Exploring the tumour extracellular matrix by in vivo Fast Field Cycling relaxometry after the administration of a Gadolinium‐based MRI contrast agent

1H Fast Field Cycling NMR (FFC‐NMR) relaxometry is proposed as a powerful method to investigate tumour stroma in vivo upon the administration of a Gd‐based contrast agent. To perform this study, an FFC‐NMR equipment endowed with a wide bore magnet was used for the acquisition of Nuclear Magnetic Resonance Dispersion profiles on healthy muscle and tumour tissue in living mice. At magnetic field strengths < of ca. 1 MHz, the differences in the relaxation rates of the intra and extracellular compartment become of the same order of magnitude of the exchange rate across the cellular membranes. Under this condition, the water exchange rate between the two compartments yields to a biexponential magnetization recovery that can be analysed by fitting the experimental data with the two‐Site eXchange (2SX) model. Using this model, it was possible to obtain, for the two compartments, both relaxation properties and water kinetic constants for water exchange across cell membranes. The method allowed us to determine the effect of the “matrix” on the water proton relaxation times and, in turn, to get some insights of the composition of this compartment, till now, largely unknown.


| INTRODUCTION
Magnetic Resonance Imaging (MRI) has got a key role in the field of oncology since few decades. Its prominent role relies on the superb spatial and temporal resolution of the images it can provide. The relaxation time T 1 , a measure of how fast the nuclear spin magnetization returns to its equilibrium state after the application of a radiofrequency pulse, is a very important source of contrast in MRI. However, at the magnetic field strength of the currently available MRI scanners, changes in tissue T 1 do not appear sensitive enough to report on some peculiar aspects of the tumour. [1] The noninvasive differentiation between benign and malignant lesions (in breast cancer for example), relies on the use of Dynamic Contrast Enhancement MRI (DCE-MRI) protocols, based on the use of Gd-based contrast agents. This procedure provides important functional information related to the enhanced permeability of blood vessels in the tumour region. Diffusion-weighted imaging and MR spectroscopic imaging protocols have been implemented into the clinical practice to obtain additional functional parameters that may be simultaneously analysed in the so-called multiparametric imaging approach. [2] However, there is widespread opinion that, at low magnetic field strength, the marked increase of R 1 (=1/T 1 ) observed in biological tissues might be used to improve the diagnostic potential of MRI in tumour phenotyping. [3][4][5] Although examples of 1/T 1 Nuclear Magnetic Relaxation Dispersion (NMRD) profiles [6] acquired in the 0.01-20 MHz range on fresh or thawed surgical specimens were reported since the early 80s of the past century, [7] the interpretation of these data remains almost completely empirical due to the lack of suitable model theories for their correct and exhaustive analysis. Moreover, the use of ex vivo samples has the flaw that it cannot take into account the dynamics of water mobility, that is a key determinant of the NMRD profile. Fast Field Cycling (FFC) is the only practicable way of measuring T 1 -dispersion; it involves switching the magnetic field between different field strengths, during the measurement procedure. Recently, the prognostic potential of the NMRD analysis has been highlighted by our group, using a prototype FFC relaxometer endowed with a wide bore magnet with a dedicated transmitter/receiver solenoid detection coil placed around the mouse's leg. Water proton 1/T 1 NMRD measured in vivo on implanted mammary tumours showed a marked T 1 elongation at low magnetic fields (<0.2 T) with respect to healthy tissues. [8] The differences observed in the NMRD profiles allowed the discrimination between mammary tumours characterised by different metastatic potential such as 4 T1, TS/A, and 168FARN models. These differences are directly correlated to the exchange rate of water molecules across the membrane of the tumour cells. In fact, the observed T 1 is the result of the mixing between the relaxation rates of the intracellular (R 1in ) and extracellular compartments (R 1ex ), being V ex and V in the respective volume fractions (Figure 1). Differences in the relaxation rates of the two compartments are inversely proportional to the magnetic field strength, becoming of the same order of magnitude of the exchange rate across the cellular membranes at magnetic fields of ca. 0.025 T (i.e., at the Larmor frequency of ca. 1 MHz). Under this condition the water exchange rate between the two compartments yields to a biexponential magnetization recovery that can be analysed by fitting the experimental data with the two-Site eXchange (2SX) model. [9][10][11][12][13][14] Thus, the measurement at magnetic fields <1 MHz allow the exploration of water dynamics and the separation of the intracellular contribution from the one of the extracellular compartment. This is not possible at high magnetic fields because the similar values of the relaxation rates in the extra and intracellular compartments cause the occurrence of the fast exchange regime that makes the magnetization recovery mono-exponential with a relaxation rate that is an average between the values of the two compartments. Landis and coworkers [13] showed that, at the clinical imaging fields, the biexponentiality of the M z -curve can be recovered if the extracellular relaxation rate R 1ex is "artificially" increased by the presence of paramagnetic contrast agents.
The aim of our work was to explore the role a paramagnetic contrast agent (CA), for example, Gd-HPDO3A (ProHance®, Bracco Imaging S.p.A., Milan, Italy), may have on the NMRD profiles acquired in vivo on murine models. On may expect that the paramagnetic agent may act as a reporter of the physicochemical characteristics of the "matrix" that defines the extracellular compartment ( Figure 1). In the literature, some example of NMRD profiles acquired after the administration of paramagnetic agents on fresh surgical specimens, [15][16][17] are reported. However, the use of ex vivo samples has the drawback that it cannot take into account the dynamics of the CA distribution as well as the water mobility, that are the key determinants of the NMRD profile. Six-week-old female BALB/c mice were inoculated in their muscle hind limb with about 1 million of 4 T1 (Group 1, n = 4) or TS/A (Group 2, n = 4) tumour cells suspended in 100 μl of phosphate-buffered saline. A third group of untreated mice (n = 6) was used as control. Both cell lines were initially derived from a spontaneous breast tumour growing in a BALB/c mouse. 4 T1 (ATCC® CRL-2539™ purchased from American Type Culture Collection [ATCC, USA]) and TS/A (kindly provided by prof. F. Cavallo's group, University of Turin) were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 U/ml Penicillin (P) with 100 μg/ml Streptomycin (S) and 4 mM Glutamine (Gln). Cells were cultured in 5% CO 2 /95% air at 37°C in a humidified chamber and were split every 2 to 3 days. All cells were tested negative for mycoplasma by MycoAlert™ Mycoplasma Detection Kit. All materials were purchased from Lonza (Basel, Switzerland).

| MRI
T 1 -weighted images of the mouse limb region were acquired on a 7 T Bruker AV300 spectrometer equipped with a Micro 2.5 microimaging probe and a birdcage resonator with 30-mm inner diameter. The animals were anesthetised before MRI examination as described above. Images were recorded 1 day before the acquisition of the NMRD profiles, before and after the injection of ProHance (0.2 mmol/Kg). The distribution of ProHance in the tumour region was followed by measuring T 1 before and 7′, 13′, 20, 32′ after the contrast agent injection, by means of a saturation recovery sequence (TE = 3.3 ms; number of slices = 3; slice thickness = 2 mm; FOV 30 × 30 mm; matrix 32 × 32). The tumour volume was measured from T 2weithed MRI images obtained by using a rapid acquisition with refocused echoes sequence protocol (TR = 5,000 ms; TE = 28 ms; number of slices = 11; slice thickness = 1 mm; FOV 50 × 50 mm; matrix 168 × 160).
Assuming the occurrence of a fast exchange regime (i.e.,|R 1in − R 1ex |« k in + k ex ), [18] at 7 T, the effective intratumour ProHance concentration, [Gd] eff , was determined by Equation 1 Gd Where R 1POST and R 1PRE are the relaxation rates measured after and before the CA injection; r 1 is the millimolar relaxivity measured at 7 T of ProHance for a 1-mM solution in Matrigel (r 1 = 4.86 mM −1 s −1 ) used as extracellular matrix model. Matrigel is the trade name for a gelatinous protein mixture secreted by Engelbreth-Holm-Swarm mouse sarcoma cells produced and marketed by Corning Life Sciences and BD Biosciences.
The [Gd] eff is the effective value of the tissue concentration, as it refers to the sum of the intracellular and extracellular volumes. Then, the ProHance concentration in the extracellular volume fraction (V ex ) is given by Equation 2 Gd

| RESULTS AND DISCUSSION
The two-Site eXchange (2SX) model describes the time evolution of M Z that depends on the absolute values of the relative size of the "relaxation" term, |R 1in − R 1ex |, and of the "exchange" term k in + k ex , (where k in = 1/τ in is the water exchange rate from the extracellular to the intracellular compartment and k ex = 1/τ ex is the water exchange rate in the opposite direction). This relationship was previously defined as the NMR "shutterspeed." [9,12,13] It is important to note that both R 1in and R 1ex depend on the magnetic field strength; their values are very similar at high magnetic field strength and progressively diverge on going to low magnetic field strengths, where R 1in > R 1ex . [8] R 1ex is influenced by the CA extravasation from blood capillaries to the extracellular space, [19] yielding R 1ex = R 1ex 0 + r 1 [Gd] ex , where R 1ex 0 is the contribution in the absence of exchange and CA and r 1 and [Gd] ex are the extracellular CA relaxivity and millimolar concentration, respectively. The values of the rate constants, k in and k ex , depend mainly on temperature, osmotic pressure, and cell metabolism.
In the range 0.01-1 MHz, that represents the proton Larmor frequencies at which the NMRD profiles were acquired, the predominant contribution arises from R 1in , and the condition │R 1in − R 1ex │~k in + k ex is expected to be met, even in the presence of ProHance (that causes the increase of R 1ex ). The system is in an intermediate-exchange regime, and the longitudinal 1 H 2 O relaxation exhibits a biexponential experimental decay associated to two apparent contributions, one with the larger and one with the smaller longitudinal relaxation time. By the 2SX analysis of the M z evolution, it is possible to dissect the two contributions and assess the ProHance r 1 value for the extracellular compartment. This method gives the possibility, for the first time, to explore the tumour extracellular matrix that regulates tumour cell migration and invasion, through the detection of the r 1 . This parameter is a good reporter of the interactions the paramagnetic complex may have with the biomolecules of the microenvironment in which it distributes.
Thereafter, two mouse mammary adenocarcinoma cell lines, namely, TS/A and 4 T1, were injected in the muscle of the hind limb to obtain the corresponding tumour xenografts suitable for in vivo studies. In order to set-up the in vivo acquisition time interval of the NMRD profiles, ProHance was administered intravenously to mice bearing tumours with a volume that is >65% of the total leg + tumour tissue. The CA biodistribution dynamics were evaluated by measuring T 1 on a 7-T MRI scanner after the injection of ProHance at 0.2 mmol/kg dose. The experiment was carried out the day before the NMRD profile acquisition in order to deal with similar anatomical/functional conditions. The use of a high magnetic field strength ensured the occurrence of a fast exchange regime, that is, | R 1in − R 1ex |« k in + k ex , that makes the relaxation rate measurement independent on the presence of intra and extracellular compartments. Figure 2 shows that the Gd concentrations ([Gd] eff ) found in both mammary tumour xenografts were significantly higher than those found in the healthy muscle. This is due to an increase of permeability of the neoformed vessels, often called enhanced permeation and retention, effect shown by many tumour types. By analysing the contrast dynamic behavior, the appropriate NMRD acquisition interval was set between 7 and 20 min after the ProHance injection (i.e., an acquisition time of 13 min, the grey area in Figure 2). During this time, the Gd complex concentration showed a variation of less than 30% in the tumour tissue (more precisely between 0.136 ± 0.018 mM and 0.123 ± 0.017 mM for 4 T1; between 0.140 ± 0.009 mM and 0.115 ± 0.006 mM for TS/A). The fast elimination of ProHance from the tumour tissue forced us to measure R 1 at a maximum of six magnetic field strengths, in the 0.01-1 MHz proton Larmor frequency range. Figure 3 shows the observed relaxation rate R 1 values obtained by the monoexponential fitting of magnetization recovery for 4 T1 (on the left) and TS/A (on the right) tumours before and after the injection of ProHance. As expected, the relaxation rate measured after ProHance injection is higher than the PRE contrast value at any magnetic field strength. The enhancement of the relaxation rate (i.e., [R 1POST -R 1PRE ]/R 1PRE ) brought by ProHance markedly decreases as the applied magnetic field strength decreases. Basically, this is the consequence of the simultaneous increase of tissue relaxation rates (mainly due to the R 1in contribution as shown above) and the nearly constant ProHance relaxivity in this magnetic field range (Figure 4). A good model for the determination of T 1 in the presence of ProHance in the extracellular matrix is obtained by dissolving the CA in Matrigel, a gelatinous protein mixture secreted by Engelbreth-Holm-Swarm mouse sarcoma cells. It is a model of the extracellular environment found in many tissues and used as a substrate for cell cultures. [20] Figure 4 shows the NMRD profile of ProHance dissolved in water and in Matrigel, and Table 1 reports the parameters obtained by the fitting for the inner-and outer-sphere contributions (Solomon-Bloembergen-Morgan and Hwang-Freed theory, [23,24] respectively). An increase of the rotational correlation time, τ R , value of about 60% accounts for the different behavior, as a consequence of the higher viscosity of Matrigel. For both profiles, the considered τ M is the value determined in water. τ M was not determined in Matrigel as this medium changes the physicochemical state from gel to liquid at T < 22C°. However, being τ M determined by the enthalpy of the Gd-Ow bound, one can reasonably assume that it is not significantly affected on going from neat water to Matrigel. [25,26] Next, the fitting of the magnetization recovery curves acquired as a function of the magnetic field strength on the two tumour models before the ProHance injection, was then performed according to the 2SX model taking into account the biexponential behavior of the magnetization. Data were simultaneously analysed, sharing the V ex end τ in parameters maintaining R 1ex 0 fixed to the Matrigel values obtained in a separated experiment (Figure 4). The V ex was allowed to vary within a reasonable range, in accordance with results already reported in the literature (0.09-0.19 for healthy mouse hind limb, 0.15-0.5 for tumour mouse hind limb. [13,[27][28][29] The R 1in admitted range was the one that previously reported on the same type of tumours. [8] The results obtained from the analysed NMRD profiles (six field values and 32 tau) on healthy mouse and tumour bearing one were comparable with those ones reported in our previous studies under slightly different conditions (10 field values and 48 tau, total time 32 min). [8] As expected, the tumour extracellular volume (V ex ) was higher than in healthy tissue, going from 0.13 ± 0.02 to 0.22 ± 0.01 and 0.26 ± 0.03 for healthy, TS/A, and 4 T1 tissues, respectively. [13,[27][28][29] The opposite trend for the intracellular water residence time was confirmed (1.24 ± 0.1 s for healthy tissue, 1.12 ± 0.14 s for TS/A tumour model and 0.68 ± 0.16 s for the more aggressive 4 T1 tumour model).
In principle, the fitting of the NMRD data acquired after the ProHance injection involves a high number  The considered τ M is the average value determined in water weighted over the relative isomeric population the ratio for the two possible coordination geometries SAP (capped square antiprismatic geometry) and TSAP (capped twisted square antiprismatic geometry) is 75% and 25%, respectively [21,22] ; b the value was fixed taking into account the relative differences (14%) of the apparent diffusion coefficient measured by DWI-MRI (see supplementary material) in water and in Matrigel, respectively.
*Fixed during the fitting procedure. . The resulting R 1in and R 1ex obtained applying this fitting protocol are reported in Figure 5. The slight increase of R 1ex observed at ca. 1 MHz may be due to that, in the 2SX model, the compartment with lower R 1 is no longer the extracellular one as a consequence of the decrease of the intracellular relaxation rate. This can cause a slight error in the R 1ex estimation. The r 1 values obtained from the fitting ( Figure 6), referred to the extracellular compartment, are reasonably similar to those one obtained for ProHance in Matrigel. This is the first time that an NMRD profile of a paramagnetic contrast agent is obtained in vivo after its distribution in the extracellular matrix of mammary tumour, taking into account the water exchange rate with the intracellular compartment. This finding gives further support to the view that the extracellular matrix in these type of tumours shows a slightly increased viscosity that, however, does not cause any dramatic change in the ProHance r 1 . This could be due to the presence of a limited amount of proteins or other polymeric species (the "matrix") in these aggressive and partially necrotic tumours. In fact, the mobility of water appears similar to the one shown in a gel-like matrix as Matrigel that can be definitively considered a good model for the extracellular matrix in this kind of tumours. Finally, in the presence of clinical doses of Gd-CA, the magnetization recovery, in the investigated magnetic field strength range, can be well described by the 2SX model.

| CONCLUSION
Herein the first in vivo NMRD profiles for tumour tissues, obtained in the presence of a paramagnetic Gd-complex are reported. Upon the analysis of the experimental data obtained in the 0.01-1 MHz range by means of the 2SX procedure before the CA injection, it has been possible to extract the relaxation rate values for the extracellular and intracellular compartment (R 1ex 0 and R 1in ), the intracellular water residence time (τ in ) and the intracellular water fraction (V in = 1-V ex ). Fixing these values, it was then possible to extract the r 1 relaxivity of the paramagnetic agent in the extracellular compartment where it distributes. The obtained values are similar to those obtained when ProHance is dissolved in Matrigel that, therefore, results to be an excellent model to mimic the characteristics of the extracellular compartment of the mammary tumours considered in this work. The method, by assessing the relaxation properties of the extracellular space, will allow us to determine the effect of the "matrix" on the water proton relaxation times and, in turn, to get some insights of the composition of this compartment, till now, largely unknown. This open new horizon in the use of the fast field cycling MRI technologies that are currently under development for clinical use. [30,31]