Using paired-agent principles and widefield fluorescence optical projection tomography to rapidly identify and localize lymph node micro metastases

Abstract

The proof-of-concept presented here utilizes a micro metastatic lymph node mode, where porcine lymph nodes are first implanted with a 200-micron human cancer cell spheroid. The lymph nodes are then infused with 40 nM ABY-029 and IRDye 700Dx-Affibody isotype conjugate, followed by widefield multi-channel imaging of 700 nm, 800 nm fluorescence and transmission images across 12 projections, followed by tomographic reconstruction of each channel. Absorption and fluorescence sensitivity matrices needed for reconstructions are created through monte-carlo simulations of photon propagation through the imaging system. Each channel is reconstructed separately at a resolution of 87x87x65 voxels, followed by calculation of binding potential (BP) maps as the ratio of the two fluorescence reconstructions, minus 1. The BP reconstruction is then overlaid onto the absorption reconstruction and the location of the EGFR+ tumor spheroid is isolated in the 3D reconstruction through inspection. These procedures are proposed for intraoperative use.

1. INTRODUCTION

Head and neck squamous cell carcinoma is the most prevalent of the head and neck cancers¹ and can be considered a curable disease if two conditions are met: 1) the entirety of the primary tumor is removed through surgical resection and 2) the cancer has not spread to adjacent tissues or distant organs. However, this disease has a high propensity for metastasis from the primary tumor,² where typically the first location of metastasis is thought to be the lymph nodes draining the primary tumor. As such, if metastasis is suspected a surgeon may elect to do a radical neck dissection where all draining lymph nodes are isolated from the head and neck tissue and screened for the presence of metastasis. This screening process is done via standard tissue histopathology through either frozen-sectioning or sectioning formalin-fixed, paraffin-embedded (FFPE) tissue.

To accomplish this, the lymph nodes are first bread-loaf sliced into 2-mm thick sections, where each of the bread loaf sections are then processed and a single 3–5-micron slice is then taken and stained via hematoxylin and eosin followed by inspection for metastasis by a trained pathologist. This process is both time-consuming and can only be used to screen <1% of the lymph node volume, where any metastases smaller than 2-mm have a chance of being missed due to sampling error. To address this, our group has proposed a novel staining and infusion protocol³ to rapidly stain lymph nodes intraoperatively followed by fluorescence optical projection tomography to rapidly screen the entirety of the lymph node volume for the presence of metastatic disease.^4,5 Herein we demonstrate a proof-of-concept using a previously described metastatic lymph node model including infusion of the paired-agent cocktail and subsequent imaging through multiple fluorescence projections followed by reconstruction using an algebraic tomographic reconstruction algorithm. We illustrate the uses of this methodology for the detection of 200-micron human cancer cell spheroids through 3D binding potential maps.

2. METHODS

The lymph node model and infusion protocol used here has been described previously by our group.^3,6–8 Here we describe these procedures briefly:

2.1. Human cancer cell spheroid culture

A previously described methyl cellulose-based spheroid culture protocol was used in this project.^6,9 Briefly, human squamous cell carcinoma cells (FaDu) are cultured until 80% confluence, followed by trypsinization and suspended in 0.25% methyl cellulose:DMEM solution in non-tissue culture treated transparent round-bottom 96-well plates. Cells were seeded at 1,000 cells per well and allowed to culture for 1-2 days until spheroids reached approximately 200-300-microns in diameter. These spheroids were then gently aspirated into a 1000 uL micropipette tip and transferred to a 1.5 uL conical vial, followed by rinsing 3x with 1x PBS. The rinsed spheroids were then gently loaded into a 22-gauge needle tip attached to a 1-mL syringe, ready to be implanted into a porcine lymph node. Significantly more spheroids were cultured (n = 48) than needed in these experiments to allow for selection of the most desirable spheroids (i.e. properly condensed spheroids with no fibers in the wells).

2.2. Micro metastatic lymph node model

Fresh porcine heads were harvested from freshly sacrificed pigs via a local animal butchering facility (Park Packing Co. Chicago IL). These porcine heads were immediately brought back to the laboratory and dissected, removing the tissue surrounding the mandible and neck intact. These tissue pieces were then palpated to find lymph nodes, followed by fine dissection to remove mandibular and cervical lymph nodes. Special care was taken to remove lymph nodes without interrupting the capsular membrane while still removing as much attached adipose tissue as possible. Of the resulting lymph nodes, only those with a diameter less than 1.2 cm were utilized for these experiments to accommodate the limited field-of-view of our prototype dual-channel fluorescence-emission tomography system. These lymph nodes were then implanted with the cultured spheroid described in section 2.1, where upon implantation the needle tip was allowed to sit inside the lymph node for 1-minute to reduce the chances of spheroid ejection upon removal of the syringe.

2.3. Paired-agent infusion

The paired-agent infusion protocol has been described previously by our group.⁶ In brief, the spheroid-loaded lymph node was punctured on both sides with 28-gauge butterfly needles. The lymph node was first rinsed with 2% bovine serum albumin (BSA) in 1x phosphate-buffered saline (PBS) for 5 minutes at a flow rate of 300 uL/minute. Immediately afterward, the lymph node was then infused with a cocktail of an anti-EGFR affibody molecule, ABY-029^10,11 at a concentration of 40 nM alongside 40 nM IRDye700DX-Affibody isotype control for 1 minute at a flow rate of 300 uL/minute. The infused spheroid was then allowed to sit for 5 minutes to allow adequate time binding, followed by a second round of rinsing with 2% BSA in 1X PBS for 5 minutes at a flow rate of 300 uL/minute.

2.4. Lymph node mounting

The spheroid-loaded and agent-infused lymph node was then mounted in a cylinder of 1% agarose. To do so, 1 gram of biotechnology-grade agarose powder (AMRESCO, Solon OH, USA) was dissolved in 100 mL of deionized (DI) water, using a common microwave as the heat source. The resulting solution was the pipetted into a 1.2 x 3 cm cylindrical mold and allowed to cool until it reached 40 degrees Celsius in temperature. The lymph node was then placed into the cooling gel with forceps and held in place for approximately 2-minutes while allowing the gel to fully solidify. Once the gel was solidified, the mold was then placed in the refrigerator for 5-minutes to ensure the gel was fully cooled before proceeding with imaging.

2.5. Fluorescence optical projection tomography

The optical system used here has been described previously.^4,5 In brief, the system consists of 2 LEDs and collimator assemblies (660 nm & 780 nm) with corresponding bandpass excitation filters (680 +/− 5 nm & 780 +/− 5 nm) connected to a beam splitter to bring both light sources in line with each other. The subsequent beam path passes through a water bath, in which the agarose block is suspended while connected to a 360-degree motorized rotating stage. On the detection side, fluorescence light is collected via a 100 mm focal length plano-convex lens, which focuses light through an aperture stop, followed by collimation by a 75 mm plano-convex lens. The final numerical aperture (NA) of the system was calculated to be 0.06. The collimated emission light then passes through either a 712 long-pass or 800 nm long pass filter, depending on the channel being imaged, and passes onto a cooled CCD Coolsnap HQ2 camera sensor (Andor Newton, Belfast Ireland). All optical components apart from the camera were purchased from Thorlabs (Newton NJ, USA). 12 equally spaced projections were obtained in both the 700 nm and 800 nm fluorescence channels for subsequent reconstruction. An additional transmission image was taken at each projection using the 700 nm light source with no emission filter to allow for an absorption reconstruction as well.

3D reconstruction was done using an in-house server system that consisted of 192 GB of RAM and a 24-core Intel Xeon processor. Algebraic tomographic reconstruction was done via a traditional maximum-likelihood expectation-maximization (MLEM) algorithm with 50 iterations per reconstruction. The system matrix was constructed using montecarlo simulations of a homogenous cuboid with optical properties approximating that of a lymph node (μ_a = 0.2 cm⁻¹, μ_s’ = 10 cm⁻¹, g = 0.9), as previously described by our group.¹² Monte carlo simulations were done using the publicly available MCMatlab toolbox¹³ and accelerated on a NVIDIA GeForce RTX 2080Ti GPU. 10¹⁰ photons were launched for each widefield simulation while 10⁶ photons were launched for each individual detector simulation. To accommodate the memory-constraints of the system, projections were binned down to 65x87 elements prior to reconstruction and the simulations were set up to accommodate these binned down image sizes.

Binding potential maps were calculated from the reconstructions using the following equation (1), where BP corresponds to binding potential, R_t corresponds to the targeted-agent reconstruction (ABY-029), and R_u corresponds to the untargeted agent reconstruction.

$$\text{BP}=\frac{{\text{R}}_{\text{t}}}{{\text{R}}_{\text{u}}}-1$$ (1)

3. RESULTS

Results of the 3D reconstruction can be seen in Figure 1, where the absorption reconstruction (thus the lymph node volume) can be seen in gray. Slices transecting the spheroid reconstruction were selected and displayed showing the corresponding binding potential maps.

Figure 1:

Lymph node absorption (top left) and selected slices displaying binding potential reconstructions (top right, middle right, and bottom).

4. DISCUSSION

This proof-of-concept successfully demonstrates the feasibility of fluorescence optical projection tomography for the localization of micro metastases in a clinically useful time frame. The reconstruction presented here (n=1) demonstrates that a 200-micron human cancer cell spheroid can be successfully detected and reconstructed using the described in-house system and reconstruction approach. Future work will consist of 1) refinement of the reconstruction algorithm to reduce the impact of imaging artifacts on the final reconstruction and 2) correlation of results with subsequent histopathology data to verify spheroid localization in the final reconstruction.