Preclinical Evaluation of Robotic-Assisted Sentinel Lymph Node Fluorescence Imaging

Abstract

An ideal substance to provide convenient and accurate targeting for sentinel lymph node (SLN) mapping during robotic-assisted surgery has yet to be found. We used an animal model to determine the ability of the FireFly camera system to detect fluorescent SLNs after administration of a dual-labeled molecular imaging agent.

Methods

We injected the footpads of New Zealand White rabbits with 1.7 or 8.4 nmol of tilmanocept labeled with ^99mTc and a near-infrared fluorophore, IRDye800CW. One and 36 h after injection, popliteal lymph nodes, representing the SLNs, were dissected with the assistance of the FireFly camera system, a fluorescence-capable endoscopic imaging system. After excision of the paraaortic lymph nodes, which represented non-SLNs, we assayed all lymph nodes for radioactivity and fluorescence intensity.

Results

Fluorescence within all popliteal lymph nodes was easily detected by the FireFly camera system. Fluorescence within the lymph channel could be imaged during the 1-h studies. When compared with the paraaortic lymph nodes, the popliteal lymph nodes retain greater than 95% of the radioactivity at both 1 and 36 h after injection. At both doses (1.7 and 8.4 nmol), the popliteal nodes had higher (P < 0.050) optical fluorescence intensity than the paraaortic nodes at the 1- and 36-h time points.

Conclusion

The FireFly camera system can easily detect tilmanocept labeled with a near-infrared fluorophore at least 36 h after administration. This ability will permit image acquisition and subsequent verification of fluorescence-labeled SLNs during robotic-assisted surgery.

Sentinel lymph node (SLN) detection and excision is becoming incorporated in breast cancer and melanoma surgery (1,2). However, there has been less progress with genitourinary cancers largely because of the limitations in available agents with favorable timing parameters from injection to imaging and surgery. For example, prostate cancer has considerable variation regarding the decision to perform, or the extent of, pelvic lymph node dissection based on preoperative risk factors (3–5). The previously limited pelvic lymph node dissection usually does not include high-risk nodes, causing the current risk of positive lymph nodes to be lower than predicted (5). SLN biopsy in prostate cancer has been shown to accurately predict positive lymph nodes and, pending further research, may parallel the excellent results of SLN biopsy in breast cancer and melanoma (6,7).

Limitations in previous studies include logistics and cumbersome intraoperative γ detectors. However, the endoscope used in minimally invasive surgery has provided a platform for fluorescence-guided surgery. The United States Food and Drug Administration approved ^99mTc-labeled tilmanocept (8), which is the first receptor-targeted radiopharmaceutical approved for lymphatic mapping (9–12). The benefits of receptor targeting allow the agent to bind and remain within SLN, providing logistic flexibility. Radiolabeled tilmanocept can be used to provide a preoperative roadmap using imaging with SPECT/CT (11) or PET/CT (13); however, a laparoscopic γ detector would still be necessary (14). The ideal agent for robotic surgery would be the use of tilmanocept also labeled with a fluorophore for intraoperative visualization by the currently used FireFly endoscope (Da Vinci Si surgical robot; Intuitive Surgical) and the option of preoperative imaging.

In preparation for a phase 1 clinical trial, we investigated the use of fluorescence-labeled tilmanocept for SLN imaging using the FireFly endoscope. Our primary goal was to confirm the ability of the robotic camera to visualize the fluorescence accumulated within the SLN. We tested this ability at different molar doses of fluorescent tilmanocept and at different times after administration. Confirmation of SLN fluorescence imaging by the FireFly camera will provide us with the approximate molar dose and timing parameters with which to perform robotic-assisted fluorescence-based mapping of pelvic SLNs many hours after an ultrasound-guided transrectal injection of fluorescence-labeled tilmanocept into the canine prostate gland.

MATERIALS AND METHODS

Experimental Design

These experiments were designed to test the ability of the Da Vinci Si FireFly camera system to detect fluorescent SLNs resulting from the accumulation of 800CW-tilmanocept. Our goal was to determine the amount of fluorescence-labeled tilmanocept that could be detected by the FireFly system but would not saturate the SLN and carry fluorescence to downstream lymph nodes, which would be falsely identified as SLNs. We elected to use a rabbit model that was used during the development of ^99mTc-tilmanocept (15–17). In this model, a hind-limb administration of the fluorescence-labeled radiotracer would define the popliteal lymph node as the sentinel node and the paraaortic lymph nodes as the downstream lymph nodes. At a predetermined time point, the rabbits were euthanized before fluorescence-guided dissection using the FireFly endoscope attached to the Da Vinci Si robotic surgical system. Images were acquired with the standard clinical software and a modification that permitted the visualization of only the green fluorescence image in the operator console of the surgical robot. This modification, which will be incorporated into future versions of the da Vinci software, essentially turned off the white light of the camera system. After euthanasia, images were acquired over the popliteal fossa before surgical incision and after surgical exposure, when the SLN can be viewed without attenuation by overlying tissue. Four rabbits were used for each of 2 dose levels (1.7 and 8.4 nmol) and each of 2 imaging time points (1 and 36 h), for a total of 16 rabbits. After euthanasia, the popliteal and paraaortic lymph nodes were excised via open dissection.

Imaging Agent Preparation and Administration

The preparation and quality control of ^99mTc-labeled 800CW-tilmanocept was performed using a previously described method (18). The 800CW-tilmanocept used in this study consisted of an average of 1.5 IRDye800CW molecules covalently attached to each tilmanocept molecule. The resulting molecular weight was approximately 18,000 g/mol. Quality control before each animal study was performed using high-performance liquid chromatography and instant thin-layer chromatography; both methods used fluorescence and radioactivity detection. Acceptance criteria required that both the radiochemical purity and the fluorescence purity be greater than 97%.

Our animal protocol was approved by the University of California, San Diego, Institutional Animal Care and Use Committee, and the research staff was appropriately trained. We anesthetized New Zealand White rabbits with ketamine (Fort Dodge Animal Health) and xylazine (Vedco Inc.). The 0.10-mL injections each contained 3.7–7.4 MBq (0.10–0.20 mC) of ^99mTc and 1 of 2 quantities of fluorescence-labeled tilmanocept (1.7 or 8.4 nmol). Each volume was administered to both hind footpads of each rabbit with a 28-gauge needle. We monitored the rabbits after anesthesia visually and with an oximeter that measures heart rate and functional oxygen saturation.

FireFly Fluorescence Imaging Endoscope

The FireFly fluorescence imaging endoscope is an optional component of the DaVinci Si robotic-assisted surgical system. The endoscope uses near-infrared light with an 805-nm-wavelength excitation laser source. We investigated the ideal visualization characteristics of 800CW-tilmanocept by turning off the background white light of the FireFly system. We also tested the detection ability of the camera system using the standard clinical firmware and hardware.

Nuclear Counting

Each node was placed in a plastic scintillation vial and assayed for radioactivity using a γ well counter (100- to 200-keV energy window, Gamma 9000; Beckman Instruments). The nodes were compared with a counting standard that was prepared from a known dilution of the injected material. The accumulation of ^99mTc-800CW-tilmanocept within each SLN was expressed as percentage injected dose, which was calculated by comparing the tissue counts with the counting standard. The amount of fluorescent dye, IRDye800CW, was calculated by multiplying the percentage injected dose times the number of dyes per tilmanocept times the amount of 800CW-tilmanocept administered to each hind paw times 0.01. The percentage extraction E_SLN of popliteal lymph nodes for each animal was calculated by Equation 1, where the cpm_LP, cpm_RP, and cpm_PA represent the counts per minute within the left popliteal, right popliteal, and paraaortic lymph nodes, respectively.

$$E_{\mathit{SLN}}=\frac{100({\mathit{cpm}}_{LP}+{\mathit{cpm}}_{RP}-{\mathit{cpm}}_{PA})}{{\mathit{cpm}}_{LP}+{\mathit{cpm}}_{RP}+{\mathit{cpm}}_{PA}}$$ Eq. 1

Fluorescence Imaging

Ex vivo optical imaging was performed with a time-correlated single-photon fluorescence imaging system (Optix MX2; Advanced Research Technologies). Images were acquired for each excised lymph node using a scan resolution of 1 mm² and a 758-nm pulsed laser diode with an 800-MHz frequency and a 0.3-s integration time. Emitted photons were collected through a pair of 780-nm long-pass filters and a 782-nm band-pass filter, both with a 20-nm bandwidth. For each node, the total fluorescence intensity (counts per minute) was determined by an operator-defined region of interest and reported in kilocounts per second per microwatt (kcts·s⁻¹·μW⁻¹).

Statistical Analysis

The Friedman test for nonparametric paired samples (right popliteal, left popliteal, and paraaortic lymph nodes) was used, with a P value of less than 0.05 needed for statistical significance. Data were analyzed using the SPSS statistical package (version 21; IBM) and reported as mean ± SD.

RESULTS

Visualization of Fluorescent SLNs

The FireFly endoscope attached to the Da Vinci Si robotic surgical system visualized all of the popliteal lymph nodes at 1 and 36 h after administration of 1.7 or 8.4 nmol of fluorescence-labeled tilmanocept. Fluorescence visualization of all lymph nodes was possible with the FireFly endoscope using both the factory settings and the modified software, which turned off the white-light image at the viewing consol. Visualization of afferent lymph channels was possible 1 h after injection. Surgical exposure was not required for visualization of the SLN fluorescence when the duration between imaging and injection was 36 h. Four lymph nodes of the 1-h studies were also visualized behind the muscle layer without the need for surgical exposure.

Figure 1 demonstrates the fluorescence detection of an SLN and its afferent lymph channels by the FireFly camera. This was accomplished at the lower of the 2 doses, 1.7 nmol of fluorescence-labeled tilmanocept, and without surgical exposure of the lymph node or the afferent lymph channels. Figure 1A is a bright-field image of a rabbit hind-limb acquired 1 h after a 1.7-nmol injection of 800CW-tilmanocept. After the switch to fluorescence mode (Fig. 1B), fluorescence could be visualized traveling up the lymphatic channel (arrow) to the SLN located in the popliteal fossa (arrow in Fig. 1C). Figure 1D is a fluorescence-mode image of the SLN (arrow), which accumulated 15 pmol of fluorescence-labeled tilmanocept to which 25 pmol of IRDye800CW was covalently attached. After excision, the SLN produced a fluorescence intensity of 8.6 kcts·s⁻¹·μW⁻¹. The paraaortic lymph nodes contained 0.14 pmol of 800CW-tilmanocept (0.20 pmol IRDye800CW), which produced a fluorescence intensity of 0.16 kcts·s⁻¹·μW⁻¹. The percentage extraction was 98.3%. The fluorescence images (Figs. 1B and 1D) use a modification of the console software that allows visualization of only the fluorescence image from the FireFly camera.

FIGURE 1.

The FireFly endoscope attached to Da Vinci Si robotic surgical system visualized all popliteal lymph nodes at 1 and 36 h after administration of 1.7 or 8.4 nmol of ^99mTc-labeled fluorescent tilmanocept. (A) Bright-field image of rabbit hind-limb was acquired 1 h after 1.7-nmol injection of 800CW-tilmanocept. (B) After switch to fluorescence mode, lymphatic channel (arrow) was visualized. (C and D) On repositioning of camera to visualize popliteal region (arrow, C), fluorescence mode visualized SLN (arrow, D) without surgical exposure. SLN accumulated 0.87% of injected dose (25 pmol of IRDye800CW).

SLN Accumulation

Table 1 lists the means and relative SDs of the various measurements of SLN accumulation—percentage injected dose, amount of fluorescent dye, percentage extraction, and fluorescence intensity—for each of the experimental groups (1 h after a 1.7-nmol 800CW-tilmanocept injection, 36 h after a 1.7-mol injection, 1 h after an 8.4-nmol injection, and 36 h after an 8.4-nmol injection).

TABLE 1.

SLN Accumulation

Number of SLNs	Location of SLNs	Time (h)	Tilmanocept injected (nmol)	Percentage injected dose*	IRDye800CW (nmol)	Extraction (%)*	Fluorescence intensity (kcts s−1 μWminus;1)†
8	Popliteal	1.0	1.7	0.95 ± 0.78	0.024 ± 0.020‡	99.4 ± 4.0	6.22 ± 2.45‡
8	Popliteal	36.0	1.7	0.56 ± 0.39‡	0.014 ± 0.010‡	98.5 ± 0.9	8.62 ± 3.72‡
8	Popliteal	1.0	8.4	1.64 ± 0.77	0.21 ± 0.10	95.1 ± 5.2	19.1 ± 9.23
8	Popliteal	36.0	8.4	1.21 ± 0.33	0.15 ± 0.04	94.2 ± 5.9	11.6 ± 4.24

^*Measured by γ well counter (100–200 keV).

^†Measured by Optix MX2.

^‡P < 0.05 for comparison with data from same time but different amount of injected tilmanocept.

Data are mean ± SD.

SLN accumulation was higher at 1 h after injection than at the later time point. At 1 h, the mean and SD of the percentage injected dose from the 8 SLNs receiving the 1.7-nmol dose of 800CW-tilmanocept was 0.95% ± 0.78%; at 36 h the percentage injected dose was 0.56% ± 0.39%. The 8.4-nmol dose delivered 1.64% ± 0.77% at 1 h and 1.21% ± 0.33% at 36 h. 800CW-tilmanocept delivered more fluorescent dye using the higher dose (8.4 nmol) at 1 h from injection (0.21 ± 0.10 vs. 0.024 ± 0.020 pmol; P < 0.05) and at 36 h after injection (0.15 ± 0.04 vs. 0.014 ± 0.0.010 pmol; P < 0.05). The fluorescence intensity of all the paraaortic nodes was less than 10% of the popliteal lymph nodes. At both doses (1.7 and 8.4 nmol), the popliteal nodes had higher (P < 0.050) optical fluorescence intensity than the paraaortic nodes at the 1- and 36-h time points. The percentage SLN extraction for all the experimental groups was greater than 95%, indicating the lack of SLN saturation at both low and high doses (1.7 and 8.4 nmol) of 800CW-tilmanocept. The 8.4-nmol dose produced higher fluorescence intensity with the SLNs at both time points: 19.1 ± 9.23 kcts·s⁻¹·μW⁻¹ 1 h after injection and 11.6 ± 4.24 kcts·s⁻¹·μW⁻¹ 36 h after injection. The 1.7-nmol dose of 800CW-tilmanocept produced a fluorescence intensity of 6.22 ± 2.45 kcts·s⁻¹·μW⁻¹ at 1 h and 8.62 ± 3.72 kcts·s⁻¹·μW⁻¹ at 36 h. Increasing the amount of 800CW-tilmanocept injected produced a significantly higher (P < 0.05) fluorescence intensity at both time points.

DISCUSSION

This study demonstrated that the FireFly endoscopic camera system of the Da Vinci Si surgical system can image fluorescent SLNs resulting from an 8.4-nmol injection of 800CW-tilmanocept and that this can be accomplished without receptor saturation within the SLN. Moreover, when imaging is performed at the appropriate time, the FireFly camera can visualize fluorescence from afferent lymph channels.

Previous animal studies using the New Zealand White rabbit model have shown that without a molecular or physiologic mechanism to retain the imaging agent within the lymph node, nearly all ^99mTc had passed through the SLNs by 6 h after injection (15–17). Using ^99mTc-labeled 800CW-tilmanocept, we show that radioactivity levels can be detected in popliteal lymph nodes at 1 and 36 h after footpad injection in a rabbit model. The molecule is retained within the popliteal lymph node and does not significantly migrate to the next level of lymph nodes (paraaortic) or significantly dissipate over time. These results suggest that the receptor binds within 1 h and will stay localized up to 36 h, allowing for flexibility in image acquisition with nuclear medicine scanning before surgery. The use of imaging before surgery to identify SLNs has been shown to reduce operative time when compared with γ probe–guided lymph node dissections during radical prostatectomy (19,20).

Pelvic lymphadenectomy during radical prostatectomy provides staging information that may guide adjuvant therapy (21–25). The extent and definition of the pelvic lymph node dissection are controversial (5,26,27). Studies have shown that limited pelvic lymph node dissection resects only 35% of the primary landing zones of prostate-draining lymph nodes (28). Therefore, the SLN dissection has become an emerging concept during radical prostatectomy to tailor surgery and decrease the risk-to-benefit ratio of lymph node dissection by resecting only high-yield nodes (7,29,30). Moreover, this technique may reduce the complications of increasingly extensive lymph node dissections (i.e., lymphocele, lymphedema, deep venous thrombosis, and pulmonary embolism) (31). One of the initial studies of SLN dissection was performed by Wawroschek et al. by injecting ^99mTc-labeled colloid into the prostate and performing dynamic lymphoscintigraphy followed by surgery with γ-probe guidance the next day (29). This concept was later confirmed in a study with over 1,000 prostatectomies. That study noted that only 1% of metastases were found in patients with negative SLNs; however, the study did not provide intraoperative visual confirmation with fluorescence (32).

Previous investigations have added indocyanine green (ICG) to fluorescently image lymph channels and SLNs. One study used ^99mTc-labeled Nanocoll (GE Healthcare) to image before radical prostatectomy followed by a separate ICG injection at the time of surgery (33). A separate study used a mixture of ICG and ^99mTc-labeled Nanocoll imaged by SPECT/CT 2 h after injection, followed by surgery within the hour (20). The use of an ICG-radiolabeled Nanocoll mixture assumes that the dye binds to Nanocoll with adequate affinity for valid SLN mapping. Both studies using ICG detection showed accurate results (34); however, the use of ICG in this manner requires multiple injections or an extra compounding step by a pharmacist. Moreover, Nanocoll is prepared from human blood and can have similar risks to transfusion. Lastly, Nanocoll is not available in the United States. Therefore, using a non–blood-based dual-labeled receptor-targeted molecule such as 800CW-tilmanocept may have better logistic properties with reduced risk. In addition to determining the benefit of the 36-h interval, we also established that the larger dose (8.4 nmol) provided improved fluorescence without saturation and migration to next-level lymph nodes.

CONCLUSION

^99mTc-tilmanocept covalently tagged with IRDye800CW can provide fluorescence detection of SLNs by the FireFly endoscope used during robotic surgery. Detection can be achieved with and without background white light. The molecule remains in the SLN at the 2 tested doses for at least 36 h to allow image acquisition and subsequent surgical verification via fluorescence guidance. Further in vivo (35) and subsequent human studies are needed to provide more information on the feasibility and accuracy of this technology during robotic surgery.