Anthranilic acid analogues as diamagnetic CEST (diaCEST) MRI contrast agents that feature an IntraMolecular-bond Shifted HYdrogen (IM-SHY)

Abstract

Diamagnetic chemical exchange saturation transfer (diaCEST) agents are a new class of imaging agents, which have unique magnetic resonance (MR) properties similar to agents used for optical imaging. Here we present a series of anthranilic acid analogues as examples of diaCEST agents that feature an exchangeable proton shifted downfield, namely, a IntraMolecular-bond Shifted Hydrogen (IM-SHY), which produce significant and tunable contrast at frequencies of 4.8 – 9.3 ppm from water. Five analogues of N-sulfonyl anthranilic acids are all highly soluble and produced similar CEST contrast at ~ 6 – 8 ppm. We also discovered that flufenamic acid, a commercial non-steroidal anti-inflammatory drug, displayed CEST contrast at 4.8 ppm. For these N-H IM-SHY agents, the contrast produced was insensitive to pH making these complementary to existing diaCEST probes. This initial IM-SHY library includes the largest reported shifts for N-H protons on small organic diaCEST agents, and should find use as multi-frequency MR agents for in vivo applications.

Introduction

Chemical exchange saturation transfer (CEST) contrast agents, first introduced in 2000(1), are an alternative to traditional magnetic resonance (MR) contrast agents, which rely on direct enhancement of water relaxivity. The CEST mechanism involves saturation of labile protons on the agents via selectively-irradiation at their resonance frequencies. The signal loss is then transferred to surrounding bulk water through chemical exchange, leading to a reduction in water signal (2–4). This water signal loss (CEST contrast) results in an amplification of the signal from low-concentration protons through the multiple exchange events occurring during the saturation pulse. Because the CEST contrast is derived from irradiation at a specific proton frequency, it is easier to discriminate from other sources of signal change than T1 or T2* contrast. This frequency dependence of contrast also allows the simultaneous detection and discrimination of multiple agents within an image(5–7). Diamagnetic CEST (diaCEST) and paramagnetic CEST (paraCEST) agents have been the subjects of several recent reviews(8–11). DiaCEST agents, such as glucose(12–14), glycogen(15), myo-inositol(16), glutamate(17), creatine(18,19), L-arginine(20,21), glycosaminoglycans(22,23) and peptides(5,24–26) are attractive biocompatible materials, but compared with paraCEST agents(27), they suffer from reduced sensitivity due to the relatively small chemical shift difference between their exchangeable protons and those of water (1–5.0 ppm). To address this issue, diaCEST agents with protons of increased chemical shift have been reported, including the thymidine analogues (5.5 ppm)(28) and iopamidol (4.2 and 5.5 ppm)(29,30). Most recently, we reported that the C2-OH in 2-hydroxybenzoic acid analogues resonates between 8.7 – 10.8 ppm from water, with solute-to-water exchange rates (k_sw) that are well suitable for CEST imaging(31). Building upon that report, here we describe the anthranilic acid analogues: N-aryl derivatives, N-acyl derivatives and N-sulfonyl derivatives, as another class of IntraMolecular-bond Shifted Hydrogens exchangeable proton (IM-SHY) diaCEST agents, based on the exchange of N-H protons instead of O-H (Scheme 1).

Scheme 1.

CEST frequency [ppm], contrast [%] and k_sw, [kHz] of anthranilic acid and its analogues. Experimental conditions: pH 7.1 –7.5, using t_sat=3 s, B₁=3.6 µT. For Z-spectra, see Tables S1 and S2. All the MR experiments were performed at 37°C.

Results and Discussion

Salicylic acid (1) displays CEST contrast at 9.3 ppm (31) (Fig. 1). This dramatic chemical shift derives from the low barrier hydrogen bond between the exchangeable phenolic proton and the carboxylate anion at neutral pH(32,33). We also determined that similar CEST signals could be observed in other compounds with the 2-hydroxybenzoic acid scaffold, representing a powerful new type of CEST agent, based on the principle of IM-SHY(31). We were interested in preparing similar agents with labile anthranilic rather than phenolic protons to explore further the capabilities of the benzoic acid core for generating CEST contrast. However, anthranilic acid (2), an N-H analogue of salicylic acid, failed to produce contrast (Scheme 1, Fig. 1). To understand why, we measured the CEST contrast properties of a wide range of common anthranilic acid analogues, including those with N-alkyl, N-aryl, N-acyl, and N-sulfonyl substitutions (Scheme 1). Interestingly, significant contrast was observed in N-phenylanthranilic acid (4), although the labile protons resonate at 4.8 ppm, which is much lower than the 9.3 ppm observed in 1. At a relatively low saturation field strength (B₁) = 3.6 µT, 4 showed a broader peak in the CEST MTR_asym spectrum than that of 1 and 12 (Fig. 1b), indicating a faster exchange. Using the QUESP experiment (34) we measured k_sw = 2.0 kHz (supplemental Fig. S1), which is slightly too fast to obtain optimal CEST contrast using the 3 – 5 µT saturation pulses we are able to employ on our clinical scanners. Comparing the CEST signal between 4 and 2, the loss of CEST signal in 2 indicates that k_sw is too high. This is possibly due to the presence of the additional non-hydrogen-bonded C2 N-H proton, which might undergo a fast intramolecular exchange with the hydrogen-bonded proton. In addition, if we modify 2 through substitution of a methyl group for one of the amine protons (3), the CEST contrast is still absent, which implies stereoelectronic influences are also important (Scheme 1). It is worth mentioning that N- phenylanthranilic acid analogues are commonly used as non-steroidal anti-inflammatory drugs (NSAIDs). The CEST properties were measured on five commercially available drugs, including flufenamic acid (5), meclofenamic acid (6), mefenamic acid (7), tolfenamic acid (8) and niflumic acid (9). Their water solubility is generally low (~ 10 mM or lower). As shown in Scheme 1, flufenamic acid (5) showed similar CEST properties to 4. The exchangeable proton resonates at 4.8 ppm, with k_sw = 1.0 kHz. The CEST data of 6, 7 and 8 indicated the importance of steric interaction on the proton exchange rate with water. Adding the chloro group ortho to the exchangeable N-H (6) reduced its water accessibility and the CEST contrast dropped to 1%. This is presumably because the exchange is too slow, however, it’s difficult to quantify k_sw because of the small contrast. Increasing the steric hindrance through addition of methyl (7 and 8) eliminated the CEST signal. Niflumic acid (9), the pyridine analogue of 5, did not display any CEST contrast. One possible explanation is that the presence of the pyridine nitrogen tends to strongly hydrogen bond to water and alters the proton exchange of the IM-SHY -NH.

Figure 1.

CEST contrast curves for representative salicylic acid (1) and anthranilic acid derivatives (2, 4, 11 and 12) at concentrations of 25 mM (pH 7.1–7.4) using B₁ = 3.6 µT, t_sat = 3 s. The gray box indicates this group of agents includes a new frequency region for amide and sulfonamide protons.

We next determined the detection limits of 5 with CEST, because it could potentially be translated into clinical applications(35). The solubility of 5 is quite poor at pH values below 7, however 10 mM could be achieved in PBS buffer at pH above 7.2. As shown by the QUESP data in Fig. 2a, the contrast is near maximal at B₁ > 6 µT, with a smaller k_sw (1.0 kHz) than that of 4. The peaks in the Z-spectrum and the MTR_asym spectrum are also sharper than those of 4 (Table S1), which is also due to a slower k_sw. The contrast of 5 is nearly linearly dependent with concentration over a range from 0.75 mM to 10 mM (36) (pH 7.4), with 1.2% contrast observed at a concentration of 1.5 mM (Fig. 2b).

Figure 2.

CEST properties of 5. a) QUESP data at 10 mM at pH = 7.4, with k_sw = 1.0 kHz where the data are shown as points and the solid line representing the best fit after numerically solving the 2-pool Bloch equations; b) CEST contrast at 4.8 ppm as a function of concentration using B₁ = 3.6 µT. (Solid line: linear fitting)

In an attempt to increase the chemical shift further to fit the slow to intermediate detection window of CEST (k_sw < Δω) while still keeping k_sw slow enough for achieving efficient saturation using a B₁ suitable for the MR hardware used in our in vivo scans, we investigated the C2 amide analogues of anthranilic acid. Amide N-H protons tend to be shifted further than amine protons, although they also tend to exchange with water slower as well (5). As expected, 10 did not show any CEST contrast presumably because k_sw is too slow (Fig. 3a, Scheme 1). However, after modification of the structure to 11, an example of a more acidic N-H proton, we observed CEST contrast with the labile proton resonating at 9.3 ppm indicating a strong hydrogen bond interaction in water. The contrast produced by 11 is relatively low (6 % at 25 mM, B₁ = 3.6 µT), because k_sw is relatively slow (0.3 kHz, see supplementary Figs. S2, S3 for QUESP/pH details). Further increasing the acidity through 2-(methyl-sulfonamido) benzoic acid (12) results in more substantial contrast at 7.3 ppm (~15 % at 25 mM, B₁ = 3.6 µT), based on adjusting the proton exchange of the IM-SHY -NH. According to our QUESP measurements, 12 displays a k_sw = 0.6 kHz at pH = 7.1, which is quite similar to salicylic acid (31) and barbituric acid (supplemental Fig. S5). Maximum contrast is achieved using B₁ = 6 µT or higher with ~90% of this contrast available at B₁ = 3.6 µT (Fig. 3c), which is near the maximum power we can apply using a parallel transmit body coil on our clinical scanners. More interestingly, the contrast and k_sw of 11 and 12 remained almost constant between the pH values 6 – 8 (Figs. 3d, S2, S3, S4). For comparison, salicylic acid (1), an alternative IM-SHY agent, possesses protons with k_sw that decrease dramatically over this range (k_sw = 2.4 kHz at pH 6.5, k_sw = 0.4 kHz at pH 7.8). This pH independence makes 11 and 12 ideal IM-SHY probes for in vivo quantification purposes. As expected, a nearly linear relationship between contrast and concentration was observed for 12 (Fig. 3b), with 1% CEST contrast produced at a concentration of 1.5 mM. Although the chemical shift is not as large as 1 or 11, 12 represents the first diaCEST agent with labile N-H protons resonating at 7–8 ppm from water that produces significant contrast. This compound should be useful for multiple frequency detection and complementary to other existing diaCEST probes.

Figure 3.

CEST properties of 10 – 16. a) Z-spectra and MTR_asym for 10 – 12 at 25 mM, pH = 7.2, t_sat = 3 s and B₁ = 3.6 µT; b) CEST contrast of 12 at 7.5 ppm as a function of concentration, using B₁ = 3.6 µT; c) QUESP data of 12 at 25 mM, pH = 7.1, with k_sw =0.6 kHz; d) pH dependence of % contrast for 12; e) Analogues of 12 with different CEST peak frequencies from 6 – 8 ppm.

Encouraged by the result from 12, we studied several commercially available analogues to check if the CEST contrast of this scaffold would tolerate chemical modification. As shown in Scheme 1 and Fig. 3e, similar contrast was obtained upon chemical modification of the aniline ring (13–15), with the CEST frequency varying from 6 ppm to 7.3 ppm. Placing a strong electron donating -NH₂ group (15) at the para-position to the C2-NH reduced the CEST frequency to 6.3 ppm, which is quite similar to the electronic effects we observed previously (31). Placing a –Cl at the para-position of the C2-NH (13) leads to faster k_sw (1.0 kHz), and as a result a higher CEST contrast (~20%). Substitution of a phenyl for the methyl (16) resulted in deshielding with the chemical shift increased to 7.8 ppm. In comparison, replacing the methyl group in 12 with a -CF3 (17) results in loss of CEST contrast. As this group of agents, 12–16, generated similar contrast to 1 in phantoms, we further chose to monitor in vivo the contrast in kidneys after administration into the tail vein of mice of the most sensitive, 13 (Fig. 4). The contrast was monitored over time, and compared to the pre-injection images (Fig.4b), we observed a 2–3% increase in the CEST contrast 7.5 minutes after injection integrating from 7.0 – 7.6 ppm (Figs. 4b,c). The histogram in Fig. 4d indicates the pixelwise distribution of MTR_asym values for mouse 1 pre- and post-injection. A negative MTR_asym was observed as baseline for the kidneys, which is presumably due to strong relayed NOE transfer of signal loss to water (37,38). As shown in Fig.4e, for both mice the contrast reaches maximum at ~7.5 mins. post-injection

Figure 4.

In vivo contrast for 13. a) T2w image; b) overlay MTR_asym map pre-injection for mouse 1; c) overlay MTR_asym map at 10 min post-injection for mouse 1; d) histogram displaying the distribution of MTR_asym for Mouse 1 pre- and post-injection (Figs.4c,d). e) dynamic time course of ∆MTR_asym based on ROIs enclosing both left and right kidneys for the two mice using ω₁= 3.6 µT (circle: Mouse 1, triangle: Mouse 2, solid line, average value of Mouse 1 and Mouse 2).

As is shown above, anthranilic acid IM-SHY probes have larger shifts for their exchangeable protons than spherical lipoCEST agents(10), and similar shifts to those found for paraCEST probes such as Yb-DO3A-oAA(39). The shifts are not nearly as large as some of the Yb, Eu, Tm or Dy complexes described previously(40–43) or the cryptophane cages used for hyperCEST(43) however because k_sw can be tuned as slow as 0.5 – 1 kHz through structure changes and is insensitive to pH in the physiologically relevant range, these USHY probes are well suited for detection using saturation pulses attainable on clinical scanners. A more detailed investigation of the steric and electronic factors for this scaffold is ongoing.

Conclusion

We have demonstrated that anthranilic acid provide a suitable scaffold for tunable IM-SHY diaCEST agents. Labile protons in N-aryl anthranilic acids (4–6) resonate at 4.8 ppm while for N-sulfonyl anthranilic acids (12–16) these resonate between 6 – 8 ppm and for 11 labile protons resonate at 9.3 ppm. Anthranilic acid analogues could be used for multi-color MR imaging, with one NSAID, 5, already administered to patients, having been identified among these analogues. The 2-sulfonamidobenzoic acid scaffold has been shown to allow chemical modification with labile protons that exchange in a non-pH dependent manner, which could be advantageous for in vivo quantification. Additional studies are ongoing to improve our understanding of the relationship between CEST properties and molecular structure for these and other IM-SHY diaCEST agents.

Experimental Section

Phantom Preparation and data acquisition

Compounds 1 – 12 were purchased from Sigma Aldrich (St. Louis, MO). Compounds 13–17 were purchased from Enamine Ltd (Monmouth, NJ). Samples were dissolved in 0.01 M phosphate-buffered saline (PBS) at several concentrations from 1.5 mM up to 25 mM depending on the solubility, and titrated using high concentration HCl/NaOH to various pH values ranging from 6 to 8. The solutions were placed into 1 mm glass capillaries and assembled in a holder for CEST MR imaging. They were kept at 37 °C during imaging. Phantom CEST experiments were performed on a Bruker 11.7 T vertical bore MR scanner, using a 20 mm birdcage transmit/receive coil. CEST images were acquired using a RARE (RARE factor = 8) sequence with a continuous wave (CW) saturation pulse length of 3 s and saturation field strength (B₁) from 1.2 µT to 14.4 µT. The CEST Z-spectra were acquired by incrementing the saturation frequency every 0.3 ppm from −15 to 15 ppm; TR /effective TE = 6s/17 ms with linear phase-encoding, matrix size = 64*48 and slice thickness = 1.2 mm. For determining k_sw using QUESP, Z-spectra were collected at B₁ = 1.2 µT, 2.4 µT, 3.6 µT, 5.4 µT, 7.2 µT, 10.8 µT and 11.4 µT.

In vivo mouse imaging

To evaluate whether the N-sulfonyl derivatives, 12–16, could be detected after administration into live animals, we injected two mice with 60 µL of a 0.25 M solution of compound 13 and collected CEST images. Images consisting of a single axial slice containing both kidneys were collected. To improve the temporal resolution and able to correct the B₀ shift, we collected a partial z-spectrum every five minutes by incrementing ∆ω over ten frequencies: [±8.2 ppm, ±7.6 ppm, ±7.3 ppm, ±7 ppm, ±6.6 ppm], and an average MTR_asym at [±7.6 ppm, ±7.3 ppm, ±7 ppm]. The imaging sequence employed is the same as for the phantoms, with the following parameters:B₁ = 3.6 µT, T_sat = 3 s, TR/effective TE = 5 s/16 ms with linear phase-encoding, matrix size 96×64.

Post-processing

CEST contrast was quantified using MTR_asym = (S(−Δω)–S(+Δω))/S₀for phantom and 1- S(+Δω)/ S(−Δω) for in vivo to increase the temporal resolution and reduce the motion where S(+Δω) represents water signal intensity with a saturation pulse applied at the frequency +Δω and S₀ represents the water signal without a saturation pulse. The Z-spectra were corrected pixel by pixel using a B0 map acquired using WASSR as described in detail previously (9). To indicate the kinetics of CEST contrast upon injection of the agents, we subtracted the MTR_asym values at each time-point with a reference MTRasym(0) at pre-injection, i.e. ∆MTR_asym (t) = MTR_asym (t) – MTR_asym (0), and plotted the averaged ∆MTRasym (t) of the whole kidney as a function of minutes post-injection. The solvent to water exchange rate (k_sw) was calculated according to the QUEST and/or QUESP methods (34), which were considered as a simple and robust method for estimating k_sw, especially for the slow to intermediate exchange regime(44,45). In particular we numerically solvedthe 2-pool model Bloch equations to fit the measured MTR_asym values as a function of different T_sat or B₁ as described previously (34), with the parameters for the fittings: R_2w = 0.9 s⁻¹, R_1s = 0.71 s⁻¹, R_2s = 39 s⁻¹, T_sat = 3 s. R_1w was allowed to float between 0.33 – 0.40 s⁻¹ to obtain the best fit. The QUESP/QUEST fittings are shown in supplemental Figs. S1–S5.