The careful selection of zwitterionic nanoparticle coating results in rapid and efficient cell labelling for imaging-based cell tracking

Abstract

The increased clinical application of cell-based therapies has resulted in a parallel increase in the need for non-invasive imaging-based approaches for cell tracking, often through labelling with nanoparticles. An ideal nanoparticle for such applications must be biologically compatible as well as readily internalized by cells to ensure adequate and stable cell loading. Surface coatings have been used to make nanoparticle trackers suitable for these purposes, but those currently employed tend to have cytotoxic effects. Zwitterionic ligands are known to be biocompatible and antifouling, however the head-to-head evaluation of specific zwitterionic ligands for cell loading has not yet been explored. Magnetic particle imaging (MPI) detects superparamagnetic iron oxide nanoparticles (SPIONs) using time varying magnetic fields. Because MPI can produce high contrast, real-time images with no tissue depth limitation, it is an ideal candidate for in vivo cell tracking. In this work, we have conjugated hard (permanently charged) and soft (pKa-dependently charged) biomimetic zwitterionic ligands to SPIONs and characterized how these ligands changed SPION physicochemical properties. We have evaluated cellular uptake and subcellular localization between zwitterions, how the improvement in cell uptake generated stronger MPI signal for smaller numbers of cells, and how these cells can be tracked in an animal model with greater sensitivity for longer periods of time. Our best-performing surface coating afforded high cell loading within 4 hr, with full signal retention in vivo over 7 days.

Graphical Abstract

1. Introduction:

Cell therapies are garnering increasing attention in both research and clinical development due to their potential to treat currently untreatable diseases and complex injuries.^[1,2] As of 2021, cell therapies comprised nearly 400 active clinical trials, but less than 10 FDA approved products.^[3,4] Cell therapy involves the administration of autologous or allogenic cells into a patient for a therapeutic purpose, the most common of which include immune cells for targeting cancer, as well as hematopoietic and mesenchymal stem cells for regenerative purposes.^[5–7] Following their administration into a patient, these cells are expected to travel to their specific therapeutic destination (tumors and lymph nodes for immune cells, areas of injury for stem cells) to deliver their medicinal payload.^[8,9] Survival and oftentimes proliferation of these cells are paramount for eliciting and maintaining their therapeutic action.^[10]

In consideration of the in vivo environment these highly concentrated cells will face during their journey from injection to their therapeutic site, it is unsurprising that several factors will contribute to an insignificant therapeutic effect. This includes major challenges such as large-scale cell death following administration (sometimes greater than 95%),^[11] cells failing to reach their therapeutic destination,^[12] and failure to proliferate.^[13] These challenges introduce significant uncertainty in cell status and location and limit the therapeutic potential, necessitating a need for better cell tracking for monitoring cell therapies.

Cell tracking plays a crucial role in enabling clinicians to monitor the localization and proliferation of cell-based therapies, maximizing their effective therapeutic potential.^[2] The cell tracking process is always initiated by the labelling of the therapeutic cells, either in situ or more commonly ex vivo.^[14–17] These labelled cells can then be imaged and monitored using traditional medical imaging modalities, with the potential for determining both location and status of the administered cells. There are several different imaging techniques that have been used for cell tracking, including positron emission tomography^[18] and optical imaging,^[19] but one of the most frequently used modalities is magnetic resonance imaging (MRI).^[20] MRI cell tracking relies on labelling cells with paramagnetic species, most commonly metal chelates (Gd or Mn)^[21–23] or Fe-based nanoparticles (NPs).^[24] Fe-based NP solutions, more specifically superparamagnetic iron oxide NPs (SPIONs) have been the more widely used labelling strategy for cell tracking by MRI,^[20] mostly due to intracellular cytotoxicity concerns associated with Gd³⁺ in comparison to Fe²⁺.^[25] Unlike Gd which provides contrast by image brightening, SPIONs produce an image darkening effect (i.e. negative contrast) in MRI,^[26] which is non-specific and cannot be directly quantified. Magnetic nanoparticle imaging (MPI) was first developed in the early 2000s as a tracer-based, tomographic technique that allows for specific tracking and quantification of SPIONs.^[27] MPI relies on a gradient magnetic field that saturates SPIONs and then sweeps a magnetic field-free region over the image acquisition area to detect excited SPIONs and produce a hotspot image.^[28] In comparison to MRI, the MPI signal is much more specific as it only detects SPIONs and has little to no background signal due to lack of endogenous superparamagnetic iron, and has high spatial resolution of around 1 mm. ^[29] With current instrument implementations, however, the sensitivity of MRI and MPI for SPIONs is about equal.^[30] All these properties associated with a tracer-based modality makes MPI an extremely attractive imaging modality for cell tracking and has been where the majority of MPI clinical application research has been focused.^[31]

A significant amount of research and development efforts surrounding SPIONs as MPI tracers have been focused on improving SPION magnetization intensity to increase the limit of tracer detection.^[32–34] Most of these SPIONs utilize an anionic surface coating, commonly citric acid or carboxydextran, to facilitate SPION cellular uptake and improve nanoparticle aqueous solubility.^[35–37] The nature of SPION coating and its role in cellular uptake has been largely understudied. The negative charges help attract SPIONs to positively charged components of the extracellular membrane, but the high hydrophilicity of these strong anions precludes meaningful interaction with most of the cell surface and reduces intracellular uptake.^[38] Some neutral (polyethylene glycol) and cationic (trialkylammonium-based) surface coatings have been evaluated, but these readily form large protein coronas in blood and serum, masking SPIONs from cellular interaction.^[39,40] Many cationic coatings are often cytotoxic.^[41]

Zwitterionic molecules are charged species with a net neutral charge, and comprise many of the biomolecules endogenous to cells, such as amino acids, proteins, and lipids.^[42] Zwitterionic molecules also possess excellent antifouling properties^[43] and given their biomimetic electronic state, have been shown to facilitate high intracellular nanoparticle uptake while remaining highly cytocompatible.^[44] These properties make zwitterionic molecules an excellent potential surface coating for SPIONs to achieve high levels of non-cytotoxic cell loading leading to greater signal per cell. There have been some reports of zwitterionic SPIONs for MRI based cell tracking,^[45,46] but we are aware of none that have systematically evaluated different zwitterionic classes, and none for optimizing ex vivo cellular loading with SPION for MPI. In this work, we have functionalized previously described SPIONs possessing excellent MPI magnetic properties^[47] and modified them to bear a polymeric (polymaleic anhydride, PMAO) surface coating that was functionalized with “hard” and “soft” zwitterions with permanent or pKa-derived charge-states, respectively. Zwitterion functionalization of originally anionic PMAO-coated SPIONs did not disrupt the measured physicochemical properties or affect their MPI signal generating capacity but did improve stability in cell culture media. Finally, we have shown that the zwitterion-coated SPIONs had much higher levels of intracellular uptake than the anionic PMAO-coated SPIONs in several cell types, including human mesenchymal stem cells (MSCs). The higher degree of uptake observed with zwitterionic SPIONs led to greater MPI signal for longer periods of time than the PMAO-coated nanoparticles when labelled MSCs were injected in mice for in vivo tracking by MPI. These zwitterionic SPIONs may be a key contributor to unlocking the potential for effective cell tracking by MPI.

2. Results and Discussion:

2.1. Synthesis and Dry Characterization of Hard and Soft zwitterionic SPIONs:

The single-core SPIONs were synthesized following the procedure outlined in S. Liu et al.^[48] The SPION core exhibits an MPI signal that is ~3-times better than commercially available ferucarbotran (Resovist) and is stabilized by a coating of poly(maleic anhydride-alt-1-octadecene, PMAO). In addition to the strong MPI signal generated by the nanoparticle core, the PMAO coating on these nanoparticles readily hydrolyzes in water to polymaleic acid. The presentation of surface carboxylic acids serves as a chemical handle for facile zwitterionic functionalization.^[49]

Two biomimetic molecules of the cellular membrane outer leaflet, sulfobetaine (SB) and pyridinium sulfobetaine (PSB), were synthesized and characterized (Supplementary Scheme 1 & Supplementary Fig. 1–9) as our chosen “hard” zwitterions. SB and PSB possessed a permanent positive charge on their nitrogen (dimethylammonium nitrogen for SB and pyridinium nitrogen for PSB), and an effectively permanent negative charge in biological environments on the terminal sulfonate group, due to the negative pKa. Both molecules have been previously reported showing excellent biocompatibility and improvement in cellular uptake when functionalized on nanoparticle surfaces.^[50–54] Functionalization of Fe@PMAO was achieved using standard coupling chemistry. The carboxylic acids on the SPION were activated with1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) to drive amide bond formation to link SB or PSB to Fe@PMAO, resulting in the hard zwitterion functionalized Fe@SB and Fe@PSB SPIONs (Fig. 1A, top scheme). Endogenous natural and semi-natural amino acids L-cysteine and DL-homocysteine were chosen as the soft zwitterionic molecules. Both cysteine (Cys) and homocysteine (HCys) can enter cells as substrates of the transporter ASCT2 and can also diffuse through SLC transporters.^[55,56] The amino acid backbone of these molecules is zwitterionic between the pH ranges of ~3 – ~10. EDC coupling was used to install a maleimide group on Fe@PMAO, followed by a change in buffer and addition of either Cys or HCys (Fig. 1A, bottom scheme). Maleimide allowed the selective use of the thiol as the nucleophilic coupling partner at a controlled pH, preserving the amino acid backbone of both molecules and resulting in the soft zwitterion functionalized Fe@Cys and Fe@HCys SPIONs.

Figure 1. Synthesis and dry characterization of hard and soft zwitterionic superparamagnetic iron oxide nanoparticles (SPIONs).

A) Synthetic scheme for the synthesis of hard and soft zwitterion surface functionalized SPIONs. B) Transmission electron microscopy images of nanoparticles before and after functionalization. Scale bar represents 100 nm. C) Labelled Fourier-transformed infrared spectroscopy of SPIONs before and after functionalization.

Nanoparticle morphology before and after functionalization was evaluated by transmission electron microscopy (TEM) to detect any morphological changes to the nanoparticle core as a result of chemical functionalization (Fig. 1B). Both hard and soft zwitterion-functionalized nanoparticle morphology appeared unchanged in comparison to Fe@PMAO.

Surface functionalization with zwitterions was evaluated by Fourier-transformed infrared spectroscopy (FTIR). Fe@SB and Fe@PSB (Fig. 1C, left panel) showed characteristic FTIR signals that were not observed for Fe@PMAO: S=O stretch from 1030 – 1070 cm⁻¹, sulfonic acid stretch from 1120 – 1230 cm⁻¹, a decrease in the intensity of the COO⁻ stretch ~1600 cm⁻¹, and a formation of a conjugated anhydride peak at 1770 cm⁻¹. All signals were in agreement for similar sulfobetaine molecules.^[57] The formation of the anhydride peak was of particular interest, and was the result of an iminolization-cyclization reaction following the first amide bond formation between the polymaleic acid and SB or PSB.^[49,58,59] The formed imide is known to be more hydrolysis-resistant than the amide counterpart, and will therefore improve the hydrolytic resistance of these functionalized nanoparticles. FTIR analysis of Fe@Cys and Fe@HCys (Fig. 1C, right panel) proved more challenging given the lack of new and distinct functional groups following functionalization in comparison to those already present in Fe@PMAO. No change in the COO⁻ stretch ~1600 cm⁻¹ was observed, but an increase in the C=O stretch at 1700 cm⁻¹ was observed. This stretch is associated with new carboxylic acid groups formed by the reaction between the maleimide and thiol of Cys or HCys, the new carbonyls on the hydrolyzed maleimide, formed polymaleimide, and on Cys or HCys. The appearance of the conjugated anhydride peak attributed to polymaleimide formation was again observed at 1770 cm⁻¹.

2.2. Wet characterization of hard and soft zwitterionic SPIONs:

Nanoparticle properties before and after functionalization were evaluated in solution to examine their suitability for cell uptake and to further determine if their initial magnetic properties had changed following functionalization. Changes in magnetic properties would have indicated significant changes to the SPION core, or aggregation behavior in solution.

Zeta potential measurements for all SPIONs were acquired at varying pHs to determine both aqueous dispersibility and the effect of the zwitterionic molecules on the electric double layer (EDL). (Fig. 2A). At basic pH = 11, the surface charge of all nanoparticles remained highly negative, with zeta potentials of −49.91 ± 1.47 mV, −46.86 ± 2.34 mV, −43.82 ± 6.62 mV, −47.44 ± 4.32 mV, and −47.93 ± 3.31 mV for Fe@PMAO, Fe@SB, Fe@PSB, Fe@Cys, and Fe@HCys, respectively. At pH = 11, the functional groups of all five nanoparticles were net negative, contributing to the highly negative value and the high dispersibility of all five nanoparticles. The charged nitrogen in Fe@SB and Fe@PSB had little effect on the overall zeta potential at this pH given the non-terminal position of the positive charge having less contribution on the EDL than the terminal sulfonate. This effect was observed in previous work with nanoparticles also possessing the SB moiety.^[60] At physiological pH = 7.4, the zeta potential of all five SPIONs remained similar and negative, though slightly more positive than at pH = 11, with values of −42.57 ± 8.22 mV, −43.50 ± 1.21 mV, −42.22 ± 3.22 mV, −41.78 ± 6.76 mV, and −46.16 ± 1.92 mV for Fe@PMAO, Fe@SB, Fe@PSB, Fe@Cys, and Fe@HCys, respectively. No significant change in nanoparticle zeta potential between pH = 11 and pH = 7.4 was expected. For Fe@PMAO, Fe@SB, and Fe@PSB, the net charge of these functional groups remained the same for pH = 11 and pH = 7.4. For Fe@Cys and Fe@HCys, while the terminal primary amine became positively charged, the carboxylic acid adjacent to the thioether linkage that resulted from maleimide hydrolysis remained negatively charge, likely contributing to the still negative value zeta potential, though this does not interfere with the ability for the terminal amino acid backbone of cysteine/homocysteine to remain zwitterionic. However, zeta potential values at this pH were similar to other cysteine-functionalized nanoparticles reported previously.^[61] At pH = 3, below the pKa of the carboxylic acids, the zeta potentials all nanoparticles became significantly more positive compared to pH = 11 and pH = 7.4 (p<0.05) for all five nanoparticles. Fe@PMAO, Fe@SB, and Fe@PSB possessed similar zeta potentials of −24.38 ± 2.46 mV, −21.83 ± 9.74 mV, and −25.71 ± 3.28 mV, respectively, while Fe@Cys and Fe@HCys possessed significantly more positive charges than the other nanoparticles at pH = 3 (−7.64 ± 6.36 mV and −5.25 ± 4.64 mV, respectively. p<0.05). Interestingly, the soft zwitterion nanoparticles had a much more positive zeta potential value compared to the hard zwitterion nanoparticles. The zeta potential data, in addition to FTIR characterization, supported the successful conjugation of ligands to the PMAO surface through amide bond formation.

Figure 2. Wet characterization of hard and soft zwitterionic superparamagnetic iron oxide nanoparticles (SPIONs).

A) Zeta potential measurements of zwitterion surface functionalized SPIONs at varying pHs in 1 mM KNO₃. B) Z-average measurements of zwitterion surface functionalized SPIONs in phosphate-buffered saline (pH 7.4, PBS). C) Photo of Fe@PMAO and zwitterion surface functionalized SPIONs dispersed in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin and left on the benchtop overnight. D) Magnetic relaxometry of SPIONs in PBS. Sensitivity and resolution were derived from the relaxometry curves. Data in A) and B) are presented are boxplots of n = 3 individual nanoparticle syntheses. Statistical analysis was done by one-way ANOVA followed by a Tukey post-hoc test. *p<0.05, **p<0.01. Data in D) are presented as lines or bars of n = 3 replicate measurements from one nanoparticle synthesis.

Hydrodynamic nanoparticle size (Z-average) was also measured by DLS in phosphate buffered saline (pH = 7.4, PBS) to observe deviations in hydrodynamic size following functionalization, as well as to examine the existence of any aggregation behavior (Fig. 2B). Prior to functionalization, the hydrodynamic size of Fe@PMAO was 81.60 ± 16.83 nm. Hydrodynamic size following hard zwitterion functionalization was not significantly changed, with Fe@SB and Fe@PSB having hydrodynamic sizes of 93.75 ± 25.59 nm and 118.95 ± 17.16 nm, respectively. Functionalization of SPIONs with soft zwitterions, Fe@Cys and Fe@HCys, significantly increased the hydrodynamic size of the nanoconjugates to 278.84 ± 81.51 nm and 275.46 ± 34.00 nm, respectively (p<0.01). This was likely the result of a small degree of aggregation behavior caused by inter- or intraparticle interactions between the terminal opposing charges in PBS. The dispersion of these nanoparticles in a more ionically complex system (cell culture media supplemented with both fetal bovine serum (FBS) and penicillin/streptomycin (P/S)) resulted in all four zwitterion-functionalized formulations remaining well-dispersed (Fig. 2C). Importantly, while zwitterion functionalized SPION remained well-dispersed, Fe@PMAO aggregated and settled to the bottom of the tubes (Fig. 2C). Though polymaleic acid makes Fe@PMAO highly soluble given the strong negative surface charges, they are more susceptible to forming large protein coronas that cause them to crash out of solution.^[62]

The magnetic properties of the nanoparticle formulations were evaluated by magnetic relaxometry in PBS in order to evaluate if surface functionalization would impact their effective MPI contrast properties. Magnetic relaxation curves between all five SPIONs were superimposable (Fig. 2D, top panel). Sensitivity, the maximum peak intensity, remained similar for Fe@PMAO, Fe@SB, and Fe@PSB (52.29 ± 0.27 mg⁻¹ Fe, 46.64 ± 0.43 arb. Units/mg Fe, and 59.26 ± 0.37 mg⁻¹ Fe, respectively). Fe@Cys and Fe@HCys showed a slight decrease in sensitivity (32.67 ± 0.11 mg⁻¹ Feand 30.47 ± 0.35 mg⁻¹ Fe, respectively), likely due to peak broadening due to slight nanoparticle aggregation in PBS, agreeing with DLS measurements. Specificity, the full width at half maximum intensity (FWHM), remained similar between Fe@PMAO and the functionalized nanoparticles (0.293 ± 0.002 T/mg Fe, 0.328 ± 0.001 T/mg Fe, 0.280 ± 0.002 T/mg Fe, 0.403 ± 0.003 T/mg Fe, and 0.303 ± 0.001 T/mg Fe, for Fe@PMAO, Fe@SB, Fe@PSB, Fe@Cys, and Fe@HCys, respectively), suggesting that changes to the surface chemistry do not alter the MPI contrast production properties of the PMAO.

The culmination of these wet characterization results showed that zwitterion functionalization of Fe@PMAO improved the solubility characteristics of the nanoparticles while maintaining their already excellent magnetic properties.

2.3. Cellular uptake, localization, and performance of zwitterionic surface functionalized SPIONs:

All five nanoparticles were evaluated for cell uptake in three different cell lines representative of cell types typically tracked by imaging methods, including MPI: cancer cells (triple negative human breast cancer MDA-MB-231), stem cells (human mesenchymal stem cells MSC), and immune cells (mouse bone marrow macrophage RAW 264.7). Optimization of nanoparticle incubation concentration and incubation time was first performed, ultimately leading to the use of a nanoparticle concentration of 50 μg/mL Fe for 4 hr incubation for effective cell uptake (Supplementary Fig. 10 & 11). Where previous reports for SPION loading into cells required 24 to 48 hr of incubation to reach uptake maximum, the use of zwitterionic ligands resulted in more rapid uptake, reaching plateau by 4 hrs.^[63,64] No transfection reagents, rocking, or additional reagents beyond the nanoparticles and culture media were used. A histological method using both Perl’s Prussian Blue (PPB, visualization of Fe in blue) and Nuclear Fast Red (visualization of cytoplasm and nucleus in pink) was employed to qualitatively evaluate nanoparticle uptake (Fig. 3).

Figure 3. Identifying cellular uptake of hard and soft zwitterion functionalized superparamagnetic iron oxide nanoparticles (SPIONs).

SPIONs functionalized with or without different zwitterionic functional groups were incubated with each cell type grown to 80% confluency (50 μg/mL Fe concentration for 4 hr at 37°C, 5% CO₂, and humidified environment). Cells deposited on a glass slide using a Shandon^™ Cytospin^™ were washed, fixed, and stained with Perl’s Prussian Blue (blue stain) to detect Fe, and with Nuclear Fast Red to (pink stain) to detect cellular components prior to imaging. All scalebars represent 50 μm.

The loading of MDA-MB-231 cells (Fig. 3, left panels) resulted in low Fe@PMAO, Fe@SB, and Fe@PSB uptake, with higher levels of uptake observed during incubation with Fe@Cys and Fe@HCys. The increase in the soft zwitterion nanoparticle uptake was likely the result of the higher levels of expression of amino acid transporters to support higher metabolic demands of the cancer cell line.^[65] Incubation of RAW 264.7 cells (Fig. 3, center panels) with SPION formulations appeared to show high levels of uptake with Fe@PMAO, but the majority of the observed PPB stain was in between the cells and not within the cytoplasm (i.e. limited blue-pink overlay). This extracellular accumulation was the result of interactions between the PMAO coating, highly proteinaceous cell culture media, and the positively charged glass coverslips used for culture. From a practical point of view, these interactions are other factors that limit internalization of these highly anionic nanoparticles and make them less reliable for cell tracking. The hard zwitterion nanoparticles Fe@SB and Fe@PSB, in contrast to Fe@PMAO, resulted in higher uptake in RAW 264.7 cells. The punctate cytoplasmic appearance of these nanoparticles is suggestive of uptake via intracellular vesicles. The soft zwitterionic nanoparticles Fe@Cys and Fe@HCys qualitatively showed the highest uptake in this cell line overall, again demonstrating a punctate Fe-stain pattern within the cytoplasmic areas of the cells. For all zwitterionic nanoparticle formulations, there was little to no Fe stain seen outside of the cells, a positive indication of their ease of washing and resistance to aggregation in cell culture media. The loading of MSC (Fig. 3, right panels) was qualitatively similar to the other two cell types: very little intracellular staining of Fe@PMAO. Overall higher and punctate PPB staining was seen in all four zwitterion-functionalized nanoparticles, with the highest qualitative uptake observed again with Fe@Cys and Fe@HCys. Our results show that it will be important to optimize SPION concentration and incubation times for any specific cell type being tracked.

To evaluate the method of transport of nanoparticles into cells, cell loading experiments were repeated with incubations at 4°C, a temperature at which active cellular transport is significantly reduced (Fig. 4)^[66]. No Fe staining was observed in MDA-MB-231 cells (Fig. 4, left panels), indicating that nanoparticle uptake in these cells is energy-dependent. For both RAW 264.7 macrophages (Fig. 4, center panels) and MSC (Fig. 4, right panels), neither Fe@PMAO nor the hard zwitterion-functionalized nanoparticles showed any Fe staining, again indicating no nanoparticle uptake in the absence of active transport. However, some intracellular Fe staining, although less than at 37°C, was observed in both RAW 264.7 and MSC incubated with the soft zwitterion-functionalized nanoparticles. Given that both soft zwitterions are amino acid functional groups, uptake was likely the result of facilitated diffusion, a process that commonly occurs with critical metabolites, such as amino acids and glucose, where carrier proteins and pores are able to translocate these molecules.^[67] That soft zwitterion-functionalized nanoparticles could generate cell labelling even in energetically unfavorable cellular conditions is a potentially powerful feature for these formulations.

Figure 4. Identifying cellular uptake of hard and soft zwitterion functionalized superparamagnetic iron oxide nanoparticles (SPIONs) with limited cellular active transport.

SPIONs functionalized with or without different zwitterionic functional groups were incubated with each cell type grown once grown to 80% confluency (50 μg/mL Fe concentration for 4 hr at 4°C, 5% CO₂, and humidified environment). Cells deposited on a glass slide using a Shandon^™ Cytospin^™ were washed, fixed, and stained with Perl’s Prussian Blue (blue stain) to detect Fe, and with Nuclear Fast Red to (pink stain) to detect cellular components prior to imaging. Left and center panel scalebars represent 50 μm and the right panel scalebar represents 35 μm.

In order to confirm the nanoparticle uptake observed through histology and further evaluate their subcellular localization, SPION-loaded cells were evaluated by TEM (Fig. 5). In MDA-MB-231 cells (Fig. 5, left panels), all five nanoparticle formulations result in the formation of endocytotic vesicles encapsulating the nanoparticles. As observed by histology, Fe@PMAO resulted in mostly extracellular nanoparticles very close to the cell surface. Those Fe@PMAO nanoparticles found intracellularly appear to be mostly within multivesicular bodies (MVB). Fe@SB and Fe@PSB were more extensively found in the intracellular spaces, with Fe@PSB localizing primarily within MVBs, and Fe@SB localizing to early endo/lysosomes. Likewise, both Fe@Cys and Fe@HCys were more extensively localized intracellularly relative to Fe@PMAO, with localization for both nanoparticle formulations occurring in the late endo/lysosome. TEM investigation of RAW 264.7 macrophages (Fig. 5, center panels) incubated with Fe@PMAO confirmed what was observed by histology (Fig. 3), with majority of nanoparticles remaining extracellular. Those nanoparticles that were intracellular were localized within the macrophage phagosome. Incubation with all of the zwitterionic-functionalized ligands resulted in a substantially higher degree of nanoparticle intracellular localization than Fe@PMAO. Fe@PSB localized to the phagosome and MVBs, whereas Fe@SB was found to be confined within the phagosome and the early endo/lysosomes. Similarly, Fe@Cys and Fe@HCys were observed within late endo/lysosomes with some phagosome localization. The phagosome uptake observed in all five of these nanoparticle formulations support the greater overall uptake seen in RAW 264.7 cells by histology (Fig. 3). For MSC (Fig. 5, right panels), very little Fe@PMAO uptake was observed. Higher uptake for all zwitterionic nanoparticles was observed in comparison to Fe@PMAO As was observed in the histology, there were no identified sites of Fe@PMAO uptake in MSCs. The subcellular localization of the respective zwitterionic nanoparticles were similar to what was observed with the other two cell lines: Fe@PSB and Fe@SB primarily localized within MVBs and early endo/lysosomes, while Fe@Cys and Fe@HCys showed overall higher uptake primarily localized within late endo/lysosomes. The observed subcellular localization of the functionalized nanoparticles is corroborated by previous reports of nanoparticle transport into cells, primarily through endo- or pinocytosis.^[68,69] While endocytosis of all nanoparticles will likely lead to lysosomal catabolism, the high level of lysosomal localization for Fe@Cys and Fe@HCys is an interesting feature. Amino acids are known activators of RAG proteins on the lysosomal surface as an initiation to recruitment of mTORC1, promoting cellular anabolic activity and potentially driving further cell uptake of the amino-acid labelled nanoparticles.^[70,71]

Figure 5. Intracellular localization following cellular uptake of hard and soft zwitterion functionalized superparamagnetic iron oxide nanoparticles (SPIONs).

SPIONs functionalized with or without different zwitterionic functional groups were incubated with each cell type grown to 80% confluency on glass slides (50 μg/mL Fe concentration for 4 hr at 37°C, 5% CO₂, and humidified environment). Cells were washed, fixed, sectioned, and stained for imaging by transmission electron microscopy. M: multivesicular body. E: endo/lysosome. P: phagosome. Large scalebars represent 1 μm and small scalebars represent 150 nm.

2.4. In vivo Cellular Tracking by MPI:

Given the class-wise similarities in both magnitude of cellular uptake, as well as the mode of transport and subcellular localization, we chose to evaluate the best performer from each of the soft and hard zwitterion categories: Fe@SB for the hard zwitterions due to its more favorable intracellular localization into early endo/lysosomes compared to Fe@PSB, and Fe@HCys for the soft zwitterions due to its overall qualitatively higher level of cell uptake. Cellular uptake experiments were performed once again, except that following the wash step the cells were pelleted and imaged by MPI to quantitatively evaluate nanoparticle uptake into cells (Fig. 6). The MPI signal of the largest cell pellet for each cell type was imaged by MPI (Fig. 6A) and resulted in similar extents of nanoparticle loading as the respective nanoparticle-cell combinations qualitatively observed by light microscopy in Figure 3. Fe@PMAO and Fe@HCys resulted in ~2.2x higher levels of signal intensity as compared to Fe@SB when loaded into MDA-MB-231 cells (Fig. 6B, left panel). Labelling with Fe@PMAO and Fe@HCys generated MPI signal above the limit of detection (LOD), with as little as 19,000 cells. Both Fe@SB and Fe@HCys produced higher MPI signal as compared to Fe@PMAO (1.4x and 2x respectively) when loaded into RAW 264.7 cells (Fig. 6B, center panel). Both hard and soft zwitterion-functionalized nanoparticles generated MPI signal above the LOD with as little as 25,000 cells. Similarly, both Fe@SB and Fe@HCys resulted in higher MPI signal than Fe@PMAO (2.2x and 1.4x respectively) when loaded into MSC (Fig. 6B, right panel). Both Fe@SB and Fe@HCys generated MPI signal above LOD with as little as 15,625 cells. For each cell type, the signal intensity on a per cell basis was compared (Fig. 6C) to determine the labelling efficacy (LE) of the nanoparticles within each cell population. Fe@PMAO and Fe@HCys both had an improved LE over Fe@SB (0.215 ± 0.021 arb. Units/cell and 0.235 ± 0.023 arb. Units/cell compared to 0.112 ± 0.032 arb. Units/cell, respectively. p < 0.0001) in MDA-MB-231 cell. However, when incubated with RAW 264.7 cells, Fe@HCys had a significantly greater LE than both Fe@SB and Fe@PMAO (0.364 ± 0.035 arb. Units/cell compared to 0.236 ± 0.035 arb. Units/cell and 0.176 ± 0.014 arb. Units/cell, respectively. p < 0.0001), though the LE for Fe@SB was also significantly greater than that of Fe@PMAO (p < 0.05). Both Fe@SB and Fe@HCys again resulted in higher LE than Fe@PMAO (0.152 ± 0.045 arb. Units/cell and 0.109 ± 0.023 arb. Units/cell compared to 0.067 ± 0.020 arb. Units/cell, p < 0.01 for Fe@PMAO versus Fe@SB) upon incubation with MSC. Ultimately, given the receptor and transporter expression differences between different cell types, it would be valuable to further evaluate a specific cell uptake mechanism for hard versus soft zwitterion-modified nanoparticles prior to use in potential cell therapy.

Figure 6. Magnetic nanoparticle imaging (MPI) of cell pellets labelled with hard and soft zwitterion functionalized superparamagnetic iron oxide nanoparticles (SPIONs).

A) MPI images of cell pellets (150,000, 500,000, and 250,000 cells for MDA-MB-231, RAW 264.7, and MSC, respectively) following labelling with nanoparticles (50 μg/mL Fe concentration for 4 hr at 37°C, 5% CO₂, and humidified environment). B) MPI signal of cell pellets containing decreasing numbers of cells from the initial pellet after dilution with phosphate-buffered saline (pH 7.4). LOD represents the limit of detection of the MPI, defined by the signal generated by unlabelled cells. Data are presented as line graphs of continual dilution of a single population of pelleted, labelled cells. C) The labelling efficacy of the nanoparticles for each cell type defined by the signal intensity generated per cell. Data as boxplots of n = 5 (MDA-MB-231) or n = 3 (RAW 264.7, MSC) labelled cell signal pellet signal intensities. Statistical analysis was done by two-way ANOVA followed by a Tukey post-hoc test. *p<0.05, **p<0.01, and ****p<0.0001.

The totality of the results obtained from the histology, TEM, and in vitro MPI experiments led us to conclude that the soft zwitterion-functionalized nanoparticles, Fe@HCys, was the best aggregate performer for loading cells with SPION across the three cell lines evaluated. In addition to generating the highest contrast across all three MPI cell labelling experiments, the ability for these nanoparticles to achieve intracellular localization in the absence of active transport is a powerful property from a cell tracking perspective, as it ensures labelling can occur even in suboptimal conditions. Because of this, we chose to evaluate the performance of Fe@HCys in comparison to the non-zwitterionic Fe@PMAO by tracking one of the most therapeutically relevant cell types, MSCs, across 7 days post-injection (Fig. 7). Importantly, Fe@HCys did not change MSC viability even after 24 hr incubation (Supplementary Fig. 12). Previous work has also demonstrated that SPION only the chondrogenic differentiation of MSCs is impacted by SPION loading, which is attributable to the Fe-content itself, not the nanoparticle surface ligand.^[72]

Figure 7. Comparing mesenchymal stem cell tracking in mice between soft zwitterion functionalized and anionic superparamagnetic iron oxide nanoparticles (SPIONs).

A) Location of labelled SPION (white arrow) injection in the MPI scan region. and the associated MPI images of the region-of-interest following injection on day 0 and repeat scans on day 3 and 7. Red arrows denote gastronintenstinal signal. B) Signal intensity of the labelled cells for each timepoint. Data are presented as cell pellet signal intensities for n = 3 mice. Statistical analysis was done by one-way ANOVA followed by a Tukey post-hoc test. **p<0.01, and ***p<0.001.

In vivo MPI tracking of MSC labeled with Fe@PMAO or Fe@HCys over 7 days is shown in Fig. 7. There was no outwardly observable toxic response to cells labeled with Fe@PMAO or Fe@HCys in any of the mice over the 7 days of monitoring. At all 3 imaging timepoints, the MPI signal was significantly higher for MSC labeled with Fe@HCys compared to MSC labeled with Fe@PMAO (Fig. 7C). In agreement with the in vitro MPI results (Fig. 6), overall MPI labelling signal by Fe@HCys was stronger than by Fe@PMAO (Fig. 7C). In fact, MPI signal generated from Fe@HCys labelled cells, remained significantly higher than Fe@PMAO throughout the course of the experiment: ~2.0x higher on day 0 (p<0.01), ~2.4x higher on day 3 (p<0.001), and ~2.1x higher on day 7 (p<0.001).

Cells labelled with either nanoparticle were able to maintain SPION-generated signal over 7 days without any significant change in signal intensity, though Fe@HCys was able to achieve this with better labelling efficacy and higher MPI signal. Low intensity signal outside of the injection site is observed (red arrows, Fig. 7B), which appears in both mice. In mice administered Fe@PMAO this signal remains stable over the course of 7 days, however is decreases over the course of 7 days in mice administered Fe@HCys. These results highlighted the better labelling and higher MPI signal provided by Fe@HCys, resulting in higher signal retention in cells in vivo over time. The ability for the treated particles to maintain this signal over a period of 7 days also demonstrates the stability of the internalized particles; any cell death would have shown a diffusion and decrease in MPI signal, while degradation of the paramagnetic core of Fe@HCys would have shown a decrease in signal over time.

3. Conclusion:

The functionalization of SPION with either hard or soft zwitterionic ligands resulted in nanoparticles of similar characteristics and magnetic properties as the precursory Fe@PMAO nanoparticles. However, the functionalized nanoparticles were more colloidally stable when dispersed in cell culture media, while Fe@PMAO aggregated and precipitated out. Cell labelling experiments with these nanoparticles qualitatively (PPB) and quantitatively (MPI) showed higher uptake and more extensive intracellular localization with zwitterionic functionalization. Most interestingly, the amino acid functionalized soft zwitterionic nanoparticles were able to achieve intracellular localization in metabolically unfavorable conditions that other nanoparticle formulations could not. Overall, the zwitterion functionalized nanoparticles resulted in higher MPI signal for smaller numbers of cells, giving a broader dynamic range by which cells can be confidently identified and tracked. High signal-to-noise tracking of labelled MSCs in mice over 7 days was demonstrated, where Fe@HCys was able to give significantly higher signal over the course of cell tracking than those labelled with Fe@PMAO.

In summary, this study highlights the strength of zwitterionic ligands in enhancing intracellular cell labeling with nanoparticles. While anionic groups contribute to the dispersibility of nanoparticles and draw their proximity to cell surfaces, they fall short in promoting substantial cell uptake. On the other hand, zwitterionic ligands not only maintain outstanding dispersibility in vitro, which is key to enhancing cell labelling, but also exhibited greater cell culture media solubility. Additionally, their biomimetic properties facilitate superior intracellular localization. This work has also provided a new route for unlocking the full potential of cellular MPI: focusing on improving the stabilizing functional groups on the nanoparticle surface to maximize rapid, transfection reagent-free cell uptake. The surface functional groups and conjugation chemistries utilized in this work are simple, cost-effective, and scalable to be employed for future pre-clinical evaluation. In essence, zwitterions emerge as a promising solution for advancing MPI applications in cell tracking.

Experimental Section

All animal studies were conducted under (AUP 2023–113) and were approved by the University of Western Ontario Animal Care and Use Subcommittee. Mice were housed doubly under a 12 h light/dark cycle, and ambient temperature of 20–24°C and 45 to 65% humidity. All rats were provided access to food (Rodent Laboratory Chow) and water ad libitum.

General reagents:

All chemical reagents were purchased from Sigma-Aldrich and used as is unless otherwise reported. All solvents were HPLC grade, except for water (18.2 MΩ cm Millipore water).

Experimental procedures:

All MPI imaging was performed using a MOMENTUM^™ MPI (Magnetic Insight Inc., Alameda CA). Non-MPI image analyses and quantifications were performed using Fiji and all MPI image analyses and quantifications were performed using Horos. GraphPad Prism 10.0 was used to generate all graphs, graphical figures, and statistical analyses.

PMAO coating of iron oxide nanoparticles:

Particle coating using PMAO was performed according to published work with some modifications.^[73] A solution of PMAO in chloroform at an initial concentration of 0.137 M was added to a solution of oleic acid coated SPIOs in chloroform, at a concentration of 0.1 μM, to reach 300 polymer monomer units per nm² of nanocrystal surface. The particle-polymer solution was sonicated for 1 h in a bath sonicator followed by slow evaporation of the solvent under controlled pressure, using a rotary evaporator. A solution of bis(hexamethylene)triamine in chloroform (0.02 M), was added to the PMAO-coated SPIOs to reach a ratio of 10 crosslinker molecules per nm² of nanocrystal surface. Chloroform was added to the solution to reach a final concentration of nanoparticles of 0.5 μM Fe₃O₄. The mixture of PMAO-coated SPIOs with crosslinker was bath sonicated for 30 minutes followed by rotary evaporation for 1 h to ensure complete removal of solvents. A thin film of nanoparticles on the flask wall indicated complete solvent removal. Sodium borate buffer (50 nM, pH 9) was added to cover the thin film of particles in the flask, followed by water bath sonication for 2 h at 60 °C or until the particles detached from the flask wall. Ultrasonication was then performed to obtain a suspension of single nanoparticles. PMAO-coated nanoparticles were purified by magnetic separation using magnetic columns (Myltenyi Biotec, LS Columns) and washed 3x in the magnetic column using the borate buffer to remove free polymer and crosslinker. Particles were then collected in deionized water.

General method for functionalized nanoparticle synthesis:

In a typical synthesis, a volume of Fe@PMAO equivalent to 1.4 mg Fe from the provided stock was centrifuged at 13,000 xg for 30 min to pellet the nanoparticles. The supernatant was aspirated, and the nanoparticles were resuspended by brief sonication with 500 μL of 0.5 mM 2-(N-morpholino)ethanesulfonic acid (MES) buffer adjusted to pH = 5.9 with sodium carbonate containing a dissolved 1 mg of EDC. This solution was shaken on a tube shaker for 30 min at 40°C to activate the carboxylates of the PMAO. For hard zwitterion functionalized nanoparticles, 4 mg of either SB or PSB was added to the tube, and for soft zwitterion functionalized nanoparticles 3 mg of aminoethyl maleimide was added. In all cases, the tube was briefly sonicated and left on the shaker at 40°C overnight. The nanoparticles were placed on a 1 T magnet to remove the nanoparticles from solution and the supernatant was aspirated and 1 mL of water was added. This was repeated three times to wash any unbound ligand from the nanoparticles. For hard zwitterions, nanoparticles were resuspended in 500 μL of PBS and stored at 4°C.For soft zwitterions, nanoparticles were resuspended in 3-(N-morpholino)propanesulfonic acid (MOPS) buffer adjusted to pH = 6.9. Two milligrams of either L-cysteine or DL-homocysteine was added to this solution and sonicated briefly. Nanoparticles were again shaken overnight at 40°C and washed using the same procedures as previously described. Soft zwitterion-coated nanoparticles were resuspended in 500 μL of PBS and stored at 4°C.

Determining nanoparticle concentration by inductively coupled plasma optical emission spectroscopy (ICP-OES):

To determine the concentration of nanoparticles following functionalization, 50 μL of the functionalized nanoparticle stock solution in PBS was added to 3 mL of aqua regia (3:1 HCl:HNO₃) in a glass vial and stirred at 80°C for 4 hr. The vial was cooled to room temperature, diluted with 3 mL of Milli-Q water, and analyzed for Fe concentration against a previously determined standard curve by ICP-OES using an Agilent 7000.

TEM analysis of nanoparticle morphology:

Nanoparticles were diluted with Milli-Q water, drop cast, and dried on the carbon-coated side of a Cu TEM grid with mesh size 300 placed on parafilm. TEM images were acquired using an FEI Tecnai G2 Spirit Twin TE.

FTIR analysis of nanoparticle surface functional groups:

Ten microliters from the nanoparticle stock solutions were drop cast on polyethylene plastic sheets and dried overnight. The dried nanoparticle concentrate was analyzed by FTIR using a Nicolet 6700 FTIR-ATR with background correction performed using polyethylene plastic without any nanoparticles.

Nanoparticle zeta potential and hydrodynamic size evaluation by DLS:

For zeta potential measurements, 10 μL of nanoparticles were added to 990 μL of 1 mM KNO₃ buffer adjusted to pH = 3 with HNO₃, and pH = 7.4 or pH = 11 with KOH. This solution was added to Malvern Zetasizer cuvettes (DTS1070) and zeta potential was acquired using a Malvern Zetasizer Nano-ZS for triplicate batches. Hydrodynamic size was acquired by dispersing 10 μL of nanoparticles in 990 μL of PBS. The solution was added to Malvern Zetasizer cuvettes (DTS1070), and Z-average size was acquired using a Malvern Zetasizer Nano-ZS for triplicate batches.

Magnetic relaxometry of nanoparticles:

The RELAX module (MOMENTUMTM) was used for relaxometry measurements. Nanoparticle solutions from their stocks were added to a 0.5 mL microcentrifuge tube (final Fe content of 50 μg) and placed at the center of the sample holder (0 position) in the MPI scanner. The amount of iron in the sample was carefully chosen to avoid saturation of the detector (< 5 arb. Units). Magnetic relaxation curves were acquired by sweeping the magnetic field from −150 to 150 mT to produce relaxation curves. Full width at half maximum (FWHM) and amplitude corresponding to specificity and sensitivity, respectively, were measured directly from these curves.

Dispersibility of nanoparticles in cell culture media:

A stock solution of nanoparticles was added to supplemented Dulbecco’s modified Eagle’s medium (10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S), supplemented DMEM) to a final concentration of 50 μg/mL Fe, briefly sonicated, and left in an incubator under cell culture conditions overnight (37°C, 5% CO₂, humidified). A picture was taken of the solutions the following day.

Cellular loading with nanoparticles:

MDA-MB-231 cells and RAW 264.7 bone marrow macrophages were grown in supplemented DMEM, while MSCs were grown in α modified Eagle’s medium supplemented with 10% FBS and 1% P/S (supplemented αMEM) under cell culture conditions. These 3 cell lines are adherent. Cells were grown to 80% confluency and passaged three times before being seeded into 24-well plates and grown again to 80% confluency.

Once confluent, each nanoparticle was dispersed in the respective supplemented cell culture media to a final concentration of 50 μg/mL Fe and briefly sonicated. The growth media was replaced with this nanoparticle-containing supplemented media. Cells were incubated for 4 hr under cell culture conditions at 37°C or 4°C (control, block active cell uptake). Following incubation, media was aspirated, and adhered cells were washed three times with warm (37°C) PBS. Cells were removed from the plate by incubation with 0.25% trypsin-EDTA, diluted with supplemented media and centrifuged at 400 xg (5 min, 4°C). Cell pellets were resuspended in 5 mL of warm PBS and centrifuged at 400 xg (5 min, 4°C). This was repeated two additional times to further remove unbound nanoparticles from cells. The final cell pellet was resuspended in 100 μl of warm PBS and an aliquot of this solution was diluted 1:1 with trypan blue to be read by a ThermoFisher Countess II cell counter. Cells were diluted further with warm PBS according to the cell count to a concentration 1.25 mil cells/mL. One hundred microliters of this solution was placed in a CytoSep^™ funnel connected to a coverslip, placed in a Shandon^™ CytoSpin³, and spun at 1000 RPM for 5 min. Funnels were removed from the coverslips, dried for 1 min by air, then fixed in a 3:1 v/v methanol:acetic acid solution for 10 min. The slide was washed with distilled water then incubated in a modified PPB solution for 20 min (20mL solution containing 20% v/v HCl and 20% w/v potassium ferrocyanide). This was washed with distilled water and counterstained with freshly prepared Nuclear Fast Red (20 mg Nuclear Fast Red, 1 g aluminum sulfate, 20 mL distilled water) for 10 min. After washing with distilled water, slides were dehydrated by 5 min incubations in solutions of distilled water with increasing ethanol concentrations (70%, 95%, and 99%), followed by clearing with xylenes (5 min and 10 min in fresh xylenes each time). Slides were air dried and sealed with coverslips using Epredia CytoSeal^™. Slides were imaged using a Nikon NiU Ratiometric Microscope (40x objective).

TEM analysis of nanoparticle subcellular localization:

Cellular uptake was performed as previously described in the methods, except after washing cells in the 24-well plates, they were fixed in 2.5% glutaraldehyde diluted with 0.1 M sodium cacodylate buffer for 2 hr. After scraping from the plate and placing into microcentrifuge tubes, cells were rinsed with 0.1 M sodium cacodylate buffer for 10 minutes, three times, followed by fixation for 2 hr with 1% osmium tetroxide. Cells were washed again with distilled water and placed in 0.25% uranyl acetate at 4°C overnight. Cells were then dehydrated using increasing concentrations of acetone in water (50%, 70% x2, 95% x2, 100% x3, 10 minutes each) and infiltrated with Epon-Araldite resin (1:3 with acetone for 3 hr, 3:1 with acetone overnight, and then 100% Epon-Araldite resin for 3 hr twice). Samples were then embedded in 100% Epon-Araldite resin and placed in a 60°C for 48 hr to cure.

Cured samples were sectioned using a Reichert-Jung Ultracut E ultramicrotome with a 100 nm diamond knife and placed on Cu grids of mesh size 300. Grids were stained with aqueous 2% uranyl acetate for 10 min, rinsed with distilled water for 5 min twice, stained with lead citrate for 4 min, rinsed with distilled water, and air dried. Images of sectioned cells was acquired using a JEOL JEM 1230 TEM.

Evaluating MPI signal of labelled cell pellets:

Cell labeling was performed as previously described in the methods, except the final, washed cell pellet (varying number of cells based on cell line) was resuspended in 50 μL of PBS and placed into PCR tubes. An aliquot of this sample was counted using a ThermoFisher Countess. The PCR tubes containing labelled cells were placed in a tube holder and loaded into the bed of the MPI scanner and imaged individually using the following parameters: FOV = 12 cm × 6 cm, gradient strength = 3.0 T/m, dual-channel acquisition (X and Z), excitation amplitude = 20 mT (X channel) and 26 mT (Z channel), imaging time = 1.5 min. After imaging, the pellet was diluted 1:1 with PBS and imaged again. This was repeated until the cell pellet was undetectable.

Cell tracking of labelled MSCs in mice following intramuscular injection:

Eight-week-old athymic nude mice were purchased from Charles River. All animals were housed in conventional breeding cages, fed standard rodent chow, received water ad libitum, and were housed in a facility with a 12-h light cycle.

Cellular uptake was performed as previously described with MSCs labelled with either Fe@PMAO or Fe@HCys. The cell pellet was resuspended in 300 μl warm, sterile PBS. This labelled cell suspension (500,000 cells) was injected into the hindlimb of three anesthetized mice (isoflurane) mice for Fe@PMAO and Fe@HCys, each. The opposite hindlimb was injected with 300 μl of warm, sterile saline as a control. Following injection, mice were anesthetized by isoflurane and placed in the MPI scanner. MPI images of the hindlimb area were acquired under anesthesia using thesame imaging parameters described above. These scans were repeated on day 3 and day 7 to follow the change in signal of the labelled cells over time.

Statistical Analyses:

Statistical analyses for experiments are reported in the corresponding methods section. Normality was assumed where appropriate for all data sets. Prior to ANOVA, Levene’s test was used to confirm equal variance, and visual quantile-quantile plot analysis was used to confirm homoscedasticity.

Supplementary Material

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.