Tracking cells in vivo with imaging technologies is highly important for biology, pathology, and medicine (Hong et al., 2011; Kircher et al., 2011; Bulte and Daldrup-Link, 2018). Such imaging technologies may provide real-time, noninvasive, quantitative, and multidimensional information about the cells for a better understanding of the biological and pathological processes on a cellular level, thus promoting accurate diagnosis and effective therapy. For example, tracking the circulating cancer cells in vivo with imaging technology may shed light on cancer metastasis mechanisms and therapeutic strategies (Xia et al., 2012; Nedosekin et al., 2013; Lee et al., 2018; Dutta et al., 2019; Deán-Ben et al., 2020), while monitoring macrophages may help visualize the immune response to inflammation (Stoll et al., 2012; Bouvain et al., 2019).

In recent years, considerable attention has been given to developing novel imaging technologies for cell tracking. Among the imaging technologies, fluorescent protein-based fluorescence imaging (FLI) (Chudakov et al., 2003; Sun et al., 2009; Ghosh et al., 2019), fluorescence dye-based FLI (Bulte et al., 2004; Maška et al., 2013; Tian et al., 2020), paramagnetic nanoparticle-based ¹H MRI (Cromer Berman et al., 2011; Ahrens and Zhong, 2013; Li et al., 2013), and fluorinated nanoemulsion-based ¹⁹F MRI (Ferreira et al., 2008; Bulte, 2009; Wang et al., 2013) are among the most used ones. Since each imaging technology has its strengths and weaknesses, tracking cells with multimodal imaging is highly preferred, which takes advantage of each imaging technology and provides accurate, detailed, and multidimensional information. During cell tracking, the cells are usually labeled in vitro and tracked in vivo, and some special cells, such as macrophages, may also be labeled and tracked in vivo. FLI is advantageous for in vitro cellular study due to its convenience, high sensitivity, and resolution. Meanwhile, ¹⁹F MRI perfectly overcomes the tissue-depth limit of FLI during the in vivo animal study and provides noninvasive, quantitative, and background-free “hot-spot” cell images without ionizing radiation. Thus, ¹⁹F MRI-FLI dual imaging systems have shown significant potential for clinical application.

Although some ¹⁹F MRI-FLI dual imaging systems have recently been developed for cell tracking (Ahrens et al., 2005; Gaudet et al., 2017; Peng et al., 2018), many issues remain unaddressed. On the one hand, the low sensitivity of ¹⁹F MRI requires high ¹⁹F MRI agent-loading per cell and a large number of ¹⁹F-labeled cells to achieve clear cell images (Ahrens et al., 2005; Gaudet et al., 2017; Peng et al., 2018). Furthermore, the non-symmetric allocation of fluorine atoms in most perfluorocarbon (PFC)-based imaging agents leads to severe signal splitting, low ¹⁹F MRI sensitivity, and chemical shift imaging artifacts. On the other hand, the peculiar physicochemical properties of PFCs, such as high hydrophobicity, low polarity, and low solubility in hydrocarbons, usually result in complex formulation and severe internal organ retention (Meyer et al., 1992; Kimura et al., 2004; Janjic et al., 2008). Notably, the extremely low solubility and interaction of fluorescent dyes in PFCs may cause aggregation-induced quenching (ACQ) of fluorescence, difficulties in encapsulation, and early release of fluorescent dyes (Würthner et al., 2011; Battistelli et al., 2016). Moreover, the high density of PFCs and their nanoemulsions may cause low cell uptake, cell damage, and misleading results (Patel et al., 2013). To address these issues, we recently used partially fluorinated agents, hydrofluorocarbons (HFCs), as alternatives for PFCs (Chen et al., 2022; Wu et al., 2022). These commercially available HFCs had a unified ¹⁹F NMR peak, considerable solubility toward fluorescent dyes and functional agents, relatively low density and high ¹⁹F MRI sensitivity, and efficient cell uptake. However, significant cytotoxicity was found in the HFC nanoemulsions, which promoted the synthesis of highly sensitive and biocompatible HFC agents.

Herein, we have developed novel partially fluorinated nanoemulsions for efficient ¹⁹F MRI cell tracking through synthesizing, screening, and formulating a series of HFCs. Because of their high ¹⁹F MRI sensitivity, biocompatibility, and stability (Wu et al., 2021), perfluoro-tert-butyl benzyl ethers 1–7 were designed as HFCs (Figure 1). Three strategies have been used to improve the ¹⁹F MRI sensitivity. First, 1–6 symmetrical PFTB groups were introduced into the benzyl cores to generate an intense unified singlet ¹⁹F NMR peak from 9 to 54 equivalent ¹⁹F atoms in HFCs 1–7, that is, the utility of every ¹⁹F atom for sensitive ¹⁹F MRI. Second, the π–π interactions among the benzene cores of HFCs 1–7 may slow their movements and thus shorten the longitudinal relaxation time (T₁), which would shorten the data collection time during ¹⁹F MRI and thus improve the signal-to-noise ratio per unit time for sensitive ¹⁹F MRI. Third, perfluoro-tert-butoxylated diketone chelators 8 and 9 may capture paramagnetic ions into the HFCs and further shorten T₁ through the paramagnetic relaxation enhancement (PRE) effect for highly sensitive ¹⁹F MRI. Due to their similar hydrofluorocarbon structures, chelators 8 and 9 may be highly soluble in HFCs 1–7 and provide a unified ¹⁹F NMR peak from the mixture. In these ways, a unified ¹⁹F NMR peak with an ultrashort T₁ would be generated for highly sensitive ¹⁹F MRI cell tracking. The near-infrared fluorescence of the resulting fluorinated nanoemulsion may be realized by aza-BODIPY 11. Compared to highly fluorinated PFCs, HFCs 1–7 with a benzene core and lower fluorine content may have much higher solubility toward aza-BODIPY 11 and thus improve the FLI capability. Moreover, the co-solubility and improved interactions among HFCs 1–7, chelators 8 and 9, and aza-BODIPY 11 may stabilize the nanoemulsion for high biocompatibility. Perfluoro-15-crown-5 10, a widely used PFC in ¹⁹F MRI, was employed as a control to illustrate the impact of π–π interactions and enhanced solubility.

With these ideas in mind, the synthesis of HFCs 1–7 and fluorinated chelators 8 and 9 was then carried out (Scheme 1). First, from a series of commercially available benzyl bromides 12–18, HFCs 1–6 were synthesized in just one step by nucleophilic substitution of the bromides with potassium perfluoro-tert-butoxide in N, N-dimethylformaldehyde (DMF). Under the optimized conditions, HFCs 1–6 were prepared in good yields on multi-gram scales. Notably, pure HFC 1 was conveniently obtained by collecting the lower phase of the water-quenched reaction mixture, and HFCs 2–6 were obtained by filtration from the mixture, followed by washing with water. Because of the high steric hindrance, many attempts failed to synthesize HFC 7 with 6 perfluoro-tert-butoxyl groups, which delivered complex mixtures. Second, the synthesis of fluorinated chelator 8 commenced with the construction of fluorinated ester 22 and ketone 25. The commercially available benzoic acid derivative 19 was first transformed into the corresponding methyl ester 20, which was then radically di-brominated into bromide 21. Through the potassium perfluoro-tert-butoxide substitution reaction mentioned earlier, fluorinated ester 22 was prepared on a gram scale with a 36% yield over three steps. Similarly, commercially available ketone 23 was di-brominated and then substituted with potassium perfluoro-tert-butoxide to give fluorinated ketone 25 with a 37% yield over two steps. However, the Claisen condensation between the fluorinated ester 22 and ketone 25 led to a complex mixture, probably due to the high steric hindrance and peculiar fluorous properties in the reactants. To address these issues, we replaced the fluorinated ketone 25 with commercially available 1-(4-methoxyphenyl)ethan-1-one 26 and successfully prepared partially fluorinated chelator 9 under anhydrous conditions.

The structures of HFCs 1–6 and chelator 9 were confirmed by their ¹H/¹³C/¹⁹F NMR and mass spectra (see Supplementary Material). The structures of known compounds 1–5 were further confirmed by comparing their ¹H NMR spectra with those of Zhao et al. (2012). As expected, all the fluorinated compounds gave a sharp singlet ¹⁹F NMR peak from their multiple symmetrical fluorines, respectively (Figure 2A). HFCs 2–6 are clear solids at room temperature. Notably, solid fluorinated agents usually have an ultra-short transverse relaxation time (T₂) and a quenched ¹⁹F signal. Thus, it is necessary to disperse solid HFCs 2–6 in surfactants or liquid agents to extend the T₂ for sensitive ¹⁹F MRI. Interestingly, solid HFCs 2–6 showed high solubility in liquid HFC 1 due to their similar chemical structures, delivering fluorinated solutions with a pseudo singlet ¹⁹F NMR peak and tunable fluorine content (F%). HFCs 5 and 6, with the highest fluorine contents, showed heavy fluorous properties, for example, low solubility in most hydrocarbon solvents but high solubility in fluorous solvents like perfluoro(methylcyclohexane) (PFMCH). Thus, the ¹⁹F relaxation times of HFCs 1–6 were measured in PFMCH (Figure 2B). A molecular weight (MW)-dependent pattern of ¹⁹F relaxation time was observed from HFCs 1–6, that is, the larger the MW, the shorter the relaxation times. Compared to many perfluorocarbons, HFCs 1–6 had relatively short relaxation times and small T₁/T₂ ratios. During ¹⁹F MRI, short T₁ and long T₂, that is, a small T₁/T₂ ratio, are beneficial for rapid data collection and thus sensitive ¹⁹F MRI. It was noteworthy that the π–π interactions among the benzene cores were successfully observed in the concentration-dependent ¹⁹F NMR spectra of HFCs 1–6 in PFMCH (Figure 2C). When increasing the concentrations, the ¹⁹F NMR peaks of HFCs 1–6 were shifted to the low field as far as 0.184 ppm in HFC 3. In contrast, neglectable apparent chemical shift changes, about 0.005 ppm, were observed in the corresponding solutions of perfluoro-20-crown-5 10, which has no benzene structure. Furthermore, HFC 6 with four bulky PFTB groups gave a minimal ¹⁹F chemical shift change of 0.045 ppm among the HFCs, in which the high steric hindrance severely reduced the π–π interactions. In pure HFC 1 and its HFC 2–6 solutions, the π–π interactions were supposed to reduce the T₁ for sensitive ¹⁹F MRI. Therefore, HFCs 1–6 were conveniently synthesized on relatively large scales as valuable and sensitive ¹⁹F MRI agents with a high F%, a sharp singlet ¹⁹F NMR peak, significant π–π interactions among the benzene cores, and relatively short relaxation times.

Because of its liquid form, very cheap precursor, and convenient multigram-scale synthesis, HFC 1 was selected to formulate partially fluorinated nanoemulsions for ¹⁹F MRI cell tracking. Due to their similar chemical structures, fluorinated chelator 9 was quite soluble in HFC 1, dramatically simplifying the formulation of paramagnetic nanoemulsions. S75 and egg yolk lecithin together with F68 were, respectively, identified as the optimal surfactants for the formulation, providing nanoemulsions of HFC 1, chelator 9, and aza-BODIPY 11 using a thin-film dispersion method (Table 1, Entries 1 and 2). Interestingly, adding a small amount of soybean oil delivered highly monodisperse nanoemulsions E1 and E3 (Table 1, Entries 3 and 5). After adding FeCl₃ to the nanoemulsion solution, the apparent color change indicated the chelation of Fe³⁺ by fluorinated chelator 9, providing a paramagnetic nanoemulsion with negligible particle size and PDI changes. After the removal of unchelated Fe³⁺ in the solutions by dialysis, paramagnetic nanoemulsions E2 and E4 were prepared with high monodispersity (Table 1, Entries 4 and 6).

The high monodispersity of nanoemulsions E1–E4 was confirmed by dynamic light scattering (DLS) with low polydispersity indexes (PDI, Figure 3A). Furthermore, DLS detected only slight changes in the particle size and PDI of E1 solutions during 14 days (Figure 3B), indicating its high stability. In contrast, a considerable increase in the PDI of E3 between day 8 and day 14 showed less efficacy of lecithin and F68 in formulating HFC 1. NIR dye aza-BODIPY 11 was successfully encapsulated into the nanoemulsions with high efficiency (E1: 93%, E2: 96%, E3: 93%, and E4: 97%) and a loading content of about 0.45%. Compared to aza-BODIPY 11, the nanoemulsions gave slightly red-shifted maximum UV-Vis absorption peaks around 653 nm (Figure 3C) and a maximum FL emission peak around 682 nm (Figure 3D). Moreover, the encapsulation did not compromise the fluorescence intensity of aza-BODIPY (Supplementary Figure S3). It was found that the Fe³⁺ in nanoemulsions E2 and E4 considerably lowered the maximum UV-Vis absorption and FL emission intensities, probably due to the photoinduced electron transfer between aza-BODIPY and Fe³⁺ (El-Khouly et al., 2004; Doose et al., 2009; Fukuzumi et al., 2012). Each fluorinated nanoemulsion gave a singlet ¹⁹F NMR peak of around −72.4 ppm (Figure 3E). The nanoemulsions E1 and E3 had relatively long ¹⁹F T₁, while the Fe³⁺ in nanoemulsions E2 and E4 significantly shortened the ¹⁹F T₁ by 3.6 and 5.1 folds, respectively, through the PRE effect (Figure 3E). The short T₁ facilitated a fast ¹⁹F T₁-dependent ¹⁹F MRI scan for high-quality images. The nanoemulsions E1–E4 were imaged by ¹⁹F MRI at a ¹⁹F concentration as low as 5 mM using a ¹⁹F density-dependent MRI method with a data collection time of 384 s (Figure 3F). As expected, the ¹⁹F MRI sensitivity was significantly improved by the encapsulated Fe³⁺ using a ¹⁹F T₁-dependent MRI method, in which paramagnetic nanoemulsions were imaged at a low concentration of 2.5 mM with a data collection time of 389 s (Figure 3G). Notably, the logarithm of ¹⁹F signal intensity (SI) was proportional to the ¹⁹F concentration in both density-dependent ¹⁹F MRI and T₁-dependent ¹⁹F MRI (Figures 3F, G), which facilitates the quantification of ¹⁹F concentration with ¹⁹F SI. Thus, fluorinated nanoemulsions E1–E4 with high monodispersity, stability, and unified and paramagnetic-enhanced ¹⁹F signals were formulated as sensitive and quantitative ¹⁹F MRI-FLI dual imaging agents.

Next, nanoemulsions E1–E4 were investigated in cells. First, the biocompatibility of nanoemulsions E1–E4 was investigated in human lung cancer A549 cells, human breast cancer MCF-7 cells, and human triple-negative breast cancer MDA-MB-231 cells using the CCK-8 cytotoxicity assay. High cell viabilities were observed in the nanoemulsion-treated cells with a ¹⁹F dose as high as 45 mM after 12 h of incubation (Figures 4A–C), far beyond the ¹⁹F MRI-detectable concentrations mentioned earlier. Second, due to their relatively high stability, a quantitative ¹⁹F NMR study was carried out on the nanoemulsion E1- and E2-treated A549 cells. After incubating 1 × 10⁷ A549 cells with nanoemulsions E1 and E2 at a ¹⁹F concentration of 18 mM for 12 h, the cell lysates were quantitatively analyzed with ¹⁹F NMR, which gave a strong ¹⁹F NMR peak of around −71.3 ppm, respectively (Figure 4D). After calibrating the ¹⁹F NMR SI with an internal standard of sodium triflate, each cell was found to take up 10¹¹ fluorine atoms (Figure 4E). Interestingly, A549 cell lines showed significantly higher uptake of paramagnetic nanoemulsion E2 than diamagnetic nanoemulsion E1, while the reason for this phenomenon is still unclear. Third, the cell uptakes of nanoemulsions E1 and E2 were studied with confocal microscopy in A549 cells (Figure 4F). After 12 h of incubation, the red fluorescence of aza-BODIPY was visible in the cytoplasm around the nucleus dyed with blue-fluorescent DAPI, indicating the successful uptake of nanoemulsions E1 and E2. Notably, the Fe³⁺ partially quenched the FL of aza-BODIPY in nanoemulsion E2, resulting in a weaker FL intensity of E2-treated cells than that of E1-treated cells. In addition, no notable hemolysis was observed for nanoemulsions E1 and E2 in the blood (Supplementary Figure S2), and no death of BALB/c mice was observed after intravenous injection of nanoemulsion (100 μl, ¹⁹F concentration of 450 mM) in an acute toxicity test, indicating the good biocompatibility of nanoemulsions E1 and E2. Finally, ¹⁹F MRI was carried out on the E1- and E2-treated A549 cells. About 1 × 10⁷ A549 cells were clearly imaged by both ¹⁹F density-dependent and T₁-dependent ¹⁹F MRI with short data collection times of 384 s and 389 s, respectively (Figure 4G). Moreover, due to the PRE effect of Fe³⁺, the SI of T₁-dependent ¹⁹F MRI was 2.2-fold higher than that of ¹⁹F density-dependent ¹⁹F MRI in the mice injected with paramagnetic nanoemulsion E2-treated A549 cells (Figure 4H). Therefore, the partially fluorinated nanoemulsions E1 and E2 had high biocompatibility, quantifiable ¹⁹F NMR, efficient cell uptake, and high ¹⁹F MRI sensitivity, which were suitable for quantitative ¹⁹F MRI-FLI dual imaging cell tracking.

Finally, in vivo ¹⁹F MRI-FLI dual-modal cell tracking with partially fluorinated nanoemulsions E1 and E2 was carried out in BALB/c nude mice. On the one hand, the tracking of lung cancer A549 cells was investigated by subcutaneous injection of 1 × 10⁷ E1- and E2-labeled A549 cells to the right and left flanks of mice, respectively (Figure 5A). The injected cells were first tracked with in vivo FL, in which paramagnetic nanoemulsion E2-treated A549 cells showed weaker FL intensity than that of E1-treated cells due to the FL-quenching Fe³⁺ in nanoemulsion E2 (Figure 5B). Then, the A549 cells were tracked with ¹⁹F MRI using ¹⁹F density-dependent and T₁-dependent MRI sequences (Figure 5C). Notably, the cells injected into the left flank gave a 1.9-fold higher ¹⁹F MRI SI than that of the right flank using a ¹⁹F density-dependent sequence, while the ratio of SI was further improved to 3.3-fold using a T₁-dependent sequence (Figure 5D), showing the high efficacy of Fe³⁺ in improving ¹⁹F MRI sensitivity. On the other hand, macrophage cells were labeled in vivo and tracked in a carrageenan-induced inflammation model. After tail vein injection of 100 μl nanoemulsions E1 and E2 at a ¹⁹F concentration of 72 mM into the mice with right paw inflammation, the time-dependent FLI and ¹⁹F MRI were recorded, respectively. The time-dependent FL images showed that the nanoemulsion-labeled macrophages accumulated mainly in the liver, lung, and kidney at 2 h post-injection and selectively accumulated in the right paw at 4 h post-injection, while neglectable FL was observed in the left paw (Figure 5E). The time-dependent FL intensity in the right paw region indicated that E2-labeled macrophages were eliminated much faster than E1-labeled macrophages in this inflammation model (Figure 5F). The FL images of internal organs collected at 4 h post-injection illustrated the nanoemulsion mainly accumulated in the liver, while the accumulations in the heart, spleen, lung, and kidneys were not detected at this time (Figure 5G). Unfortunately, many attempts to track the nanoemulsion-labeled macrophages with ¹⁹F MRI could not provide clear images of the inflamed paw, probably due to the low dose of E1 and E2.

In summary, we have developed partially fluorinated paramagnetic nanoemulsions from readily available chemicals for ¹⁹F MRI-FLI dual-modal cell tracking. To address the challenge of ¹⁹F MRI sensitivity in cell tracking, we have used many valuable strategies, including the efficient synthesis of fluorinated molecules with multiple symmetrical fluorine atoms and a strong ¹⁹F signal, unifying the ¹⁹F signals from all components in the nanoemulsions and shortening T₁ through the intermolecular interactions and the PRE effect of paramagnetic iron (III). Indeed, the partially fluorinated paramagnetic nanoemulsions exhibit high ¹⁹F MRI sensitivity, which was detected at a low fluorine concentration of 2.5 mM with a short data collection time of 389 s. Further improvement in ¹⁹F MRI sensitivity, for example, using HFCs 5 and 6 as ¹⁹F signal molecules, is still needed to achieve more sensitive in vivo ¹⁹F MRI cell tracking as indicated in the inflammation mice model. Meanwhile, efforts have been made to improve partially fluorinated nanoemulsions’ physicochemical and biological properties. Through screening surfactants and adding soybean oil, highly monodispersed and stable nanoemulsions were obtained. The significant intra-molecular interactions of perfluoro-tert-butyl benzyl ethers have been shown to facilitate the successful encapsulation of functional molecules, including fluorinated chelators and aza-BODIPY, into stable nanoemulsions, which enables their dual-modal imaging and high biocompatibility toward a series of cells. Because of their peculiar physicochemical properties, the application of perfluorocarbon nanoemulsions in ¹⁹F MRI cell tracking has been hampered by issues of sensitivity, biocompatibility, and multimodal compatibility. The strategy developed here, lowering the fluorine content while maintaining sensitivity and optimizing physicochemical properties, may address some of these issues and push ¹⁹F MRI cell tracking into further clinical application.