pii: s1470-2045(02)00848-3

For personal use. Only reproduce with permission from The Lancet Publishing Group.

THE LANCET Oncology Vol 3 September 2002 http://oncology.thelancet.com546

We have developed a way of imaging metastases in miceby use of tumour cells expressing green fluorescentprotein (GFP) that can be used to examine fresh tissue,both in situ and externally. These mice present many newpossibilities for research including real-time studies oftumour progression, metastasis, and drug–responseevaluations. We have now also introduced the GFP gene,cloned from bioluminescent organisms, into a series ofhuman and rodent cancer-cell lines in vitro, which stablyexpress GFP after transplantation to rodents withmetastatic cancer. Techniques were also developed fortransduction of tumours by GFP in vivo. With thisfluorescent tool, single cells from tumours and metastasescan be imaged. GFP-expressing tumours of the colon,prostate, breast, brain, liver, lymph nodes, lung, pancreas,bone, and other organs have also been visualisedexternally by use of quantitative transcutaneous whole-body fluorescence imaging. GFP technology has alsobeen used for real-time imaging and quantification ofangiogenesis.

Lancet Oncol 2002; 3: 546–56

The study of microscopic cancer is essential for theunderstanding and control of cancer dormancy, growth, andcolonisation of distant sites.1 Several approaches involvingtumour-cell labelling have been developed for visualisingtumour cells in vivo. The Escherichia coli beta-galactosidase(lacZ) gene has been used to detect micrometastases.2

However, lacZ detection requires extensive histologicalpreparation and sacrifice of the tissue or animal. Therefore,other techniques are required for real-time imaging andstudy of tumour cells in viable fresh tissue or living animals.

Use of skin-fold chambers, exteriorisation of organs, andsubcutaneous windows inserted with semi-transparentmaterial3–5 has yielded insights into microscopic tumourbehaviour. However, these techniques are only suitable forectopic models3 or for investigations with short periods ofobservation.4,5 The difficulties of maintaining windows orother devices that are made with heavy, stiff, or other types offoreign materials limit the length of time that these tools canbe used in vivo.3,4 Cutaneous windows made with polyvinylchloride film become opaque or detached after time, whichprecludes their use in long-term studies.4 However, in micewith green fluorescent protein (GFP)-expressing tumours(figure 1), gene expression, angiogenesis, and physiologicalproperties can be studied by use of the dorsal skin-chambercombined with multiphoton confocal microscopy.3 A

disadvantage to this technique is that the chamber limitsinvestigation to the ectopic primary tumour.

Mice bearing GFP-expressing microscopic tumours onexteriorised organs can be examined with intravitalmicroscopy. However, with this approach, the mice tend notto survive long enough to enable spatial–temporal studies oftumour dormancy, progression, and metastasis.5

Weissleder and colleagues6,7 infused tumour-bearinganimals with probes that fluoresce at an infrared frequencywhen activated by protease activity. Tumours with

Review GFP imaging in vivo

RMH is President of AntiCancer Inc and Professor of Surgery at theUniversity of California, CA, USA.

Correspondence: Dr Robert Hoffman, AntiCancer Inc, 7917 OstrowStreet, San Diego, CA 92111, USA. Tel: +1 858 654 2555. Fax: +1 858 268 4175. Email: [email protected]

Green fluorescent protein imaging of tumourgrowth, metastasis, and angiogenesis in mousemodels

Robert M Hoffman

Figure 1. GFP transfectants in veins and capillaries. To study the limit of detection of GFP transfectants in vivo, a nude mouse was sacrificed 2 minutes after Chinese hamster ovary-GFP cells were injected into thetail vein. The fresh organ tissues removed from the mice were examinedby use of fluorescence microscopy. GFP-expressing clone-38 cellsformed emboli in a capillary of the right adrenal gland. Bar=100 �m.26


THE LANCET Oncology Vol 3 September 2002 http://oncology.thelancet.com 547

appropriate proteases activated theprobes and could be seen externally.This system is limited by very highliver-to-tumour background fluores-cence, which means that metastasis tothe liver–among the most importantmetastatic sites–cannot be studied.Furthermore, the time limit for studieswas 96 hours, so growth and efficacystudies were not possible. Tumoursmust have appropriate protease activityto be detectable, and the probes mustbe delivered selectively to the tumour.

Another approach for visualisingtumour cells in vivo involves insertionof the luciferase gene into tumour cellscausing them to emit light.8,9 However,once transferred to mammalian cells, luciferase enzymesrequire exogenous delivery of their luciferin substrate, whichis an impractical procedure in an intact animal. Because ofthe low image resolution and signal achieved with thisapproach, it takes a substantial amount of time to collectsufficient photons to form an image from an anesthetisedanimal.8 Also, whether luciferase genes can function stably intumours over long periods and in the metastases derivedfrom them, is not known.

Green fluorescent proteinIn order to externally image and follow the natural course orimpediment of tumour progression and metastasis, highspecificity, a strong signal, high resolution, and goodphysiological conditions are necessary. The GFP gene, clonedfrom the bioluminescent jellyfish Aequorea victoria,10 waschosen to satisfy these conditions since it has great potentialfor use as a cellular marker.11,12 GFP cDNA encodes a 283aminoacid monomeric polypeptide with a molecular weightof 27 kDa13,14 that requires no other Aequorea proteins,substrates, or cofactors to fluoresce.15 Recently, gain-of-function bright mutants expressing the GFP gene have beengenerated by various techniques16–19 and have been humanisedfor high expression and signal.20 Red fluorescence proteins(RFP) from the Discosoma coral have also been described andshould prove useful for in vivo imaging studies.21–23

Stable transduction of tumour cells with GFPSeveral groups have selected tumour cell lines to stably expressGFP at high levels both in vitro and in vivo. These cells can betransplanted into animals and visualised in situ in fresh tissues(figure 1).24–28 Furthermore, tumour cells expressing GFP havebeen visualised with or without subsequent colonisation in all the major organs including liver, lung, brain, spinal cord, axial skeleton, and lymph nodes.24–28 GFP models ofmetastatic disease have been developed for lung cancer,27,29–32

prostate cancer,33,34 melanoma,35 colon cancer,36 pancreaticcancer,37,38 breast cancer,39–43 ovarian cancer,26,44,45 and braincancer.46,47 This review shows that tumour cells transfectedwith the GFP gene are a powerful tool for in vivo visualisationof tumour growth, angiogenesis, dormancy, dissemination,invasion, and metastasis.

Ex vivo imaging of tumours in animal modelsOvarian cancerStudies in which GFP-expressing Chinese hamster ovarytumour fragments (CHO-K1-GFP) of about 1 mm3 wereimplanted into the ovarian serosa of nude mice by surgicalorthotopic implantation (SOI) resulted in the developmentof ovarian tumours.26,35 The tumours, which were stronglyfluorescent, subsequently spread throughout the peritonealcavity, including the colon, cecum, small intestine, spleen,and peritoneal wall. GFP fluorescence was used to tracktumour spread; numerous micrometastases were detectedon the lungs of all mice and multiple micrometastasis werealso detected on the liver, kidney, contralateral ovary,adrenal gland, para-aortic lymph node, and pleuralmembrane. Single fluorescing cells could be visualised withGFP that could not be detected with standard histologicaltechniques.26,44

Additional studies, in which CHO-K1-GFP cells wereinjected into the tail vein of nude mice, showed thattumours, and even single cells, could be detected by GFPfluorescence in the peritoneal wall vessels.26 These cellsformed emboli in the capillaries of the lung, liver, kidney,spleen, ovary, adrenal gland, thyroid gland, and brain.

Lung cancerGFP expressing cells from the highly metastatic human cellline ANIP were implanted by SOI onto the left lungs of nudemice.27,28 The advancing margin of the tumour, which spreadthrough the ipsilateral lung, was visualised by GFPfluorescence. All the animals showed evidence of chest-wall invasion and local and regional spread. Metastaticcontralateral tumours involved the mediastinum,contralateral pleural cavity, and the contralateral visceralpleura. Whereas the ipsilateral tumour had a continuous andadvancing margin, the contralateral tumour seemed to havebeen formed by multiple seeding events. Involvement ofcontralateral hilar and cervical lymph nodes was alsovisualized by GFP expression. When non-GFP-transfectedANIP was compared with GFP-transformed ANIP formetastatic capability similar results were seen.27,28

Mice implanted by SOI with GFP-expressing human lungcancer cells (H-460-GFP) were sacrificed after 3–4 weeks

ReviewGFP imaging in vivo

Figure 2. Whole-body external images of murine melanoma (B16F0-GFP) metastasis in brain. Themetastasis were imaged by GFP expression under fluorescence microscopy. Clear images ofmetastatic lesions in the brain can be seen through the scalp and skull. (a) 14 days after injection ofGFP-expressing tumour cells. Bar=1280 �m. (b) 20 days after injection. Bar=1280 �m. (c) 25 daysafter injection. Bar=1280 �m.36



when there was a significant decline in performance status.GFP fluorescence showed that there were metastases in the leftlung, the contralateral lung, chest wall, and the skeletalsystem. The vertebrae were the most involved skeletal site ofmetastasis (figure 3); skull, tibia, and femur marrowmetastasis were also observed by GFP fluorescence.29

Paris and co-workers30 described a model of spontaneouslung metastases of H-460M-GFP in nude mice. Nude micewere subcutaneously inoculated with H-460M-GFP cells and,after 2 weeks, 75% of animals had fluorescent spontaneouslung micrometastases. From the third week onwards, all theanimals had metastatic disease, mainly in the lungs. By 7 weeks, the number of lung metastases had increased toabout 400 in all mice.

Chishima and co-workers injected human ANIP-GFPlung-cancer cells into the tail vein of nude mice, which weresubsequently sacrificed after either 4 or 8 weeks.28 In both

groups, numerous micrometastaticcolonies in the lung were detected byGFP expression in fresh tissue. 8 weeksafter injection, most of the colonieswere no more developed than those inmice sacrificed at 4 weeks.Several smallcolonies with 10 or fewer cells weredetected at the surface of the lung inboth groups. Dormant tumour cellswere seen in the lung along with thosethat were actively colonising,28 which isan extremely interesting finding sincemicrometastatic dormancy is one ofthe most important processes forunderstanding tumour progression.After 8 weeks, metastases were presentin the brain, the submandibular gland,lung, pancreas, bilateral adrenal glands,

peritoneum, and pulumonary hilum lymph nodes.28

Lewis lung carcinoma cells have been used widely formany important studies, but many groups use subcutaneoustransplant or orthotopic cell-suspension injection models,which do not allow the tumour to fulfill its metastaticpotential. However, in studies involving a new highlymetastatic model of Lewis lung carcinoma, which has beendeveloped by use of SOI and GFP-expressing tumours,100% of animals had metastases on the ipsilateraldiafragmatic surface, contralateral diafragmatic surface,contralateral lung, and in mediastinal lymph nodes. Heartmetastases were seen in 40% of the animals, and brainmetastases were present in 30%.31

Hastings and co-workers investigated the effects oftreatment with a neutralising antibody to parathyroidrelated hormone (PTHrP) on the growth of BEN cells—ahuman lung squamous-cell carcinoma line that expresses

PTHrP and its receptor. BEN-GFP cellswere implanted into athymic miceproducing orthotopic lung tumours.The mice received either subcutaneousPTHrP antibodies or irrelevantimmunoglobulin. After 30 days, six of10 mice receiving antibodies to PTHrPhad GFP-expressing lung tumours,whereas only one of the 10 controlmice had developed cancer.32

Prostate cancerGFP-labelled human prostate cancercells (PC-3-GFP) were implanted bySOI in the dorsolateral lobe of theprostate in nude mice.33 GFP fluores-cence indicated the presence of skeletalmetastasis in several places includingthe skull, rib, pelvis, femur, and tibia.The tumours also metastasised to thelung, pleural membrane, kidney, liver,adrenal gland, brain, and spinal cord.33

Severe combined immunodeficient(SCID) mice orthotopically inoculated


Figure 3. External image of bone metastasis of AC3488-GFP. External images of tumours in theskeletal system including (a) the skull, (b) scapula, and (c) spine in a dorsal view of live intact nudemouse.36

Figure 4. External imaging of single U-87 glioma cells in microscopic tumour colonies on the brainimaged through a skin flap in a nude mouse. Single cancer cells and small colonies of cancer cellscan be seen. Bar=80 �m.54



with LNCaP-GFP cells and metastases to distant organs werechronologically examined.34 LNCaP-GFP cells were detectedin the lung under fluorescence microscopy as early as 4 weeks after inoculation. There was no difference in growthrates and androgen-responsiveness between LNCaP-GFPand LNCaP cells observed in vitro. LNCaP-GFP cellsinoculated in SCID mice produced prostate-specific antigen.

MelanomaMalignant melanoma cell lines that highly express GFP(B16-GFP) were used to visualise skeletal and multiorganmetastases in two types of mouse. C57BL/6 mice receivedan intravenous injection of B16-GFP cells and nude micewere given an intradermal injection of human melanomaLOX.35 The animals were sacrificed after 3 weeks andextensive B16F0 metastases to the bone and bone marrowwere visualised by use of GFP expression. For both celllines, metastases were present in several organs includingbrain, lung, pleural membrane, liver, kidney, adrenalgland, lymph nodes (figure 7), skeleton, muscle, and skin.Moore and colleagues50 also implanted B16-GFP into thehind leg of mice; all animals developed para-aortic lymphnode metastases.

Pancreatic cancerGFP-expressing metastases of two human pancreaticcancer cell lines, MIA-PaCa2-GFP and BxPC-3-GFP, wereexamined in orthotopic models. The cells weretransplanted by SOI to the pancreas of nude mice and GFPexpression was used to visualise the chronology ofpancreatic tumour growth and metastasis. BxPC-3-GFPtumours developed rapidly and spread to the spleen andretroperitoneum as early as 6 weeks after inoculation.However, distant metastases were rare. By contrast, MIA-PaCa-2-GFP tumours grew slowly in the pancreas, butrapidly metastasised to distant sites including the liver andportal lymph nodes. Regional metastases with thesetumours were rare. These studies showed that pancreaticcancers have highly specific and individual “seed–soil”interactions, which governs the chronology and sites ofmetastases.37

Brain cancerDaoy medulloblastoma-GFP cells were stereotacticallyinjected into the frontal cortex of nude mice by MacDonaldand colleagues.46 The primary tumour margins were clearlyvisible in brain tissue sections by GFP fluorescence; localinvasion and micrometastases, including single cells, couldalso be visualised. However, the micrometastases could notbe detected by routine hematoxylin and eosin staining andimmunohistochemistry.

Breast cancerSchmidt and colleagues implanted GFP-expressing humanbreast tumour cells (MDA-MB-435-GFP) orthotopically inthe mammary fat pad of SCID mice.39 Metastatic cellsmigrated to the lungs of the animals very early afterorthotopic implantation and the numbers of migrating cellscontinued to increase until death, about 8 weeks later. By


Figure 5. External imaging of angiogenesis of orthotopic BxPC-3 humanpancreatic tumour through a skin flap. The blood vessels can be seen bycontrast with the GFP-expressing tumour.54 (a) Macroimaging of BxPC-3-GFP human pancreatic tumour angiogenesis through the abdominal wallskin flap window. (b) Microscopic imaging of BxPC-3-GFP angiogenesisthrough the abdominal-wall skin-flap window (magnification X20). (c)Microscopic imaging of Bx PC-3-GFP angiogenesis through theabdominal-wall skin-flap window (magnification X40).



use of flow cytometry, one GFP-fluorescent tumour cell per200 000 host cells could be detected in dissociated tissues.Circulating tumour cells in blood could also be visualised byGFP fairly soon after implantation.39

Human MDA-MB-435-GFP breast cancer cells, whichare tumorigenic but slow-growing in vivo, were transfectedwith extracellular matrix metalloproteinase inducer cDNA.The transfected cells were injected orthotopically intomammary tissue of female NCr nude mice41 and tumourgrowth was compared with empty plasmid-transfectedcancer cells. GFP fluorescence confirmed that the cDNA-transfected cells were considerably more tumorigenic andinvasive than the plasmid-infected cells.

GFP-expressing cells have also been used to visualisetumour-cell motility in orthotopic models of breast tumoursin rats.40 A metastatic rat breast-cancer cell line was

established that constitutively express-ed GFP. These cells were injected intothe mammary fat pad of femaleFischer 344 rats and, after skin hadbeen removed from the animals,primary and metastatic tumours werevisualised by GFP fluorescence. By useof a laser scanning confocal micro-scope, intravital imaging of theprimary tumour in situ enabled themotility patterns of the tumour cells tobe visualised.40

Cortactin is a substrate of the Src-related tyrosine kinases, which is known to enhance tumourprogression in MDA-MB-231 breastcancer cells. When nude mice were injected through their cardiacventricles with MDA-MB-231 cellsoverexpressing GFP-cortactin, theanimals showed bone osteolysis at arate that was about 85% higher than inanimals injected with cells expressingthe vector alone.43

Intravital imaging of GFP-expressing cellsIntravital videomicroscopy can beused to visualise sequential steps inmetastasis by use of CHO-K1 cellsthat stably express GFP. In mouseliver, the stages from the initial arrestof cells in the microvasculature up tothe growth and angiogenesis ofmetastases have been recorded.5

Individual non-dividing cells, as wellas micro and macrometastases werevisualised and quantified; additionalcellular details such as pseudopodialprojections were also detected. Micrometastases were found topreferentially grow and survive nearthe liver surface. A small population

of single cells persisted over the 11-day observation periodand the investigators believe these cells may representdormant tumour cells.5

Rat tongue carcinoma cell lines expressing GFP havebeen used to investigate the formation of micrometastasis.The cells were injected into the portal vein and then trackedby use of intravital video microscopy.51 The two cell types—LM-GFP metastatic and E2-GFP non-metastatic tonguecarcinoma cells—immediately got stuck in the sinusoidalvessels near terminal portal venules. The E2-GFP cellsdisappeared from the liver sinusoid within 3 days, whereas asubstantial number of LM-GFP cells remained in the liver—possibly because these cells formed stable attachments to thesinusoidal wall. Upon examination of the process with aconfocal laser scanning microscope, only LM-GFP cells wereshown to grow in the liver.


Figure 6. Whole-body fluorescence imaging of lymphoma dissemination. A mouse lymphoma arisingin a Em-myc transgenic mouse was transduced with GFP and introduced to syngenic mice.71

(a) Lymphoma disseminated to lymph nodes, bone, and brain can be seen. (b) The rendered figureeliminates all non-GFP-expressing areas of the mouse with the z-direction indicating intensity.



GFP techniques have also shown that CT-26 (mousecolon carcinoma) cells inhibit the development of livermetastasis in BALB/c mice that receive intraportal injectionsof GFP-expressing tumour cells.52 Intravital microscopy ofthe livers of these mice showed that growth of primarytumours promoted dormancy of single tumour cells for upto 7 days. Immunohistological staining for Ki-67 confirmedthat these solitary cells were indeed dormant. By contrast, inthe absence of a primary tumour, GFP-expressing tumourcells quickly developed into micrometastases. Thus, primaryCT-26 tumour implants seem to inhibit tumour metastasisby promotion of a state of single-cell dormancy.

The ras oncogene promotes growth of micrometastasesinto macroscopic metastases. Two types of cell withconstituitively active ras, NIH 3T3 and T24 H-ras-transformed (PAP2) fibroblasts, both of which wereexpressing GFP, were injected into mouse liver. Subsequentexamination by GFP intravital imaging established that onlythe micrometastases formed by ras-transformed cells wenton to produce macroscopic metastases; most NIH 3T3micrometastases just disappeared. Furthermore, PAP2 met-astases had a significantly higher proportion of proliferatingcells than apoptotic cells, whereas NIH 3T3 metastases hadlow proliferation and a high proportion of apoptotic cells.53

Whole-body imaging of tumour growth andmetastasisExternal whole-body imaging of mice with primary andmetastatic tumours that are genetically labelled with thefluorescent proteins GFP and RFP is a simple but powerfultool for investigating tumour development. The technologyis based on the bright intrinsic fluorscence of GFP and RFP,which is partly caused by the high quantum yield of thesefluorophores.19,23 For tumour cells to be visualised with thistechnique, they must be transduced with GFP or RFP genes,such that they become brightly fluorescent. This can beaccomplished by in vitro26,36 and in vivo64 selection of suchfluorescent tumour cells. To produce metastasis in mice, thegenetically-fluorescent tumours should be transplantedorthotopically.26,27,29,31–34,37,38,44,48 Once the GFP-expressingtumours have developed and metastases have formed,individual tumour cells can be detected in the live mouse byuse of whole-body imaging with fairly simple equipment. Afluorescence light box with fibre-optic lighting at about490 nm and appropriate filters, placed on top of the lightbox, can be used to image large tumours and can be viewedwith the naked eye.36 Alternatively, the light box can belinked to a camera with an appropriate filter to enableimages to be displayed on a monitor and digitally stored.36 Inorder to visualise smaller tumours and metastases, theanimal can be put on a fluorescence dissecting microscopewhich incorporates a light source and filters for excitation atabout 490 nm. Fluorescence emission can be observedthrough a 520 nm long-pass filter.36 The animals can beirradiated at 490 nm for long periods without harming themor bleaching the GFP or RFP fluorescence. Images can beprocessed with standard software and the imagingprocedures can be repeated as often as necessary withoutharming the animal. Therefore, with these techniques real-

time tracking of tumour growth and metastasis is feasible.Reversible skin-flaps can also be introduced onto differentparts of the animal in order to look at single tumour cells orsmall colonies on internal organs.54 The skin-flaps arerendered reversible by simple suturing. External imaging canthen be done through the relatively transparent body walls ofthe mouse, which include the skull, by use of a fluorescencedissecting microscope. Blood vessels growing on tumourscan also be observed using skin flaps because they contrastwith the fluorescence of the tumours.54 Examples of findingsfrom studies of these techniques are described below.

Whole-body imaging of metastatic lesionsTransplanted mice with metastatic lesions of GFP-expressing tumours in the colon, brain (figure 2), liver,lymph nodes, and bone (figure 3) have been used to produceimages of metastasis. These images are used for real-timequantitative measurement of primary and metastatictumour growth for each of these organs. GFP-expressingcells emit a bright fluorescence signal compared withbackground fluorescence from other tissue. The signal is sostrong and selective that external images of GFP-expressingtumours and their metastases can be obtained in freely-moving animals.29

In another study, GFP-expressing human pancreatictumour cell lines were introduced as tissue fragments intothe pancreases of nude mice by SOI. As the tumours weregrowing, the investigators used whole-body optical imagingtechniques to track, in real-time, the growth of the primarytumour and the formation of metastatic lesions thatdeveloped in the spleen, bowel, portal lymph nodes,omentum, and liver. The images were used forquantification of tumour growth in each of these organs.38

Using a different approach, human ovarian tumour cells(SKOV3.ip1) were made to express GFP by infection with areplication-deficient adenoviral (Ad) vector encoding GFP.45

The infected cells showed high GFP fluorescence, and whenimplanted into mice, intraperitoneal tumours as small as 0.2mm in diameter could be detected externally.

Peyruchaud and colleagues established a GFP-expressingbone-metastasis subclone of MDA-MB-231 (B02/GFP.2) byrepeated in vivo passages in bone, by use of the heart injectionmodel. When injected into the tail vein of mice, the selectedcells grew preferentially in bone. Whole-body fluorescenceimaging of the live mice showed that bone metastases could bedetected about 1 week before radiologically distinctiveosteolytic lesions developed. Furthermore, when the tumour-bearing mice were treated with a bisphosphonate, progressionof established osteolytic lesions, and the expansion of breastcancer cells within bone, were inhibited.42

Opening a reversible skin-flap in the light path greatlyreduces attenuation of the fluorescent signal, therebyincreasing the sensitivity of tumour detection by manytimes. This procedure also greatly increases the depth atwhich tumour cells can be observed. Single GFP-expressingtumour cells can thus be seen on numerous internal organs.GFP glioma cells seeded on the brain can be visualisedthrough a scalp skin-flap (figure 4). GFP lung tumourmicro-foci, which represent a few malignant cells, can be




viewed through a skin-flap over the chest wall andcontralateral micrometastases can also be examined by useof the corresponding skin-flap. Peritoneal wall skin flaps canbe used to look at GFP-pancreatic tumours and theirangiogenic micro-vessels (figure 5). For the liver, a skin-flapallowed imaging of physiologically relevant micro-metastases that had originated in an orthotopically-implanted GFP tumour. Single tumour cells on the liver,which had been introduced through an intraportal injection,were also detectable.54

Ilyin and co-workers visualised glioma cells in rats byinserting a fibre-optic endoscope through a pre-implantedguide cannula. Tumour monitoring was coupled to confocalmicroscopy so that visualisation of the fluorescent signalsfrom the C-6 glioma-GFP cells that had been preimplantedin the brain, was very sensitive.47

Whole-body and intravital imaging ofangiogenesis and intravascular tumour cellsTumour angiogenesis can also be visualised by use of GFPtechniques. The footpads of mice are quite transparent withfew resident blood vessels and are therefore ideal forquantitative imaging of tumour angiogenesis in intactanimals. Vessels can be seen because of their strikingcontrast to the GFP fluorescence of the tumour tissue.48

Yang and colleauges injected GFP-expressing Lewis lungcarcinoma cells subcutaneously into the footpad of nudemice and, using whole-body imaging, they found thatcapillary density increased linearly over 10 days. Similarly,when GFP-expressing MDA-MB-435 cells were ortho-topically transplanted to the mouse fat pad, whole-bodyoptical imaging showed that blood vessel density increasedlinearly over about 20 weeks. As described above, reversibleskin flaps can be used to visualise angiogenesis on GFP-expressing tumours transplanted onto internal organs (figure 5), in addition to examining the tumouritself.54 These angiogenesis mouse models can be used for real-time in vivo evaluation of agents inhibiting or promoting tumour angiogenesis.48,54

Imaging GFP tumour cells in blood vesselsFollowing injection of tumour cells stably expressing GFP into the tail vein of mice, it was possible to visualise singletumour cells in blood vessels.26 With intravital microscopy,Naumov and colleagues visualised GFP tumour cellscolonising various organs after extravasation.5 Huang,55 Li,56

and their respective co-workers visualised GFP tumour-cell–vessel interaction by use of skin window chambers inrodents and observed angiogenic effects very early in tumourcolony formation. In an orthotopic mammary-pad injectionmodel, Wyckoff and colleagues65 also visualisedtumour–vessel interaction by looking at the fluorescence ofthe tumour cells. Al-Mehdi1 and co-workers saw what theyclaimed to be intravascular tumour colony formation bytracking GFP expression in a lung perfusion study. Inaddition, Moore and colleauges have also visualised vessels ina GFP-expressing rodent cell line.49

Wong and colleagues57 showed that death of trans-formed, metastatic, rat embryo cells, which were expressing

GFP, occurred via apoptosis in the lungs 24–48 h after in-jection into the circulation. The researchers established thatBCL-2 overexpression was causing inhibition of apoptosis inculture, and this mechanism also conferred resistance in vivofor 24–48 h after injection. This inhibition of apoptosis ledto a greater number of macroscopic metastases.

Large detectable metastases did not form after athymicmice were given an intravenous injection of chromosome 6-transduced tumour cells expressing GFP.58 However,fluorescence microscopy revealed micrometastases (singlecells or clusters of fewer than 10 cells) in the lungs,suggesting that these cells managed to lodge themselves inthe lungs, but failed to proliferate. Cells isolated from mouselungs up to 60 days after the injection were able to grow inculture and formed tumours when injected into skin;therefore, the cells were still viable, but dormant. This resultimplies that the gene(s) on chromosome 6 interferespecifically with growth regulatory response in the lung, butnot in the skin.

Chang and co-workers59 used CD31 and CD105 toidentify endothelial cells and GFP-labelling of tumour cellsand showed that colon carcinoma xenografts had mosaicvessels with focal regions where no CD31/CD105immunoreactivity was detected and tumour cells appearedto contact the vessel lumen.

GFP labeling of VEGF in vivoBrown and colleagues3 showed that multiphoton laser-scanning microscopy could provide high resolution three-dimensional images of angiogenesis gene expression andthat this techniques could be used to investigate deeperregions of GFP-expressing tumours in dorsal skin-foldchambers. To monitor the activity of the vascularendothelial growth factor (VEGF) promoter, Fukumura andcolleagues made transgenic mice that express GFP undercontrol of the VEGF promoter.60,61 Multiphoton laserscanning microscopy showed that the tumour was able toinduce activity of the VEGF promoter; GFP-positive stromalcells were seen at least 200 �m into the tumour with thistechnique.

Fukumura’s group also developed a novel in vivomicroscopy technique to simultaneously measure VEGFpromoter activity, pO2, and pH.62 To monitor VEGFexpression in vivo, the investigators engineered humanglioma cells that expressed GFP under the control of theVEGF promoter. The cells were subsequently implanted intocranial windows in SCID mice and VEGF promoter activitywas assessed by GFP imaging. Findings from these studiessuggest that VEGF transcription in brain tumours isregulated by both tissue pO2 and pH via distinct pathways.

Use of GFP to monitor potential expression ofheat-shock proteins in vivoMouse melanoma cells were stably transfected with aplasmid containing the GFP gene linked to a heat shockprotein gene promoter, which reacts to physiological stress.63

At physiological temperature, GFP expression in experi-mental mouse tumours was undetectable. However, thetumour cells started to fluoresce brightly in response to an




increase in temperature (heat shock) both in vitro and invivo. Thus, GFP could be a useful marker for studies ofmammalian heat shock proteins.

Clinically applicable models of GFP tumourimagingSeveral studies have focused on delivering the GFP geneselectively to tumours in order to provide a marker for thedevelopment of new metastases. Hasegawa and colleaguesadministered the GFP gene to nude mice with humanstomach tumours growing intraperitoneally, in order tovisualise future regional and distant metastases.64 GFPretroviral supernatants were injected intraperitoneally fromday 4 to day 10 following implantation of the cancer cells. Alaparotomy was done every other week so that tumourgrowth and metastasis formation could be visualised by GFP expression. No normal tissues were transduced by the GFP retrovirus. 2 weeks after retroviral GFP delivery,GFP-expressing tumour cells were observed in the gonadalfat, greater omentum, and intestine, indicating that thetumours were efficiently transduced by the GFP gene andcould be detected by its expression. Second and thirdlaparotomies were done at weeks 4 and 6, respectively. GFP-expressing tumour cells were found spreading to lymphnodes in the mesentery. The fourth laparotomy, at 8 weeks,showed widespread tumour growth including metastasis tothe liver. Thus, reporter-gene transduction of the primarytumour enabled detection of its subsequent metastasis. Thisreporter-gene therapy model could be applied to primarytumours before resection or other treatment and thusprovide an early detection system for metastasis andrecurrence.64

To determine whether the carcino-embryonic antigenpromoter could direct the GFP gene, human MKN45-GFPstomach-cancer cells were injected into the peritoneal cavityof BALB/c nude mice.65 A CEA-EGFP plasmid was thenintroduced in the peritoneal cavity using liposomes. GFP-fluorescent tumour nodules were subsequently detected byfluorescence stereomicroscopy.

In another approach, GFP was conjugated to transferrinin order to target disseminated tumours in vivo.66 WhenGFP-gene conjugates were systemically administeredthrough the tail vein to nude mice that had beensubcutaneously inoculated with tumour, GFP expressionwas only detected only in the tumour.

Varda-Bloom and colleagues67 developed a tissue-specific gene therapy to the angiogenic blood vessels oftumour metastasis using an adeno-based vector containingthe murine preproendothelin-1 (PPE-1) promoter drivingGFP. High and specific activity of PPE-1was achieved bysystemic administration of the adenoviral vector to micebearing Lewis lung carcinoma tumours. This effect wasdetected by GFP expression in the new vasculature ofprimary tumours and lung metastasis. The highest area ofexpression was in the angiogenic endothelial cells of themetastasis.

Recently, GFP on an HSV-1/EBV vector has beenadministered to tumour-bearing animals. However,persistant GFP expression was not achieved in this study.68

Fluorescent reporter gene for human T cellsNormal, human, peripheral-blood T lymphocytes weretransduced with a retroviral HSV-TK-GFP (vGFPTKfus)and nucleus-restricted green fluorescence was observed.Sorting allowed for selection of GFP-expressing T lymphocytes. The ability to target GFP-expressing T lymphocytes to tumours could have many clinical uses.69

Bone-marrow protection by transfer of drug-resistance genes coupled to GFP A retroviral vector expressing human O6-methylguanine-DNA methyltransferase (MGMT) and GFP was developedfor stem-cell protection in a murine transplant model. Micetransplanted with transduced cells showed significantresistance to the myelosuppressive effects of temozolomide,a DNA-methylating drug, and O6-benzylguanine, a drug thatinhibits MGMT. Following drug treatment, increases inGFP-positive peripheral blood cells were seen. Secondarytransplant experiments proved that selection had occurred atthe stem-cell level. Such an approach could be used clinicallyin the future to protect bone marrow against chemo-therapy.70

p53 tumour-suppressor functions were visualised in vivoby GFP imaging (figure 6). By use of the antiapoptotic geneBCL-2 or a dominant-negative caspase 9 (C9DN), GFPwhole-body imaging established that disruption of theapoptosis pathway downstream of p53 leads to Eµ-myc-GFPlymphoma expansion that phenocopies p53 loss in Eµ-myctransgenic mice. This finding shows that GFP whole-bodyimaging can be used to identify individual genes that affecttumour growth. Such information could be used to identifygenes predicting aggressive tumour growth.71

A senescence programme controlled by p53and p16INK4a affects chemotherapyThe GFP primary lymphomas derived from Eµ-myctransgenic mice respond to chemotherapy by undergoingapoptosis and engaging a premature senescence programmecontrolled by p53 and p16INK4a. Therefore, tumours respondpoorly to cyclophosphamide therapy if their p53 orINK4a/ARF genes are disrupted; this can be seen with GFPwhole-body imaging. It has also been shown that micebearing tumours capable of drug-induced senescence have amuch better prognosis after chemotherapy than thoseharboring tumours with senescence defects. These findingsindicated that premature senescence can contribute totreatment outcome in vivo and provide new insights into themolecular genetics of drug resistance, which can be appliedclinically.72

ConclusionsTumour cells stably expressing GFP in vivo are a powerfulnew tool for cancer research. Stability of expression hasbeen studied by Naumov and colleagues5 who noted that allthe CHO-K1-GFP cells used in their study were stablyfluorescent (measured by flow cytometry) even after 24 daysof growing in medium where they were deprived of selectivepressure. This finding implies that GFP can be stablyexpressed in cells in vivo. This feature has proved true for all




cells studied so far and is exemplified by the generatation ofextensive GFP-expressing metastases.

The use of GFP-expressing tumourcells in transplanted mice, fresh tissue,or live animals5,26 has provided newinsights into the real-time growth andmetastatic behaviour of cancer. Severalindependent studies, which include anextensive comparison between metas-tases of GFP-transduced rat mammarycarcinoma cell and the parental cellline,26,28,40 have shown that GFP transduction and expressiondoes not affect metastatic behaviour.40

GFP can be transfected into any cell type of interest andused as a cytoplasmic marker to show the general outlines ofcells in vivo and fine morphological details such as longslender pseudopodial projections.5

The development of tumour cells that stably express GFPat high levels has enabled investigation of tumour andmetastastic growth in a completely non-invasive manner, byuse of whole-body imaging.36 For the first time, tumourgrowth and metastatic studies, including drug evaluations,can be done and quantified in real-time in individualanimals—the potential of this technology is very high.

A further advantage of GFP-expressing cells is theincreased contrast between brightly fluorescent tumourtissue and blood vessels within it. The ability to visualise andquantify blood-vessel development in metastases in vivo willgreatly facilitate studies of angiogenesis and the testing ofeffects of antiangiogenic agents on metastatic deve-lopment.5,48

The GFP approach has several important advantagesover other optical approaches to imaging (table 1). Incomparison with the luciferase reporter, GFP has a muchstronger signal and therefore can be used to imageunrestrained animals. The fluorescence intensity of GFP isvery strong since the quantum yield is about 0·9.15–19 Theprotein sequence of GFP has also been “humanised”, whichenables it to be highly expressed in mammalian cells.20 Inaddition, GFP fluorescence is fairly unaffected by theexternal environment since the chromaphore is protectedby the three-dimensional structure of the protein.13 Theexcitation wavelength is quite long at 490 nm,16–19 whichdoes not quench the fluorescence and therefore long-termmeasurements can be made. In vivo, GFP fluorescence is mainly limited by light scattering which, as noted above, can be overcome by skin flaps54 and endoscopes47

such that single cells can be imaged externally. Longerwavelength fluorescence proteins, such as RFP, can also be

used to reduce scatter. The luciferase reporter technique,however, requires that animals are anesthetised andrestrained so that sufficient photons to construct an imagecan be collected. Furthermore, this process must be carried out in a virtually light-free environment andanimals must be injected with the luciferin substrate, whichhas to reach every tumour cell in order to be useful. This limitation precludes studies that would be perturbedby anesthesia, restraint, or substrate injection and alsomakes high-throughput screening unfeasible. Expression offirefly luciferase (Luc) can be used to visualise tumour-cellgrowth and regression in response to various therapies inmice. However, detection of Luc-labelled cells in vivo waslimited to 103 human tumour cells.8,9 The high intensitysignal produced by GFP, however, allows unrestrainedanimals to be imaged without any perturbation orsubstrate—irradiation with non-damaging blue light is theonly step needed. Images can be captured with fairly simpleapparatus and there is no need for darkness.

Near-infrared probes activated by the action ofproteases6,7 can also be used for optical imaging of tumours.This approach requires substrate injection and the tumourmust contain a specific protease that cleaves the substrate.Tumours on normal tissues, eg, the liver, that contain theseproteases can not be visualised because background signalsare too high, so that there is not sufficient distinctionbetween tumour and normal tissue to obtain useful externalimages.73

For clinical application of GFP developments, futurestudies may make use of the approach of Hasegawa andcolleagues,64 who used retroviral GFP vectors that wereselectively targeted to tumours for the purpose of identifyingfuture metastasis. GFP labelling of tumour-infiltratinglymphocytes69 and bone marrow transduced with GFP-linked genes that confer chemoresistance,70 also haveintriguing clinical potential.

Conflict of interestRMH is President of AntiCancer Inc—the company that commerciallyexploits two mouse models (the MetaMouse and the AngioMouse) forvisualisation of tumour spread and angiogenesis by use of GFP-expression. Many of the studies discussed in this review make use ofthis technology.

References1 Al-Mehdi AB, Tozawa K, Fisher AB, et al. Intravascular origin of

metastasis from the proliferation of endothelium-attached tumorcells: a new model for metastasis. Nat Med 2000; 6: 100–02.

2 Lin WC, Pretlow TP, Pretlow TG 2nd, Culp LA. Bacterial lacZ geneas a highly sensitive marker to detect micrometastasis formationduring tumor progression. Cancer Res 1990; 50: 2808–17.


Search strategy and selection criteriaReferences for this review were identified from the results ofsearches of PubMed and selected from the references ofrelevant articles. The search terms used were “greenfluorescent protein”, “cancer”, and “in vivo”. Searches werelimited to papers published in English between 1997 and 2002that discussed imaging investigations.

Table 1. Comparison of opitical imaging methods for tumour growth andmetastasis

Method Minimum Minimum Need for Need for Method of Multi-cells cells substrate? anesthesia? visualisation colourimageable imageable imagingin vitro in vivo

GFP 1* 1*† No† No§ Direct Imaging‡ Yes†

Luciferase 300|| 3000¶ Yes¶ Yes¶ Photon counting(pseudo-colour) No

*Reference 26; †reference 54; ‡reference 36; §reference 74; ||reference 75; ¶reference 76.



3 Brown EB, Campbell RB, Tsuzuki Y, et al. In vivo measurement ofgene expression, angiogenesis and physicological function in tumorsusing multiphoton laser scanning microscopy. Nat Med 2001; 7:864–68.

4 Ciancio SJ, Coburn M, Hornsby PJ. Cutaneous window for in vivoobservations of organs and angiogenesis. J Surg Res 2000; 92:228–32.

5 Naumov GN, Wilson SM, MacDonald IC, et al. Cellular expressionof green fluorescent protein, coupled with high-resolution in vivovideomicroscopy, to monitor steps in tumor metastasis. J Cell Sci1999; 112: 1835–42.

6 Weissleder R, Tung CH, Mahmood U, Bogdanov A Jr. In vivoimaging of tumors with protease-activated near-infrared fluorescentprobes. Nat Biotechnol 1999; 17: 375–78.

7 Bremer C, Tung CH, Weissleder R. In vivo molecular targetassessment of matrix metalloproteinase inhibition. Nat Med 2001; 7:743–48.

8 Sweeney TJ, Mailander V, Tucker AA, et al. Visualizing the kineticsof tumor-cell clearance in living animals. Proc Natl Acad Sci USA1999; 96: 12044–49.

9 Contag CH, Jenkins D, Contag PR, Negrin RS. Use of reportergenes for optical measurements of neoplastic disease in vivo.Neoplasia 2000; 2: 41–52.

10 Prasher DC, Eckenrode VK, Ward WW, et al. Primary structure ofthe Aequorea victoria green-fluorescent protein. Gene 1992; 111:229–33.

11 Chalfie M, Tu Y, Euskirchen G, et al. Green fluorescent protein as amarker for gene expression. Science 1994; 263: 802–05.

12 Cheng L, Fu J, Tsukamoto A, Hawley RG. Use of green fluorescentprotein variants to monitor gene transfer and expression inmammalian cells. Nat Biotechnol 1996; 14: 606–09.

13 Cody CW, Prasher DC, Welstler VM, et al. Chemical structure ofthe hexapeptide chromophore of the Aequorea green fluorescentprotein. Biochemistry 1993; 32: 1212–18.

14 Yang F, Moss LG, Phillips GN Jr. The molecular structure of greenfluorescent protein. Nat Biotechnol 1996; 14: 1246–51.

15 Morin J, Hastings J. Energy transfer in a bioluminescent system. J Cell Physiol 1971; 77: 313–18.

16 Cormack B, Valdivia R, Falkow S. FACS-optimized mutants of thegreen fluorescent protein (GFP). Gene 1996; 173: 33–38.

17 Crameri A, Whitehorn EA, Tate E, Stemmer WPC. Improved greenfluorescent protein by molecular evolution using DNA shuffling.Nat Biotechnol 1996; 14: 315–19.

18 Delagrave S, Hawtin RE, Silva CM, et al. Red-shifted excitationmutants of the green fluorescent protein. Biotechnology 1995; 13:151–54.

19 Heim R, Cubitt AB, Tsien RY. Improved green fluorescence. Nature1995; 373: 663–64.

20 Zolotukhin S, Potter M, Hauswirth WW, et al. A ‘humanized’ greenfluorescent protein cDNA adapted for high-level expression inmammalian cells. J Virol 1996; 70: 4646–54.

21 Gross LA, Baird GS, Hoffman RC, et al. The structure of thechromophore within DsRed, a red fluorescent protein from coral.Proc Natl Acad Sci USA 2000; 97: 11990–95.

22 Fradkov AF, Chen Y, Ding L, et al. Novel fluorescent protein fromDiscosoma coral and its mutants possesses a unique far-redfluorescence. FEBS Lett 2000; 479: 127–30.

23 Matz MV, Fradkov AF, Labas YA, et al. Fluorescent proteins fromnonbioluminescent Anthozoa species. Nat Biotechnol 1999; 17:969–73.

24 Chishima T, Yang M, Miyagi Y, et al. Governing step of metastasisvisualized in vitro. Proc Natl Acad Sci USA 1997; 94: 11573–76.

25 Chishima T, Miyagi Y, Li L, et al. The use of histoculture and greenfluorescent protein to visualize tumor cell host interaction. In VitroCell Dev Biol 1997; 33: 745–47.

26 Chishima T, Miyagi Y, Wang X, et al. Cancer invasion andmicrometastasis visualized in live tissue by green fluorescent proteinexpression. Cancer Res 1997; 57: 2042–47.

27 Chishima T, Miyagi Y, Wang X, et al. Metastatic patterns of lungcancer visualized live and in process by green fluorescent proteinexpression. Clin Exp Metastasis 1997; 15: 547–52.

28 Chishima T, Miyagi Y, Wang X, et al. Visualization of the metastaticprocess by green fluorescent protein expression. Anticancer Res1997; 17: 2377–84.

29 Yang M, Hasegawa S, Jiang P, et al. Widespread skeletal metastaticpotential of human lung cancer revealed by green fluorescentprotein expression. Cancer Res 1998; 58: 4217–21.

30 Paris S, Chauzy C, Martin-Vandelet N, et al. A model ofspontaneous lung metastases visualised in fresh host tissue by greenfluorescent protein expression. Clin Exp Metastasis 1999; 17:817–22.

31 Rashidi B, Yang M, Jiang P, et al. A highly metastatic Lewis lungcarcinoma orthotopic green fluorescent protein model. Clin ExpMetastasis 2000; 18: 57–60.

32 Hastings RH, Burton DW, Quintana RA, et al. Parathyroidhormone-related protein regulates the growth of orthotopic humanlung tumors in athymic mice. Cancer 2001; 92: 1402–10.

33 Yang M, Jiang P, Sun FX, et al. A fluorescent orthotopic bonemetastasis model of human prostate cancer. Cancer Res 1999; 59:781–86.

34 Maeda H, Segawa T, Kamoto T, et al. Rapid detection of candidatemetastatic foci in the orthotopic inoculation model of androgen-sensitive prostate cancer cells introduced with green fluorescentprotein. Prostate 2000; 45: 335–40.

35 Yang M, Jiang P, An Z, et al. Genetically fluorescent melanomabone and organ metastasis models. Clin Cancer Res 1999; 5:3549–59.

36 Yang M, Baranov E, Jiang P, et al. Whole-body optical imaging ofgreen fluorescent protein-expressing tumors and metastases. ProcNatl Acad Sci USA 2000; 97: 1206–11.

37 Bouvet M, Yang M, Nardin S, et al. Chronologically-specificmetastatic targeting of human pancreatic tumors in orthotopicmodels. Clin Exp Metastasis 2000; 18: 213–18.

38 Bouvet M, Wang J-W, Nardin SR, et al. Real-time optical imaging of primary tumor growth and multiple metastatic events ina pancreatic cancer orthotopic model. Cancer Res 2002; 62:1534–40.

39 Schmidt CM, Settle SL, Keene JL, et al. Characterization ofspontaneous metastasis in an aggressive breast carcinoma modelusing flow cytometry. Clin Exp Metastasis 1999; 17: 537–44.

40 Farina KL, Wyckoff JB, Rivera J, et al. Cell motility of tumor cellsvisualized in living intact primary tumors using green fluorescentprotein. Cancer Res 1998; 58: 2528–32.

41 Zucker S, Hymowitz M, Rollo EE, et al. Tumorigenic potential ofextracellular matrix metalloproteinase inducer. Am J Pathol 2001;158: 1921–28.

42 Peyruchaud O, Winding B, Pecheur I, et al. Early detection of bonemetastases in a murine model using fluorescent human breastcancer cells: application to the use of the bisphosphonate zoledronicacid in the treatment of osteolytic lesions. J Bone Miner Res 2001;16: 2027–34.

43 Li Y, Tondravi M, Liu J, et al. Cortactin potentiates bone metastasisof breast cancer cells. Cancer Res 2001; 61: 6906–11.

44 Yang M, Chishima T, Wang X, et al. Multi-organ metastaticcapability of Chinese ovary cells revealed by green fluorescentprotein (GFP) expression. Clin Exp Metastasis 1999; 17: 417–22.

45 Chaudhuri TR, Mountz JM, Rogers BE, et al. Light-based imagingof green fluorescent protein-positive ovarian cancer xenograftsduring therapy. Gynecol Oncol 2001; 82: 581–89.

46 MacDonald TJ, Tabrizi P, Shimada H, et al. Detection of braintumor invasion and micrometastasis in vivo by expression ofenhanced green fluorescent protein. Neurosurgery 1998; 43:1437–42.

47 Ilyin SE, Flynn MC, Plata-Salaman CR. Fiber-optic monitoringcoupled with confocal microscopy for imaging gene expression invitro and in vivo. J Neurosci Methods 2001; 108: 91–96.

48 Yang M, Baranov E, Li XM, et al. Whole-body and intravital opticalimaging of angiogenesis in orthotopically implanted tumors. ProcNatl Acad Sci USA 2001; 98: 2616–21.

49 Moore A, Marecos E, Simonova M, et al. Novel gliosarcoma cell lineexpressing green fluorescent protein: a model for quantitativeassessment of angiogenesis. Microvasc Res 1998; 56: 145–53.

50 Moore A, Sergeyev N, Bredow S, Weissleder R. A model system toquantitate tumor burden in locoregional lymph nodes duringcancer spread. Invasion Metastasis 1998–99; 18: 192–97.

51 Ito S, Nakanishi H, Ikehara Y, et al. Real-time observation ofmicrometastasis formation in the living mouse liver using a greenfluorescent protein gene-tagged rat tongue carcinoma cell line. Int JCancer 2001; 93: 212–17.

52 Guba M, Cernaianu G, Koehi G, et al. A primary tumor promotesdormancy of solitary tumor cells before inhibiting angiogenesis.Cancer Res 2001; 61: 5575–79.

53 Varghese HJ, Davidson MT, MacDonald IC, et al. Activated rasregulates the proliferation/apoptosis balance and early survival of




developing micrometastases. Cancer Res 2002; 62: 887–91.54 Yang M, Baranov E, Wang J-W, et al. Direct external imaging of

nascent cancer, tumor progression, angiogenesis, and metastasis oninternal organs in the fluorescent orthotopic model. Proc Natl AcadSci USA 2002; 99: 3824–29.

55 Huang Q, Shan S, Braun RD, et al. Noninvasive visualization oftumors in rodent dorsal skin window chambers. Nat Biotechnol1999; 17: 1033–35.

56 Li CY, Shan S, Huang Q, et al. Initial stages of tumor cell-inducedangiogenesis: evaluation via skin window chambers in rodentmodels. J Natl Cancer Inst 2000; 92: 143–47.

57 Wong CW, Lee A, Shientag L, et al. Apoptosis: an early event inmetastatic inefficiency. Cancer Res 2001; 61: 333–38.

58 Goldberg SF, Harms JF, Quon K, Welch DR. Metastasis-suppressedC8161 melanoma cells arrest in lung but fail to proliferate. Clin ExpMetastasis 1999; 17: 601–07.

59 Chang YS, di Tomaso E, McDonald DM, et al. Mosaic blood vesselsin tumors: frequency of cancer cells in contact with flowing blood.Proc Natl Acad Sci USA 2000; 97: 14608–613.

60 Fukumura D, Xavier R, Sugiura T, et al. Tumor induction of VEGFpromoter activity in stromal cells. Cell 1998; 94: 715–25.

61 Fukumura D, Yuan F, Monsky WL, et al. Effect of hostmicroenvironment on the microcirculation of human colonadenocacinoma. Am J Pathol 1997; 151: 679–88.

62 Fukumura D, Xu L, Chen Y, et al. Hypoxia and acidosisindependently up-regulate vascular endothelial growth factortranscription in brain tumors in vivo. Cancer Res 2001; 61: 6020–24.

63 Wysocka A, Krawczyk Z. Green fluorescent protein as a marker formonitoring activity of stress-inducible hsp70 rat gene promoter.Mol Cell Biochem 2000; 215: 153–56.

64 Hasegawa S, Yang M, Chishima T, et al. In vivo tumor delivery ofthe green fluorescent protein gene to report future occurrence ofmetastasis. Cancer Gene Ther 2000; 7: 1336–40.

65 Kaneko K, Yano M, Yamano T, et al. Detection of peritoneal

micrometastases of gastric carcinoma with green fluorescent proteinand carcinoembryonic antigen promoter. Cancer Res 2001; 61:5570–74.

66 Sato Y, Yamauchi N, Takahashi M, et al. In vivo gene delivery totumor cells by transferrin-streptavidin-DNA conjugate. FASEB J2000; 14: 2108–18.

67 Varda-Bloom N, Shaish A, Gonen A, et al. Tissue-specific genetherapy directed to tumor angiogenesis. Gene Ther 2001; 8: 819–27.

68 Qi J, Link CJ, Wang S. Direct observation of GFP gene expressiontransduced with HSV-1/EBV amplicon vector in unfixed tumortissue. Biotechniques 2000; 28: 206–08.

69 Paquin A, Jaalouk DE, Galipeau J. Retrovector encoding a greenfluorescent protein-herpes simplex virus thymidine kinase fusionprotein serves as a versatile suicide/reporter for cell and genetherapy applications. Hum Gene Ther 2001; 12: 13–23.

70 Sawai N, Zhou S, Vanin EF, et al. Protection and in vivo selection ofhematopoietic stem cells using temozolomide, O6-benzylguanine,and an alkyltransferase-expressing retroviral vector. Mol Ther 2001;3: 78–87.

71 Schmitt CA, Fridman JS, Yang M, et al. Dissecting p53 tumorsuppressor functions in vivo. Cancer Cell 2002; 1: 289–98.

72 Schmitt CA, Yang M, Fridman JS, et al. Senescence programcontrolled by p53 and p16INK4a contributes to the outcome of cancertherapy. Cell 2002; 109: 335–46.

73 Hirsch FR, Prindiville SA, Miller YE, et al. Fluorescence versuswhite-light bronchoscopy for detection of preneoplastic lesions: arandomized study. J Natl Cancer Inst 2001; 93: 1385–91.

74 Dusich JM, Oei YA, Purchio T, Jenkins DE. In vivo detection oflung colonization and metastasis using luciferase-expressing humanA549 lung cells. Proc Am Assoc Cancer Res 2002; 43: 1059 (Abstr5244).

75 Vooijs M, Jonkers J, Lyons S, Berns A. Noninvasive imaging ofspontaneous retinoblastoma pathway-dependent tumors in mice.Cancer Res 2002; 62: 1862–67.


pii: s1470-2045(02)00848-3

Documents