in theory this procedure can be applied multiple times to obtain mutants with even better iron mineralization properties. Since the starting sequence of the screen may have a strong impact on the outcome, it may be worthwhile to perform the screening procedure on Ft sequences from various species in the future. Alternatively, this system could be used to improve iron mineralization capability of Ft fusion constructs that show limited iron mineralization function. As demonstrated in Chapter 3, Ft can be made into a sensor by modifying its N- terminus with amino acid sequences that interact with particular molecules of interest. However, in many cases when Ft is fused with an extra domain, it does not accumulate much iron most likely due to structural changes or steric issues. Using our yeast genetic screen, we may be able to obtain mutant fusion Fts that exhibit iron accumulation ability of wild-type. Finally, the genetic screening strategy can be generalized to improve the mineralization capabilities of other types of metalloprotein besides Ft.
Although we demonstrated that the engineered sensor, SPFt-L55P accumulates almost twice as much iron as wild-type nanoparticles and shows significantly higher sensitivity in MRI when expressed in yeast, its relaxivity is still far less than that of similarly sized SPNs. If we
76 could engineer n Ft to spontneouslv acuiimidnte ferromnanitic iron minerak found in SPNs, utility of Ft-based MRI contrast agent for in vivo imaging would increase dramatically. Toward this goal, we developed magnetic cell sorting screen to selectively capture yeast cells that express mutant Fts with higher magnetic moment (see Appendix B).
Potential applications of SPFt-L55P nanoparticles include non-invasively measuring gene expression by MRI for a long-term to track transplanted cells and monitor progression of gene therapy in living subjects. As a first step towards these applications, we have virally delivered the gene encoding SPFt-L55P into the rat brains. Although the project is still in its early stage, we observed a promising sign of successful transfection of brain cells with SPFt-L55P according to immunohistochemistry (IHC) results. In the future, we would like to take MRI measurements of rat brains infected with SPFt-L55P and correlate them with the IHC results at several different time points to show that expression of SPFt-L55P induces changes in MRI contrast in vivo. Once the performance of SPFt-L55P as a genetically encoded contrast agent is validated, we could for example, insert a cell type specific promoter to control the expression of SPFt-L55P to image the distribution of specific types of cells in a live animals with MRI. We could also use it as a gene marker to track transplanted cells.
5.2. Ft-based dynamic gene reporter
In Chapter 3, we showed that the Strep-tagged Ft nanoparticles, SPFt-L55P can be used to non-invasively monitor dynamic changes in gene expression levels in yeast cells via aggregation mechanism. Although the effect was modest (~25% signal changes), this was the first demonstration where aggregation of magnetic nanoparticles induces MRI contrast changes in a cellular environment. Results of this proof-of-principle experiment are promising for its
77 future applications in mammalian cells and live animals. One way to improve the utility of this gene reporter is by improving the sensitivity of the nanoparticles by the various methods mentioned in earlier sections. Another way to improve the system is to make the aggregation process reversible. Currently, SPFt-L55P nanoparticles interact with the crosslinker protein STm and form a large cluster of nanoparticles, but we have not been able to reverse the aggregation. If the sensor could reversibly aggregate in response to the crosslinker expression level, it will improve the temporal resolution of the sensor and also allow us to make measurements on genes whose expression levels fluctuates relatively rapidly.
There are many potential applications with the aggregation sensor. Since the nanoparticles and the crosslinker are both fully genetically encodable, their genes can be virally delivered for a long-term monitoring of dynamic changes in gene expression levels in transplanted cells or endogenous tissues. Moreover appropriate promoter sequences can be coupled to the gene encoding STm, which will result in an aggregation-based sensor that responds to specific cellular environmental such as low oxygen levels, starvation, and oxidative stress.
78 Appendix A: Metalloprotein-based MRI probes (reviewT article)
Abstract
Metalloproteins have long been recognized as key determinants of endogenous contrast in magnetic resonance imaging (MRI) of biological subjects. More recently, both natural and engineered metalloproteins have been harnessed as biotechnological tools to probe gene expression, enzyme activity, and analyte concentrations by MRI. Metalloprotein MRI probes are paramagnetic and function by analogous mechanisms to conventional gadolinium or iron oxide- based MRI contrast agents. Compared with synthetic agents, metalloproteins typically offer worse sensitivity, but the possibilities of using protein engineering and targeted gene expression approaches in conjunction with metalloprotein contrast agents are powerful and sometimes definitive strengths. This review summarizes theoretical and practical aspects of metalloprotein- based contrast agents, and discusses progress in the exploitation of these proteins for molecular imaging applications.
79 Introduction
Metalloproteins are essential to life. Roughly a third of proteins are associated with metals, most frequently magnesium, zinc, iron, and manganese, in order from most to least abundant 100 .
Metalloprostheses enable many of these proteins to play important roles in biological processes, prominently including photosynthesis, respiration, and various oxidation reduction reactions'"1.
A particularly impressive example, the cytochrome b6f from plants and photosynthetic bacteria, contains four types of heme, a magnesium porphyrin, and an [2Fe-2S] cluster all in a single supramolecular assembly designed to drive proton gradient formation using energy from light absorbed by the metal complexes0 2 . Both optical and electronic properties of the metal centers thus contribute to a reaction that ultimately gives rise to the nutrients used by most organisms on the planet. A far simpler but also famously vital metalloprotein is hemoglobin (Hb). Hb contains four polypeptides, each bound to a heme group. Oxygen binding to the metal centers induces a change in the coordination geometry, which in turn drives a global conformation change that favors further oxygen binding 0 2. Via this mechanism, the chemistry of metal-ligand interactions governs the ability of Hb to bind and release oxygen in physiologically appropriate concentration ranges.
Biophysical properties of metalloproteins and protein-metal interactions have led to a number of biotechnological applications. The strength and specificity of metal chelation by polypeptides gives rise to the well-known Ni2 affinity purification method used to isolate polyhistidine-tagged proteins 103 . Catalytic oxidation of chlorinated alkanes by a diiron active center in methane monooxygenase from Methylosinus trichosporium OB3b has been used for bioremediation of contaminated groundwater 104 105. A redox-active metalloprotein, azurin, has
80 been used as a biotransistor106 and component of a protein-based biomemory device107
Metalloproteins are also a basis for monitoring physiology in living animals. The most famous example is provided again by Hb, which due to the oxygen-dependent spectral properties of its heme groups, and the relative transparency of tissue to wavelengths differentially absorbed by
oxy- and deoxy-Hb, is the basis for vital signs monitoring in virtually every hospital in the world'0 ".
It has been recognized more recently that magnetic properties of metalloproteins also
provide important biotechnological capabilities. Many metalloproteins contain ions with
unpaired electrons, rendering them paramagnetic and detectable or manipulable by magnetic
tools. Accompanying the excitement around optogenetics'09 has been a particular interest in
using metalloproteins such as ferritin (Ft) to provide magnetic "handles" on cell function. In a
recent example, implanted cells overexpressing a Ft derivative were induced by application of
oscillating magnetic fields to secrete insulin in mice5 5 . There is also a suggestion that protein-
catalyzed paramagnetic metal accumulation in cells could be used for magnetic pull-down
assays , 1m. Finally, paramagnetic metalloproteins have been used as contrast agents for
magnetic resonance imaging (MRI). Here, the prospect of finding MRI-detectable analogs to
green fluorescent protein (GFP) has been an important inspiration; a hope is that suitable proteins
could be used as gene reporters and sensors analogous to the various fluorescent and luminescent
proteins that have transformed research in cell biology over the past two decades.
Although the majority of MRI contrast agents have historically been based on small
3 23 organic molecules such as Gd chelates , 24, metalloprotein MRI contrast agents present a
variety of advantages, each of which may be important in distinct contexts. Thanks to the
advancement of molecular biology techniques, proteins are much easier to synthesize and modify
81 than organic molecules; numerous protein engineering techniques may be applied to tune metalloprotein properties for MRI (reviewed in25 ). Due to their larger size, protein-based contrast agents are retained in the blood pool for a longer time than small molecule agents, allowing for
26, 27 longer imaging times in some types of experiments 27. In some cases, metalloproteins can indeed be targeted and expressed using gene delivery methods, a la GFP 76, 112. Such reporters generate relatively static contrast, but can allow particular types of cells to be tracked in vivo over time39 40, 113. By furthermore sensitizing metalloprotein contrast agents to analytes" 4' "5, additional information about physiologically relevant signals can be obtained. Development of such environmentally-sensitive agents is underway but largely in its infancy. In this article, we discuss the characteristics of metalloprotein-based MRI contrast agents and review recent progress in the development and applications of magnetically active proteins in these new spheres of investigation.
Theoretical basis of MRI contrast agents
In proton MRI 2, the form of MRI most commonly applied in laboratories and clinics, populations of nuclear spins arising from hydrogen nuclei primarily in water are perturbed and monitored to generate images. At thermal equilibrium in a strong magnetic field (Bo), the water proton spins align weakly with the applied field and give rise to a net "longitudinal" magnetization aligned with Bo. This magnetization is unobservable, but can be detected following application of radiofrequency energy pulses which tilt the magnetization vector off of the BO axis and give rise to a nonzero "transverse" magnetization component. After excitation, the transverse magnetization component decays away with a time constant T2 (the transverse relaxation time) and the overall magnetization returns to thermal equilibrium with a time
82 constant T, (the longitudinal relaxation time)- The shorter T, is- the more frequently an MRI signal can be repeatedly measured per unit time; areas of a specimen with short T, therefore give rise to a larger average MRI signal. Conversely, the shorter T2 is, the more rapidly the observable component of magnetization disappears, and the lower the MRI signal becomes.
Paramagnetic species influence contrast in MRI by reducing the T, and T2 relaxation times4'5 (Figure 1). In both cases, accelerated relaxation arises from coupling between the magnetic dipole of the contrast agent and the nuclear spins of water molecules that interact with the agent through bonds ("inner sphere" interactions) or through space ("outer sphere").
Relaxation rates R1 (= 1/T,) and R 2 (= 1/T2) are generally linear with contrast agent concentration. The slopes of these relationships are referred to as the T, and T2 relaxivities, ri and r2, respectively, which measure the strength of the contrast agent and are expressed in units of mM' s -. Greater relaxivity is beneficial to an MRI contrast agent, because the agent can then be applied at lower doses or to greater effect at any given concentration. Both ri and r2 vary
strongly with Bo and depend on physical parameters including the electron spin number (S) or magnetic moment of the contrast agent, the number of coordinated inner sphere water molecules
(q), the time constant for inner sphere water exchange (rM), and the rotational correlation time of the agent (rR). Inner sphere contributions to relaxivity are described by the theory of Solomon,
Bloembergen, and Morgan4' 6'8 , and apply to metalloproteins with adjustments to account for
slow rotation in macromolecules9 . Outer sphere contributions are described by related theories
applicable to mononuclear 10 and particulate 1 -1 3 metal complexes. Determinants of relaxivity are
summarized in the appendix to this article, and are also thoroughly discussed in a number of
secondary references14, 15, 116, 117. Relaxivity determinants are important not only because they
83 explain how contrast agents may be optimized, but also because they provide potential mechanisms for designing MRI-detectable sensors.
Most MRI contrast agents or sensors, including paramagnetic proteins, tend to affect T- weighted MRI scans more than T2-weighted scans, or vice versa, and are correspondingly referred to as T or T2 agents. T1 agents have an ri/r2 ratio of 1-2 and generally contain one or a small number of paramagnetic ions. Classical T agents are exemplified by complexes of Gd3 with small chelators like diethylenetriaminepenaacetic acid (DTPA)16' 17, and are the most commonly applied agents in clincial MRI; analogous proteins can include porphyrin prostheses or directly bound metal ions (Figure IA). At typical field strengths for clinical or preclinical MRI
1 (> 1 T), most T agents have r1 values from 1-10 mM's- . In biological samples with T1 values typically near 1 s, T, agents need to be applied at concentrations near 100 pM to induce substantial contrast effects; 100 pM of a contrast agent with r1 = 5 mM1 s~I, for instance, would induce an MRI signal increase of ~20% against a background with a T of 1 s under conditions of optimal signal to noise ratio. Inner sphere r1 contributions are usually particularly important for
T1 agents, and Ti-based sensors typically undergo analyte-dependent changes in the parameters that most affect inner sphere relaxation mechanisms, such as q and rR.
T2 agents have r2/rl ratio greater than ~10 (a necessary condition because T2 values are much shorter than T1 values in vivo) and are best exemplified by superparamagnetic
18-20 nanoparticles (SPNs) -2, contrast agents that incorporate discrete crystalline domains that exhibit highly cooperative magnetic behavior. A biosynthetic analog to SPNs is ferritin (Ft)13 , an iron storage protein that accumulates minerals in a 12 nm shell like structure formed from 24 polypeptide chains (Figure 1 B). SPN T2 agents produce high r2/rl ratio because they become magnetized by the BO field and create microscopic magnetic perturbations experienced by
84 diffusing water molecules in solition These perturbations affect T2 relaxation more than T, particularly for particles with highly magnetic mineral cores over ~3 nm in diameter and at high
Bo strengths (> 1 T), where SPN magnetizations tend toward an asymptotic "saturation" point2'
The physics of the interaction between diffusing protons and SPNs also leads to a strong dependence of r2 on R2/D, where R is the mineral core radius and D is the solvent self-diffusion constant (see appendix); SPNs with larger size shorten T2 more effectively, up to a so-called static dephasing limit2 2 near -50 nm for typical SPNs. Most SPN contrast agents incorporate iron oxide; synthetic iron-containing SPNs usually have r2 values of 50-500 (mM Fe)ls-1. Given typical background T2 values near 100 ms in tissue, synthetic SPNs applied at concentrations of
1-10 ptM Fe can produce substantial effects under optimal T2-weighted imaging conditions. For instance, an agent with r 2 = 200 mM's-1 could produce a -20% decrease in MRI signal at a concentration of 10 tM Fe. Because of the dependence of r2 on particle size for SPN agents, sensors can be constructed by coupling analyte concentration to the clustering of these particles 98,118 , which results in defacto size changes"19
Relaxivity of protein contrast agents
Protein contrast agents have distinct advantages and disadvantages compared with
conventional synthetic contrast agents in terms of relaxivity. In most cases, proteins are handicapped by incorporating metal ions with electronic properties that are suboptimal for R, or
R2 enhancement 120. Naturally-occurring paramagnetic proteins tend to contain Cu2+, Mn2+
Mn 3± , Fe 2± , or Fe 3±*ions, ranging in spin number from 1/2 to 5/2. On the other hand, Gd 3+ ions
used in most synthetic small molecule agents provide S= 7/2. Since relaxivity values are
approximately proportional to S(S+1), the maximum relaxivity of a gadolinium agent is about
85 twice what would be achieved in principle with a transition metal-containing protein with otherwise equivalent relaxivity parameters. Electronic relaxation times Tie and T2e have a great influence on inner sphere relaxivities and also tend to be less advantageous (shorter) for transition metals than for gadolinium. Short electronic relaxation times limit relaxivity when they are less than the molecular motion timescale TR (10 ns), as has been reported for some metalloproteins. A further limitation on the relaxivity of metalloproteins, compared with small metal complexes, can arise from the relative inaccessibility of protein-coordinated metal ions to outer sphere water molecules. At field strengths above 1.5 T, outer sphere interactions account
14 for a sizeable fraction of the r1 of typical small molecule contrast agents '121, a contribution that could be reduced due to steric effects in a macromolecule.
The limitations of metalloprotein contrast agents are partially offset by properties that are predicted to benefit relaxivity. At moderate magnetic field strengths and for molecules with sufficiently long electronic relaxation times, inner sphere relaxivity tends to be greater for molecules with longer TR 117 Approximate TR values can be estimated by the Stokes-Einstein relationship 122 and are proportional to molecular size; values of 10 ns or above are typical of proteins (> 20 kD), whereas small molecules generally have TR less than 1 ns. Water structure around metalloproteins may also be conducive to greater relaxivity. Many paramagnetic proteins provide more than one water-accessible coordination site per metal ion (q > 1), compared with q
= 1 for conventional Gd3+-containing contrast agents; this is a substantial benefit because inner sphere relaxivity scales with q. Although complexes with higher q bind metal ions less tightly, the potential health risk from complex dissociation is mitigated by the fact that naturally abundant transition metals are far less toxic than Gd3 at low doses. The presence of additional bound water molecules associated with metalloproteins, including "second sphere" waters near
86 hut not directly coordinated to paramagnetic ions may confer a further relaxivity gain 123 A finl relaxivity-related advantage of protein contrast agents over some small molecule agents is their
relative solubility. Most cytosolic or secreted metalloproteins are evolved to remain in solution
or interact with well-defined ligands; this limits the potential for biological environments to
substantially degrade relaxivity or analyte sensing capabilities by adversely affecting water
proton interaction parameters.
The preceding discussion of advantages and disadvantages relates most directly to
synthetic vs. metalloprotein T1 agents, but analogous criteria differentiate synthetic T2 agents
(SPNs) from Ft as well. Synthetic iron oxide SPN contrast agents contain magnetite or
maghemite, magnetic materials with saturation magnetization (Ms) values of 92-100 or 60-80
emu/g, respectively 42, whereas Ft naturally contains a hydrated iron oxide called ferrihydrite,
which has a reported Ms of only 0.9-1.2 emu/g 41 . Because the expected T2 effects of iron oxide
cores are proportional to their magnetization, the relaxivity of Ft at saturating Bo is in principle
only about 1% that of synthetic iron oxides for equivalent core sizes. The situation with
ferrihydrite-loaded Ft may be more complex, however, as its r2 appears to increase linearly with
field at least up to 11.7 T, as opposed to reaching a saturating value around 1 T like most SPNs 1 .
Further, it has been shown that Ft r2 is more directly dependent on stored iron concentration than
on core size per se 57, suggesting an important role for inner sphere relaxivity mechanisms akin to
those of conventional paramagnetic, as opposed to superparamagnetic, contrast agents. Practical
advantages of Ft compared with synthetic SPNs include its regular structure and predictable size,
both of which can aid in characterization or engineering the relaxivity of Ft-based contrast
agents.
87 Natural metalloproteins in MRI
The first substantial investigation of magnetic properties in a metalloprotein centered on hemoglobin, the oxygen transporting heme protein in blood. Pauling and Coryell reported in a
1936 paper that hemoglobin is paramagnetic in the absence of ligand but diamagnetic when bound to oxygen 2 4 . This discovery eventually led to the development of blood oxygenation level dependent (BOLD) technique for functional neuroimaging (fMRI) studies,2 5, which is by far the most commonly exploited example of metalloprotein-induced contrast in MRI. In the BOLD effect, T2-related signal changes are produced by variations in the oxygenation of hemoglobin in blood vessels. Although deoxyhemoglobin can act as a contrast agent through direct interactions with water molecules, its dominant contribution in the BOLD effect is to change the overall magnetic susceptibility of erythrocytes and blood vessels in their entirety, converting these structures into cellular-scale "contrast agents" that alter MRI signal by outer sphere effects analogous to those of SPN agents94 . The BOLD effect is enabled by the large concentration of hemoglobin, -150 g/L in whole blood126, giving rise to -4 mM Fe, which ensures that even a relatively low percentage of deoxygenation translates into a formidable deoxyhemoglobin concentration. In BOLD fMRI, brain activity-induced increases in the blood supply to affected areas induces transient drops in the concentrations of deoxyhemoglobin, which are detectable as localized MRI signal increases27129. BOLD imaging has been applied to detect hemodynamic changes in other tissues as well'13, 131, and a BOLD-like effect produced by paramagnetic deoxymyoglobin in muscle can also be used to monitor aspects of muscular physiology 32
More recently, interest in exploiting natural metalloproteins for MRI contrast has revolved largely around the potential for expressed paramagnetic proteins to act as gene reporters. An early effort to express myoglobin in transgenic mice did not produce substantial
88 MRI contrast changesP33 bit contrast changes have been achieved by overexpressing Ft in rells and animals. Ft has been reported to bind up to -4500 Fe atoms per 24-mer 29, providing a stoichiometric advantage of over a hundred, compared with heme proteins, in terms of sheer iron accumulation. Although many of the iron atoms in Ft are organized into "antiferromagnetic" ferrihydrite domains 33 , the limited per-iron relaxivity of Ft is still compensated for by the number of atoms that can be stored. The effect of overexpressing Ft was first demonstrated in C6 glioma cells transfected with murine heavy chain Ft (one of two isoforms, heavy and light, expressed in mammals) 34. A 2005 study then showed for the first time that ectopic (adenovirus
driven) Ft overexpression leads to detectable T2 contrast in rodent brains3 5, and later papers
37 39 40 reported T2 effects in transgenic mice and transplanted cells in vivo , , 113
A requirement for efficacy of the Ft reporter approach is that highly regulated
endogenous iron dynamics must be effectively perturbed by overexpression of the metalloprotein
without toxic effects on cells; this criterion might be satisfied to varying extents in different
tissues or cell types. Indeed, as potential gene reporters, metalloproteins in general are limited by
the kinetics of metal ion transport and processing. Iron uptake by cells, for instance, depends on
the import rate and availability of extracellular iron sources, and can require exposure times on
the order of hours to achieve saturation 134-136. To address the potentially restrictive role of iron
import machinery, one study proposed co-expressing Ft with the transferrin receptor, a natural
participant in cellular iron transport 36. A conceptually related strategy involves using
overexpression of paramagnetic metal ion transporters themselves to induce MRI contrast;
contrast changes in rodents have been reported using the magnetotactic bacterial protein MagA5 3
and the mammalian divalent cation transporter DMT 1137, but it is not clear which specific
metalloproteins might be involved in binding the metal ions once they have entered cells.
89 Engineering protein-based contrast agents to detect biological targets
One of the greatest strengths of protein contrast agents with respect to small molecules is in the area of biomolecular target detection, the objective of "molecular imaging." The large surface area and numerous functional groups presented by protein interfaces render these macromolecules especially suited to binding potential ligands with high affinity and specificity.
Although the potential utility of natural metalloproteins for MRI is far from fully explored, however, it is unusual to be able to find a naturally occurring paramagnetic molecule that spontaneously fits the demands of a particular biosensing application. Here another strength of proteins-their amenability to engineering approaches-comes into play. Whereas modifying small molecule agents often requires complicated synthetic schemes, proteins may be engineered using simple DNA-level changes or straightforward post-translational approaches.
The simplest examples of how proteins may be modified to facilitate analyte detection include the use of bioconjugation strategies to attach protein targeting domains to metal compounds, creating metal-protein hybrids with desirable binding and relaxivity properties. This approach was first applied to attach Gd3 ions to antibodies esterified to DTPA or similar chelating groups138-140, in the expectation that the resulting complexes could be used to home in on antigens in vivo and facilitate their visualization by MRI. A recent achievement using this type of approach involved the modification of the neuronal tract tracer cholera toxin subunit B with Gd3+-tetraazacyclododecanetetraacetic acid (Gd-DOTA) complexes141. The resulting molecules were injected into rat brains, where they were taken up by neurons and transported to distal areas connected to the injection site, exposing patterns of connectivity in the brain. Protein bioconjugation strategies have also been extensively applied in combination with SPNs, which
90 provide substantially greater contrast per metal atom than paramagnetic T, agents' . Individual
SPN particles can be attached to multiple targeting proteins, adding the potential for improved targeting through avidity effects. SPN-protein conjugates have been produced to target genetically-expressed reporters44, markers of apoptosis 142, vascular pathology1 43, and cancer
44 cells1 ' 145. Analyte sensors have also been actuated by proteins that bring about reversible
clustering of SPNs in the presence of various targets43 146. In one instance, a calcium ion sensor was made by attaching SPNs to two proteins that bind to each other in the presence but not the
absence of Ca2+43. Calcium-dependent SPN aggregation was observed and response properties
could be tuned by further engineering of the protein domains.
Canonical metalloproteins have also been engineered to act as MRI probes. Heme
proteins are particularly attractive bases for MRI sensors because of their high metal affinity and
diverse ligand binding functionality, but their properties typically need to be adjusted. For
instance, Hb is the quintessential metalloprotein-based sensor , but its use as an explicit probe
for measuring tissue oxygen pressure (p02) by MRI is hindered by the fact that its natural
binding midpoint (EC5o) is near a PO2 of 8 mmHg, below concentrations normally found in
tissue. Crosslinking Hb with glutaraldehyde improves its stability and shifts to the EC5o to 38
47 mmHg1 , allowing it to sense physiological relevantpO 2 levels in tissue with a Ar 2 of ~7 mM's-
1 at 14.1 T 148 . In a far more generalizable example of heme protein engineering for MRI, a
bacterial cytochrome P450 heme domain (BM3h) that normally binds unsaturated fatty acids was
retuned through the process of directed evolution149 to bind and sense neural signaling
molecules" 4 in which binding of the analytes competes with inner sphere water to bring about a
change in q (Figure 2A). Over five rounds of random mutagenesis and selection based on an
optical titration screen, BM3h variants were obtained with micromolar affinity for dopamine, a
91 neurotransmitter involved in reward-related signaling in the brain. These sensors had Ar of 1 mM-1 s~1 at 4.7 T (Figure 2B), a modest change, but one that nevertheless permitted detection of stimulus-induced dopamine release in rats. The same directed evolution strategy could also be applied to generate sensors for other targets, such as the neurotransmitter serotonin" 4
Ft is an obvious platform for engineering MRI probes because of its similarity to SPNs and the evidence that endogenous- or ectopic35,37 Ft expression affects MRI contrast in vivo.
The most natural way to engineer Ft without disrupting its metal storage functionality is to
"display" targeting moieties on the protein surface. In one example of this approach, Uchida et al. introduced a tumor targeting peptide, RGD-4C, at the N-terminus of human heavy chain Ft and confirmed that the engineered protein (RGD4C-Fn) can still mineralize iron oxide in its cavity 13. RGD4C-Fn was shown to bind melanoma cells better than unmodified Ft. The solvent exposed Ft N-terminus can also be modified to create responsive MRI contrast agents. Following this strategy, Shapiro et al. developed a protein kinase A (PKA) activity sensor based on Ft, which produces aggregation-dependent T2-weighted MRI contrast changes analogous to those obtained using synthetic SPNs 65 . The PKA sensor contains two populations of engineered Ft particles, one fused with the kinase inducible domain (KID) of the transcription factor CREB, and the other with KIX domain of the protein CBP. Upon phosphorylation of the KID domain by
PKA, KID and KIX domains interact with each other1 5 4 and induce clustering of KID-Ft with
KIX-Ft, increasing the r2 of the sensor (Figure 2C-D). These sensors have not yet been applied in vivo, but may have the potential to allow genetically targeted functional imaging of cell or tissue types because of the all-protein nature of the Ft-based sensors and the important role of kinases in mediating cell signaling processes. Responsive Ft derivatives have also been constructed by chemically crosslinking Ft to reversible polymerizing domains77.
92 Engineering metalloproteins for high relaxivity
The chief limitation of engineered metalloproteins for targeting, sensing, and gene reporting applications in MRI is their low relaxivity. For this reason, there is great interest in engineering metalloproteins with higher r1 and r2 than natural proteins. One approach to this problem is to mutate metalloprotein polypeptide sequences and examine effects on relaxivity. In one instance of this approach, a library of P450 BM3h domains selected for neurotransmitter
1 4 1 affinity was examined for r 1 variation . A range of values from 0.7 to 1.9 mM s- at 4.7 T was found among mutants that differed primarily in residues near the ligand binding site proximal to the heme. The shapes of so-called nuclear magnetic relaxation dispersion (NMRD) curves, which plot relaxation rate as a function of Bo field strength, combined with X-ray crystallographic
analysis, suggested that r1 differences arose from minor variations among multiple relaxivity
determinants, with subtle changes in the metal-proton distance for inner sphere water (2.6-3.0 A)
having greatest effect on r1 (Figure 3). Unfortunately it was not possible to verify the inferred
distance changes at the resolution of the crystal structures, but it could be possible in the future to
apply rational design principles to bring about similar changes. The NMRD results more
generally indicate that nontrivial enhancement of metalloprotein relaxivity is possible through
mutagenesis and screening-based approaches. A gain in the MRI contrast induced by human Ft
expression has also been achieved by amino-acid level changes. Functionality of mammalian Ft
is normally quite sensitive to the balance between the Ft heavy and light chains. lordanova et al.
fused H and L chains together and showed that the resulting chimera improved relaxation rates in
U2OS cells were improved by roughly 50%64. Although it is unclear whether any difference in r2
93 per iron or per Ft was achieved, cells harboring the construct appeared to contain significantly more iron than cells expressing Ft heavy and light chains separately.
Some of the most severe limits on metalloprotein relaxivity come from the characteristics of naturally bound metal ions themselves, and in particular from the spin numbers (S < 5/2) of bound transition metal ions. For this reason, a second strategy for creating metalloproteins with enhanced relaxivity is to engineer the metal content of proteins, rather than simply modifying their polypeptide sequences. This strategy was applied to BM3h, which normally contains a low
3 3 spin (S= 1/2) Fe ion 155. By substituting the native ferric heme with a Mn + protoporphyrin complex (S= 2), a gain in relaxivity by a factor of 2.5 was obtained. In principle, the strategy should have resulted in an eight-fold improvement in ri, all else being equal, suggesting that further relaxivity enhancements might be possible by using mutagenesis approaches in parallel to
3 metal substitution. Non-native Mn + incorporation into BM3h was made possible within bacterial cells by coexpressing BM3h with a porphyrin transporter called ChuA. This facilitates large scale production of metal-substituted protein in bacteria, as well as possible applications requiring intracellular compartmentalization of the Mn 3+-containing protein variants.
Metal substitution has also been possible with Ft, which can be engineered to contain mineral species with higher magnetization than the natural ferrihydrite core material. The most impressive example of this approach has been the creation of "magnetoferritin," an Ft complex
5 6 in which magnetite is mineralized in the protein shell , 156, 157. This can be accomplished by incubating purified apoferritin under controlled pH and oxygenation conditions in the presence of iron salts, and results in particles with a reported r 2 of 78 (mM Fe)~'s~1 at 1.5 T and 37 'C, comparable to synthetic SPNs and two orders of magnitude higher than the r2 of ferrihydrite- loaded Ft. Yet higher relaxivity has been reported following mineralization of gadolinium oxide
94 (r2= 240 mM s-I at 1.5 T) 15 in Ft variants, and moderate T, relaxivitv has been reported
following mineralization of manganese oxyhydroxide (ri = 6 mM- s at 0.47 T) 159. This strategy has also been adapted to convert viral capsids into protein-based MRI contrast agents 160-162
Finally, Ft has also been used as a compartment for entrapment of soluble Gd-DOTA-related T1
agents163. This approach takes advantage of the protein's shell structure and availability of
exchangeable protons at the protein surface, but not its mineral nucleation capabilities per se.
1 Very high ri values, near 80 mM s 1, were reported at 0.47 T. A variant of this strategy was also
2-i- 164. used to entrap Mn ions in a Ft variant engineered to contain metal ions better than native Ft
documented r, and r2 values were 10 mM- s and 74 mMI s 1, respectively, at 3 T. A
disadvantage of most metal substitution approaches for Ft and other proteins is that they cannot
be implemented in cells expressing a metalloprotein, with examples like the BM3h/ChuA
strategy being occasional exceptions. Metal substituted proteins could nevertheless be useful
contrast reagents for exogenous application in MRI experiments, however.
A different technique for engineering metalloprotein contrast agents is to graft metal
binding functionality onto proteins or polypeptides that do not normally bind metals. One version
of this route was taken with the gadolinium compound MS-325, which was designed to bind
noncovalently to serum albumin in the bloodstream 26 . The resulting complex has higher
relaxivity at B0 values because of the long TR of the albumin protein. Karfeld et al. took a
somewhat different approach by conjugating a large number of gadolinium chelators covalently
165, 166 via lysine side chains in a family of engineered repetitive polypeptides 6. The authors found
that the amino acid sequence of the polypeptide could be modified to produce substantial
changes in Ti relaxivity. A parent sequence with -22 kD molecular weight bearing eight Gd3
per molecule displayed a per gadolinium ri of 8.8 mMI s at 1.4 T, but bioconjugates with
95 denser spacing of conjugation sites and more Gd3+ ions per polypeptide achieved r, values up to
12.1 mM's'. Relaxivities of up to 14.6 mM- s could also be obtained by increasing the molecular weight of the complexes up to -80 kD.
Polypeptide contrast agents that chelate lanthanides without the need for bioconjugation have also been devised. In one example, an immunoglobulin domain was mutated to introduce
167 3 multiple acidic sidechains in a cluster at the protein surface . One variant formed a Gd + complex with a reported r, = 117 mM-Is-I at 1.5 T, and with an apparent Kd < 1 pM and q = 2.
Several efforts to construct artificial metalloprotein complexes have made use of the EF hand motif 6 8 , 169, a sequence derived from calcium binding proteins such as parvalbumin. Caravan et al. inserted an EF hand motif into a DNA-binding helix-loop-helix motif to create a chimeric
3 DNA-sensing contrast agent . The Gd + complex has reported r1 values of 21 and 42 mM s at
1.4 T in the absence and presence of DNA, respectively. In another study, the inherent promiscuity of metal binding by the EF hand motif was used to generate a sensor that functions by calcium-dependent displacement of Mn 2+ ions associated with the protein calmodulin'77 1 .
Longitudinal relaxivity values of 11 and 8 mM-1s-1 at 4.7 T were reported in the absence and presence of 1 mM Ca 2+, respectively. To improve the affinity and specificity of EF hand motifs for lanthanide binding, another group used a luminescence assay to screen libraries of troponin-
3 172 derived peptides for Tb binding , 173. T1 relaxivities were subsequently measured from Gd3+ complexes of several variants of an optimized peptide sequence174 . Values of up to 5.9 mM~ls- at 14 T were observed, despite an apparent q = 0 for the sequence with the highest relaxivity.
Several of the lanthanide binding tags also retained r1 values of 2.3-5.0 mM' s - when fused to the small protein ubiquitin, and crystallographic analysis of the fusion proteins reveals that the
Gd 3 -binding peptide domains are capable of forming q = 1 chelate complexes similar to Gd-
96 D10TA (Figure 4). Although r, values of these domains do not exceed those of conventional MRI contrast agents, the ability to incorporate high affinity Gd 3 -binding moieties into genetically encodable proteins may facilitate application of protein engineering techniques to further enhance relaxivity or target analytes of interest.
Conclusions
The continued development and exploitation of metalloprotein contrast agents for MRI poses unique opportunities in the fields of molecular imaging and protein engineering. It is still uncertain what applications protein-based MRI probes will be most successful for, but utility for measuring macromolecular biokinetics and sensing a variety of ligands in vivo seems within reach. Only metalloproteins that do not require chemical modification or artificial transmetallation are applicable as endogenously expressed reporters, but efforts to improve relaxivity of these agents may ultimately yield valuable biotechnological tools, perhaps alongside recently-engineered diamagnetic proteins detectable by chemical exchange saturation contrast 175
Further analysis of structure-activity relationships in metalloprotein contrast agents will be interesting from a basic chemical perspective, and could also guide design of synthetic contrast agents with enhanced properties. In each of these contexts, the array of powerful genetic techniques for engineering and expressing proteins will be a tremendous benefit, and helps make continued research in this area a potential rewarding investment.
Acknowledgements
The authors were supported by NIH grants DP2-OD002114, RO I-DA028299, and RO 1-
NS076462 and DARPA grant W91 1NF-10-0059.
97 Figure captions
Figure 1. Mechanisms of MRI contrast enhancement by paramagnetic metalloproteins. (A)
Protein-based T, contrast agents operate largely through an inner sphere relaxation mechanism dependent on metal-coordinated water molecules in fast exchange with bulk solvent. In the example shown (top), the heme iron (brown) of an engineered P450 BM3 variant (gray ribbon structure) interacts with an axial inner sphere water ligand (blue ball, arrowhead) 1 4. Bound water protons (not shown) experience dipole coupling with the metal ion and undergo relaxation.
T1 relaxivity of the complex may also be promoted by the presence of second sphere water molecules (additional blue balls) that participate in weak dipole coupling with the iron atom and exchange with water in the inner sphere position. The T, contrast affects the saturation of MRI signals during repeated application of a pulse sequence (schematics at bottom; pulses in gray and raw MRI signal in black). In the absence of the contrast agent, the MRI signal tends to decline due to saturation over time (top trace). The T1 agent relieves this effect (bottom trace), resulting in a enhanced signal and relative hyperintensity in images. (B) The best-characterized protein T2 contrast agent is ferritin (Ft, ribbon structure), a shell-shaped oligomer of 24 subunits (one shown in purple) that contains a paramagnetic hydrated iron oxide core. The core induces a dipolar field perturbation (field lines shown) over a length scale comparable to the core diameter, and water molecules (blue balls) undergo transverse relaxation by diffusing through the dipole field. The amplitude of an MRI signal obtained using a T2-weighted pulse sequence (schematics at bottom) depends on the amount of T2 relaxation that has occurred prior to acquisition of the signal with each repetition of the pulse sequence. A T2 contrast agent like Ft promotes T2 relaxation and
98 leads to attenuation of the signal (bottom trace) and relative hypointensity in MRI scans. Ft structural model from reference 176
Figure 2. Engineered metalloprotein-based MRI sensors. (A) Metalloprotein-based T, contrast agents can sense analytes via a so-called "q-modulation" mechanism. In this mechanism,
exemplified by BM3h dopamine sensors, inner sphere water molecules bound to the paramagnetic center in the ligand free structure (left) are displaced by analyte binding (right).
For the BM3h-based sensors, neurotransmitter binding reduces q of the heme iron atom from one
to zero. (B) The ligand dependent change in q induces a sharp drop in the r1 of BM3h variants,
from roughly 1 to 0.2 mM's'1 for the best two dopamine-binding variants, BM3h-B7 and -8C8,
identified in reference" 5 . The relaxivity decrease upon dopamine binding also leads to a
reduction in the corresponding Ti-weighted MRI signal intensities (inset). (C) Metalloprotein-
based T2 sensors can be constructed by modifying protein domains in Ft to include analyte-
sensitive moieties. In the example of reference 6, a kinase-sensitive Ft-based probe was
constructed by genetically fusing the KID domain of the protein CREB and the KIX domain of
the protein CBP to Ft light chain to make chimeric KID-Ft (blue) and KIX-Ft (magenta) variants.
In the presence of protien kinase A (PKA), KID domains are phosphorylated and tend to bind
KIX domains, leading to clustering of the multivalent KID-Ft and KIX-Ft proteins. (D) Kinase-
dependent Ft clustering leads to a change in per-particle r2 values. Relaxivities measured from
prephosphorylated KID-Ft (pKID) mixed with KIX-Ft or from KID-Ft mixed with KIX-Ft in the
presence of PKA (middle two bars) are approximately twice the r2 values measured from KID-
Ft/KIX-Ft in the absence of phosphorylation (left bar), or in the presence of the ATP phosphate
99 source but not the kinase (right bar). Corresponding MRI image intensities are shown in the inset.
Figure 3. Nuclear magnetic relaxation dispersion of BM3h variants. (A) Mutant low spin (S
1/2) BM3h proteins show differing T, relaxivities as a function of magnetic field strength' 4 .
NMRD curves were fit to a Solomon-Bloembergen model equation with an iron proton distance. r, water exchange time constant um, and field-independent electronic relaxation time rs. Best fit values for five protein variants are shown in the inset, along with corresponding measured relaxivity values (circles) and fitted curves (solid lines), color coded by mutant. Substantial relaxivity variation among mutants is apparent, showing that relaxivity changes are accessible by mutagenesis of the metalloprotein, even without changing the nature of the bound metal complex. (B) Location of amino acid substitutions in the BM3h variants of panel A. Ca positions of mutated residues are denoted by blue balls in the protein backbone trace (gray) 77 , with the heme shown in pink and the native fatty acid ligand shown in black. Color-coded dots denote which residues are mutated with respect to the wild type protein in each of the variants listed in panel A. Mutations were selected to alter ligand binding near the heme site and most are clustered in the ligand binding region of the protein.
Figure 4. Structures of lanthanide binding sites. (A) Structure of europium-DOTA, a compound isomorphous to the canonical contrast agent Gd-DOTA178 . The structure shows coordination of the lanthanide (magenta ball) by carboxylate oxygens and amine nitrogens on the chelator, with a single water molecule (cyan) bound at a Eu-O distance of 2.5 A. (B) The structure of a polypeptide lanthanide binding tag fused to ubiquitin, in complex with gadolinium
100 (magenta), is surprising similar to the Eu-DOTA structure- The lanthanide is again coordinated by a mixture of oxygen and nitrogen ligands and exhibits a q of 1, with a single inner sphere water molecule bound at a Gd-O distance of 2.9 A and a second sphere bound water, which may also contribute to ri, at a distance of 5.4 A from the lanthanide.
101 Figure 1. Mechanisms of MRI contrast enhancement by paramagnetic metalloproteins
AL
~~aAAA*~.
.1.11I I I
102 Figure 2. Engineered metalloprotein-based MRI sensors
A B L 1dparnrnw1.2 OI 57
0anrj 9hM ~ WWI dopainifl
C KID-Ft D 4000 3200 t 2400 z1600 800
KID-Ft KID/ pKID/ PKA ATP -~ KIX KIX
103 Figure 3. Nuclear magnetic relaxation dispersion of BM3h variants
A 10 r Tw 't's B 2G9C6 7-6 0.05 0.71 -2G9 28 0.07 0.72 - OD7 2.7 0.12 0.45 B7 3.0 &1 0.76 3DB10 28 20 6 0.60. B ***00 4 00*
0 0-2 10-1 1O 10 H Larmor Frequency (MHz)
104 Figure 4. Structures of lanthanide binding sites
A
0215.4 A
A
105 Appendix B: Magnetically enhanced mutant ferritins by magnetic cell sorting screen.
Background and motivation
The main goal of my thesis is to develop protein-based MRI contrast agents which will be broadly applicable as genetically-controlled tools for in vivo imaging. As described in
Chapter 1, Ft is a promising platform to engineer such agents; however its low relaxivity limits its application in vivo. The utility of Ft as a reporter gene or as a genetically encodable sensor would be significantly enhanced by improving its ability to induce MRI contrast changes. One way to improve the potency of Ft as a MRI contrast agent is by inducing it to accumulate a ferromagnetic iron oxide core of a highly magnetic nature instead of its standard ferrihydrite core in a physiological environment. Under conditions of elevated pH and temperature, highly magnetic iron oxide (magnetite) can be formed within the cavity of horse spleen ferritins (HSF) that have been previously demineralized 179. Ft containing magnetite is denoted
"magnetoferritin" (MgFt), and was shown to have more than 100 times higher relaxivity than that of natural Ft, allowing it to be detected by MRI at 100 times lower concentration than Ft 156
MgFt is made by chemical synthesis, and is not thought to form spontaneously in nature, at least on a regular basis. Sporadic reports of naturally occurring MgFt may be found in the literature; however, including electromicroscopic evidence that a small proportion of Ft cores naturally consist of magnetite 180. A neurological condition called neuroferritinopathy, associated with C- terminally altered Ft variants 181, 182, has been associated with accumulation of magnetite in the human brain, but it is not clear that the mineral is contained within Ft oligomers. In addition, several instances of magnetite formation in biological contexts independent of Ft have been well established. So-called magnetotactic bacteria form several regularly shaped magnetite crystals of
106 high chemical purity in their cytosols 183 Magnetite is also naturally formed in the tissues of
some vertebrate animals, such as in the upper beak tissue of homing pigeons 184 and in the
ethmoid tissue of the 185', 186. Collectively, these facts support the notion that it might be possible
to create artificial Ft variants with a tendency to nucleate magnetite, rather than ferrihydrite, in
vivo.
Results and discussions
Our approach for obtaining mutant Fts with desired magnetic properties is though
directed evolution (Fig. 1). A large population of yeast cells expressing mutant Fts were created
by PCR-based random mutagenesis followed by homologous recombination into a yeast
expression vector. The library size was about 10 million cells. As a starting point, we used Ft
from Pyrococcusfuriosuswith N-terminal strep-tagIl (SPFt), which was developed and
characterized in detail in chapter 3. Ft expression was induced in an iron-supplemented medium
to facilitate iron loading of Ft in vivo, we were not able to capture cells on the magnetic column.
Major challenge of this project has been the low sensitivity of magnetic cell sorting procedures.
The magnetic force that holds cells on the column is described by the following equation:
F = -VU =(pV )B V1V (B -V)B p0
B : Magnetic induction -flux density
Zvo : Magnetic susceptibility per unit volume of yeast cells
p : Magnetic moment
V: Volume of yeast cells
U: Magnetic potential energy
107 F: Magnetic force
First, we manipulated the magnetic susceptibility of the cells by coexpressing high- affinity iron transporter (FET3/FTR1) to facilitate the iron accumulation in the yeast. Although the total iron content of the cells increased, iron stored in Ft did not, so we abandoned this approach. Second, we manipulated the cell volume. We used a temperature sensitive mutant of yeast, cd28-4 187, which grows to about 10 pim in diameter when grown in 37 'C where normal yeast grows to about 5 pm. Although the oversized yeast cells were captured on the column, there was no difference in retention rates between cells with or without Ft. Furthermore, once the cells were induced to grow to a large enough size to be held on a magnetic column, a large portion of them died.
We then turned our attention to the flow rate, which also affects the sensitivity of the magnetic cell sorting experiment. By using a column with a significantly slower flow rate, we were able to capture regular-sized yeast cells with SPFt, but not the yeast cells without SPFt (Fig.
2A). Furthermore, the retention rate of cells on the column was proportional to the applied magnetic field strength (Table 1), which indicates that the dynamic range of the assay can be adjusted by changing the magnetic field strength. We then conducted a test sorting experiment where 1:1 ratio of yeast cells expressing SPFt and hypermagnetic SPFt mutant, L55P (mentioned in Chapter 2) were sorted together and top 9 %most magnetic cells were collected and subjected to sequencing analysis. Out of 24 samples, 20 showed mutant sequence, which is promising that this assay can distinguish cells with varying magnetic moment due to Ft.
Yeast library of mutagenized Ft was constructed by error-prone PCR with SPFt L55P as a starting template sequence. The yeast library was sorted on the magnetic column and top 5-10 % most magnetic cells were collected and enriched in the growth medium supplemented with iron.
108 Five rounds of screening and enrichment procedures resulted in progressively improved retention, suggesting that the screening process for cells with higher magnetic moment is indeed working. When individual clones were sequenced after five rounds of sorting and enrichment, we obtained two clones that were enriched MI (A21E/F39S/K44E/L55P) and M2
(F3 1L/L55P/K136R). The plasmids containing these mutations were retransformed into the base stain and the clones were individually grown in the medium with iron supplement and subjected to magnetic cell sorting. Unfortunately the retention rates of these mutants were much lower than that of the starting clone, L55P. One possibility is that these cells may have accumulated background mutations in the genomic DNA that made them more magnetic during the five rounds of panning.
Future directions
The growth conditions of the yeast cells may need further optimization. It appears that both the concentrations and the types of iron supplements have a great impact on the amount of iron accumulated in the cell. It is possible that the temperature, aeration, and duration of
incubation would affect the amount and moreover the types of iron mineral formed in Ft inside yeast cells. Finding a condition that allows sufficient iron accumulation without introducing random background mutations in the genomic DNA would minimize false positives. Once yeast
cells with improved magnetic moment are identified, further characterization and iteration of this
screening approach should be performed. It will be essential to quantify the content of iron in Ft
and to analyze the nature of the mineral by magnetometry, MRI and electron microscopy in order
to claim that the mutant Fts truly form hypermagentic iron mineral. After verifying the magnetic
109 property of the iron mineral, these hypermagnetic variants will be studied in mammalian cell culture and virally expressed in rat brains.
Materials and methods
Yeast strain and library construction
Yeast strain and procedures to create Ft library in yeast in this study is described in
Chapter 3.
Magnetic cell sorting
High gradient magnetic separations of yeast cells were performed with Frantz Canister
Separator Model L1-CN (S. G. Frantz Company Inc., Tullytown, PA) and LD columns (Miltenyi
Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. Yeast cells were grown in YPAD medium supplemented with 20 mM ferric citrate from OD 600 of 0.4 for 12 h at 30 C. Cells were harvested and washed twice with PBS with 10mM EDTA. 2 x 108 cells were sorted on the magnetic column at various magnetic fields. The retention rate was calculated by dividing the number of cells eluted off the column (OD600 x elution volume x cell number/OD600 ) by the total number of cells applied on the column. For sorting library yeast, the same procedures were used except that the eluted cells were resuspended in the iron rich medium and incubated for 12 h at 30 C for the subsequent round of screening.
110 Figures and tables
i E: > 1, *
Figure 1. Schematic of directed evolution process, which includes mutant DNA library construction, transformation of yeast with the library by electroporation, and magnetic cell sorting based screen.
111 0.8 ...... B 80
0.6 Vec 60 E - C S0.4 L40 SPFt 00.2 20 10% 13% 13% 5% 7%
0 0AAAAA FT FT W W W EL EL EL EL Lib 51S2 53 S4 55
Figure 2. Magnetic cell sorting experiments with yeast cells. (A) Eluted fractions including flow through (FT), wash (W), and elution (EL) of magnetically sorted yeast cells are monitored by absorbance at 600 nm. Yeast cells transformed with SPFt gene showed some retention on the column whereas cells with empty vector (Vec) came off the column mostly in the FT and W fractions. (B) Yeast cells with a library of mutated L55P DNA were subjected to five rounds of magnetic screen and enrichment. Each population of yeast cells was subjected to the magnetic sorting experiment as in (A) and retained fraction of the cells was calculated. The stringency of the screen of every round is shown (red) as the percentage of cells collected.
112 Appled(A) crren VecSPFt PqPPI ICU LuI. ICL ks~v Vec P
0.6 5.8% 10.0% 1 4.0% 17.8% 1.5 4.2% 31.2%
Table 1. The retention rate of magnetic cell sorting experiments conducted at varying applied magnetic field induced by the applied current. Yeast cells expressing SPFt showed increased retention rate as the magnetic field is increased as predicted. However, the retention rate of yeast cells with empty vector (Vec) did not change when applied magnetic field was increased, indicating that the cells' magnetic moment is too small to be captured.
113 Appendix C: A novel protein-based kinase activity sensor for MRI
Background and motivation
Although a protein-based MRI sensor holds great promise as described in earlier sections, very few natural proteins have been identified and verified with their ability to generate MRI contrast. These include Ft 35'37, transferrin receptor (TfR)36, the mammalian divalent cation transporter DMT 1137, and the magnetotactic bacterial protein MagA . While Ft generates MRI contrast by itself, TfR, DMT 1, and MagA enhance cellular MRI contrast indirectly by upregulating the iron transport into the cytosol. Aside from Ft, very few metalloproteins have been explored to make MRI contrast agents. An example of this is a bacterial heme protein that was evolved to perform as a MRI sensor for an important neurotransmitter dopamine 188 . List of currently available protein-based MRI sensors are very limited, and it would be beneficial to have a variety of contrast agents available for imaging with different purposes. Depending on the molecular target of the interest, it would be convenient to have a contrast agent with a particular affinity, sensitivity, localization and specificity. For example, if the target molecule exists in the extracellular space, the agent needs to be secreted outside of the cell after it has been synthesized by the cells. There is variety of molecular targets one might want to study; thus contrast agents with specificity towards various metabolites and signaling molecules need to be developed.
My approach to find novel protein-based contrast agents involves making a short list of candidate metalloproteins (below).
The list of Mn containing proteins
- Phosphatase
o Inorganic pyrophosphatase
114 o Bacteriophage lambda protein phosphatase
o Sweet potato purple acid phosphatase
- Catalase
o Manganese catalase from Lactobacillus plantarum
- Ribonucleotide reductase
- Arginase
- Phosphotriesterase
- Aminopeptidase
- Exonuclease
- Endonuclease
- Phospholipase D
- Xylose isomerase
- Aminoacyl-tRNA synthetase
o Aspartyl-tRNA synthetase
- Concanavalin A
The list of iron containing protein
- Heme containing proteins
o Cytochrome a, c
o hemoglobin
O hemocyanin
o myoglobin
o neuroglobin
115 o cytoglobin
o leghemoglobin
o peroxidase
o ligninase
- Iron-sulfur protein
o Rubredoxin
o Ferrodoxin
- Cage-like proteins
o Listeria innocua: ferritin-like structure (12subunits)
o Small heat shock protein of M.jannaschii (Hsp20)
o ssDps (Sulfolobus solfataricus Dps)
o PfDps (Pyrococcus furiosus Dps)
- Others
o Lipoxygenases
o Tyrosine 3-monooxygenase
o Purple acid phosphatase
o Uteroferrin
o Catechol 2,3-dioxygenase
o Mandelate 4-monooxygenase
o Methane monooxygenase
o Anthranilate 3-monooxygenase (deaminating); from Aspergillus niger
One of the most promising proteins from this list was bacteriophage lambda protein phosphatase
(kPP) because it contains two manganese ions and binds three water molecules (Fig. lA and B)
116 189 which may contribute to MRI signal. In addition- XPP has a natural affinity to phosphorylated peptides making it an ideal candidate as a MRI kinase activity sensor.
Results and discussions
XPP with C-terminal his-tag (ChiskPP) was cloned and expressed in E.coli at a high level and purified through affinity purification method. There were three issues with the protein preparation; (1) the purified protein showed significant amount of degradation product when analyzed on SDS-PAGE (Fig. 2A), (2) the amount of Mn to protein ratio fluctuated between experiments, and (3) there was high level of iron contamination. In order to mitigate these issues, the expression was conducted in M9 minimum medium instead of LB medium. Cells induced in the minimum medium supplemented with 100pM manganese exhibited much less insoluble fraction upon lysis buffer with manganese, and resulted in a clean product without degradation
(Fig. 2B). The new purification protocol yielded about 100 mg of protein per 1 L of culture and the purity was above 95 %. The purified protein was subjected to inductive coupled plasma (ICP) analysis to measure the manganese content, which usually resulted in 1~1.2 Mn ions per protein.
The purified ChiskPP showed relaxivity of 7.3 ± 0.2 mM' s-1 (n=4) (Fig. 2C). The protein has three water molecules coordinated to the two manganese ions at the active site. Both of these
Mn 2+ ions are reported to be high spin, so its relaxivity could be as high as 30 mM- s1 . The fact that there was only about 1~1.2 Mn ions per protein is consistent with the modest relaxivity.
The next step was to make mutations on ChiskPP such that it still binds to phosphorylated
peptide but does not hydrolyze it. In order eliminate its catalytic activity of XPP, four mutants
(D20N, D52N, R53A, and H76N) which have been documented previously 190 were made. Out
of these mutations, ChiskPP with R53A (abbreviated as R53A for simplicity) appeared most
117 promising with high protein yield as well as high relaxivity (8.6 ± 1.0 mM-1 s-1) and very low catalytic activity (< 0.1% of wild type) (Table 1).
To test see if ChiskPP's relaxivity can be perturbed by binding of a substrate, we incubated 30 pM of ChiskPP and R53A with a competitive inhibitor, sodium orthovanadate
(non-hydrolysable phosphate analog) at various concentrations and measured T1 relaxation rates
(Fig. 3). We observed the decrease in relaxation rate by addition of increasing amounts of orthovanadate. Even though the change in relaxivity was relatively modest (~20%), the effect was reproducible and therefore promising. The signal did not go back to baseline even with 10 fold excess of orthovanadate (300 pM), indicating that water molecules still have access to the active site Mn ions in the presence of orthovanadate. The dissociation constant of orthovanadate is 0.7 ± 0.2 pM 191 is low enough that at the concentration that I added, active site of ChiskPP and R53A must have been fully occupied by the inhibitor. However, it is unclear how many water molecules are displaced by the binding of the orthovanadate and it is likely that larger substrate such as phosphopeptide would have greater effect on the Ti relaxation rate.
We then focused on finding an appropriate phosphopeptide substrate that would perturb
ChiskPP's T1 relaxation rate upon binding. P protein of human repiratory syncytial virus (RSVP) is a natural substrate of LPP 192. We first cloned N-terminally his-tagged RSVP but it was rapidly degraded during purification, so we worked with RSVP which had maltose binding protein fused to its N-terminus (abbreviated as RSVP for simplicity), which is known to be more stable 193
Purified RSVP still showed some degradation but it was stable enough to perform further experiments (Fig. 4A). We first tested to see if we can phosphorylate the protein by visualizing the phosphorylated protein with Pro-Q@ Diamond Phosphoprotein Gel Stain. Incubation with casein kinase II (CKII) phosphorylated RSVP whereas protein kinase A (PKA) phosphorylated
118 RSVP at much lower efficiency (Fig. 4B). Unphosphorylated RSVP showed no signal under phosphorstain as predicted.
To see if the phosphorylated RSVP (RSVP-P) blocks the active site of R53A and alter its relaxation rate, we added 10 fold excess of RSVP or RSVP-P into R53A with and without additional Mn. Unfortunately there were no differences between the relaxation rates of R53A with RSVP and RSVP-P (Fig. 4C). It was not clear whether chimeric RSVP-P with MBP is still a substrate of ChiskPP, so we incubated RSVP-P with ChiskPP and visualized the phosphorylation status with phosphostain. RSVP-P incubated with ChiskPP was hydrolyzed whereas RSVP-P incubated with R53A was not hydrolyzed as expected (Fig. 4D). From this result, we concluded that RSVP-P is still a substrate for ChiskPP but does not occupy the active site of R53A effectively to induce the change in relaxation rate.
Future directions
Ultimate goal of this project is to produce a chimeric protein consisting of metalloprotein and substrate peptide as a simple one component sensor, but it will require some thoughts into the length and the flexibility of the linker between the two domains. By tethering the substrate to
R53A, the local concentration of the substrate would be much higher and may benefit in blocking the active site upon phosphorylation. Initial efforts to fuse RSVP to the C-terminus of
R53A using a 20 amino acid long linker did not yield an expressible construct. The C-terminus of R53A is located on the other side of protein from the active site (Fig. IA), so it is necessary to
introduce a long linker so that RSVP domain could reach the active site when it is fused to the C- terminus of R53A. Obviously one could try fusing a substrate peptide on the N-terminus of
R53A which would require much shorter linker. If none of these strategies succeed, since there
119 is no other obvious candidate for a substrate, one could try high-throughput methods such as yeast display and phage display to obtain such peptide sequence. Such screen has to be carefully planned in order to identify peptide sequences that only when they are phosphorylated show affinity to R53A and not when dephosphorylated. Once the MRI kinase sensor based on a chimeric protein is developed, it will be studied in mammalian cell culture system and virally expressed in rat brains.
Materials and methods
Cloning, expression and purification of Chis kPP
C-terminal histag was introduced to XPP by using polymerase chain reaction (PCR) with
High-Fidelity Phusion master mix (New England Biolabs, Ipswich, MA) and primer set 1 and plasmid, pT7-7-XPP as a template (a kind gift from Dr. Reiter). The PCR product was subcloned into NdeI/EcoRI sites of T7-7 plasmid resulting in pT7-7 Chis XPP. All the mutations on XPP gene were introduced by QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent
Technologies, Santa Clara, CA).
The plasmid was then transformed into BL2 1 DE3 cells for expression of the protein. 2 ml of overnight culture was diluted into 200 ml of M9 minimum medium with 100 ptM MnCl2 and
400 mg/ml ampicillin and grown for about 3 h at 37 'C until OD 60 0 read about between 0.8 and
1. Protein expression was induced by the addition of IPTG at final concentration of 0.4 mM at room temperature for 16 h. Cell were harvested and lysed with 10 ml BugBuster reagent (EMD
Millipore, Billerica, MA) supplemented with protease inhibitor cocktail III and VII, Lysonase
Bioprocessing Reagent (EMD Millipore, Billerica, MA), and MnCl2 at a final concentration of
0.5 mM for 1 h at 4 C. Insoluble fractions were removed by centrifugation at 16,000 g for 40
120 min. at which point there was almost not visible pellet. Incubate the soluble fraction of the lysate and 4 ml of Ni-NTA resin (Qiagen, Valencia, CA) supplemented with imidazol at final concentration of 15 mM for 1 h at 4 C. Affinity purification of the protein was carried out as described by manufacturer's instructions, and the eluted fraction was buffer exchanged using
Nap- 10 column (GE healthcare, Buckinghamshire, United Kingdom) into the working buffer (50 mM HEPES and 100 mM NaCl).
Expression and purification of RSVP
A plasmid containing a fusion gene MBP-RSVP, pMal-P was kindly provided by Dr
Ventura and freshly transformed into BL2 1DE3 plysS strain of E.coli for expression study. Since
RSVP is toxic to bacteria, the frozen stock of such strain cannot be maintained at -80 'C. 1 ml of overnight culture of the cells is diluted into 200 ml of 2xTY with 100 ptg/ml ampicillin and 34
pg/ml chloramphenicol, and incubated at 37 C. Once OD 600 reached 0.5, the culture was then
induced with IPTG at a final concentration 0.3 mM at room temperature for 3 h, and then harvested. The cells were lysed with 12.5 ml BugBuster reagent (EMD Millipore, Billerica, MA)
with protease inhibitor cocktail set III (EMD Millipore, Billerica, MA) and Lysonase
Bioprocessing Reagent (EMD Millipore, Billerica, MA) at 4 C for 1 h. After removing the
insoluble fraction by centrifugation at 16000 g for 40 min, the cleared lysate was then mixed
with 6 ml of amylose resin (New England Biolabs, Ipswich, MA) and incubated for 1.5 h in 4 'C.
The affinity purification was carried out according to manufacturer's instructions. The eluted
protein fractions were pooled together and buffer exchanged into Tris buffer (200 mM Tris/HCl
and 200 mM NaCl).
121 Relaxivity measurements of ChisXPP and the mutants
To test magnetic relaxation behavior of the proteins, we prepared XPP samples (60 pl) in the wells of microtiter plates and placed them in a 40-cm-bore Bruker Avance 4.7 T MRI scanner. Unused wells in the plates were filled with PBS. We used a Ti-weighted spin echo pulse sequence; echo time (TE) was 10 ms, and repetition times (TR) were 150, 200, 400, 800 ms, 1.5,
3, and 5 s. We calculated relaxation rates by fitting the image intensity with the following exponential function:
I = k 1 - exp (-)](T1
I: observed MRI signal intensity k: constant
We computed the relaxivity (rj) of proteins by linear fitting to a plot of I/T1 against protein concentration, typically with 5 data points ranging from 0 to 120 pM.
Phosphatase activity measurement
The reaction mixture for phosphatase activity assay contained 1 mM MnCl 2, 1xPMP buffer (New England Biolabs, Ipswich, MA), Ix pNPP (New England Biolabs, Ipswich, MA) and the protein at appropriate concentrations. The reaction is mixed and incubated for 5 min in
30 'C and stopped with addition of 1 ml of 0.5 M EDTA solution. The specific activity of a protein (U) was calculated by the following equation:
3 U _ A x 1.05 X 10- L x 1 x 30min U 1 1 1 9 mg 16000 M- cm- t min 1x10- moles CMX 1 0-3M
A: Absorbance at 405nm
T: Reaction time
C: Concentration of the protein 122 Phosphorylation assays and detection of phosphopeptide
RSVP was phosphorylated by 5000U of CKII (New England Biolabs, Ipswich, MA) or
PKA (New England Biolabs, Ipswich, MA) in the presence of 1 mM ATP and 1x PKA buffer
(New England Biolabs, Ipswich, MA) for 1.5 h at 30 C. The phosphorylated protein samples were then desalted on Zeba Spin Desalting Columns (Thermo Scientific, Waltham, MA) according to the manufacturer's instructions. Presence of phosphorylated peptide was detected by running the samples on two separate SDS-PAGE gels and one was stained with Coomassie blue and the other with ProQ Diamond Phosphoprotein Gel Stain (Life Technologies, Carlsbad,
CA) according to the manufacturer's instructions.
123 Figures and tables
A B
Figure 1. X-ray crystal structure of bacteriophage XPP 189 (A) Stereo ribbon diagram of XPP and the two Mn2+ ions (purple) (B) Active site of XPP interacting with terminally bound sulfate ion (green). Important active site residues (ball-and-stick representation), water molecules (red balls), Mn2+ ions, sulfate ion, and side chains of the proteins are shown. The hydrogen bonding interactions are shown as dotted black lines and coordination bond to Mn2+ ions are shown in solid black lines.
124 A B ka kDa kDa 1 < O\\C O -
40 40 30
20 20 9<
r1=7.3±0.2 D 8r,=.6±1.0 1.5 1.5
0.5 W I 0. I 0 Li 0 0.05 0.1 0.15 0 0.05 0.1 0.15 LPP concentration [mM] LPP concentration [mM]
Figure 2. Affinity purified ChisXPP and R53A and their relaxivity measurements. (A) SDS-
PAGE analysis of Ni-NTA purification procedure of ChiskPP showing degradation product (<).
(B) SDS-PAGE analysis of ChisXPP and R53A purified using the improved protocol confirming no degradation. (C), (D) Representative plots of linear fit for relaxity measurements with
ChiskPP and R53A. The error of the relaxavity value is s.e.m. of 4 independent experiments.
125 1.0 I
0.8 70.6 ~0.4 0.2 0
Figure 3. Competitive inhibitor reduces T, relaxation rates of phosphatase sensors. T, relaxation rate modulation upon addition of a competitive inhibitor sodium orthovanadate (1) into
ChisXPP protein samples. Error bars show s.e.m. of two independent experiments.
126 A B C kDa Std RSVP
116-1 2 3 kDa Std 1 2 3 0.6
66 0.4 60 66 66 50 0.2 40 45 ,- 45 Coomassie blue Phosphostain 0 4Q 1. RSVP 2. RSVP+ CKI X 20 3. RSVP+ PKA D kDa Std 1 2 3 4 5 6 7 Std 1 2 3 4 5 6 7 116 1. RSVP-P 2. RSVP-P + WT 66 3. RSVP-P + WT + Mn(ll) 45 4. RSVP-P + R63A 5. RSVP-P + R63A + Mn( 1I) 24 6. WI 7. R63A %.#%Ri'UN I MOOM IJ Ufa ~ F- I ic" j Rjeial I 1
Figure 4. Experiments with a potential phosphopeptide substrate of XPP, RSVP. (A) SDS-
PAGE analysis of purified MBP fused RSVP showing modest amount of degradation products.
(B) SDS-PAGE of RSVP and RSVP-P were stained with Coomassie blue and ProQ phosphostaining showing RSVP was successfully phosphorylated by Casein Kinase I (CKII) but not by Protein Kinase A (PKA). (C) T, relaxation rates of the mixtures of R53A and RSVP or
RSVP-P showing no effect was observed upon addition of RSVP-P into R53A. (D)
Dephosphorylation assay was carried out and visualized on SDS-PAGE gels stained as in (B).
RSVP-P was rapidly dephosphorylated by ChisXPP (WT) but not by R53A with or without additional Mn2+ ions.
127 Enzyme activity T, relaxivity (%/WT) [mM-1 s-1] D20N <0.001 1.5± 0.5 D52N 0.7 ± 0.2 3.5 ± 0.2 R53A 0.001± 0.0002 8.6 ± 1.0 H76N 0.4± 0.07 2.8 * 0.2
Table 1. Enzymatic activity and T1 relaxivity of ChisXPP mutants
128 References
1. Massoud, T.F. & Gambhir, S.S. Molecular imaging in living subjects: seeing
fundamental biological processes in a new light. Genes & development 17, 545-580
(2003).
2. Lauterbur, P.C. Image Formation by Induced Local Interactions - Examples Employing
Nuclear Magnetic-Resonance. Nature 242, 190-191 (1973).
3. Grewe, B.F. & Helmchen, F. Optical probing of neuronal ensemble activity. Curr Opin
Neurobiol 19, 520-529 (2009).
4. Bloembergen, N., Purcell, E.M. & Pound, R.V. Relaxation Effects in Nuclear Magnetic
Resonance Absorption. PhysicalReview 73, 679-712 (1948).
5. Bloch, F., Hansen, W.W. & Packard, M. The Nuclear Induction Experiment. Physical
Review 70, 474-485 (1946).
6. Bloembergen, N. & Morgan, L.O. Proton Relaxation Times in Paramagnetic Solutions
Effects of Electron Spin Relaxation. Journal of Chemical Physics 34, 842-& (1961).
7. Bloembergen, I. Proton relaxation times in paramagnetic solutions. J. Chem. Phys. 27,
572-573 (1957).
8. Solomon, I. & Bloembergen, N. Nuclear Magnetic Interactions in the Hf Molecule.
Journalof Chemical Physics 25, 261-266 (1956).
9. Kowalewski, J., Kruk, D. & Parigi, G. in Advances in inorganic chemistry, Vol. 57 41-
104 (Elsevier, 2005).
10. Freed, J.H. Dynamic effects of pair correlation functions on spin relaxation by
translational diffusion in liquids. II. Finite jumps and independent TI processes. J. Chem.
Phys. 68, 4034-4037 (1978).
129 11. Roch, A., Muller, R.N. & Gillis, P. Theory of proton relaxation induced by
superparamagnetic particles. Journalof Chemical Physics 110, 5403-5411 (1999).
12. Koenig, S.H. & Kellar, K.E. Theory of 1/T1 and 1/T2 NMRD profiles of solutions of
magnetic nanoparticles. Magn Reson Med 34, 227-233 (1995).
13. Gillis, P. & Koenig, S.H. Transverse relaxation of solvent protons induced by magnetized
spheres: application to ferritin, erythrocytes, and magnetite. Magn Reson Med 5, 323-345
(1987).
14. Caravan, P., Farrar, C.T., Frullano, L. & Uppal, R. Influence of molecular parameters and
increasing magnetic field strength on relaxivity of gadolinium- and manganese-based TI
contrast agents. Contrast media & molecular imaging 4, 89-100 (2009).
15. Lauffer, R.B. Paramagnetic Metal-Complexes as Water Proton Relaxation Agents for
Nmr Imaging - Theory and Design. Chemical Reviews 87, 901-927 (1987).
16. Caravan, P., Ellison, J.J., McMurry, T.J. & Lauffer, R.B. Gadolinium(III) Chelates as
MRI Contrast Agents: Structure, Dynamics, and Applications. Chemical Reviews 99,
2293-2352 (1999).
17. Carr, D.H. et al. Intravenous chelated gadolinium as a contrast agent in NMR imaging of
cerebral tumours. Lancet 1, 484-486 (1984).
18. Renshaw, P.F., Owen, C.S., McLaughlin, A.C., Frey, T.G. & Leigh, J.S., Jr.
Ferromagnetic contrast agents: a new approach. Magn Reson Med 3, 217-225 (1986).
19. Olsson., M.B.E., Persson., B.R.B., Salford., L.G. & Schr6der, U. Ferromagnetic particles
as contrast agent in T2 NMR imaging. MRI 4, 437-440 (1986).
20. Mendonca Dias, M.H. & Lauterbur, P.C. Ferromagnetic particles as contrast agents for
magnetic resonance imaging of liver and spleen. Magn Reson Med 3, 328-330 (1986).
130 21. Jun, Y.W., Seo, J.W. & Cheon, J. Nanoscaling laws of magnetic nanoparticles and their
applicabilities in biomedical sciences. Acc Chem Res 41, 179-189 (2008).
22. Gillis, P., Moiny, F. & Brooks, R.A. On T(2)-shortening by strongly magnetized spheres:
a partial refocusing model. Magn Reson Med 47, 257-263 (2002).
23. Weiner, E.J., Datta, A., Jocher, C.J. & Raymond, K.N. High-relaxivity MRI contrast
agents: where coordination chemistry meets medical imaging. Angew Chem Int Ed Engl
47, 8568-8580 (2008).
24. Toth., E., Helm., L. & Merbach., A.E. in The chemistry of contrast agents in medical
magnetic resonance imaging., Edn. 1 45-1202001).
25. Hsieh, V. & Jasanoff, A. Bioengineered probes for molecular magnetic resonance
imaging in the nervous system. ACS chemical neuroscience 3, 593-602 (2012).
26. Caravan, P. et al. The interaction of MS-325 with human serum albumin and its effect on
proton relaxation rates. Journalof the American Chemical Society 124, 3152-3162
(2002).
27. Schmiedl, U., Ogan, M.D., Moseley, M.E. & Brasch, R.C. Comparison of the Contrast-
Enhancing Properties of Albumin-(Gd-Dtpa) and Gd-Dtpa at 2.0-T - an Experimental-
Study in Rats. Am JRoentgenol 147, 1263-1270 (1986).
28. Shi, H.F., Bencze, K.Z., Stemmler, T.L. & Philpott, C.C. A cytosolic iron chaperone that
delivers iron to ferritin. Science 320, 1207-1210 (2008).
29. Fischbach, F.A., Harrison, P.M. & Hoy, T.G. The structural relationship between ferritin
protein and its mineral core. J Mol Biol 39, 235-238 (1969).
131 30. Gossuin, Y., Roch, A., Muller, R.N., Gillis, P. & Lo Bue, F. Anomalous nuclear magnetic
relaxation of aqueous solutions of ferritin: an unprecedented first-order mechanism.
Magn Reson Med 48, 959-964 (2002).
31. Gossuin, Y., Roch, A., Muller, R.N. & Gillis, P. Relaxation induced by ferritin and
ferritin-like magnetic particles: the role of proton exchange. Magn Reson Med 43, 237-
243 (2000).
32. Gossuin, Y., Gillis, P., Hocq, A., Vuong, Q.L. & Roch, A. Magnetic resonance relaxation
properties of superparamagnetic particles. Wiley InterdiscipRev Nanomed
Nanobiotechnol 1, 299-310 (2009).
33. Shonat, R.D. & Koretsky, A.P. Expression of myoglobin in the transgenic mouse brain.
Advances in experimental medicine and biology 530, 331-345 (2003).
34. Cohen, B., Dafni, H., Meir, G., Harmelin, A. & Neeman, M. Ferritin as an endogenous
MRI reporter for noninvasive imaging of gene expression in C6 glioma tumors.
Neoplasia 7, 109-117 (2005).
35. Genove, G., DeMarco, U., Xu, H., Goins, W.F. & Ahrens, E.T. A new transgene reporter
for in vivo magnetic resonance imaging. Nat Med 11, 450-454 (2005).
36. Deans, A.E. et al. Cellular MRI contrast via coexpression of transferrin receptor and
ferritin. Magn Reson Med 56, 51-59 (2006).
37. Cohen, B. et al. MRI detection of transcriptional regulation of gene expression in
transgenic mice. Nat Med 13, 498-503 (2007).
38. Vande Velde, G. et al. Evaluation of the specificity and sensitivity of ferritin as an MRI
reporter gene in the mouse brain using lentiviral and adeno-associated viral vectors. Gene
Ther 18, 594-605 (2011).
132 39. Naumova. A.V. et al. Ferritin overexpression for noninvasive magnetic resonance
imaging-based tracking of stem cells transplanted into the heart. Mol Imaging 9, 201-210
(2010).
40. lordanova, B. & Ahrens, E.T. In vivo magnetic resonance imaging of ferritin-based
reporter visualizes native neuroblast migration. Neuroimage 59, 1004-1012 (2012).
41. Mohie-Eldin, M.E.Y., Frankel, R.B. & Gunther, L. A comparison of the magnetic
properties of polysaccharide iron complex (PIC) and ferritin. Journal of Magnetism and
Magnetic Materials 135, 65-81 (1994).
42. Rebodos, R.L. & Vikesland, P.J. Effects of oxidation on the magnetization of
nanoparticulate magnetite. Langmuir 26, 16745-16753 (2010).
43. Atanasijevic, T., Shusteff, M., Fam, P. & Jasanoff, A. Calcium-sensitive MRI contrast
agents based on superparamagnetic iron oxide nanoparticles and calmodulin. Proc Natl
Acad Sci USA 103, 14707-14712 (2006).
44. Weissleder, R. et al. In vivo magnetic resonance imaging of transgene expression. Nat
Med 6, 351-355 (2000).
45. Louie, A.Y. et al. In vivo visualization of gene expression using magnetic resonance
imaging. Nat. Biotechnol. 18, 321-325 (2000).
46. Li, W.H., Fraser, S.E. & Meade, T.J. A calcium-sensitive magnetic resonance imaging
contrast agent. Journalof the American Chemical Society 121, 1413-1414 (1999).
47. Etoc, F. et al. Subcellular control Rac-GTPase signalling by magnetogenetic
manipulation inside living cells. Nat Nanotechnol 8, 193-198 (2013).
48. Cho, M.H. et al. A magnetic switch for the control of cell death signalling in in vitro and
in vivo systems. Nature materials 11, 1038-1043 (2012).
133 49. Huang, H., Delikanli, S., Zeng, H., Ferkey, D.M. & Pralle, A. Remote control of ion
channels and neurons through magnetic-field heating of nanoparticles. Nat Nanotechnol
5, 602-606 (2010).
50. Bartelle, B.B., Szulc, K.U., Suero-Abreu, G.A., Rodriguez, J.J. & Turnbull, D.H.
Divalent metal transporter, DMT1: A novel MRI reporter protein. Magn Reson Med 70,
842-850 (2012).
51. Nishida, K. & Silver, P.A. Induction of biogenic magnetization and redox control by a
component of the target of rapamycin complex 1 signaling pathway. Plos Biol 10,
e1001269 (2012).
52. Kim, T., Moore, D. & Fussenegger, M. Genetically programmed superparamagnetic
behavior of mammalian cells. Journal ofBiotechnology 162, 237-245 (2012).
53. Zurkiya, 0., Chan, A.W. & Hu, X. MagA is sufficient for producing magnetic
nanoparticles in mammalian cells, making it an MRI reporter. Magn Reson Med 59,
1225-1231 (2008).
54. Theil, E.C. Ferritin - Structure, Gene-Regulation, and Cellular Function in Animals,
Plants, and Microorganisms. Annu Rev Biochem 56, 289-315 (1987).
55. Stanley, S.A. et al. Radio-Wave Heating of Iron Oxide Nanoparticles Can Regulate
Plasma Glucose in Mice. Science 336, 604-608 (2012).
56. Meldrum, F.C., Heywood, B.R. & Mann, S. Magnetoferritin: in vitro synthesis of a novel
magnetic protein. Science 257, 522-523 (1992).
57. Vymazal, J., Zak, 0., Bulte, J.W., Aisen, P. & Brooks, R.A. TI and T2 of ferritin
solutions: effect of loading factor. Magn Reson Med 36, 61-65 (1996).
134 58. Yamaguchi-Iwai, Y., Ueta, R., Fukunaka, A. & Sasaki, R. Subcellular localization of
Afti transcription factor responds to iron status in Saccharomyces cerevisiae. J. Biol.
Chem. 277, 18914-18918 (2002).
59. Yamaguchi-Iwai, Y., Stearman, R., Dancis, A. & Klausner, R.D. Iron-regulated DNA
binding by the AFTI protein controls the iron regulon in yeast. The EMBOjournal15,
3377-3384 (1996).
60. Strochlic, T.I., Setty, T.G., Sitaram, A. & Burd, C.G. Grdl9/Snx3p functions as a cargo-
specific adapter for retromer-dependent endocytic recycling. The Journalof cell biology
177, 115-125 (2007).
61. Tatur, J., Hagedoorn, P.L., Overeijnder, M.L. & Hagen, W.R. A highly thermostable
ferritin from the hyperthermophilic archaeal anaerobe Pyrococcus furiosus.
Extremophiles 10, 139-148 (2006).
62. Stefanini, S. et al. Thermal stability of horse spleen apoferritin and human recombinant H
apoferritin. Archives of biochemistry and biophysics 325, 5 8-64 (1996).
63. Tatur, J., Hagen, W.R. & Matias, P.M. Crystal structure of the ferritin from the
hyperthermophilic archaeal anaerobe Pyrococcus furiosus. J Biol Inorg Chem 12, 615-
630 (2007).
64. lordanova, B., Robison, C.S. & Ahrens, E.T. Design and characterization of a chimeric
ferritin with enhanced iron loading and transverse NMR relaxation rate. J Biol Inorg
Chem 15, 957-965 (2010).
65. Shapiro, M.G., Szablowski, J.O., Langer, R. & Jasanoff, A. Protein nanoparticles
engineered to sense kinase activity in MRI. JAm Chem Soc 131, 2484-2486 (2009).
135 66. Wade, V.J. et al. Influence of Site-Directed Modifications on the Formation of Iron Cores
in Ferritin. J.Mol. Biol. 221, 1443-1452 (1991).
67. Levi, S. et al. Mutational analysis of the channel and loop sequences of human ferritin H-
chain. The Biochemicaljournal264, 381-388 (1989).
68. Lawson, D.M. et al. Identification of the Ferroxidase Center in Ferritin. Febs Lett 254,
207-210 (1989).
69. Uchida, M., Kang, S., Reichhardt, C., Harlen, K. & Douglas, T. The ferritin superfamily:
Supramolecular templates for materials synthesis. Biochimica et biophysica acta 1800,
834-845 (2010).
70. Wach, A., Brachat, A., Pohlmann, R. & Philippsen, P. New heterologous modules for
classical or PCR-based gene disruptions in Saccharomyces cerevisiae. Yeast 10, 1793-
1808 (1994).
71. Volschenk, H. et al. Malolactic fermentation in grape musts by a genetically engineered
strain of Saccharomyces cerevisiae. Am JEnol Viticult 48, 193-197 (1997).
72. Chao, G. et al. Isolating and engineering human antibodies using yeast surface display.
Natureprotocols 1, 755-768 (2006).
73. Benatuil, L., Perez, J.M., Belk, J. & Hsieh, C.M. An improved yeast transformation
method for the generation of very large human antibody libraries. Protein engineering,
design & selection: PEDS 23, 155-159 (2010).
74. Tamarit, J., Irazusta, V., Moreno-Cermeno, A. & Ros, J. Colorimetric assay for the
quantitation of iron in yeast. Analytical biochemistry 351, 149-151 (2006).
75. Ebrahimi, K.H., Hagedoorn, P.L., Jongejan, J.A. & Hagen, W.R. Catalysis of iron core
formation in Pyrococcus furiosus ferritin. JBiol Inorg Chem 14, 1265-1274 (2009).
136 76. Gilad, A A_ et al MRI renorter genes .Journal nfnur1ear medicine: official publication,
Society ofNuclear Medicine 49, 1905-1908 (2008).
77. Bennett, K.M., Shapiro, E.M., Sotak, C.H. & Koretsky, A.P. Controlled aggregation of
ferritin to modulate MRI relaxivity. Biophysicaljournal95, 342-351 (2008).
78. Gallizia, A. et al. Production of a soluble and functional recombinant streptavidin in
Escherichia coli. Protein expression andpurification 14, 192-196 (1998).
79. Voss, S. & Skerra, A. Mutagenesis of a flexible loop in streptavidin leads to higher
affinity for the Strep-tag II peptide and improved performance in recombinant protein
purification. Protein Eng 10, 975-982 (1997).
80. Hussey, S.L., Muddana, S.S. & Peterson, B.R. Synthesis of a beta-estradiol-biotin
chimera that potently heterodimerizes estrogen receptor and streptavidin proteins in a
yeast three-hybrid system. Journalof the American Chemical Society 125, 3692-3693
(2003).
81. Athavankar, S. & Peterson, B.R. Control of gene expression with small molecules:
biotin-mediated acylation of targeted lysine residues in recombinant yeast. Chem Biol 10,
1245-1253 (2003).
82. von der Haar, T. Optimized protein extraction for quantitative proteomics of yeasts. PLoS
ONE 2, e1078 (2007).
83. Tamarit, J., Irazusta, V., Moreno-Cermeno, A. & Ros, J. Colorimetric assay for the
quantitation of iron in yeast. Anal. Biochem. 351, 149-151 (2006).
84. Sosnovik, D.E. & Weissleder, R. Emerging concepts in molecular MRI. Curr Opin
Biotechnol 18, 4-10 (2007).
137 85. Perez, J.M., Josephson, L. & Weissleder, R. Use of magnetic nanoparticles as
nanosensors to probe for molecular interactions. Chembiochem 5, 261-264 (2004).
86. Brooks, R.A. T(2)-shortening by strongly magnetized spheres: a chemical exchange
model. Magn Reson Med 47, 388-391 (2002).
87. Brooks, R.A., Moiny, F. & Gillis, P. On T2-shortening by weakly magnetized particles:
the chemical exchange model. Magn Reson Med 45, 1014-1020 (2001).
88. Muller, R.N., Gillis, P., Moiny, F. & Roch, A. Transverse relaxivity of particulate MRI
contrast media: from theories to experiments. Magn Reson Med 22, 178-182; discussion
195-176 (1991).
89. Jensen, J.H. & Chandra, R. NMR relaxation in tissues with weak magnetic
inhomogeneities. Magn Reson Med 44, 144-156 (2000).
90. Jensen, J.H. & Chandra, R. Strong field behavior of the NMR signal from magnetically
heterogeneous tissues. Magn Reson Med 43, 226-236 (2000).
91. Kiselev, V.G. & Posse, S. Analytical model of susceptibility-induced MR signal
dephasing: effect of diffusion in a microvascular network. Magn Reson Med 41, 499-509
(1999).
92. Yablonskiy, D.A. & Haacke, E.M. Theory of NMR signal behavior in magnetically
inhomogeneous tissues: the static dephasing regime. Magn Reson Med 32, 749-763
(1994).
93. Roch, A., Gossuin, Y., Muller, R.N. & Gillis, P. Superparamagnetic colloid suspensions:
Water magnetic relaxation and clustering. Journalof Magnetism and Magnetic Materials
293, 532-539 (2005).
138 94. Boxerman, J.L. et al. The intravascular contribution to fMRI signal change: Monte Carlo
modeling and diffusion-weighted studies in vivo. Magn Reson Med 34, 4-10 (1995).
95. Kennan, R.P., Zhong, J. & Gore, J.C. Intravascular susceptibility contrast mechanisms in
tissues. Magn Reson Med 31, 9-21 (1994).
96. Hardy, P.A. & Henkelman, R.M. Transverse relaxation rate enhancement caused by
magnetic particulates. Magn Reson Imaging 7, 265-275 (1989).
97. Shapiro, M.G., Atanasijevic, T., Faas, H., Westmeyer, G.G. & Jasanoff, A. Dynamic
imaging with MRI contrast agents: quantitative considerations. Magn Reson Imaging 24,
449-462 (2006).
98. Josephson, L., Perez, J.M. & Weissleder, R. Magnetic nanosensors for the detection of
oligonucleotide sequences. Angew Chem Int Edit 40, 3204-+ (2001).
99. Yigit, M.V. et al. Smart "turn-on" magnetic resonance contrast agents based on aptamer-
functionalized superparamagnetic iron oxide nanoparticles. Chembiochem 8, 1675-1678
(2007).
100. Waldron, K.J., Rutherford, J.C., Ford, D. & Robinson, N.J. Metalloproteins and metal
sensing. Nature 460, 823-830 (2009).
101. Lu, Y., Yeung, N., Sieracki, N. & Marshall, N.M. Design of functional metalloproteins.
Nature 460, 855-862 (2009).
102. Voet, D. & Voet, J.G. Biochemistry. (J. Wiley & Sons, New York; 2004).
103. Hochuli, E., Bannwarth, W., Dobeli, H., Gentz, R. & Stuber, D. Genetic Approach to
Facilitate Purification of Recombinant Proteins with a Novel Metal Chelate Adsorbent.
Bio-Technol 6, 1321-1325 (1988).
139 104. Hazen, T.C. in Handbook of Hydrocarbon and Lipid Microbiology. (ed. K.N. Timmis)
2505-2514 (Springer Berlin Heidelberg, Berlin, Heidelberg; 2010).
105. Fox, B.G., Borneman, J.G., Wackett, L.P. & Lipscomb, J.D. Haloalkene oxidation by the
soluble methane monooxygenase from Methylosinus trichosporium OB3b: mechanistic
and environmental implications. Biochemistry 29, 6419-6427 (1990).
106. Alessandrini, A., Salerno, M., Frabboni, S. & Facci, P. Single-metalloprotein wet
biotransistor. Applied Physics Letters 86, 133902-133904 (2005).
107. Yagati, A.K., Lee, T., Min, J. & Choi, J.W. A robust nanoscale biomemory device
composed of recombinant azurin on hexagonally packed Au-nano array. Biosens
Bioelectron 40, 283-290 (2013).
108. Kram, H.B. Noninvasive tissue oxygen monitoring in surgical and critical care medicine.
The Surgical clinics of North America 65, 1005-1024 (1985).
109. Boyden, E.S., Zhang, F., Bamberg, E., Nagel, G. & Deisseroth, K. Millisecond-timescale,
genetically targeted optical control of neural activity. Nature neuroscience 8, 1263-1268
(2005).
110. Nishida, K. & Silver, P.A. Induction of Biogenic Magnetization and Redox Control by a
Component of the Target of Rapamycin Complex 1 Signaling Pathway. Plos Biol 10
(2012).
111. Kim, T., Moore, D. & Fussenegger, M. Genetically programmed superparamagnetic
behavior of mammalian cells. J.Biotechnol. 162, 237-245 (2012).
112. Westmeyer, G.G. & Jasanoff, A. Genetically controlled MRI contrast mechanisms and
their prospects in systems neuroscience research. Magn Reson Imaging 25, 1004-1010
(2007).
140 113 Vande Velde. G. et a. Quantitative evaluation of MRI-based tracking of ferritin-labeled
endogenous neural stem cell progeny in rodent brain. Neuroimage 62, 367-380 (2012).
114. Brustad, E.M. et al. Structure-guided directed evolution of highly selective p450-based
magnetic resonance imaging sensors for dopamine and serotonin. J Mol Biol 422, 245-
262 (2012).
115. Shapiro, M.G. et al. Directed evolution of a magnetic resonance imaging contrast agent
for noninvasive imaging of dopamine. Nat Biotechnol 28, 264-270 (2010).
116. Robert N. Muller, A.R., Jean-Marie Colet, Assia Ouakssim, Pierre Gillis in The
chemistry of contrast agents in medical magnetic resonance imaging2001).
117. Toth, E., Helm, L. & Merbach, A.E. in Contrast agents I, Vol. 221 61-101 (Springer
Berlin Heidelberg, Berlin, Heidelberg; 2002).
118. Perez, J.M., Josephson, L., O'Loughlin, T., Hogemann, D. & Weissleder, R. Magnetic
relaxation switches capable of sensing molecular interactions. Nat Biotechnol 20, 816-
820 (2002).
119. Matsumoto, Y. & Jasanoff, A. T2 relaxation induced by clusters of superparamagnetic
nanoparticles: Monte Carlo simulations. Magn Reson Imaging 26, 994-998 (2008).
120. Schwert, D.D., Davies, J.A. & Richardson, N. in Contrast agents I, Vol. 221 165-199
(Springer Berlin Heidelberg, Berlin, Heidelberg; 2002).
121. Bonnet, C.S., Fries, P.H., Crouzy, S. & Delangle, P. Outer-sphere investigation of MRI
relaxation contrast agents. Example of a cyclodecapeptide gadolinium complex with
second-sphere water. The journal ofphysical chemistry. B 114, 8770-8781 (2010).
141 122. Einstein, A. Iber die von der molekularkinetischen theorie der warme geforderte
bewegung von in ruhenden flussigkeiten suspendierten teilchen. Ann. Phys. 17, 549-560
(1905).
123. Diakova, G., Goddard, Y., Korb, J.P. & Bryant, R.G. Water-proton-spin-lattice-relaxation
dispersion of paramagnetic protein solutions. JMagn Reson 208, 195-203 (2011).
124. Pauling, L. & Coryell, C.D. The Magnetic Properties and Structure of Hemoglobin,
Oxyhemoglobin and Carbonmonoxyhemoglobin. Proc Natl Acad Sci U S A 22, 210-216
(1936).
125. Ogawa, S., Lee, T.M., Kay, A.R. & Tank, D.W. Brain magnetic resonance imaging with
contrast dependent on blood oxygenation. Proc Natl Acad Sci U S A 87, 9868-9872
(1990).
126. Kapff, C.T. & Jandl, J.H. Blood, atlas and sourcebook of hematology. (Little, Brown,
Boston; 1991).
127. Bandettini, P.A., Wong, E.C., Hinks, R.S., Tikofsky, R.S. & Hyde, J.S. Time course EPI
of human brain function during task activation. Magn Reson Med 25, 390-397 (1992).
128. Ogawa, S. et al. Intrinsic signal changes accompanying sensory stimulation: functional
brain mapping with magnetic resonance imaging. Proc Natl Acad Sci U S A 89, 5951-
5955 (1992).
129. Kwong, K.K. et al. Dynamic magnetic resonance imaging of human brain activity during
primary sensory stimulation. Proc Natl Acad Sci U S A 89, 5675-5679 (1992).
130. Prasad, P.V. & Priatna, A. Functional imaging of the kidneys with fast MRI techniques.
Europeanjournal of radiology 29, 133-148 (1999).
142 131. Carlier, P.G. Skeletal muscle perfusion and oxygenation assessed by dynamic NMR
imaging and spectroscopy. Advances in experimental medicine and biology 701, 341-346
(2011).
132. Tran, T.K., Sailasuta, N., Hurd, R. & Jue, T. Spatial distribution of deoxymyoglobin in
human muscle: an index of local tissue oxygenation. NMR in biomedicine 12, 26-30
(1999).
133. Harrison, P.M., Fischbach, F.A., Hoy, T.G. & Haggis, G.H. Ferric oxyhydroxide core of
ferritin. Nature 216, 1188-1190 (1967).
134. Kaplan, J., Jordan, 1. & Sturrock, A. Regulation of the transferrin-independent iron
transport system in cultured cells. The Journalof biological chemistry 266, 2997-3004
(1991).
135. Karin, M. & Mintz, B. Receptor-mediated endocytosis of transferrin in developmentally
totipotent mouse teratocarcinoma stem cells. The Journal of biological chemistry 256,
3245-3252 (1981).
136. Rothman, R.J., Serroni, A. & Farber, J.L. Cellular Pool of Transient Ferric Iron,
Chelatable by Deferoxamine and Distinct from Ferritin, That Is Involved in Oxidative
Cell Injury. Mol. Pharmacol.42, 703-710 (1992).
137. Bartelle, B.B., Szulc, K.U., Suero-Abreu, G.A., Rodriguez, J.J. & Turnbull, D.H.
Divalent metal transporter, DMT1: A novel MRI reporter protein. Magn Reson Med
(2012).
138. Lauffer, R.B. & Brady, T.J. Preparation and water relaxation properties of proteins
labeled with paramagnetic metal chelates. Magn Reson Imaging 3, 11-16 (1985).
143 139. Unger, E.C. et al. Magnetic resonance imaging using gadolinium labeled monoclonal
antibody. Invest Radiol 20, 693-700 (1985).
140. Shreve, P. & Aisen, A.M. Monoclonal antibodies labeled with polymeric paramagnetic
ion chelates. Magn Reson Med 3, 336-340 (1986).
141. Wu, C.W. et al. Development of a MR-visible compound for tracing neuroanatomical
connections in vivo. Neuron 70, 229-243 (2011).
142. Zhao, M., Beauregard, D.A., Loizou, L., Davletov, B. & Brindle, K.M. Non-invasive
detection of apoptosis using magnetic resonance imaging and a targeted contrast agent.
Nat Med 7, 1241-1244 (2001).
143. Ruehm, S.G., Corot, C., Vogt, P., Kolb, S. & Debatin, J.F. Magnetic resonance imaging
of atherosclerotic plaque with ultrasmall superparamagnetic particles of iron oxide in
hyperlipidemic rabbits. Circulation 103, 415-422 (2001).
144. Cerdan, S., Lotscher, H.R., Kunnecke, B. & Seelig, J. Monoclonal antibody-coated
magnetite particles as contrast agents in magnetic resonance imaging of tumors. Magn
Reson Med 12, 151-163 (1989).
145. Tiefenauer, L.X., Kuhne, G. & Andres, R.Y. Antibody-magnetite nanoparticles: in vitro
characterization of a potential tumor-specific contrast agent for magnetic resonance
imaging. Bioconjug Chem 4, 347-352 (1993).
146. Sun, E.Y., Weissleder, R. & Josephson, L. Continuous analyte sensing with magnetic
nanoswitches. Small 2, 1144-1147 (2006).
147. Chang, T.M.S. Blood substitutes : principles, methods, products, and clinical trials.
(Karger Landes Systems, Basel; New York; 1997).
144 148. Sun, P.Z., Schoening, Z.B. & Jasanoff, A. In vivo oxygen detection using exogenous
hemoglobin as a contrast agent in magnetic resonance microscopy. Magn Reson Med 49,
609-614 (2003).
149. Bloom, J.D. & Arnold, F.H. In the light of directed evolution: pathways of adaptive
protein evolution. Proc Natl Acad Sci US A 106 Suppl 1, 9995-10000 (2009).
150. Vymazal, J. et al. The quantitative relation between TI-weighted and T2-weighted MRI
of normal gray matter and iron concentration. JMagn Reson Imaging 5, 554-560 (1995).
151. Mani, V., Briley-Saebo, K.C., Hyafil, F. & Fayad, Z.A. Feasibility of in vivo
identification of endogenous ferritin with positive contrast MRI in rabbit carotid crush
injury using GRASP. Magn Reson Med 56, 1096-1106 (2006).
152. Yao, B. et al. Susceptibility contrast in high field MRI of human brain as a function of
tissue iron content. Neuroimage 44, 1259-1266 (2009).
153. Uchida, M. et al. Targeting of cancer cells with ferrimagnetic ferritin cage nanoparticles.
JAm Chem Soc 128, 16626-16633 (2006).
154. Shaywitz, A.J., Dove, S.L., Kornhauser, J.M., Hochschild, A. & Greenberg, M.E.
Magnitude of the CREB-dependent transcriptional response is determined by the strength
of the interaction between the kinase-inducible domain of CREB and the KIX domain of
CREB-binding protein. Molecular and cellular biology 20, 9409-9422 (2000).
155. Lelyveld, V.S., Brustad, E., Arnold, F.H. & Jasanoff, A. Metal-substituted protein MRI
contrast agents engineered for enhanced relaxivity and ligand sensitivity. JAm Chem Soc
133, 649-651 (2011).
156. Bulte, J.W. et al. Magnetoferritin: characterization of a novel superparamagnetic MR
contrast agent. J Magn Reson Imaging 4, 497-505 (1994).
145 157. Jordan, V.C., Caplan, M.R. & Bennett, K.M. Simplified synthesis and relaxometry of
magnetoferritin for magnetic resonance imaging. Magn Reson Med 64, 1260-1266
(2010).
158. Sanchez, P. et al. MRI relaxation properties of water-soluble apoferritin-encapsulated
gadolinium oxide-hydroxide nanoparticles. Dalton Trans, 800-804 (2009).
159. Kalman, F.K., Geninatti-Crich, S. & Aime, S. Reduction/dissolution of a beta-MnOOH
nanophase in the ferritin cavity to yield a highly sensitive, biologically compatible
magnetic resonance imaging agent. Angew Chem Int Ed Engl 49, 612-615 (2010).
160. Allen, M. et al. Paramagnetic viral nanoparticles as potential high-relaxivity magnetic
resonance contrast agents. Magn Reson Med 54, 807-812 (2005).
161. Liepold, L. et al. Viral capsids as MRI contrast agents. Magn Reson Med 58, 871-879
(2007).
162. Qazi, S. et al. P22 viral capsids as nanocomposite high-relaxivity MRI contrast agents.
Mol Pharm 10, 11-17 (2013).
163. Aime, S., Frullano, L. & Geninatti Crich, S. Compartmentalization of a gadolinium
complex in the apoferritin cavity: a route to obtain high relaxivity contrast agents for
magnetic resonance imaging. Angew Chem Int Ed Engl 41, 1017-1019 (2002).
164. Sana, B., Poh, C.L. & Lim, S. A manganese-ferritin nanocomposite as an ultrasensitive
T2 contrast agent. Chem Commun (Camb) 48, 862-864 (2012).
165. Karfeld, L.S., Bull, S.R., Davis, N.E., Meade, T.J. & Barron, A.E. Use of a genetically
engineered protein for the design of a multivalent MRI contrast agent. Bioconjug Chem
18, 1697-1700 (2007).
146 166. Karfeld-Sulzer, L.S., Waters, E.A., Davis, N.E., Meade, T.J. & Barron, A.E. Multivalent
protein polymer MRI contrast agents: controlling relaxivity via modulation of amino acid
sequence. Biomacromolecules 11, 1429-1436 (2010).
167. Yang, J.J. et al. Rational design of protein-based MRI contrast agents. JAm Chem Soc
130, 9260-9267 (2008).
168. Grabarek, Z. Structural basis for diversity of the EF-hand calcium-binding proteins. J
Mol Biol 359, 509-525 (2006).
169. Gifford, J.L., Walsh, M.P. & Vogel, H.J. Structures and metal-ion-binding properties of
the Ca2+-binding helix-loop-helix EF-hand motifs. The Biochemicaljournal405, 199-
221 (2007).
170. Caravan, P., Greenwood, J.M., Welch, J.T. & Franklin, S.J. Gadolinium-binding helix-
turn-helix peptides: DNA-dependent MRI contrast agents. Chem Commun (Camb), 2574-
2575 (2003).
171. Atanasijevic, T., Zhang, X.A., Lippard, S.J. & Jasanoff, A. MRI sensing based on the
displacement of paramagnetic ions from chelated complexes. Inorganic chemistry 49,
2589-2591 (2010).
172. Franz, K.J., Nitz, M. & Imperiali, B. Lanthanide-binding tags as versatile protein
coexpression probes. Chembiochem 4, 265-271 (2003).
173. Nitz, M., Franz, K.J., Maglathlin, R.L. & Imperiali, B. A powerful combinatorial screen
to identify high-affinity terbium(LII)-binding peptides. Chembiochem 4, 272-276 (2003).
174. Daughtry, K.D., Martin, L.J., Sarraju, A., Imperiali, B. & Allen, K.N. Tailoring
encodable lanthanide-binding tags as MRI contrast agents. Chembiochem 13, 2567-2574
(2012).
147 175. Gilad, A.A. et al. Artificial reporter gene providing MRI contrast based on proton
exchange. Nat Biotechnol 25, 217-219 (2007).
176. Hempstead, P.D. et al. Comparison of the three-dimensional structures of recombinant
human H and horse L ferritins at high resolution. J Mol Biol 268, 424-448 (1997).
177. Li, H. & Poulos, T.L. The structure of the cytochrome p450BM-3 haem domain
complexed with the fatty acid substrate, palmitoleic acid. Nature structuralbiology 4,
140-146 (1997).
178. Spirlet, M.R., Rebizant, J., Desreux, J.F. & Loncin, M.F. Crystal and molecular structure
of sodium aqua(1,4,7,1 0-tetraazacyclododecane- 1,4,7,1 0-tetraacetato)europate(III)
tetrahydrate Na+(EuDOTA.H20)-.4H20, and its relevance to NMR studies of the
conformational behavior of the lanthanide complexes formed by the macrocyclic ligand
DOTA. Inorganic chemistry 23, 359-363 (1984).
179. Meldrum, F.C., Heywood, B.R. & Mann, S. Magnetoferritin: in vitro synthesis of a novel
magnetic protein. Science 257, 522-523 (1992).
180. Galvez, N. et al. Comparative structural and chemical studies of ferritin cores with
gradual removal of their iron contents. Journalof the American Chemical Society 130,
8062-8068 (2008).
181. Curtis, A.R. et al. Mutation in the gene encoding ferritin light polypeptide causes
dominant adult-onset basal ganglia disease. Nature genetics 28, 350-354 (2001).
182. Vidal, R. et al. Intracellular ferritin accumulation in neural and extraneural tissue
characterizes a neurodegenerative disease associated with a mutation in the ferritin light
polypeptide gene. Journalof neuropathology and experimental neurology 63, 363 -3 80
(2004).
148 183. Bazylinski, D.A. & Frankel, R.B. Magnetosome formation in prokaryotes. Nature
reviews. Microbiology 2, 217-230 (2004).
184. Kirschvink, J.L. & Gould, J.L. Biogenic magnetite as a basis for magnetic field detection
in animals. Bio Systems 13, 181-201 (1981).
185. Mann, S., Sparks, N.H., Walker, M.M. & Kirschvink, J.L. Ultrastructure, morphology
and organization of biogenic magnetite from sockeye salmon, Oncorhynchus nerka:
implications for magnetoreception. The Journal of experimental biology 140, 3 5-49
(1988).
186. Diebcl, C.E., Proksch, R., Green, C.R., Neilson, P. & Walker, M.M. Magnetite defines a
vertebrate magnetoreceptor. Nature 406, 299-302 (2000).
187. Goranov, A.I. et al. The rate of cell growth is governed by cell cycle stage. Genes &
development 23, 1408-1422 (2009).
188. Shapiro, M.G. et al. Directed evolution of a magnetic resonance imaging contrast agent
for noninvasive imaging of dopamine. Nat. Biotechnol. 28, 264-270 (2010).
189. Voegtli, W.C., White, D.J., Reiter, N.J., Rusnak, F. & Rosenzweig, A.C. Structure of the
bacteriophage lambda Ser/Thr protein phosphatase with sulfate ion bound in two
coordination modes. Biochemistry 39, 15365-15374 (2000).
190. Zhuo, S.Q., Clemens, J.C., Stone, R.L. & Dixon, J.E. Mutational Analysis of a Ser/Thr
Phosphatase - Identification of Residues Important in Phosphoesterase Substrate-Binding
and Catalysis. JournalofBiological Chemistry 269, 26234-26238 (1994).
191. Reiter, N.J., White, D.J. & Rusnak, F. Inhibition of bacteriophage lambda protein
phosphatase by organic and oxoanion inhibitors. Biochemistry 41, 1051-1059 (2002).
149 192. Ansai, T., Dupuy, L.C. & Barik, S. Interactions between a minimal protein
serine/threonine phosphatase and its phosphopeptide substrate sequence. The Journal of
biological chemistry 271, 24401-24407 (1996).
193. Simabuco, F.M., Carromeu, C., Farinha-Arcieri, L.E., Tamura, R.E. & Ventura, A.M.
Production of polyclonal antibodies against the human respiratory syncytial virus
nucleoprotein and phosphoprotein expressed in Escherichia coli. Protein expression and
purification53, 209-215 (2007).
150