CEST theranostics: label-free MR imaging of anticancer drugs

Image-guided drug delivery is of great clinical interest. Here, we explored a direct way, namely CEST theranostics, to detect diamagnetic anticancer drugs simply through their inherent Chemical Exchange Saturation Transfer (CEST) MRI signal, and demonstrated its application in image-guided drug delivery of nanoparticulate chemotherapeutics. We first screened 22 chemotherapeutic agents and characterized the CEST properties of representative agents and natural analogs in three major categories, i.e., pyrimidine analogs, purine analogs, and antifolates, with respect to chemical structures. Utilizing the inherent CEST MRI signal of gemcitabine, a widely used anticancer drug, the tumor uptake of the i.v.-injected, drug-loaded liposomes was successfully detected in CT26 mouse tumors. Such label-free CEST MRI theranostics provides a new imaging means, potentially with an immediate clinical impact, to monitor the drug delivery in cancer.


S1. The effect of saturation parameters on the CEST contrast of cytidine based anticancer drugs
To determine the optimal saturation pulse, we acquired the Z-spectra and MTR asym spectra of the cytidine based anticancer drugs, gemcitabine (dFdC), cytarabine (araC), decitabine (Dec), and azacitidine (Aza) and their natural analog deoxycytidine (dC) using, 1) varied B 1 : 1.2, 2.4, 3.6, 4.7 and 5.9 µT with a fixed T sat = 4 seconds; or 2) varied pulse duration (T sat ) of 0.5, 1, 1.5, 2.5, 4, 6 seconds with a fixed B 1 = 4.7 µT. All samples were prepared at a concentration of 20 mM and pH 7.4. Triplicate samples were prepared at each condition for each agent. The results are shown in Figure S1. Note that the offset frequencies of hydroxyl protons were found to differ slightly between different agents. For the plots shown in c and d, we used 1.2 ppm for Aza and Dec.
These results can be used to estimate the practically useful RF pulse parameters for acquiring maximum CEST MRI. For example, the good saturation parameters for detecting dFdC and dC, are B 1 =3.6 µT and T sat = 3 seconds. Further increasing saturation power and length won't increase the CEST contrast much. Instead it will increase water direct saturation and competing semi-solid magnetization transfer effects, as well as the SAR.

S2. The pH dependence of gemcitabine and other cytidine-based anticancer drugs
The pH dependence of other cytidine-based anticancer drugs between pH 6.5 and pH 8 was measured and the results are shown in Figure S2. We also used the frequency-labeled exchange (FLEX) transfer method [1] as previously described [2] to determine the exchange rate of amino protons for each drug.
In brief, FLEX experiments were conducted at 17.6 T on a Bruker Avance III spectrometer using a micro2.5 microimaging probe. Each FLEX acquisition consisted of 600 LTMs, each LTM containing excitation pulses with flip angle = 90 degrees, duration = 0.15 ms, and the carrier frequency was placed 8.6 ppm downfield from the water resonance. The t exch was set to 8 ms for a total preparation time of 5 s. The FLEX labeling period was followed by a single-shot fast spin-echo (FSE) imaging readout with TR/TE = In total, 100 images were acquired with t evol ranging from 0 ms to 3 ms.
FLEX data was processed using a collection of python packages (www.numpy.org, www.scipy.org, and www.matplotlib.org). The estimated exchange rates of the amino protons on each compound are listed in Table S1. It should be noted that due to the small difference in the chemical shifts of NH 2 and OH, the OH protons could have a strong effect on the measurement of NH 2 protons. We also used the same FLEX method to determine the exchange rate of dFdC at different pH. As shown in Table S2, the change in exchange rate is consistent with the trend of CEST contrast change shown in Figure S2c.

S3. Definition of CNR in the CEST MRI detection
In the current study, the CNR (between pre-and post-contrast) was determined using the equation below [3], Where SNR is the signal-to-noise Ratio and SNR S0 is the SNR determined using the S 0 image (no saturation). Noise was estimated from the standard deviation (δ) of background noise in the S 0 image. S +Δω and S -Δω are the MRI signal intensities after saturation at negative and positive values of the offset frequency Δω from the water proton frequency (which is set at 0 ppm by convention).
The difference in CEST contrast between a gemcitabine-containing sample and its corresponding reference sample then was defined as ∆CNR,

∆CNR= CNR agent -CNR control
where CNR agent is the CNR of a agent-containing samples and CNR control is the CNR of a reference sample (either PBS or gel).

S.4. Comparison of CEST and 19 F MRI measurements on detecting dFdC
Because dFdC contains two 19 F atoms in its structure, it can also be detected by 19 F MRI [4]. We therefore performed a comparative study of CEST MRI and 19 F MRI on the same set of phantoms. The result is shown in Figure S5.
MRI experiments were performed on a vertical 17. 6 Table S4.

S6. Characterization of liposomal drugs
To determine the stability of the prepared liposomes, we used both CEST and UV absorbance (OD268.8) to assess the retained dFdC in liposome over time. In brief, liposomes were filtered immediately after formation using a G50 column to remove unencapsulated drug. Six samples of 0.3 ml liposome solution were placed in a 0.5 ml dialysis cassette (3.5 KD MW cut off, Thermo scientific), and each of them was immersed in 40 mL PBS buffer under continuous stirring, which was sufficient to produce a good sink condition considering that the solubility of gemcitabine in PBS is approximately 2 mg/ml. At each time point, i.e., 0, 1, 2, 3, 4, and 24 hours, one sample was removed for measurement of CEST contrast. After the CEST MRI, the sample was suspended in a 10% v/v Triton X-100 solution and thoroughly agitated using a water bath sonicator at 42 °C, followed by centrifugation (21,000 ×g, 10 min), followed by UV absorbance measurement at 268.8 nm using a UV-Vis spectrophotometer [7].
To determine the stability in serum, filtered liposomes were mixed with an equal volume of fetal bovine serum (FBS) to a final volume of 450 µl in a dialysis cassette (3.5 KD MW cut off) [8]. The dialysis cassette then was immersed in 60 mL PBS. At each time point, i.e., 0, 1, 2, 3, 4, and 24 hours, a 3 mL sample of dialysis solution was taken

S7. In vivo MRI results
The results for all the animals are shown in Figure S5 and Figure   Histogram of the MTR asym values within the tumor regions, before and after the injection of liposomes.

In vivo fluorescence imaging of the tumor uptake of liposomes
Fluorescence imaging was performed and analyzed using a Spectrum/ CT IVIS® in vivo imaging system with the Living Image® software (PerkinElmer, Waltham, MA).
Fluorescence signal (emission = 620 nm, excitation = 570 nm) was quantified as radiant efficiency. The exposure time for each image acquisition was 1 s. Images were scaled to the same maximum intensity using the supplier's software. Because rhodamine-B-PE is incorporated into the liposome bilayer, the fluorescent signal is an excellent representation of the location and amount of liposomes [9,10].

Biodistribution assessed by ex vivo fluorescence
Immediately after MRI measurements and in vivo fluorescence imaging, mice were sacrificed by cervical dislocation and tumor, brain, liver, spleen, kidneys, and lung were collected and assembled on a Petri dish for image acquisition. All images were scaled to the same maximum intensity for direct comparison. For quantification, regions of interest (ROI) were drawn over the organs displayed in ex vivo images (n = 3) and fluorescence signal intensity of the organs was calculated using the supplier's software.