Xiaoming Yang, MD, PhD Digital Optical of Hong Liu, PhD Dechun Li, MD, PhD Green Fluorescent Proteins for Xianzheng Zhou, MD, PhD William C. Jung, MS Tracking Vascular Gene Abby E. Deans, BS Yan Cui, PhD Expression: Feasibility Study Linzhao Cheng, PhD in Rabbit and Human Cell

Index terms: 1 Animals Models Carotid arteries, 942.1299, 942.91, 942.99 Experimental studies Genes and genetics PURPOSE: To investigate the feasibility of using a sensitive digital optical imaging Molecular analysis technique to detect green fluorescent protein (GFP) expressed in rabbit vasculature Proteins and human arterial smooth muscle cells. 2001; 219:171–175 MATERIALS AND METHODS: A GFP plasmid was transfected into human arterial smooth muscle cells to obtain a GFP–smooth muscle cell solution. This solution was Abbreviations: GFP ϭ green fluorescent protein imaged in cell phantoms by using a prototype digital optical imaging system. For in SMC ϭ smooth muscle cell vivo validation, a GFP-lentivirus vector was transfected during surgery into the carotid arteries of two rabbits, and GFP-targeted vessels were harvested for digital optical imaging ex vivo. 1 From the Departments of Radiology (X.Y., H.L.), Anesthesiology and Criti- RESULTS: Optical imaging of cell phantoms resulted in a spatial resolution of 25 cal Care (D.L.), and Oncol- ogy (X.Z., Y.C., L.C.) and the Center ␮m/pixel. Fluorescent signals were detected as diffusely distributed bright spots. At for Medical Optics and Electronic Im- ex vivo optical imaging of arterial tissues, the average fluorescent signal was signif- aging (H.L., W.C.J., A.E.D.), the Johns icantly higher (P Ͻ .05) in GFP-targeted tissues (mean Ϯ SD, 9,357.3 absolute units Hopkins University School of Medi- cine, Outpatient Center, Rm 4243, of density Ϯ 1,001.3) than in control tissues (5,633.7 absolute units of density Ϯ 601 N Caroline St, Baltimore, MD 985.2). Both fluorescence microscopic and immunohistochemical findings con- 21287-0845. Received March 27, firmed these differences between GFP-targeted and control vessels. 2000; revision requested May 13; re- vision received July 27; accepted Au- CONCLUSION: The digital optical imaging system was sensitive to GFPs and may gust 29. Supported in part by the potentially provide an in vivo imaging tool to monitor and track vascular gene Cardiovascular and Interventional Radi- ology Research and Education Foun- transfer and expression in experimental investigations. dation (CIRREF) and a National Insti- tutes of Health grant CA 70209. Ad- dress correspondence to X.Y. (e-mail: [email protected]). © RSNA, 2001 Atherosclerotic cardiovascular disease remains the leading cause of mortality in the United States (1). Gene therapy is a rapidly expanding field with great potential for the treatment of cardiovascular diseases. Many genes have been shown to be useful for preventing acute thrombosis, blocking postoperative restenosis, and stimulating growth of new blood vessels (angiogenesis) (2–5). However, precise monitoring of gene delivery into and ade- quate follow-up of gene expression in targeted atherosclerotic plaques are two challenging tasks. To date, most investigations about imaging of gene therapy have been focused Author contributions: primarily on noncardiovascular systems, and, to our knowledge, no in vivo imaging Guarantors of integrity of entire study, modalities are currently available for vascular gene therapy (6). X.Y., H.L.; study concepts and design, The recent emergence of a viable marker, green fluorescent protein (GFP), has opened X.Y., H.L.; literature research, X.Y., H.L.; experimental studies, all authors; the door for the convenient use of intact living cells and organisms as experimental data acquisition and analysis, all au- systems in fields ranging from cell biology to biomedicine (7,8). GFP has been widely used thors; statistical analysis, X.Y., H.L.; as a sensitive marker that can be detected with fluorescence microscopy or flow cytometry manuscript preparation and editing, (9,10). The native protein of GFP is fluorescent in living cells, which allows in situ X.Y., H.L.; manuscript definition of in- tellectual content, revision/review, detection in the living animal. Moreover, the fluorescent signal emitted from GFP can be and final version approval, all authors. detected with optical imaging. This property of GFP has been widely used in in vitro investigations with living cells and has been experimentally tested in vivo to track the

171 distribution of gene transfer in the eye Animal Health, Fort Dodge, Iowa), ace- (11), tumors (12), and the airway (13). promazine maleate (1.1 mg/kg; Fermenta These results demonstrate the potential Animal Health, Kansas City, Mo), and at- application of optical imaging of GFP to ropine sulfate (0.05 mg/kg; American Re- monitor gene transfer and track gene ex- gent Laboratories, Shirley, NY). An ear pression. To our knowledge, however, vein was cannulated to permit mainte- the ability to optically image GFP in the nance of anesthesia. Pentobarbital so- cardiovascular system has not yet been dium (20 mg/kg; Abbott Laboratories, explored. North Chicago, Ill) was later adminis- The objective of this study was to in- tered intravenously to bring the animal vestigate the feasibility of using a highly to a surgical plane of anesthesia. Animals sensitive digital optical imaging tech- were intubated and mechanically venti- nique to detect GFPs expressed from the lated by using a respirator (model 55- Figure 1. Surgery-based gene delivery into vasculature. 0798; Harvard Apparatus, South Natick, the rabbit carotid artery (solid arrow). After Mass). The animals were also adminis- isolation of the artery by means of two Senti- MATERIALS AND METHODS tered heparin (100 IU per kilogram body nel loops (arrowheads), the GFP–lentivirus vec- weight). Anesthesia was monitored for tor solution (open arrow) is injected into the In Vitro Experiments the duration of the experiment by using carotid artery. We used an N-terminal–enhanced GFP regular tests of the eyelid reflex and mild plasmid vector (pEGFP-N1; Clontech, paw compression. Surgery-based GFP vector delivery.—Us- Palo Alto, Calif). Human pulmonary ar- optical imaging and another for immedi- ing an arteriotomy approach (14), we terial smooth muscle cells (SMCs) (Clo- ate frozen sections and subsequent fluo- first exposed an approximately 1.5-cm- netics, San Diego, Calif) were cultured in rescence microscopic examination, as long portion of the bilateral carotid arter- SMC growth medium (SmGM-2; Clonet- well as immunohistochemical confirma- ies. The exposed arterial portion on the ics, Walkersville, Md). They were then tion. seeded into a six-well plate at a concen- right side was isolated with sutures and 5 then harvested (prior to administration tration of 10 cells per well and were in- Digital Optical Imaging of any GFP vector in the left side) to serve cubated at 37°C (with 5% CO2) for 24 hours. Then, GFP plasmid vector (1 ␮g) as control. The exposed arterial portion Digital optical imaging system.—To im- was transfected for 30 minutes into SMCs on the left side was isolated by tempo- age SMC-GFPs in vitro and GFP-targeted with a transfection reagent (lipofect- rarily tightening the target vessel proxi- vessel tissues ex vivo, a digital optical AMINE PLUS Reagent; Life Technologies, mally and distally with two Sentinel imaging prototype was developed in our Rockville, Md). Subsequently, the non- loops (Sherwood Medical, St Louis, Mo) laboratory. The system consists of an ex- transfected plasmid was removed, and (Fig 1). By inserting a 24-gauge catheter ternal source, a fiberoptic light the plate was further incubated in the (Quik-Cath; Baxter, Marion, NC) into the guide, wavelength selective optics, a sam- 37°C incubator for an additional 2 days isolated segment of the left carotid artery, ple holder, a custom relay lens, and a to allow sufficient GFP expression. After we drew off the blood from the vessel highly sensitive charge-coupled–device this, the cells were harvested by means of lumen and thereafter directly injected a detector (Fig 2) (15). trypsinization and washing and were an- GFP-containing lentiviral vector solution The detector module uses a mechani- alyzed by using flow cytometry (Decton (developed and produced in our labora- cal shutter. An optical hood is used to Dickinson, Immunocytometry Systems, tory in January 2000) to fill the isolated shield the detector from ambient light. San Jose, Calif) to quantify the transfec- vessel portion for 1 hour. A silk tie was The charge-coupled–device arrays con- tion rates of the cell phantom. placed in the adjacent tissue to identify sist of 1,024 ϫ 1,024 pixels, and each the vessel segment to be harvested. After pixel measures 0.024 ϫ 0.024 mm with a the transfection, we drew off the GFP 0.999 fill factor. The spatial resolution of In Vivo and ex Vivo Experiments vector solution, removed the catheter, the overall digital optical imaging system Animals.—We used the bilateral ca- closed the puncture point with cyanoac- is 20 line pairs per millimeter (16). The rotid arteries from two New Zealand rylate glue (Superglue; Elmer’s, Colum- camera is operated at a temperature of white rabbits (Robinson Services, Clem- bus, Ohio), loosened the Sentinel loops, Ϫ25°C to reduce thermal electron noise. mons, NC), each approximately 5 kg in closed the arteriotomy incision with su- The low temperature is achieved with a weight. All animals were treated accord- tures, and kept the rabbit alive for 4 days. compact thermoelectric cooler. This pro- ing to the “Principles of Laboratory Ani- After the surgery-based gene delivery, we totype provides 14-bit digitization. mal Care” of the National Society for administered a postoperative analgesic During image acquisition, the cell Medical Research and the Guide for the (0.01 mg/kg buprenorphine, Buprenex; phantoms or tissue samples were placed Care and Use of Laboratory Animals (Na- Reckitt & Coleman Pharmaceuticals, on top of the holder. The GFPs were ex- tional Institutes for Health publication Richmond, Va) with an intramuscular in- cited by means of an external light source no. 80–23, revised 1985). The Animal jection every 12 hours. At day 5 after GFP with a selected wavelength of 475 nm Ϯ Care and Use Committee of the Johns vector transfection, the rabbit was anes- 20 (SD). The GFP fluorescence (which Hopkins University approved the experi- thetized, and the GFP-targeted segment peaked at 513 nm) was then guided to mental protocol. of the left carotid artery was harvested. the photosensitive surface of the detector Animals were sedated with an intra- We then euthanized the animals with a through a series of optical filters. The muscular injection of a mixture of ket- dose of 100 mg/kg pentobarbital sodium. wavelength selective optics and the digi- amine hydrochloride (22 mg per kilo- Each of the harvested arterial specimens tal acquisition and calibration procedure gram of body weight [mg/kg]; Fort Dodge was cut into two equal pieces: one for were carefully designed to minimize sys-

172 ⅐ Radiology ⅐ April 2001 Yang et al were fixed in 100% ethanol for 20 min- utes. After air drying, the slides were treated with 3% hydrogen peroxide in phosphate-buffered saline (pH 7.4) for 20 minutes to quench endogenous peroxi- dase activity. After blocking of nonspe- cific sites with 10% goat serum in phos- phate-buffered saline for 60 minutes, the slides were washed with phosphate- buffered saline three times for 5 min- utes and incubated at 4°C overnight with a 1:250 dilution of specific mono- clonal antibody for GFP (Roche, India- napolis, Ind). After unbound primary Figure 2. Schematic shows setup of the digital optical imaging sys- antibodies were washed off with phos- tem. Light enters the system from the light source through a fiberop- phate-buffered saline, the slides were tic guide. The excitation light is reflected up to the antireflective- incubated with biotinylated anti-mouse coated sample plate. Light that is absorbed by the sample and re- antibody (1:500 dilution) for 1 hour. Spe- emitted as fluorescence is transmitted back through the dichroic and cific binding was detected by using a emission filters to the digital detector. 1:100 dilution of an avidin-biotin–horse- radish peroxidase complex (Vector, Bur- lingame, Calif) applied for 1 hour and a substrate solution of hydrogen peroxi- dase and diaminobenzidine, according to the manufacturer’s instructions. The slides were then counterstained with hematoxylin, dehydrated with graded concentrations of alcohol and xylene, mounted with coverslips, and examined with a microscope (Vanox AHBS3; Olym- pus). Negative controls were carried out with normal mouse immunoglobulin G, according to the primary antibodies used Figure 3. Graphs show results of flow cytometry analysis of GFP expression after transfecting for the staining. SMCs. Left: Nontransfected (control) SMCs. Right: SMCs transfected with GFP plasmid expressing 8% GFPs. At 2 days after transfection, SMCs were harvested and analyzed with flow cytometry. Relative green fluorescent signals were detected with the FL1 channel of the fluorescent-activated Statistical Analysis cell sorter. GFP-positive cells were identified by gating the marked region (M1) on the basis of the negative control fluorescence (Յ1% GFP-positive cells). To compare the difference between flu- orescent signals emitted from the GFP- targeted tissues and the control tissues, tem noise. The electronic output of the tissue on the bottom of an optical win- we measured and recorded the fluores- detector was then routed to a computer dow (Labglass 9395; Corning). The opti- cent signal values obtained from each of for analysis, storage, and display. cal window was then placed in the holder the wet tissue samples nine times in a Optical imaging.—For in vitro optical of the digital optical imaging system to randomly selected region of interest, imaging of cell phantoms, we used three detect fluorescent signals emitted from with 280 pixels in each region of interest. 96-well special optics plates (COSTAR GFPs in the tissue. The imaging parame- Data are expressed as means plus or mi- 3614; Corning, Acton, Mass). In each of ters were the same as for the in vitro nus the SD. A Student t test was applied these three plates, we deposited 0.3 mL of experiments. for the comparison. A probability level of SMC-GFP solution into three wells and less than .05 was considered to indicate a SMC-only solution into adjacent wells to significant difference. Fluorescence Microscopy serve as controls. Each well of the plate and Immunohistochemistry was 7 mm in diameter. The plate was RESULTS then placed in the holder of the digital Fresh carotid arterial tissue was imme- optical imaging system. We separately diately frozen in liquid nitrogen, and 14- The quantitative measurement achieved imaged each well to detect the fluores- ␮m-thick sections were obtained for slides. with flow cytometry confirmed that we cent signals emitted from GFPs in the For fluorescence microscopy, the slides obtained a solution of 8% GFPs expressed SMCs, by using multiple exposure times were air dried for 30 minutes and then in SMCs (Fig 3), which was used for in (0.1–10.0 seconds) and a pixel-binning observed by using a fluorescence micro- vitro digital optical imaging with the cell technique. scope (Olympus IMT-2; Olympus, Tokyo, phantoms. In vitro experiments resulted In the in vivo experiments for imaging Japan) linked to a charge-coupled–device in optical images of the cell phantoms the tissues ex vivo, we first opened the camera (model 3 CCD; Toshiba, Tokyo, with a spatial resolution of 25 ␮m/pixel. target vessel portion longitudinally to Japan). The fluorescent signals, emitted from the obtain an approximately 0.4 ϫ 0.6 cm For immunohistochemical detection solution of 8% GFPs expressed in SMCs, sample of wet tissue. We placed the wet of GFP in the tissue sections, the slides were detected as diffusely distributed

Volume 219 ⅐ Number 1 Digital Optical Imaging to Track Vascular Gene Expression ⅐ 173 bright spots, whereas no such signals were found in the control non-GFP (SMCs-only) solution (Fig 4). This phe- nomenon was repeatedly visualized in different wells and on various plates. In ex vivo optical imaging of the arterial tissues, we could distinguish higher fluo- rescent signal enhancement on the GFP- targeted tissues than on the control tis- sues (Fig 5). The average fluorescent signal value was significantly higher (P Ͻ .05) in GFP-targeted tissues (9,357.3 ab- solute units of density Ϯ 1,001.3) than in control tissues (5,633.7 absolute units of density Ϯ 985.2). At fluorescence microscopic examina- tion of GFP-targeted vessels, we detected a stronger fluorescent signal in the entire Figure 4. Digital optical images of a cell phantom, which holds human arterial SMCs. A, Con- intimal layer, including endothelial cells trol phantom contains nontransfected cells. B, Phantom contains SMCs transfected with 8% GFP and the internal elastic laminae, whereas plasmid. The fluorescent signals emitted from the solution of 8% GFPs expressed in SMCs were in the control tissues the weak fluores- detected as diffusely distributed bright spots (B), whereas no such signals were found in the cent signal was detected along the endo- control non-GFP solution (A). Scale bar ϭ 1 mm. thelium due to autofluorescence (Fig 5). Immunohistochemical findings corre- lated well with the findings from fluores- cence microscopy, confirming that GFPs, manifested as a brown-colored precipi- tate, were localized in both the endothe- lium and the internal elastic laminae of GFP-targeted vessels, whereas no such precipitate was found in the control ves- sels (Fig 5).

DISCUSSION

Medical optical imaging detects light sig- nal transmitting/reflecting or emitting from media, such as tissues, to help determine the interior structure and chemical content. The prominent advan- tages of the optical imaging technique include (a) capability for bedside or oper- ating room monitoring, (b) lack of expo- Figure 5. Rabbit carotid arterial tissues without (A, B, C) and with (D, E, F) GFP transfection. sure to ionizing radiation, (c) relatively A, D, Digital optical images. B, E, Corresponding fluorescence microscopy images. (Original magni- low cost, and (d) ability to directly detect fication, ϫ200.) C, F, Corresponding photomicrographs of immunohistochemically analyzed signals emitted from fluorescent materi- tissue sections. (Original magnification, ϫ200.) In the digital optical images (A, D), fluorescent als such as GFPs. Different optical tech- signal enhancement is seen in the GFP-targeted tissue (D), and no such enhancement is present in the non-GFP-targeted tissue (A). Scale bar ϭ 1 mm. In B and E, GFPs were detected as a strong niques that do not use GFP as an “imag- green-colored fluorescence located in the endothelium and internal elastic laminae (arrows in E), ing marker” have been reported (17–19) which was confirmed at immunohistochemical analysis (C, F), where GFP manifested as a for diagnostic imaging or for imaging of brown-colored precipitate (arrowheads in F). B, C, No such manifestations were found in non– gene expression. GFP-transfected vessels. Autofluorescence of the arterial structures, such as elastic laminae and To date, to our knowledge, in vivo mo- endothelium, is seen in the normal carotid artery (B). C, F, At immunohistochemical analysis, the lecular imaging of vascular gene therapy vessel tissues are stained with anti-GFP antibody with diaminobenzidine as substrate. has not been explored because of the lack of appropriate imaging modalities to ad- dress this promising new avenue of ther- sensitive to GFPs. With a spatial resolu- could verify the GFP signals emitted from apeutic management. We hypothesized tion of 25 ␮m/pixel, this system can de- the intima of GFP-targeted vessels, al- that digital optical imaging could be used tect GFPs as diffusely distributed bright though there is overlap of fluorescent sig- to detect vascular GFPs, which were ex- spots. This digital optical imaging tech- nals from autofluorescent structures such pressed within the GFP-targeted vessel nique may provide a useful molecular as the endothelium and elastic laminae. wall. The in vitro investigation of the imaging tool for in vitro imaging of GFPs These in vitro and in vivo results present study shows that the digital op- expressed in living cells. On the ex vivo should establish the groundwork for the tical imaging system we used was highly optical images of wet arterial tissues, we future innovative development of intra-

174 ⅐ Radiology ⅐ April 2001 Yang et al vascular and external digital optical im- teins (transfected into the target tissues) the green fluorescent protein. Nucleic Ac- aging techniques to monitor cardiovas- from intrinsic autofluorescence (emitted ids Res 1996; 24:4592–4593. 8. Misteli T, Spector D. Applications of the cular gene therapy. An intravascular digital by the tissue structures themselves). green fluorescent protein in cell biology optical imager (probe) could be used to de- In conclusion, the digital optical imag- and biotechnology. Nat Biotechnol 1997; tect GFP gene expression in deep-seated ing system developed in our laboratory is 15:961–964. vessels such as the aorta and the iliac sensitive to GFPs, which may potentially 9. Chalfie M, Tu Y, Euskirchen G, Ward W, arteries, and an external digital optical provide an in vivo imaging tool to mon- Prasher D. Green fluorescent protein as a marker for gene expression. Science 1994; imager could be used to detect GFP gene itor and track vascular gene transfer and 263:802–805. expression in surface vessels such as the expression. 10. Zolotukhin S, Potter M, Hauswirth W, carotid and brachial arteries. These ulti- Practical applications: In vitro fluo- Guy J, Muzyczka N. A “humanized” green mate goals will require a “real-time” GFP rescence microscopic imaging of vascular fluorescent protein cDNA adapted for GFPs may potentially be extended to in high-level expression in mammalian cells. vector in which GFPs could be expressed J Virol 1996; 70:4646–4654. in advance, before delivery into the tar- vivo digital optical imaging of vascular 11. Moritz O, Tam B, Knox B, Papermaster D. get vessel wall. Thus, once the real-time GFPs. The current study may lead to ad- Fluorescent photoreceptors of transgenic GFP vector has been transfected into the ditional innovative advancements, in- Xenopus laevis imaged in vivo by two target vessel, we could use these digital cluding (a) an intravascular digital opti- microscopy techniques. Invest Ophthal- mol Vis Sci 1999; 40:3276–3280. optical imagers to immediately acquire cal imaging technique that could detect 12. Yang M, Baranov E, Jiang P, et al. Whole- baseline information about the success, either real-time or long-term GFP gene body optical imaging of green fluorescent level, and location of gene transfection. expression in deep-seated vessels and protein-expressing tumors and metasta- Another important issue is to find a long- (b) an external digital optical imaging ses. Proc Natl Acad SciUSA2000; 97: technique that could detect GFP gene ex- 1206–1211. term GFP vector in which GFPs could be 13. Flotte T, Beck S, Chesnut K, Potter M, expressed for a substantial amount of pression in surface vessels. Poirier A, Zolotukhin S. A fluorescence time after transfection into the target ves- video-endoscopy technique for detection sel wall. Thus, as the long-term GFP is Acknowledgments: The authors thank of gene transfer and expression. Gene Xiaojian Yuan, MD, PhD, for kindly providing expressed, we could use an intravascular Ther 1998; 5:166–173. human arterial SMCs and Mary McAllister, MS, 14. Yang X, Bradley D, Bolster J, Kraitchman or external digital optical imager to track for her editorial assistance. DL, Atalar E. Intravascular MR-monitored vascular gene therapy. balloon angioplasty: an in vivo feasibility For digital optical imaging of fluores- References study. J Vasc Interv Radiol 1998; 9:953– cent signals from wet tissues, there is 1. American Heart Association. 1998 heart 959. 15. Liu H, Karellas A, Harris L, D’Orsi C. Op- overlap of signals due to autofluores- and statistical update. Dallas, Tex: American Heart Association, 1998. tical properties of fiber tapers and their cence emitted from vessel structures such 2. Nabel E. Gene therapy for cardiovascular impact on the performance of a fiberop- as endothelium and elastic laminae. Fur- disease. Circulation 1995; 91:541–548. tically coupled CCD x-ray imaging sys- ther efforts must be focused on improve- 3. Isner J. Manipulating angiogenesis against tem. Proc SPIE 1993; 1894:136–147. 16. Liu H, Jiang H, Fajardo LL, Karellas A, ment of the sensitivity of the digital op- vascular disease. Hosp Pract (Off Ed) 1999; 34:69–74. Chen WR. Lens distortion in optically tical imaging technique for in vivo 4. Stephan D, Nabel E. Gene and other bio- coupled digital x-ray imaging. Med Phys detection of GFPs. One strategy is to de- logical therapies for vascular diseases. 2000; 27:906–912. sign wavelength-selective optics, which Fundam Clin Pharmacol 1997; 11:97– 17. Mahmood U, Tung C, Bogdanov A, would filter out autofluoresence from tis- 110. Weissleder R. Near-infrared optical imag- 5. Thomas J, Kuo M, Chawla M, et al. Vas- ing of protease activity for tumor detec- sues. Another strategy is to find other cular gene therapy. RadioGraphics 1998; tion. Radiology 1999; 213:866–870. fluorescent materials, such as blue, cyan, 18:1373–1394. 18. Huang D, Swanson E, Lin C, et al. Optical and yellow fluorescent proteins. These 6. Weissleder R. : explor- . Science 1991; 245: variants would emit fluorescence with a ing the next frontier (editorial). Radiol- 1178–1181. 19. Contag CH, Spilman SD, Contag PR, et al. wavelength different from those of cur- ogy 1999; 212:609–614. 7. Yang T, Cheng L, Kain S. Optimized Visualizing gene expression in living mam- rent GFPs, which may benefit in the dif- codon usage and chromophore muta- mals using a bioluminescent reporter. ferentiation of extrinsic fluorescent pro- tions provide enhanced sensitivity with Photochem Photobiol 1997; 66:523–531.

Volume 219 ⅐ Number 1 Digital Optical Imaging to Track Vascular Gene Expression ⅐ 175