In vivo capture and label-free detection of early metastatic cells

Scaffolds for in vivo capture of metastasizing cells

An orthotopic model of human breast cancer metastasis was employed to investigate the capture of metastatic cells at a biomaterial implant. tdTomato- and luciferase-expressing MDA-MB-231BR (231BR) cells, a highly metastatic variant of the MDA-MB-231 cell line²¹, were transplanted into the right mammary fat pads of female NOD/SCID-IL2Rγ^−/− (NSG) mice. One week after tumor inoculation, microporous PLG scaffolds (5 mm diameter, 2 mm height, Supplementary Fig. 1a–c) were implanted into the peritoneal fat pads, a site to which 231BR cells are not reported to colonize, yet supports the vascularization of PLG scaffolds. Bioluminescence imaging (BLI) and histological analysis of peritoneal fat pads removed 28 days after tumor inoculation demonstrated the presence of tumor cells within the implanted PLG scaffold (Fig. 1a,d) and the absence of tumor cells in fat pads without scaffolds (Fig. 1b,e), indicating that the local environment generated by implantation of the scaffold enabled recruitment of metastatic cells. Primary tumor growth was not affected by either implantation of scaffolds or a mock surgery (Supplementary Fig. 2). Staining for fibronectin, a matrix protein reported to be involved in establishment of the pre-metastatic niche⁸, indicated that fibronectin was present in scaffolds implanted in both healthy and tumor-bearing mice as early as 7 days post-implantation (Supplementary Fig. 1d). Interestingly, recruitment of cells to the scaffold was not site-specific, as tumor cells were detected in scaffolds implanted in the subcutaneous tissue (Supplementary Fig. 3).

Scaffolds reduce tumor burden in solid organs

We subsequently investigated whether capturing tumor cells in scaffolds would reduce colonization of standard metastatic sites, such as the lung and liver. At 28 days post-tumor inoculation, the relative abundance of tumor cells, reported as the ratio of tdTomato-positive tumor cells to total cells, was determined. For mice that received scaffolds, the relative abundance of tumor cells in the lung was 1:5,400, compared to 1:645 for mice receiving a mock surgery (Fig 2a, Supplementary Fig. 4). Thus, the presence of a scaffold reduced the tumor burden for the lung by 88 ± 7% (average ± SEM). Histological analysis of lung sections confirmed a reduction in the tumor cell burden with scaffold implantation (Fig. 2b,c), with an average of 1.7 ± 0.5 metastatic lesions per section observed in the lungs of scaffold-bearing mice, compared to 5.5 ± 1.7 lesions per section in mice receiving mock surgeries. Furthermore, flow cytometric analysis of cells isolated from the liver showed detectable tumor cells in 8 out of 8 mice receiving mock surgeries, while mice receiving scaffold implants only exhibited detectable tumor cells in 2 of 8 livers (P < 0.01, Fisher’s exact test).

Early detection of tumor cells in scaffold

The potential to use scaffolds for early detection of metastasis was determined by quantifying the percentage of tumor cells in intraperitoneal and subcutaneous scaffolds compared to the lung and liver at day 14 post-tumor inoculation. In a group of eight mice, most intraperitoneal scaffolds (15/16) contained tumor cells at this time point, while none of the mice had detectable tumor cells in the lung and liver (Fig. 3a). In a separate group of mice, all subcutaneous scaffolds (10/10) contained tumor cells. The incidence of detectable metastatic disease at this early time point was lower than at day 28 post-tumor inoculation. At day 28 post tumor inoculation in scaffold-bearing mice, the lung and liver exhibited tumor cells in 8 and 2 of the 8 mice, respectively. Furthermore, for mice receiving mock surgeries instead of scaffold implants, the incidence of metastatic cells in both the lung and liver increased to 8 out of 8 mice. Importantly, at day 14 post-inoculation, while none of the lungs and livers exhibited detectable tumor cells, both intraperitoneal and subcutaneous scaffolds had a detectable percentage of tumor cells (0.019 ± 0.005% for intraperitoneal scaffolds and 0.044 ± 0.017% for subcutaneous scaffolds) (Fig. 3b, Supplementary Fig. 5). This ability to detect tumor cells in the scaffold prior to detection in the lungs and liver may enable the early detection of metastatic disease through imaging the scaffold.

Label-free detection of metastasis at the scaffold

Inverse spectroscopic optical coherence tomography (IS-OCT)²² was applied to directly visualize the scaffold architecture and provide quantitative measurement of the ultrastructural changes induced by the cancer cells. IS-OCT is a light scattering-based technique capable of non-invasive 3-D imaging of tissue morphology with micron-level resolution and millimeter-level penetration depth^23,24. In addition, for each 3-D resolution voxel (15 × 15 × 2 μm) IS-OCT also performs a spectroscopic analysis and quantifies the power of the spectra by a scattering model I(λ)∝λ^{D-4 22}. D is the shape factor that physically defines the macromolecular density correlation function for a range of length scales from ~40 nm to 350 nm²⁵, with higher D values indicating a more clumped structure. It has been demonstrated that D is a ubiquitous marker of the ultrastructural alterations in the early stages of various cancer types despite their different etiologies^26–30, with both neoplastic cells and the surrounding stroma exhibiting an increase in D in part due to chromatin condensation and collagen remodeling, respectively^27,30,31. Given the nanoscale sensitivity of measuring D and the tissue-level imaging capability, we hypothesized that IS-OCT could be an effective approach for detection of cancer cells within the scaffold.

In vitro studies were performed to demonstrate that IS-OCT could capture changes in D for cells and matrices, and the technique was subsequently applied to in situ imaging of scaffolds. IS-OCT analysis of 231BR cell pellets confirmed that they had a higher D than normal mammary epithelial cells (MCF10A) and cells isolated from lungs of tumor-free NSG mice (3.49 ± 0.12, 2.74 ± 0.15, and 3.00 ± 0.13, respectively, Supplementary Fig. 6a). Changes to collagen remodeling by 231BR cells were evaluated by culturing 231BR cells in collagen gels for 3 days, after which cells were extracted and the gels were analyzed using IS-OCT. 231BR-conditioned matrices exhibited a D value of 1.69 ± 0.08 compared to a D value of 1.29 ± 0.04 for gels cultured with media (Supplementary Fig. 6b). With confirmation of the technique in vitro, in situ IS-OCT analysis was applied to scaffolds, which demonstrated that scaffolds implanted in the subcutaneous tissue of tumor-bearing mice also had an increase in D at day 14 post-tumor inoculation compared to control scaffolds in tumor-free mice (Fig. 4), with an average D value of 5.77 ± 0.38 in tumor-bearing mice compared to 4.71 ± 0.17 in tumor-free mice. This increase in D is consistent with the changes associated with the presence of cancerous cells and the ensuing reorganization of the extracellular matrix. These results indicate that this method can be used for label-free detection of micrometastases within the scaffold at the early stages of metastatic disease.

Immune cells contribute to tumor cell recruitment

Given the critical role of various immune cells types in establishing the pre-metastatic niche^{8,10,12,16,19,20}, we hypothesized that the immune response to the scaffold was mediating recruitment of tumor cells. For analysis of the immune environment within the scaffolds, we utilized an immune-competent mouse model in addition to the NSG model in order to account for effects of both the innate and adaptive immune response. Scaffolds implanted into BALB/c mice inoculated with 4T1 mouse breast cancer cells also demonstrated metastatic cells within the scaffold, indicating that the scaffold could still achieve homing within the context of an intact immune system (Supplementary Fig. 7).

Inflammatory cells proposed to be involved in recruiting tumor cells were characterized within the peritoneal fat pads of mice in the presence and absence of a scaffold. A high density of CD45-positive leukocytes was present in histological sections of the scaffold, with no observed CD45-positive leukocytes present in fat pads of mice receiving mock surgeries (Fig. 5a,b). The ability of CD45-positive leukocytes to influence homing of tumor cells was investigated through migration assays using media conditioned by splenocytes isolated from spleens of tumor-bearing mice, as the spleen contains a large number of immune cells and has a distribution of immune cells similar to the scaffolds, with the predominant cell type being Gr1^hiCD11b⁺ cells (Supplementary Fig. 8). Migration of both cell types was significantly increased in the presence of splenocyte-conditioned media relative to unconditioned media (Fig. 5c,d), with 261 ± 35 migrating 231BR cells per well in conditioned media compared to 137 ± 20 cells in unconditioned media and 521 ± 22 migrating 4T1 cells per well in conditioned media compared to 292 ± 23 cells in unconditioned media. These results indicate that paracrine signaling from immune cell populations similar to those present in the scaffold can induce migration of tumor cells.

Before and after the introduction of cancer cells, we analyzed the local immune environment of the implanted scaffold, which was compared to that of the lung, a common site of breast cancer metastasis. Flow cytometric analysis demonstrated that during disease progression, the most notable change in the relative distribution of immune cells was an increase in Gr1^hiCD11b⁺ cells (Fig 5e–h, Supplementary Fig. 9–11). This increase was consistent in both the NSG and BALB/c mouse models, and was found in both the lungs and scaffolds. In the lungs of the NSG mice, the percentage of Gr1^hiCD11b⁺ cells increased from 45 ± 2% at day 0 (no cancer cells) to 52 ± 3% at day 14-post cancer cell inoculation, to 84 ± 1% at day 28 post cancer cell inoculation (Fig 5e), Following this trend, the percentage of Gr1^hiCD11b⁺ cells found in the scaffolds from the NSG mice increased from 21 ± 1% (tumor-free mice), to 35 ± 5% at day 14 post cancer cell inoculation, to 62 ± 2% at day 28 post cancer cell inoculation (Fig 5g). In the lungs of BALB/c mice, Gr1^hiCD11b⁺ cells increased from 1 ± 0.1% at day 0 (tumor-free mice) to 57 ± 1% at day 14-post cancer cell inoculation, to 89 ± 2% at day 28-post cancer cell inoculation (Fig. 5f). Likewise, within the scaffolds from the BALB/c mice, Gr1^hiCD11b⁺ cells increased from 0.1 ± 0.01% (no cancer cells) to 32 ± 3% at day 14-post cancer cell inoculation, to 66 ± 5% at day 28-post cancer cell inoculation (Fig. 5h). These results highlight the correlation between disease exposure time and the relative abundance of Gr1^hiCD11b⁺ cells found in both the lungs and scaffolds of affected animals, and show that the immune environment within a metastatic organ site is similar to that of an implanted scaffold.

We subsequently investigated the hypothesis that modulation of the inflammatory microenvironment within the scaffold site would influence recruitment of metastatic cells. Lentiviral vectors were delivered from the scaffold to promote localized transgene expression for the duration of the study (Supplementary Fig. 12). The chemokine CCL22 was selected for expression as it induces migration of splenocytes harvested from NSG mice (Supplementary Fig. 13a) yet does not influence the migration of tumor cells (Supplementary Fig. 13b). Flow cytometric analysis of CCL22 scaffolds at day 7 post-implantation indicated an increase in Gr1^hiCD11b⁺ cells, which have been implicated in the pre-metastatic niche^10,14,15, from 20 ± 2% to 29 ± 3% in NSG mice and from in 14 ± 1% to 19 ± 1% BALB/c mice (Fig. 6a,c). Additionally, F4/80⁺CD11b⁺ inflammatory macrophages, which are recruited to circulating metastatic cells as they undergo extravasation and begin colonization¹², increased from 28 ± 2% to 34 ± 2% in the NSG model and from 10 ± 1% to 14 ± 1% in the BALB/c model (Fig. 6a,c). Furthermore, CCL22 expression increased the number of tdTomato-positive tumor cells present in the scaffold in both mouse models at day 7 post-implantation (day 14 post-tumor inoculation), with an average of 92 ± 16 cells compared to an average of 23 ± 4 cells for β-galactosidase expression in NSG mice and an average of 62 ± 5 cells compared to an average of 20 ± 2 cells for β-galactosidase expression in BALB/c mice (Fig. 6b,d).

A cell population upregulated by CCL22 delivery, Gr1^hiCD11b⁺ cells, was next transplanted on the scaffold to directly modulate the local immune environment and further demonstrate that the immune environment mediates metastatic cell recruitment. Gr1^hiCD11b⁺ Gr1^hiCD11b⁺ cells were selected for these studies given their established role in the pre-metastatic niche, along with migration assays showing soluble factors secreted by MDSCs isolated from spleens of tumor-bearing NSG mice induced migration of 231BR cells (Fig. 6e). Interestingly, the Gr1^hiCD11b⁺ cells induced migration to a greater extent than splenocyte-conditioned media (Supplementary Fig. 14). Implantation of scaffolds seeded with Gr1^hiCD11b⁺ cells significantly increased the percentage of Gr1^hiCD11b⁺ cells in the total leukocyte population (from 20 ± 2% to 26 ± 2%, Fig. 6f) as well as the number of tumor cells (from 29 ± 10 to 74 ± 13 cells, Fig. 6g) present per scaffold, indicating that Gr1^hiCD11b⁺ cells contributed to recruitment of tumor cells to the scaffold.

Study Design

The objective of the study was to utilize biomaterials scaffolds and optical imaging techniques in order to capture and detect metastasizing breast cancer cells in vivo. We hypothesized that due to the local inflammatory response following implantation, the scaffolds become infiltrated with immune cells conducive to recruiting metastatic cells, and that a light scattering-based imaging method could detect changes to the tissue ultrastructure associated with arrival of the cancer cells. Animal studies were performed with at least 2 independent replicates of 4 to 8 female 10 to 14 week old mice per group with random assignment. Tissues were analyzed at day 14 or 28 post-tumor inoculation, with day 14 being the earliest timepoint for detecting metastasis and day 28 being the final timepoint, as the primary tumor reached the maximum allowable size. In vitro experiments had multiple samples that were independently repeated at least 2 times.

Tumor inoculation and volume measurement

Animal studies were performed in accordance with institutional guidelines and protocols approved by the Northwestern University Institutional Animal Care and Use Committee (IACUC). Tumor inoculation was performed by injecting 2 × 10⁶ 231BR (Northwestern University Developmental Therapeutics Core) or 4T1 (ATCC) cells in a volume of 50 μL PBS (Life Technologies) into the number four right mammary fat pads of 10–14 week old female NSG or BALB/c mice. NSG mice were purchased from The Jackson Laboratory or bred in house. BALB/c mice were purchased from The Jackson Laboratory. Tumor length and width were measured weekly using digital calipers (VWR) beginning at day 14 post-inoculation, and tumor volume was calculated using the formula Volume = (Width² × Length)/2.

Scaffold fabrication and implantation

Poly(lactide-co-glycolide) (PLG) scaffolds were fabricated as previously described⁴⁰. Briefly, microspheres were prepared by emulsifying a 6% solution of PLG (Lakeshore Biomaterials; 75:25 lactide:glycolide, i.v. = 0.76 dL/g) in dicholoromethane in 1% poly(vinyl alcohol) (PVA). Microspheres were washed four times with deionized water to remove residual PVA and lyophilized overnight. Next, microspheres were mixed with 250–425 μm salt particles in a 30:1 ratio and pressed in a steel die at 1500 psi. The scaffolds were then gas-foamed and salt particles were removed by washing in water for 90 minutes. Scaffolds were sterilized with 70% ethanol and rinsed with water before drying. Incorporation of a viral vector was performed by incubating 1 × 10⁹ viral particles in a volume of 10 μL PBS with the scaffold for 8 minutes. In these studies, tumors were inoculated at day 0 and scaffolds were implanted in the intraperitoneal fat pads or subcutaneous space at day 7.

Virus production

The lentiviral vector of interest was co-transfected with the packaging vectors pRSV-Rev, pIVS-VSV-G, and pMDL-GagPol⁵⁰ into HEK-293T cells grown in DMEM (Sigma) containing 10% fetal bovine serum (Life Technologies) for virus production using Lipofectamine 200 (Life Technologies). After 48 hours the supernatant was collected. Viral particles were concentrated using PEG-it (Systems Biosciences) and resuspended in PBS. Virus titer was quantified using the qPCR Lentivirus Titration Kit (Applied Biological Materials) according to the manufacturer’s instructions.

Bioluminescence imaging

Five minutes prior to imaging, mice were injected IP with D-luciferin (Molecular Imaging Products; 20 mg/mL in PBS) at 150 mg/kg body weight. For whole animal imaging, mice were anesthetized with inhaled isoflurane and in vivo luciferase expression was evaluated using the IVIS Spectrum imaging system (Caliper Life Sciences). For imaging of tissues, mice were euthanized five minutes after injection with D-luciferin and tissues were retrieved and imaged using the IVIS Spectrum.

Histological analysis and immunohistochemistry

Organs and implanted scaffolds were retrieved from mice, rinsed in PBS and then immediately flash-frozen in pre-chilled isopentane. The frozen tissue was then embedded in optimal cutting temperature (OCT; Cardinal Health) compound with 30% sucrose. The lungs were sectioned at 12 μm, and the scaffolds and fat pads were sectioned at 14 μm thick sections using a cryostat (Microm HM 525; Microm International) and stored at −20 °C until staining. In order to determine the presence of tumor cells, slides were fixed with 4% paraformaldehyde (Sigma) for 10 minutes and stained with Gill III Hematoxylin (Leica) and Eosin Y (Leica). Blinded scoring of the number of tumor clusters in each tissue section was performed by a pathologist. Two sections were analyzed per mouse, and in each of 2 separate experiments, sections from 3 mice per group were analyzed, for a total of 6 animals per condition. The sections analyzed were non-adjacent and separated by at least 5 mm. For immunohistochemistry, sections were fixed for 12 minutes in 4% paraformaldehyde and stained with a polyclonal rabbit anti-CD45 primary antibody (1:100, Abcam) followed by a CF633 anti-rabbit secondary antibody (1:500, Sigma) or a polyclonal rabbit anti-fibronectin antibody (1:250, Abcam) followed by an Alexa Fluor® 488 anti-rabbit secondary antibody (1:500, Life Technologies). Nuclei were labeled with Hoechst 33250 (1:2000, Invitrogen). Samples were imaged using a Leica SP5 II laser scanning confocal microscope.

Flow cytometry

Mice were euthanized and tissues were retrieved, minced with microscissors in a 0.38 mg/mL solution of Liberase TL or TM (Roche Applied Science) in Hank’s Balanced Salt Solution (HBSS; Life Technologies) and placed at 37 °C for 20 minutes. Liberase was neutralized with 0.125M EDTA, and cells were isolated by passing the digested tissue through a 70 μm filter (BD Biosciences) and rinsing with FACS buffer, PBS containing 0.5% Bovine Serum Albumin (BSA; Sigma) and 2mM EDTA (Life Technologies). Red blood cells were lysed using ACK lysing buffer (Life Technologies). For analysis of tdTomato-positive tumor cells, samples were resuspended in FACS buffer and analyzed using an LSR II flow cytometer (Becton Dickinson Immunocytometry Systems, BDIS). A standard curve was established for each organ by adding defined numbers of tumor cells into cells isolated from healthy lungs and livers and measuring the number of tdTomato-positive cells via flow cytometry. From this analysis, it was determined that the threshold tumor cell density of 0.002% (5 tumor cells in 250,000 total cells) was consistently and significantly above background, and thus samples were deemed to have detectable cancer cells if the tumor cell density was above 0.002%. For analysis of leukocyte populations, cells were blocked with anti-CD16/32 (1:50, eBioscience) and stained for viability using fixable blue dead cell stain kit (Life Technologies). Cells were then stained with Alexa Fluor® 700-conjugated anti-CD45 (30-F11, 1:125; Biolegend), Pacific Blue-conjugated anti-Gr-1 (RB6-8C5, 1:70; Biolegend), FITC-conjugated anti-Ly-6C (HK1.4, 1:100; Biolegend), PE-Cyanine7-conjugated anti-F4/80 (BM8, 1:70; Biolegend), APC-conjugated anti-CD11c (N418, 1:85; eBioscience), v500-conjugated anti-CD11b (M1/70, 1:100; BD Biosciences), Pacific Blue-conjugated anti-CD19 (6D5, 1:100; Biolegend), v500-conjugated anti-CD4 (RM4-5, 1:100; BD Biosciences), FITC-conjugated anti-CD8a (53-6.7, 1:25; Biolegend), and PE-Cyanine7-conjugated anti-CD49b (DX5, 1:40; Biolegend). Samples were analyzed using an LSRFortessa flow cytometer (BDIS).

Splenocyte isolation for transplants and media conditioning

Spleens were isolated from tumor-bearing mice, minced using microscissors, and incubated in a 0.38 mg/mL solution of Liberase TL (Roche Applied Science) for 20 minutes at 37 °C. The minced tissue was passed through a 70 μm strainer and red blood cells were lysed using ACK lysing buffer (Life Technologies). To collect myeloid-derived suppressor cells (MDSCs), splenocytes were labeled with FITC-conjugated anti-Gr-1 (RB6-8C5, 1:100; Biolegend), APC-conjugated anti-Ly-6C (HK1.4, 1:50; Biolegend), and APC-eFluor780-conjugated anti-CD11b (M1/70, 1:100; eBioscience), and the Gr1^hiCD11b⁺Ly6C⁻ population was collected via FACS sorting using a BD FACSAria cytometer (BDIS). The remaining splenocyte population (Gr1^hiCD11b⁺ cell-depleted) was also collected. For cell transplantation studies, scaffolds were rinsed with RPMI medium (Life Technologies) containing 5% fetal bovine serum (FBS; Life Technologies) and dried on gauze. 1 × 10⁶ Gr1^hiCD11b⁺ cells were seeded onto each scaffold in a volume of 5 μL RPMI with 5% FBS and scaffolds were implanted into the IP fat pads. For generation of conditioned media, splenocytes, Gr1^hiCD11b⁺ cells, and Gr1^hiCD11b⁺ cell-depleted splenocytes were each plated at a concentration of 1 × 10⁶ cells/mL in RPMI media. Media was conditioned for 48 hours, after which the media was sterile-filtered to remove splenocytes.

Splenocyte migration assays

Migration assays were performed in 24-well Transwell chambers with 5 μm pore size filters (Corning). 1 × 10⁶ splenocytes were placed in the top chamber in 100 μL of RPMI medium containing 10% FBS. 600 μL of RPMI medium containing 10% FBS with or without 200 ng/mL recombinant mouse CCL22 (Peprotech) was placed in the bottom chamber. After 1.5 hours, the number of cells that had migrated to the bottom chamber was counted using a hemocytometer.

Tumor cell migration assays

Tumor cells were serum-starved overnight in DMEM (Sigma) or RPMI (Life Technologies) containing 0.2% FBS (Life technologies). Migration assays were performed in 24-well Transwell chambers with 8 μm pore size filters (Corning). 5 × 10⁴ 231BR or 4T1 cells were placed in the top chamber in 300 μL of DMEM (for CCL22 assays) or RPMI (for conditioned media assays) containing 0.2% FBS. For CCL22 assays, 750 μL of DMEM containing 0.2% FBS with or without 200 ng/mL recombinant human CCL22 (Peprotech) was placed in the bottom chamber. For conditioned media, 750 μL of RPMI, splenocyte-conditioned RPMI, Gr1^hiCD11b⁺ cell-conditioned RPMI, or RPMI conditioned with Gr1^hiCD11b⁺ cell-depleted splenocytes was placed in the bottom chamber (all media was supplemented with 0.2% FBS). After 24 hours, the cells on the top of the filter were removed by gently scrubbing with a cotton swab, and the cells on the bottom of the filter were stained for 60 minutes in a 0.5% (w/v) solution of crystal violet (Sigma) in 60% PBS/40% ethanol. The inserts were rinsed in PBS and the number of migrating cells was counted.

Conditioning collagen gels with 231BR cells

A 1.6 mg/mL collagen I solution was prepared by mixing rat tail Collagen I (BD Biosciences) with deionized water, 10X PBS (Sigma, 1:10 dilution) and 1 N NaOH (Sigma, 1:1000 dilution) on ice. Collagen solution (150 μL) was placed in each well of a 48-well plate. Gels were incubated at 37 °C for 1 hour, after which 1.875E5 231BR cells in 400 μL of RPMI containing 10% FBS were seeded onto each gel. Control gels received 400 μL of RPMI containing 10% FBS. After incubation at 37 °C for 3 days, cells were extracted from the gels. The media was removed from each well and the gels were washed 3 times with 1X PBS (Life Technologies). A 20mM NH₄OH solution containing 0.5% (v/v) Triton X-100 (Sigma) was added to each well and the chamber slide was incubated at 37 °C for 5 minutes. After removal of the NH₄OH/Triton X-100 solution, the gels were washed twice with distilled water and twice with PBS. 175 μL of PBS was added to each well prior to imaging.

Scanning Electron Microscopy

Structural characteristics of scaffolds were imaged with a scanning electron microscope (Hitachi s4800-II cFEG SEM; Hitachi High-Technologies Corp.). A 15-nm gold coating was applied and the microscope was operated at 2 kV.

IS-OCT imaging and analysis

The OCT system used a Fourier domain configuration. A supercontinuum light source (NKT photonics, SuperK versa) provided a broadband laser from ~500 nm to 840 nm. The spectral range from 650 nm to 830 nm was used for OCT image construction. The laser was delivered by an optical fiber, collimated by a lens, and input into a cube beam splitter (BS; Thorlabs, CM1-BS013), by which the light was divided into a sample arm and a reference arm. The reference arm consisted of a series of glass plates for dispersion control and a mirror reflecting the light backward. The sample arm consisted of a two-dimensional scanning mirror (Thorlabs, GVSM002) and an objective lens (effective NA = 0.04; Zeiss) to focus the light onto specimens. The lateral resolution is estimated as 15 μm. The interfered light from the reference and sample arm at the BS was collected by another optical fiber and delivered to the spectrometer for spectral acquisition. The mirror was scanning at a frequency of 2.56 kHz, and a 3D acquisition lasted 25.6 s with 256 by 256 pixels in two lateral directions. The camera exposure time was 0.3 ms. With the bandwidth, the axial resolution was estimated as 2 μm.

The depth structural reconstruction was created by the following steps. The recorded spectrum was first normalized by the source spectrum, and then resampled into k domain with equal interval, and finally a Fourier transform was taken to recover the depth signal. k is the wave number, defined by k = 2π/λ. The calculation of D value was performed by the following steps. A Gaussian window with width k_w = 0.36 μm⁻¹ was applied on the spectrum, and the depth structural reconstruction process was performed. This process was then repeated with the Gaussian window scanning through the whole spectral range, and the wavelength-dependent 3-D structure was obtained. The reduced range of spectrum relaxed the axial resolution to be about 20 μm. The averaged spectra from the specimen were modeled by I(k) = k^2-D/2. The value of D was then obtained by calculating the power of the spectra.

Statistical analysis

Data are presented as mean ± standard error (SEM), and p-values were determined using a Mann-Whitney test for single comparisons. For comparison of relative numbers of mice containing metastasis to various tissues, a Fisher’s exact test was used to determine the p-value. Statistical analysis was performed using GraphPad Prism.