Multivalent Fluorinated Nanorings for On-Cell ¹⁹F NMR

Abstract

The design of imaging agents with a high fluorine content is necessary for overcoming the challenges of low sensitivity in ¹⁹F magnetic resonance imaging (MRI)-based molecular imaging. Chemically self-assembled nanorings (CSANs) provide a strategy to increase the fluorine content through multivalent display. We previously reported an ¹⁹F NMR-based imaging tracer, in which case a CSAN-compatible epidermal growth factor receptor (EGFR)-targeting protein E₁-dimeric dihydrofolate (E₁-DD) was bioconjugated to a highly fluorinated peptide. Despite good ¹⁹F NMR performance in aqueous solutions, a limited signal was observed in cell-based ¹⁹F NMR using this monomeric construct, motivating further design. Here, we design several new E₁-DD proteins bioconjugated to peptides of different fluorine contents. Flow cytometry analysis was used to assess the effect of variable fluorinated peptide sequences on the cellular binding characteristics. Structure-optimized protein, RTC-3, displayed an optimal spectral performance with high affinity and specificity for EGFR-overexpressing cells. To further improve the fluorine content, we next engineered monomeric RTC-3 into CSAN, η-RTC-3. With an approximate eightfold increase in the fluorine content, multivalent η-RTC-3 maintained high cellular specificity and optimal ¹⁹F NMR spectral behavior. Importantly, the first cell-based ¹⁹F NMR spectra of η-RTC-3 were obtained bound to EGFR-expressing A431 cells, showing a significant amplification in the signal. This new design illustrated the potential of multivalent fluorinated CSANs for future ¹⁹F MRI molecular imaging applications.

Graphical Abstract

INTRODUCTION

Molecular imaging enables the visualization, characterization, and quantification of biological events at the cellular and molecular levels.¹ While conventional medical imaging focuses on diagnosis, surgical guidance, and treatment monitoring, molecular imaging shows significant potential for early and specific disease detection and for efficient theranostic drug development.² Among the molecular imaging modalities, magnetic resonance imaging (MRI) possesses high spatial and temporal resolution,³ excellent soft tissue contrast,⁴ and noninvasiveness and avoids ionizing radiation.⁵ However, the inherent low sensitivity is the primary weakness of MRI. Considerable efforts have been made including creating hyperpolarized states,⁶ optimizing radiofrequency detectors,^7,8 improving data processing approaches,^9,10 and developing new imaging agents.¹¹ Paramagnetic contrast agents can improve sensitivity by affecting the relaxation, leading to increased contrast and reduced scan times. However, under standard clinical MRI techniques, administration of a high concentration of a lanthanide contrast agent is required (e.g., 469 mg/mL for MAGNEVIST), leading to environmental concerns of gadolinium contrast agent accumulation.¹² Additionally, incomplete clearance and cumulative deposits in the brain and other tissues in healthy subjects have been reported,^13–15 which increase concerns over toxicity.

One complementary strategy to overcome the low sensitivity of MRI agents is the development of the “hot spot” or “second color” imaging using heteronuclear atoms. For example, highly fluorinated molecules have generated increasing interest for ¹⁹F imaging applications.^{16 19}F has favorable properties including 100% natural abundance, a sensitivity close to ¹H, and a trace amount of mobilized ¹⁹F present in the human body.¹⁶ The limited detectable endogenous ¹⁹F gives rise to a minimal background signal, leading to high contrast for quantitative analysis.

¹⁹F MRI-based molecular imaging suffers an inherent low sensitivity limit, where a millimolar range of tissue concentration of an ¹⁹F MRI molecular tracer is required for in vivo detection, thus requiring highly fluorinated imaging agents. ¹⁷ An ¹⁹F MRI reagent suitable for molecular imaging is typically composed of a fluorinated imaging agent and a targeting agent. Significant efforts have been made to improve the sensitivity of ¹⁹F MRI. Besides enhancing the magnetic field and data processing^9,10 and optimizing the pulse sequence,¹⁸ developing more tailored ¹⁹F MRI imaging agents also plays an important part in boosting sensitivity. An ideal ¹⁹F MRI imaging agent has a high fluorine content, a single dominant resonance in the ¹⁹F NMR spectrum, and favorable relaxation parameters.^{19 19}F MRI imaging agents can be grouped into several categories: perfluorocarbon liquids (PFCs), fluorinated polymers, and small molecules. PFCs have a high fluorine content and fluorine weight percentage (wt %). However, for imaging applications, the significant hydrophobicity of PFCs requires the use of nanoemulsions, which are susceptible to droplet heterogeneity, limited biocompatibility, and environmental concerns.^16,20,21 Fluorinated polymers offer more flexibility in choices of the building blocks and have tunable biocompatibility, versatile functionality, and aqueous solubility, which make nanoemulsion-free options possible.²² However, fluorinated polymers can have multiple resonances in the ¹⁹F NMR spectra, which however is not always the case, and therefore are susceptible to undesired chemical shift artifacts in ¹⁹F MRI. Moreover, it is often difficult to possess a high fluorine content to reach the detection limit and at the same time maintain favorable relaxation parameters and aqueous solubility.^22,23

Fluorinated peptides are complementary alternatives to fluorinated polymers. Fluorinated peptides with their versatility in the choice of amino acids feature precise sequence-defined structures and highly tunable biocompatible properties. Previously, our group reported a new class of highly fluorinated peptides as ¹⁹F MRI imaging agents. Composed of N-ε-trifluoroacetyllysine (K^TFA) and lysine in an alternating pattern, these peptides possessed high aqueous solubility and a single degenerate signal.²⁴ One weakness of our fluorinated peptides was the low detection limit in a phantom MRI experiment. To overcome that low sensitivity, in later work, we prepared several amino acids with an increased fluorine content and incorporated them into shorter peptides of increased ¹⁹F wt %. Subsequently, we observed nonspecific binding of our lead peptides to cancer cells, which was removed through sequence modification.²⁵ However, preliminary cell-based experiments using protein-based molecular imaging constructs showed a limited signal, indicating the need to further improve the sensitivity.

Multivalency is one strategy to increase the fluorine content of ¹⁹F MRI imaging agents for sensitivity enhancement. Commonly seen in dendrimers, the multivalent construct, with well-defined, repeating branched polymeric chains emanating from a central core, allows the incorporation of more fluorine atoms.²⁶ The symmetrical structure also leads to a single resonance.²⁷ However, dendrimers can require long and elaborate synthesis and their large size can lead to concerns over clearance.¹⁶ Chemically self-assembled nanorings are alternative multivalent constructs suitable for ¹⁹F MRI molecular tracer design. CSANs formed through oligomerization of dimeric dihydrofolate reductase (DD) with a bis-methotrexate (bis-MTX) ligand possess an average size distribution of an octameric ring.^28–32 CSANs can be further functionalized through fusing various binding moieties to the DD subunits (e.g., Figure 1). In addition, CSANs can be disassembled by treatment with the FDA-approved antibiotic, trimethoprim, facilitating clearance.^31,33,34

Figure 1.

¹⁹F MRI molecular imaging using CSANs. (A) Functionalized CSANs are composed of a protein-specific targeting agent (maroon) and a fluorinated peptide imaging agent (colored circles). (B) Binding of the targeting agent to overexpressed cellular receptors allows for molecular imaging with high specificity.

Here, we studied the cellular interaction of our previously reported fluorinated molecular tracer composed of E₁-DD and a lead fluorinated peptide. Structural modification and subsequent cellular studies revealed that a balance of the fluorine content and binding affinity toward epidermal growth factor receptor (EGFR)-expressing cells was necessary for optimal tracer design. The lead monomeric species was then assembled into a CSAN. Optimal spectral performance and EGFR-specific binding affinity were successfully maintained. In a final cell-based ¹⁹F NMR experiment, through manipulating the cellular internalization, the optimized fluorine-labeled CSAN exhibited an amplified signal when compared with its monomeric moiety. These design principles using CSANs represent a new approach for future ¹⁹F molecular imaging applications.

EXPERIMENTAL SECTION

General Method for Peptide Preparation.

Solid-Phase Peptide Synthesis.

All peptides were synthesized on NovaSyn TGR resin using a CEM MARS extraction microwave reactor on a 25 μmol scale. Prior to the acylation reaction, N^α -Fmoc-protected amino acid (100 μmol, 4.0 equiv) was dissolved in 1 mL of DMF, followed with activation with HOBT (200 μL of 0.5 M solution in DMF, 4.0 equiv), HBTU (200 μL of 0.5 M solution in DMF, 4.0 equiv), and DIEA (200 μL of 1.0 M solution in DMF, 8.0 equiv). The reaction mixture was added to the resin and heated at 70 °C (400 W power, 2 min ramping, and 4 min holding cycle). The reaction mixture was then drained, and resin was washed with DMF three times, with DCM three times, and with DMF three times. The 9-fluorenylmethyoxycarbonyl (Fmoc) group was deprotected by addition of piperidine (2 mL, 20% v/v in DMF) to the resin and heating to 80 °C (400 W power, 2 min ramping, and 2 min holding cycle). The reaction mixture was then drained, and resin was washed with DMF three times, with DCM three times, and with DMF three times. The coupling and deprotection steps were repeated until the full-length peptide was synthesized. Following the final deprotection step, the N-terminus was reacted with maleimide-NHS. Prior to attaching the maleimide-NHS at the N-terminus, maleimide-NHS (100 μmol, 4.0 equiv) was dissolved in 1.4 mL of DMF, followed with DIEA (200 μL of 1.0 M solution in DMF, 8.0 equiv). The reaction mixture was added to the resin, and the mixture was reacted at room temperature overnight. The reaction mixture was then drained, and resin was washed with DMF three times, with DCM three times, and with DMF three times.

Peptide Cleavage and Deprotection.

Peptides were cleaved from the resin, and the side chains were deprotected at the same time through treatment of 1.9 mL of trifluoroacetic acid (TFA), 100 μL of thioanisole, 60 μL of triisopropylsilane (TIPS), and 40 μL of anisole at room temperature for 2 h. The resulting solution was drained from the syringe and precipitated into a cold diethyl ether. Peptides were then pelleted via centrifugation (2000g, 3 min, 4 °C). The supernatant was decanted.

RP-HPLC Purification of Peptides.

Crude peptides were dissolved in a 50:50 mixture of 0.1% TFA in H₂O/acetonitrile. The crude solution was purified via a Dionex Ultimate 3000 reversed-phase high-performance liquid chromatograph (RP-HPLC) using a C-18 column. The identity of the purified peptides was determined using an Ab-Sciex 5800 matrix-assisted laser desorption ionization–time-of-flight/TOF (MALDI-TOF/TOF) mass spectrometer utilizing an α-cyano-4-hydroxycinaminic acid matrix. The purity of the fraction was confirmed by analytical RP-HPLC (Dionex Ultimate 3000 RP-HPLC using a C-18 column). HPLC fractions containing the product were collected and lyophilized.

Cell Lines and Cell Culture.

The human cancer cell lines, A549 and A431 cells, were previously purchased from the American Type Culture Collection (ATCC) and were validated via the STR fingerprinting service at the Cytogenetics and Cell Authentication Core (CCAC) in the Department of Genetics at the MD Anderson Cancer Center. The cell lines were tested and certified as mycoplasma-free using a PCR mycoplasma detection kit (Applied Biological Materials Inc., Cat: G238). Raji cells were provided by the courtesy of the Wagner group (University of Minnesota). A549 and A431 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 4.5 g/L glucose and l-glutamine and supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C with 5.0% CO₂. Raji cells were cultured in Roswell Park Memorial Institute (RPMI) medium with l-glutamine and supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C with 5.0% CO₂.

Confocal Fluorescence Microscopy.

100,000 A431 cells were plated into a 35 mm μ-dish in a total volume of 2 mL. Cells were incubated overnight at 37 °C until they reached 70% confluency. Medium was aspirated and cells were rinsed two times with DMEM medium and were dosed with 500 nM RTC-3* or η-RTC-3* in DMEM (0.5% DMSO). Cells were incubated for 60 min at 37 °C and 5% CO₂ followed by aspiration and rinsing three times with Dulbecco’s phosphate-buffered saline (DPBS). Cells were fixed with 1 mL of 4% paraformaldehyde with light shaking for 10 min. Paraformaldehyde was removed and cells were rinsed three times with DPBS. 1 mL of 1 μg/mL Hoechst 34580 was added to the cells for nuclear staining with light shaking for 10 min. Hoechst was removed and cells were rinsed three times with DPBS. Cells were then imaged using Zeiss-Widefield & TIRF microscopy.

Protein Expression and Purification.

E₁-DD-CVIA fusion proteins were produced in T7 Express competent Escherichia coli cells. E. coli cell glycerol stock is provided through the generosity of the Wagner group (University of Minnesota). The E. coli cells were cultured at 37 °C (250 rpm) to the point where the OD₆₀₀ reached 0.6–0.8, and then, the protein was expressed at 37 °C for 6 h by the addition of IPTG (0.5 mM). Cells were centrifuged down. 40 mL of lysis buffer (50 mM phosphate, 300 mM NaCl, pH 7.4) and 40 mg of phenylmethanesulfonyl fluoride (PMSF) were added to the cell pellet and stirred for 30 min at room temperature. Cells were then put on ice and sonicated in 30 s intervals followed by 60 s of cooling for a total of 12 min sonication time. The lysed cells were centrifuged at 100,000g for 30 min. The supernatant was filtered. Ni affinity purification was done using a Ni HisTrap FF 5 mL column (GE Healthcare) on an AKTA fast protein liquid chromatography (FPLC) system by monitoring the absorbance at 280 nm. Proteins were eluted with a 0–100% gradient of wash buffer (50 mM phosphate, 100 mM NaCl, 40 mM imidazole, pH 7.4) and elution buffer (50 mM phosphate, 100 mM NaCl, 400 mM imidazole, pH 7.4) across 20 column volumes. Buffer exchange into phosphate-buffered saline (PBS) buffer (50 mM PBS, pH 7.4) was followed through a HiPrep desalting column (GE Healthcare) equilibrated with 1 column volume of buffer. Purified protein was analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and liquid chromatography-MS (LC-MS) using an Orbitrap Elite Hybrid mass spectrometer. The LC-MS data were analyzed by Thermo Scientific protein deconvolution.

Bioconjugation.

1.4 μL of 32 mM TCEP in pH 7.3 PBS buffer was added to 280 μL of 32 μM E₁-DD protein in pH 7.3 PBS buffer at 4 °C. After 1 h, 1.9 μL of 23 mM peptide in DMSO was first added to 280 μL of pH 7.3 PBS buffer and then added to the E₁-DD protein solution. Bioconjugation was carried out at 4 °C for 24 h. The reaction mixture was then purified through a PD-10 desalting column (Sephadex G-25 resin). Protein was collected and concentrated. The bioconjugated proteins were characterized by SDS–PAGE electrophoresis and LC-MS using an Orbitrap Elite Hybrid mass spectrometer. The LC-MS data were analyzed by Thermo Scientific protein deconvolution.

Nanoring Formation and Characterization.

A 1.8 μL portion of 2 mM dimerizer, bis-MTX DMSO stock, was added to 100 μL of pH 7.3 PBS buffer. Bis-MTX solution was added to 100 μL of 10 μM bioconjugated E₁-DD solution in an Eppendorf tube. The reaction mixture was reacted at room temperature for 40 min. Nanoring formation was characterized by SEC using a Superdex 200 Increase 10/300 gel filtration column (GE Healthcare Life Sciences, Cat: 2899094). The hydrodynamic diameters of the nanoring η-RTC-3 were measured with a Punk dynamic light scattering unit (Unchained Laboratories).

Binding Affinity Assay.

Binding affinity of the bioconjugated E₁-DD complex and nanoring was studied by flow cytometry. A549 cells were harvested and washed with Dulbecco’s phosphate-buffered saline (DPBS) three times. Aliquots of 10 × 10⁴ cells were then resuspended in DPBS solution containing the bioconjugated E₁-DD complex or nanoring at concentrations of 0, 5, 20, 100, 200, 500, and 1000 nM and incubated for 1 h at 4 °C. Cells were then pelleted, washed with DPBS three times, and resuspended in 50 μL of Alexa Fluor 647 anti-His-tag antibody solution and incubated for 30 min at 4 °C in the dark. Cells were then pelleted, washed with DPBS for 3 times, and resuspend with 1 mL of cold DPBS for analysis through a BD LSR Fortessa X-20 cell analyzer at the University Flow Cytometry Resource (UFCR). Data analysis and $K_{\mathrm{d}}$ determination were processed with GraphPad Prism8. Data were fit with “one site-specific binding” under eq 1

$$Y=\frac{B_{\mathrm{m}\mathrm{a}\mathrm{x}}\times X}{K_{\mathrm{d}}+X}$$ (1)

Binding Selectivity Assay.

Binding affinity of RTC-3 and η-RTC-3 was studied by flow cytometry. Raji was selected as the non-EGFR-expressing cell line; A549 was selected as the mid EGFR-expressing cell line; A431 was selected as the high EGFR-expressing cell line. Cells were harvested and washed with Dulbecco’s phosphate-buffered saline (DPBS) three times. Aliquots of 10 × 10⁴ cells were then resuspended in DPBS solution containing RTC-3 or η-RTC at a concentration of 500 nM and incubated for 1 h at 4 °C. Cells were then pelleted, washed with DPBS three times, and resuspended in 50 μL of Alexa Fluor 647 anti-His-tag antibody solution and incubated for 30 min at 4 °C in the dark. Cells were then pelleted, washed with DPBS three times, and resuspended with 1 mL of cold DPBS for analysis through a BD LSR Fortessa X-20 cell analyzer at the University Flow Cytometry Resource (UFCR).

In Vitro ¹⁹F NMR Experiment.

For in-cell ¹⁹F NMR experiments, A431 cells were first trypsinized and were then washed with DMEM three times to inhibit trypsin completely. 10,000,000 cells were added to a 75 mL culture flask. The appropriate amount of the protein complex dissolved in 10 mL of warmed DMEM (0.5% DMSO) was added to cells to a final concentration of 500 nM. Cells were incubated for 30 min at 37 °C and 5% CO₂. After incubation, cells were recovered by pipetting and then were transferred to Eppendorf tubes and spun at 300 rpm for 5 min. Pellets were washed two times with DPBS and one time with L-15 medium (10% D₂O). Cells were resuspended with 150 μL of L-15 medium (10% D₂O) and transferred to a 3 mm Shigemi NMR tube. The ¹⁹F NMR spectrum was obtained using a Bruker 600 MHz Avance NEO with a CryoProbe 5 mm TCI cryoprobe. Spectra were acquired at 37 °C with an acquisition time of 0.05 s and relaxation delay of 3 s, with the number of scans set to 1000 for both RTC-3 and η-RTC-3 samples.

For on-cell ¹⁹F NMR experiments, A431 cells were first trypsinized and were then washed with DMEM three times to inhibit trypsin completely. 10,000,000 cells were added to a 75 mL culture flask. The appropriate amount of the protein complex dissolved in 10 mL of cold DMEM (0.5% DMSO) was added to cells to final concentration of 500 nM. Cells were incubated for 30 min at 4 °C. After incubation, cells were recovered by pipetting and then were transferred to Eppendorf tubes and spun at 300 rpm for 5 min. Pellets were washed two times with cold DPBS and one time with cold L-15 medium (10% D₂O). Cells were resuspended with 150 μL of L-15 medium (10% D₂O) and transferred to a 3 mm Shigemi NMR tube. A ¹⁹F NMR spectrum was obtained using a Bruker 600 MHz Avance NEO with a CryoProbe 5 mm TCI cryoprobe. Spectra were acquired at 4 °C with an acquisition time of 0.05 s, relaxation delay of 3 s, receiver gain of 64, and number of scans of 3000 for the RTC-3 sample and 1024 for the η-RTC-3 sample.

A leakage check was followed with the sample being centrifuged down, and the supernatant was transferred to another 3 mm Shigemi NMR tube. A ¹⁹F NMR spectrum was obtained using a Bruker 600 MHz Avance NEO with a CryoProbe 5 mm TCI cryoprobe. Spectra were acquired at 37 °C with an acquisition time of 0.05 s, relaxation delay of 3 s, receiver gain of 64, and number of scans of 3000 for the RTC-3 sample and 1024 for the η-RTC-3 sample.

T₁ Determination.

The T₁ of RTC-3 and η-RTC-3 was determined through an inversion recovery experiment with delay times of 0.25, 0.5, 0.8, 1, 2, 4, 8, and 12 s. Eight spectra were obtained with a 90° pulse of 15 μs and a D1 of 8 s. The number of scans was set to 12 for the individual acquisition time of 0.3 s and a relaxation time of 8 s was analyzed using the relaxation fitting functions within software Topspin 4.1.4 (Bruker).

RESULTS AND DISCUSSION

In our previous studies, a prototype protein complex RTC-1 was prepared as a ¹⁹F MRI molecular tracer (Figure 2A),²⁵ composed of an EGFR-targeting E₁-DD fusion protein containing a 4-residue C-terminal cysteine containing the sequence, CVIA. While previous research has used this recognition motif for enzyme-mediated farnesylation reactions,³³ due to the single cysteine in the entire protein, we elected to conduct a bioconjugation reaction with a maleimide-terminated highly fluorinated, peptide 1 (Figure S1). RTC-1 manifested a favorable ¹⁹F NMR spectrum, for which a single degenerate signal was observed with a narrow resonance with a full width at half maximum (fwhm) of 54 Hz (Figure 2B). However, in a preliminary cell-based ¹⁹F NMR experiment, a very limited signal was observed when RTC-1 was incubated with A431 squamous carcinoma cells. Our working hypothesis was that the fluorine content of the cellular sample was not high enough to be detected by ¹⁹F NMR, requiring higher fluorine labeling. Alternatively, the dynamics associated with being attached to the cell surface or internalization could be leading to unfavorable resonance broadening. Finally, a third hypothesis was that the peptide bioconjugation perturbed the affinity of the protein construct for the EGFR. To address these questions, a systematic cellular interaction study was needed to improve our designs.

Figure 2.

Sequences and spectra of the E₁-DD conjugates. (A) Sequences of E₁-DD conjugates composed of highly fluorinated peptides with the respective fluorine content and binding affinity to A549 cells. The standard errors of RTC-3 in the table were calculated from three independent experiments. (B) ¹⁹F NMR spectra of the E₁-DD conjugates RTC-1, RTC-2, and RTC-3.

Binding affinity studies were first performed using flow cytometry to evaluate the functional effects of fluorinated peptide bioconjugation on RTC-1 interacting with A549 cells. The affinity of RTC-1 was determined to be significantly weaker ($K_{\mathrm{d}}>1000\mathrm{n}\mathrm{M}$, Figures 2A, S19, and S29) when compared with parental E₁-DD ($K_{\mathrm{d}}=121\mathrm{n}\mathrm{M}$) (Figure S30). With a significantly reduced binding affinity, the initial concentration of RTC-1 (500 nM) in the cell-based ¹⁹F NMR experiment could be one reason for the limited observed signal.

In an attempt to recover the binding affinity, structural modification of RTC-1 was performed. High fluorination has a tendency to alter the physicochemical properties of biomolecules.³⁵ As such, the bio-conjugation with highly fluorinated peptide 1 may interfere with the binding characteristics of E₁-DD to A549 cells. Hence, we modified the structure of 1, where the number of K^TFA residues was reduced and replaced with glycine. Three maleimide-terminated peptides, 2, 3, and 4, were synthesized with the total fluorine content being reduced to 3, 15, and 24, respectively (Figure S1). The charges on the peptide side chains remained unchanged to favor aqueous solubility, to promote the unfolded state, and to minimize the nonspecific binding as previously described.²⁵ The alternating pattern of K^TFA peptide 1 was also maintained to favor signal degeneracy in the ¹⁹F NMR spectrum. These three peptides were each bioconjugated to E₁-DD, forming RTC-2, RTC-3, and RTC-4 (Figure 2A) through thiolmaleimide reactions. Purification was conducted via size exclusion chromatography (SEC) to remove any unreacted peptide, and the successful bioconjugations were confirmed by both LC-MS and SDS–PAGE (Figures S7–S9 and S22–S24).

¹⁹F NMR was performed on the structurally modified proteins to evaluate the spectral performance prior to the nanoring assembly. RTC-2, RTC-3, and RTC-4 each showed a single degenerate resonance (Figure 2B). The highly overlapping resonances avoid undesired chemical shift artifacts and give rise to a high signal for future ¹⁹F MRI applications. The spectral performance also reflects the high flexibility of the peptide construct, where a degenerate resonance with an acceptable line width at half height (fwhm) was maintained after bioconjugation to the 52 kDa E₁-DD protein. These results support the use of sequence-defined peptides for other molecular imaging applications beyond EGFR targeting.

Binding affinity studies were conducted on the three modified proteins by using flow cytometry. Encouragingly, both RTC-2 and RTC-3 maintained binding affinities close to the parental E₁-DD with a $K_{\mathrm{d}}=140.2\mathrm{n}\mathrm{M}$ and 188.9 nM, respectively (Figures 2A and 3A). However, RTC-4, despite a lower fluorine content (24 ¹⁹F) than that of RTC-1 (30 ¹⁹F), showed a low binding affinity with $K_{\mathrm{d}}>1000\mathrm{n}\mathrm{M}$ (Figure 2A). These affinity studies indicated that for the E₁-DD protein conjugates to retain a good affinity for EGFR-expressing cells, a fluorine content of 15 fluorine atoms is optimal for this particular peptide sequence. The intermediate fluorine content with one–two additional K^TFA may also prove beneficial but were not evaluated in this case.

Figure 3.

Binding characteristics of RTC-3. (A) Binding affinity determination of RTC-3 to A549 cells using flow cytometry. $K_{\mathrm{d}}$ measured as 188 ± 51 nM (n = 3), squares, triangles, and dots indicating an individual independent experiment. (B) Flow cytometry specificity experiment of RTC-3 with A431, A549, and Raji cells. (C) Fluorescence microscopy of RTC-3* with A549 cells obtained at 10× magnification. The punctate pattern indicates the internalization of RTC-3* into A549 cells. In the DMSO control, background fluorescence was limited and localized in the nucleus.

Binding specificity of the best-performing RTC-3 to EGFR-expressing cells was then assessed by flow cytometry. Three cell lines with different EGFR expression levels were chosen, A431 cells (~2 × 10⁶ copies per cell), A549 cells (~0.1 × 10⁶ copies per cell), and non-EGFR-expressing Raji cells. RTC-3 bound selectively to A431 cells over A549 cells with limited binding to Raji cells (Figure 3B).

To gain further insights into the cellular interaction between bioconjugated proteins and EGFR-expressing cells, confocal fluorescence microscopy was carried out. RTC-3 was selected for fluorescence imaging due to its good binding affinity, desired spectral performance, and the highest fluorine content. To enable confocal imaging, peptide 8, an analogue of 3, was synthesized where the lysine of 3 at the C-terminus was replaced with (5-FAM)lysine (Figure S1). Fluorescein-tagged RTC-3* was prepared through the bioconjugation of 8 to E1-DD. After incubation with A549 cells, RTC-3* was shown to bind and internalize. The observed punctate patterns indicate active uptake of RTC-3* into A549 cells (Figure 3C). The fast rate of internalization of RTC-3* is anticipated to help accumulate the fluorinated protein construct for MRI imaging instead of being limited to the number of receptors on the cell surface if the protein complex can bind only to the receptor (EGFR) without subsequent internalization.

Extrapolating from the studies of RTC-1 and the subsequent modification, we next attempted to discover more bioconjugated E₁-DD proteins containing other fluorinated amino acids. Previously, we reported a molecular tracer that was composed of an EGFR-targeting peptide moiety and a short K^FF-containing peptide as the imaging agent. A single degenerate signal was observed, and a high signal-to-noise ratio (SNR) was obtained with 50 scans in the cell-based ¹⁹F NMR experiment. Considering the acceptable spectral performance, we prepared three peptides, 5, 6, and 7, which had sequences similar to those of the aforementioned short K^FF-containing imaging agent and differed in only the number of lysine residues. Only one K^FF was incorporated into each peptide to keep the fluorine content under 15 ¹⁹F (Figure S1). A PEG₃ linker was also incorporated into those peptides for additional flexibility to favor signal degeneracy and to improve the aqueous solubility. Three new peptides were bioconjugated to E₁-DD individually using the maleimide chemistry described above to form RTC-5, RTC-6, and RTC-7 (Figure 2A).

Binding studies were performed to assess any potential functional perturbations of the proteins. Compared with parental E₁-DD, RTC-5 had a similar affinity of 144 nM (Figure 2A), while both RTC-6 and RTC-7 showed slightly higher affinities of 352 and 306 nM (Figure 2A), respectively. In the context of the ¹⁹F NMR spectral performance, none of the proteins showed signal degeneracy with resonance dispersion spanning from 192 to 1290 Hz. While the absence of degeneracy was surprising, the poor spectral performance may be due to the interaction between the bis(trifluoromethyl)benzyl groups on the K^FF side chain and protein surface of E₁-DD. In this case, rotation of the bis(trifluoromethyl)benzyl groups may be hindered and may lead to multiple observable conformational states. With the need for the individual bioconjugated E₁-DD having both a degenerate ¹⁹F NMR signal and high affinity to EGFR-expressing cells, none of those three proteins showed competing characteristics to the best-performing peptide RTC-3.

To further increase the fluorine content for higher sensitivity, the best-performing protein bioconjugate, RTC-3, was selected to be engineered into a multivalent nanoring. RTC-3 was incubated with the chemical dimerizer, bis-MTX, for 30 min to form the CSAN η-RTC-3 (Figure 4A) as similarly reported by Wang et al.³³ Formation of the CSAN was characterized by SEC where a complex of larger size eluted at 16 min (Figure S18). Compared with the parental octameric CSAN, η-RTC-3 had a similar retention time in SEC (16.0 vs 16.5 min), indicating the formation of the nanoring of comparable size. The protein assembly was further confirmed by dynamic light scattering (DLS) with a hydrodynamic radius of 20 ± 2 nm (Figure S19) close to 18.1 nm of the parental CSAN and thus consistent with the average octameric oligomeric size, as previously reported. We attribute the earlier retention and larger hydrodynamic radius to the extra size from the bioconjugated peptides. With the assembly into an average octameric CSAN, the fluorine content of η-RTC-3 was boosted to 120 ¹⁹F atoms, eightfold over the fluorine content of RTC-3.

Figure 4.

Assembly, spectrum, and binding characterization of η-RTC-3. (A) CSANs were assembled from the bioconjugated E₁-DD in the presence of bis-MTX. (B) ¹⁹F NMR spectrum of η-RTC-3 with fwhm of 178 Hz was observed. (C) Binding affinity determination of η-RTC-3 to A549 cells using flow cytometry. $K_{\mathrm{d}}$ was measured as 168 ± 42 nM (n = 3). (D) Binding specificity experiment of η-RTC-3. Flow cytometry results showed that η-RTC-3 bound selectively to A431 cells with limited binding to Raji cells. (E) Fluorescent microscopy of η-RTC-3* with A431 cells obtained at 40× magnification. The punctate pattern indicates the internalization of η-RTC-3* into A431 cells, while costaining shows that a limited amount of η-RTC-3* entered the nuclei.

After the CSAN assembly was confirmed, both cellular binding and NMR performance were assessed to ensure that η-RTC-3 was suitable for cell-based imaging studies. Spectral analysis was carried out using ¹⁹F NMR. In this case, two partially overlapped resonances were observed with a fwhm of 178 Hz (Figure 4B). We attribute the increase in the fwhm of η-RTC-3 versus the monomer to a decreased tumbling rate of the larger CSAN assembly. The partial loss in signal degeneracy may be due to different spatial orientations of the fluorinated peptides within the CSAN.

Cellular binding studies toward A549 cells were examined through flow cytometry and confocal fluorescence microscopy. The affinity of η-RTC-3 was determined to be 168 ± 42 nM (Figures 4C and S28), close to the affinity of the monomeric RTC-3, 189 ± 51 nM, indicating a high affinity toward A549 cells after nanoring assembly. The affinity determined (Figure 4C) was the true monomeric dissociation constants $K_{\mathrm{d},1}$, and the affinity of η-RTC-3, $K_{\mathrm{d},N}$, was determined to be 2.6 ± 0.7 nM, based on eq 2 as previously reported,³⁶ where the valency of the octameric nanoring here was set to 8.

$$\text{apparent}{K}_{\mathrm{d},\mathrm{N}}=\frac{K_{\mathrm{d},1}}{{\text{valency}}^2}$$ (2)

With η-RTC-3 showing a similar binding profile as RTC-3 to A549 cells, binding specificity between the high EGFR-expressing A431 cells and non-EGFR-expressing cells was examined. As expected, η-RTC-3 showed selective binding to A431 cells, while very limited binding to Raji cells was observed (Figure 4D). Binding affinity and selectivity toward high EGFR-expressing cells were maintained, similar to the monomeric construct. These results demonstrated the suitability of η-RTC-3 as a molecular tracer for cell-based imaging studies.

The binding of η-RTC-3 toward EGFR-overexpressing cells was also investigated by confocal fluorescence microscopy. η-RTC-3* was assembled from RTC-3* under the same conditions as those for η-RTC-3 assembly. Octameric CSAN formation was confirmed by SEC with a retention time similar to that of η-RTC-3 (Figures S16 and S17). η-RTC-3* was next incubated with the A431 cells. The binding of η-RTC-3* to A431 cells with a punctate pattern was again observed, indicating that η-RTC-3* shared similar internalization characteristics as the monomeric RTC-3* (Figure 4E).

With η-RTC-3 possessing high cellular affinity and selectivity, cellular internalization, and acceptable spectral performance, a cell-based ¹⁹F NMR experiment was performed to examine whether the modified RTC-3 and the multivalent CSAN, η-RTC-3, could generate a signal suitable for future imaging studies. A431 cells were incubated with 500 nM RTC-3 and 62.5 nM η-RTC-3 (corresponding to 500 nM monomeric RTC-3) individually. However, no NMR signal was observed in either cell sample, similar to our previous results with RTC-1. η-RTC-3, with a fully recovered affinity and a fourfold increase in the fluorine content compared to that of RTC-1, showed no noticeable improvement in the NMR signal from RTC-1. This indicated that the fluorine content or EGFR affinity is not attributed to the loss of signal.

Prior studies have shown that increased viscosity and nonspecific binding within the cell can significantly affect NMR performance and thus could be significant factors affecting RTC-3 and η-RTC-3.³⁷ Increased chemical exchange from nonspecific interactions and decreased diffusion are anticipated to both dramatically broaden the ¹⁹F NMR signal, leading to loss of signal. Based on such a hypothesis, we tested whether the signal could be recovered by adjusting the incubation temperature from 37 to 4 °C to stop internalization. This time, for both RTC-3 and η-RTC-3 samples, the ¹⁹F NMR signal was successfully observed. In the RTC-3 cell sample, a signal with an acceptable SNR was observed after 3000 scans (Figure 5B). However, signal degeneracy was lost, showing multiple resonances with a fwhm of 187 Hz. In contrast, in the η-RTC-3 cell sample, a high SNR signal was observed with a ~3-fold reduced number of scans (1100 scans, Figure 5C). In addition, signal degeneracy was successfully maintained with a reduced fwhm of 141 Hz. In comparison, a limited signal was observed when non-EGFR-expressing Raji cells were incubated with η-RTC-3 under the same internalization-blocked conditions (Figure S40), highlighting the selectivity of η-RTC-3 toward EGFR-expressing cells.

Figure 5.

On-cell ¹⁹F NMR. (A) Illustration of RTC-3 and η-RTC-3 bound to the EGFR on the cells. (B) On-cell ¹⁹F NMR spectra of RTC-3 and η-RTC-3 with 187 and 141 Hz, respectively. RTC-3 showed multiple resonances, while η-RTC-3 showed a degenerate signal.

Due to the long relaxation times of fluorine nuclei, which can affect signal intensity, we also sought to determine the T₁ relaxation constant. Therefore, the T₁ of both RTC-3 and η-RTC-3 cell samples was measured by inversion–recovery experiments. RTC-3 had a T₁ of 2.46 s, while the η-RTC-3 cell sample had a more favorable T₁ of 1.73 s. This shorter T₁ allows for shorter scan times to improve ¹⁹F MRI sensitivity. We conclude from these experiments that the overall spectral performance of η-RTC-3 was the most promising for further imaging applications.

With the success of the cell-based ¹⁹F NMR experiment (600 MHz) on η-RTC-3, an initial in vitro phantom MRI experiment was also attempted with a 9.4T (400 MHz) magnet at a concentration of 62.5 nM in the presence of A431 cells. Unfortunately, in this preliminary experiment, the insufficient signal prevented further imaging. While it is useful for cellular applications via NMR, to reach the detection limit for MRI under a weaker magneti field and a less sensitive probe, future work will focus on further engineering the CSAN η-RTC-3. These constructs will need to increase the fluorine content through new multivalent display approaches or by adjusting the spectral properties such as T₁ and T₂ through contrast agents. Given our prior success conducting in-cell NMR with EGFR-targeting fluorinated peptides,²⁵ installing cleavable linkers onto our modified CSANs to release the fluorinated peptides provides an alternative approach for accumulating fluorinated tracer in the cell to increase the signal.

CONCLUSIONS

In these studies, we described the rational design of a multivalent molecular tracer for guiding the development of future ¹⁹F MRI molecular imaging agents. These studies build on our earlier results of a monomeric bioconjugated E₁-DD, which despite the advantages of high fluorine content and favorable spectral performance had a limited signal in cell-based ¹⁹F NMR experiments. Cellular interaction studies revealed a balance of fluorine content and binding affinity to the targeted biomarker, EGFR. In our latest design, structure-optimized RTC-3 exhibited a degenerate resonance in ¹⁹F NMR as well as high affinity and specificity for targeting EGFR-expressing cells. To further increase the fluorine content, for the first time, we assembled RTC-3 into a multivalent CSAN, η-RTC-3, and the binding characteristics of that construct were maintained with good ¹⁹F NMR behavior. EGFR-expressing cancer cells could be identified by ¹⁹F NMR spectroscopy by reducing η-RTC-3 internalization. In this case, an amplified degenerate signal was observed via cell-based ¹⁹F NMR in a short scan time. Given the modularity of this design, fluorinated imaging tags could also be applied to targeting a variety of other extracellular receptors, including those with significantly slower rates of internalization. EpCAM is one such alternative, which we have shown to internalize 24-fold slower than EGFR when targeting with CSANs while maintaining high expression on a variety of solid tumor cells.³⁸