The tri-modality imaging system has a modular design. The PET, CT, and FMI modules were initially developed separately and then integrated into the same gantry. The system includes five major components: gantry, CT, PET detector ring, FMI module, and the animal bed ( Figure 1 ).

The CT scanning geometry uses a rotated source-detector pair when imaging living animals. It consists of a flat-panel CMOS detector (1512, DEXELA Inc., USA) and an X-ray tube (SB-80-250, Source-ray Inc., USA) with a 33 μm focal spot. The X-ray tube and the CMOS detector are mounted on a high precision translation stage which is mounted on the rotatable gantry. The PET module is made up of 22 detector modules arranged in a ring-like geometry with an interior diameter of 100 mm and an axial FOV of 32 mm. The FMI module included two common fluorescence imaging modes: fluorescence molecular tomography (FMT, 3D whole-body imaging of mice) and fluorescence reflectance imaging (FRI, 2D imaging of subcutaneous tissues). Both imaging modes rely on a tungsten-halogen lamp (71LT250/71PT250A, Beijing 7-Star, China) as an illumination source, while an sCMOS (ORCA-Flash 4.0, Hamamatsu, Japan) is used as the detection device. The whole setup is mounted on a surface plate which is also connected to the gantry module.

FMI and CT imaging modules share one rotation gantry. In addition, all three modalities (PET/CT/FMI) images were scanned on the same animal bed. The animal bed was mounted on a high precision motor-controlled translation stage which carries the imaging objects along with the different imaging modalities. Multi-modal image fusion was precisely done by co-registered positional data overlay. An anesthetics vaporizer and thermostatic equipment were reserved during the experiment for animal safety reasons.

For system testing and calibration purposes, a phantom was constructed to evaluate all three imaging modalities. The solid, cylindrical phantom with an outer diameter of 30 mm and a length of 70 mm (made of polyformaldehyde) contained two 2 mm diameter holes of 35 mm depth. The center-to-center distance between the two holes was 10 mm. The phantom had an absorption coefficient of µa = 0.02 cm^-1 and a reduced scattering coefficient of µs = 10.0 cm^-1, which perfectly mimics tissue properties. The phantom images were used to calibrate the co-registration of the images obtained from the tri-modality system. For this, prior to the experiment, two capillary tubes were separately filled with indocyanine green (ICG, Merck Sigma-Aldrich, 10μL with 10 µM) and fluorodeoxyglucose (FDG, 50µL with 70µCi), were inserted into the holes of the phantom. The reconstructed CT, PET, and fluorescence images were used for registration.

PET images were first acquired over a time period of approximately 5 min using energy and coincidence windows of 350 – 650 keV and 12 ns, respectively. Subsequently, the phantom was moved to the field of view of the CT module. A CT acquisition was performed by employing the following parameters: X-ray voltage: 80 kVp, anode current: 450 μA, and exposure time of each of the 360 rotational steps was 50 ms. Finally, the fluorescence images were collected from 36 projections evenly along with the 360° rotation, and the integration time of the sCMOS camera was 4 s at each projection angle.

PET data were reconstructed after attenuation correction by using ordered-subsets expectation maximization (OSEM). A CT data were reconstructed by using a filtered back-projection algorithm. FMT data were reconstructed by the method of L1 norm regularization with modified FOCUSS (L1-FOCUSS). The obtained PET and FMT images were registered with the CT images, respectively.

ML-10 (2-(5-fluoro-pentyl)-2-methyl-malonic acid, MW=206 Da) is the prototypical member of a novel family of small-molecule detectors of apoptosis. The fabrication process of ¹⁸F-ML-10 was performed according to a previously described approach (22, 23).

¹⁸F-ML-10 was synthesized from its respective precursor, ML-10-mesylate, by a nucleophilic substitution reaction. For radiolabeling, ¹⁸F-fluoride delivered from a cyclotron was trapped on a quaternary ammonium cartridge, and excess 18H₂O was removed. 18F was eluted with Kryptofix222 (Sigma-Aldrich) and azeotropic drying at 90°C under argon with acetonitrile resulting in dry ¹⁸F-fluoride. The precursor ML-10 mesylate in anhydrous acetonitrile was added for a 15-min reaction at 90°C. After dilution with acetonitrile, a silica Sep-Pak (Waters) was used to separate t-butyl–protected ¹⁸F-ML-10. Hydrolysis of the protecting groups was accomplished by treatment with trifluoroacetic acid (TFA):H₂O (9:1) at room temperature for 15 min. Dilution with deionized water followed by semipreparative reversed-phase high-performance liquid chromatography (HPLC) (mobile phase of water:acetonitrile: TFA, 80:20:1 [v/v/v]; octyldecyl silane column [Phenomenex]; flow rate of 10 mL/min) provided purified ¹⁸F-ML-10. The product eluted at a retention time of 15 min. Organic solvents were removed after dilution with deionized water, and ¹⁸F-ML-10 was trapped on a Sep-Pak C18 column (Waters) under elution with absolute ethanol. The HPLC product fraction containing ¹⁸F-ML-10 was mixed with deionized water containing 1% acetic acid (100 mL), and the mixture was passed over a Sep-Pak C18 column. Flushing with absolute ethanol (1 mL) yielded the final product. The radiochemical yield of ¹⁸F-ML-10 was 40%255% at the end of synthesis (EOS), with a collected product yield of 30%240% after purification steps (EOS). The radiochemical purity was greater than 99%, and the specific activity was greater than 40.7 GBq/mmol (EOS). The overall time required for radiolabeling was about 75 min.

11.1 MBq (215 µL with 0.3 mCi) of ¹⁸F-ML-10 was injected through the tail vein to investigate the tumor apoptotic area. The specific activity of ¹⁸F-ML-10 is 51.6 MBq/g.

Thirty specific pathogens free (SPF) grade BALB/c-nu/nu nude mice (6-8 weeks, weighing 18-22 g, male) were housed in an environment with a temperature of 22 ± 1 °C, relative humidity of 50 ± 1%, and a light/dark cycle of 12/12 hr, and were given water and food ad libitum. All animal studies (including the mice euthanasia procedure) were done in compliance with the regulations and guidelines of Peking University institutional animal care and conducted according to the AAALAC and the IACUC guidelines. Body weight was monitored and recorded weekly.

Thirty mice were used for imaging and Western Blot (WB). Among them, half of them (15) was assigned to the control group and 15 to the experimental group. The procedure of animal imaging consisted of the following steps: probe injection, probe uptake time, anesthesia procedure, and multi-modality image collection, as shown in Figure 2 . The imaging process has been optimized in order to minimize the total experiment time.

In this study, the tri-modality imaging system was used to investigate the therapeutic effect (apoptotic and necrotic tumor induction) of a new recombinant bacterial strain on mouse head and neck xenografts in vivo. Briefly, two imaging probes were used for this purpose: firstly, in order to be able to investigate the necrotic tumor area, an ICG solution (8.0 mg/kg) was injected into the tail vein 24 h prior to the experiment. 8 mg/kg is a dose that takes into account both animal safety and fluorescence signal-to-noise ratio. Secondly, 11.1 MBq (0.3 mCi) of ¹⁸F-ML-10 was injected into the tail vein to detect the tumor apoptotic area. The timeline for the multi-modality imaging experiment was as follows ( Figure 2 ): mice were abdominally injected with 2.5% avertin anesthetic (0.1ml/5g) and placed on the animal bed in a dorsal position. After 50 min, the fluorescence images were acquired with an exposure time of 1 s using appropriate filters (ex. filter with 769 nm, em. filter with 832 nm cut-off, Semrock). The raw data were acquired under the control of a baseline clamp and processed with the help of Andor SOLIS software. Immediately after fluorescence image recording, whole-body PET/CT images were acquired.

The recombinant Escherichia coli–Bifidobacterium shuttle vectors pLW5 and pLW9, which include IL24 (835 bp) and green fluorescent protein (GFP, 930bp) fragments, respectively, were used to obtain B. breve-IL24 (pLW5-breve) and B. breve-GFP (pLW9- breve) transformant. B.breve-IL-24 transformants were obtained through electrotransformation. Transformed strains expressing IL-24 were analyzed by RT-PCR before starting the experiment, as described previously (18). Briefly, he WSU-HN6 cell line was used as reagents, which was authenticated by short tandem repeat profiling (STR) and free of mycoplasma contamination. ( Supplementary Figure S3 ). 1x10⁶ WSU-HN6 cells were then subcutaneously injected into the mouse back (in 30 mice). When the tumor tissues reached 2-3 mm in diameter, mice were randomly assigned to a treatment or control group (10 mice in each group). The treatment group received 200 µL solution containing 1x10⁸ B.breve-IL-24 bacteria for the treatment group, while a 200 µL saline solution was given to the control group; injections were given by the tail vein twice per week for a total of 2 weeks. Tumor size was measured twice a week post-injection by using a standard caliper tumor volume calculation method: V_tumor= (l x w²)/2. After treatments, all mice were imaged according to the in vivo imaging protocol mentioned above (see section” Animal imaging”). Then, the mice were euthanized, and tumor tissues were dissected and analyzed by Western blot and a histopathology assay. Detailed materials and methods for Western blotting and the histopathology assay are described in the “Electronic Supplementary Material” section. Cell necrosis and apopsos could be observed by HE staining. To be specific, volume expansion and deformation of groups of cells or some tissues can be observed. Cell membrane was observed damaged. Chromatin does not aggregate and appears flocculent. Organelles are enlarged or broken. The nuclei of apoptotic cells were pyknotic and fragmented, showing blue-black color, and their cytoplasm was light red. Normal cell nuclei are uniformly blue. While necrotic nuclei appear very pale blue or disappear.

Tumor size and real-time PCR results were expressed as mean ± standard deviation (SD). All calculations and analyses were performed using SPSS 19.0 software (SPSS Inc., Chicago, IL, USA). A P value <0.05 was considered as being statistically significantly different. All continuous variables were first tested for normality, and then the mean difference was tested according to the group t-test. The rank-sum test was performed when continuous variables did not pass the normality test. Categorical variables were tested by using the chi-square test. A p value <0.05 was considered to be statistically significant for all statistical methods.

Reconstructed PET, FMT, CT, and tri-modality fusion images of the cylindrical phantom are shown in Figure 3 . Three different color maps are used for the representation of the three different data sets in Figure 3A . The integrated tri-modality three-dimensional images are shown in Figure 3B . The ICG solution within the left capillary tube is displayed in green. The ¹⁸F-FDG within the right capillary tube is shown by a hot spot representation. Taken together, our results indicate that our multi-modality imaging system is able to provide fused images with a misalignment of less than 1 mm).

After bacteria treatment, Balb/c nude mice bearing a subcutaneous WSU-HN6 xenograft were subjected to multi-modality imaging in order to examine the apoptotic and necrotic induction within the tumor, as well as to evaluate tumor growth compared to the control group in vivo. Figure 4A shows transverse, coronal, and sagittal slices of the small-animal PET/CT, as well as the X-ray image of the ear, bone, and tumor tissues in the lower, left rib, and several nodes of a single mouse.

To obtain better subcutaneous fluorescence data, the FRI mode was used in the animal study. Each rotational angle of the acquisition was recorded so that the FRI image could be successfully projected and merged with the CT images. The results indicated cell death in the center part of the xenograft WSU-HN6 ( Figures 4B, C ). The additional ICG signal, detected outside the tumor, could have been caused by the tumor or by bacteria obstructing a skin vessel.After the administration of the bacteria, apoptosis induction was examined in real-time by using an ¹⁸F-ML-10 tracer. The merged real-time images showing the apoptotic and necrotic tumor areas, including the whole-body animal structure, are shown in Figure 4D .

To further evaluate the antitumor effect of B.breve-IL-24 in vivo, a xenograft mouse model was established. When tumors reached 8 to 10 mm in diameter, B.breve-IL-24 or saline (control) was injected. Results showed higher tumors grew in the control group compared to B.breve-IL-24-treated mice, which suggests that the B.breve-IL-24 strain can inhibit tumor growth over time ( Figure 5A ). Tumor size was measured to monitor the growth inhibition efficiency in vivo. Saline group was used as blank control and B. breve-GFP was used as negative control. A higher tumor growth inhibition was observed in B. breve-IL24 compared with B. breve-GFP group and higher tumor growth inhibition in B. breve-GFP compared with the Saline group. Although no significance between B. breve-GFP and B. breve-IL24 group were observed. Nevertheless, the average tumor size (including slower tumor growth rate) in B. breve-IL24 was smaller compared with the B. breve-GFP group.

Interleukin-24 (IL-24) is a novel tumor suppressor cytokine that selectively induces apoptosis in a wide variety of tumor types. To further substantiate an activation of the apoptosis pathway, we analyzed the effect of B.breve-IL-24 on caspase-3 activation in tumor tissues. We found that B.breve-IL-24 induced activation of caspase-3 and Bim in tumor tissues, the results could be found in our previous study (18).

The PET/CT scanning results of the tumors treated with B.breve-IL-24 showed a significantly higher apoptosis induction as detected by the higher radioactivity (135.35 µCi) compared to the control group (20.09 µCi) in vivo ( Figure 5B ). These results were confirmed by Western blotting of pro- and anti-apoptotic proteins. We could show that Bim and cleaved caspase-3 were highly expressed by B.breve-IL-24, while low Bim expression was detected in this strain, indicating a low Bcl-2 expression (18).

Finally, the necrotic tumor area was analyzed by bright field microscopy after histopathologic tissue staining. In contrast to the apoptosis induction results, a larger necrotic tumor area was observed in the control group ( Figures 5C–E ). The reason for the higher necrosis in the control group may be explained by the faster growth of the tumor compared to the treatment group.

In summary, our multi-modality system provides more comprehensive information about the tumor behavior and the therapeutic effect in vivo than single modality approaches.