Monitoring polymers in vivo plays a pivotal role in the development of materials for biomedical applications, such as drug delivery and tissue regeneration1,2,3,4. MRI is a powerful 1H-based anatomical imaging technique, which is free of ionizing radiation and not limited by tissue penetration depth. Heteronuclear “hot spot” MRI opened new horizons, introducing the imaging of other nuclei, currently focusing on 19F5,6,7. While conventional 1H MRI agents modulate tissue contrast, heteronuclear agents are highly specific and act as new color, uniquely providing quantitative information along with the spatial localization and monitoring over time. Therefore, “hot spot” 19F MRI became effective in imaging of cardiovascular diseases, cancer, monitoring cellular therapies, nanomedicines, and biomaterials, such as artificial tissues8,9,10,11,12,13,14. However, common 19F MR agents are formulations of per- and polyfluoroalkyl substances (PFAS or perfluorocarbon PFC), which are under discussion recently15. Here, we introduce an alternative solution based on biocompatible and biodegradable polymers, i.e., polyphosphoesters, opening the route to more sustainable MRI-traceable polymer materials.

Currently, polymeric materials, such as nanocarriers and hydrogels can be labeled for heteronuclear MRI by encapsulation of PFAS. However, PFAS are hydrophobic, lipophobic (amphiphobic), and chemically inert, which makes their stabilization in physiological media and functionalization challenging6,16. Moreover, PFAS often show prolonged organ deposition, which is a disadvantage for example in repeated imaging sessions and can lead to misdetection, e.g., when the labeled material itself was degraded17. Lastly, due to the high chemical stability which makes PFAS basically undegradable, they are classified as environmental pollutants15. Current approaches focus on improving the properties of fluorinated compounds for 19F MRI, e. g. developing of partly fluorinated polymers which are stable in physiological milieu, or inorganic nanoparticles13,18,19,20,21,22,23,24,25. Instead, we decided to develop polymers that can be imaged via 31P nuclei in their backbone instead of common 19F, to take advantage of the biocompatible properties of polyphosphoesters26,27.

31P is the natural, 100% abundant, NMR-active isotope of phosphorus. However, the development of background-free 31P MRI agents and 31P MRI-traceable materials has been hampered by several factors, including the intrinsic background from natural phosphates, the low MR sensitivity of 31P of 7% compared to 1H, and other unfavorable MR characteristics, including too short or too long relaxation times and J coupling. For example, most of biomolecules yield weak 31P NMR signals due to these reasons. Nevertheless, 31P MR spectroscopic approaches were used for pH-sensing and detection of metabolites, such as endogenous phosphocreatine and ATP, as well as non-biodegradable synthetic polymers28,29,30, but are limited due to the required long acquisition times.

Here, we introduce biodegradable nanocarriers that are made of bioinspired polyphosphoesters with a high 31P content and variable chemical shifts to overcome the low sensitivity of 31P MRI. PPEs are versatile polymers with highly diverse chemistry, which make them interesting candidates for several biomedical applications, particularly as a stealth coating26,27. Uniquely, we synthesized biodegradable, 31P MRI agents based on amphiphilic gradient copolymer micelles which can encapsulate hydrophobic therapeutic payload as their second inherent function, acting as both background-free imaging agents and carrier material for drugs. This work will lead to the development of MRI-traceable biomedical polymers for a broad range of applications, from MR imaging agents to label-free 31P MRI-traceable drug delivery and materials for tissue engineering.

Results and discussion

Proof of concept that polyphosphonate nanocarriers can be imaged with 31P MRI

We selected the following strategy to solve the fundamental challenges in MRI of 31P:

(1) to ensure the background-free imaging, we tuned the molecular environment of 31P. Specifically, we used polyphosphonates (PPn, general structure Fig. 1A), which contain a P-C-bond, which is not present in endogenous phosphates resulting in a chemical shift that is separated by more than 10 ppm from other biomolecules to allow selective and specific imaging. The shift can be further adjusted by varying the chemical nature of the side chain in Fig. 1A, generating different resonance frequencies and thereby give access to additional MRI ‘colors’.

Fig. 1: Proof-of-concept that PPn colloids can be used in 31P MRI.
figure 1

(A) General chemical structure of PPn. The composition of the main and side chains can be adjusted to tune the properties. (B) Chemical structures of block copolymers and schematic of colloidal nanostructures PPnCORE that contains phosphorus-31 in the core (left) PPnSHELL (right) with phosphorus-31 in hydrophilic PPn shell, and the respective NMR spectra. The relaxation times were suitable for imaging (NMR spectroscopy in H2O/D2O 9:1 v:v, decoupled). (C) 31P MR images of aqueous dispersions of PPnSHELL and PPnCORE, (polymer concentration 4 wt.-% polymer correspond to 0.1 M 31P for PPnCORE and 0.05 M 31P in PPnSHELL), TAcq 17 min, FOV 20×20 mm2, matrix 64×64, 9.4 T.

(2) We prepared PPE-copolymers that can self-assemble into colloidal nanostructures to enhance the local 31P content of within the agent, as the MR signal linearly correlates with the local concentration of 31P.

(3) We developed polymers with a low glass transition temperature (Tg < body temperature) to adjust the MR relaxation times in a favorable manner: Usually, a shorter transverse relaxation time T1 and a longer longitudinal relaxation time T2 lead to higher signal-to-noise ratios (SNR)31. However, solid polymers typically display a very short T2, but polymers above the Tg are in a “liquid-like” and not in a solid state, which should enhance both T2 and SNR.

Our first goal was to obtain proof-of-concept that 31P MRI is possible with colloidal PPn nanostructures despite the low sensitivity of the 31P nucleus. Our first attempt were nanoparticles prepared by an emulsion formulation utilizing hydrophobic polyphosphonates, which, however, displayed a too short T2 of 0.017 s (Supplementary Fig. 1). To adjust the relaxation times, we expanded to amphiphilic block-copolymer microstructures. We focused on polymers with a 31P-free MR-invisible second block (Fig. 1B) to obtain a single resonance signal for conventional spin-echo MRI sequences. For this strategy, we synthesized an amphiphilic block copolymer of hydrophobic PPn with phenyl side groups (Ph-PPn) and hydrophilic polyethylene glycol that self-assemble into micelles in aqueous solution with PPn in the core (PPnCORE, Fig. 1B, Supplementary Table 1). Additionally, we prepared poly(styrene-b-ethyl phosphonate)32 with the hydrophilic PPn with ethyl side chains (Et-PPn) in the shell of the micelles PPnSHELL (Fig. 1B, Supplementary Table 1). The hydrophobic PPnCORE already displayed a longer T2 of 0.049 s in micelles. PPnSHELL showed an even higher T2 due to the higher mobility of the hydrophilic Ethyl-PPn. As a result of improved T2, both nanostructures were successfully imaged with 31P MRI using a spin-echo sequence within acquisition time of 17 min (TAcq; Fig. 1C). These results provided the first proof that polyphosphonates can be imaged with MRI.

Gradient microstructures improve MRI characteristics

Based on these initial results, we hypothesized that a gradient microstructure could further beneficially affect the MR properties, as it should enable an even higher mobility of the macromolecules resulting in longer T2 compared to block copolymer with a single phase boundary. Moreover, the 31P content and consequently the sensitivity should increase by copolymerization of two 31P-containing monomers. Therefore, we developed a single-step synthesis approach to amphiphilic 31P MRI agents by anionic ring-opening copolymerization of a mixture of phenyl- and ethyl-functional monomers (Fig. 2A, Supplementary Table 1).

Fig. 2: Synthesis and 31P MRI of gradient copolymer micelles PPnGRAD.
figure 2

A Synthesis of gradient copolymer by anionic ring-opening copolymerization. B Copolymerization kinetics: Monomer concentration in solution as a function of the total conversion of monomers Et-Pn and Ph-Pn. C Monte-Carlo simulated microstructure for 10 representative polymer chains, calculated from the determined reactivity ratios. D Average composition of monomer fraction F plotted against total conversion. E Schematic of PPnGRAD micelles with NMR spectrum (158 MHz, D2O). F 31P MRI of PPnGRAD micelles (top row). The signals of both Ph- and Et-units were added to a sum image, resulting in the highest SNR. The images of block copolymer micelles PPnCORE and PPnSHELL are shown at the same intensity scale for comparison. Polymer content 4 wt% (31P 0.1 M in PPnCORE, 2x0.1 M in each block of PPnGRAD; and 0.05 M 31P In PPnSHELL; similar polymer concentration was chosen. FOV 20 × 20 mm2, matrix 64 × 64, 9.4 T. G PPnGRAD micelles could be imaged and localized with 31P MRI after injection in physalis berry. 1H MRI (matrix 512 × 512), 31P MRI and an overlay image (31P MRI in hot iron, matrix 64 × 64) are shown. Arrow indicates the injection site, mCSSI, TAcq 17 min, FOV 30 × 30 mm2, 9.4 T. Color scale in arbitrary units (a.u.).

The resulting copolymers with Ph-PPn and Et-PPn repeat units displayed a gradient microstructure, as confirmed by both kinetic measurements (Fig. 2B, Supplementary section 1.3,) and Monte Carlo simulations (Fig. 2C, D, further characterization Supplementary Figs. 3, 4). In aqueous dispersion, they self-assembled to micelles PPnGRAD of 9 nm radius with a narrow size distribution (PDI < 0.1) and a low critical micelle concentration of 14 mg L−1 similar to conventional nonionic surfactants (Fig. 2E, see Supplementary Figs. 59 for further characterization including stability and atomic force microscopy images).

Indeed, the gradient structure led to improved MR properties (Fig. 2E, F, cf. Supplementary Table 3 for relaxation times of different batches). The T2 of hydrophilic Et-PPn-block was twice as long compared with the hydrophilic shell in the above-mentioned block copolymer micelles PPnSHELL (0.9 s versus 0.5 s; see Supplementary Section 2.1 for fitting and discussion). Similarly, the T2 of Ph-PPn in the core increased to 0.13 s. As the gradient microstructure creates a broader interfacial region in the micelles, the mobility of macromolecules and the T2 of hydrophobic core increase compared to the other nanostructures. Furthermore, both 31P-containing blocks lead to a higher 31P content in a micelle of 20 wt-%, which along with longer T2 should lead to higher SNR when a similar polymer dose is injected.

To take full advantage of both the increased T2 and 31P concentration in the micelles, we implemented and optimized multi-chemical selective imaging (mCSSI; spin-echo type of sequence) for 31P MRI. Usually, conventional MRI of compounds with several NMR signals results in chemical shift artefacts, making a reliable signal assignment and localization impossible. The mCSSI sequence allows for simultaneous, artefact-free imaging of multiple signals and subsequent summation of the individual images that results in increased SNR9. Indeed, mCSSI of the different PPn-colloids at the same polymer dose in aqueous dispersion revealed a three-fold higher SNR for the PPnGRAD micelles compared to the both block-copolymers (Fig. 2F, TAcq 17 min). The signal intensity decreased linearly, when the PPnGRAD dispersion was diluted (Supplementary Fig. 11). A comparison to 19F MRI using an emulsion of perfluoro-15- crown 5 ether using 3.8 mol L−1 19F, vs 1.5 mol L−1 31P showed approx. 5 times higher SNR for 19F (Supplementary Fig. 11). However, these samples are not directly comparable due to their different physicochemical properties. The lowest detectable concentration of polymer was comparable to the one reported for fluorinated polymers previously13, indicating sufficient signal intensity for in vivo studies by 31P MRI.

To demonstrate that gradient micelles can be reliably imaged in tissues, we injected PPnGRAD micelles via the stem base into a physalis berry (Fig. 2G left, arrow). 31P mCSSI led to artefact-free images without endogenous background and merge with the ‘anatomical’ 1H MRI demonstrated the distribution of PPnGRAD micelles from their injection site into the core of the physalis (Fig. 2G).

In vivo 31P MRI and biodegradation of polyphosphonate micelles

After encouraging 31P MRI in vitro, we confirmed that 31P mCSSI of gradient micelles PPnGRAD can be used in vivo. We used Manduca sexta caterpillars (Fig. 3A); M. sexta has a hemolymph volume of 1–2 mL, comparable to the blood volume of mice, which makes it a suitable, alternative animal imaging model instead of mammalians according to 3 R principles of animal-friendly testing33,34.

Fig. 3: 1H/31P MRI in Manduca sexta caterpillars in vivo and degradation of the agents (FOV 20 × 20 mm2, 1H MRI matrix 256 × 256, 31P MRI matrix 64 × 64, 9.4 T).
figure 3

A Photograph of the M. sexta caterpillar, orange square denotes the region shown on MR images. B 1H MRI showing the anatomy of caterpillar. C 31P MRI overlaid on 1H MR image after the injection of the agents into the dorsal vessel (heart). PPnGRAD agents circulate in hemolymph. D After 24 h the agents can still be found in the circulation, as expected for micelles with stealth surface. E, F Degradation of PPnGRAD in vivo. E After injection of the agents into the gut, overlay of 31P MR on 1H MR image confirms their localization. F 31P NMR spectrum of feces (D2O, 158 MHz) collected after 24 h shows the characteristic signals of degradation products confirming that PPnGRAD agents are biodegradable. Color scale in a.u.

First, we demonstrated that PPnGRAD micelles can be imaged during the circulation in hemolymph (i.e., the blood of insects). Therefore, the micelles were injected into the dorsal vessel of the animals. In spite of the lower sensitivity of the 31P nucleus compared to 19F, we selected an overall colloid dose typical for intravenous injection of 19F MRI agents (15 mg in 100 µL). All animals tolerated these injections well and continued their development as usual. The micelles distributed in the hemolymph homogenously after injection, as shown by overlaying the morphological 1H MRI and 31P mCSSI data (Fig. 3C, Supplementary Fig. 12, TAcq 17 min). After 24 h, the 31P signal was still located in the hemolymph and decreased slightly (Fig. 3C). Thus, only small amounts of the polymers were excreted or degraded under these conditions; the biodistribution should be studied in the future in different animal models. We assume that the previously reported stealth effect of Et-PPn prolonged the circulation time, allowing one to detect the micelles in hemolymph after 24 h similar to PEGylated nanocarriers35.

Finally, we confirmed the biodegradation of the PPnGRAD micelles in the M. sexta model. The agents were directly injected into the anterior part of the gut; 31P and 1H MRI confirmed the localization of the agents within the target organ (Fig. 3E). After 24 h, we collected the feces of these animals and used 31P NMR spectroscopy to analyze the degradation products (Fig. 3F). Indeed, the 31P NMR spectrum of the aqueous extracts proved the presence of characteristic sharp resonances of low molar mass degradation products at 16 and 31 ppm next to the broader signals of the injected agent at 20 and 38 ppm, respectively, with the chemical shifts as expected for the two degradation products (Fig. 3F, see Supplementary Fig. 13 for further animals). In M. sexta, the gut passage takes ca. 1.5 h, which is obviously insufficient for complete biodegradation. PPEs degrade mainly by a backbiting mechanism36; as Et-PPn is the outer block, it degrades first, followed by Ph-PPn, thereby resulting in a stronger signal of degradation product of Et-PPn (Fig. 3F). Hence, these results confirm that PPnGRAD agents are degradable in vivo.

Polyphosphonate micelles act as drug nanocarriers

Besides in vivo 31P MRI, we proved that PPnGRAD micelles can be used to formulate hydrophobic therapeutics for drug delivery, which is currently the main biomedical application of amphiphilic copolymer micelles37. We selected Proteolysis Targeting Chimera (PROTAC, ARV-825) as a model payload38. PROTACs are emerging drugs that induce the degradation of pathogenic proteins by the cellular degradation machinery, holding the potential to treat currently undruggable targets, such as resistant tumors38. However, many PROTACS are large and poorly water-soluble molecules that need to be formulated to be stable in body fluids.

We used a simple nanoprecipitation technique to formulate PROTAC-loaded PPnGRAD micelles. The resulting aqueous dispersion appeared translucent, with micelles of 10 nm radii similar to the non-loaded micelles, as determined by DLS. The aqueous dispersion was stable over at least two months according to DLS (Fig. 4A right vial and 4B). Conversely, when PROTAC was formulated without the amphiphilic polymer, the hydrophobic drug precipitated and formed polydisperse aggregates immediately after the addition of the organic solution into water (Fig. 4A left vial, and 4B). The PROTAC-PPnGRAD micelles reduced the viability of cancer cells (HeLa Fig. 4C) and induced early apoptosis (Fig. 4D) similar to free PROTAC, suggesting that PPnGRAD can be to used formulate hydrophobic drugs. Overall, these results suggest that PPnGRAD can be used to develop theranostic agents in the future.

Fig. 4: Stabilization of hydrophobic payload by PPnGRAD 31P MRI agents applied to PROTAC ARV-825.
figure 4

A Photograph of PROTAC-loaded PPnGRAD micelles (right vial) and a control PROTAC suspension prepared without the addition of PPnGRAD (left vial). PROTAC-loaded micelles form a clear dispersion. Conversely, in the control PROTAC (yellow) precipitated immediately after the preparation. B DLS of PROTAC-loaded PPnGRAD micelles. The micelles had a low polydispersity and were stable for at least two months. In contrast, PROTAC suspension displayed a broad size distribution. The size measurement of PROTAC-only suspension can be affected by sedimentation. C Viability of cancer cells (HeLa) was reduced upon the administration of PROTAC-loaded micelles to a comparable level as with free PROTAC. Unloaded PPnGRAD micelles were used as a control. The viability values are normalized to untreated controls. n = 3 (data presented as mean ± SD). D Flow cytometry shows that cancer cells treated with PROTAC-loaded micelles (1 µM ARV-825) start entering an early apoptosis phase.

To conclude, we developed biocompatible, biodegradable micelles for 31P MRI based on polyphosphoesters, overcoming the low sensitivity of the 31P MRI. Therefore, we have synthesized and compared different colloidal nanostructures. Based on these results we have developed amphiphilic gradient copolymers using polyphosphonates, which self-assemble into well-defined micelles with advantageous 31P MRI characteristics. The PPE micelles were injected into M. sexta and successfully imaged in vivo using 31P MRI. The in vivo biodegradation of the PPE-micelles was underlined from collected feces by 31P NMR. These amphiphilic micelles can encapsulate and deliver hydrophobic payload as their second inherent function in parallel to their function as 31P MRI agents, as we have exemplarily shown for PROTAC. This unique polyphosphoester platform acts simultaneously as the carrier material for therapeutics and as the imaging agent and we believe pave the way for future MRI-traceable polymers.


Size exclusion chromatography (SEC) measurements were performed in DMF (containing 1 g·L−1 of LiBr) at 60 °C and a flow rate of 1 mL min−1 with a PSS SECcurity as an integrated instrument, including three PSS GRAM column (100/1000/1000 g mol−1) and a refractive index (RI) detector. Calibration was carried out using polystyrene standards supplied by Polymer Standards Service. The SEC data were plotted with OriginPro 9 software from OriginLab Corporation.

NMR spectroscopy was measured at Brucker Avance III 400 MHz spectrometer equipped with a PA BBO 400S1 BBF-H-D-05 Z SP probe at 298 K (driven by TopSpin 8). As deuterated solvents CDCl3, CD2Cl2, or D2O were used. The proton spectra were calibrated against the solvent signal (CDCl3: δ H = 7.26 ppm, CD2Cl2: δ H = 5.32 ppm, D2O: δ H = 4.79 ppm). Longitudinal relaxation times T1 were measured using an inverse recovery sequence. For the measurements of transverse relaxation times T2 the Carr-Purcell-Meiboom-Gill (CPMG) sequence was used. At least 10 data points were acquired, which were then used for data fitting. An interscan delay d1 was set to 5 × T1. In case of samples in water, deuterium oxide (10 vol%) was added to the samples for locking. Data analysis was performed using Mestrenova14 from Mestrelab and OriginPro2019b.

Dynamic light scattering (DLS) was done at a Zetasizer Lab from Malvern UK at a scattering angle of 90°, and 295 K. The samples were diluted with ultrapure water or physiological saline so that the attenuator was at step 10-11 (set automatically by the device). Data analysis was done with ZSxplorer software from Malvern Panalytical.

Fluorescence spectroscopy to determine the critical micelle concentration (CMC) using pyrene assay39 was measured at spectrometer FL6500 spectrometer from Perkin Elmer, equipped with a Pulse Xenon lamp. A concentration series of polymer in ultrapure water between c = 1.1 mg mL−1 and 7.7 × 10-5 mg mL−1 was prepared and mixed with pyrene stock solution in methanol (1.2 × 10−4 mol L−1) so that the final concentration of pyrene was 6 × 10−7 mol L−1; typically 5 µL of pyrene stock solution were added to 995 µL of the polymer solution. The emission spectra were recorded approximately 15 min after preparation of the solutions using an excitation wavelength of 333 nm, and emission range of 350–550 nm, scan speed of 50 nm min−1. To calculate the CMC, the ratio of intensities I(372 nm) I3 (383 nm) was plotted versus the logarithm of concentration ln(c). The CMC was obtained as an intersection of linear fits of the low c region, where I1/I3\(\approx\)const, and the region where I1/I3 decreases.

For atomic force microscopy (AFM), the micelles were deposited from a solution (0.2 mg mL-1) on freshly prepared silicon wafers by spin-coating (2000 rpm, 60 s). Silicon wafers were cleaned at two bath steps: ultrasonication in acetone and Piranha solution treatment for 20 min. AFM images were obtained in the air and at room temperature using a MultiMode 8 AFM instrument with a NanoScope V controller (Bruker). The AFM was operated in the PeakForce Quantitative Nanomechanical Mapping mode (PF-QNM) to record force-distance curves and to further processed them in the NanoScope Analysis software (version 1.9). The ScanAsyst setting was set to “on” in order to apply optimized scanning parameters, particularly the feedback loop and the applied load, for imaging soft micelles. The force-distance curves were collected following a sine-wave sample-tip trajectory with a frequency of 2 kHz and utilizing a peak-force amplitude value of 30 nm. Soft AFM cantilevers were chosen with a nominal spring constant of 0.4 N/m and a tip with a nominal radius of 2 nm (Bruker, ScanAsyst-air). The AFM optical sensitivity (deflection sensitivity) was calculated based on the thermal tune method based on the nominal spring constant40.

Differential scanning calorimetry (DSC) measurements were performed using a Trios DSC 25 series thermal analysis system, running with the Trios DSC Software 5. The temperature range was from −80 °C to 50 °C under nitrogen and heating rate was 10 °C min−1. All glass transition temperatures (Tg) were obtained from the second heating ramp of the experiment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.