Illuminating Cancer Systems With Genetically-Engineered Mouse Models and Coupled Luciferase Reporters In Vivo

Abstract

Bioluminescent imaging (BLI) is a powerful non-invasive tool that has dramatically accelerated the in vivo interrogation of cancer systems and longitudinal analysis of mouse models of cancer over the past decade. Various luciferase enzymes have been genetically engineered into mouse models (GEMMs) of cancer which permit investigation of cellular and molecular events associated with oncogenic transcription, post-transcriptional processing, protein-protein interactions, transformation and oncogene addiction in live cells and animals. Luciferase-coupled GEMMs ultimately serve as a non-invasive, repetitive, longitudinal, and physiological means by which cancer systems and therapeutic responses can be investigated accurately within the autochthonous context of a living animal.

Introduction

Transitioning into the Proper Context

Genomic lesions within incipient cancer cells in collaboration with alterations in the microenvironment contribute to neoplastic progression (1-3). Tumor cells can modulate the surrounding microenvironment to promote the progression of cancer through intrinsic oncogenic pathways. Furthermore, key genetic lesions have a profound impact on cancer cell migration, invasion and regulation of the immune system through tumor-extrinsic manipulation of the microenvironment (4, 5). The importance of the host microenvironment in neoplastic progression, independent of tumor manipulation, is also underscored by studies demonstrating that fibroblasts, among many other stromal and immune cell types, stimulate growth of pre-neoplastic and neoplastic cells along with promoting drug resistance (6-8). Given these observations, understanding the complex interactions between genomic lesions and tumor microenvironment in mouse models is crucial to uncovering new anti-cancer therapies. Thus, implementation of molecular imaging within basic research and pre-clinical mouse models of cancer has become an essential tool for interrogating these hallmarks of cancer and monitoring tumor progression within the proper physiologic context.

Currently, the most commonly used types of mouse models of cancer can be grouped into primary tumor cells, tumor cell lines and their associated tumor engraftments, or genetically-engineered mouse models (GEMMs) of spontaneous cancer. Xenograft models entail subcutaneous or orthotopic transplantation of human cell lines or primary tumors into an immunodeficient mouse while mouse allografts similarly employ orthotopic or subcutaneous host implantation. Although traditional cell line xenografts and mouse allografts have yielded limited clinical correlations (9-11), the robust ability of human ‘xenopatient’ models (12) and newly adapted “human-in-mouse” (HIM) cancer models for accurately modeling patient disease and predicting patient response have been encouraging (11, 12, 13). However, due to the inherent nature of xenograft and HIM models, immune compromised mice are required, and thus the contribution of the immune system and the autochthonous tumor stroma cannot be fully interrogated. Additionally, these tumors are implanted as a population of late-stage tumorigenic cells and do not accurately recapitulate all steps of tumorigenesis. In contrast, GEMMs permit investigation of the proper tumor microenvironment, model tumor development from the initial genetic alteration in situ to subsequent neoplastic progression to metastasis, and enable tissue-relevant drug pharmacodynamics (13). Constitutive or conditional GEMMs of cancer (transgenic, knock-out or knock-in) as well as chimeric or non-germline GEMMs have proven to be of significant interest for cancer biology research as well as accurate predictive models of human cancers for pre-clinical drug development (14-16).

Molecular Imaging with Genetically-Encoded Reporters

Regardless of the mouse model, molecular imaging techniques (nuclear, fluorescence, and bioluminescence) at both macroscopic and microscopic scales make it possible to explore the consequences of the interactions between tumor cells and microenvironment during tumor progression in vivo, in real time. This expanding set of molecular probes, detection technologies, and imaging strategies, collectively termed molecular imaging, now provides researchers and clinicians alike, new opportunities to visualize gene expression, biochemical reactions, signal transduction, protein–protein interactions, regulatory pathways, cell trafficking, and drug action noninvasively and repetitively in their normal physiological context within living organisms in vivo (17-21). In particular, integration of genetically-encoded imaging reporters into live cells and, more importantly, whole animal mouse models of cancer has provided powerful tools to monitor cancer-associated molecular, biochemical, and cellular pathways in vivo (22). Traditional means of interrogating these oncogenic-associated biological processes and characterizing new anti-cancer therapeutics have relied on invasive techniques that are often laborious and only provide a static window of analysis. Microscopic fluorescence imaging with green fluorescent protein (GFP) provided pioneering studies of biological activities and cellular processes at high resolution (23). Concurrently, molecular probes, contrast agents, exploitation of fundamental tissue characteristics, and development of multi-spectral fluorescent and bioluminescent (luciferase) proteins and highly sensitive instrumentation, have revolutionized non-invasive and longitudinal imaging of cancer biology at the whole organism level.

These various imaging modalities and strategies acquire macroscopic information in vivo through two basic strategies: injected agents or genetically-encoded reporters. Injected agents have contributed significantly to pre-clinical cancer research and also have great potential for translation, but require significant optimization and characterization depending on the experimental model, biological target, background noise, instrumentation, route of administration, and, for human use, are impacted by similar regulatory hurdles as therapeutic agents (21, 22). An inherent constraint to the development of conventional injectable agents is that the details of synthesizing, labeling and validating a new and different ligand for every new receptor or protein of interest impose long cycle times on development. However, genetically-encoded reporters offer more modular tools for preclinical research, which once cloned into appropriate vectors and biologically confirmed, can be quickly applied to a broad array of applications with minimal modification (22, 24). While genetically-encoded imaging reporters are under development for use in humans, the potential for immunogenicity and transduction inefficiencies raise unique challenges (25). However, genetically-encoded imaging reporters represent a technically and biologically robust means of monitoring the dynamics of tumor biology with relatively high temporal resolution and various levels of spatial resolution when coupled with GEMMs.

Imaging of biological processes using genetically-encoded reporters relies on the ability of the reporter gene to produce a measureable signal that can be detected and quantified by extrinsic instrumentation. Reporter expression and thus signal output is controlled by a regulatory element such as constitutive or conditional DNA-promoter system, or subsequent peptide fusion that regulates posttranslational modulation of the reporter. Most commonly used genetically-encoded imaging reporters produce signal through optical imaging strategies, but magnetic resonance imaging (MRI) and radiopharmaceutical (PET/SPECT) approaches have been explored. Optical imaging of genetically-encoded reporters can provide image contrast through, 1) reporter-mediated enzymatic activation of an optically silent substrate (e.g., light-producing luciferase-based oxidation of D-luciferin in the presence of Mg²⁺, ATP, and O₂) (22, 26), 2) photo-excitation signal production (e.g., fluorescent proteins) (23), or 3) reporter-mediated enzymatic release/trapping of optically-tuned leaving groups (e.g., β-glucuronidase-mediated hydrolysis of glucuronide groups coupled to NIR imaging dyes (27)). Nuclear imaging of genetically-encoded reporters can utilize, 1) enzyme-mediated modification of a labeled substrate causing intracellular accumulation or proximal cell association (e.g., HSV1-TK-mediated phosphorylation of radiolabelled nucleosides for PET imaging) (28, 29), or 2) direct import of a labeled tracer (e.g., sodium iodide transporters/radioiodines for PET/SPECT) (22, 30). An early innovation for MRI was use of a galactopyranose blocking group coupled to a gadolinium-based relaxivity agent that rendered the MRI contrast agent sensitive to expression of the reporter gene β-galactosidase (31).

Genetically-encoded reporters with optical outputs, specifically fluorescence or bioluminescence, are most commonly used for cancer research in mouse models due to overall modest cost, sensitivity and lack of technical restrictions and required regulatory barriers often encountered with other approaches. Whole animal fluorescence imaging in vivo suffers from low signal-to-noise as a result of background auto-fluorescence, modeling-dependent photon quantification, photo-bleaching, low tissue penetration and low resolution (26). In addition, fluorescent proteins are known to generate reactive oxygen species (ROS) that can induce significant cellular stress under selected conditions (23). However, computed image analysis along with laser-induced fluorescence has increased the sensitivity of non-invasive fluorescence imaging in vivo (32). Also, compared to other genetically-encoded reporters, fluorescent proteins are independent of substrate delivery and pharmacokinetics, are amenable to high resolution microscopic analysis, and several new far red-shifted fluorescent proteins have been developed that enhance the penetration of photons in vivo (21).

Bioluminescence imaging has emerged as an invaluable optical imaging tool and has become widely adapted to molecular imaging of cancer models in vivo. The major advantages of luciferase reporter systems in vivo include essentially zero background signal, high signal-to-noise imaging, relative ease of signal acquisition, modest cost, user-friendly instrumentation and direct measure of live cell mass (ATP-dependent activity). Moreover, luciferase enzymes have a shorter half-life (~3-5 hrs for native North American P. Pyralis firefly luciferase and Renilla luciferase versus 12-26 hrs for native GFP variants) and are rapidly folded and functional post-translationally, thereby providing a more robust readout of kinetic processes such as transcriptional activation, protein degradation, reversible protein-protein interactions and other rapid biological processes (22, 33, 34). Also, red-emitting firefly and click beetle luciferases with relatively higher photon outputs have advanced luciferase imaging beyond the original long wavelength luciferase variants, providing further advantages over Renilla/ Gaussia luciferases and related mutants that emit at blue or blue-green wavelengths, which, while perhaps useful, remain suboptimal for imaging in vivo. However, luciferase enzymes in general are dependent on substrate pharmacokinetics, and furthermore, the Renilla/ Gaussia substrate (coelenterazine) is transported by P-glycoprotein and auto-luminesces due to auto-oxidation from serum albumin (35, 36), which can confound analysis in vivo. Additionally, due to overall low photon output, luciferase reporters traditionally have been limited to macroscopic imaging analysis. However, recent advances in low-light microscopy technologies have permitted the interrogation of live cells and live bioluminescent tissues ex vivo at high magnifications (37-39), a notable advance that extends the capacity of BLI.

Similar to studies in cultured cells, genetically-encoded bioluminescent reporters in mice offer the ability to non-invasively monitor transcriptional regulation, post-transcriptional and post-translational events, as well as transformation and neoplastic progression. When coupled with an oncogenic protein or signaling pathway of interest, these properties of BLI allow various features of cancer to be interrogated such as chemoresistance, inflammation, angiogenesis, DNA maintenance, apoptosis, therapeutic response and oncogene addiction. Additionally, BLI reporter mice can simultaneously and directly assess tumor burden through constitutive or conditional expression of luciferase. With the sensitivity of BLI reporter mice, one can also non-invasively survey and monitor small, non-palpable tumors as well as metastases in a relatively fast and efficient manner. While a considerable amount of early effort was directed toward fluorescent GEMMs of cancer, only more recently have mouse models been advanced with the versatility of BLI. Thus, the development and utility of BLI in GEMMs of cancer is the focus of this review and the reader is referred elsewhere for overviews of molecular imaging in non-GEMM models (19-22, 40). More specifically, this review highlights GEMMs of cancer reported in the last half-decade that utilize genomically-encoded bioluminescence reporters for investigating tumor biology and associated signaling pathways inherent to the cancer system or pathway of interest, summarizes notable models, and suggests future directions for BLI-coupled GEMMs of cancer. In addition to the models highlighted in detail below, an extensive referenced list of cancer-related luciferase-coupled GEMMs according to mode of luciferase regulation (Table I) and cancer type (Table II) are included for the general reader.

Table I.

Cancer-Associated Processes Observable in Genetically-Encoded Luciferase Reporter Mice.

Mode of Regulation	Cancer Biology Target	Genetic Strategy	Ref.
Transcriptional
	Cell cycle	KI(p21 promoter->Luc)	(37)
	B-cell specific imaging	KI(CD19 promoter->Luc)	(50)
	Various	KI(Cox-2 promoter->Luc)	(81)
	Various	Tg(Egr1 promoter-> Luc)	(82)
	Chemoresistance	KI(Mdr1a promoter->Luc)	(83)
	Lymphangiogenesis	KI(IRES-EGFP-Luc downstream of Vegfr3)	(51)
	Apoptosis	Tg(Birc5 promoter-> Luc)	(84)
	Telomerase regulation	Tg(hTERT BAC->Rluc)	(85)
	Angiogenesis	Tg(Vegf promoter->TSTA-Luc)	(86)
	Angiogenesis	Tg(Vegfa promoter->EGFP-Luc)	(87)
	Angiogenesis	Tg(Vegfr2 promoter->Luc)	(88)
	Various	Tg(NF-κB RE-> Luc)	(89)
	Various	Tg(Smad 2/3 RE-> Luc)	(90)
	Cre mediated Rb inactivation in pituatary gland.	Tg(Pomc promoter-> Luc)	(91)
	PDGF induced inactivation of Rb pathway	Tg(E2f1 promoter->Luc)	(67)
	PDGF-induced activation of Gli1/2	Tg(Gli1/2 responsive promoter->Luc)	(92)
	Prostate specific imaging	Tg(Pbsn promoter + androgen response elements->Luc)	(41)
	Systemic response to anti-androgen therapy	Tg(Slp-androgen response elements -TK->Luc)	(93)
	Afp activation in DEN induced HCC	KI(Afp promoter->TK-IRES-Luc)	(94)
	Afp activation in DEN induced HCC	Tg(Afp promoter -> Luc)	(95)
	Androgen deprivation in normal prostate, TAg, and JOCK models	Tg(syn. Pbsn promoter-> Luc)	(43)
	Androgen deprivation in normal prostate, TAg, and JOCK models	Tg(syn. Psa promoter -> Luc)	(43)
	Androgen deprivation in normal prostate and TAg model	Tg(Psa promoter-> Luc)	(42)
	Androgen deprivation in normal prostate and TAg model	Tg(Psa promoter-> Luc)	(44)
	Δ16Her2 induced mammary gland dysplasia	Tg(MMTV-Δ16Her2-IRES-Luc)	(96)
	SV40 ER induced pancreatic cancer monitoring	Tg(Rip1 promoter->SV40 ER-IRES-Luc)	(97)
	Estrogen receptor activation	Tg(Estrogen RE->Luc)	(98)
Post-Transcriptional
	ER Stress	Tg(CMV->XBP1-STOP-SA-Luc)	(55)
Post-Translational
	Hypoxia	Tg(Hypoxic RE->ODD-Luc)	(62)
	Hypoxia	KI(Rosa26 promoter->ODD-Luc)	(61)
	Hypoxia, Neu/ Beclin1+/fl induced mammary gland dysplasia	Tg(MMTV->neu; ODD-Luc); Beclin1fl/+	(60)
Protein-Protein Interactions
	Modular nuclear protein-protein interactions	Tg(Gal4 promoter-> Luc)	(64)
Transformation Cre-Activation
	Cre Recombination	Tg(Pomc promoter->Cre; Pomc promoter->Luc)	(91)
	Cre Recombination	Tg(β-actin promoter->flx-GFP-pA-flx-Luc-pA)	(68), (71)
	Cre Recombination	KI(Rosa26 promoter ->flx-STOP-flx-Luc)	(99), (70)
	Cre Recombination	Tg(CAG promoter->flx-STOP-pA-flx-TAg-Luc)	(100)
	Cre Recombination	KI(Notch1 promoter ->Notch1-Cre;->Notch1flx; Rosa26 promoter->flx-STOP-flx-CBR)	(72)
Tet-Regulated
	Coupled to MYCN expression	Tg(Glt1 promoter->tTA; tet-o-Mycn-Luc)	(79)
	TBX3 induced mammary gland dysplasia	Tg(MMTV->rtTA; tet-o-myc-Tbx3-IRES-Luc)	(101)
	Wnt1 induced mammary adenocarcinoma	Tg(MMTV->rtTA; tet-o-Wnt1-IRES-Luc)	(102)
	PyMT induced pancreatic cancer monitoring	Tg(Pdx1-rtTA; tet-o-PyMT-IRES-Luc)	(78)
	PyMT induced pancreatic cancer monitoring	Tg(Rip7 promoter->tTA; tet-o-PyMT-IRES-Luc)	(78)
	HPV16 E7 induced cervical cancer	Tg(Krt5 promoter->rtTA; tet-o-Luc- E7)	(103)

Tg = transgenic mouse; KI= knock-in mouse; -> indicates promoter driving gene expression; RE = response elements multiple; TK= thymidine kinase; gene loci are separated by a semicolon (;).

Table II.

Luciferase-Coupled Reporter Mice Utilized in GEMMs of Cancer.

Cancer Type	Reporter Regulation	Target Gene/ Process Investigated	Reporter Strategy	Ref.
Brain
	Transcriptional	Rb inactivation and tumor development in pituitary gland	Tg(Pomc promoter-> Luc)	(91)
	Transcriptional	PDGF inactivation of Rb pathway	Tg(E2f1 promoter->Luc0	(67)
	Transcriptional	PDGF activation of Gli1/2 and gliomagenesis	Tg(Gli1/2 responsive promoter-> Luc)	(92)
	Translational	MYCN driven medulloblastoma	Tg(Glt1 promoter->tTA; tet-o-Mycn-Luc)	(79)
Prostate
	Transcriptional	Androgen deprivation in normal prostate, TAg, and JOCK models	Tg(Pbsn promoter-> Luc)	(43)
	Transcriptional	Androgen deprivation in normal prostate, TAg, and JOCK models	Tg(Psa promoter -> Luc)	(43)
	Transcriptional	Androgen deprivation in normal prostate	Tg(Pbsn promoter + androgen response elements->Luc)	(41)
	Transcriptional	Androgen deprivation in normal prostate and TAg model	Tg(Psa promoter-> Luc)	(42)
	Transcriptional	Androgen deprivation in normal prostate and TAg model	Tg(Psa promoter-> Luc)	(44)
	Cre Recombination	Conditional Pten loss	Tg(β-actin promoter->flx-GFP-pA-flx Luc-pA)	(71)
	Cre Recombination	Conditional Pten loss	KI(Rosa26 promoter->flx-STOP-flx-Luc)	(70)
	Transcriptional	Systemic response to anti-androgen therapy	Tg(Slp-androgen response elements -TK->Luc)	(93)
Liver
	Transcriptional	Afp activation in DEN induced HCC	KI(Afp promoter->TK-IRES-Luc)	(94)
	Transcriptional	Afp activation in DEN induced HCC	Tg(Afp promoter -> Luc)	(95)
Breast
	Tet-on	TBX3 induced mammary gland dysplasia	Tg(MMTV->rtTA/; tet-o-myc-TBX3-IRES-Luc)	(101)
	Transcriptional	Δ16Her2 induced mammary gland dysplasia	Tg(MMTV->Δ16Her2-IRES-Luc)	(96)
	Transcriptional	Neu, Beclin+/- induced mammary gland dysplasia	KI(Rosa26 promoter ->ODD-Luc)	(60)
	Tet-on	Wnt1 induced mammary adenocarcinoma	Tg(MMTV->rtTA; tet-o-Wnt1-IRES-Luc)	(102)
Pancreas
	Tet-on	PyMT induced pancreatic cancer monitoring	Tg(Pdx1-rtTA; tet-o-PyMT-IRES-Luc)	(78)
	Tet-off	PyMT induced pancreatic cancer monitoring	Tg(Rip7 promoter->tTA; tet-o-PyMT-IRES-Luc)	(78)
	Transcriptional	SV40 ER induced pancreatic cancer monitoring	Tg(Rip7 promoter->SV40 ER-IRES-Luc)	(97)
Lung
	Transcriptional	Krasv12 induced lung tumorigenesis	Tg(β-actin promoter->flx-GFP-pA-flx-Luc-pA)	(68)
Lymphoma
	Transcriptional	B-cell lymphoma	KI(CD19 promoter->Luc)	(50)
Cervical
	Tet-off	HPV16 E7 induced cervical cancer	Tg(Krt5 promoter->tTA; tet-o-Luc- E7)	(103)
Spontaneous
	Transcriptional	Ubiquitous TAg induced tumorigenesis	Tg(CAG promoter->flx-STOP-pA-flx-TAg-Luc)	(100)
	Cre Recombination	Whole animal stochastic Notch1 loss of heterozygosity	KI(Notch1 promoter ->Notch1-Cre;->Notch1flx; Rosa26 promoter->flx-STOP-flx-CBR)	(72)

Tg = transgenic mouse; KI= knock-in mouse; -> indicates promoter driving gene expression; RE = response elements multiple; TK= thymidine kinase; gene loci are separated by a semicolon (;).

Regulation of Luciferase in GEMMs of Cancer

Transcriptional

Transcriptionally-regulated luciferase enzymes provide a robust tool to monitor tissue-specific tumor burden or interrogate biological processes in tumors in vivo. Transcriptional systems are simple in design and consist of a composite or endogenous promoter sequence upstream of luciferase and introduced into the mouse genome either through transgenic or targeted knock-in approaches. Conventionally, genetic regulatory elements derived from cytomegalovirus (CMV) or simian vacuolating virus 40 (SV40) provide robust protein or reporter expression in the cell or tissue harboring the construct. However, at the whole animal level, to track spontaneous tumor progression, researchers have utilized promoters or cis acting regulatory regions from endogenous or viral genes activated specifically in neoplastic cells. For example, prostate growth and development is largely governed by androgen signaling, thus offering an avenue to specifically image prostate cells in physiological or pathological states. With this in mind, a plethora of transgenic luciferase mice have been developed utilizing composite promoters from human kallikrein 2, probasin, prostate-specific antigen and various forms of concatenated minimal androgen response elements (Table I). Other transgenic models also use endogenous androgen-responsive promoters derived from rat probasin and human prostate specific antigen (41, 42). Baseline signal with varying intensities is confined to the prostate with minimal promoter activity outside of the prostate for most models, allowing non-invasive prostate-specific BLI. Prostate bioluminescent signal from these promoters correlates with normal prostate development and decreases upon castration or androgen ablation. However, in only one transgenic model (Tg(PSA->Luc)) did the reporter mouse demonstrate an increase in prostate bioluminescence when crossed to a previously characterized TRAMP model that expresses oncogenic SV40 small and large T antigen in the prostate via a minimal rat probasin promoter (42). Two other similar prostate transcriptional luciferase reporter models failed to show a consistent increase in prostate signal when crossed to TRAMP models despite histologically confirmed tumor progression (43, 44). It was suggested that the inability of these reporter models to show a tumorigenic increase in prostate bioluminescence, as seen in Tg(PSA->Luc; rPB-Tag) mice, was due to the androgen independence of these aggressive, neuroendocrine carcinomas that are characteristically observed on a FVB background (45). However, this observation is unsatisfying given the fact that if the tumors were truly AR-independent, the relatively high prostate-specific signal (>10⁶ photons) would be, by default, representative of tumor mass only and would have been expected to increase along with the tumor, which was not observed despite tumor progression (43, 44). These discrepancies point out the potential pitfalls of using transgenic strategies with regulated promoter-based reporters for readout of tumor burden which can be confounded by gene locus effects, gene silencing, or tumor evolution, independent of promoter- or gene-dependent luciferase expression. This discordance can be investigated by correlating bioluminescence with tumor burden as measured by an alternative means (caliper measurements, MRI, etc.). Conversely, bioluminescent promoter systems in transgenic GEMMs of cancer that are intended to interrogate gene-associated oncogenic processes can ultimately become a measure of tumor burden alone, completely separate from the original gene-associated biological intent. Thus, care should be taken when developing a reporter mouse using luciferase (or any reporter gene) to monitor tumor hallmarks when coupled to a gene of interest.

Transcriptional bioluminescent reporters knocked into an endogenous locus or immediately downstream of a start codon also provide a means to monitor tumor biology and development in vivo with decreased signal variability compared to that often observed between transgenic founder lines. Although technically more difficult, knock-in strategies maintain the entire promoter regulatory region adjacent to the luciferase gene and thus, in principle, provide a more accurate measure of gene transcription within the native context of the genome. This is exemplified in a p21^Waf1/CIP knock-in luciferase model (p21^FLuc) in which firefly luciferase was genetically introduced into the endogenous p21 locus downstream of the native promoter (Fig. 1A) (37). p21 is a critical regulator of cell cycle progression, is a direct transcriptional target of p53 among many other signaling pathways and is frequently altered in human cancers (46). As expected, p53-dependent activation of the p21 promoter in response to external beam irradiation (IR) could be non-invasively and repetitively monitored over time in p21^FLuc mice (37) using surgically implanted micro-osmotic pumps that constantly delivered D-luciferin substrate (47) (Fig1. B). Previous attempts to non-invasively monitor p21 levels in vivo utilized transgenic transcriptional luciferase or lacZ reporter mice that were regulated by short fragments of the p21 promoter (<5 kb) and only produced robust signal when strains harbored multiple copies of the reporter (up to 23) (48, 49). In the knock-in reporter strain, baseline bioluminescence levels three logs higher than the luciferase transgenic strain allowed Tinkum et al. (37) to identify specific organs that contained high levels of p21 independent of p53 status. Additionally, select organs showed dramatic regional differences in p21-luciferase activity that was identified using high-resolution bioluminescence microscopy to localize live sub-organ structures and specific cell populations with high-level expression of p21 (Fig. 1C), providing new insight into future lines of investigation.

Figure 1. Whole animal imaging of p21 promoter activity.

(A) Schematic representation of p21^FLuc reporter mice with luciferase knocked into the endogenous p21 locus. (B) Non-invasive, whole animal imaging of p53-dependent p21 promoter activity in response to radiation in p21^+/FLucTrp53^+/+ and p21^+/FLucTrp53^fl/fl mice. (C) Low-light, bioluminescence microscopy of p21 promoter activity in various p21^+/FLuc live tissues, including the villi from the small intestine, throughout the liver, in the epithelial cell layer below the keratinized penile spines, as well as the epithelial cell layer of the vagina. Bars, 200 μm; 50 μm for the penis. Images from Tinkum, et al, (37) were modified and reprinted with permission from Molecular and Cellular Biology.

Similar to organ-specific imaging in the aforementioned prostate transgenic models, sub-cellular whole animal imaging of tumor burden can be accomplished using transcriptional reporter mice as well. Scotto et al. (50) introduced a mCherry-Luciferase fusion into the endogenous CD19 locus. This enabled non-invasive longitudinal imaging and microscopic analysis ex vivo of the B-cell lineage under normal and pathologic conditions when crossed to a λ-MYC transgenic mouse model of spontaneous B-cell lymphoma (50). Knock-in strategies appear to offer more refined imaging of whole organs or distinct cell lineages in terms of sensitivity and specificity when it comes to analysis of GEMMs of cancer at the macroscopic scale.

However, a caveat is that knock-in strategies can potentially disrupt expression of the endogenous locus by usurping its promoter function or preventing expression of the targeted locus, thereby disrupting the process under investigation. In this regard, several groups have adapted viral internal ribosome entry sites (IRES) to couple translation of luciferase enzymes to the transcriptional activation of an upstream gene in a bicistronic fashion while maintaining expression of the targeted gene. One group utilized this strategy by knocking-in an IRES-EGFP-luciferase fusion downstream of the endogenous vascular endothelial growth factor receptor 3 stop codon (Vegfr3^EGFPluc) (Fig. 2A) (51). VEGFR3 is a potent regulator of angiogenesis, lymphangiogenesis and metastasis and is emerging as an alternative target in combination with other VEGF anticancer therapeutics (52-54). The authors used this reporter mouse to quantify the association between inflammation and lymphangiogenesis during wound healing and in response to a contact hypersensitivity inflammation model. Microscopic analysis of the coupled EGFP reporter allowed the authors to show that Vegfr3^EGFPluc luciferase intensity correlated with increased lymphatic network density. Additionally, tumor-activated lymphangiogenesis was observed in DMBA/TPA-skin papillomas and at lymph nodes distant from subcutaneous injection of B16-V5 melanoma cells (Fig. 2B-C). Importantly, the sensitivity and longitudinal capability of luciferase imaging permitted identification of tumor-activated lymphangiogenesis at distant lymph nodes that preceded tumor metastasis (Fig. 2C).

Figure 2. Imaging inflammation and tumor-associated lymphangiogenesis.

(A) Schematic representation of the Vegfr3^EGFPluc reporter knocked in downstream of the endogenous Vegfr3 locus, which utilizes an internal ribosome entry site (IRES) for monitoring Vegfr3 expression. (B) DMBA/TPA-induced skin papillomas in Vegfr3^EGFPluc/+ reporter mice displayed localized lymphangiogenesis as indicated by the black arrows. (C) Whole body imaging of tumor-activated lymphangiogenesis over time in a B16-V5 melanoma xenograft model at distant lymph nodes (red arrows) prior to metastasis of the primary tumor xenograft in female Vegfr3^EGFPluc/+ reporter mice. Images from Martinez-Corral, et al, (51) modified and reprinted with permission from Proceedings of the National Academy of Sciences.

Post-Transcriptional

The luciferase enzyme can also be coupled to post-transcriptional mechanisms that monitor mRNA modifications associated with oncogenic signaling. This can be accomplished by coupling or fusing a luciferase enzyme to a coding sequence such that modification or interaction of the mRNA or protein results in modulation of luciferase signal. This strategy was applied to investigate the effect of the tumor microenvironment on tumor unfolded protein response (UPR) and endoplasmic reticulum (ER) stress in live animals by visualizing alternative splicing (55). The heterogeneous tumor microenvironment imposes ER stress upon the tumor through hypoxia, acidic pH and low nutrients (56). These constraints, along with deregulated translation and proteotoxicity place evolutionary pressure on developing cancer cells, which can respond by augmenting several key steps of the UPR pathway for survival. Thus, UPR is an emerging target for anti-cancer therapies (56). Upon loss of ER homeostasis, activation of inositol-requiring 1α (IRE1α), one of the three UPR pathways, results in unconventional splicing of a 26 bp intron from IRE1α-X-box binding protein 1 (XBP-1), thereby incorporating an extended open reading frame that increases protein stability and augments XBP-1 transcriptional activation of essential UPR response genes (57, 58). Spiotto et al. (55) created a transgenic mouse that harbors a CMV->XBP1-Luc transgene which is regulated in a manner similar to that of the endogenous XBP1 and thus, luciferase expression is a direct readout of this splicing event (55). Under normal physiological conditions, the reporter mouse maintained background photonic levels, but appropriately displayed tumor-specific signal once crossed to breast cancer models such as, Tg(MMTV-TAg) and Tg(MMTV-Her2), that localized with ER stress markers upon microscopic analysis of tumors. Importantly, the authors observed no correlation between tumor size and bioluminescent signal suggesting XBP1-Luc was a measure of the tumor-intrinsic ER stress and not overall tumor load. Tumors arising within the same mouse possessed variable signal intensities, which portrayed the heterogeneic nature of tumor metabolism as further indicated by differential glucose uptake and hypoxia, and the contribution of the unique tumor microenvironments within the same mouse. These observations serve as an example of the sensitivity of BLI compared to other optical imaging modalities that were employed in a similar transgenic Tg(XBP1-GFP) mouse, which had very low signal that at the time only allowed endpoint analysis of XBP1 activation in a few extracted organs (59).

Post-Translational

Cancer-associated post-translational modifications and the subsequent effects on protein processing and protein-protein interactions are amenable to luciferase reporter mice and BLI in general. Previous designs and biologically affirmed reporters have transitioned from cell culture to provide the framework for whole animal preclinical evaluation of anti-cancer drugs in the proper physiological context. Interrogation of inhibitors of the hypoxia inducible transcription factor 1α (HIF-1α) have been aided by the development and application of various transgenic mouse models that fused the oxygen-dependent degradation domain of HIF-1α to firefly luciferase (Tg(ODD-Luc) and Tg(Hypoxic RE->ODD-Luc)) (60-62). Under normoxic conditions, endogenous HIF-1α protein is retained at low levels due to hydroxylation, polyubiquitination and subsequent proteasomal degradation and correspondingly, luciferase background signal levels are low in these luciferase reporter mice during normoxia (63). During hypoxia or hydroxylase inhibition, HIF-1α is stabilized and appropriately, bioluminescence intensity of these reporter mice increases as a result, thereby indirectly monitoring HIF-1α-dependent responses to acute or chronic hypoxia. When crossed to spontaneous Tg(MMTV-neu);Beclin1^+/fl or carcinogen susceptible RasH2 cancer models, increased bioluminescence was detected in hypoxic tumors, highlighting the ability to monitor both tumor growth and tumor hypoxia non-invasively with this reporter mouse (60, 62). Additionally, inclusion of cis acting hypoxia response elements before a minimal CMV promoter in the Tg(HRE->ODD-Luc) mice inherently provided interrogation of the transcriptional phase of the HIF-1α positive feedback loop, which attempts to recapitulate HIF-1α-dependent transcriptional activation of its own mRNA (62).

The analysis of druggable oncogenic protein-protein interactions can also be interrogated non-invasively using luciferase reporter mice. This notion is exemplified in a proof-of-principle mouse model utilizing the Gal4-VP16 “two hybrid” interaction system in which nuclear interaction of the DNA binding domain of the yeast transcription factor Gal4 with the transactivation domain from herpes simplex virus VP16 protein can functionally lead to the transcriptional activation of a Gal4 responsive reporter gene (64). If each component is individually fused to a set of proteins known to interact, visualization of their proximity and interaction can be indirectly assessed through Gal4-responsive reporter activation. Pichler et al. (64) generated a transgenic mouse harboring a luciferase reporter regulated by Gal4 response elements which could indirectly monitor protein-protein interactions by hydrodynamic somatic gene transfer of constructs expressing Gal4 fused to p53 and VP16 fused to the SV40 large T antigen. Using this reporter mouse, abrogation of p53-TAg interaction due to loss of p53 was readily observed in mice upon shRNA-mediated knockdown of p53 in vivo. Adapting this modular system to other models of protein interactions could aid in the preclinical evaluation of modulators of oncogenic protein-protein interactions longitudinally in whole animals (65).

Conditional Transformation

Exquisite genetic techniques have enabled researchers to initiate and follow transformation, progression, invasion, metastasis, therapeutic response and oncogene dependence in spontaneous or conditional GEMMs of cancer. Luciferase reporter mice can be genetically coupled to these molecular and biological events, thus permitting longitudinal and non-invasive imaging of a relatively small cohort of mice that can provide statistically meaningful results since each animal serves as its own control. In one example, inactivation of the retinoblastoma (RB) tumor suppressor pathway in response to platelet-derived growth factor (PDGF)-induced oligodendrogliomas was indirectly monitored through activation of a E2F1 promoter driving luciferase in an engineered transgenic reporter mouse (66, 67). Direct monitoring of genetic deletion of tumor suppressors or activation of oncogenes is also possible through the use of several floxed firefly luciferase transgenic mice in which Cre-mediated excision of an upstream floxed-stop cassette allows downstream luciferase expression. Previously, Lyons et al. generated a transgenic strain in which the beta actin promoter-driven luciferase expression is regulated by removal of a floxed GFP-polyA transgene (68). Crossing the reporter strain with a floxed Kras2^v12 followed by adeno-Cre inhalation induced lung adenocarcinomas that could be simultaneously monitored using BLI for over 100 days. Using the same reporter mouse crossed to a conditional prostate-specific Cre-expressing strain, Tg(PB-Cre4);Pten^fl/fl, another group was able to monitor for over 400 days spontaneous prostate adenocarcinoma initiation, progression, response to castration and subsequent development of castration-resistant prostate cancers (CRPC) reliably in a small cohort of animals (69). The emergence of CRPC was not observed and potentially may never be observed through coupled-reporter gene imaging in the androgen-dependent transcriptional prostate carcinoma reporter models discussed earlier. This highlights the utility of Cre/loxP approaches, which can mark tumor cell lineage prior to biologic-specific functions, thereby improving upon the complexities of transcriptional reporters, such as the prostate cancer reporter mice discussed above. The authors observed that the non-recombined β-actin promoter-driven reporter was leaky and read-through could be observed in muscle tissue using reverse transcriptase PCR of tissue mRNA. This was potentially due to the strength of this promoter in regions of high actin expression and/or from the variability associated with the loci or extent of genomic integration of the reporter. Also, the authors observed pronounced luciferase signal at intraperitoneal sites of repeated luciferin injection, which could be remedied through tail vein injection of the luciferase substrate. Svensson et al. also utilized a conditional Rosa-Luc knock-in reporter mouse to monitor prostate carcinoma progression on the less aggressive C57BL/6 background (70), using the same strategy as Liao et al. (Tg(PB-Cre4);Pten^fl/fl) (71). Despite the now well-characterized differences in prostate development due to the genetic backgrounds of the mouse used (e.g., C57BL/6 versus aggressive FVB/N), Svensson et al. noted a dramatic reduction in luciferase signal variability over time compared to Liao et al. These differences potentially stem from the Rosa26 locus and/ or the fact that the Rosa-Luc mouse was strategically backcrossed onto an albino C57BL/6 background, thereby greatly reducing signal attenuation due to coat color. Additionally, analysis of Tg(PB-Cre4);Pten^fl/fl prostate tumors indicated the BLI was a more accurate readout of prostate tumor burden because ex vivo analysis revealed massive fluid retention in the anterior prostate that could be misinterpreted as tumor mass when analyzed via MRI.

Cre/LoxP-luciferase reporter mice have also been used for following stochastic neoplastic genetic lesions and marking distinct cell lineages when coupled to other fluorescent or LacZ reporter strains. Liu et al. implemented this strategy to identify organ susceptibility and monitor subsequent tumor progression in a sophisticated whole animal Notch1 loss-of-heterozygosity (LOH) model crossed to a floxed stop cassette-Rosa26-click beetle red luciferase (Rosa->CBR) knock-in reporter mouse (Fig. 3) (72). The authors generated a mouse harboring a knock-in non-functioning Notch1-Cre fusion on one allele along with a conditional Notch1 knockout cassette on the second allele (Notch1^fl), which was further crossed with either the conditional Rosa->CBR mouse or other conditional Rosa->LacZ and Rosa->EYFP reporter mice strains (Fig. 3A). Following the first round of embryonic Notch1 expression, any second endogenous ligand-dependent activation of Notch signaling in Notch1-Cre;Notch1^fl mice results in cleavage and subsequent nuclear translocation of the Notch-intracellular-domain (NCID)-Cre fusion, which in turn results in excision of the remaining floxed Notch1 allele throughout all active Notch1-signaling cells (Fig. 3A). Thus, stochastic LOH and the subsequent development of highly vascularized tumors could be indirectly monitored at the macroscopic (Rosa->CBR) and at the cell lineage level (Rosa->LacZ/EYFP) by Cre-mediated excision of the co-engineered lox-stop-lox cassette inserted upstream of the Rosa locus on each allele in these reporter mice (Fig. 3B, C). The Rosa->CBR mouse has since been crossed with the Rosa->LacZ mouse to create a conditional Rosa->CBR/ LacZ dual-modality reporter mouse that has been extensively backcrossed onto the albino C57BL/6 background (Piwnica-Worms, D. et al. unpublished). This dual reporter mouse has the potential to provide a robust and powerful readout of Cre-activation and carcinogenesis through the high, red-emitting photonic output of CBR and the microscopic utility of LacZ staining.

Figure 3. Imaging whole animal Notch1 loss-of-heterozygosity (LOH) and surveying subsequent tumor development.

(A) Schematic representation of Notch1-Cre;Notch^fl; Rosa->CBR cells prior to Notch1 ligand activation. (B) Upon activation, cleavage of NOTCH1 (between S2 and S3) permits translocation of the Notch-intracellular-domain (NCID)-Cre fusion into the nucleus where it excises both the remaining wildtype Notch1^fl allele, and the floxed stop cassette preceding CBR. Through these series of Cre-mediated excisions, lineage tracking of Notch1 LOH can be longitudinally imaged via the genetically-coupled floxed Rosa->CBR reporter. (C) Whole animal imaging of the development and progression of Notch1 LOH-induced angiosarcomas of the liver as imaged in Notch1⁺or Notch1^fl mice crossed to Notch1-Cre;Rosa->CBR reporter mice. Images from Liu, et al, (72) modified and reprinted with permission from Journal of Clinical Investigation.

Tet-Regulated Systems

Conditional cancer mouse models using genetically-coupled luciferase reporters can also utilize tetracycline (or the more stable analogue, doxycycline)-regulated expression systems or tamoxifen-inducible systems for spatial and temporal induction and reversion of oncogene biology in mice. However, tetracylcine (tet)-systems are more commonly used due to toxicities associated with tamoxifen and the Cre-estrogen receptor fusion observed in tamoxifen-inducible mice (73, 74). In the tet-on system, expression is dependent on binding of a reverse tetracycline-regulated transactivator (rtTA) to the tetracycline-response promoter elements in the tet-operator (tet-o) engineered upstream of the coding sequence of interest in the presence of tetracycline (75). The tet-off system uses a different tetracycline-regulated transactivator protein (tTA) which cannot bind tet-o in the presence of tetracycline, effectively silencing expression only when tetracycline is present (75-77). Regardless of the system, expression of the tetracycline transactivator (and hence the gene of interest) can be regulated through promoters specific to a tissue or cell type of interest. Ultimately, these systems have been instrumental in determining the extent of oncogene dependence in spontaneous mouse tumor models within the proper context in vivo. Additionally, conditional repression of an oncogene mimics therapeutic inhibition in a targeted molecular pathway and coupled luciferase reporters allow longitudinal imaging confirmation of this ‘therapeutic inhibition’. In one example, Du et al. generated a conditional tet-o-polyoma middle T antigen (PyMT)-IRES-luciferase reporter mouse, Tg(tet-o-PyMT-IRES-Luc), to investigate the cell-specific effect of oncogene induction on the development and progression of pancreatic cancer using two previously reported conditional Tet-system mice Tg(Rip7-rtTA and Pdx1-tTA) (78). Within 1 day of doxycycline removal, non-invasive BLI imaging allowed confirmation of subsequent oncogene withdrawal in Tg(Rip7-rtTA; tet-o-PyMT-IRES-Luc) mice. Interestingly, this had no effect on the established hyperplastic β cell islets, indicating oncogene independence, which was also conversely confirmed using the Pdx1-tTA mice. Additionally, 10% of Tg(Pdx1-tTA; tet-o-PyMT-IRES-Luc) mice developed aggressive acinar cell carcinomas as a result of activation of the Pdx1 promoter in early pancreatic progenitor cells. Regression of these tumors when deprived of PyMT, confirmed as soon as 1 day after doxycycline addition by BLI, indicated a requirement or dependence on PyMT for sustained tumor maintenance. Another group generated a tet-off luciferase reporter mouse to delineate the requirement of MYCN in medulloblastoma (79). One of the two lines crossed to generate this bigenic mouse contained the cerebellum-specific glutamate transporter1 (Glt1) promoter driving expression of tTA and the other a tet-o-driven bidirectional MYCN and firefly luciferase expression cassette. Using a characterized amount of photon flux as a measure of tumor burden, the authors repressed MYCN expression in a subset of mice by feeding them doxycycline-containing chow. Within one week, they observed a dramatic drop in bioluminescence and tumor regression, confirming the requirement for sustained MYCN expression in these tumors and suggesting its potential for targeted therapy in medulloblastomas. Thus, bioluminescent reporter mice coupled to integrated conditional tet-regulated systems are invaluable tools to quickly and efficiently monitor oncogene expression in manageable cohorts of mice and validate potential therapeutic targets, which guide and inform more invasive and laborious secondary tumor analyses.

Lessons Learned and Future Considerations

As with all experiments, meticulous planning and a dose of foresight are paramount to the success of engineering a genetically-encoded luciferase mouse. There are several considerations and nuances that can dramatically reduce time and labor and expand upon the potential of a reporter mouse in terms of signal sensitivity, biological accuracy and overall utility. As the genetic background of the mouse strain can dramatically affect the biology of the cancer, for example, as seen in the differential sensitivity of C57BL/6 and FVB strains to prostate cancer models as well as other models (80), so will genetic background inherently modulate the signal strength and interpretation of coupled luciferase reporters. When possible, an albino mouse strain should be used to minimize signal attenuation. This will allow for enhanced sensitivity for gross analysis of luciferase expression, and will strongly benefit low luciferase-expressing tissues in scenarios such as low endogenous expression, metastasis or tumor regression. Commercial albino C57BL/6 embryonic stem cells are now available that have high germline rates, and are technically amenable to genomic manipulation for targeted or transgenic reporter approaches, thereby minimizing onerous backcrossing. When specifically attempting to monitor live tumor cells, a relevant genetically-encoded reporter should be specific to live cell mass alone and will likely be most accurate when using Cre/loxP-based or knock-in strategies as discussed previously. Although technically much simpler, transgenic transcriptional reporters can be overridden by tumor evolution or genomic loci effects as seen in the androgen-sensitivity GEMMs of prostate cancer discussed above. Low light microscopy, which is capable of imaging live luminescent tissues, is becoming accessible to the general researcher and is approaching the high levels of magnification and resolution necessary for subcellular inspection. Nonetheless, analysis ex vivo of live tissues and organs synergizes with the whole animal imaging capabilities of luciferase reporter mice and also can be performed using luciferase antibodies or secondary coupled reporters such as fluorescent proteins as used to observe lymphatic vessels in Vegfr3^EGFPluc mice (51).

Conclusions

Genetically-encoded luciferase reporter mice have made a profound impact on imaging tumorigenesis, cancer progression, response to therapies and the contributions of the tumor microenvironment when crossed with GEMMs of cancer. Compared to other imaging modalities, bioluminescence provides an efficient, relatively low cost, non-invasive and longitudinal means to investigate genetic alterations in the autochthonous tumor environment and its ultimate effect on tumor biology. Combining the advantages of genetically-encoded luciferase reporters with the development of new and clinically accurate GEMMs of cancer paints a bright horizon for our understanding of molecular cancer biology and the development of novel and durable anti-cancer therapies.

Statement of Significance.

• Deciphering the complex molecular interactions between cancer somatic lesions and host microenvironment are crucial to understanding cancer cell proliferation, migration, invasion, and immune evasion, and may guide new anti-cancer therapies.

• Of equal importance as digitized genomics, the next essential challenge is to functionalize the cancer genome and to correctly capture these molecular mechanisms in their proper biological context within cancer systems.

• Molecular imaging with genetically-encoded imaging reporters, especially bioluminescence, provides a dynamic and noninvasive analysis platform to resolve the cooperative genetic elements of cancer systems at various temporal and spatial scales.

• Bioluminescence reporters expressed in live cells and mouse models of cancer have provided powerful tools to monitor cancer-associated genetic circuits, signaling pathways, cell lineage, and drug-targeted protein function in real time in vivo.

Abbreviations

GEMM

genetically-engineered mouse model

BLI

bioluminescence imaging

PET

positron emission tomography

SPECT

single photon emission computed tomography

MRI

magnetic resonance imaging