2.1.

Cell Culture

The human pancreatic cancer cell-line BxPC-3 was maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum and 2-mM glutamine from Gibco-BRL, Life Technologies, Inc. (Grand Island, New York). All cells were cultured at 37°C in a 5% ${\mathrm{CO}}_2$ incubator.

2.2.

Surgical Orthotopic Implantation

After confluence, BxPC-3 human pancreatic cancer cells ($1\times {10}^6$) were injected subcutaneously into the flanks of nude mice until tumors grew between 10 and 20 mm in diameter. The tumor was harvested, and $1{\mathrm{mm}}^3$ tumor fragments were implanted orthotopically in nude mice, as previously described by our laboratory.^16–19

2.3.

Antibody-Dye Conjugation

Chimeric anti-CEA antibody (Aragen Biosciences, Morgan Hill, California) was conjugated with Alexa Fluor dyes (Life Technologies, Inc.) or DyLight dyes (Thermo Fisher Scientific, Rockford, Illinois) (Table 1) per manufacturer specifications, ensuring a minimum of at least $4:1$ dye:protein ratio. Protein:dye concentrations were confirmed using a NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, Massachusetts).

2.4.

Animal Care

Athymic nu/nu nude mice (AntiCancer, Inc., San Diego, California), between 4 and 6 weeks of age, were maintained in a barrier facility at AntiCancer, Inc., on high-efficiency particulate air-filtered racks. The animals were fed with autoclaved laboratory rodent diet (Teckland LM-485; Western Research Products, Orange, California). All surgical procedures and imaging were performed with the animals anesthetized by intramuscular injection of 0.02 mL of a solution of 50% ketamine, 38% xylazine, and 12% acepromazine maleate. When inhalational anesthesia was required, 99% isoflurane and 100% oxygen were delivered via a vaporizer. All animal studies were conducted in accordance with the principles and procedures outlined in the NIH Guide for the Care and Use of Animals under PHS licence number A3873-1.

2.5.

Experimental Protocol

After surgical orthotopic implantation of the human pancreatic cancer cell-line BxPC-3, tumors were allowed to grow for 7 days, after which the mice were injected with the antibody-dye conjugate into the retro-orbital vein (Fig. 1). Antibody amounts were maintained between 50 and 75 μg, while ensuring 1.25-μmol dye was delivered with each injection. Whole body imaging was subsequently performed using the OV100 Small Animal Imaging System (Olympus, Tokyo, Japan) after 24 h. If tumors were detectable on whole body imaging, the animal was sacrificed, and tumor characteristics including size, fluorescence intensity, weight, and depth were determined. For mice that did not demonstrate tumors on initial imaging, the tumors were allowed to grow for another 7 days, and the process repeated until the tumor was visible, at which point mice were sacrificed and tumors analyzed as above.

Fig. 1

Illustration of the experimental plan. Serial weekly imaging was designed to attempt to capture the smallest visible tumor on whole-body imaging.

2.6.

Tissue Penetration

Prelabeled tumor fragments, with either Alexa Fluor or DyLight dye–antibody conjugate, were cut into fragments of 0.5, 1, and 2 mm. Three fragments of each size were then placed on a petri dish and imaged. After confirming fluorescence labeling, the fragments were covered with a layer of skin obtained from a nude mouse and imaged to evaluate skin penetration. The tumor and peritoneum were then covered with the abdominal wall obtained from a nude mouse and subsequently imaged to evaluate for penetration through the skin and the abdominal wall.

2.7.

Photobleaching

The 488, 550, 650, and the 755 groups of dye conjugates were titrated to equal fluorescence intensity in three consecutive wells on a 64-well plate. Each well plate contained 200 μL. The conjugates were then continuously exposed to a 60-W halogen light bulb to mimic operating room settings for up to 48 h. For the 488 group, fluorescein was also added as a control. The dye conjugates were imaged at the 1-, 2-, 4-, 6-, 8-, 24- and 48-h time points and intensities were recorded.

2.8.

Hemoglobin Quenching

DyLight 488, 550, 650, and 755 dye conjugates were titrated to approximately equal fluorescence intensity on three consecutive wells on a 64-well plate. Each well contained 200 μL. Then 15 μL mouse whole blood was then sequentially added to each well, after which the wells were imaged and the fluorescence was recorded. Tumor fragments, prelabeled with each of the dyes, were then covered with a liver fragment, modeled to simulate penetration through a hemoglobin-dense organ. Imaging was performed with the OV100 to assess for penetration through the liver tissue.

2.9.

Animal Imaging

Mice were imaged using the OV100, containing an MT-20 light source (Olympus Biosystems, Planegg, Germany) and DP70 CCD camera (Olympus Corp.). The OV100 was used due to its unique ability to accomplish high-fidelity fluorescence imaging with variable magnification capabilities that allow for imaging of not only the whole animal, but also the subcellular level.²⁰ The instrument incorporates a unique combination of high numerical aperture and long working distance. Four individually optimized objective lenses, parcentered and parfocal, provide a 10⁵-fold magnification range for seamless imaging of the entire body down to the subcellular level without disturbing the animal. In addition, openings on the machine allow for delivery of inhalational anesthesia to the animal without allowing entry of outside light. This allows for capturing high-quality in vivo images with minimal morbidity to the animal by avoiding anesthetic injections. All images were analyzed using Image-J (National Institutes of Health, Bethesda, Maryland) and were processed with the use of Photoshop elements-11 (Adobe Systems Inc., San Jose, California).

2.10.

Frozen Section Analysis

Twenty-four hours after injection of 1.25 μmol of anti-CEA conjugated to DyLight 650, the orthotopic tumor was harvested. The tumor was placed in OCT compound and subsequently dipped into liquid nitrogen. Once the tumor was frozen, it was placed on a microtome, sliced into 5-μm sections, placed on a glass slide, and immediately dipped in 95% ethanol for fixation. Alternate slides were stained with hematoxylin & eosin (H&E) or 4′,6-diamidino-2-phenylindole for fluorescence imaging. Fluorescence and bright-field microscopy were performed with the Biorevo BZ-9000 microscope (Keyence Corp., Itasca, Illinois).

2.11.

Statistical Analysis

All statistical analyses were performed using SPSS software version 21 (IBM, Armonk, New York). For pairwise comparisons between groups within the wavelengths, quantitative variables were compared with the Student’s t-test and confirmed with the Wilcoxon rank-sum test. A $p$-value of $<0.05$ was considered significant. The 95% confidence intervals obtained on analysis of the data were configured into all the error bars of the appropriate figures and graphs. Multiple groups ($\ge 3$) were initially compared with one-way analysis of variance (ANOVA), and if a difference was found, pairs of interest were further compared with the Student’s t-test. For tumor size versus intensity modeling, the following $R^2$ values were obtained when a quadratic function plot was used for regression analysis: DyLight 488: 0.32; DyLight 550: 0.34; DyLight 650: 0.26; and DyLight 755: 0.41.

3.1.

Fluorescence Intensity of Dye-Antibody Conjugates

Mice with orthotopic BxPC-3 pancreatic cancers were treated with various anti-CEA antibody-dye conjugates and then imaged 24 h later with the OV100 (Fig. 1). Within the 550 and 750 group of dyes, the DyLight dyes were brighter than the corresponding Alexa Fluor dyes ($p<0.001$). Alexa Fluor 488 was significantly brighter than DyLight 488 ($p<0.001$). There was no significant difference in the 650 group of dyes ($p=0.4$). The brightest dye was DyLight 755 (Fig. 2), which was significantly brighter than the next brightest dye, Alexa Fluor 488 ($p<0.001$). There was no significant difference on ANOVA analysis between Alexa Fluor 488, DyLight 650, and Alexa Fluor 600 with $p=0.32$. DyLight 650 was, however, significantly brighter than Alexa Fluor 750 with $p=0.05$. No significant difference was noted between Alexa Fluor 750 and DyLight 488 with $p=0.18$, although Alexa Fluor 750 was significantly brighter than DyLight 550 ($p=0.01$). DyLight 550 had no significant difference when compared with DyLight 488 ($p=0.2$) but was significantly brighter than Alexa Fluor 555 with $p<0.001$. A summary of the intensities is provided in Fig. 3.

Fig. 2

Representative images of each dye used in imaging studies. (a) Alexa Fluor 488 is significantly brighter than DyLight 488 ($p<0.001$). (b) DyLight 550 is significantly brighter than Alexa Fluor 555 ($p<0.001$). (c) Although DyLight 650 appeared brighter, no significant difference was noted between Alexa Fluor 660 and DyLight 650 ($p=0.4$). (d) DyLight 755 is significantly brighter than Alexa Fluor 750 ($p<0.001$). Note the decreasing ability to discern the background with increasing wavelength (a–d).

Fig. 3

DyLight 755 was significantly brighter than Alexa Fluor 488 ($p<0.001$). There was no significant difference with ANOVA between Alexa Fluor 488, DyLight 650, and Alexa Fluor 600 ($p=0.32$). DyLight 650 was significantly brighter than Alexa Fluor 750 ($p=0.05$). There was no significant difference between Alexa Fluor 750 and DyLight 488 ($p=0.18$). Alexa Fluor 750 was significantly brighter than DyLight 550 ($p=0.01$). There was no significant difference between DyLight 550 and 488 ($p=0.2$), which were significantly brighter than Alexa Fluor 555 ($p<0.001$), which had the lowest overall intensity.

3.2.

Tumor-to-Background Ratio

The 750 and 650 group of dyes had the highest TBRs (Fig. 4) with DyLight 755 showing the highest ratio of the other dyes ($p<0.001$). ANOVA analysis showed no significant difference between the TBRs of DyLight 650, Alexa Fluor 650, and Alexa Fluor 750, all of which were significantly higher than the dye with the next highest ratio DyLight 550 ($p<0.001$). The TBR of DyLight 550 was significantly higher than that of Alexa Fluor 555 ($p<0.001$). No significant difference was noted between the Alexa Fluor 555 and 488 pair ($p=0.1$) and between the Alexa Fluor 488 and DyLight 488 pair ($p=0.07$). Alexa Fluor 555 had a TBR that was significantly higher than DyLight 488 ($p=0.007$)

Fig. 4

DyLight 755 had a significantly higher tumor-to-background ratio (TBR) compared with all the other dyes ($p<0.001$). There was no significant difference between DyLight 650, Alexa Fluor 650, and Alexa Fluor 750 with ANOVA. DyLight 550 was significantly higher than Alexa Fluor 555 ($p<0.001$). There was no significant difference between Alexa Fluor 555 and 488 ($p=0.1$) and between Alexa Fluor 488 and DyLight 488 ($p=0.07$). Alexa Fluor 555 was significantly higher than DyLight 488 ($p=0.007$).

The longer-wavelength dyes had the most contrast between tumor and background. The most background autofluorescence was noted with the 488 group, allowing for near complete visualization of the background, while little to no background was discernible with the 650 and 750 groups.

3.3.

Tumor Depth and Size

The ability to detect the smallest and the deepest tumors consistently increased with longer wavelengths [Figs. 5(a) and 5(b)]. The longer wavelength dyes (650 and 750 nm) had the best tissue penetration with the ability to visualize the smallest tumors. In general, DyLight was able to detect smaller and deeper tumors better than Alexa Fluor within each group, except for the 488 group. There was no significant difference in the ability to detect deeper tumors within the 750 and 550 groups ($p=0.5$ and 0.2, respectively). There was a significant difference, however, within the 650 and 488 groups. DyLight 650 could detect deeper tumors better than Alexa Fluor 660 ($p=0.04$). Alexa Fluor 488 could detect deeper tumors better than DyLight 488 ($p=0.006$). Between groups, DyLight 755 could be detected significantly deeper than DyLight 650 ($p=0.04$) and DyLight 650 could be detected significantly deeper than DyLight 550 ($p=0.03$). No significant difference was observed between DyLight 550 and Alexa Fluor 488 ($p=0.3$).

Fig. 5

(a) White bars represent the smallest detectable tumor size, and colored bars represent tumor depths. In analyzing the tumor depths, within groups, there was no significant difference between DyLight 755 and Alexa Fluor 750 ($p=0.5$) and between DyLight 550 and Alexa Fluor 555 ($p=0.2$). DyLight 650 enabled significantly deeper imaging than Alexa Fluor 660 ($p=0.04$), and Alexa Fluor 488 enabled significantly deeper imaging than DyLight 488 ($p=0.006$). Between groups, DyLight 755 enabled significantly deeper imaging than DyLight 650 ($p=0.04$), and DyLight 650 enabled significantly deeper imaging than DyLight 550 ($p=0.03$). There was no significant difference between DyLight 550 and Alexa Fluor 488 ($p=0.3$). With respect to tumor size, within groups, DyLight 755 detected significantly smaller tumors than Alexa Fluor 750 ($p=0.009$); DyLight 550 detected significantly smaller tumors than Alexa Fluor 555 ($p=0.01$), and Alexa Fluor 488 detected significantly smaller tumors than DyLight 488 ($p=0.01$). No significant difference within the 650 group was observed ($p=0.6$). Between groups, there were significant differences, except between DyLight 550 and 650 ($p=0.4$). DyLight 755 detected significantly smaller tumors than DyLight 650 ($p=0.006$), and DyLight 550 detected significantly smaller tumors than Alexa Fluor 488 ($p<0.001$). (b) Depiction of the relationship between tumor size, depth, and dye wavelength. With an increase in wavelength, there was an increase in the ability to detect deeper and smaller tumors.

Similarly, smaller tumors were detected at longer wavelength with DyLight dyes better than Alexa Fluor dyes, except for the 488 group. Differences within the groups were all significant except for the 650 group ($p=0.6$). Within the 750 and 550 groups, DyLight conjugates detected significantly smaller tumors ($p=0.009$ and 0.01, respectively). In the 488 group, Alexa Fluor 488 was able to detect significantly smaller tumors ($p=0.01$). Between groups, there were significant differences, except between DyLight 550 and 650 ($p=0.4$). DyLight 755 was able to detect significantly smaller tumors than DyLight 650 ($p=0.006$), and DyLight 550 was able to detect significantly smaller tumors than Alexa Fluor 488 ($p<0.001$).

Prelabeled tumor fragments of 0.5, 1, and $1.5{\mathrm{mm}}^3$ were imaged, and subsequently a layer of skin and body wall was placed over the fragments. Repeated imaging verified better tissue penetration with the longer wavelength dyes (Fig. 6). With addition of the tissue layers, the smallest of the DyLight 650 and 755 labeled fragments were still visible on imaging but were invisible in the 550 and 488 groups. Within all groups, the DyLight-labeled fragments were visible at greater depths in comparison to the Alexa Fluor dyes. This held true for all the groups except for the 488 group, where there was better tumor visualization with the Alexa Fluor 488-labeled tumor fragments.

Fig. 6

Prelabeled tumor fragments were serially covered with skin and skin + anterior abdominal wall and assessed for tissue penetration. $\mathrm{A}4=\text{Alexa Fluor}488$; $\mathrm{D}4=\text{DyLight}488$; $\mathrm{A}5=\text{Alexa Fluor}555$; $\mathrm{D}5=\text{DyLight}550$; $\mathrm{A}6=\text{Alexa Fluor}660$; $\mathrm{D}6=\text{DyLight}650$; $\mathrm{A}7=\text{Alexa Fluor}750$, and $\mathrm{D}7=\text{DyLight}755$.

3.4.

Tumor Size and Fluorescence Intensity

In assessing for a correlation between tumor size and fluorescence intensity with a constant dose of antibody-dye conjugate, the fluorescence intensity of all the DyLight dye labeled tumors and the corresponding tumor sizes were measured. In the analysis, the fluorescence intensity was set as the dependent variable and the tumor sizes as an independent variable, thus creating a model for prediction of the expected intensity given a certain tumor size (Fig. 7). Small tumors appeared to have a lower signal intensity, which would then increase to a maximum for an “optimum” tumor size and would again begin to decrease with increasing tumor size. The longer wavelength dyes were not only the brightest, but also maintained their signal intensity over a broader range of tumor sizes when compared with the shorter-wavelength dyes.

Fig. 7

Prediction model for optimum fluorescence intensity and tumor size according to dye. $\text{Blue}=\text{DyLight}755$; $\text{Red}=\text{DyLight}650$; $\text{Orange}=\text{DyLight}550$, and $\text{Green}=\text{DyLight}488$.

3.5.

Photobleaching

Photobleaching was assessed with a 60-W halogen light continuously shone on antibody-dye conjugates for a period of up to 48 h, with serial imaging and analysis of fluorescence intensity. Overall, the DyLight dyes were more photostable than the Alexa Fluor dyes [Figs. 8(a) and 8(b)]. The time point of eight hours was used as a reference point for evaluation, as this would encompass even the lengthiest of most pancreatico-duodenectomies and other oncologic surgeries. At the 8-h time point, there was no significant difference in the percentage of the original intensity in the 488 ($p=0.5$) and 550 ($p=0.3$) groups. Alexa Fluor 488 and DyLight 488 retained 96% and 90% of their original intensity, respectively. Both Alex Fluor 488 and DyLight 488 were significantly brighter than FITC, which retained only 45% of its original intensity ($p<0.001$ and 0.002, respectively). The 550 group of dyes was the most photostable, retaining approximately 95% of their original intensity for up to 48 h. With the longer-wavelength dyes, however, there was a significant difference between groups. Alexa Fluor 660 lost all of its fluorescence in 1 h. In contrast, by 8 h, DyLight 650 still had 53% of its original fluorescence ($p<0.001$). Similarly, in the 750 group, there was a more rapid decline in the fluorescence intensity of Alexa Fluor 750, and by 8-h, Alexa Fluor 750 had no discernible fluorescence DyLight 755 retained 22% of its original fluorescence intensity at 8 h. Overall, DyLight dyes were more photostable than Alexa Fluor.

Fig. 8

(a) 64-well plate with antibody-dye conjugates. Conjugates were exposed to continuous white light and imaged at different time points. $\mathrm{A}4=\text{Alexa Fluor}488$; $\mathrm{D}4=\text{DyLight}488$; $\mathrm{A}5=\text{Alexa Fluor}555$; $\mathrm{D}5=\text{DyLight}550$; $\mathrm{A}6=\text{Alexa Fluor}660$; $\mathrm{D}6=\text{DyLight}650$; $\mathrm{A}7=\text{Alexa Fluor}750$; and $\mathrm{D}7=\text{DyLight}755$. (b) $\mathrm{A}=\text{Intensity profile of the}488\text{group}$; $\mathrm{B}=\text{Intensity profile of the}550\text{group}$; $\mathrm{C}=\text{Intensity profile of the}650\text{group}$, and $\mathrm{D}=\text{Intensity profile of the}750\text{group}$. Percentages are with respect to the original intensity.

3.6.

Hemoglobin Quenching

With the assessment of hemoglobin quenching, a significant difference was noted between the longer and shorter wavelength dyes. Wavelengths of 550 nm or less showed significant quenching of the fluorescence signal upon addition of whole blood. Dyes at 650 nm or greater showed resistance to hemoglobin quenching [Figs. 9(a) and 9(b)]. With serial addition of whole blood, both DyLight 488 and 550 had $<1\%$ of their original intensity after 100 μL; while DyLight 650 and 755 had 95% and 89% of their original intensity, respectively. In a related experiment, a thin slice of liver tissue was placed over prelabeled $1\text{-}{\mathrm{mm}}^3$ fragments of tumor and imaged to evaluate for tissue penetration. Similarly, the 488 and 550 fragments were not discernible after placing the liver tissue overlay, while all three fragments in the 650 and 750 groups were visible, illustrating the difference in tissue penetration between the longer and shorter wavelength dyes.

Fig. 9

(a) 64-well plate with antibody-dye conjugates. Whole blood was serially added to each well in 15 μL. 488, 550, 650, and 750 correspond to DyLight 488, DyLight 550, DyLight 650, and DyLight 755, respectively. (b) Graphic depiction of hemoglobin quenching. 488, 550, 650, and 750 correspond to DyLight 488, DyLight 550, DyLight 650, and DyLight 755, respectively. DyLight 755 and 650 maintained their fluorescence signal, while DyLight 488 and 550 showed significant hemoglobin quenching.

3.7.

Frozen Section Analysis of Surgical Margins

Figure 10 depicts a frozen section of tumor margin prelabeled with an anti-CEA antibody conjugated to DyLight 650 alongside a corresponding H&E-stained slide. The sharp demarcation of tumor versus normal tissue interface on fluorescence imaging allows for more accurate determination of tumor margins.

Fig. 10

(a) Frozen section of tumor margin prelabeled with an anti-CEA antibody conjugated to DyLight 650. (b) Corresponding H&E-stained slide. Note sharp demarcation of the tumor-normal tissue interface seen with fluorescence imaging (arrows), allowing for more accurate determination of tumor margins.