An enzyme-activatable and cell-permeable MnIII-porphyrin as a highly efficient T1 MRI contrast agent for cell labeling

MnAMP, a cell-trappable pro-contrast agent gets enzymatically activated and accumulated intracellularly to provide a strong MRI signal for cell labeling.


Introduction
Our capability to study and utilize cell functions has greatly increased owing to advances in cellular imaging techniques. In recent years, as the research eld involving cell transplantation, such as adoptive immunotherapy 1 and stem-cell therapy, 2,3 is rapidly growing, there are increasing demands for new translational imaging methods to noninvasively label and track implanted cells in vivo. 4,5 Among conventional clinical imaging modalities, MRI has a unique combination of advantages, including true non-invasiveness, deep tissue penetration, intrinsic anatomic information with delicate tissue contrast, a large and adjustable eld of view (FOV) with sub-millimeter resolution, and an unlimited time window for repeated imaging. Therefore, MRI has become a preferred choice for monitoring spatial and temporal changes in cell localization and distribution in vivo. Since the native relaxation times (T 1 and T 2 ) of different cell types are too similar to detect by MRI, a group of implanted cells can only be visualized aer labeling in vitro prior to transplantation in vivo. 6 Cells have been successfully labeled with traditional CAs acting on 1 H-NMR relaxation enhancement or using non-proton 19 F agents [7][8][9] without background signal. Currently the most widely used CAs for MR cell imaging are T 2 agents based on superparamagnetic iron oxide nanoparticles (SPIOs), owing to high sensitivity down to single cell detection 10 and adjustable cell-permeability, which has been demonstrated in a variety of both animal studies and clinical trials. 11,12 There are however limitations inherent in the contrast mechanism utilized by these agents. The negative contrast enhancement (image darkening) generated by T 2 agents can also result from different sources, including tissues with high iron content, hemorrhages, hemochromatosis or other artefacts associated with magnetic susceptibility such as air spaces and tissue interfaces. 13,14 T 1 CAs can overcome these limitations by providing conspicuous positive contrast enhancement (brightening) on MR images, more specic to distinctly labeled cells. In fact, Gd-based T 1 CAs (GBCAs) dominate the regular clinical applications of contrast enhanced MRI. 15 While most clinical GBCAs belong to extracellular uid (EFC) agents considered to be cell-impermeable, a study led by Aime et al. demonstrated MRI-observable cellular uptake of Gd-HPDO3A, presumably via pinocytosis. High concentrations (5-100 mM) and long labeling times (12-24 h), however, were required for sufficient cell labeling, due to low permeability. 16 A more sophisticated approach is to incorporate multiple Gdcomplexes into nanocarriers, 17 such as liposomes, 18 virus capsids, 19,20 protein cargos 21 or glucan particles 22 to improve relaxivity and cell loading capacity. In addition, attachment of small molecular Gd-chelates or Gd-nanocarriers with cell penetrating peptides [23][24][25][26] or lipophilic moieties 27,28 can enhance the cell uptake. While more stable cyclic Gd-chelates are the preferred choice for cell labeling, long-term intracellular stability of GBCAs is still a concern due to toxicity of free Gd ions. 29 Another limitation of typical GBCAs is that T 1 relaxivity (r 1 ) is not optimal, especially at high eld, a favorable condition for cellular MRI. Quenching of r 1 down to below 1 mmol À1 s À1 was observed when large amounts of Gd are concentrated within the small intracellular space. Atypical T 1 agent platforms such as free MnCl 2 , 30,31 or Mn-nanoparticles 32,33 were explored for cellular MRI. In the +2 oxidation state, Mn II -based T 1 agents including nanoparticles typically show a decrease of T 1 relaxivity with increase of magnetic elds above 0.1 T, 34,35 and are accompanied by relatively strong T 2 effects (darkening) that compromise the T 1 enhancement. 30,36 In addition, Mn IIcomplexes are kinetically labile in aqueous solution. Despite the above mentioned promising results, highly sensitive, stable and biocompatible T 1 CAs with high cell permeability and retention remains an unmet challenge. To expand this repertoire we decided to explore a different platform to develop T 1 CAs for cellular MRI. Herein we report the design, synthesis, characterization and cell labeling studies of a novel cell-permeable and esterase-activatable T 1 agent based on a Mn III -porphyrin (MnP).

Molecular design and principle of strategy
The molecular design started from Mn III tetra(carboxyporphyrin), (MnTCP, 3), a novel Gd-free ECF T 1 agent we developed. 37 Optimized from the leading compound, Mn III tetra(4-sulfonatophenyl)porphyrin (MnTPPS), 38 MnTCP is the smallest water-soluble MnP synthesized to date. 37 It displays a fast in vivo clearance through renal ltration 39 and rapid extravasation ideal for cancer detection by dynamic contrast enhanced (DCE) MRI. 40 Despite a smaller size and thus shorter rotational diffusion time (s R ), MnTCP exhibits an r 1 of 7.9 mM À1 s À1 , comparable to the larger MnTPPS at a high clinical eld of 3 T. This is about twice as sensitive as typical small GBCAs of similar size. To the best of our knowledge, MnTCP has the highest r 1 at clinical magnetic eld strengths (1-3 T) among all known small T 1 CAs with molecular weight below 600 dalton. Unlike Mn II -based T 1 agents, Mn III TCP exhibits much lower negative T 2 effect (r 2 /r 1 ¼ 1.15 at 3 T). The rigidity of the porphyrin macrocycle results in a pre-organized metal binding pocket compatible for Mn III . This leads to a MnP complex with high thermodynamic and kinetic stability, reducing the likelihood of metal leakage. 41,42 By eliminating four phenyl rings from MnTPPS, the design of MnTCP minimized the hydrophobic surface areas, with four water-solubilizing carboxylates directly attached at the meso-positions of the porphyrin to increase polarity. The reduced lipophilicity together with condensed negatively charged groups helps to avoid interference with cellular molecular machinery, minimizing intracellular compartmentalization, 43 and also helps to retain MnTCP by preventing leakage through the intact cell membrane, a property well-documented for polyanionic uorescent cell tracers. 43 These characteristics of MnTCP make it an ideal precursor to design a cell labeling T 1 agent, if intracellular delivery can been realized.
To noninvasively deliver MnTCP into the cells, our strategy is to mask the polarity of MnTCP by converting polar carboxylate groups to lipophilic AM esters, a prodrug approach well-established for loading otherwise impermeable carboxylate-bearing therapeutics or imaging agents into cells. [44][45][46] Unlike most Gdchelates, of which the carboxylate groups are directly involved in Gd-binding, the four peripheral carboxylate groups in MnTCP are free and available for derivatization with little effect on Mn-affinity. As illustrated in Scheme 1, a permeable AM ester derivative of MnTCP (MnAMP) was designed for crossing the cell membrane.
The lipophilic nature and positively charged Mn IIIporphyrin core of MnAMP will facilitate the cell uptake. Furthermore, the hydrophobic MnAMP is expected to aggregate in aqueous solution resulting in a state of low relaxivity, a wellcharacterized phenomenon for a series of known hydrophobic MnPs. 47 Hydrolysis of AM esters catalyzed by intracellular Scheme 1 Proposed mechanism for cell uptake and retention. esterase releases the negatively charged carboxylate that is expected to help disaggregation, thereby inducing increases in relaxivity and facilitating the retention of the agent inside the cell.

Synthesis and characterizations of MnAMP
The stepwise total synthesis of MnAMP is summarized in Scheme 2. The precursor MnTCP, 3 was synthesized through the hydrolysis of 5,10,15,20-tetra(ethoxycarbonyl)porphyrinato manganese(III), 2 (ref. 48) under basic conditions. Since the resulted sodium salt of MnTCP is highly polar it is not soluble in DMF. Protonation of the carboxylate groups with 1.0 M HCl was necessary to impart sufficient solubility in DMF for the subsequent installation of the AM ester groups.
Synthesis of MnTAMP, 4a was accomplished with acetoxymethyl bromide (AMBr) and an organic base, diazabicyclo [5.4.0]undec-7-ene (DBU) in DMF at 55 C for 30 h. During the purication, 4a was shown to be unstable. It decomposes on silica gel during ash chromatography, to form a tris-AM ester derivative (MnTriAMP) 4b with very similar retention time as 4a. Thus, MnTAMP and MnTriAMP were isolated together in 65% yield with a ratio of about 53 : 47 as determined by HPLC-ESI MS (Fig. S5 †). Interestingly, the mono-hydrolyzed product 4b is much more stable than 4a with negligible decomposition observed on silica gel and was therefore isolated in a pure form for characterization. We hypothesized that the low stability of 4a may involve the net positive charge on the molecule due to Mn III , facilitating the nucleophilic attack of negatively charged hydroxide on the ester carbonyl carbon. The stability of MnTriAMP is owing to the balance of charge between the carboxylate anion and the metal center. To test this hypothesis and to further verify the synthesis, the Mn-free analogue of 4a was prepared starting with apo-porphyrin 1 as a precursor following the same synthetic pathway (Scheme S2 †). The following observations are in good agreement that Mn III may activate the single ester hydrolysis: (1) tetra(acetoxymethoxycarbonyl)porphyrin (TAMP), 5, is much more stable than Mn-inserted analogue 4a, since no decomposition was observed during the purication on silica gel; (2) the tetraethyl ester MnP analog, 2, also decomposed slightly during the metal insertion reaction to provide mono-hydrolyzed MnEt 3 P as the only side product, 48 while Mn-free form 1 is much more stable in ester hydrolysis. The structures of the Mn-free products along this control synthetic pathway were conrmed with NMR, since the paramagnetic Mn III makes it difficult to use NMR for routine structural characterization. All the MnPs were characterized by high resolution ESI-MS, UV-visible and Fourier transform infrared (FTIR) spectroscopy (see ESI †). The purity of the nal products was conrmed by HPLC ( Fig. S1-S3 †) and Mn atomic absorption spectroscopy (AAS).
Since the single negative charge on the carboxylate of MnTriAMP was balanced by Mn III to give an overall neutral compound, MnTriAMP was also expected to cross the cell membrane efficiently. In fact, the similar retention time of 4a and 4b on silica gel TLC as well as reverse phase (C18) HPLC suggest they have similar lipophilicity (Fig. S2 †). Upon cellular internalization and esterase hydrolysis, both 4a and 4b would be converted to polar MnTCP, with three net negative charges, thereby trapping the compounds inside the cell. All experiments below were carried out with a mixture of 4a and 4b referred to collectively as MnAMP.

Low relaxivity of MnAMP due to aggregation
Unlike MnTCP, MnAMP is readily soluble in organic solvents, due to its hydrophobic nature, a property desirable for achieving cell permeability. The signicant change, observable on the UV-visible spectrum, makes it possible to monitor the kinetics of aggregation/disaggregation processes optically, as shown in the next section. 47 To conrm the aggregation of MnAMP will lead to a "quenching" state of r 1 , the nuclear magnetic relaxation dispersion (NMRD) prole of MnAMP was acquired with a fast eld-cycling NMR relaxometer from 0.23 mT up to 1 T, and compared with that of MnTCP 37 ( however, the relaxivity of MnTCP (r 1 ¼ 9.95 mM À1 s À1 ) is more than 3-fold higher than that of MnAMP (r 1 ¼ 3.01 mM À1 s À1 ). This large difference in relaxivity suggests that MnAMP has good potential as an esterase-activatable CA.

Esterase-catalyzed hydrolysis of MnAMP
To demonstrate that MnAMP can be hydrolyzed by esterase to lead to disaggregation, partially pre-aggregated MnAMP (60 mM, HEPES buffer) was incubated with porcine liver esterase (3.1.1.1, 3 U mL À1 ), and the reaction was monitored by UVvisible spectroscopy. A gradual increase in absorption and red-shiing of the Soret band (Fig. 2a) could be continuously monitored during the incubation, as opposed to the optical response during the above-mentioned aggregation process. In the control sample without addition of esterase, the absorption continued to slowly decrease over time (Fig. 2b). These results support that esterase hydrolysis leads to disaggregation of MnAMP.
To further conrm that the enzyme-induced disaggregation was due to the stepwise AM ester hydrolysis, the partiallyhydrolyzed sample was analyzed by HPLC-MS. Along with the residual MnTriAMP, the expected intermediates, including regio-isomers of MnBiAMP (R 1 , , were all detected as separated HPLC peaks (Fig. S4 †) that were conrmed by ESI-MS (Table S4 †).
Their order of retention times on a reverse-phase (C18) column is consistent with their relative polarities. Those species were also intermediates for the synthetic step of installing AM esters on MnTCP (Scheme 2), which were detectable by HPLC-MS during the reaction. We noticed, however, that the enzymatic reaction by the isolated esterase does not proceed to the completed hydrolysis. This is not surprising due to the limited scope of substrate diversity and compromised reactivity of isolated enzymes in buffer. Because there are a variety of intracellular esterases 49 that are more active in live cells than as isolated enzymes, intracellular hydrolysis should proceed to MnTCP more efficiently under cell labeling conditions. This has been repeatedly demonstrated with a wide range of AM ester prodrugs 49,50 and uorescent tracers. 46 As a large family of enzymes, esterases are ubiquitous in mammals and are found in all kingdoms with broad substrate specicity. There are also a variety of esterases present in the cytosol. This led to the development of an assortment of commercially available cell viability and membrane integrity probes utilizing uorescent sensors with AM esters masking polar carboxylates that are trapped in the cytosol upon esterase hydrolysis. 51,52

MRI activation of MnAMP by enzymatic hydrolysis
To test that the esterase catalyzed hydrolysis induces MRI activation of MnAMP, the relaxivity (25 C, 1.5 T) was measured in the presence and absence of porcine liver esterase (20 U mL À1 ). Solutions of MnAMP (1 mM) in HEPES buffer (50 mM, pH 7.4) were prepared and allowed to pre-aggregate in the dark for 30 min. The T 1 relaxation times were measured prior to addition of esterase and aer 2, 4, and 6 h incubation with esterase at 37 C (Fig. 3).
A signicant relaxation enhancement from T 1 ¼ 1.06 s to T 1 ¼ 312 ms was observed aer 2 h incubation with esterase. Continuous increase of T 1 relaxivity at a slower rate could be monitored at 4 and 6 h. These results conrmed that MRI activation occurred upon the stepwise hydrolysis of MnAMP. The release of more polar carboxylates broke up the aggregation and increased the water accessibility to the paramagnetic Mn center. Under the current experimental condition, enzymatic activation produced a 3.5-fold increase in T 1 relaxation. Other   MRI enzyme sensing strategies have been developed 53 such as modulation of water coordination number (q), 54 tumbling rate (s R ), 55 chemical exchange saturation transfer 56 and precipitation enhanced staining. 57

Biocompatibility of MnAMP
To ensure that MnAMP is biocompatible for cell labeling purposes, we evaluated the toxicity of MnAMP in mammalian cells. A human glioma cell line, U373, was rst chosen for the safety test. The extracellular uid precursor MnTCP was used as a membrane-impermeable control. Aer incubation with 80 mM MnPs for 2 h, cell proliferation was indirectly examined with an MTT assay 24 and 48 h aer labeling. Similar to the untreated cells, proliferation remained unaffected for cells incubated with MnTCP and MnAMP (Fig. S11 †). In addition, cell viability > 96% was determined by trypan blue exclusion tests for all three samples, showing good biocompatibility for both MnTCP and MnAMP directly aer MnPs incubation, and aer 5 h extra growth in fresh medium. Similarly, both MnAMP and MnTCP did not show toxicity effects on MDA-MB-231 cells, a human breast cancer cell line (Table S7 †).

Cell labeling with MnAMP
To test cell permeability and labeling efficiency, approximately 9 Â 10 6 U373 cells were incubated with 80 mM MnAMP or MnTCP for 2 h in growth medium (DMEM), washed 3 times with Hanks Balanced Salt Solution (HBSS), detached and pelleted for MRI and relaxivity measurements. To further examine cell retention of MnAMP, an additional experiment was conducted where MnAMP labeled cells were incubated for a further 5 h in fresh medium in the absence of MnPs.
As shown in the photograph of the cell pellets (Fig. 4a), the cell uptake of the dark red MnP was clearly visible by eye in cells treated with MnAMP (I) as well as the MnAMP treated sample aer 5 h extra growth in fresh medium (II). By comparison, cells treated with MnTCP, which has a similar color as MnAMP, did not show signicant color staining (III), similar to the untreated cells (IV), suggesting little cell uptake of MnTCP. Subsequent MRI of the cell pellets was conducted on a 3 T MRI scanner using a T 1 -weighted inversion recovery fast spin-echo pulse sequence. As shown in Fig. 4b, the MR image (T 1 map) exhibited signicant positive contrast enhancement for MnAMP treated cells (I 0 ), with enhancement maintained aer 5 h extra growth (II 0 ), in comparison to unlabeled cells (IV 0 ) and MnTCP-treated control (III 0 ). The relaxation times, T 1 and T 2 , of the cell pellets were calculated either from the MR images (at 3 T) or by relaxometry (T 1 at 1 T). As summarized in Table 1, the untreated cells exhibit T 1 and T 2 similar to typical literature values. 58 A signicantly shorter T 1 of 161 AE 4 ms was determined at 3 T for cells treated with MnAMP. The 5 h retention sample resulted in a slight increase of T 1 to 272 AE 12 ms but still maintained the majority of T 1 enhancement, compared to 1134 AE 18 ms for the untreated cells. Cellular T 1 enhancement and retention of MnAMP was also conrmed by relaxivity measurement at 1 T. Notice that the systematically shorter T 1 values for all samples at 1 T compared to those measured at 3 T are consistent with the relaxation dispersion behaviors of MnTCP and unlabeled tissues. Even though T 2 enhancement was detectable for MnAMP treated cells (45% T 2 shorting), the T 1 effect is dominant. This is signicantly different from Mn II -based CAs, which show strong T 2 effects (darkening) that compromise the positive T 1 enhancement. 30 By comparison, MnTCP treated cells exhibited a negligible decrease in T 1 (at 1 T and 3 T) or T 2 (at 3 T), thereby conrming that MnTCP is cell-impermeable and its permeability was dramatically enhanced through addition of the lipophilic AM ester groups.
With a relatively low incubation concentration (80 mM) and short incubation time (2 h), the observed signicant T 1 enhancement of labeled cells demonstrated that MnAMP has high labeling efficiency (Table 1). In comparison, the known small molecule T 1 agents oen require higher concentrations and longer incubation times, up to 24 h, to achieve sufficient T 1 enhancement. 59 The most efficient cell labeling T 1 agent reported to date is a Gd-loaded glucan particle (Gd-GP). Macrophages incubated with 250 mM of Gd-GP for 24 h resulted in an R 1 ¼ 3.6 s À1 (T 1 ¼ 278 ms) at 1 T. 60 In contrast, with lower concentration and shorter incubation time, MnAMP labeled cells reached even higher T 1 enhancement with an R 1 ¼ 10.6 s À1 (95 ms) at 1 T. To the best of our knowledge MnAMP is the most efficient T 1 agent developed for cell labeling to date.
To conrm whether MnAMP is converted to MnTCP intracellularly, the cells were lysed and the cytosolic and membrane fractions were isolated and analyzed by HPLC and Mn-AAS, respectively. As expected all of the MnAMP was completely converted to MnTCP (Fig. S14-S17 †) and was mainly found in the cytosol. In contrast, no MnP was detected in the membrane fraction.

Quantitative cell uptake of MnAMP
For quantitative analysis of cell uptake, all the cell pellets were digested with 60% nitric acid in an ultrasonic bath at 40 C for   (Table 1). Unlike Gd, Mn is an essential micronutrient present in cells. The endogenous intracellular [Mn] was subtracted as background signal from all calculations. An increase in cellular Mn content 2.8 Â 10 À15 mole per cell was determined directly aer labeling, corresponding to 1.7 Â 10 9 Mn atoms per cell. Since the uptake of paramagnetic metal ion per cell is within the range of recently developed T 1 cell labeling agents, 27,60,61 the high cellular contrast enhancement likely results from the combination of high cell uptake together with higher T 1 relaxivity of MnPs at high elds compared to small molecule GBCAs and Gd-nanoparticles. 62 MnAMP also demonstrated good retention aer 5 h in fresh medium with only a slight decrease in T 1 enhancement.
In comparison the increase of Mn content for the cellimpermeable control MnTCP was minimal with only 4.3 Â 10 À17 mole per cell as expected from the lack of contrast enhancement. These results demonstrate that MnAMP is one of the most efficient T 1 cell labeling agents to date. Many previous studies demonstrated successful MRI cell labeling on phagocytes, such as macrophages and dendritic cells. These phagocytes can take up particulate MRI CAs through phagocytosis, even though the labeling agents have limited cell permeability. A variety of approaches have been developed to label non-phagocytic cells 64 such as the attachment of cell penetrating peptides, 23-26 functionalization of nanoparticles to promote receptor mediated endocytosis, 65 or by physical perturbation of the cell membrane. 66 The two cell lines used in the current study are not considered as typical highly phagocytic cell types. The high level of cell uptake of MnAMP is thus likely due to high permeability. Therefore, MnAMP is a safe and efficient T 1 agent and can potentially be applied for labeling a variety of different cell types.

Chemicals and reagents
For synthetic procedures all reagents and solvents were of commercial reagent grade and were used without further purication except where noted. Reagents were purchased from Sigma-Aldrich. Solvents were purchased from Caledon Labs. All reactions were carried out with oven dried glassware, anhydrous solvents and under argon atmosphere unless stated otherwise. Biotechnology grade (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (HEPES) buffer (typically used as 10 mM solution, pH ¼ 7.4, ionic strength ¼ 100 mM, except where noted), Hanks Balanced Salt Solution (HBSS) was purchased from Life Technologies and ammonium acetate was purchased from Fisher Scientic. Phosphate buffer saline was purchased from Sigma Life Science, sterile ltered and endotoxin tested. Ultrapure water was generated by a MilliQ system. Porcine liver esterase (3.1.1.1.) was purchased from Sigma-Aldrich (E3019) as a lyophilized powder, 17 U mg À1 . The MDA-MB-231 cell line was obtained from ATCC (American Tissue Culture Collection Manassas, VA, USA). The human glioma cell line U373 was gied to us by Dr Janusz Rak. Trypsin EDTA was purchased from Gibco (Carlsbad, CA, USA). a Data at 1 T was acquired by relaxometry by means of standard inversion recovery techniques, tted by non-linear regression with SD from the tting. b Data at 3 T was quantied from the MR images based on a pixel-by-pixel relaxation time analysis, the SD represents the variation among the pixels. c Cells labeled with MnAMP were grown in MnP-free medium for extra 5 h. d Data are shown as means AE SD of three independent experiments.

Synthesis of MnAMP
Protonated MnTCP 37 (55.6 mg, 103 mmol) was suspended in DMF (8 mL). Under continuous stirring, DBU (80 mL, 535 mmol, 5 eq.) was added dropwise. AMBr (130 mL, 1.3 mmol, 13 eq.) was added in three separated portions at 10 min, 6 h and 24 h and the progress of the reaction was monitored by TLC. The reaction temperature was maintained at 55 C for 30 h. Distillation of DMF under reduced pressure resulted in a crude dark oil. DCM was added to the crude material to dissolve the product and the mixture was ltered. The DCM layer was neutralized with water and brine twice each. The organic layer was dried over sodium sulfate and ltered prior to concentration on a rotary evaporator. Purication by ash column chromatography (eluting with 5% to 10% MeOH in DCM) on silica gel gave 55.4 mg (65%) of 4a and 4b (53 : 47 ratio by HPLC-MS) as a red-brown solid. Due to decomposition of 4a to 4b on silica gel it was not possible to isolate 4a. The purest fraction that contained 4a was further isolated by preparative TLC and was analyzed by LCMS and gave a ratio of 4a : 4b (83 : 17). Pure 4b was isolated from the mixture by ash chromatography, with the same elution solvents, 2 more times. Characterization of 4a:

UV-visible spectroscopy
UV-visible spectra were recorded on an Agilent 8453 UV-visible spectroscopy system. The extinction coefficient (3) of 3 was previously reported. 37 Absorption spectra of 3, 4b, MnAMP, and 5 were measured in HEPES buffer at 25 C.

Relaxivity measurements
The NMRD proles were acquired with a fast eld-cycling NMR relaxometer ( Zero lling interpolation was applied to the spin-echo images. The signal intensity was analyzed to give the relaxation times T 1 and T 2 of the cell pellets. Matlab was used to generate non-linear ts for each pixel to curves dened by the following equations: For T 2 : S ¼ M xy e Àt/T 2 + n off where M z is the steady state longitudinal magnetization at thermal equilibrium, M i is the magnitude of the inverted magnetization acquired during the readout, M xy is the transverse magnetization, and n off is any signal offset present in the images.

Cell lysis, HPLC and UV-Vis analysis aer labeling
Aer cell labeling and testing of viability, the cells were re-suspended in 500 mL PBS with 0.01% saponin. 67 Aer 30 min at 25 C the cells were centrifuged at 1000g for 5 min. The supernatant was collected as the cytosolic fraction. To the pellet was added 500 mL PBS followed by 50 strokes on a Dounce homogenizer and centrifugation at 15 000g. The supernatant was collected as the nuclear fraction with the remaining pellet collected as the membrane fraction. The cytosolic fractions were analysed by HPLC and UV-visible spectroscopy (Fig. S14-S18 †). The membrane fractions were digested and quantied by GFAAS.

Graphite furnace atomic absorption spectroscopy
Mn-quantication was determined with a ThermoFisher GFS 35 graphite furnace absorption spectrometer equipped with an electrothermal atomizer, an autosampler and a deuterium-lamp background correction system. A Perkin-Elmer Intensitron manganese hollow-cathode lamp was used according to the manufacturer's recommendations. Table S1 † shows the electrothermal program for the determination of manganese. A 10% w/v magnesium nitrate solution (Aldrich, Germany), was used as the chemical modier. A 1000 mg mL À1 manganese (2% HNO 3 ) Titrisol (Aldrich) was used for preparation of standard working solutions. Nitric acid Suprapure (Aldrich) was used for stabilization of samples and working standards. Standard solutions for calibration purposes were prepared by proper dilution with 2% w/v HNO 3 solution. A rinsing step was included prior to withdrawal of each aliquot. Spectroscopic analyses of samples were performed with 15 mL of standard/ sample and 5 mL modier injected sequentially into the graphite furnace atomizer. Measurements were performed in triplicate. The cellular elemental concentration was determined by dividing the total content by the number of cells. The endogenous Mn content determined in the unlabeled cells was subtracted from the MnP labeled cells.

Conclusions
In summary, we have designed, synthesized and characterized MnAMP, a cell-permeable and esterase-activatable T 1 CA specically developed for MRI cell labeling. The novel paramagnetic porphyrin, MnAMP is constructed by rational structural modication of MnTCP, a membrane-impermeable CA with high stability and high T 1 relaxivity. Reaction of MnTCP with AMBr under basic condition successfully converted the polar carboxylate groups to lipophilic AM esters to give MnAMP as a mixture of MnTAMP (4a) and MnTriAMP (4b). As a prodrug of MnTCP, MnAMP is cell-permeable and exhibits low extracellular relaxivity due to aggregation in the aqueous medium. We have demonstrated that AM ester groups in MnTAMP can be catalytically hydrolyzed by a commercially available liver esterase to release the polar carboxylates, inducing disaggregation and thereby, signicantly increasing T 1 relaxivity. We have further shown that MnAMP can effectively cross the cell membrane and is converted to MnTCP by intracellular esterase, leading to intracellular accumulation. In comparison the cell uptake of negatively charged MnTCP was negligible, conrming that installation of the AM ester groups was necessary to enhance cell permeability. Highly efficient MRI labeling of two types of non-phagocytic human cells, including a glioma cell line and a breast cancer cell line, was achieved with a relatively low concentration (80 mM) of MnAMP and a short incubation time (2 h). Unprecedentedly strong T 1 enhancement of labeled U373 cells was determined at clinical elds of 1 T (T 1 ¼ 95 ms) and 3 T (T 1 ¼ 161 ms), corresponding to about 10-fold and 7-fold T 1 shortening at both eld strengths, respectively. In contrast, the negative T 2 effect was much less signicant, suggesting MnAMP primarily acts as a positive agent. Cell viability and proliferation remained unaffected by MnAMP. Therefore, MnAMP is biocompatible and to the best of our knowledge, is the most efficient T 1 cell labeling CA available to date. The current work has demonstrated the potential of MnAMP to be widely applied to label different cell types for in vivo monitoring at the commonly used high clinical eld of 3 T. Future studies will be focused on in vivo applications involving therapeutic cells, including stem cells and dendritic cells for monitoring and optimization of adoptive immunotherapy and stem cell transplantation.