Current preclinical osteosarcoma (OS) models primarily focus on quantifying primary tumor growth. However, patients with OS more commonly die from metastatic disease, not their primary tumors [2, 3, 22]. Patients with lung metastases detected at the time of OS diagnosis have 5-year survival rates as low as 20% to 30% [10, 11, 18]. To improve OS survival, novel therapies must act against both primary and metastatic disease. However, the characterization of potential OS treatments suffers from a lack of preclinical models that can quantify changes in OS metastatic burden. There are myriad reasons for this shortcoming. In contrast to the hindlimb, which can be immobilized, the lungs change in size and dimension over the course of normal breathing. This can complicate conventional in vivo and ex vivo quantitative measurement techniques such as contrast-aided CT. Questions of orientation, precision, time, and financial expense plague positron emission tomography studies. Any method for quantifying tumor burden must be rapid, inexpensive, precise, and not interact with experimental treatments.
Members of our group have used indocyanine green (ICG) dye angiography in multiple preclinical applications [7, 8, 20]. ICG is a near-infrared cyanine fluorophore (excitation 802 nm, return 830 nm) previously FDA-approved for surgical and ophthalmologic applications that is known for its low toxicity and hepatic metabolism . After intravenous injection, ICG fluoresces only after binding tightly to plasma albumin and is therefore typically representative of purely intravascular flow. However, in environments with damaged or disorganized vasculature such as burns or a malignancy, ICG is able to extravasate, thereby persisting in a region of tissue long after intravascular signal has dissipated. This “postwashout” ICG fluorescence has been previously used to reliably visualize cancer, although not quantitatively [12, 13].
The quantification of ICG fluorescence within regions of tissue has been found to have both acute and long-term prognostic utility. Members of our group have shown that ICG fluorescence in and around burn wounds can predict long-term scarring ; the acute fluorescence of ICG within flaps can predict pedicle flap survival [6, 20]; and interval ICG fluorescence within wounds covered with acellular dermal matrix can track vascular invasion . The residual ICG signal that can persist for days in areas of gross vascular disruption (trauma, burns) or disorganized endothelium (tumors) has been poorly characterized. Although ICG has been previously shown to extravasate in tumor microvasculature , further characterization of this phenomenon has focused purely on signal precision. Quantification of this signal has rarely been attempted with no attempt to equate tumor quantification with disease natural history. We believed that ICG angiography was capable of quantitatively characterizing both the primary and metastatic tumor burden in a previously validated orthotopic hindlimb OS mouse model.
Our study had three main questions: (1) Can near-infrared ICG fluorescence be used to attach a comparable, quantitative value to the primary OS tumor in our experimental mouse model? (2) Will primary tumor fluorescence differ in mice that go on to develop metastatic lung disease? (3) Does primary tumor fluorescence correlate with tumor volume measured with CT?
Materials and Methods
All study protocols were approved by the Institutional Animal Care and Use Committee at our institution. All surgeries and fluorescence analyses were performed by the same investigator (MSF), who was kept blind as to study group throughout the experimental protocol. A total of 36 immunocompetent Balb/c 4- to 6-week-old mice (RRID:IMSR_JAX:000651) received paraphyseal injections of 0, 1 x 105, 2.5 x 105, 5 x 105, 7.5 x 105, or 1 x 106 K7M2 mouse OS cells (n = 6 per group) (RRID:CVCL_V455) into their left hindlimb proximal tibia using a previously published technique (Fig. 1) . We selected K7M2 OS cells because of their previously demonstrated prometastatic activity and their prior characterization in similar orthotopic models . Cells were diluted in phosphobuffered saline into an injectable solution of 20 μL in all cases. Overall tumor formation took place in 27 of 36 animals (75%), confirmed clinically and with CT. No differences were observed in the rate of tumor growth in any cell injection group. No control animals grew tumors.
Primary Tumor Amputation
Four weeks after OS injection, all animals underwent left hindlimb transfemoral amputations. Animals were anesthetized using 2.5% isoflurane via an induction chamber and a nose cone. Surgical tools were sterilized using a bead sterilizer at 233° C (CellPoint Scientific, Gaithersburg, MD, USA). Animals were laid supine with both hindlimbs taped in a fully extended position. All hair on their hindlimbs was removed with depilatory cream (Nair™; Church & Dwight, Trenton, NJ, USA). The skin overlying the inguinal ligament was incised, and the femoral neurovascular bundle proximal to the branching of the superficial vessels was visualized. Suture ligation of the vascular bundle was performed using 4-0 Prolene® (Ethicon, Somerville, NJ, USA). The limb was then sharply amputated through the midfemoral diaphysis using a No. 10 blade scalpel. Hemostasis was obtained and closure was performed using 7-mm surgical clips (Braintree Scientific, Braintree, MA, USA).
Study Endpoint and Lung Extraction
Ten weeks after K7M2 OS hindlimb injection, animals were euthanized by CO2 asphyxiation followed by cervical dislocation. After confirmation of euthanasia, a longitudinal incision through the animal’s thorax and abdomen was made. The inferior vena cava, superior vena cava, and mediastinal structures were cut immediately superior and inferior to the lungs. The remainder of the mediastinum was carefully removed with a hemostat in a manner that permitted the undamaged lungs to be removed as a unit.
Indocyanine Green Fluorescence Measurements
Fluorescence measurements with the SPY-Elite (Novadaq, Bonita Springs, FL, USA) were performed on the hindlimb immediately before amputation and on the lungs ex vivo immediately after extraction. ICG was administered to each animal through retroorbital injection 24 hours before each measurement with 30 µL of 2.5 mg/cc ICG (IC-Green®; Novadaq). All injections were performed while being monitored by the SPY-Elite to ensure acceptable dye uptake, although formal quantitative measurements were not performed at that time. Primary tumor measurements were taken with the animal laid prone, before sharp amputation. The on-board laser target permitted the device sensor to be placed above the animal’s dorsum immediately proximal to the base of its tail, permitting simultaneous standardized measurements of both the hindlimb and tail. Primary tumor quantitative fluorescence was calculated as the average of three standardized points within the animal’s hindlimb, as measured with the SPY-Q (Novadaq) software preloaded onto the SPY-Elite (Fig. 2A-B). The maximum fluorescence measured at the base of the animal’s tail was used as the normalization metric in an adaptation of a previously published technique developed by members of our group . Measurement of metastatic tumor burden was performed with the sensor positioned above the carina of the lungs. High-quality jpeg images were exported with the pixel contrast scaled to a midpoint of 64 arbitrary perfusion units (APU; Fig. 3A-B). The APU, a unit of measure independently defined by the manufacturer of the SPY apparatus and utilized in the detector and analysis software (SPY-Q; Novadaq), is a normalized, reproducible measurement of pixel return by the device, scaled to a measure of 0 to 255. This value has been used consistently and reproducibly in prior work by our group and others involving animal and human models of soft tissue pathology [5-9, 20]. APU is also output as a blue to red map of relative perfusion. Using the region-of-interest feature, whole lung fluorescence as shown in these images was calculated as average quantitative gray saturation with ImageJ (NIH, Bethesda, MD, USA) with the background subtracted.
CT of Hindlimbs
CT was performed on formalin-preserved hindlimbs using the FIDEX veterinary imaging apparatus. Axial, coronal, and sagittal cuts were used to measure tumor volume in orthogonal planes using OsiriX software (Pixmeo, Bernex, Switzerland). Volumetric measurements were calculated using the equation for the volume of an ellipse: 4/3*π*r1*r2*r3.
Freshly harvested leg samples were immersed in tissue-freezing medium and flash frozen with liquid nitrogen for 60 seconds in a 2-methylbutane bath. Legs with intact tumors were sectioned at 6-μm thickness on Superfrost™ Plus microscope slides (ThermoFisher Scientific, Waltham, MA, USA). Fluorescence signal was detected on unmounted slides using an Olympus IX-81 (Olympus Corporation, Shinjuku, Tokyo, Japan) outfitted with an ICG cube (excitation 775; emission 845), LB-LS/OF30R with excitation intensity of 5% at 775 nm, and Retiga EXi Aqua digital camera (Qimaging, Surrey, British Columbia, Canada). Images were captured and processed using MetaMorph® software (Molecular Devices, Sunnyvale, CA, USA).
With the aid of a blinded reader (PEA) experienced with sectioning and histologic analysis, ICG was found to localize specifically within the tumor in both primary (Fig. 4A) and metastatic disease (Fig. 4B) samples. No false-positive ICG signal within normal tissue was identified. No sample bleed was noted when the same slides were assessed 48 hours and 2 weeks after sectioning, suggesting that long-term sample preservation may not impact the localization or fluorescence of the dye.
To answer our first research question, which was to quantify the primary tumor using fluorescence angiography, we performed ICG fluorescence measurements of the primary tumor immediately before amputation 4 weeks after OS injection and correlated these measurements with clinically diagnosed disease. To answer our second research question, in which we sought to use ICG angiography to highlight lung metastases, we quantified metastatic tumor burden by performing similar fluorescence measurements ex vivo on the lungs of each mouse 10 weeks after OS injection, correlating these measurements with clinically observed OS metastases. Finally, to answer our third question, which sought to correlate primary tumor fluorescence with clinical tumor volume, we compared three-dimensional primary tumor volume measurements measured on CT with primary tumor fluorescence through linear regression.
All statistical analyses were performed by the investigators using Prism 7.0 (GraphPad, La Jolla, CA, USA). Differences in CT measurements, normalized hindlimb fluorescence, and lung fluorescence were evaluated between OS dose groups using one-way analysis of variance. Linear regressions were performed between hindlimb and lung fluorescence as well as hindlimb CT measurements and primary tumor fluorescence. In both regression analyses, all animals were considered a single group. All continuous data are presented as mean ± SD. In all cases, p < 0.05 was considered significant.
Mice with clinical tumor growth had greater hindlimb fluorescence (3.5 ± 2.3 APU) compared with those without tumors (0.71 ± 0.38 APU, -2.7 ± 0.5 mean difference, 95% confidence interval [CI] -3.7 to -1.8, p < 0.001; Fig. 5). When only animals with primary tumor growth were considered, hindlimbs injected with 1 million OS cells exhibited greater fluorescence (10.7 ± 8.8) than all other experimental groups (9.5 mean difference, 91% CI 1.2-17.9, p < 0.05; Fig. 6). No other between-group differences were observed.
Twenty-five mice were diagnosed with metastatic OS after clinical lung evaluation after ex vivo harvest (Fig. 7A-B). All had previously developed primary tumors. Lung fluorescence was assessed as a function of hindlimb tumor fluorescence. A linear trend (r2 = 0.81, p < 0.01; Fig. 8) was noted between primary tumor and lung fluorescence. However, no trend was noted when lung fluorescence was stratified by cell injection group.
There was no correlation (r2 = 0.04, p = 0.45; Fig. 9) observed between normalized tumor fluorescence and tumor volumetric measurements with CT.
OS with metastases to the lungs at the time of diagnosis is associated with a drastically lower 5-year disease survival rate. Advances that may improve our ability to treat advanced disease are limited because of our inability to reliably quantify OS in preclinical models. We sought to characterize ICG angiography as a novel modality for quantifying experimental primary and metastatic OS burden. We found that (1) ICG is capable of characterizing primary tumor burden in vivo and metastatic burden ex vivo; (2) ICG histologically localizes to the tumor; and (3) ICG hindlimb fluorescence is not dictated by actual tumor size. These findings highlight the uniqueness of our analysis technique and suggest the utility of near-infrared fluorescent imaging in both preclinical and clinical applications.
There are several limitations to our model. The first is the intrinsic two-dimensional limitation of ICG angiography as a technique. Inconsistently placed measurement points or drastically altered tumor location may lead to altered fluorescence measurements. We account for this in our model by targeting OS cell injections to the same consistent area on the dorsal hindlimb using a standardized administration technique. Although the penetration depth of ICG of up to 1 cm  is likely to appropriately account for the depth of any tumor in this model, further confirmatory analysis of consistent tumor measurements must be performed before the formal validation of our technique in other animals. Although we have shown histologically that ICG localizes specifically to the OS tumor bed, this localization appears to be unrelated to tumor size. This has been demonstrated in humans with myriad tumor specimens and subcentimeter disease but without histologic validation . Fluorescence measurements in APU do not reflect absolute values, but rather a relative pixel return normalized to a machine-dictated scale of 0 to 255. We therefore concede that we can only calculate relative relationships with ICG angiography. This may not be a true shortcoming preclinically given the anticipated use of this model to evaluate the impact of novel drugs on the same tumor environments. Absolute units would therefore be unnecessary. However, it does limit the quantitative features of ICG angiography to the purely experimental realm, in which the thickness of small animal limbs and the general anatomy of the tumor are largely consistent. We anticipate the translation of ICG in orthopaedic oncology to be largely qualitative in nature and based on its localization to the tumor bed. We propose that it may eventually serve as a qualitative, intraoperative margin detector, reducing the need for intraoperative pathology and the long delays inherent to that process. Finally, although this murine orthotopic OS model has been previously described [4, 16], no work has fully substantiated the exact route of OS cells during metastasis. The paraphyseal injection of a suspended group of cells has not been shown to localize only to the injection site. It is possible that lymphatic drainage could lead to almost immediate metastasis, which would therefore not permit a reasonable representation of human OS pathogenesis. Although this will be evaluated using green fluorescent protein-tagged OS cells in future work, this shortcoming must be understood before therapeutic analyses using our model.
Although not necessarily limitations, several characteristics of our model must also be taken into account during the planning and execution of future work using this model and ICG angiography. A tumor growth rate of 75% is acceptable for an immunocompetent model, because this reflects the expected variability secondary to an active immune system in individual animals. It does, however, require future work to include “expected tumor negatives” into their sample size calculations. This limitation may permit instead an expected “negative growth” group that can be used in additional experiments and analyses. Furthermore, only 25 of 27 mice with primary tumors developed metastatic disease. This additional “expected negative rate” will need to be applied when utilizing this model in treatment versus control studies.
We show here that ICG is capable of quantifying primary tumor burden in vivo and metastatic burden ex vivo in a previously validated orthotopic mouse model. However, the use of ICG angiography in oncologic applications is not novel. Prior work has identified the utility of residual fluorescent imaging in adenocarcinoma models, identifying 24 hours after dye administration as the critical point at which the signal:background ratio of ICG fluorescence was optimal for tumor imaging . Translation of these findings into sarcoma models was first proposed by Judy and colleagues . However, this was limited to qualitative primary tumor imaging. This limitation is representative of the specific aims of most prior studies of ICG in oncologic applications. Qualitative margin optimization has long been the priority of near-infrared tumor imaging. No prior work has focused on the prognostic importance of quantitative ICG signal in vivo.
We found that primary tumor fluorescence was linearly related to metastatic tumor burden in our OS model. In contrast, there was no relationship between tumor fluorescence and tumor size alone. This finding suggests that although tumor size is likely a partial determinant of primary tumor fluorescence, fluorescence represents a unique tumor property in and of itself. Furthermore, the traits illustrated by quantitative ICG signal may be more important to OS metastasis than primary tumor size alone. Because ICG must be able to travel to an extravasation site to exit the tumor microvasculature, it may be a reasonable assumption that regions that have greater amounts of endothelial interruption will permit a greater amount of dye extravasation. To our knowledge, this is the first time that a fluorescence relationship between OS primary and metastatic activity has been noted. The theory of endothelial disorganization as a prognostic factor in tumor metastasis has been previously studied and may also be an indicator of changes in the tumor microenvironment and extracellular matrix composition and function . Based on our findings and the conclusions of prior work, we hypothesize that OS metastasis is less dependent on tumor size than it is on tumor neoangiogenesis. Further characterization and confirmation of our hypothesis as well as a concurrent evaluation of genetic tumor alterations and differences in the phenotypic expression of pro- and antiangiogenic factors between primary and lung OS will be the subject of future work.
In conclusion, we found that ICG angiography is capable of quantifying metastatic and primary OS in a preclinical OS model. The magnitude of ICG fluorescence appears to be related to many factors, but is poorly related to clinical tumor size. Although our findings may be limited to purely preclinical, experimental work, the specific localization of postwashout ICG to the OS tumor bed portends significant potential for ICG angiography as an intraoperative positive margin detector. Future work will entail the validation of ICG’s reliable localization to all components of the tumor, and clinical translation mandates a long-term observational study to evaluate the importance of residual positive signal as prognostic of local or remote recurrence. Additional experimental work will use our quantitative model to evaluate the efficacy of novel OS small molecule inhibitors on reducing primary and metastatic disease burden, permitting the advancement of new therapies to clinical trials.
We thank Drs Dan March and Rashmi Agarwal for their help in developing the current animal model of OS.
1. Allen RC, Oetting TA. Indocyanine green toxicity. Ophthalmology. 2007;114:197; author reply 197.
2. Anderson ME. Update on survival in osteosarcoma. Orthop Clin North Am. 2016;47:283–292.
3. Bacci G, Longhi A, Versari M, Mercuri M, Briccoli A, Picci P. Prognostic factors for osteosarcoma of the extremity treated with neoadjuvant chemotherapy: 15-year experience in 789 patients treated at a single institution. Cancer. 2006;106:1154–1161.
4. Cole HA, Ichikawa J, Colvin DC, O'Rear L, Schoenecker JG. Quantifying intra-osseous growth of osteosarcoma in a murine model with radiographic analysis. J Orthop Res. 2011;29:1957–1962.
5. Fourman MS, Gersch RP, Levites HA, Phillips BT, Bui DT. Is there a right way to interpret SPY? Normalization of indocyanine green angiography readings in a burn model. Plast Reconstr Surg. 2015;136:128e–130e.
6. Fourman MS, Gersch RP, Phillips BT, Nasser A, Rivara A, Verma R, Dagum AB, Rosengart TK, Bui DT. Comparison of laser Doppler and fMcFarlane flap. Ann Plast Surg. 2015;75:102–107.
7. Fourman MS, McKenna P, Phillips BT, Crawford L, Romanelli F, Lin F, McClain SA, Khan SU, Dagum AB, Singer AJ, Clark RA. ICG angiography predicts burn scarring within 48 h of injury in a porcine vertical progression burn model. Burns. 2015;41:1043–1048.
8. Fourman MS, Phillips BT, Crawford L, McClain SA, Lin F, Thode HC Jr, Dagum AB, Singer AJ, Clark RA. Indocyanine green dye angiography accurately predicts survival in the zone of ischemia in a burn comb model. Burns. 2014;40:940–946.
9. Fourman MS, Phillips BT, Fritz JR, Conkling N, McClain SA, Simon M, Dagum AB. Laser-assisted indocyanine green dye angiography accurately predicts the split-thickness graft timing of integra artificial dermis. Ann Plast Surg. 2014;73:150–155.
10. Friebele JC, Peck J, Pan X, Abdel-Rasoul M, Mayerson JL. Osteosarcoma: a meta-analysis and review of the literature. Am J Orthop (Belle Mead NJ). 2015;44:547–553.
11. Goedhart LM, Gerbers JG, Ploegmakers JJ, Jutte PC. Delay in diagnosis and its effect on clinical outcome in high-grade sarcoma of bone: a referral oncological centre study. Orthop Surg. 2016;8:122–128.
12. Handgraaf HJM, Boogerd LSF, Hoppener DJ, Peloso A, Sibinga Mulder BG, Hoogstins CES, Hartgrink HH, van de Velde CJH, Mieog JSD, Swijnenburg RJ, Putter H, Maestri M, Braat AE, Frangioni JV, Vahrmeijer AL. Long-term follow-up after near-infrared fluorescence-guided resection of colorectal liver metastases: a retrospective multicenter analysis. Eur J Surg Oncol. 2017 May 6. [Epub ahead of print]
13. Jiang JX, Keating JJ, Jesus EM, Judy RP, Madajewski B, Venegas O, Okusanya OT, Singhal S. Optimization of the enhanced permeability and retention effect for near-infrared imaging of solid tumors with indocyanine green. Am J Nucl Med Mol Imaging. 2015;5:390–400.
14. Judy RP, Keating JJ, DeJesus EM, Jiang JX, Okusanya OT, Nie S, Holt DE, Arlauckas SP, Low PS, Delikatny EJ, Singhal S. Quantification of tumor fluorescence during intraoperative optical cancer imaging. Sci Rep. 2015;5:16208.
15. Keating J, Newton A, Venegas O, Nims S, Zeh R, Predina J, Deshpande C, Kucharczuk J, Nie S, Delikatny EJ, Singhal S. Near-infrared intraoperative molecular imaging can locate metastases to the lung. Ann Thorac Surg. 2017;103:390–398.
16. Lu P, Weaver VM, Werb Z. The extracellular matrix: a dynamic niche in cancer progression. J Cell Biol. 2012;196:395–406.
17. Luo S, Zhang E, Su Y, Cheng T, Shi C. A review of NIR dyes in cancer targeting and imaging. Biomaterials. 2011;32:7127–7138.
18. Meazza C, Scanagatta P. Metastatic osteosarcoma: a challenging multidisciplinary treatment. Exp Rev Anticancer Ther. 2016;16:543–556.
19. Mu X, Isaac C, Greco N, Huard J, Weiss K. Notch signaling is associated with ALDH activity and an aggressive metastatic phenotype in murine osteosarcoma cells. Front Oncol. 2013;3:143.
20. Nasser A, Fourman MS, Gersch RP, Phillips BT, Hsi HK, Khan SU, Gelfand MA, Dagum AB, Bui DT. Utilizing indocyanine green dye angiography to detect simulated flap venous congestion in a novel experimental rat model. J Reconstr Microsurg. 2015;31:590–596.
21. van der Vorst JR, Schaafsma BE, Hutteman M, Verbeek FP, Liefers GJ, Hartgrink HH, Smit VT, Lowik CW, van de Velde CJ, Frangioni JV, Vahrmeijer AL. Near-infrared fluorescence-guided resection of colorectal liver metastases. Cancer. 2013;119:3411–3418.
© 2018 Lippincott Williams & Wilkins LWW
22. Wang B, Tu J, Yin J, Zou C, Wang J, Huang G, Xie X, Shen J. Development and validation of a pretreatment prognostic index to predict death and lung metastases in extremity osteosarcoma. Oncotarget. 2015;6:38348–38359.