For colorectal, breast and head-and-neck cancer, we selected two cell lines with different tumour characteristics. Luciferase transfected cells were used to follow orthotopic tumour growth by bioluminescence imaging (BLI). MCF-7 and OSC-19 cells were transfected with Luciferase 2 and green fluorescent protein as described previously . All cell lines were grown in a humidified incubator at 37 °C and 5% CO2. Cells were cultured for not more than 10 passages and regularly checked for Mycoplasma infection by PCR.
EpCAM expression of HT29(−/+)luc2, COLO320, OSC-19-luc2-cGFP, FaDu-luc2, MCF-7-luc2-cGFP and MDA-MB-231 cells was evaluated by flow cytometry. Cells were cultured until 90% confluence and detached with trypsin. Viability of the cells was evaluated using trypan blue. After adjusting the number of cells to 0.5 × 106 per tube in ice cold phosphate-buffered saline (PBS), they were incubated with 0.4 μg/ml 323/A3 anti-EpCAM antibody or isotype control MOPC21 for 30 min on ice. Then cells were washed three times in ice cold PBS and incubated with a goat anti-mouse IgG1-AF488 secondary antibody (Invitrogen, 2.5 μg/ml). The cells were washed three times in ice cold PBS and resuspended in 400 μL PBS containing propidium iodide to exclude dead cells from the analysis. Flow cytometry was performed using the LSRII (BD Biosciences). The experiments were performed in duplicate and EpCAM expression was estimated as the geometric mean of fluorescence intensity measured in 10,000 viable cells. For quantitative determination of EpCAM numbers per cell type the Qifikit (Dako) was used.
Antibodies and conjugation to IRDye 800CW
EpCAM specific monoclonal chimeric antibody 323/A3 and the IgG1k isotype control monoclonal antibody MOPC21 (BioXcell, West Lebanon, USA) were used . Antibody 323/A3 has a medium high affinity (Kα = 2 × 109 M−1) for EpCAM and is directed against the EGF-like domain I epitope on the extracellular domain of the EpCAM molecule, whereas MOPC21 has an unknown specificity after testing on human and rodent tissues [33–35]. Both antibodies were covalently conjugated to NIR fluorochrome IRDye 800CW (LI-COR, Lincoln, NE, USA). λex = 773 nm, λem = 792 nm) using N-hydroxysuccinimide ester chemistry as indicated by the manufacturer. Removal of unconjugated fluorophore was accomplished by using two Zeba Spin Desalting columns (Thermo Fisher Scientific, Perbio Science Nederland B.B., Etten-Leur, The Netherlands) per protein in two sequential steps. For comparison experiments, the two conjugates i.e. the EpCAM specific (323/A3-800CW) and control (MOPC21-800CW) were complemented by the chemically inactive carboxylate version of IRDye 800CW, representing the fluorescent label without antibody control.
The stability of 323/A3-800CW in human serum was evaluated using HPLC (Biosep-SEC-s2000, Phenomenex, USA). Serum and sodium azide dilution were filtrated through a 0.22 μm filter in a 15 ml tube. A 24-wells plate (Greiner Bio-one, Germany) was prepared with 0.02% sodium azide and serum/probe in a ratio of 1:1 and PBS as control and incubated at 37 °C under 5% CO2. At 4, 24, 48 and 96 h 20 μl of sample, diluted in 40 μL PBS was evaluated using HPLC in PBS at a flow rate of 0,5 ml/min for 60 min, detected at 2 channels, 280 and 780 nm.
Cell binding study
A cell binding assay was performed to confirm the EpCAM specificity of 323/A3-800CW. HT29-luc2 (40,000 cells), COLO320 (40,000 cells), OSC-19-luc2-cGFP (25,000 cells), FaDu-luc2 (35,000 cells), MCF-7-luc2-cGFP (40,000 cells) and MDA-MB-231 cells (40,000 cells) cells were seeded in a black 96-well plate (Greiner Bio-one, Germany). At ±90% confluence the cells were washed twice with PBS. 323/A3-800CW and MOPC21-800CW were added in a concentration range of 0–8 μg/ml and incubated for 1 h at 37 °C. After incubation, the cells were washed twice with culture medium without supplements. Bound antibody was imaged with an Odyssey scanner (LI-COR), scanning at the 800 nm channel. To correct the fluorescence signal for the number of tumour cells per well a cell nucleus staining was performed: The cells were fixed/permeabilized with acetone/methanol for 10 min, washed with PBS, and incubated with TO-PRO-3 (Invitrogen) at 1:1000 for 5 min at room temperature. After washing twice with PBS, the plate was imaged with the Odyssey scanner at the 700 nm channel to detect TO-PRO-3 fluorescence. The ratio of the 800 and 700 nm fluorescence was plotted. The experiments were performed in triplicate.
Nude Balb/c female mice (Charles River laboratories, l’Arbresle, France), aged 4–6 weeks, were housed in individually ventilated cages and provided with food and sterilized water ad libitum. Their general health state was monitored by weight measurements throughout the experiments. Tumour growth was monitored longitudinally by visual inspection of the tumours, caliper measurements and/or by bioluminescence imaging. Bioluminescence imaging was performed by intraperitoneal injecting of 150 mg/kg of D-luciferin solution (SynChem, Inc, Elk Grove Village, IL) in PBS, in a total volume of 50 μL. After 10 min, mice were imaged with the IVIS Spectrum imaging system (PerkinElmer, Waltham, MA, USA). Imaging procedures were performed under isoflurane gas anaesthesia
Colon cancer models
To induce colon tumours, mice were subcutaneously injected at four sites with 5 × 105 HT29 cells in 40 μL RPMI1640 medium. Tumour growth was followed by calliper measurements and after 10 days, when the tumours reached a volume of approximately 75 mm3, imaging experiments started. Orthotopic HT29-luc2 tumours were induced as described previously .
To induce orthotopic breast tumours, 2.5 × 105 MCF-7-luc2-cGFP cells were inoculated in two contralateral mammary fat pads. Oestrogen pellets (17β-oestradiol, 0.36 mg/pellet, 60 day release) were implanted subcutaneously. Tumour growth was followed by visual inspection and bioluminescence measurements as described above.
Breast carcinoma model
Orthotopic tongue tumours were induced in the tip of the tongue through a submucosal injection of 4 × 104 OSC-19-luc2-cGFP cells. When tumours were visible and bioluminescence signal ranged between 5 × 109 and 1 × 1010 relative light units (RLU) imaging experiments started.
Multiple small MCF-luc2-cGFP tumours were induced in the peritoneum by intraperitoneal injection of 2.5 × 105 MCF-luc2-cGFP cells. Tumour growth was followed twice a week by bioluminescence as described above. Imaging experiments initiated when multiple tumour nodules were formed of various sizes.
NIR fluorescence imaging systems
Real-time NIR fluorescence imaging and operative resection of the tumours was performed using the next generation Artemis imaging system (Quest Medical Imaging, Middenmeer, the Netherlands). An earlier iteration of this system was extensively validated . This system has a freely moveable handheld camera for simultaneous acquisition of visible light and NIR fluorescence. Since the system has been improved such that the camera can be fixed in a stable position with an arm, while both camera and arm are covered with a sterile drape. The illumination efficiency and homogeneity has been improved with a ring containing eight hemispheric illumination lenses centred around a wide field imaging lens for open surgery. Illumination is provided by four visible light sources with peaks centred in the blue, cyan, green and red and a NIR laser with a peak at 785 nm for fluorescence excitation. Reflected excitation light is blocked by a 750–800 nm notch filter. Captured visible and NIR light is split using a prism containing a dichroic coating (<785 mm). Visible light additionally passes through a low-pass filter (<640 nm) and the NIR emission light is filtered with a high pass filter (>808 nm). Exposure times and sensor gains are separately adjustable for both imaging channels. The visible light channel, the NIR fluorescence channel and an adjustable overlay of both channels are simultaneously presented during the procedures.
Next to the Artemis imaging system, the Pearl Impulse small animal imaging system (LI-COR) was used as a preclinical reference to visualize tumours and calculate the tumour-to-background ratios (TBRs). Data from the Artemis and Pearl imaging systems were analysed using imageJ (W. Rasband, Bethesda, Maryland) and the Pearl Cam Software, respectively.
NIR fluorescence measurements
Subcutaneous HT29 colon tumours were used to confirm in vivo EpCAM specificity of 323/A3-800CW and to measure fluorescence over time. When the subcutaneous HT29 colon tumours were 36 ± 6 mm2, 1 nmol (≈150 μg) of 323/A3-800CW (n = 3), 1 nmol (≈150 μg) MOPC-800CW (n = 3) or 1 nmol (≈1.1 μg) of 800CW carboxylate was injected intravenously. NIR fluorescence signals were measured at 0, 4, 24, 48, 72 and 96 h after injection using the PEARL small animal imaging system and the intraoperative Artemis imaging system after which TBRs were calculated.
After in vivo confirmation of the EpCAM specificity and establishment of the optimal time frame for imaging with the subcutaneous model, the clinically more relevant orthotopic MCF-7-luc2-cGFP breast, OSC-19-luc2-cGFP tongue and HT29-luc2 colon tumours were evaluated. Hence, 1 nmol of 323/A3-800CW (n = 3) or 1 nmol MOPC21-800CW (n = 3) was intravenously injected in each group. Fluorescence imaging of mice bearing orthotopic tumours was performed 72 h after administration for optimal TBR as determined in the subcutaneous HT29 colon carcinoma model. TBRs of orthotopic tumours were measured and tumours were resected under NIR fluorescence guidance using the Artemis imaging system. Ex vivo, fluorescent measurements of resected tissue were performed on a back table. Tumours were sliced and fluorescence measurements were performed on the sections to evaluate the distribution of the probe. Resected tumours from the head-and-neck cancer and breast cancer models were assessed by BLI imaging and GFP fluorescence imaging (OSC-19-luc2-cGFP and MCF-7-luc2-cGFP, IVIS spectrum).
The MCF-luc2-cGFP peritonitis carcinomatosa tumour model was used to determine the minimal tumour sizes that could be detected by intra-operative fluorescence imaging using the EpCAM specific antibody 323/A3-800CW in combination with 2 imaging systems. Because an enhanced permeability and retention (EPR) effect in these micrometases is not expected as indicated recently by Hall et al.  no MOPC21-800CW control was used in this model. Mice were anesthetized 72 h after intravenous injection of 323/A3-800CW (1 nmol, n = 3), as described above and fluorescence imaging using the Pearl and Artemis imaging system was performed. A midline abdominal incision was made and the abdominal skin was removed. Fluorescence imaging of the mice with both imaging systems was performed followed by resection of the peritoneum and again fluorescence images were taken. Fluorescence imaging of the abdominal area was performed to search for residual intraperitoneal tumour nodules. Presence of tumour nodules and tumour specific NIR fluorescence was confirmed by bioluminescence imaging and GFP fluorescence imaging, as described before. Receiver Operator Curve (ROC) analysis was performed for the detection of micrometastases in the peritoneum. The overlay of the BLI and GFP signals was used as the ground truth for tumour metastases and the ascending TBRs as positive cut off criteria. The area under the curve (AUC), including the sensitivity and specificity rates at the optimal TBR cut off were computed. The signals of NIR fluorescence and the overlay with BLI and GFP was confirmed against pathology in the primary tumor but not in the micrometastases.
Multiple regions of interest were drawn in the tumour and in adjacent normal tissue and divided by each other to calculate TBRs. TBRs of subcutaneous colon and orthotopic breast tumours were calculated with skin overlying the tumour and adjacent normal tissue. Tongue tumours were imaged through an epithelial cell layer covering the tumour and normal tissue. For colon tumours the peritoneum was opened.
In vivo competition study
Three of the six nude Balb/C mice with bilateral orthotopic MCF-7-luc2-cGFP breast tumours were pre-injected with unconjugated 323/A3 antibody (1 mg, intraperitoneal, 100 μL). After 48 h, all six mice were intravenously injected with 1 nmol 323/A3-800CW. Then, 72 h after injection of 323/A3-800CW fluorescence imaging of all mice was performed using a Pearl imaging system. Mice were sacrificed and tumours were collected. Quantification of fluorescence was done as described before . In brief, tumours were resected and lysed with a TissueLyser II system (Qiagen, Venlo, The Netherlands) using pre-cooled Eppendorf tube holders, 5-mm stainless steel beads, and RIPA buffer supplemented with a complete EDTA-free mini tablet protease inhibitor cocktail. Homogenates were serially diluted in 96-well plates, in parallel with a probe dilution. The fluorescence intensity of both series was detected at 800 nm using the Odyssey scanner. The concentration of probe in the homogenates was extrapolated from the calibration curves and the concentration values were used to calculate the injected dose per gram of tissue (% ID/g) with standard error of the mean (SEM) indicated per group, based on the volume of the homogenate and the weight of the tumours.
Histology and NIR fluorescence microscopy
After ex vivo fluorescence measurements, tumours were snap frozen in isopentane and kept at −80 °C. Tissues were sectioned at 10 μm and fluorescence imaging was performed using the Odyssey imager. The presence of OSC-19-luc2-cGFP and MCF-7-luc2-cGFP tumour cells was confirmed by fluorescence microscopy (Nikon Eclipse e800). All histologic sections were stained with standard haematoxylin-eosin stain (HE) after acetone fixation. To confirm the presence of HT29 and HT29-luc2 cells, sections were stained with an anti-human wide-spectrum cytokeratin antibody (Abcam inc., Cambridge, MA, USA). Primary antibodies or controls were incubated for 60 min at room temperature. All slides were three times washed with PBS and incubated with Envision anti-rabbit (DAKO) for 30 min at room temperature. Subsequently, the slides were washed with PBS and staining was visualized by using 3,3-diaminobenzidine. Sections were counterstained with haematoxylin, dehydrated and mounted with pertex. Frozen OSC-19-luc2-cGFP and MCF-7-luc2-cGFP tumours were stained with an anti-cGFP staining (Evrogen, Moscow, Russia). Sections stained with anti-cGFP were fixated with 4% formalin for 10 min. After washing with PBS, cells were treated with 0,1% saponin/PBS for 10 min and incubated with the anti-GFP antibody, diluted in 0,1% saponin/PBS for 60 min at room temperature. Adjacent sections were fixated with aceton for 10 min followed by three washes with PBS and stained for cytokeratin as described before.
For statistical analysis, SPSS statistical software package (version 20.0 for Windows, IBM SPSS Inc, Chicago, USA) was used. TBRs were calculated by dividing the fluorescent signal of the tumour by fluorescent signal of surrounding healthy tissue. TBRs are reported in mean and standard deviation. A two-way repeated measurement ANOVA was used to assess the relation between TBRs in the dose groups and time points. Furthermore a paired Student’s t-test was used to calculate the overall difference between the EpCAM specific and control groups. The two-way repeated measurement ANOVA was corrected using the Bonferroni correction. A P-value equal or lower than 0.05 was considered significant.