EFFICACY OF ANTITUMORAL PHOTODYNAMIC THERAPY WITH HYPERICIN:RELATIONSHIP BETWEEN BIODISTRIBUTION AND PHOTODYNAMIC EFFECTSIN THE RIF-1 MOUSE TUMOR MODELBin CHEN

1, Yan XU2, Tania ROSKAMS

3, Els DELAEY1, Patrizia AGOSTINIS

2, Jackie R. VANDENHEEDE2 and Peter DE WITTE

1*1Laboratorium voor Farmaceutische Biologie en Fytofarmacologie, Faculteit Farmaceutische Wetenschappen, K.U. Leuven, Belgium2Afdeling Biochemie, Faculteit Geneeskunde, K.U. Leuven, Belgium3Afdeling Histochemie en Cytochemie, Faculteit Geneeskunde, K.U. Leuven, Belgium

We investigated the hypericin-mediated PDT effects onthe tumor and normal skin and in correlation with its biodis-tribution. These studies were carried out on C3H mice bear-ing RIF-1 tumors. The hypericin distribution and PDT effectswere recorded at different intervals (0.5–24 hr) after intra-venous injection of a 5-mg/kg dose of hypericin. After admin-istration, rapid biphasic exponential decay was observed inthe plasma drug concentration. It was found that hypericinwas preferentially bound to the plasma lipoproteins. Thetumor drug levels increased rapidly over the first few hoursand reached a maximum around 6 hr after injection. Incontrast, PDT efficacy was maximal when irradiation wasperformed at 0.5 hr after hypericin administration, which ledto 100% cure. The PDT efficacy decreased rapidly as theadministration-irradiation interval was prolonged. No tumorcure was obtained at the 6-hr interval, even though it was atthis time that the tumor drug level peaked. Fluorescencemicroscopic studies showed that hypericin was mainly con-fined within the tumor vasculature at 0.5 hr after injection,whereupon it rapidly diffused to the surrounding tumor tis-sue. At 6 hr, a strong hypericin fluorescence was observed inthe tumor tissue with only faint fluorescence within the vas-culature, whereas at 24 hr the fluorescence in the tumor alsodecreased and became more diffused, and no fluorescencecould be seen in the tumor vasculature. Like the tumorresponse, skin reactions were also found to be much moredramatic at short administration-irradiation intervals. Hy-pericin distribution and PDT response studies revealed aclose correlation between the plasma drug level and the PDTeffects, which suggests that vascular damage is the primaryeffect of hypericin-mediated PDT in this tumor model.© 2001 Wiley-Liss, Inc.

Key words: photodynamic therapy; hypericin; antitumor effect; vas-cular effect; biodistribution

Photodynamic therapy (PDT) of tumors involvesthesystemic ortopical administration of a photosensitizer to patients followed byirradiation of the tumor mass with light of an appropriate wave-length. In the presence of oxygen, photoactivated sensitizers gen-eratehighly reactiveoxygen species (ROS). Theoxidativedamageto various cellular organelles and functions induced by ROS leadsto direct cytotoxicity on tumor cells.1 Indirect effects, however,have also been shown to contribute to tumor destruction.2 Theseeffects originate from the PDT-induced severe vascular damageand result in the collapse of the entire microcirculation system.3Consequently, persistent tumor ischemia contributes to the overalltumor ablation. Moreover, an indirect antitumoral-specific immunereaction may also be induced by PDT as aresult of the activationof inflammatory cells. It is likely that thisphenomenon plays arolein long-term tumor control.4

Thesuccessful application of PDT in theclinic and theapparentlimitations associated with the only clinically approved photosen-sitizer, Photofrin, are stimulating the search for more efficaciousand safer phototoxic compounds. Hypericin is one of the photo-sensitizers being investigated as a potential PDT tool due to itshigh singlet oxygen yield.5 This compound is presently in Phase Iclinical trials for the treatment of cutaneous lesions.6 Unlike por-phyrin derivatives, hypericin is a polycyclic aromatic naphthodi-anthrone (chemical structure: Fig. 1). It is isolated from plants of

the Hypericum genus. In organic solvents, the compound (as amonobasic salt) forms red solutions (lmax: 548 nm and 591 nm inethanol (e591: 45.000 M21 cm21)) that exhibit red fluorescence(lmax: 594 nm and 642 nm in ethanol).7

Recently, we and other investigators have been testing the invitro photocytotoxicity of hypericin.8–12 From the published data,it can be concluded that hypericin invariably exerts a potentphototoxicity on cells of different histological origin. Because thecompound is lipophilic, it has been postulated that membranes arethe principal targets of hypericin, where it likely exerts its cyto-toxic effect through the locally photogenerated ROS.9,13 The for-mation of hypericin radicals or a pH drop, however, may also playan important role in the photocytotoxicity of the compound.14,15

Studying the biochemical background of apoptosis induced byphoto-activated hypericin, we demonstrated that the induction ofthe apoptotic program in HeLa cells most likely originates fromthe release of mitochondrial cytochrome c followed by a pro-caspase-3 activation.16 Furthermore, we have recently shown thatthe JNK1 and p38 MAPK pathways play an important role in thecellular resistance to PDT-induced apoptosis by hypericin.17

Thus far, there have been only a few brief reports on the in vivophotodependent antitumor activity of hypericin.18–20 The relation-ship between the tissue distribution and PDT effects on tumor aswell as surrounding tissue has not been clearly established. Fur-thermore, the different tumor models used in those studies make itdifficult to compare the results obtained with other photosensitiz-ers. The encouraging in vitro findings together with the relativepaucity of in vivo evidence prompted us to determine in moredetail thepotential of hypericin as aPDT tool. In thepresent study,we describe the efficacy of the photodynamic action of hypericinagainst a RIF-1 mouse tumor growing subcutaneously, a well-characterized tumor model used frequently in the assessment ofantitumor effects of photosensitizers. In addition, the relationshipbetween the distribution kinetics of hypericin in tumor and thesurrounding healthy tissues and the time-dependent phototoxicityon tumor and skin was also examined.

Abbreviations: CIS, carcinoma in situ; DMSO, dimethyl sulfoxide; FBS,fetal bovine serum; HDL, high-density lipoprotein; LDL, low-densitylipoprotein; mTHPC, meta-tetra (hydroxyphenyl) chlorin; PBS, phosphate-buffered saline; PDT, photodynamic therapy; RIF, radiation-induced fibro-sarcoma; VLDL , very low density-lipoprotein.

*Correspondence to: Laboratorium voor Farmaceutische Biologie enFytofarmacologie, Faculteit Farmaceutische Wetenschappen, K.U. Leu-ven, Van Evenstraat 4, B-3000 Leuven, Belgium. Fax: 132-16-323460.E-mail: peter.dewitte@farm.kuleuven.ac.be

Received 19 December 2000; Revised 14 February 2001; Accepted 2March 2001

Published online 20 April 2001

Int. J. Cancer: 93, 275–282 (2001)© 2001 Wiley-Liss, Inc.

Publication of the International Union Against Cancer

MATERIAL AND METHODS

Synthesis and preparation of hypericinHypericin was synthesized and purified with silica and Seph-

adex LH-20 column chromatography as described elsewhere.10 Astock solution of hypericin (5 mg/ml) was prepared in polyethyl-ene glycol 400 and stored at220°C in the dark. Immediatelybefore injection into the animals, it was diluted 5 times in PBS(Gibco BRL, Paisley, Scotland) to obtain a final concentration of1 mg/ml.

Animals and tumor systemFemale C3H/Km mice (10–14 weeks old, weight range 21–25

g) obtained from the KU Leuven Animal Facility were usedthroughout this study. The RIF-1 murine tumor line (kindly pro-vided by Dr. F. Stewart, The Netherlands Cancer Institute) wasmaintained and passaged according to establishedin vivo/in vitroprocedures.21 Approximately 13 105 cells were inoculated sub-cutaneously on the depilated lower dorsum of the mice. Tumorgrowth was documented regularly by caliper measurements inthree orthogonal dimensions. Tumors were used for experimenta-tion 7 to 12 days after inoculation when reaching a surface diam-eter of 4 to 6 mm and a thickness of 2 to 3 mm. All aspects of theanimal experiment and husbandry were carried out in compliancewith national and European regulations and were approved by theAnimal Care and Use Committee of the KU Leuven.

Analysis of tissue distributionTumor-bearing animals were sacrificed at 0.5 hr, 1 hr, 2 hr, 4 hr,

6 hr, 8 hr and 24 hr after i.v. injection of a 5 mg/kg dose ofhypericin. Plasma, tumor, surrounding skin and muscle were har-vested, weighed and frozen at220°C until the determination of thehypericin content. The same tissue samples were also taken fromcontrol mice receiving no drug. The extraction and quantificationof tissue hypericin concentrations were performed as follows. Theminced tissue and plasma samples were homogenized with tetra-hydrofuran in a tissue homogenizer. The resulting tissue suspen-sions were vortexed and sonicated for 30 min, and the supernatantswere then evaporated to dryness under reduced pressure. Theresidues were dissolved in 0.3 ml DMSO before fluorescencedetermination by means of a microplate fluorescence reader(FL600, Bio-Tek Instruments, Winooski, VT) with excitation andemission filters of 530 nm (bandwidth 25 nm) and 645 nm (band-width 40 nm), respectively, as described elsewhere.20 The back-ground fluorescence from control samples was subtracted. Thehypericin concentrations were calculated from the calibrationcurve and expressed asmg/g of wet tissue ormg/ml of plasma.

Plasma protein distributionThe plasma distribution property of hypericin was studied with

both human plasma incubated with 10mg/ml hypericin for 2 hr at

37°C and mouse plasma sample obtained from C3H/Km mice 2 hrafter i.v. injection of 5 mg/kg of hypericin. Plasma proteins wereseparated by density-gradient ultracentrifugation with a KBr/NaClgradient and centrifuging for 26 hr (90,0003 g, 4°C) with aBeckman SW-40 Ti rotor in a Beckman Optima LE-80K ultracen-trifuge. The density gradient was created by careful, sequentiallaying KBr/NaCl solutions of densities of 1.063, 1.019 and 1.006g/ml on top of plasma brought to a density of 1.21 g/ml, whichcorresponds to the conventional boundaries of four classes ofproteins: VLDL (d , 1.006), LDL (1.006, d , 1.063), HDL(1.063 , d , 1.21) and heavy proteins, mainly albumin (d .1.21).22 The position of the different lipoproteins in the gradientwas further confirmed by absorption measurements at 280 nm.23

Immediately after centrifugation, 0.5 ml aliquots were removedsequentially with a peristaltic pump from the bottom of the tube,and the fluorescence of hypericin was determined with a micro-plate fluorescence reader as mentioned above.

Intratumoral localizationTo determine the intratumoral localization of hypericin, the

tumors were excised at 0.5 hr, 6 hr and 24 hr after the adminis-tration of hypericin (5 mg/kg, i.v.). The samples were immediatelymounted in medium (Tissue Tek embedding medium, Miles Inc.,Elkhart, IN) and immersed in liquid nitrogen. Two serial cryostatsections (5-mm slices) were taken from each tumor, one section forstaining with haematoxylin and eosin (H&E) and the other forfluorescence microscopic analysis. The fluorescence microscopicstudy was carried out with Diaplan fluorescence microscopy(Leica, Heerbrugg, Switzerland) equipped with a 100-W high-pressure mercury lamp, a 535-nm bandpass excitation filter (band-width 50 nm), and a 610-nm bandpass emission filter (bandwidth75 nm). The fluorescence image was recorded with a light-sensi-tive charge-coupled device (CCD) digital camera (Leica DC 200).The comparisons were made between the fluorescence images andthe corresponding light microscopy images after H&E staining todetermine the histological tumor distribution of hypericin.

PDT treatment of tumorTumor-bearing mice were given an i.v. injection of 5 mg/kg

hypericin. At 0.5 hr, 1 hr, 2 hr, 4 hr, 6 hr and 24 hr afteradministration, the animals were anesthetized by i.p. injection of90 mg/kg sodium pentobarbital. The tumors were then givenexternal light treatment. For the irradiation, the light beam emittedby a Rhodamine 6G dye laser (375B, Spectra Physics, MountainView, CA, USA) pumped by a 4-W argon laser (Spectra Physics)was coupled to a fiberoptic cable fitted with a microlens (RareEarth Medical, West Yarmouth, MA) to provide ultra-uniformillumination. The laser was set at a wavelength of 595 nm for PDTwith hypericin. This wavelength was verified by a monochromator(DMC1-02, Optometrics UK Ltd., Leeds, England). The light dose(120 J/cm2) was delivered at a fluence rate of 100 mW/cm2, asmeasured by an IL 1400A photometer (International Light, MA).The irradiation spot, centered on the tumor mass, was adjusted to1 cm in diameter. The control group consisted of tumor-bearingmice without treatment. Each group included 7–10 animals.

Assessment of tumor responseAfter treatment, tumor growth, as measured with a caliper, was

recorded each day for 1 month and then weekly up to 60 days. Thetumor response was evaluated by tumor regrowth delay and tumorcure as described elsewhere.24 The growth delay was calculated asthe time taken for a tumor to increase by 2 mm in mean diameterfrom pre-treatment size (T12 mm). Cures were defined as novisible or palpable tumor at 60 days after treatment.

PDT-mediated normal skin reactionThe normal skin response to PDT with hypericin was also

evaluated in C3H mice without tumor. The right hind leg of thesenormal mice was depilated 2 days before irradiation to avoid skinirritation caused by this procedure and then treated with PDT in theexactly same manner as used in the tumor response study. In

FIGURE 1 – The structure and absorption spectrum of hypericin (30mM in PBS with 10% FBS).

276 CHEN ET AL.

addition, the same time intervals between drug and light were usedfor the evaluation. A numerical skin scoring system (Table I) withminor modifications was adopted from Gomer and Razum25 toassess the overall skin response. Two independent observersscored the skin reaction 3 times a week and the final score was theaverage of the 2 individual evaluations. Examination and scoringwere started from the day after the treatment until the skin healedand the hair regrew.

StatisticsThe tumor responses were presented as Kaplan-Meier curves in

which the percentage of animals with tumor sizes less than theendpoint is plotted against the days after treatment. The cures wereincorporated as censored observation. Log rank analysis was usedand significance accepted atp , 0.05. All the analyses, includingnonlinear curve fitting and linear regression, were carried out usingGraphPad software (GraphPad, San Diego, CA)

RESULTS

Absorption spectrum of hypericinThe absorption of hypericin in PBS supplemented with 10%

FBS is shown along with its chemical structure in Figure 1. Thepolycyclic aromatic structure of this compound results in a strongabsorption throughout the visible spectrum with a peak at 595 nm.Consequently, this wavelength was selected for PDT treatment.

Tissue distributionThe kinetic distribution profile of hypericin in the tumor tissue

and the surrounding skin and muscle after intravenous injection (5mg/kg) is shown in Figure 2. The tumor drug concentration in-creased rapidly over the initial hours, peaked approximately 6 hrafter administration, and gradually decreased thereafter. After in-crease in the first few hours, the drug level in the surroundingnormal skin and muscle did not change markedly, which indicatesthat the drug removal is slow during the first 24 hr. As comparedwith the tumor tissue, the concentrations in these tissues weremuch lower. The plasma clearance rate of hypericin was rapid,however, fitting to 2-phase exponential decay (r2 5 0.9819) withhalf-life values of 0.08 hr and 1.4 hr during the first 24-hr periodafter injection (Fig. 3a). The calculated ratios between the hyperi-cin concentrations found in the tumor and plasma, muscle and skinare shown in Figure 4 as a function of time after injection. Thetumor/muscle and tumor/skin ratios amounted to a factor of 5 to 30and 2 to 6, respectively, within the 24-hr period. The tumor/plasmaratios varied considerably, ranging from 0.057, 1.4 and 4.7 at 0.5hr, 4 hr and 8 hr, respectively, after drug administration. Figure 5bshows a RIF-1 tumor under UV365 irradiation 6 hr after injection,the time with the highest tumor drug concentration. It can be seenthat the tumor was much more fluorescent than the surroundingnormal tissue.

Plasma protein distributionThe binding of hypericin to human plasma incubated with 10

mg/ml hypericin for 2 hr at 37°C and mouse plasma obtained from

mice 2 hr after i.v. injection of 5 mg/kg hypericin was assessed byseparating the different plasma proteins by means of density-gradient ultracentrifugation. A concentration of 10mg/ml hyperi-cin for human plasma incubation was chosen because approxi-mately this concentration is reached 2 hr after i.v. injection of mice(5 mg/kg). The fluorescence profile of different plasma constitu-ents (Fig. 6) clearly shows that most of the hypericin was associ-ated with LDL in human plasma with relatively small amounts

TABLE I – SKIN SCORING SYSTEM

Score Observation

Reaction appearing0 Normal1 Slight swelling2 Marked edema3 Slight skin necrosis with exudation4 Extensive necrosis with large area

hemorrhageReaction subsiding

3 Scab formation2 Papery skin appearance1 Slight abnormal appearance0 Normal

FIGURE 2 – Concentration of hypericin, expressed asmg/g of wettissue, in the tumor and the surrounding skin and muscle as a functionof the time after i.v. injection at a dose of 5 mg/kg to the C3H micewith RIF-1 tumors. Each point represents the mean6 SD of 3 animals.

FIGURE 3 – (a) Plasma concentration of hypericin, expressed asmg/ml of plasma, as a function of the time after i.v. injection at a doseof 5 mg/kg to the C3H mice with RIF-1 tumors. Each point representsthe mean6 SD of 3 animals. (b) Tumor response after PDT withhypericin at different time points after drug administration. The RIF-1tumor-bearing mice were irradiated with 120 J/cm2 dose of light (100mW/cm2, 595 nm) at 0.5 hr, 1 hr, 2 hr, 4 hr, 6 hr, 24 hr after injectionof hypericin (5 mg/kg, i.v.). The tumor response, expressed as thepercentage of cure, was determined from animals that were tumor free60 days after treatment. Both curves were fitted to 2-phase exponentialdecay with regression coefficientr2 values of 0.9819 and 0.9886,respectively.

277DISTRIBUTION AND PDT EFFECT OF HYPERICIN

bound to HDL, VLDL and heavy proteins (mainly albumin),whereas in the case of mouse plasma, most of the hypericin wasbound to HDL and heavy proteins.

Intratumoral localizationFigure 7 shows fluorescent images and the corresponding

H&E staining micrographs from sections of RIF-1 tumor takenat 0.5 hr, 6 hr and 24 hr after i.v. injection of hypericin (5mg/kg). At 0.5 hr, intense hypericin fluorescence is clearly seenwithin the tumor blood vessel, with a small amount of fluores-cence in the perivascular region. Conversely, at 6 hr, strongfluorescence is apparent in both tumor interstitial tissue andtumor cell cytoplasm, whereas only faint fluorescence is ob-served in the tumor vasculature. The tumor cell nuclei show nofluorescence. The connective tissue surrounding the normalmuscle fiber also displays fluorescence, but no fluorescence isfound in the muscle fiber itself. At 24 hr, a weak but more diffuselydistributed hypericin fluorescence is observed throughout the tu-mor tissue, although the tumor vasculature exhibits no fluores-cence at this time. It is of interest that hypericin also accumulatedin necrotic regions present in tumor tissue, especially after longertime intervals (not shown).

Tumor response to PDT and correlation with plasmaconcentration

Tumor response to PDT performed at various intervals after i.v.injection of hypericin (5 mg/kg) is shown as a Kaplan-Meier curvein Figure 8. Compared with the control group, PDT treatments atdifferent intervals all have significant antitumor effects (p , 0.05).Initial tumor regression together with pronounced edema wasobserved shortly after treatment, and the tumors generally becamenecrotic and flat within 2 days. Significantly, the PDT at 0.5 hrproduced a 100% cure as evaluated 60 days after PDT. Thistreatment condition, however, also led to extensive skin damagewith heavy edema and large scar formation. The percentage ofcures decreased very rapidly as a function of the interval betweendrug administration and light treatment, and the skin damagebecame less severe as the drug-light interval increased. No curecould be observed at the 6-hr interval, even when at this intervalthe tumoral concentration of hypericin peaked. PDT at 6 hr and 24hr could only delay tumor regrowth, as compared with the controlgroup (p , 0.05). Treatments with hypericin or light alone exhib-ited no antitumor effect (data not shown).

FIGURE 4 – The mean concentration ratios of tumor to the plasmaand surrounding muscle and skin at different time points after i.v.injection of hypericin (5 mg/kg) to the C3H mice with RIF-1 tumors.

FIGURE 5 – Preferential localization of hypericin in RIF-1 tumor.The tumor-bearing C3H mouse was i.v. injected with a 5 mg/kg doseof hypericin and photographed 6 hr later under normal light conditions(a) and UV365 irradiation with a 450 nm long-pass filter (b).

FIGURE 6 – Fluorescence profiles of hypericin bound to humanplasmain vitro (h) and mouse plasmain vivo (■). Human plasmaincubated with 10mg/ml hypericin and mouse plasma obtained fromnormal C3H mice at 2 hr after i.v. injection of 5 mg/kg hypericin wereseparated by density-gradient ultracentrifugation using a KBr/NaClgradient.

278 CHEN ET AL.

Similar to the plasma concentration, the percentage of cures as afunction of the time interval between hypericin administration andlight irradiation could also be fitted very well to 2-phase exponentialdecay (r2 5 0.9886) (Fig. 3b). Consequently, linear regression anal-ysis indicated a high correlation (r2 5 0.9473) between the tumorresponse (expressed as the percent cure) and the plasma concentration

(Fig. 9). No significant correlation between the tumor response andthe tumor drug level was found (r2 5 0.1629).

Time-dependent skin phototoxicityThe normal skin damage induced by hypericin PDT was exam-

ined under conditions similar to the ones used in the tumor re-

FIGURE 7 – Fluorescence (a,c,e) and corresponding H&E staining (b,d,f) photomicrographs of 5mm RIF-1 mouse tumor sections sampled at0.5 hr (a,b), 6 hr (c,d) and 24 hr (e,f) after i.v. injection of 5 mg/kg hypericin. Hypericin is mainly confined within the tumor blood vessel (v)at 0.5 hr. At 6 hr, strong fluorescence was found in both tumor interstitial tissues and tumor cell cytoplasm, whereas only faint fluorescence wasassociated with the vasculature. Note the normal muscle fiber (m) is not fluorescent. At 24 hr, a weak but more diffused fluorescence is spreadthroughout the tumor tissue. Tumor vasculature shows no fluorescence at this time. Scale bar5 25 mm.

279DISTRIBUTION AND PDT EFFECT OF HYPERICIN

sponse study. The results are shown in Figure 10. It is apparent thatirradiation at 0.5 hr and 1 hr led to the most severe skin damage,followed in degree by 2 hr-interval PDT. Skin reactions after the4 hr-, 6 hr- and 24-hr intervals were somewhat comparable. Gen-erally, skin blanching followed by marked edema was seen withina few hours after light treatment. At the 0.5-hr and 1-hr intervalsPDT, extensive hemorrhage was observed before scar formationthe following days. Both treatment conditions led to extensive skinnecrosis and did not heal until 37 and 30 days, respectively, aftertreatment. Skin necrosis was also observed in the 2 hr-intervalPDT group, but not as severe as in the other cases. PDT at 4-hr,6-hr and 24-hr intervals generally induced marked edema withoutapparent skin necrosis. The extent of skin damage expressed by themaximum score reached after treatment correlated much more

closely with the plasma concentration (r2 5 0.9119, Fig. 9) thanwith the skin drug level (r2 5 0.6566).

DISCUSSION

We used a RIF-1 tumor model to evaluate the antitumoraleffects of hypericin-mediated PDT. Our results show that hyperi-cin exhibits a very effective photodependent antitumoral effectagainst subcutaneously growing RIF-1 tumor. Because the RIF-1tumor cell line is not immunogenic in C3H mice, the number ofliving cells required to generate tumors in 50% animals (TD50) isonly about 10.21,31 Consequently, a small number of viable cellsremaining after treatment are always able to expand, which makesRIF tumors difficult to cure.32 In the current study, however, a highcure percentage was obtained under specific PDT conditions. Thesame tumor model has been used with other photosensitizers suchas Photofrin,31,32 silicon phthalocyanine33 and mTHPC.24 Com-pared with the results obtained with these photosensitizers, thepresent data are definitely among the best yet obtained. Althoughit has been suggested that an ideal photosensitizer should havestrong light absorption at wavelength above 630 nm,30 the wave-length of 595 nm used to activate hypericin does not limit the PDTefficacy, at least in this subcutaneously transplanted tumor model.

Significantly, complete tumor eradication was only achievedwhen high intravascular hypericin concentrations were present atthe time of photo-activation. Indeed, a linear correlation could befound between the percentage of animals cured and the hypericinplasma concentration present at the moment of PDT. Conversely,PDT performed when the tumoral concentration of hypericinpeaked (6 hr after injection) produced only tumor growth delay butnot cure. The importance of the localization of hypericin as acrucial determinant in the PDT outcome is further demonstrated bythe finding that average tumor drug levels at 0.5 hr and 24 hr afteri.v. injection were roughly similar. Fluorescence microscopy, how-ever, revealed that hypericin was exclusively located in the vas-cular network of the tumor 0.5 hr after injection, whereas thecompound had almost completely disappeared from the vesselsand was located in the tumor cells at 24 hr. Significantly, PDTperformed at 0.5 hr and 24 hr after administration resulted in a100% and 0% cure, respectively.

The finding that the PDT efficacy of hypericin is highly depen-dent upon the circulating drug level suggests that vascular damageis primarily responsible for the tumor eradication. A similar rela-tionship between the circulating sensitizer level and PDT tumor

FIGURE 8 – Kaplan-Meier curve of RIF-1 tumor regrowth after PDTtreatment with hypericin at different time points. The tumor-bearingmice were irradiated with 120 J/cm2 dose of light (100 mW/cm2, 595nm) at 0.5 hr (Œ), 1 hr (‚), 2 hr (F), 4 hr (E), 6 hr (■), 24 hr (h) afteri.v. injection of 5 mg/kg hypericin. Controls (�) received no treatment.The percentage of tumors less than 2 mm more in mean diameter frompre-treatment size (T12 mm) is plotted against the time after treat-ment. The hanging points represent the percentage of animals cured asobserved 60 days after PDT. Seven to 10 animals were included ineach group.

FIGURE 9 – Correlation between the tumor response (■), skin pho-tosensitivity (�) and plasma concentration. Both the tumor response(expressed as cure percentage) and skin photosensitivity (expressed asthe highest skin grading score) to PDT (120 J/cm2, 595 nm) withhypericin performed at different time points after injection (5 mg/kg,i.v.) correlate closely with the plasma concentration with correlationcoefficient values (r2) of 0.9473 and 0.9119, respectively.

FIGURE 10– Normal skin response after exposing C3H mice toirradiation (120 J/cm2 at 100 mW/cm2, 595 nm) at 0.5 hr (Œ), 1 hr (‚),2 hr (F), 4 hr (E), 6 hr (■), 24 hr (h) after i.v. injection of 5 mg/kghypericin. Each point represents the average score of 3 animals. TheSD values of most points were less than 20%.

280 CHEN ET AL.

response has been documented for other vasculature-targeting pho-tosensitizers like mono-L-aspartyl chlorin e6,26 bacteriochlorins,27

mTHPC24 andbenzoporphyrin derivative.28 It has been shown thatvascular collapse could be caused by damage to sensitive vasculartargets such as endothelial cells, plateletsand red blood cells.3,29

These targets all are directly exposed to hypericin in circulation.Therefore, illumination at high plasma drug concentrations (atshort intervals after drug injection) results in more severe vasculardamage. Subsequent to vascular collapse, tumor cells are mainlykilled by the induced tissue ischemia.

Irradiation at longer drug-light intervals would decrease thisvascular effect due to reduced plasma concentrations and thereforedecrease the overall PDT efficacy. In this study, it was found thatPDT at 6 hr and 24 hr after drug administration led only to tumorregrowth delay without cure. It has been reported that extendingthe drug-light interval increases the contribution of direct tumortoxicity because the drug has extravasated to the tumor tissue atthese time points30 as observed in this study. Because PDT de-pends on the presence of oxygen, however, tissue hypoxia as aresult of the acute vascular effect together with photochemicalconsumption of oxygen during the PDT process might limit thedirect tumoral killing effect.3 That the hypericin-mediated PDTeffect does not correlate well with its tumor concentration furtherindicates that direct tumor cell kill plays a less significant role inthe overallin vivo PDT effect in this tumor model.

Similar to the tumor response, hypericin-mediated skin photo-toxicity also follows the plasma pharmacokinetics more closelythan the skin drug level, as reported with other photosensitizerslike Photofrin34 and benzoporphyrin.35 The high percentages oftumor cure after PDT with hypericin at short intervals also in-curred a significant normal tissue damage. Even though destroyinga margin of normal tissue has been shown to be essential for tumorcure,36,37 future attempts should be made to achieve the maximalantitumor effects with minimal damage to the surrounding tissues.For instance, it has been shown that thresholds of both light anddrug doses exist for the PDT effect, and this threshold may behigher for normal tissue than for tumor.24 This difference in PDTresponse between tumors and normal tissues due to differences invascular integrity might provide a way to achieve better tumorselectivity. Additional investigations are, therefore, warranted tomanipulate the drug and light doses and the time interval betweendrug injection and light delivery to balance the tumor ablation andnormal tissue damage for the highest selective therapeutic effects.

Interestingly, the data presented here indicate that hypericinpreferentially accumulates in tumor tissue, which is in agreementwith our previous study.20 Because this compound features a highfluorescent quantum yield,7 this finding suggests that hypericincould also be used as a cancer diagnostic tool. In fact, we have

recently introduced hypericin in the clinic for the selective detec-tion of flat bladder CIS.38

The precise mechanism involved in this preferential tumor ac-cumulation, however, is not understood at present. As suggestedfor other photosensitizers,2,39 it may be due to the interaction ofseveral different factors including physicochemical properties ofthe photosensitizer, plasma protein binding, pharmacokinetic be-havior as well as tumor tissue properties. Among them, tumoruptake of photosensitizers through receptor-mediated endocytosisof LDL has been shown to be an important determinant for thepreferential accumulation, because a high LDL receptor expressionwas found in tumor cells.40 The present study shows that hypericinhas a high affinity for LDL in human plasma and for HDL inmouse plasma. This apparent discrepancy can be explained interms of species differences because human plasma contains muchmore LDL than HDL whereas the opposite pattern is found inmouse plasma.39 It has been shown, however, that redistribution toLDL binding can occur in the case where a drug binds predomi-nantly to other lipoproteins in plasma.39 Whether this can happenwith hypericin as well as the relative contribution of this pathwayto the preferential accumulation of hypericin has yet to be deter-mined. Finally, the different drug-binding behavior observed in thehuman and mouse in this study might suggest that, to determinedrug distribution in a more relevant situation, the tumor modelusing animals with similar levels of plasma protein constituents ashuman (for example, the hamster tumor model) would be pre-ferred.

In conclusion, we found that hypericin is an effective PDT tool,the effectiveness of which depends largely on the circulating ratherthan the tumor drug level. Vascular damage seems to contributesignificantly to the overall PDT effect, and PDT at a short intervalcan be chosen to maximize this effect for the treatment of canceror other diseases in which vessel ablation is desired, such asage-related macular degeneration (AMD). Furthermore, the pref-erential tumor accumulation of hypericin demonstrated in thisstudy in combination with its high fluorescence yield could beexploited for tumor diagnosis. Because tumor response and drugmetabolism might vary between species and in tumors with dif-ferent vascularity and histological origin, however, translationfrom this animal study to the human situation has to be undertakenwith great care. Future efforts are warranted to determine whetherthese results are applicable in the clinical situation.

ACKNOWLEDGEMENTS

We thank Dr. Fiona Stewart and Hugo Oppelaar (The Nether-lands Cancer Institute) and Gerda Luyckx (Department of Pathol-ogy) for helpful discussions and excellent technical assistance.

REFERENCES

1. Oleinick NL, Evans HH. The photobiology of photodynamic therapy:cellular targets and mechanisms. Radiat Res 1998;150:S146–56.

2. Dougherty TJ, Gomer CJ, Henderson BW, Jori G, Kessel D, KorbelikM, et al. Photodynamic therapy. J Natl Cancer Inst 1998;90:889–905.

3. Fingar VH. Vascular effects of photodynamic therapy. J Clin LaserMed Surg 1996;14:323–8.

4. Korbelik M. Induction of tumor immunity by photodynamic therapy.J Clin Laser Med Surg 1996;14:329–34.

5. Racinet H, Jardon P, Gautron R. Formation d’oxygene singulet 1 deltag photosensibilisee par 1’hypericine,etude cinetique en milieu micel-laire non ionique. J Chim Phys 1988;85:971–7.

6. Sharman WM, Allen CM, van Lier JE. Photodynamic therapeutics:basic principles and clinical applications. Drug Discov Today 1999;4:507–17.

7. Yamazaki T, Ohta N, Yamazaki I, Song PS. Excited-state propertiesof hypericin: electronic spectra and fluorescence decay kinetics. JPhys Chem 1993;97:7870–5.

8. VanderWerf QM, Saxton RE, Chang A, Horton D, Paiva MB, Ander-son J, et al. Hypericin: a new laser phototargeting agent for humancancer cells. Laryngoscope 1996;106:479–83.

9. Hadjur C, Richard MJ, Parat MO, Jardon P, Favier A. Photodynamic

effects of hypericin on lipid peroxidation and antioxidant status inmelanoma cells. Photochem Photobiol 1996;64:375–81.

10. Vandenbogaerde AL, Delaey EM, Vantieghem AM, Himpens BE,Merlevede WJ, de Witte PA. Cytotoxicity and antiproliferative effectof hypericin and derivatives after photosensitization. Photochem Pho-tobiol 1998;67:119–25.

11. Fox FE, Niu Z, Tobia A, Rook AH. Photoactivated hypericin is ananti-proliferative agent that induces a high rate of apoptotic death ofnormal, transformed and malignant T lymphocytes: implications forthe treatment of cutaneous lymphoproliferative and inflammatorydisorders. J Invest Dermatol 1998;111:327–32.

12. Johnson SA, Dalton AE, Pardini RS. Time-course of hypericin pho-totoxicity and effect on mitochondrial energies in EMT6 mousemammary carcinoma cells. Free Radic Biol Med 1998;25:144–52.

13. Chaloupka R, Petit PX, Israel N, Sureau F. Over-expression of Bcl-2does not protect cells from hypericin photo-induced mitochondrialmembrane depolarization, but delays subsequent events in the apo-ptotic pathway. FEBS Lett 1999;462:295–301.

14. Diwu Z, Lown JW. Photosensitization with anticancer agents. 17.EPR studies of photodynamic action of hypericin: formation ofsemiquinone radical and activated oxygen species on illumination.Free Radic Biol Med 1993;14:209–15.

281DISTRIBUTION AND PDT EFFECT OF HYPERICIN

15. Miccoli L, Beurdeley TA, De Pinieux G, Sureau F, Oudard S, Dutril-laux B, et al. Light-induced photoactivation of hypericin affects theenergy metabolism of human glioma cells by inhibiting hexokinasebound to mitochondria. Cancer Res 1998;58:5777–86.

16. Vantieghem A, Assefa Z, Vandenabeele P, Declercq W, Courtois S,Vandenheede JR, et al. Hypericin-induced photosensitization of HeLacells leads to apoptosis or necrosis. Involvement of cytochrome c andprocaspase-3 activation in the mechanism of apoptosis. FEBS Lett1998;440:19–24.

17. Assefa Z, Vantieghem A, Declercq W, Vandenabeele P, VandenheedeJR, Merlevede W, et al. The activation of the c-Jun N-terminal kinaseand p38 mitogen-activated protein kinase signaling pathways protectsHeLa cells from apoptosis following photodynamic therapy withhypericin. J Biol Chem 1999;274:8788–96.

18. Thomas C, Macgill RS, Neill P, Pardini RS. Thein vitro and in vivophotoinduced antineoplastic activity of hypericin. Proc Am AssocCancer Res 1992;33:500.

19. Vandenbogaerde AL, Geboes KR, Cuveele JF, Agostinis PM, Mer-levede WJ, de Witte PA. Antitumor activity of photosensitized hy-pericin on A431 cell xenografts. Anticancer Res 1996;16:1619–25.

20. Chen B, de Witte PA. Photodynamic therapy efficacy and tissuedistribution of hypericin in a mouse P388 lymphoma tumor model.Cancer Lett 2000;150:111–7.

21. Twentyman PR, Brown JM, Gray JW, Franko AJ, Scoles MA, Kall-man RF. A new mouse tumor model system (RIF-1) for comparisonof end-point studies. J Natl Cancer Inst 1980;64:595–604.

22. Cuomo V, Jori G, Rihter B, Kenney ME, Rodgers MA. Liposome-delivered Si(IV)-naphthalocyanine as a photodynamic sensitizer forexperimental tumors: pharmacokinetic and phototherapeutic studies.Br J Cancer 1990;62:966–70.

23. Hopkinson HJ, Vernon DI, Brown SB. Identification and partialcharacterization of an unusual distribution of the photosensitizermeta-tetrahydroxyphenyl chlorin (temoporfin) in human plasma. Pho-tochem Photobiol 1999;69:482–8.

24. Veenhuizen R, Oppelaar H, Ruevekamp M, Schellens J, Dalesio O,Stewart F. Does tumor uptake of Foscan determine PDT efficacy? IntJ Cancer 1997;73:236–9.

25. Gomer CJ, Razum NJ. Acute skin response in albino mice followingporphyrin photosensitization under oxic and anoxic conditions. Pho-tochem Photobiol 1984;40:435–9.

26. Ferrario A, Kessel D, Gomer CJ. Metabolic properties and photosen-sitizing responsiveness of mono-L-aspartyl chlorin e6 in a mousetumor model. Cancer Res 1992;52:2890–3.

27. Kessel D, Smith KM, Pandey RK, Shiau FY, Henderson B. Photosensi-tization with bacteriochlorins. Photochem Photobiol 1993;58:200–3.

28. Fingar VH, Kik PK, Haydon PS, Cerrito PB, Tseng M, Abang E, et al.Analysis of acute vascular damage after photodynamic therapy usingbenzoporphyrin derivative (BPD). Br J Cancer 1999;79:1702–8.

29. Menezes-da-Silva FA, Newman EL. Time-dependent photodynamicdamage to blood vessels: correlation with serum photosensitizer lev-els. Photochem Photobiol 1995;61:414–6.

30. Stewart F, Baas P, Star W. What does photodynamic therapy have tooffer radiation oncologists (or their cancer patients)? Radiother Oncol1998;48:233–48.

31. Henderson BW, Fingar VH. Relationship of tumor hypoxia and re-sponse to photodynamic treatment in an experimental mouse tumor.Cancer Res 1987;47:3110–4.

32. Henderson BW, Waldow SM, Mang TS, Potter WR, Malone PB,Dougherty TJ. Tumor destruction and kinetics of tumor cell death in2 experimental mouse tumors following photodynamic therapy. Can-cer Res 1985;45:572–6.

33. Anderson CY, Freye K, Tubesing KA, Li YS, Kenney ME, MukhtarH, et al. A comparative analysis of silicon phthalocyanine photosen-sitizers for in vivo photodynamic therapy of RIF-1 tumors in C3Hmice. Photochem Photobiol 1998;67:332–6.

34. Bellnier DA, Dougherty TJ. The time course of cutaneous porphyrinphotosensitization in the murine ear. Photochem Photobiol 1989;49:369–72.

35. Lin GC, Tsoukas ML, Lee MS, Gonzalez S, Vibhagool C, AndersonRR, et al. Skin necrosis due to photodynamic action of benzoporphy-rin depends on circulating rather than tissue drug levels: implicationsfor control of photodynamic therapy. Photochem Photobiol 1998;68:575–83.

36. Fingar VH, Henderson BW. Drug and light dose dependence ofphotodynamic therapy: a study of tumor and normal tissue response.Photochem Photobiol 1987;46:837–41.

37. Star WM, Marijnissen HP, Berg-Blok AE, Versteeg JA, Franken KA,Reinhold HS. Destruction of rat mammary tumor and normal tissuemicrocirculation by hematoporphyrin derivative photoradiation ob-served in vivo in sandwich observation chambers. Cancer Res 1986;46:2532–40.

38. D’Hallewin MA, de Witte PA, Waelkens E, Merlevede WaBL. Flu-orescence detection of flat bladder carcinomain situ after intravesicalinstillation of hypericin. J Urol 2000;164:349–51.

39. Kessel D, Woodburn K. Biodistribution of photosensitizing agents.Int J Biochem 1993;25:1377–83.

40. Jori G, Reddi E. The role of lipoproteins in the delivery of tumour-targeting photosensitizers. Int J Biochem 1993;25:1369–75.

282 CHEN ET AL.