Polymerizable Gd(iii) building blocks for the synthesis of high relaxivity macromolecular MRI contrast agents

A new synthetic strategy for the preparation of macromolecular MRI contrast agents (CAs) is reported. Four gadolinium(iii) complexes bearing either one or two polymerizable methacrylamide groups were synthesized, serving as monomers or crosslinkers for the preparation of water-soluble, polymeric CAs using Reversible Addition–Fragmentation Chain Transfer (RAFT) polymerization. Using this approach, macromolecular CAs were synthesized with different architectures, including linear, hyperbranched polymers and gels. The relaxivities of the polymeric CAs were determined by NMR relaxometry, revealing an up to 5-fold increase in relaxivity (60 MHz, 310 K) for the linear polymers compared with the clinically used CA, Gd-DOTA. Moreover, hyperbranched polymers obtained from Gd(iii) crosslinkers, displayed even higher relaxivities up to 22.8 mM−1 s−1, approximately 8 times higher than that of Gd-DOTA (60 MHz, 310 K). A detailed NMRD study revealed that the enhanced relaxivities of the hyperbranched polymers were obtained by limiting the local motion of the crosslinked Gd(iii) chelate. The versatility of RAFT polymerization of Gd(iii) monomers and crosslinkers opens the doors to more advanced polymeric CAs capable of multimodal, bioresponsive or targeting properties.


Introduction
Magnetic Resonance Imaging (MRI) provides 2-and 3-dimensional anatomical information on tissues and organs in a noninvasive manner.MRI displays submillimeter spatial resolution, unlimited penetration depth, and excellent so tissue contrast imaging, enhancing the diagnostic potential for neurological, cardiovascular and oncological imaging. 1 However, MRI suffers from intrinsic low sensitivity, and image contrast can be enhanced by using contrast agents (CAs) to increase the relaxation rate of water protons.Most commercial CAs are based on discrete, low molecular weight gadolinium(III) complexes, such as Gd-DOTA and Gd-DTPA. 2,3The ability of discrete Gd(III) complexes to enhance image contrast is measured by their relaxivity (r 1 ), which is determined by the number of water molecules coordinated to the metal (q), the water exchange lifetime (s M ) and the rotational correlation time (s R ).The majority of commercial CAs display relaxivities around 4-5 mM À1 s À1 (20 MHz, 298 K), far from the theoretical maximum value (r 1 ¼ 100 mM À1 s À1 for complexes where q ¼ 1). 4 An exciting prospect in MRI CA design is the development of macromolecular systems that possess signicantly higher relaxivities.Numerous strategies have been pursued, [5][6][7][8] including the conjugation of one or several Gd(III) complexes to polymers, 9-12 dendrimers, [13][14][15][16][17][18] micelles [19][20][21][22][23] or nanoparticles, [24][25][26][27][28][29][30][31] or via the non-covalent association with biomolecules (e.g.serum albumin protein) 32,33 or nano-assembled capsules. 34,35ome macromolecular systems have been shown to possess signicantly higher relaxivities compared with commercial CAs (up to 40 mM À1 s À1 ), attributed predominantly to the slower tumbling of the macromolecules, and the incorporation of several Gd(III) complexes within a single system. 4,6espite these advances, the challenge still remains to develop high molecular weight CAs wherein the global motion of the macromolecule is effectively coupled to the motion of the paramagnetic centre. 34,36One way to approach this challenge involves positioning a single Gd(III) chelate at the barycentre of the macromolecule (e.g.dendritic systems). 17,37The rotational correlation time of these high molecular weight Gd(III) chelates is thus dened by the motion of the macromolecule.Polymeric CAs have been developed which incorporate multiple Ln(III) chelates, linked via a single exible arm to the polymeric scaffold.1][12]38 A further challenge involves the development of new methods to control the macromolecular size, structure and shape, as this could lead to well-dened second spheres of hydration, whilst allowing fast-inner sphere water exchange.The majority of polymeric CAs are prepared by classical conjugation of appropriate ligands to a polymer bearing reactive pendant groups (e.g.maleimide group, ester activated monomers).Typically, a protected ligand is covalently attached to a synthesized polymer and the resulting conjugated ligand is deprotected, followed by complexation with Gd(III). 9,10,39In related work, Sherry and co-workers have presented a strategy for polymeric PARACEST agents in which non-metal containing ligands are directly incorporated by the free radical polymerisation of acrylamide functionalised ligands, followed by complexation with Eu(III). 38However, this approach has limitations, because it is difficult to quantify both the extent of ligand conjugation and the degree of lanthanide(III) complexation.Furthermore, the removal of residual lanthanide(III) ions can be challenging, but is crucial if such macromolecules are to be considered for biomedical or clinical applications.
Herein we present a new synthetic approach to macromolecular MRI CAs, involving the single-step incorporation of kinetically stable, monomeric Gd(III) complexes within well-dened macromolecular CAs.We have synthesized four DOTA-like Gd(III) complexes (Gd$L 1-4 , Fig. 1), bearing one or two pendant methacrylamide arms, which serve as monomers and crosslinkers respectively, for the synthesis of linear and hyperbranched polymers using reversible addition fragmentation chain transfer (RAFT) polymerization.Gd$L 1 contains a single methacrylamide arm, capable of forming linear polymers, whereas complexes Gd$L 2-4 contain two methacrylamide arms, serving as crosslinkers to create hyperbranched polymers, in which the motion of the Gd(III) centre is effectively coupled to the slowly tumbling macromolecule.Complex Gd$L 2 possesses two trans-related polymerizable arms, whereas for Gd$L 3 and Gd$L 4 the polymerizable arms are in cis-geometry, with Gd$L 3 possessing shorter arms.The impact of these structural and geometric modications on the relaxivity and tumbling motion of the resulting polymers was evaluated.Each complex possesses a DOTA-like core, to confer maximal thermodynamic and kinetic stability. 3,40Moreover, Gd$L 1-4 possess a single coordinated water molecule (q ¼ 1) and are negatively charged overall, thereby allowing relatively fast water exchange to overcome the limitations of some previously reported macromolecular CAs. 7,8

Results and discussion
Synthesis of monomeric and crosslinker complexes Gd$L 1-4 The syntheses of the monomeric and crosslinker complexes Gd$L 1-4 were optimized to give sufficient quantities of each complex (approximately 500 mg) to allow optimization of the polymerization reactions.Representative syntheses of Gd$L 2 and Gd$L 4 are presented in Scheme 1 and full details of the synthesis of Gd$L 1-4 can be found in the ESI (Schemes S1-S4 †).Gd$L 1-4 were synthesized from tert-butyl protected derivatives of 1,4,7,10-tetraazacyclododecane (cyclen), including DO3A(O t Bu) 3 for Gd$L 1 , trans-DO2A(O t Bu) 2 (2) for Gd$L 2 , and cis-DO2A(O t Bu) 2 (3) for Gd$L 3-4 .][43] For complexes Gd$L 1-3 , mono or bis-alkylation of the macrocyclic free amines with a-bromoester 1, prepared from diazotization and bromination of Cbz-protected L-ornithine (Scheme S2 †), led to formation of the fully protected macrocyclic ligands (e.g. 3 for Gd$L 2 , Scheme 1).The yields of the alkylations were signicantly improved by the addition of potassium iodide to the reaction mixture (K 2 CO 3 /acetonitrile), allowing iodide/bromide exchange. 44It is possible that partial racemisation of the alkylating agent 1 or racemisation during the alkylation reaction occurred, 45 leading to the formation of the protected ligands as a mixture of stereoisomers.The methacrylamide arms were introduced by Cbz deprotection of the ornithine sidechains to give the bis-amine (e.g.macrocycle 4), followed by coupling with N-hydroxysuccinimide methacrylate ester 5. Next, the tert-butyl esters were deprotected using triuoroacetic acid, followed by the addition of GdCl 3 in water at pH 7, to afford the water soluble Gd(III) complexes Gd$L 1-3 aer purication by preparative reverse-phase HPLC (Fig. S1-S3 †).The Gd(III) complexes of a given isomer of ligand L 1-3 will have further elements of chirality arising from the sign and torsion angles of the cyclen NCCN chelate rings, and the NCCO chelates dening the helicity of the pendant arms. 3,36As such, the Gd(III) complexes will exist as a mixture of stereoisomers in solution, which may interconvert by either cyclen ring inversion or arm rotation.The separation of stereoisomers was not attempted in this work.
The synthesis of Gd$L 4 involved initial bis-alkylation of cis-DO 2 A(O t Bu) 2 (8) with a-bromoester 7, prepared from L-glutamic acid (Scheme S2 †) to give protected ligand 9. Again, it is possible that partial racemisation of a-bromoester 7 occurred, or racemisation during the alkylation reaction, resulting in a mixture of stereoisomers of protected ligand 9. Subsequent deprotection of the tert-butyl esters of 9 using triuoroacetic acid, followed by the addition of slight excess of GdCl 3 in water at pH 7, gave the precursor Gd(III) complex 10.Finally, the methacrylamide groups were introduced via coupling the terminal carboxylic acids of 10 to N-(3-aminopropyl)-methacrylamide, using the coupling reagent TNTU (2-(5-norborene-2,3-dicarboximido)-1,1,3,3-tetramethyl-uronium tetrauoroborate), to give the water soluble complex Gd$L 4 aer purication by reverse-phase HPLC.Analysis of the puried complexes Gd$L 1-4 by analytical reverse-phase HPLC revealed a single peak in each case, and high-resolution mass spectral data conrmed formation of the desired complexes (Fig. 2 and S1-S4 †).A major signal corresponding to the negatively charged molecular ion, [M] À , was observed in each case, and the isotopic distribution was in excellent agreement with the theoretical data.

Polymerization of monomeric and crosslinker complexes Gd$L 1-4
RAFT polymerization was utilized to generate copolymers incorporating the monomeric and crosslinker complexes Gd$L 1-4 .RAFT enables access to reproducible polymers with low dispersity and control over the chain length, molecular weight and polymer architecture. 46,47Additionally, RAFT can be used in a wide range of conditions, including different temperature, solvents, co-solvents, various additives, and with different monomers (e.g.acrylates, methacrylates, acrylamides, styrenes, vinyl esters and vinyl amides). 46,48,49Importantly, it has been shown that RAFT is suitable for the polymerization of charged monomers, 50 hence we postulated that direct polymerization of the Gd(III) complexes would be possible.Before attempting polymerization with complexes Gd$L 1-4 , we veried that RAFT polymerization of N-acryloylmorpholine (NAM) was possible in the presence of the negatively charged Gd(III) complex, Gd-DOTA.NAM was chosen as the monomer because P(NAM) displays several desirable properties for MRI applications, including good bio-compatibility, very low toxicity, remarkable stealth properties, and prolonged blood residence time. 51,52Pleasingly, RAFT polymerization of NAM in the presence of Gd$DOTA was well controlled, with the homopolymer displaying low dispersity and close to target molecular weight (Table 1, entry 1).
Next, the synthesis of linear P(NAM-r-Gd$L 1 ) copolymers was investigated using different molar proportions of Gd$L 1 , ranging from 1 to nearly 17 mol%.The polymerizations were conducted in a mixture of DMSO/water (80 : 20) at 80 C (Scheme 2), and the polymers were puried by dialysis through a semi-permeable membrane against distilled water (15 MU cm À1 ).Successful synthesis of the target linear P(NAM-r-Gd$L 1 ) copolymers was conrmed by SEC analysis (Fig. 3 and S7 †): polymers with number average molecular weights (M n ) between 7600 and 14 200 g mol À1 were formed with low dispersity (Đ) values, ranging from 1.08 to 1.25 (Table 1, entries 2-8).Due to the charged nature of the Gd(III) complex and the difference between the PNAM and the standard used (PS), the polymers displayed lower molecular weights than the theoretical values.Standard 1 H NMR analysis to conrm M n for the polymers was not possible due to severe line broadening imposed by the paramagnetic Gd(III) ion; however, polymerization of NAM in the presence of free Gd-DOTA resulted in polymers with the expected M n by 1 H NMR spectroscopy.
Analysis of the bulk magnetic susceptibility (BMS) shi of each sample and ICP-MS analysis of the copolymers aer dialysis enabled an approximation of the number of Gd(III) ions per polymer chain (Tables 1, S2 and S3 †).As expected, the number of Gd$L 1 units incorporated into the polymer increased with increasing molar ratio of Gd$L 1 /NAM monomers used for the polymerization.In the highest case, eight Gd$L 1 units were incorporated per polymer chain (Table 1, entry 8).
Having demonstrated that linear copolymers can be prepared in a controlled manner using the monomers Gd$L 1 and NAM, we turned our attention to the crosslinker complexes Gd$L 2-4 .In order to prevent gelation from occurring and based on some of our previous work and that of Flynn et al., 50,53 the crosslinked polymers were obtained by xing the concentration ratio of the crosslinker (Gd$L 2-4 ) relative to the chain transfer agent (CTA), such that [Gd$L 2-4 ]/[CTA] ¼ 0.9.Initially, a range of water soluble hyperbranched polymers were synthesized by RAFT polymerization of NAM and the crosslinker Gd$L 2 at different initial concentrations (1-2 M) in DMSO, using poly(ethylene glycol) methyl ether 2-(dodecylthiocarbono-thioylthio)-2-methyl-propionate as the CTA (Table 2, entries 1-4).For each reaction, SEC analysis of the molecular weight distribution indicated the formation of hyperbranched polymers, evident from the high molecular weight shoulder (Fig. 3, centre).The broad molecular weight distribution was reected in the high dispersities, Đ, ranging between 4.2 and 19.9 (Table 2), consistent with crosslinking of the growing polymer chains during RAFT polymerization.Further evidence for crosslinking of Gd$L 2 was given by analysis of the BMS shi and ICP-MS  S2) .
i Not determined due to the high content of Gd$L 1 .
Scheme 2 Synthesis of linear (DP 100 and m between 1 to 17) and crosslinked (DP 100 and m ¼ 0.9) polymeric MRI CAs using NAM and Gd$L 1-4.
analysis (Tables 2 and S3 †), which conrmed an increasing number of crosslinked polymer chains at higher initial monomer concentration.For example, when [NAM] 0 was 1.00 M, the average number of crosslinker Gd$L 2 units was estimated to be 7.0 per hyperbranched polymer chain, whereas this increased to approximately 34 units per chain when [NAM] 0 was increased to 2.00 M. A higher initial monomer concentration led to higher molecular weight and dispersity, with values up to M n ¼ 744 000 g mol À1 and Đ ¼ 19.9 when [NAM] 0 ¼ 2.00 M.However, polymerization conducted at higher initial monomer concentration, such that [Gd$L 2 ]/[CTA] > 1, consistently led to the formation of gels (Table 2, entry 5).This is in accordance with the work of Perrier and co-workers, who showed that EGDMA/ CTA ratios of over 1 led to gelation in a RAFT system. 54he optimal conditions found for the synthesis of the hyperbranched P(NAM-r-Gd$L 2 ) polymers were [NAM] 0 ¼ 2.00 M, [Gd$L 2 ]/[CTA] ¼ 0.9 (entry 4).These parameters were applied to the synthesis of hyperbranched polymers using crosslinkers Gd$L 3 and Gd$L 4 , bearing cis-related polymerizable arms (entries 6 and 7).Hyperbranched polymers P(NAM-r-Gd$L 3 ) and P(NAM-r-Gd$L 4 ) were successfully formed.Notably, they displayed lower molecular weights and dispersities compared with those obtained using Gd$L 2 under similar conditions ([NAM] 0 ¼ 2.00 M, [Gd]/[CTA] ¼ 0.9), suggesting that these polymerizations could be achieved at higher initial monomer concentration (>2.00 M).
The average hydrodynamic diameter D h and dispersity of representative examples of linear and hyperbranched polymeric CAs were estimated by diffraction light scattering (DLS) Table 2 Conditions, monomer conversions and SEC data of hyperbranched copolymers P(NAM-r-Gd$L 2-4 ), prepared by RAFT polymerization of NAM and Gd$L 2-4
[NAM] 0 , mol L À1   S3).measurements.From the number weighted particle size distribution, the linear polymer P(NAM-2%-Gd$L 1 ) has a D h of 4.3 AE 0.9 nm.Crosslinked systems with Gd$L 3 or Gd$L 4 display D h of 13.2 AE 4.6 nm and 15.1 AE 4.3 nm, respectively (Fig. 4).This indicates that the polymers exist primarily as unimers in solution.The sizes of the hyperbranched polymers are comparable to previously reported hydrophilic and charged hyperbranched polymers (D h between 10 to 20 nm for MW between 100 to 500 kDa). 50The Gd(III) monomers are charged and therefore hydrophilic, and as expected, do not direct the assembly of these polymers into higher order structures.

H and 17 O NMR relaxometric studies
The millimolar water proton longitudinal relaxation rates ( ) of a Gd(III)-chelate, both in monomeric or polymeric forms, depends on the magnetic eld strength, temperature and on several important structural and dynamic molecular parameters that describe the magnetic coupling between the water protons and the paramagnetic ion.As shown in Table 3, the r 1 values for the discrete complexes Gd$L 1-4 were found to be in the range 4.5-6.6 mM À1 s À1 at 60 MHz (310 K, pH 7.4), each higher than that measured for Gd-DOTA under the same experimental conditions (2.9 mM À1 s À1 ). 55This increase in relaxivity is consistent with the slightly higher molecular weights of Gd$L 1-4 relative to Gd-DOTA, and hence the increase in the rotational correlation time, s R . 56pon copolymerization of Gd$L 1 with NAM by RAFT, the resulting linear polymers P(NAM-r-Gd$L 1 ) possessed signicantly higher relaxivities (Table 4) in the range 12.6-13.5mM À1 s À1 at 60 MHz, and 14.3-15.4mM À1 s À1 at 20 MHz (310 K, pH 7.4), up to ca. 5 times higher than Gd-DOTA.These relaxivity values are similar to those obtained previously by Davis, Boyer and coworkers, 10,57 for both discrete core crosslinked star polymers and hyperbranched polymers, each containing Gd(III) chelates attached via a single pendant arm.
The hyperbranched polymers P(NAM-r-Gd$L 2-4 ), obtained using crosslinkers Gd$L 2-4 ,possessed even higher relaxivities in the range 18.6-22.8mM À1 s À1 at 60 MHz, and 23.0-33.5 mM À1 s À1 at 20 MHz (310 K, pH 7.4) (Table 4).The enhancements in relaxivity of the polymeric CAs relative to the reference agent Gd-DOTA are shown in Fig. 6.Notably, we observe a substantial 9 to 10-fold increase in relaxivity for the crosslinked polymers based on Gd$L 2 and Gd$L 3 , relative to Gd-DOTA at 20 MHz, and a 2-fold increase relative to the linear polymers.Such high gains in relaxivity can be ascribed to the role of Gd$L 2-3 crosslinkers, which reduce the rate of tumbling of the Gd(III) chelate in the resulting hyperbranched polymers.This limited rotational exibility leads to much higher relaxivity. 4It is also clear from Fig. 6 that the gains in relaxivity for the linear polymers are essentially the same at 20 and 60 MHz, whereas for the crosslinked polymers based on Gd$L 2 and Gd$L 3 , the relaxivity gains are greater when measured at 20 MHz.In contrast, for the crosslinked polymer based on Gd$L 4 the relaxivity enhancement at 20 MHz is less substantial.This can be explained by the faster local tumbling motion of the paramagnetic centre of Gd$L 4 , due to the longer and more exible crosslinking arms.
Nuclear Magnetic Resonance Dispersion (NMRD) proles, i.e. the variation of relaxivity (r 1 ) as a function of the applied magnetic eld strength, were measured at 298 and 310 K and at pH 7.4 in the proton Larmor frequency range 0.01-70 MHz (0.000234-1.64T).Representative examples of the NMRD proles of the monomeric complexes, and of the linear and crosslinked polymers are presented in Fig. 5.The monomeric complexes Gd$L 1-4 displayed proles typical for fast tumbling, low molecular weight complexes, each characterized by a steady decrease in their relaxivity at low magnetic eld (<1 MHz), a drop in relaxivity between 1 MHz to 10 MHz, followed by a second plateau in the high magnetic eld region (>10 MHz), governed by the rotational correlation time, s R .In the high eld region, the relaxivity is similar for the four monomers Gd$L 1-4 as expected, since their similar molecular weight, size and charge results in similar rotational dynamics.Raw data and a To t 1 H NMRD data at 298 K, the following parameters were xed in the tting procedure:  a To t the 1 H NMRD data at 298 K, the following parameters were xed: 25 Â 10 À5 cm 2 s À1 .b Estimated from VT NMR relaxivity proles (Fig. 5) at xed magnetic eld (20 MHz).tted NMRD proles are reported in the ESI (Tables S5-S8 and Fig. S9 †).
9][60] Certain parameters were xed to reasonable values according to previously reported examples: 25,56,61 a hydration number of q ¼ 1 was assumed, and the distance between the Gd(III) ion and a bound water molecule proton, r GdH , was set to 3.0 Å, based on crystallographic data for Gd-DOTA.The closest approach of the bulk water molecules, a GdH , was set to 4.0 Å, and the diffusion coefficient of a water proton away from the Gd(III) centre was assumed equal to D GdH ¼ 2.25 Â 10 À5 cm 2 s À1 at 298 K, or 3.1 Â 10 À5 cm 2 s À1 at 310 K.
To provide better estimates of the rotational correlation times of Gd$L 1-4 , by tting of the NMRD proles, the water residence lifetimes (s M ¼ 1/k ex ) were determined by 17 O NMR relaxometry.Thus, the temperature dependence of the transverse relaxation rate (R 2 ) and chemical shis (Du r ) were determined at 11.75 T at neutral pH using relatively concentrated solutions of Gd$L 1-4 (Fig. 5 and S10, Table S9-S12 †) and the proles were tted according to the Swi-Connick theory for 17 O relaxation. 62,63For complexes Gd$L 1-4 the s M values were similar (s M ¼ 119-166 ns) and in line with the values determined for similar Gd-DOTAGA (DOTAGA ¼ 2-(4,7,10-triscarboxymethyl-1,4,7,10-tetraazacyclododecan-1-yl)pentanoic acid) derivatives under the same conditions. 25However, it appears that the complexes Gd$L 3 and Gd$L 4 , bearing cismethacrylamide arms, exhibit slightly faster water exchange.
The NMRD proles obtained for the linear and crosslinked polymers revealed a signicant increase in relaxivity over the entire proton Larmor frequency range (Fig. 5), compared with monomeric complexes Gd$L 1-4 (raw data and tted NMRD proles are reported in Tables S13-S19 and Fig. S11 and S12 †).The prole shapes were also distinctly different, indicating successful incorporation of the monomeric complexes into higher molecular weight macromolecules.The proles of the linear copolymers P(NAM-r-Gd$L 1 ) displayed a broad relaxivity peak in the Larmor frequency range of 10-60 MHz, whereas in the case of the hyperbranched (crosslinked) polymers, the relaxivity peak was sharper in the 20-50 MHz region, indicating slower tumbling of the crosslinked polymers compared with the linear polymers.
To obtain more accurate tting of the NMRD proles of the polymeric systems, s M values were estimated by tting of the variable temperature 1 H NMR proles at 20 MHz for the linear polymers containing between 1-17 mol% Gd$L 1 theoretically, and for all hyperbranched polymers (Table S22-S24 †).In fact, the relatively low concentration of Gd(III) in the polymeric systems prevented the acquisition of 17 O NMR data, which typically require Gd(III) concentrations in excess of 5 mM.The water exchange rate of the coordinated water molecule of the polymerized Gd(III) complexes were determined to be two times slower than the corresponding monomeric complexes (s M values between 300-330 ns).Slower water exchange kinetics has been observed previously for other macromolecular Gd(III) systems, 25,27 and in the current study this may be tentatively ascribed to weak non-covalent interactions between the Gd(III) chelate and polymer backbone.Alternatively, a reduced rate of water diffusion through the polymer could also contribute to the slower water exchange rate. 10,64For linear polymers with 5 and 9 mol% of Gd$L 1 , the water residence lifetime was assumed to be 330 ns, since these polymers have similar molecular weights, dimensions and relaxivities to those containing 1 and 17 mol% Gd$L 1 .
The NMRD proles for the linear and crosslinked polymers were tted based on SBM theory and modied with the Lipari-Szabo approach for the description of the rotational dynamics of Gd(III) chelates covalently linked to macromolecules (see equations in Section 5). 65,66In particular, the contributions of the fast local tumbling motion of the Gd-chelate (s RL ) were separated from the slower global tumbling of the macromolecule (s RG ).s RL and s RG are associated with the order parameter, S 2 , which describes the level of interconnectivity between the local and the global motions (i.e., if S 2 ¼ 0 the motions are independent, if S 2 ¼ 1 the motions are fully linked).
The NMRD proles of the linear polymers were tted over the entire range of magnetic elds investigated (0.01 to 70 MHz, Fig. 5-B2 and S11 †), whereas for the hyperbranched systems, the proles were better tted using only the high eld data, i.e. above 1 MHz (Fig. 5-B3 and S12 †), ‡ as commonly performed for large macromolecular MRI CAs. 1 Compared with the discrete complexes Gd$L 1-4 , the linear polymers displayed slower local (s RL greater than 2-fold longer) and slow global (s RG greater than 20-fold longer) reorientation correlation times (Table 4).This can be ascribed to the incorporation of Gd$L 1 into polymer chains via the pendant arm of the macrocyclic ligand.The order parameter, S 2 , obtained for the linear polymers ranged between 0.12 and 0.18, which is reasonable for macromolecules containing Gd-chelates conjugated via a single exible linker, which allows relatively fast local tumbling.The hyperbranched polymers containing Gd$L 2-4 showed a large increase in relaxivity (18.6-22.8mM À1 s À1 at 60 MHz), primarily attributed to the slower tumbling of the crosslinked Gd(III) complexes.In fact, both s RL and s RG increased relative to the linear polymers (Table 4), and the order parameter S 2 , was also much higher than for the linear polymers (S 2 z 0.60 for P(NAM) containing 0.9% Gd$L 3 ), consistent with a more restricted motion of the Gd(III) chelate within the crosslinked systems.
Further inspection of the relaxivity data revealed that for the crosslinked polymers prepared from Gd$L 2 and Gd$L 3 , there is very little difference in relaxivity at 298 K and 310 K (Fig. 5-A3).This cannot be explained by differences in water exchange rate, since this parameter is very similar for the different polymers synthesised (Table 4).Rather, this can be attributed to more effective coupling of the local and global tumbling motion (S 2 up to 0.60) in the crosslinked systems, hence we do not lose any relaxivity gains at higher temperature.In the case of the crosslinked polymer synthesised from Gd$L 4 , the local and global motion is less effectively coupled (S 2 ¼ 0.40), and consequently the relaxivity at 310 K is approximately 11% lower than at 298 K.
A similar decrease in relaxivity is observed for the linear polymers at 310 K, where the motion of the Gd(III) centre is less efficiently coupled to the more slowly tumbling polymer (S 2 ¼ 0.12-0.18).

Comparison with previous macromolecular CAs
A comparison of the relaxivities and NMRD parameters obtained in this work with previously reported linear and crosslinked systems (acquired at the same magnetic eld, temperature and pH) is given in Table 5.Our linear polymers display similar or slightly higher relaxivities (12.6-13.5 mM À1 s À1 at 60 MHz) and similar NMRD parameters to comparable macromolecular or nanoscale systems (11-13 mM À1 s À1 ), involving Gd(III) complexes attached via single exible linker. 25,27he hyperbranched polymers containing Gd$L 2-4 showed a large increase in relaxivity (18.6-22.8mM À1 s À1 at 60 MHz).Interestingly, the values of the rotational correlation times and of S 2 are comparable to those reported for micellar aggregates obtained by self-assembly of a Gd$DOTAGA 2 complex bearing two C 12 aliphatic chains in cis-position, 25 as seen in Table 5.Also in that example, the restricted local motion of the Gd(III) complex was responsible for a strong relaxivity enhancement with respect to analogous micelles embedding a Gd(III) complex bearing only one aliphatic chain.Our hyperbranched polymers also displayed similar relaxivities and NMRD parameters to those obtained for hyperbranched dendrimers, conjugated with Gd$DOTA-pBn via a single arm (r 1 z 25 mM À1 s À1 , 298 K). 67 Comparing our hyperbranched polymers at 60 MHz and 310 K (Table 4), the system prepared from crosslinker Gd$L 3 , bearing the shortest pendant arms in a cis orientation, displayed a higher relaxivity (22.8 mM À1 s À1 ), than systems prepared from the trans-oriented crosslinker Gd$L 2 (20.7 mM À1 s À1 ) or the cis-oriented crosslinker Gd$L 4 with longer arms (18.6 mM À1 s À1 ).This indicates that the combination of shorter crosslinker arms in a cis-geometry is most ideal for limiting the local motion of the Gd(III) complex within the hyperbranched macromolecules.
Of the very few reported macromolecular CAs wherein the Gd(III) complex behaves as a crosslinker, 68-70 our system is the only example which takes advantages of the crosslinking to efficiently reduce the local rotational tumbling.This leads to higher relaxivity than those previously reported (e.g.crosslinked acrylamide nanogels 68 bearing Gd-DOTA or DTPA like ligands show r 1 ¼ 9.7-17.6mM À1 s À1 , 60 MHz, 310 K).To the best of our knowledge, only one other crosslinked system displays a slightly higher relaxivity (24.1 mM À1 s À1 at 60 MHz, 310 K), 69 which we propose is due to the slower global tumbling of the nanoparticles (average diameter of 65 nm), and faster water exchange (since all the Gd(III) complexes are on the outside of the nanoparticle), despite exhibiting faster local tumbling.

Conclusions
We have developed a strategy for the efficient synthesis of high molecular weight macromolecular contrast agents, via RAFT polymerization of kinetically stable Gd(III) monomers and crosslinkers Gd$L 1-4 , each based on a DOTA-like core bearing one or two pendant methacrylamide arms.
Copolymerization of Gd$L 1 , bearing a single methacrylamide arm, with NAM led to the formation of linear polymers with higher relaxivities (12.6-13.5 mM À1 s À1 at 60 MHz) and slower tumbling compared with the discrete monomeric complexes Gd$L 1-4 .Moreover, hyperbranched polymers prepared via the incorporation of crosslinked Gd(III) chelates Gd$L 2-4 displayed signicantly higher relaxivities and slower tumbling compared with the linear polymers.Analysis of the NMRD proles revealed that the higher relaxivities of the hyperbranched polymers is due to the restricted motion of the crosslinked Gd(III) chelates, which approximately doubles the local and global reorientation correlation times relative to the linear polymers containing Gd$L 1 .Crucially, the global motion of the hyperbranched polymers was more effectively coupled to the motion of the paramagnetic centre.This is apparent from an increase in the order parameter, S 2 , relative to the linear polymers, thus showing that the rotational exibility was signicantly reduced for polymers containing crosslinkers Gd$L 2-4 .
Hyperbranched polymers prepared from Gd$L 3 displayed the highest relaxivity (22.8 mM À1 s À1 at 60 MHz), suggesting that the combination of shorter polymerizable arms in a cis-orientation is optimal for limiting the local motion of the Gd(III) complex within a hyperbranched polymer.In comparison, hyperbranched polymers prepared from Gd$L 4 , bearing more exible arms in a cis-geometry, or from Gd$L 2 bearing two transrelated polymerizable arms, showed lower relaxivities (18.6-20.7 mM À1 s À1 at 60 MHz).The results obtained herein will guide the design of second generation Gd(III) monomers possessing three or four polymerizable arms, in order to access macromolecules with even higher relaxivities.
Our synthetic approach to macromolecular CAs combines the simplicity of a single polymerization step (with no postpolymerization modication) and scalability.The monomeric complexes Gd$L 1-4 serve as building blocks for the construction of more complex polymeric MRI CAs possessing responsive or theragnostic properties. 71Further, polymeric CAs capable of in vivo targeting may be addressed through the copolymerization of Gd$L 1-4 with monomers containing water-soluble sugar moieties or small peptide sequences, which modulate the second sphere of hydration.Work in this regard is ongoing in our laboratories.
Part C. Deprotected ligand 6b (354 mg, 0.54 mmol, 1.00 equiv.)and gadolinium(III)chloride hexahydrate (221 mg, 0.60 mmol, 1.10 equiv.)were stirred in deionized water (10.0 mL) for 24 hours at room temperature.Over the course of the reaction, the pH was adjusted to 7 by addition of aq.HCl or aq.NaOH.The reaction mixture was lyophilized and the resulting solid was redissolved in deionized water, centrifuged twice, and the supernatant was ltered through a syringe lter (220 nm cut-off).The crude product was puried by preparative HPLC (gradient 0-100% acetonitrile in 25  Part A -(S)-2-bromopentanedioic acid.(S)-2-bromopentanedioic acid was synthesized according to literature. 25L-Glutamic acid (9.21 g, 62.6 mmol, 1.00 equiv.)and KBr (22.4 g, 188 mmol, 3.00 equiv.)were dissolved in aq.HBr (1.00 M, 143 mL, 143 mmol, 2.30 equiv.) at room temperature.The reaction mixture was stirred for 5 minutes and cooled to À10 C. A solution of NaNO 2 (10.8 g, 157 mmol, 2.50 equiv.) in water (15 mL) was added dropwise over 2 hours.Aer the addition, the reaction mixture was stirred at room temperature overnight.H 2 SO 4 (4.00 mL) was added slowly and the resulting aqueous solution was extracted with Et 2 O (3 Â 100 mL).The combined organic layers were washed with brine (2 Â 75 mL), dried over MgSO 4 , ltered and concentrated under reduced pressure.The crude (S)-2-bromopentanedioic acid (4.00 g) was used directly in the next step without further purication.

Dynamic light scattering (DLS)
DLS measurements were performed with a Malvern Zetasizer Nano ZS using Zetasizer (version 7.12).The Zetasizer system uses a Diode-pumped solid-state laser operating at a wavelength of 532 nm and an avalanche photodiode (APD) detector.The scattered light was detected at an angle of 175 .The temperature was stabilized to AE0.1 C of the set temperature (25 C).All aqueous polymer solutions were ltered prior to measurement, using a nylon syringe lter with 220 nm cut-off.

Relaxometry measurements
NMRD.The observed water protons longitudinal relaxation rate constant (R obs 1 ) values were measured as a function of the magnetic eld strength in non-deuterated aqueous solutions on a Fast Field-Cycling Stelar SmarTracer relaxometer over a continuum of magnetic eld strengths from 0.00024 to 0.25 T (corresponding to 0.01-10 MHz proton Larmor frequencies) at 25 and 37 C by using the standard inversion recovery pulse sequence with 4 scans for each acquired data point.The relaxometer operates under computer control with an absolute uncertainty in 1/T 1 of AE1%.To complete the data set, 6 ESI data † points were obtained by measurements at higher magnetic elds (precisely 20, 30, 40, 50, 60 and 70 MHz) on a Stelar relaxometer with a Spinmaster console connected to a Bruker WP-80 magnet (80 MHz/2 T) adapted to variable-eld measurements.The temperature was set and controlled with a Stelar VTC-91 airow heater and measured by a substitution technique using a copper-constantan thermocouple (error AE 0.1 C).The exact concentration of Gd(III) was determined by measurement of bulk magnetic susceptibility shis of a tBuOH signal, 72 or by inductively coupled plasma mass spectrometry.The variable temperature 1 H NMR proles were obtained by measuring the relaxation rate at different temperature from 5 to 75 C (12 to 16 acquisition points) at a xed magnetic eld intensity (20 MHz or 30 MHz) using an inversion recovery method with a 90 pulse.
17 O NMR measurements.Variable-temperature 17 O NMR measurements were recorded on a 500 MHz Bruker Avance III spectrometer (11.75 T) equipped with a 5 mm probe and standard temperature was regulated by air or nitrogen ow controlled by a Bruker BVT 3200 control unit.The samples were analyzed at 278 K and from 280 to 350 K with a 5 K increment (16 measurements).Concentrated aqueous solutions of complexes (10-20 mM) at physiological pH (7.4) and containing 2.0% of the 17 O isotope (Cambridge Isotope) were used.The observed transverse relaxation rates (1/T 2 ) were measured from the peak width at half-height.The tting parameters were D 2 , s v , the s M value at 298 K, its enthalpy of activation DH M , and the scalar Gd-17 O w coupling constant A/ħ.

Fig. 3
Fig. 3 SEC molecular weight (MW) distribution of selected linear and hyperbranched copolymers demonstrating: (left) low Đ of linear copolymers and increasing MW as the ratio of Gd$L 1 : NAM increases; (centre) high MW shoulder and broad distribution of MW when a crosslinker is introduced; (right) increase in high MW content as initial concentration of monomer is increased.

Fig. 4
Fig. 4 Representative examples of DLS number weighted particle size distribution of linear and hyperbranched polymeric CAs obtained in deionized water at 25 C after filtration with a syringe filter (220 nm cut-off).Associated correlograms and detailed DLS results are presented in the ESI (Fig. S8 and Table S4 †).
25 Â 10 À5 cm 2 s À1 .b From the tting of the 17 O NMR data, with the xed value E V ¼ 1 kJ mol À1 and E R ¼ 20 kJ mol À1 .

Fig. 6
Fig. 6 Relaxivity gains observed for our linear or hyperbranched polymers, relative to Gd$DOTA (3.4 and 2.9 mM À1 s À1 at 20 MHz and 60 MHz, respectively), determined at 310 K at 20 MHz and 60 MHz.

Table 1
Monomer conversions and SEC data of linear P(NAM-r-Gd$L 1 ) copolymers RAFT polymerization in the presence of Gd-DOTA.b Theoretical number of NAM and Gd$L 1 units per chain depending of the initial polymerization reaction composition.c Conversion determined by 1 H NMR spectroscopy.d Expected M n ¼ ([NAM] 0 M NAM Conv NAM + [Gd$L 1 ] 0 M Gd$L1 Conv Gd$L1 )/[CTA] 0 + M CTA , with Conv Gd$L1 ¼ 1. e Obtained by SEC analysis (CH 3 Cl/triethylamine 98 : 2 v/v, RID detectors).Obtained by SEC analysis (H 2 O/MeOH 80/20 v/v with 0.1 M NaNO 3 , RID detector).g Gd(III) concentration determined by ICP-MS based on mass spectral signal of 157 Gd isotope.h N Gd Chain ¼ number of Gd(III) ions per polymer chain, estimated from ICP-MS data (ESI, Section 2, Table a f