In vivo three-dimensional evaluation of tumour hypoxia in nasopharyngeal carcinomas using FMT-CT and MSOT

All solvents and chemicals were obtained from commercial sources and used without further purification. CAIX targeting agent 1 was synthesised in the laboratory as previously described [29]. IRDye 800 N-hydroxysuccinimide ester (NHS) (Biotium Inc., LI-COR, Fremont, CA) and agent 1 were conjugated following a previously reported method [30]. Briefly, agent 1, IRDye 800 NHS, and trimethylamine (in a 1:1:6 M ratio) were mixed in dimethylformamide and stirred at room temperature for 2 h. After the solvent was removed under vacuum, the product was purified by high-performance liquid chromatography. CAIX-800 was purified using an Inertsil C18 Luna 46 × 150-mm column on a 1260 Infinity LC system (Agilent, Santa Clara, CA). Mass spectroscopy was used to characterise the conjugates of the probe. The CAIX-FITC used for the in vitro cell binding assay was prepared and tested in the same way as CAIX-800.

Two cell lines, i.e., 5-8F (a CAIX positive control [31]) and C666-1 (a CAIX negative control [32]) were provided by Southern Medical University. The cells were cultured at 37 °C and 5% CO₂ in Dulbecco’s modified Eagle’s medium supplemented with 10% foetal bovine serum and with penicillin and streptomycin (Gibco Invitrogen, Carlsbad, CA, USA). The cells were incubated on confocal plates (2 × 10⁵ cells/plate) for 24 h. After removal of the medium, CAIX-FITC and free FITC were incubated separately with the cells for 4 h at a final concentration of 10 nM. The cells were then washed with phosphate-buffered saline three times and fixed with 4% paraformaldehyde for 15 min at 37 °C. The cytoskeleton was first stained with rhodamine phalloidin for 30 min and the nucleus was stained with 15 μg/ml of DAPI (4, 6-diamidino-2-phenylindole) for another 8 min at room temperature. All images were acquired using a confocal laser scanning microscope (LSM-710, Carl Zeiss, Oberkochen, Germany). Imaging processing was performed using ZEN 2.3 lite (Zeiss, Germany). Fluorescence quantification was analyzed using ImageJ 2.X (LOCI, University of Wisconsin).

Four-week-old male BALB/c nude mice (Vital River Laboratory Animal Technology Co. Ltd., Beijing, China) were acclimated for 1 week before the study. The animals were kept in a specific pathogen-free unit. All surgical procedures were performed using a sterile hood. Two types of tumour models, i.e., subcutaneous and orthotopic NPC xenografts, were established. Briefly, 200 μL of phosphate-buffered saline (0.01 mol/L, pH 7.2) containing a suspension of 1.8 × 10⁶ 5-8F cells or 1.2 × 10⁷ C666-1 cells were injected subcutaneously into the lower left flank (n = 5 mice per group). The experiments started when the tumour diameter was 6–8 mm, an average tumour volume of approximately 126 mm³. Orthotopic cells were implanted as previously described [33]. Five mice per cohort were injected with 25 μl of 1.1 × 10⁵ 5-8F-fLuc cells into the nasopharynx using a 28-gauge needle. The animals were allowed to recover postoperatively for 30 min on a heated pad. Orthotopic implantation was confirmed by whole-body fluorescence imaging (PhotonIMAGER; Biospace Lab, Nesles-la-Vallée, France). All mice received a weekly intraperitoneal injection containing 3 mg of luciferin bioluminescent substrate (D-luciferin potassium salt, Perkin Elmer, Waltham, MA, USA) 4 min before imaging to monitor growth of the orthotopic tumour. Luciferase activity was quantified using region of interest analysis of the entire tumour.

The orthotopic tumours were located by 1.5-T MRI (M3TM, Aspect Imaging, Shoham, Israel) using a head and neck protocol that included axial and coronal T2-weighted imaging with the following parameters: repetition time 6000 ms, echo time 50 ms, slice thickness 0.7 mm, and slice spacing 0.2 mm. During acquisition of the images, all mice were anaesthetised using an isoflurane-oxygen gas mixture (500 ml/min, Matrx VMR Small Animal Anesthesia Machine, Matrx, USA).

At 24-h post-injection, the mice were euthanised by cervical dislocation. The heart, lungs, pancreas, spleen, liver, kidneys, small intestines, and tumours were then collected for ex vivo fluorescence imaging. The harvested tumours were immediately snap-frozen at − 80 °C in a deep freeze. A series of sequential 10-μm-thick frozen sections were then cut. An inverted microscope (Leica, Wetzlar, Germany) was used to confirm the difference in fluorescence signals between the different types of tumour. The same frozen sections were then used for immunohistochemistry. A standard immunohistochemical protocol was used to stain the tissue sections with anti-CAIX antibody (1:100 dilution, ProteinTech Group Inc., Rosemont, USA), as described elsewhere [34].

At 2-week post-implantation, the orthotopic mouse models (n = 5) were used for NPC-specific multimodality imaging. First, whole-body fluorescence imaging (PhotonIMAGER) was performed at different time points (2, 4, 6, 8, and 12 h) after intravenous injection of CAIX-800 with an excitation wavelength of 749 nm and a 776-nm filter. A video based on the fluorescence signal at the 8-h time point was reconstructed using a commercial M3 Vision system (PhotonIMAGER, Bio space Lab). MSOT was then performed in the same mice using an inVision 128 small animal imaging system (iThera Medical GmbH, Munich, Germany). The mice with the implanted 5-8F orthotopic tumours were scanned at pre (0-h), 4-h, 8-h, and 12-h intervals using transverse slices with a 0.5-mm step from the nasal portion to the cervical portion at wavelengths of 710, 730, 740, 760, 770, 780, 790, 800, 810, and 850 nm for each position. The mice were anaesthetised with 1.6% isoflurane inhalant delivered in 0.8 L of medical air and 0.1 L of oxygen. An adequate depth of anaesthesia was maintained during image acquisition with the mice oriented dorsal side up in the animal holder. The tumours were identified by a live-feed screen preview multispectral signal. Images were reconstructed using the multispectral processing along with ViewMSOT software (iThera Medical GmbH, Munich, Germany) as previously described [22]. The data were optimised using high-resolution (75 μm) back projection reconstruction.

To further investigate the above findings, the mice were scanned by micro-CT using a customised hybrid optical-CT imaging system to obtain the anatomical structure [35]. Anaesthesia was managed using the approach described earlier. Images were acquired with a 2 × 2 binning and 1-s exposure time setting on an EMCCD (electron multiplier charge-coupled device) with an emission filter of 810 nm (XBPA810, Asahi Spectra, Tokyo, Japan) when the lighting turned off. A photograph was then acquired using the EMCCD with an exposure time of 10 ms with the lighting turned on. The photographs helped to locate the mice and register the optical images with the CT volume. Next, 360 frames of CT projections were acquired in cone beam projection mode. The imaging settings were as follows: 120 μm thickness, 120 × 120 μm² base resolution, 40 kV, and 0.8 mA. Reconstruction and volume-rendered 3D images were obtained using 3D Med 4.6 software. Finally, the excised head and neck tissues were embedded in tissue freezing medium (Leica) and stored at − 80 °C. The tissue sections were cut at a thickness of 10 μm, fixed with 10% buffered formalin, and stained with haematoxylin-eosin.

At 6-week post-implantation, ipsilateral lesions suspicious for nodal metastases were detected in the five orthotopic NPC-bearing mice during monitoring of tumour growth. To investigate the feasibility of imaging NPC hypoxia at an advanced stage, planar FMI was performed to detect primary NPC and lesions suspicious for lymph node metastasis after intravenous injection of CAIX-800. Two-dimensional cross-sectional images were then obtained from the same mice using the above-mentioned MSOT inVision 128 small animal imaging system and protocol. The images were reconstructed by the ViewMSOT software. Maximum intensity projections were obtained after reconstruction. The head and neck tissues were excised for immunopathology.

The data are shown as the mean ± standard deviations. Statistically significant differences between groups were identified using independent two-samples t tests. Statistical analysis was conducted using GraphPad Prism 6 (GraphPad software, San Diego, California). P value < 0.05 was considered statistically significant.

CAIX targeting agent 1 was labelled using an IRDye 800 NHS ester (Supplementary Fig. 1). Conjugation was confirmed to be successful by mass spectroscopy (Supplementary Fig. 2). The purity was determined to be > 95% by high-performance liquid chromatography (Supplementary Fig. 3). CAIX-FITC was synthesised and characterised as reported elsewhere [29].

To validate the targeting efficacy of CAIX, 5-8F (CAIX-positive) and C666–1 (CAIX-negative) cells were fixed and stained with CAIX-FITC to visualise CAIX ligand-induced endocytic trafficking by confocal laser scanning microscopy (Fig. 1). Uptake of CAIX-FITC was much higher in 5-8F cells than in C666-1 cells, and abundant localization of CAIX-FITC was found in the cytomembrane and cytoplasm of 5-8F cells (Fig. 1a). In contrast, there was no apparent uptake of free FITC in either the 5-8F or C666-1 cells in the control groups (Fig. 1b), which confirmed the effective CAIX receptor targeting ability of CAIX-800 in vitro.

A biodistribution experiment was performed to evaluate the in vivo tumour uptake of CAIX-800 in the subcutaneous NPC mouse model. Four hours after the i.v. injection with equivalent CAIX-800, the 5-8F group started to show significantly higher probe accumulation and better optical contrast in tumours compared with the C666-1 group (P < 0.01 in 4-, 6-, 8-, 12-, and 24-h time points Fig. 2a, b). The highest optical signal in 5-8F tumours was observed 8 h after injection, but the maximum TBR of the 5-8F group reached 8.982 ± 0.85 at the 24-h time point, which was 10.6-fold higher than that of the C666-1 group (maximum TBR = 0.75 ± 0.38). Furthermore, the blocking 5-8F group that received free 1 before injection of CAIX-800 showed significantly less probe accumulation in tumours and a smaller TBR (P < 0.01 at 6, 8, 12, and 24 h) than the non-blocking 5-8F group; the control group did not show noticeable tumour uptake of IRDye 800 (Fig. 2a, b). The 3D volumetric optoacoustic imaging findings at 8-h time point were consistent with the planar FMI results (Fig. 2c). Uptake of CAIX-800 in the tumours was significantly higher in the 5-8F group than that in the C666-1 group at 8 h (P < 0.001). However, optoacoustic imaging did not identify any significant between-group difference at 4, 12, and 24 h (Fig. 2d).

Ex vivo fluorescence imaging of the organs and tumours resected from the 5-8F tumour-bearing mice revealed high probe accumulation in the kidney and tumour tissue (Fig. 2e), which is in agreement with an earlier report on the use of related radiotracers [29], because low-molecular-weight probes follow renal excretion. This also explained why there was a high contrast of the fluorescence signal in both tumour and kidney areas in the in vivo fluorescence imaging (Fig. 2a, white dotted circles and orange arrows). Near-infrared (NIR) fluorescence microscopy also confirmed that the CAIX-800 signal was stronger in 5-8F tumour specimens than in C666-1 tumour specimens (Fig. 2f). Furthermore, quantitative comparison of the fluorescence intensity in the resected organs and tumours at 24-h time point indicated a significant difference in probe uptake between the 5-8F and C666-1 tumours (P < 0.001; Fig. 2g), but there was no other significant difference in biodistribution of the probe. Finally, immunohistochemical staining revealed strong CAIX positivity in the membrane and cytoplasm of 5-8F tumours and negative expression of CAIX in C666-1 tumours.

The orthotopic NPC mouse model was constructed by injecting 5-8F-fLuc cells into the nasopharynx in nude mice (Fig. 3a, b). Growth of the NPC tumours at the orthotopic site was observed by whole-body bioluminescence imaging (PhotonIMAGER; Fig. 3c). A video based on bioluminescence signals showed the tumour location on 3D rendering (Supplementary Video 1), which was reconstructed using a commercial M3 Vision system. The mouse skeleton was simulated by the software without performing CT scanning. Quantitative analysis of tumour growth by weekly measurement of luciferase activity revealed an exponential increase in optical intensity during the following 5 weeks (Fig. 3d). The orthotopic NPC tumours were also confirmed on coronal T2-weighted MRI (Fig. 3e). These results confirmed that we had successfully established an orthotopic mouse model of NPC, lack of which used to be a major obstacle in preclinical research of NPC.

The mice with orthotopic NPC were injected with CAIX-800 and scanned using the multimodality imaging approach. Comparison of observations on reference whole-body bioluminescence imaging and on continuous FMI demonstrated that CAIX-800 selectively accumulated in the region of the NPC (Fig. 4a). However, due to the deeper depth of the orthotopic NPC, severe scattering of the fluorescence signal resulted in relatively low contrast between the tumour and the surrounding tissues. In contrast, the FMT reconstruction provided by M3 Vision at the 12-h time point allowed much better visualisation of the distribution of CAIX-800 (Supplementary Video 2), which confirmed that there was indeed selective retention of CAIX-800 in the nasopharynx region. To further validate our findings, we also performed continuous MSOT on the same mice (Fig. 4b). The results demonstrated that the pharmacokinetics of the probe in normal tissue were very different from those in tumour tissue. NPC showed much longer retention of this small optical probe, resulting in clearer tumour contrast from the 8 to 12-h time points. T2-weighted MRI confirmed that MSOT consistently located the tumours.

Given that 3D FMT using the commercial M3 Vision system only provides simulated data, we performed micro-CT and fluorescence imaging hybrid FMT-CT to quantitatively visualise the NPC [36]. The multi-angle NIR fluorescence images and CT images were acquired simultaneously so that the optical flux could be perfectly mapped on the CT data (Fig. 5a). The fluorescent source beneath the skin was reversely reconstructed in 3D (Fig. 5b) [37]. This approach demonstrated an acceptable combination of anatomical structure and the NPC area. FMT-CT showed even higher 3D contrast when compared with MSOT. These imaging results were validated pathologically in haematoxylin-eosin-stained cryosections of orthotopic tumours (Fig. 5c).

Six weeks after implantation of the tumours, orthotopic NPC-bearing mice (n = 5) spontaneously developed ipsilateral cervical lymph node metastases. Planar FMI showed high accumulation of CAIX-800 in the bulged region (Fig. 6a), but could not differentiate primary NPC from lymph node metastasis because of low spatial resolution caused by optical scattering. Unlike the fluorescence imaging approach, MSOT offered high spatial resolution for observing the distribution of CAIX-800 in 3D (Fig. 6b). Therefore, the primary tumours and lymph node metastases were able to be captured in the rendered 3D image and in different tomographic slices (Fig. 6c, d). Furthermore, the multispectral analysis overlaid the signal of HbO₂, Hb, and CAIX-800 with the background in MSOT images (Fig. 6d), which provided a comprehensive and quantifiable evaluation of the hypoxic distribution in the tumours. We found that both Hb and CAIX-800 were richly and heterogeneously distributed inside the metastatic lymph nodes whereas HbO₂ was only concentrated in the peripheral area. Finally, the histological analysis showed that the lymph nodes were diffusely infiltrated by a population of large atypical cells, confirming metastatic NPC, and the immunohistochemical analysis revealed strong expression of CAIX in areas of metastasis (Fig. 6e).