Multiparametric chemical exchange saturation transfer MRI detects metabolic changes in breast cancer following immunotherapy

Background With metabolic alterations of the tumor microenvironment (TME) contributing to cancer progression, metastatic spread and response to targeted therapies, non-invasive and repetitive imaging of tumor metabolism is of major importance. The purpose of this study was to investigate whether multiparametric chemical exchange saturation transfer magnetic resonance imaging (CEST-MRI) allows to detect differences in the metabolic profiles of the TME in murine breast cancer models with divergent degrees of malignancy and to assess their response to immunotherapy. Methods Tumor characteristics of highly malignant 4T1 and low malignant 67NR murine breast cancer models were investigated, and their changes during tumor progression and immune checkpoint inhibitor (ICI) treatment were evaluated. For simultaneous analysis of different metabolites, multiparametric CEST-MRI with calculation of asymmetric magnetization transfer ratio (MTRasym) at 1.2 to 2.0 ppm for glucose-weighted, 2.0 ppm for creatine-weighted and 3.2 to 3.6 ppm for amide proton transfer- (APT-) weighted CEST contrast was conducted. Ex vivo validation of MRI results was achieved by 1H nuclear magnetic resonance spectroscopy, matrix-assisted laser desorption/ionization mass spectrometry imaging with laser postionization and immunohistochemistry. Results During tumor progression, the two tumor models showed divergent trends for all examined CEST contrasts: While glucose- and APT-weighted CEST contrast decreased and creatine-weighted CEST contrast increased over time in the 4T1 model, 67NR tumors exhibited increased glucose- and APT-weighted CEST contrast during disease progression, accompanied by decreased creatine-weighted CEST contrast. Already three days after treatment initiation, CEST contrasts captured response to ICI therapy in both tumor models. Conclusion Multiparametric CEST-MRI enables non-invasive assessment of metabolic signatures of the TME, allowing both for estimation of the degree of tumor malignancy and for assessment of early response to immune checkpoint inhibition. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-023-04451-6.

Suppl.Table 1 Descriptive statistics of in vitro cell extracts.
Suppl.Table 2 Analysis of statistical significance of in vitro cell extracts.
Suppl.Table 3 Descriptive statistics of longitudinal in vivo study.

Suppl. Table 4
Analysis of statistical significance of longitudinal in vivo study.

Suppl. Table 5
Descriptive statistics of in vivo ICI therapy.

Suppl. Table 6
Analysis of statistical significance of in vivo ICI therapy Suppl.Method 1: Cell culture conditions.

67NR and T-cells.
For MR imaging of cell extracts, water-soluble metabolites of either 4T1, 67NR or T-cells were prepared using a dual-phase extraction method.Cells were cultured as monolayers until 80% confluency.After washing the cells twice with PBS, 4 mL ice-cold methanol were added.The cells were harvested and 4 mL chloroform and 4 mL water were added, leading to a final chloroform:methanol:water ratio of 1:1:1 (v/v/v).To separate the methanol-water phase containing the water-soluble cellular metabolites from the chloroform phase containing the cellular lipids, the samples were centrifuged for 5 minutes at 5000 rpm and the upper methanolwater phase (8 mL) was carefully removed.Methanol was eliminated from the methanol-water sample using a rotary evaporator, followed by lyophilization.For MR imaging, the lyophilized sample was dissolved in 2 mL PBS to ensure a physiological pH of 7.2 ± 0.1.

Suppl. Method 3: Preparation of tumors for ex vivo analysis.
After MR imaging, tumor were removed for further ex vivo analysis.For MALDI-2-MSI and immunohistochemistry, tumors were cryo-sectioned in 16 µm tissue slices using a rotary cryomicrotome (Leica Microsystems, Nussloch, Germany).For 1 H-NMR spectroscopy of tumor metabolites, extracts of snap-frozen tumors were prepared with the dual-phase extraction method as described above.Tumors were homogenized in a chloroform:methanol:water mixture (2:2:1.8)using a tissue homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France).After centrifugation (-20°C, 5000 rpm, 5 min), the upper methanol-water phase was separated and methanol was removed using a rotary evaporator, followed by lyophilization.
NMR samples were dissolved in 1.0 mL D2O (99.98%,Sigma-Aldrich) containing 1 mM maleic acid.Samples were transferred to a 5 mm glass tube for 1 H-NMR analysis.Spectra were recorded at 600 MHz using an Agilent DD2 600 spectrometer (Agilent Technologies, Santa Clara, California, USA). 1 H-NMR chemical shifts are given relative to tetramethylsilane and are referenced to the solvent signal (HDO, d = 4.66 ppm).The data were recorded using the manufacturer's software and processed with MestReNova (version 14.2.0,Mestrelab Research, Santiago de Compostela, Spain).An internal standard of 1 mM maleic acid was used for quantitative analysis (singlet occurring at a chemical shift between 6.2 and 6.4 ppm).
Concentrations of glucose were quantified from signals at 3.4 ppm, creatine at 2.95 ppm and amide protons as integrals from 6.8 to 9 ppm.For analysis of intratumoral metabolites, concentrations were normalized to the tumor volume, assessed by T2-weighted imaging.

Suppl. Method 5: Immunohistochemical co-staining of GLUT1, GLUT3 and CD3.
After fixation and rehydration, tumor sections were incubated in blocking buffer (2% horse serum and 0.1% fetal calf serum in phosphate-buffered saline, PBS) for 60 minutes at room temperature.Subsequently, the sections were incubated in a humidified chamber overnight at 4°C with primary antibodies against GLUT1 (sheep, 112AP, FabGennix, Frisco, Dallas, USA), GLUT3 (rabbit, 20403-I-AP, proteintech, Manchester, England) and CD3 (allophycocyaninconjugated, rat, 17A2, BioLegend, San Diego, California, USA), each at a dilution of 1:150 in blocking buffer.After washing the slides three times with blocking buffer, they were incubated for 90 minutes at room temperature with anti-sheep Alexa Fluor 488 (A-11015, Thermo Fisher Scientific, Waltham, Massachusetts, USA) and anti-rabbit Alexa Fluor 568 (A-10042, Thermo Fisher Scientific), each at a dilution of 1:500.Finally, the sections were washed with PBS, stained with 20 μM Hoechst 33342 dye (62249, Thermo Fisher Scientific) for 10 minutes at room temperature, followed by one last washing step with PBS.Microscopy was conducted with an LSM 800 (Carl Zeiss Microscopy, Oberkochen, Germany) and the software ZEN 2.6 (blue edition; Carl Zeiss Microscopy).

Suppl. Method 6: Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-2-MSI).
Conventional mass spectrometry imaging approaches for the quantitative analysis of glucose are hampered by the presence of (near-) isomeric molecules such as inositol.Here, we therefore used a tandem-MS based approach in negative ion mode that utilizes characteristic ring cleavages to identify glucose unambiguously.To achieve signal intensity for deprotonated glucose ions sufficient for tandem-MS directly from tissue we used MALDI-2-MSI and introduce a novel split pixel approach to acquire tandem-MS spectra for glucose and the respective isotopically labeled internal standard in a spatially resolved fashion.
Tissue sections were dried in an evacuated desiccator, vacuum-sealed and stored at -80 °C.At the time of measurement, slides were brought to room temperature in an evacuated desiccator and coated with norharmane matrix for MALDI-2 analysis using a pneumatic spray robot (SunCollect MALDI Sprayer, SunChrome, Germany) with parameters as follows: 3,5 mg/mL norharmane solution in 50:50 (% v/v) acetonitrile:water; total of 20 passages, flow rate: 1 st passage 15 µL/min; 2 nd passage 20 µL/min; 3 rd passage 30 µL/min; all following: 50 µL/min; scan speed: 600 mm/min, line distance 2 mm.Matrix solution was spiked with 4 mg/mL 13 C6glucose as internal standard.MALDI-2-MS imaging analysis was conducted on a timsTOF fleX MALDI2 (Bruker Daltonics, Bremen, Germany) in the negative ion mode with microgrid enabled.Acquisition parameters were as follows: Spot size: 14 x 29 µm², step size (x and y) 30 µm, shots per pixel: 150; m/z range: 50-500; trigger delay for MALDI-2: 10 µs.Two MALDI-2-MS images were collected from the same region of interest (ROI) of each section with step size of 30 µm in x-and y-direction.For each individual image, a pristine area of 14 x 29 µm² within a 30 x 30 µm² pixel was probed by shifting the ROI by 15 µm in x-direction between measurements.Both runs were collected in tandem-MS mode selecting the de-protonated ion species of glucose at m/z 179 ( 12 C6-glucose) and 185 ( 13 C6-glucose) as precursors, respectively using a selection window of 1 Da.Precursor ions were fragmented using 8 eV of activation energy in low-energy collision induced dissociation (CID).To compare relative intensities between labeled and unlabeled glucose, a characteristic ring cleavage yielding fragment ions at 89.02 for 12 C6-glucose and 92.05 for 13 C6-glucose was exploited for unambiguous identification.Images of the respective fragment ions were constructed and analyzed using SCiLS Lab MVS (SCiLS, Bremen, Germany).Quantitative MS images were generated using the SCiLs lab API and a script, written in Python.For this, fragment signal intensity of the unlabeled glucose was normalized using the fragment signal intensity of the internal standard of the same split pixel and multiplied with the respective amount of internal standard applied per unit area (80.3 pmol/mm²).Because the internal standard is not subject to analyte extraction, the presented quantitative values represent a minimum value for 12 C6-glucose content in the respective sample.After analysis, matrix was washed off using ethanol.Subsequently, Hematoxylin and eosin (H&E) staining of tumor sections was performed according to standard protocols.The ROIs for further analysis of intratumoral glucose were identified based on bright field microscopy of the H&E-stained samples.Average signal intensity for the ROI of the respective characteristic fragments was generated in SCiLS from the respective tandem-MS data and average quantitative data was generated as described above.

Suppl. Method 7: Dynamic contrast-enhanced MRI (DCE-MRI).
After CEST imaging, dynamic contrast-enhanced (DCE) MRI using Magnevist (Gd-DTPA, 0.3 mmol/kg) was performed to exclude that glucose-weighted CEST results are dominated by perfusion effects.The contrast agent was injected via a tail vein catheter (Klinika Medical GmbH, Usingen, Germany), using a perfusion pump (World Precision Instruments, Sarasota, Exemplary comparison of intratumoral 12 C glucose concentrations assessed by MALDI-2-MSI.

Florida
, USA) at a rate of 240 µL/min.The injection was initiated one minute after starting a fast low-angle shot (FLASH) scan (TR = 24.6 ms, TE = 1.5 ms, 15° flip angle, 1 average, 610 repetitions, 18 x 15 mm 2 FOV, 96 x 96 matrix, acquisition time = 20:00 min:s).Dynamic assessment of contrast enhancement after Magnevist injection was used to derive the tumor perfusion parameter volume transfer constant Ktrans.Calculation of Ktrans was performed with an in-house developed software based on the PkModeling extension for 3D Slicer (https://github.com/millerjv/PkModeling),using a three-parameter Tofts model (extended Tofts model)[2] and a population-based arterial input function[3].

Table 1 : Descriptive statistics of in vitro cell extracts.
All values are presented as mean ± standard deviation.

Table 2 : Analysis of statistical significance of in vitro cell extracts. Suppl. Table 3: Descriptive statistics of longitudinal in vivo study.
All values are presented as mean ± standard deviation.

Table 4 : Analysis of statistical significance of longitudinal in vivo study. Suppl. Table 5: Descriptive statistics of in vivo ICI therapy.
All values are presented as mean ± standard deviation.