Animal models and molecular imaging tools to investigate lymph node metastases

Abstract

Lymph node metastasis is a strong predictor of poor outcome in cancer patients. Animal studies of lymph node metastasis are constrained by difficulties in the establishment of appropriate animal models, limitations in the noninvasive monitoring of lymph node metastasis progression, and challenges in the pathologic confirmation of lymph node metastases. In this comprehensive review, we summarize available preclinical animal cancer models for noninvasive imaging and identification of lymph node metastases of non-hematogenous cancers. Furthermore, we discuss the strengths and weaknesses of common noninvasive imaging modalities used to identify tumor-bearing lymph nodes and provide guidelines for their pathological confirmation.

Introduction

Lymphatic metastasis is a predictor of poor outcome in many solid malignancies [1–4]. The presence of lymph node metastases decreases the 5-year survival of melanoma patients independent of other prognostic factors of the primary tumor [5]. Likewise, the number of metastatic lymph nodes resected correlates with survival in breast cancer patients [6]. Despite the importance of lymph node metastasis (LNM), as highlighted by its inclusion in cancer staging systems, preclinical evaluation of lymph nodes (LN) in small animal models remains limited. When performing LNM studies, investigators face multiple challenges including: (1) the paucity of valid animal models in different animal species, (2) difficulties in identifying and serially monitoring metastatic disease by noninvasive imaging techniques, and (3) challenges of harvesting and pathologically confirming tumor replaced LN in small animal models. Herein, we review animal models, detail noninvasive imaging modalities, and provide recommendations for imaging, harvesting, and pathologically confirming LNM.

Preclinical LNM animal models

In this section, we review the most common, reproducible, murine models for LNM studies in non-hematogenous cancers (Table 1) and highlight the roles of: (1) immunocompetent versus immunodeficient hosts, (2) xenogeneic or syngeneic tumor models, and (3) orthotopic versus heterotopic tumor implantation. Employing syngeneic tumor cell lines or germline-encoded tumors in immunocompetent models of cancer provide a tumor microenvironment with compatible stromal, vascular, and immunological components [7]. Immunodeficient models utilize either human cancer cell lines or tumor explants as xenografts. These models lack a complete immunologic milieu for LNM but allow the use of well-characterized human cancer cell lines that may metastasize to regional nodal basins [8]. An understanding of the LN microenvironment and biology in commonly used immunodeficient mouse strains is essential for researchers investigating LNM as it is influenced by the immunologic milieu, therapy resistance, and the development of spontaneous cancers after a certain age (Supplemental Table 1). In addition to recognizing the benefits and limitations of the immune environment in animal models, it is essential to select a model that permits high-yield LNM for the specific cancer of interest. With an emphasis on reproducibility by multiple investigators, we summarize solid cancer animal models and cell lines useful for LNM studies (Tables 1 and 2) and highlight common murine LNM solid cancer models in the following sections.

Table 1.

Animal models of cancer used in lymph node metastasis imaging studies

Primary tumor	Animal model	Cell lineb	Model type	Route	Time to LN metastases	Identification method	Lymph nodes evaluated	Yield/sensitivity/specificity
Melanoma	C57BL/6 mice [91]	B16	IC, S, O	Flank	NR	LS	Axillary
	C57BL/6 mice [13, 14]	B16-BL6 (5×106)	IC, S, O	Auricular	20 days	FL	Cervical	LN yield: 6/7 LN
	C57BL/6 mice [15]	B16-F10 (2×106)	IC, S, O	Thigh	14–21 days	LS	NRa	Yield, 17/17 mice
	C57BL/6 mice [16]	B16-F10 (2×106)	IC, S, O	Hind leg	7–28 days	MRI	Para-aortic/pelvica	LN yield, 100% mice
	C57BL/6 mice [92]	B16-F10 (1×105)	IC, S, O	Abdominal SQ	15 days	MRI	Axillary, cervical, inguinala
	C57BL/6J-TyrC-2J/J mice C57BL/6 mice [9–12, 93]	B16-F10 (2×105)	IC, S, O	Footpad	21 days	BLI, MRI, PET	Popliteal, inguinal, brachial axillarya
	MeLiM Swine [20]	Naturally occurring	IC, S, O	Spontaneous	NR	PET	NR
	Nude athymic mice [19]	M21	ID, X, O	Thigh	NR	FL	Axillary, inguinal
	Nude athymic mice [18]	B16CXC4	ID, X, O	Lip	NR	NIRF	Cervical
	Nude athymic mice [17]	B16-F10 (1×106)	ID, X, O	Auricular	7 days	US	Cervicala
	Sinclair swine [21, 22]	Naturally occurring	IC, S, O	Spontaneous	NR	LS, US	Axillary, cervical, inguinal	Yield, 6/6 swine (27/31 LN)
Breast	BALB-neuT mice [25]	Transgenic	TG	–	21 weeks	MRI	Axillarya
	C3(1) SV40 tag mice [94]	Transgenic	TG	–	12 weeks	MRI	Axillarya	Yield, 12/12 mice
	Lewis rats [95]	Disaggregated rodent tumor	IC, S, O	Mammary fat pad	NR	PET	Popliteal
	New Zealand white rabbits [96, 97]	VX2 (1×107)	IC, S, O	Intramammary pocket	30–50 days	MRI, PET, US, XR	Axillary, mediastinal, abdominal	MRI sensitivity, 78%, PET/CT; sensitivity, 68%, MRI/PET/CT; sensitivity, 82%; US sensitivity, 23.5%; XR sensitivity, 0%
	Nude athymic mice [23, 24, 66, 98–100]	KPL1, MCF-7, MDA-MB231, MDA-MB435, PT-18, T47D (1×106–1×107)	ID, X, O	Mammary fat pad	15–90 days	BLI, FI, MRI	Brachial, axillary, cervical, lateral thoracic, iliac, abdominal, inguinala	Yield, 100% mice by week 8
	Nude athymic mice [26, 27]	MDA-MB435 (2×106)	ID, X, O	Intramammary pocket	70–100+days	FI	Axillarya	Yield, 11/20 LN sensitivity/specificity, 80/96%
	Nude athymic mice [100]	MDA-MB231 (1×105 )	ID, X, H	Intracardiac	35 days	BLI	Brachial, axillarya	Yield, 5/8 mice
Head and neck	BALB/c nu/nu mice [101]	SAS-L, HSC-3 (1×105)	ID, X, O	Tongue	24 days	FL	Head/necka	Yield, 13/14 LN; sensitivity/
	C57/B Bk5myrAkt mice [102]	Spontaneous	IC, S, O	–	21 days	PET	NR	specificity, 92%/87%
	Nude athymic mice [29]	DM14 (2×105)	ID, X, O	Tongue	15 days	MRI, NIR	Cervicala	Yield, 4/6 LN
	NZ white rabbits [75, 76]	VX2 (1–2×107)	IC, S, O	Auricular	21–28 days	MRI, NIR	Cervical, mandibulara	Sensitivity/specificity,
	SCID mice [63]	OSC-19 (2×105), SCC1 (2×106)	ID, X, O	Tongue	14 days	FL	Yield, 8/8 mice	72–100%/83–85%
Lung/mesothelioma	Japanese white rabbits [103]	VX1	IC, S, H	Transbronchial	34 days	XR	Pulmonary	Yield, 1/22 rabbits
	Nude athymic mice [30]	MSTO-211 H (1×105)	ID, X, H/O	Preauricular, pleural	28–56 days	FL	Cervical, mediastinal, abdominal, retoperitoneala
	SCID mice [32]	MA44-33 (2×106)	ID, X, O	Intraparenchymal	12 days	FL	Mediastinal	Yield, 5/6 mice
	BALB/c nude mice [104]	HT-29 (5×106)	ID, X, O	Submucosal anorectal	35 days	FL	Para-aortica	Sensitivity/specificity, 100%
	b-estriol pre-treated athymic mice [105]	LS174T (4×106)	ID, X, H	Footpad	24 days	LS	Axillary, popliteala	Yield, 66%
	b-estriol pre-treated athymic mice [105]	LS174T (4×106)	ID, X, H	IV	35 days	LS	Axillary, popliteala	Yield, 66%
	C57BL/6 mice [36]	Bx-PC (1×106)	IC, S, O	Portal vein	36 days	FL	Portal veina
Gastrointestinal	F344 rats, Lewis rats [106]	RCN-H4 (5×106)	IC, S, O	Ileocolic vein	50 days	BLI	Mesenteric	Yield, 3/10 mice
	F344/N Jcl-rnm nude rats [34, 35]	OCUM-2M (1×106)	ID, X, O	Subserosal gastric	NR	FL	Gastric/hepatica	Yield, 100% rats
	NMRI nude mice [107]	AR42J (1×106)	ID, X, O	Amphicrine	NR	PET, CT	Mesenterica
	Nude mice [36, 37]	MIA-PaCa2 (SOI)	ID, X, O	Pancreatic	36 days	BLI, FL	Portal, intestinala
	Nude mice [108]	XPA1-RFP (5×105)	ID, X, H	Inguinal LN	7 days	FL	Axillarya
	SCID/bg mice [87]	HT-29(1–3×106)	ID, X, H	SQ flank	14–21 days	BLI	Axilliarya	Yield, 9/18 mice
	nu/nu mice, NOD/SCID [38, 39]	PC3M, PC3M-SV1, PC3M-LN4 (5×105)	ID, X, O	Prostate	7–14 days	BLI, FL	Pelvic	Yield, 20–100%
Prostate	C57LB/6xFBV mice [109, 110]	Transgenic	TG	–	16–24 weeks	BLI, FL, LS	Pelvic	Yield, 40%
	Copenhagen rats [111]	R3327-MATLyLu (1×106)	IC, S, H	Flank	14 days	MRI	Popliteala	Yield, 18/21 LN
	Copenhagen rats [112]	R3327-MATLyLu (1×106)	IC, S, H	Footpad	7–23 days	LS	Axillarya	Yield, 3/12 LN

IC immunocompetent, ID immunodeficient, S syngeneic, X xenogeneic, TG transgenic, H heterotopic, O orthotopic, BLI bioluminescent imaging, CT computer tomography, FL fluorescence imaging/microscopy, LS lymphoscintigraphy, MRI magnetic resonance imaging, NIR near IR fluorescence, PET positron emission tomography, US ultrasound, XR X-ray

^aMicrometastatic disease

^bCancer cell line (number of cells typically injected)

Table 2.

Tumor cell lines predisposed to LNM

Cancer	Cell line	Parent cell line	Origin	Histology	Yielda
Melanoma	B16-BL6 [113]	B16	C57BL/6J	Cutaneous	42–72% mice
	B16-F10 [113]	B16	C57BL/6J	Cutaneous	18–82% mice
Breast	MDA-MB-231(LM1) [114]	MDA-MB-231	Human	Adenocarcinoma	92% LN
	MDA-MB-435(LN435) [26]	MDA-MB-435	Human	Adenocarcinoma	NR
Head and neck	GDC185 [115]	–	Human	Squamous carcinoma	56–81% LN
	HSC-3-M3 [116]	HSC-3	Human	Squamous carcinoma	90% mice
	OSC-19 [117]	–	Human	Squamous carcinoma	81% mice
	MG14 [118]	SCC7	C3H/HeJ	Squamous carcinoma	NR
Lung	H460SM [119]	H460	Human	Adenocarcinoma	100% rats
	LLC [120]	–	C57BL/6	Adenocarcinoma	93% mice
	Ma44-3 [32]	Ma44	Human	Squamous carcinoma	83% mice
Gastrointestinal	MIA-PaCa2 [121]	–	Human	Ductal adenocarcinoma	NR
	OCUM-2M-LN [122]	OCUM	Human	Scirrhous carcinoma	100% mice
Prostate	PC3-LN4 [38]	PC-3	Human	Adenocarcinoma	100% mice

LN lymph node, NR not recorded

^aPathologically confirmed metastases

Melanoma

A common model for studying lymphatic metastasis in melanoma involves orthotopic injection of syngeneic B16 melanoma cells into the footpad [9–12], auricular tissue [13, 14], thigh [15], or abdomen [16] of C57BL/6J mice (Table 1). LNM drain to local nodal basins resulting in popliteal and inguinal metastases from footpad tumors; cervical metastases from auricular tumors; and axillary, popliteal, or inguinal metastases from ventral and flank tumors in >90% of mice within 2–4 weeks. In this murine model, the primary tumor can be resected with minimal morbidity or mortality. Large, pigmented, superficial tumor-bearing nodes, which subsequently develop, facilitate noninvasive imaging and ex vivo analysis of involved nodes without requiring mouse sacrifice. Table 1 details all melanoma models available for LNM imaging studies. Immunodeficient murine models allow for the study of human and murine melanomas in an immunological naive background [17–19]. In addition to providing a large animal model of LNM, spontaneous swine melanoma models are used to examine practical limitations of current imaging techniques [20–22].

Breast

Animal models for breast cancer lymphatic metastasis are based upon orthotopic implantation of tumor explants or cancer cells into mammary fat pads (Table 1). Such models allow for noninvasive LNM imaging of superficial LN after surgical resection of primary tumor. The most common murine models utilize immunodeficient mice injected with cells derived from human breast cancer patients, including those with a high propensity for LN metastasis (Table 2). Tumor cells are injected into the mammary fat pad and LNM are noted within 3–8 weeks in the axillary, brachial, cervical, inguinal, and popliteal basins depending on the proximity of the mammary fat pad inoculated with tumor [23–25]. Another method for establishing an orthotopic model with high tumor engraftment rates and consistent anatomic location involves the surgical implantation of a human breast cancer into the mammary fat pad of a female nude mouse [26, 27]. The resulting tumor is resected, trimmed free of necrotic tissue, homogenized, and surgically implanted into the mammary fat pad of another nude mouse. This technique results in axillary LN (55% of animals at day 106 post-implantation), lung (15%), and liver (10%) metastases [27].

Head and neck

To identify regional LNM associated with oropharyngeal squamous cell carcinoma, an orthotopic mouse model developed by Myers et al. [28] is used in LNM imaging studies. Xenogeneic tumor cells are injected into the anterior tongue of an immunodeficient host and regional metastases to cervical and mandibular nodes were noted within 2 weeks (Table 1). Superficial nodal basins identified by noninvasive imaging techniques are reproducible and easily accessible for surgical biopsy via a small cervical incision and permit histopathologic confirmation of metastatic LN [29].

Pleural/lung

In pleural cancer models, involved mediastinal nodal basins are difficult to image due to limitations in depth of penetration for many imaging modalities. Heterotopic preauricular models of mesothelioma metastasize to cervical LN and facilitate imaging of superficial nodal metastases while providing an easy method to study metastatic spread following primary tumor resection [30]. Our laboratory has developed an orthotopic xenograft model of human malignant pleural mesothelioma that accurately reflects human disease. Following the direct injection of cancer cells into the pleural space, bilateral pleural disease develops as the left and right pleural cavities communicate in mice. Nodal metastases to cervical, mediastinal, and retroperitoneal basins occur within 3 weeks of tumor inoculation. This model allows for noninvasive monitoring of tumor progression and quantitative bioluminescent imaging (BLI). BLI can identify tumor-bearing LN basins but does not permit the identification of specific LNM [31]. Tomographic methods, such as CT and MRI, are able to localize LNMs but are logistically challenging due to time constraints and the number of mice able to be screened.

In an orthotopic model of lung cancer described by Takizawa et al. [32], tumor cells suspended in Matrigel are directly injected into the lung parenchyma following a thoracotomy. Macroscopically detectable tumor typically develops at the injection site within 9 days with mediastinal LNM subsequently observed by 12 days [33]. Using this model, we have successfully developed protocols utilizing BLI to facilitate the identification and progression of mediastinal LNM. To histologically confirm mediastinal nodal involvement, LN must be harvested by carefully dissecting and mounting suspected nodal basins, even in the absence of macroscopic evidence of disease (as discussed in a later section). To differentiate pleural tumor nodules from LNM, which grossly appear similar on harvesting, a detailed pathologic analysis by an experienced animal pathologist is the only useful confirmatory method.

Gastrointestinal/pancreatic

As with pleural and lung cancer models, gastrointestinal and pancreatic cancer can be challenging for LNM studies due to deep nodal basins. Commonly used models of these tumors are summarized in Table 1. A model of orthotopic gastric cancer described by Koyama et al. [34, 35] reproducibly develops LNM in nude rats following sub-serosal injection of the anterior stomach with a human scirrhous gastric carcinoma cell line with a predilection for lymphatic metastases (OCUM-2M LN). To further promote LNM, rats were depleted of natural killer cells with 3 Gy of irradiation 3 days prior to tumor inoculation [34, 35]. In animals with established tumor, fluorescent laparoscopy was used to identify lymphatic metastases, which were subsequently confirmed histologically. Orthotopic pancreatic cancer models [36, 37] utilize surgical implantation of human PaCa-2 cells into body of the pancreas of immunodeficient mice. On fluorescent imaging, mice exhibited significant primary tumor growth and metastatic spread within 2 weeks after surgical tumor implantation. In these studies, autopsy confirmed metastases in the spleen (100%), intestinal nodes (100%), portal nodes (90%), liver (80%), retroperitoneum (60%), diaphragm (50%), kidney (30%), and lungs (10%).

Prostate

LNM imaging studies for prostate cancer utilize heterotopic and orthotopic tumor implantation into host animals (Table 1). However, only orthotopic models recapitulate human disease with nodal metastases to pelvic and retroperitoneal LN [38, 39]. In this model, a highly metastatic human prostate cancer cell line (PC3-LN4) is injected into a lobe of the prostate and can result in 100% of mice with pelvic metastases as determined by BLI and fluorescent imaging.

Lymph node metastasis imaging in animal models

The successful development and use of LNM animal models are based on the ability to noninvasively, serially image and monitor LNM involvement. Even with a clear understanding of lymphatic anatomy, it is often difficult to grossly identify lymph nodes in small animal cancer models. Here, we review preclinical LN imaging modalities that direct the investigator to lymph nodes potentially harboring metastases by identifying nodal basins with an increased probability for metastases. We also describe techniques aimed at visualizing specific LNM.

Although radiocolloid and dye-based tracer agents are used in small animal models, their use for detecting LNM [40] can be limited by: (1) invasive surgical exploration or ex vivo analysis of specimens, (2) suboptimal spatial–temporal resolution in small animal models, and (3) high background signal from nearby tumors [41]. Newer modalities and tracer agents are being developed to overcome these limitations. Table 3 provides a summary of available LN imaging modalities used to identify both lymphatic architecture (Table 3a) and LNM (Table 4b).

Table 3.

Preclinical lymphatic/SLN imaging

Imaging modality	Animal model	Lymph node tracer	Route	Tracer administration to imaging time	Tracer sizea	Depth of penetration	Nodal specificityb	Clinical translatability and comments
Visual	Dogs, normal lung [40]	Isosulfan Blue + Tc-99 m	Intrapulmonary	5–20 min	100–1000 nm	N/R	5/6 dogs, identified pulmonary, tracheobronchial, and mediastinal LN	Approved for human use; current standard in breast cancer and melanoma; limited use in other tissue such as lung
	Pigs, normal mammary glands [47]	Blue dye liposomes	Mammary gland	3–24 h	100–1,000 nm	N/R	SLN detected in all animals up to 24 h	Approved for human use; improved nodal staining compared to free blue dye
Radioisotope	Rabbits, hind leg popliteal nodes [50]	Tc-99m-labeled liposomes	SQ	0.5–24 h	~130 nm	N/R	N/R	Approved for human use; therapeutic potential
	Dogs, normal lung [123]	I-131-labeled mob	Bronchosopic delivery	14–36 h		N/R	N/R	Human toxicities unknown; able to direct tracers to specific LN basins
Ultrasound	Dogs, normal cervical and popliteal nodes [52]	Nanobubbles	SQ	0–120 min	500–1,000 nm	2 cm	34/40 LN 30/32 LN (near micron/submicron bubbles)	Nanobubbles not approved for human use; concern for thermal injury to normal tissue
	Mice, footpad [41]	Evans blue dye	Intradermal	8 s–1.5 h		3 cm	N/R	Blue dye and ultrasound approved for human use
	Dogs, normal axillary and inguinal lymph nodes [51]	Gold nanotubes	SQ	19–46 h	10×400 nm	> 1.8 nm	N/R	Sonovue approved for human use
Fluorescent imaging	Rats, footpad and pigs, GI tract [124]	Fluorescent microspheres	SQ footpad, submucosal	0.2–6 h	100–1,000 nm	N/R	5/6 esophageal and small bowel SLN; 1/6 gastric and large bowel SLN	Not approved for human use
	Pigs, normal colonic mesenteric nodes [125]	ICG:HAS	Colonic parenchyma	~5 min	~7 nm1	<4 cm	Mesenteric SLN identified in 5/5 pigs	ICG:HAS approved for human use; FLARE system in clinical trial for breast cancer SLN mapping
	Mice, subcutaneous melanoma and teratocarcinoma [19]	QDs	IT	0–40 min	22–40 nm1	N/R	N/R	QDs not approved in humans, toxicity unknown
	Rats and pigs, pleural lymphatics [58]	QDs	Intrapleural	4–24 h	15 nm1	N/R	N/R	QDs not approved for human use; intraoperative NIR fluorescent imaging system easily adaptable for human use
	Pigs, lung [43]	QDs	Lung parenchyma	1 min–4 h	15–20 nm1	1.5–5 cm	16/16 SLN	QDs not approved for human use
OFDI with intravital microscopy	Mice, dorsal skin fold sarcoma [61]	No tracer	N/A	N/A		1–2 mm	N/R	Identified cellular masses within peritumoral lymphatics

FLARE fluorescence-assisted resection and exploration, ICG indocyanine green, HAS human serum albumin, IT intratumoral, QDs quantum dots, LN lymph node, SLN sentinel lymph node, SQ subcutaneous, N/R not recorded, OFDI optical frequency domain imaging, GI gastrointestinal

^aSize of tracer agent as reported

^bNumber of LN correctly identified or number of animals with positive LNs

¹Hydrodynamic diameter

Table 4.

Preclinical LNM imaging

Imaging modality	Animal model	Lymph node tracer	Route	Tracer administration to imaging time	Tracer sizea	Depth of penetration	Nodal specificityb	Clinical translatability and comments
Fluorescent imaging	Mice, inguinal and axillary nodes from subcutaneous tumor model [62]	Fluorescent-labeled LYVE-1 mAb; RFP-pancreatic cancer cells	SQ	4–72 h		N/R	N/R	Not approved for human use; monoclonal LYVE-1 antibody targets lymphatics and co-localizes with RFP-expressing metastatic cancer cells
	Mice, oral squamous carcinoma [63]	Cy5.5-labeled anti-EGFR mAb	IV	72 h		N/R	8/8 biopsies	Cy5.5 not approved for human use
	Mice, breast cancer [26]	GFP-expressing oncolytic herpes virus NV1066	IT	48 h		N/R	4/5 LN	GFP not approved for human use; therapeutic potential
	Mice, orthotopic pleural mesothelioma [30]	GFP-expressing oncolytic herpes virus NV1066	Intrapleural	48 h		N/R	80% sensitivity	Lymph node metastases identified by GFP–fluorescent thoracoscope; therapeutic potential
	Mice, orthotopic rectal cancer [72]	GFP-expressing telomerase-specific oncolytic adenovirus	IT	5 days		N/R	12/13 LN	GFP not approved for human use; therapeutic potential
	Mice, hindpaw colon cancer model [74]	VT680-labeled lymphocytes	IV	7–9 days		5 mm	N/R	Not approved for human use
MRI	Mice, B16 melanoma footpad [126]	SPIO-labeled tumor cells	SQ	0–9 days		N/A	N/R	SPIO approved for use in humans
	Mice, breast cancer [25]	Gd-labeled dendrimers	IT	12–36 min	9–13 nm	N/A	6/6 mice; identified metastatic LN	MRI and gadolinium approved for human use; dendrimers not approved for human use
	Mice, normal mammary glands [127]	Gd/Cy5.5 dual-labeled dendrimers	Mammary gland	20–30 min	9 nm	N/A	N/R	Dendrimers not approved for human use; dual-imaging technique with NIR fluorescence
PET	Rats, subcutaneous ovarian and lymphoma xenografts [128]	FDG PET	IV	20 min–2 h		N/A	N/R	FDG PET approved for human use
	Mice, footpad melanoma [9]	FEAU; oncolytic herpes virus NV1023	IT	2 h		N/A	8/8 LN	PET approved for human use; limited by high background PET signal in primary tumor and GI tract
BLI	Mice, mammary gland [66]	D-Luciferin; firefly luciferase-expressing adenoviral vector	IT	20 min		N/A	N/R	D-Luciferin not approved for human use
	Mice, pleural tumors [79]	D-Luciferin; firefly luciferase-expressing tumors	IP	20 min		N/A	N/R	D-Luciferin not approved for human use

BLI bioluminescent imaging, FDG 2-deoxy-2[¹⁸ F]Fluoro-D-Glucose, FLARE fluorescence-assisted resection and exploration, GFP green fluorescent protein, IP intraperitoneal, IT intratumoral, IV intravenous, mAb monoclonal antibody, MRI magnetic resonance imaging, NIR near infrared, PET positron emission tomography, SCID severe combined immunodeficiency, SLN sentinel lymph node, SQ subcutaneous, N/R not recorded, OFDI optical frequency domain imaging, SPIO small paramagnetic iron oxide

^aSize of tracer agent as reported

^bNumber of LN correctly identified or number of animals with positive LNs

Optimization of tracer agent size is fundamental for lymphatic mapping. Biokinetic and lymphoscintigraphy studies have shown that particle size impacts lymphatic entry and nodal retention of tracer molecules [42, 43]. An inverse relationship between particle size and LN interstitial clearance exists. Particles greater than 500 nm typically fail to reach tumor-draining lymph nodes (TDLN) due to slow interstitial clearance, while particles less than 5 nm can freely enter capillaries and pass directly through the LN. Intratumoral administration of appropriately sized tracers (~5–150 nm) migrate into lymphatics, demonstrate retention within tumor draining LN [42, 44], and provide the basis for newer lymphatic tracer molecules, such as liposomes, dendrimers, and quantum dots.

Lymphatic nodal basins that can harbor LNM are best localized with imaging techniques that utilize exogenous tracer agents to define tumor-draining lymphatic beds (Table 3a). While specific LNM are best visualized with imaging techniques taking advantage of morphologic and functional changes in metastatic LN (Table 4b). In the following sections we describe tracer agents and imaging modalities for this purpose and detail characteristic properties including route of administration, imaging times, depth of penetration, and nodal specificity.

Liposomes—facilitate visual, radiographic, and ultrasonographic lymphatic imaging

Liposomes are lipid-based, vesicular, molecules that can be conjugated to or encapsulate various imaging agents, dyes, or radiocolloids. Ranging in diameter from 30 nm to several micrometers, the size and lipophilicity of liposomes promote efficient lymphatic entry and retention [45, 46]. Hirnle and colleagues [47–49] showed reliable SLN identification using blue dye-encapsulated liposomes administered subdermally in rabbit and pig models. Similarly, Phillips et al. [50] developed a blue dye/Tc-99m radio-colloid dual-labeled liposome, which allowed SLN identification more than a week following subdermal injection into rabbits. Dye-labeled liposomes migrate to lymphatic beds and are identified by direct visualization, surgical, or laparoscopic exploration. Radiocolloid-labeled liposomes require scintillation counters to localize tumor-draining lymphatic beds. To verify potential metastases, surgical intervention and tissue biopsy are needed. As an extension of dye-contrast ultrasonography, 5–10-μm microbubble-laden liposomes facilitate the identification of SLN by ultrasound. The tracer agent is injected subcutaneously, migrates through lymphatics, and identifies 85–95% of SLN—common sites of LNM [51, 52]. The disadvantages of US LN imaging include the limited use of identifying thoracic and retroperitoneal LN, operator experience, and ex vivo analysis of LN to confirm LNM. Although not currently utilizing liposomes/microbubbles, photoacoustic tomography (PAT) extends the ultrasonic depth of penetration by taking advantage of photoacoustic properties inherent to dye-contrast agents [53]. In mice and rats, PAT accurately detects LNs that accumulate low concentrations of methylene blue (~0.28 nM) at depths greater than 2 cm in both two- and three-dimensions [41, 54].

Quantum dots—fluorescent markers engineered for lymphatic migration and nodal retention

Quantum dots (QDs) are nearly spherical semiconductors with size-dependent fluorescence making these nanoparticles candidates for lymphatic imaging. QDs are engineered between 15 and 20 nm in diameter to optimize lymphatic migration and nodal retention [43]. QDs posses improved brightness, narrow emission spectra tunable across broad ranges, and enhanced photostability in comparison to organic fluorophores [19, 55, 56]. The physical properties of QDs are comprehensively reviewed by Biju et al. [57] and are beyond the scope of this review. In conjunction with a near infrared imaging (NIR) system, Parungo et al. [58, 59] showed that QDs demonstrate high yield localization of SLN in preclinical models of pleural and esophageal lymphatic drainage in rats. Similarly, Soltesz et al. [43] identified SLN following QD injection into lung parenchyma in 12/12 pigs within 5 min of injection at depths to 5 cm. QDs can accurately identify tumor-draining lymph nodes in animal models with no signs of acute systemic toxicities [19]. Potential problems of utilizing QD composed of heavy metals (e.g., cadmium, selenium) include cytotoxicity, tracer stability, and bioaccumulation which are catalogued by Hardman [60].

Optical frequency domain imaging with intravital microscopy—tracer-free lymphatic mapping

Although not validated for the identification of lymphatic metastases, emerging imaging techniques such as optical frequency domain imaging (OFDI) with intravital microscopy (IVM) offer high sensitivity and rapid acquisition times needed for large-volume lymphovascular morphometry [61]. Limited to superficial tumors modeled on dorsal skin folds, OFDI can perform high-resolution imaging of tumor vasculature, tracer-free functional lymphangiography, and characterization of tissue viability in the context of tumor burden and therapy. A priori knowledge of lymphatic anatomy is required to locate the lymph nodes and lymphatics draining from the dorsal skin fold model. OFDI allows for the identification of lymphovascular anatomy 1–2 mm from the surface, a more than two-fold increase from previous methods. Furthermore, the dynamic monitoring of lymphatic vasculature during cancer progression and lymphedema offers a potential method for evaluating existing and emerging therapies [61].

LNM-specific imaging techniques

The lymphatic tracer agents described above provide a means of lymphatic mapping but are not tumor cell specific. These techniques provide a roadmap to harvest cancer-draining lymph nodes for subsequent analysis for metastatic foci. Organic fluorophores conjugated to monoclonal antibodies targeting lymphatic-associated antigens have been utilized for the purposes of lymphatic mapping in combination with fluorescently labeled cancer cells to image metastatic LN. An anti-LYVE-1 monoclonal antibody (targeting the lymphatic endothelium) conjugated to AlexaFluor co-localized with fluorescently labeled meta-static pancreatic cancer cells migrating from inguinal to axillary LN through anterior abdominal wall lymphatics [62]. Similarly, Gleysteen et al. [63] identified cervical LNM in immunodeficient mice with oral squamous cell carcinoma using a tumor-targeted fluorescent anti-EGFR monoclonal antibody. This tumor-targeted strategy allows for specific imaging of tumor-involved LN. Furthermore, this method permits the study of LNM biology as demonstrated in a recent study showing that lymphangiogensis within TDLN is an early event in promoting LNM [64]. While targeted LN imaging using monoclonal antibodies has appeal, it can be hindered by non-specific uptake in visceral organs when administered intravenously [65] and by fluorophore properties. The long-term use of organic fluorophores is limited by poor tissue penetration, high levels of overlapping emission from background tissue autofluorescence and photobleaching [43, 45].

Oncolytic viruses—permit tumor-specific labeling with fluorescent proteins and identify LNM

Biological agents such as attenuated viruses [30, 66] and bacteria [67] selectively infect tumor cells, drain to TDLN, and help identify lymphatic tumor spread [30, 68, 69]. Oncolytic viruses selectively infect, replicate within, and lyse cancer cells. The virions released from lysed cancer cells infect neighboring cancer cells. When a single low dose of oncolytic virus is injected into a primary tumor, the virus traffics similar to tumor cells—from the primary tumor to the TDLN. When these draining nodes are superficial (axillary, femoral, or popliteal), LNMs can be identified by fluorescent microscopy. The replication-competent recombinant virus NV1066 (derived from HSV-1) carries a reporter gene for green fluorescent protein selectively infects cancer cells thereby facilitating the detection of nodal metastases with in vivo fluorescent imaging [30]. Our laboratory has shown the utility of identifying LNMs in the mediastinal region of pleural or lung cancer in an orthotopic mouse model utilizing fluorescent thoracoscopy. Furthermore, NV1066 was utilized to demonstrate the draining SLN in breast cancer metastases. In our experience, virally guided fluorescent imaging facilitates easy identification of metastatic LN from normal LN prior to histology and immunohistochemistry (Fig. 1). This modality offers decreased time and effort needed to identify and harvest LNMs. Oncolytic virus-based LNM imaging with virally guided fluorescent reporters not only allows imaging and identification of LNM, but may also act as a therapeutic modality since the viruses selectively infect and lyse cancer cells. In addition to fluorescent imaging, this method offers a platform for in vivo LNM imaging and responses to therapy using viruses transduced to express BLI or positron emission tomography reporter genes [9, 26, 30, 66, 70–72].

Fig. 1.

Oncolytic virus-mediated detection of lymph node metastases using either fluorescent imaging, BLI, or PET. a Fluorescence microscopy detects LNM in cervical LN following injection of GFP-expressing herpes virus (left panel). Histology confirms microscopic focus of LNM as denoted by an arrow (right panel) (reprinted with permission from Adusumilli et al. [30], © 2006 American Association for Thoracic Surgery). b BLI signal is seen in the axillary region of mice following injection of a firefly luciferase containing recombinant adenovirus in mice containing prostate cancer LNM (reprinted with permission from Macmillan Publishers Ltd., Burton et al. [70]). c Similarly, a prostate-specific PET reporter vector facilitated PET identification of LNM following primary tumor resection in conjunction with the injection of ¹⁸FHBG tracer (reprinted with permission from Macmillan Publishers Ltd., Burton et al. [70])

Fluorescent-mediated molecular tomography—detects LNM and details immune interactions

Newer techniques are emerging which combine data from microscopic studies using intravital microscopy and mesoscopic imaging with fluorescent-mediated molecular tomography (FMT). These methods allow for serial, quantitative, and three-dimensional assessments of fluorophore concentrations in whole body imaging of murine cancer models [73]. Attenuation of signal from VT680-labeled (NIR-fluorescent probe) cells by surrounding tissue is decreased since hemoglobin, water, and lipids have their lowest absorption coefficients in the NIR region (650–900 nm). Although the tissue exhibits the lowest absorption coefficient in the NIR spectrum, there is still substantial absorption and scatter that limits depth of penetration. Using this technique, Swirski et al. tagged lymphocytes with VT680 and were adoptively transferred to a murine hind paw colon cancer model [74]. Mice are then serially imaged with FMT and IVM. FMT allows researchers to monitor T cells trafficking from a tumor focus to tumor-draining inguinal LN, at depths to 6 mm. Since VT680 does not lose fluorescence in resting cells, it is possible to quantify the number of adoptively transferred T cells within the (~7×10⁴) tumor and the (~4×10⁴) TDLN—consistent with ex vivo flow cytometric data (~4–10×10⁴) [74]. Combined with IVM, this multi-modal method cannot only detect tumor-bearing LN, but also details the antigenic specificity of immunologic molecules in vivo.

Iron oxide nanoparticles—specifically detect LNM

Nanoparticles comprised of iron oxide can be used in MRI studies to accurately and sensitively detect LNM. The tracer agents ultra-small paramagnetic iron oxide or SPIO nanoparticles are administered intravenously and are selectively phagocytosed by macrophages. Based upon the T2*-weighted effects of iron oxide particles, normal functional lymphatic tissue with macrophages appears dark, while the nonphagocytic metastatic tissue appears bright. Although possible LNM in large, tumor-involved, LN, may be identified without tracer agents, studies by Lee et al. and Choi et al. combined gadolinium contrast agents with iron oxide nanoparticles to detect LNM larger than 0.5 cm with a sensitivity of 70–100% and specificity of 83–85% in rabbit models of head and neck cancer [75, 76].

Bioluminescent imaging

BLI sensitively assesses tumor progression and detects tumor metastases in live animal models [77, 78]. Tumor cells transduced to express luciferase emit light following the addition of substrate. The photon emissions from the cancer cells are detected by a cooled charge-coupled device camera, quantified, and represented as a photo overlay to identify tumor cell location. Our laboratory has applied BLI to detect LNM in mouse models of mesothelioma [31, 79] and lung [79] (Fig. 2) and prostate cancers [80]. Metastatic LN are best detected with mouse positioned ventrally (mediastinal nodes), laterally (cervical nodes), or dorsally (retroperitoneal/para-aortic nodes). Bioluminescent tumor models require optimization to maximize LNM identification and minimize technical limitations. Firefly luciferase is typically preferred in tumor models over wild-type click beetle luciferase due to a stronger bioluminescent signal [81]. Mutant click beetle luciferases engineered to emit specific wavelength photons overcome the limitation of enzyme heterogeneity and can produce bright bioluminescent signals in vivo [82]. Marine luciferases, such as Gaussia and Reinella, are less favored due to rapid enzymatic kinetics and high background emissions. Transmission of bioluminescent signal depends upon the depth, viability, and vascularity of the tumor in addition to the scattering properties of blood and overlying tissue. Low tumor burden in metastatic LN decreases bioluminescent signal, possibly requiring longer exposure times (approximately three- to fourfold longer exposure, ~2 min), while large primary tumors may obscure signals from local LNM (as in the case of cervical LNM from models of head and neck cancers). Thus, parameters that may vary between animal models and require optimization include: substrate dose, route of substrate injection, time to imaging following substrate injection, image acquisition time, and mouse positioning. Furthermore, BLI has poor spatial resolution (1–2 mm) due to diffusely scattered photons and results in challenges to precisely localize tumor foci. Multi-modality imaging techniques, such as PET/CT data complemented by BLI, are being developed to address this limitation [83, 84]. As well, differential substrate specificities for beetle and marine luciferases (oxidizing D-luciferin or coelenterazine, respectively) allow for dual reporter systems. This system can be used for applications aimed at understanding biological mechanisms involved in metastases, (i.e., simultaneously tracking tumor and immune cells) [85]. Quantitative BLI of tumor burden is often limited by inherent properties due to the animal model used and extend to imaging LNM. Techniques aimed at the validation of quantitative BLI in various tumor models are described by several groups [31, 79, 86, 87]. Thus, BLI provides a sensitive and non-invasive technique for identifying and mapping LNM in addition to a platform for studying LNM biology.

Fig. 2.

BLI detects mediastinal lymph node metastases in an orthotopic mouse model of primary lung cancer. eGFP-firefly luciferase-expressing lung cancer cells were directly injected into lung parenchyma to create an orthotopic lung cancer primary tumor (arrows). Within 2–3 weeks, mediastinal BLI signal reveals lymph node metastases in the typical lymphatic drainage pathway (arrowheads).

Mediastinal LNM was confirmed in each mouse upon necropsy using H&E and IHC for tumor markers

Histopathological evaluation of LNM in mice

The imaging techniques described above provide accurate and efficient methods to noninvasively identify LNM in animal models. However, identification by bioimaging ultimately requires pathologic confirmation. The pathologic examination of murine LNM can be difficult, particularly in the setting of immunodeficient species and should involve an experienced comparative pathologist.

Gross evaluation of LN size and weight is not a reliable indicator of neoplastic involvement [88]. Systematic evaluation of specific TDLN basins can provide high yields of tumor-involved nodes. For a comprehensive overview of the anatomic details of the murine lymphatic system, we direct the reader to a study on mouse lymphatic anatomy by Van den Broeck et al. [89]. In this study, the authors catalog and illustrate the localization of murine LN after injection of India ink—providing a roadmap for LN harvesting required to confirm suspected metastases. Superficial LN (inguinal, axillary, or cervical/mandibular) are particularly easy to surgically resect and amenable to a variety of LN imaging modalities. However, identification and dissection of intracavitary nodes (intestinal, retroperitoneal, mediastinal, or pulmonary), often require fluorescent or tomographic imaging for identification. An “en bloc” dissection is often required to remove all the tissue in a nodal basin with subsequent LNM confirmation under high-power magnification using H&E staining and immunohistochemistry. Imaging techniques used to identify LNM in vivo can be extended to specifically identify the sites of metastases within an enlarged LN ex vivo [26] (Fig. 3).

Fig. 3.

Fluorescently guided identification of metastatic foci within enlarged lymph nodes. a Orthotopic implantation of breast cancer results in enlarged axillary LN. b Ex vivo analysis of an enlarged axillary LN with fluorescent imaging reveals metastatic foci [26] (Eisenberg et al., “Real-time intraoperative detection of breast cancer axillary lymph node metastases using a green fluorescent protein-expressing herpes virus”, Ann Surg, 243(6), 824–30)

Proper confirmation of tumor-involved LN requires evaluation of the size and cellularity of LN functional compartments, the presence of apoptosis, necrosis, and inflammation, hemorrhage, sinus histiocytosis and fat necrosis, and pigment accumulation [88, 90]. LN specimens exhibit considerable variation in histopathologic appearance due to differences in collection, embedding, and sectioning, as well as in individual differences depending on anatomic location and degree of inherent antigenic stimulation. The importance of simultaneously evaluating control mice cannot be overstated. In comparison to the ease of identifying metastatic foci in preserved human and murine LN, investigators are often faced with the challenge of differentiating a LN totally or near totally replaced by tumor from a solitary tumor nodule (Fig. 4). In the setting of lymphatogenous metastases, total or near total replacement of LN by tumor can occur. Verification of LNM should include tumor immunohistochemistry as the gross appearance and architecture of LN may be distorted. Attention should be paid to the periphery of putative tumor-involved LN identifying islands of immune cells and preserved LN architecture with immunostaining. LNs from immunodeficient mouse strains have markedly different immunostaining patterns and architectures (Fig. 5).

Fig. 4.

Metastatic human and murine lymph nodes exhibiting a spectrum of involvement. a Human metastatic breast cancer with focal lymph node involvement (arrow) at low (left) and high power (right) showing preservation of lymph node architecture. b Metastatic melanoma replacing a regional lymph node and obscuring architectural features at low (left) and high (right) powers. c Murine lymph node replaced by tumor, H&E (left), immunohistochemistry (right)

Fig. 5.

Immunohistochemical characterization of murine and human lymph nodes to assist in the identification of lymph nodes in murine models. a H&E and immunohistochemistry for B cells (B220) or T cells (CD3) in immunocompetent mouse strains (C57/B6 and BALB/c) and immunodeficient strains (nude athymic and NOD/SCID gamma). b H&E and immunohistochemistry for CD20 (B cells) and CD3 (T cells) in human lymph nodes

Conclusions

The goal of LNM imaging studies is to identify and monitor the progression of metastases in animal models of cancer. Here, we provide an overview for selecting high-yielding LNM cell lines for representative common solid cancers and delineate their biokinetics and LNM yield in murine cancer models (Tables 1 and 2). Furthermore, we survey non-invasive imaging modalities appropriate for LNM cancer models, discuss their strengths and limitations, and describe the use of lymphatic specific and metastasis-specific tracer agents as well as emerging LNM imaging technologies. We provide recommendations to improve the sensitivity and specificity of identifying LNM and illustrate the potential of BLI to identify LNM based upon our laboratory’s experience with murine models of mesothelioma, lung cancer, and prostate cancer. BLI provides a cost-effective means to screen large numbers of animals for LNM, and we highlight the use of tomographic methods in tandem with BLI to overcome its limitations in spatial–temporal resolution in order to precisely identify LNM. Finally, we detail the steps to ensure high yields of LNM identification and describe the precise spatial localization of characteristic immune cell lymph node architecture for a given species of mice—providing guidance in identifying LNM, even in lymph nodes completely replaced with tumor. In summary, this review provides a single platform for translational researchers to identify high-yielding LNM mouse models, non-invasive imaging modalities most commonly used to localize LNM, and the critical steps necessary to maximize yield and pathologic confirmation of LNM.