Proc. Natl. Acad. Sci. USAVol. 88, pp. 1908-1912, March 1991Medical Sciences

Skin toxicity determined in vitro by three-dimensional, native-state histoculture

(intact tissue architecture/vital dyes/confocal fluorescent microscopy)

LINGNA LI*, LEONID B. MARGOLIS*tt, AND ROBERT M. HOFFMAN*t§*AntiCancer, Inc., 5325 Metro Street, San Diego, CA 92110; tLaboratory of Cancer Biology, University of California, San Diego, 0609F, La Jolla, CA92093-0609; and tA. N. Belozersky Laboratory of Bioorganic Chemistry and Molecular Biology, Moscow State University, Building A, Moscow 119899, USSR

Communicated by Israel M. Gelfand, November 2, 1990

ABSTRACT We describe a gel-supported in vitro systemfor culturing skin samples in a three-dimensional native state.All the cell types of skin remain viable and maintain their nativearchitecture for at least 10 days. The culture system is used fortoxicity measurements by ascertaining cell viability using twofluorescent dyes: 2',7'-bis-(2-carboxyethyl)-5-(and -6)carboxy-fluorescein acetoxymethyl ester, specific for living cells, andpropidium iodide, specific for dead cells. Cell staining with thedyes is measured throughout the tissue block by confocalscanning fluorescence microscopy. The dose-response to threeagents-ethanol, doxorubicin, and sodium hypochlorite-isshown and, in the case of sodium hypochlorite, compared to invivo skin toxicity with a high correlation. We also demonstratethat the end point of [3H]thymidine incorporation measured byhistological autoradiography can be used to measure toxicity.Our results with the [3Hlthymidine end point demonstrate thatthe hair follicle cells are the most sensitive to doxorubicin. Thenative-state model for skin may be an effective replacement foranimal systems and superior to the dispersed skin cell systemsused previously. It can allow rapid, inexpensive measurementsof the effect of manufactured products, drugs, and pollutantson skin.

Determining the toxicity of various substances in animals isexpensive, time-consuming, and sometimes misleading be-cause of the complexity of in vivo systems. The need for asuitable in vitro system for the study of skin toxicology haslong been recognized (1). In vitro systems have the benefit ofallowing the rapid, inexpensive study of a large variety ofsubstances, as well as replacing animals (2).A number of attempts to develop in vitro systems for

toxicology for skin have been reported (3-5). Based onmonolayer cell cultures, a toxicity test system has certainadvantages. However, the model is oversimplified and dis-tant from the in vivo situation in which total effects are, inpart, mediated by complex cell-cell interactions not occur-ring in monolayers. A three-dimensional reconstructed skinmodel has been established (6) by seeding dermal fibroblastsonto a nylon mesh to form a dermal equivalent. Melanocytesand keratinocytes are inoculated onto the dermal equivalentand remain above the dermal-epidermal junction. Althoughthis model is closer to the in vivo state than monolayers, it stillis markedly dissimilar from true skin. To more closelyapproximate the in vivo state, skin samples from the mouse,rat, rabbit, guinea pig, marmoset, and human have beencultured but these remain viable for only 24 hr (7).An adequate in vitro model for toxicology testing requires

relatively long-term culture of human and animal skin. Theimportant cellular components must remain viable and growwhile maintaining their native tissue architecture. Such aculture system permits study of long-term effects of various

agents, such as toxicity, and even malignant transformation.We report here the development of a gel-supported, three-dimensional histoculture system that maintains viability, cellproliferation, and intact tissue architecture of all componentsof mouse skin for at least 10 days.

Staining with fluorescent dyes allows visualizing all cellswithin the skin and determines their viability. The stainingthroughout the skin sample is visualized by using the verylarge working distances afforded by confocal scanning mi-croscopy. The system described provides an easily quantifiedtoxicity assay by using one dye to measure viable cells andanother to determine dead cells. Alternatively, we demon-strate here that [3H]thymidine incorporation end points mea-sured by histological autoradiography can determine theeffects of toxins on cell proliferation. The in vitro effects oftoxins are demonstrated, and those of sodium hypochloriteare compared to the results of the agent in vivo.

MATERIALS AND METHODSHistoculture of Skin. We have used the method developed

earlier by Hoffman et al. (8-12) based on the work ofLeighton (13) to culture various tumor and normal tissues.Small pieces of intact athymic nude mouse skin (--2 x 5 mm2and 2.0 mm thick) were cut with scissors under a dissectingmicroscope and put onto collagen-containing sponges or gelsin histoculture in medium as soon as possible. The use of skinfrom nude mice eliminated the need to shave the animals.Eagle's minimum essential medium supplemented with 10%ofetal bovine serum and gentamycin was used and cultureswere maintained at 370C in a gassed incubator with a mixtureof 95% air/5% CO2.

Confocal Microscopy. The confocal microscopy system waswith an MRC-600 confocal imaging system (Bio-Rad) mountedon a Nikon Optiphot using a 1Ox PlanApo objective.

Fluorescence Microscopy. For standard microscopy, a Ni-kon fluorescent microscope, equipped with fluorescein andrhodamine cubes, was used.

Fluorescent-Dye Labeling of Live and Dead Cells. Viablecells were selectively labeled with the dye 2',7'-bis-(2-carboxyethyl)-5-(and -6)carboxyfluorescein acetoxymethylester (BCECF-AM), which is activated to fluorescence bynonspecific esterases present only in living cells (14). Non-viable cells, whose plasma membranes are leaky, were la-beled with propidium iodide (PI), a dye that enters only cellswith nonintact membranes (14). Since the emission spectra ofthese two dyes were different they could be used simultane-ously on the same specimen. Both dyes were used at aconcentration of 5 1LM. The double dye-treated cultures wereanalyzed by fluorescence and confocal microscopy within 30min of staining.

Abbreviations: BCECF-AM, 2',7'-bis-(2-carboxyethyl)-5-(and-6)carboxyfluorescein acetoxymethyl ester; PI, propidium iodide;P11, primary irritation index.§To whom reprint requests should be addressed.

1908

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 88 (1991) 1909

[3lHJThymidine Labeling of Proliferating Cells. In additionto cytotoxicity, the effects of substances can be measured byinhibition of [3H]thymidine incorporation and subsequenthistological autoradiography. Briefly, cells within the three-dimensional cultures capable of proliferation were labeled byadministration of [3H]thymidine at 4 ,uCi/ml (1 Ci = 37 GBq)for 3 days. Cellular DNA was labeled in any cells undergoingreplication within the tissues. After 3 days of labeling, thecultures were washed with phosphate-buffered saline, placedin histology capsules, and fixed in 10% formalin. The cultureswere then dehydrated, embedded in paraffin, and sectionedby standard methods. After the slides were deparaffinized,they were prepared for autoradiography by coating withKodak NTB-2 emulsion and exposed for 5 days, after whichthey were developed (8-12). After rinsing, the slides werestained with hematoxylin and eosin. The slides were thenanalyzed by determining the percentage of cells undergoingDNA synthesis in treated vs. untreated tumor cultures, usinga Nikon or Olympus photomicroscope fitted with epi-illumination polarization. Replicating cells were identified bythe presence of silver grains, visualized as bright green in theepi-polarization system, over their nuclei due to exposure ofthe NTB-2 emulsion by radioactive DNA (11).

Toxicity Testing. The toxicities of ethanol, doxorubicin, andsodium hypochlorite were used for toxicity testing in dose-response measurements on histocultured mouse skin. After 2days of incubation, the histocultured skin was stained withBCECF-AM and PI as described above before treatment toobtain the untreated self control. After the initial observation,the skin preparations were exposed to various concentrationsof ethanol for 5 min, to doxorubicin for 24 hr, and to sodiumhypochlorite for 2 hr. After treatment, the culture medium wasremoved and replaced with fresh medium and the culture wasrestained with BCECF-AM and PI. The same area examinedbefore treatment was then observed post-treatment. The celltypes (epidermal, dermal, and follicle cells) within the intactskin cultures were observed for toxicity. The fraction of killedcells (PI-positive cells) with respect to total cells [PI-positive(red) and BCECF-AM-positive (green) cells] in treated cul-tures, compared to untreated controls, was measured byconfocal fluorescence scanning laser microscopy to determinecytotoxicity. The percentage of killed cells was calculatedfrom the following formula: % of killed cells = (Ra - Rb)/[totalcells (Ga + Ra)], where Ra is the number of red cells aftertreatment, Rb is the number of red cells before treatment, andGa is the number of green cells after treatment.To determine cytostatic effects, inhibition of [3H]thymi-

dine incorporation by histological autoradiography (seeabove) was carried out after doxorubicin treatment for 24 hr.The percentage cell proliferation relative to control wascalculated with a Nikon or Olympus photomicroscope fittedwith an epi-illumination polarization lighting system to de-termine cytostatic effects of substances (10).

Irritation Test in Vivo. To compare with the toxicity testingin vitro, primary skin irritation tests were performed on nudemice with sodium hypochlorite. Small pieces of gauze paper(=1.5 cm2) that were soaked in different concentrations ofsodium hypochlorite were tightly put on the dorsal surface ofthe mice. There were three sites on the dorsal surface ofeachmouse for each concentration. Skin sites were scored on ascale of 1-4 for erythema and edema 0.5 hr and 1 hr afterremoval of the sodium hypochlorite. The primary irritationindex (PII) for each concentration of sodium hypochloritewas derived from the following formula (15, 16):

total score (erythema and edema)P11 =

number of sites x number of observations

RESULTSLong-Term Intact Skin Cultures. We have used standard

and confocal microscopy to analyze the skin histoculturesgrown for various periods of time. As shown in Figs. 1-3, themouse skin histocultured on a floating collagen-containing gelin Eagle's minimum essential medium preserves native-statethree-dimensional tissue architecture of all the major celltypes in the viable state. The hair follicles (Fig. 1), epidermalcells (Fig. 2), and dermal fibroblasts (Fig. 2) are readilyvisualized and appear to have normal morphology. Thehistocultures are viable for at least 10 days.The cells in the skin histoculture retain their proliferative

activity. DNA synthesis and replication are shown by histo-logical autoradiography to occur in hair follicles, epidermal,and dermal cells after 6 days of histoculture (Fig. 3). With a3-day labeling period with [3H]thymidine, the percentages oflabeled cells for the follicle, epidermal, and dermal cells were38.6%, 75.2%, and 10.2%, respectively.

FIG. 1. (A) Ethanol toxicity dose-response of hair follicles ofmouse skin histocultured for 2 days and double-stained withBCECF-AM (green, live cells) and PI (red, dead cells) before (leftimage) and after (right image) treatment. Top images are 0.5%ethanol treatment. Bottom images are 5% ethanol treatment. Con-focal scanning laser microscopy. (B) Ethanol toxicity dose-responseof hair follicles of mouse skin histocultured for 2 days and double-stained with BCECF-AM and PI before (left image) and after (rightimage) treatment. Top images are 30% ethanol treatment. Bottomimages are 50o ethanol treatment. Confocal scanning laser micros-copy. (XlOOO.)

Medical Sciences: Li et al.

Proc. Natl. Acad. Sci. USA 88 (1991)

FIG. 2. (A) Mouse skin histocultured for 2 days and double-stained with BCECF-AM (green, live cells) and PI (red, dead cells)before ethanol treatment. Note that most of the cells are viable(green). Note the distinct epidermal and dermal layers. Confocalscanning laser microscopy. (B) Same specimen as in A. BCECF-AMand PI restaining after 70% ethanol treatment for 5 min. Note that allepidermal and dermal cells are dead (red). Confocal scanning lasermicroscopy. (x 1000.)

Fluorescence Toxicity Test. Two fluorescent dyes, BCECF-AM and PI, staining viable and nonviable cells, respectively,were used to study the in vitro toxic effects of ethanol,doxorubicin, and sodium hypochlorite. The dose-responsefor sodium hypochlorite was compared to an in vivo animaltest (see below).

Fig. 1 demonstrates the viability of hair follicles in histo-cultured mouse skin before and after treatment with ethanol.Increasing concentrations ofethanol cause greater PI stainingand less BCECF-AM staining, indicating loss of viability.Fig. 2 shows the toxicity of ethanol on the epidermal anddermal cells of the histocultured mouse skin. Before ethanoltreatment most of the cells were viable, but treatment withlarger amounts of ethanol caused more epidermal and dermalcells to be nonviable by the dual-fluorescence test. Fig. 4A isthe dose-response curve of histocultured mouse skin toethanol.The doxorubicin toxicity dose-response curve of histo-

cultured mouse skin demonstrates that increasing amounts ofdoxorubicin cause increasing percentages of nonviable cellsas measured by the dual-fluorescence test shown in Fig. 4B.

FIG. 3. (A) Mouse skin histocultured for 6 days in Eagle'sminimum essential medium and labeled with [3H]thymidine for days3-6. Tissue was fixed and processed for autoradiography and stainedwith hematoxylin and eosin. Autoradiograms were observed withbright-field and polarizing light. [3H]Thymidine-labeled nuclei ap-pear bright green. Note the high labeling index of hair follicle cells.(B) Same as A. Note the labeled epidermal cells. (A, x1000; B,x2000.)

Autoradiographic Toxicity Test. Another assay of toxicityuses [3H]thymidine incorporation and subsequent histologi-cal autoradiography. A 24-hr exposure of histocultured skinto doxorubicin inhibited DNA synthesis of histocultured skincells as shown in Fig. 4C. The percentage of cell proliferationrelative to control was greatly decreased with increasingconcentrations ofdoxorubicin. The different cell types in skinhad different responses to doxorubicin. DNA synthesis in thehair follicle cells was inhibited more than the epidermal ordermal cells and, therefore, they are most sensitive to thistoxic agent (17).

Irritation Test in Vivo Correlated to in Vitro Toxicity. Asshown in Fig. 4D, dose-response to sodium hypochlorite(bleach) was measured in vitro and in vivo. A high in vitro-invivo correlation was found. In particular, the threshold oftoxicity was similar in vitro and in vivo. Both the percentageof killed cells in vitro and the PIT in vivo increased with asimilar increasing concentration of sodium hypochlorite.

DISCUSSIONThere are at least three problems in developing adequate invitro toxicity test systems for the skin.

(i) If cell monolayers are used most of the specific cell-cellcontact-dependent effects are lost, which can compromisethe accuracy of modeling of tissue.

(ii) The three-dimensional cultures of the skin cells de-scribed so far either are constructed in a model not fully

1910 Medical Sciences: Li et al.

'41

Proc. Natl. Acad. Sci. USA 88 (1991) 1911

z 0....................................................................... ..........................::...........................................D. ............................................. .....................................................................................

..................................................

5 -------- ................ .........................................................

30 - ---- -- -------------------------- - ----------- -- ----------- - - --------- - ----------- - --- ----------------------------

1' t......................................................................................................................................10 -.

10-..............................................................................................................................................0 10 20 30 40 50

Ethanol concentration, %

~~~~~~100...~~~~~~~~~~~~~~~~~~~~~.............

.............................................................................................................................................I

...............................................................................................................................................

................................................................................................................................................

................................................................................................................................................

.....................................................................................................................................

5:::::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::C:.::s:U I.

....................................................

................................................p.................................. IE dermal Cells

: : : : :- ........................................I.......................... ..................................................................

.F.llipal Ce~l29.290.725.................................s .$

29 290 725

B

in

~0

4-0

el

2

D

Ca)

~0

4-0

Doxorubicin concentration, ng/ml

Doxorubicin concentration, ng/ml

0.06 0.6 6.0Sodium hypochlorite concentration, %

FIG. 4. (A) Ethanol toxicity dose-response ofmouse skin histocultured for 2 days was measured by BCECF-AM and PI staining after ethanoltreatment for 5 min. A percentage of killed cells (PI-positive cells) to total cells (BCECF-positive + PI-positive cells) in treated cultures comparedto untreated controls is calculated with confocal fluorescence scanning microscopy to determine cytotoxicity. The percentage of killed cellsincreased along with an increasing concentration of ethanol. (B) Doxorubicin toxicity dose-response of mouse skin histocultured for 4 days wasmeasured by BCECF-AM and PI staining after doxorubicin treatment for 24 hr. A percentage of killed cells (PI-positive cells) to total cells(BCECF-positive + PI-positive cells) in treated cultures compared to untreated controls is calculated with confocal fluorescence scanningmicroscopy to determine cytotoxicity. Note that the percentage of killed cells increased with an increasing concentration of doxorubicin. (C)Doxorubicin toxicity dose-response of the different cell types of mouse skin histocultured for 2 days was measured by [3H]thymidineincorporation and subsequent histological autoradiography after doxorubicin treatment for 24 hr. The percentage of cell labeling relative tocontrol is calculated to determine cytotoxicity. Note that the percentage decreased with an increasing concentration of doxorubicin and the hairfollicle cells had the greatest inhibition of DNA synthesis. (D) Dose-response to sodium hypochlorite toxicity of histocultured mouse skincompared with dose-response to sodium hypochlorite irritation ofmouse skin in vivo. Note that both percentages of killed cells in vitro (brokenline) and P11 in vivo (solid line) increased with an increasing concentration of sodium hypochlorite.

representative of the in vivo state or fail to preserve the cellsin the living state in intact skin cultured in vitro for more thana short period.

(iii) If the cells are growing in a three-dimensional system,standard microscopy cannot resolve the individual cellsbelow the outer layer. Histological sectioning remains arather complex and time-consuming procedure and does notallow for three-dimensional studies or real-time studies.Histological sectioning requires fixation of the sample, thusexcluding the prolonged observation of the response ofindividual cells.Here we describe a test system to evaluate skin toxicity

that overcomes the above-described problems; a three-dimensional gel-supported, long-term histoculture that pre-serves native tissue architecture. Confocal laser fluorescentmicroscopy can observe individual fluorescent cells through-out the tissue. Finally, two fluorescent dyes were used tosimultaneously evaluate the viability state of individual cells,

or [3H]thymidine incorporation was measured by histologicalautoradiography to determine cell proliferation.BCECF, when applied as the acetoxymethyl ester, pene-

trates the cell membrane and fluoresces only if cellularesterases are active, thus reporting on cell viability. PI doesnot penetrate intact plasma membranes and thus stains onlymembrane-damaged cells (14). In preliminary experiments,we have shown that PI stains at least all the cells stained inthe standard dye-exclusion tests with trypan blue (data notshown). After microscopic analysis, the stained specimenscan be cultured further and restained for analysis. We havedetermined that analysis with these two dyes is complete;that is, every cell in the culture is stained by one of the dyes.We have never observed cells stained by the two dyessimultaneously.We have used three substances as models to show that the

test is quantitative, dose responsive, and in vivo-like. Two ofthe substances were ethanol and the chemotherapeutic drug

A

C.

Cr

.1.4

0 "...-4Imo 0

0&I 1.4

Medical Sciences: Li et al.

Proc. Natl. Acad. Sci. USA 88 (1991)

doxorubicin, which is used to treat various malignancies suchas breast cancer in spite of its high toxicity. In both cases, we

observed a dose-dependent toxic effect according to our

method. While ethanol kills cells independent of type, dox-orubicin is preferentially toxic for hair-follicle cells. Thiseffect of the drug may be related to the alopecia it causes invivo. Thus, hair follicle cells could probably serve as the mostsensitive indicators for in vitro toxicology of skin.The third toxic agent, sodium hypochlorite, was measur-

able both in vitro and in vivo. Although the two measurableeffects are different, they are compared here only to showthat onset and saturation occur at similar concentrations ofthe toxic agent, suggesting a correlation between the in vitroand in vivo results for sodium hypochlorite.

In experiments using inhibition of [3H]thymidine as an endpoint, we have shown that doxorubicin preferentially inhibitsfollicle cells, which mirrors the in vivo effect of doxorubicin.Inhibition of [3H]thymidine incorporation may provide an

intermediate end point with respect to cell death to measure

the effects of substances at high levels of sensitivity.

We gratefully acknowledge the skills of Ms. Polly Jayne Pomeroyin preparing the manuscript. We thank Mr. George P. Himel for hisexcellent photographic work. This paper is dedicated to the 60thbirthday of Professor Sheldon Penman for his unique, deep, andlasting contributions to our understanding of cell biology. L.B.M.was supported by the Yamagawa-Yoshida Memorial InternationalCancer Research Grant funded by the Japan National Committee forthe International Union Against Cancer (U.I.C.C., Switzerland) andthe Olympus Company (Tokyo).

1. Goldberg, A. M. (1989) in Alternative Methods in Toxicology,ed. Goldberg, A. M. (Leibert, New York), Vol. 7, pp. V-VI.

2. Bell, E., Parenteau, N. L., Haimes, H. B., Gay, R. J., Kemp,P. D., Fofonoff, T. W., Mason, V. S., Kagan, D. T. &Swiderek, M. (1988) in Alternative Methods in Toxicology, ed.Goldberg, A. M. (Leibert, New York), Vol. 6, pp. 15-25.

3. Babich, H., Martin-Alguacil, N. & Borenfreund, E. (1989) inAlternative Methods in Toxicology, ed. Goldberg, A. M. (Leib-ert, New York), Vol. 7, pp. 153-167.

4. Gales, Y. A., Gross, C. L., Krebs, R. C. & Smith, W. J. (1989)in Alternative Methods in Toxicology, ed. Goldberg, A. M.(Leibert, New York), Vol. 7, pp. 169-174.

5. LaMont, G. S., Bagley, B. M., Kong, B. M. & De Salva, S. J.(1989) in Alternative Methods in Toxicology, ed. Goldberg,A. M. (Leibert, New York), Vol. 7, pp. 175-181.

6. Naughton, G. K., Jacob, L. & Naughton, B. A. (1980) inAlternative Methods in Toxicology, ed. Goldberg, A. M. (Leib-ert, New York), Vol. 7, pp. 183-189.

7. Kao, J., Patterson, F. K. & Hall, J. (1985) Toxicol. Appl.Pharmacol. 81, 502-516.

8. Freeman, A. E. & Hoffman, R. M. (1986) Proc. Natl. Acad.Sci. USA 83, 2694-2698.

9. Vescio, R. A., Redfern, C. H., Nelson, T. J., Ugoretz, S.,Stem, P. H. & Hoffman, R. M. (1987) Proc. Natl. Acad. Sci.USA 84, 5029-5033.

10. Hoffman, R. M., Connors, K. M., Meerson-Monosov, A. Z.,Herrera, H. & Price, J. H. (1989) Proc. Natl. Acad. Sci. USA86, 2013-2017.

11. Vescio, R. A., Connors, K. M., Youngkin, T., Bordin, G. M.,Robb, J. A., Umbreit, J. N. & Hoffman, R. M. (1990) Proc.NatI. Acad. Sci. USA 87, 691-695.

12. Vescio, R. A., Connors, K. M., Bordin, G. M., Robb, J. A.,Youngkin, T., Umbreit, J. N. & Hoffman, R. M. (1990) CancerRes. 50, 6095-6099.

13. Leighton, J. (1951) J. Natl. Cancer Inst. 12, 545-561.14. Haugland, R. P. (1989) in Handbook ofFluorescent Probes and

Research Chemicals (Molecular Probes, Eugene, OR), pp.105-128.

15. Swisher, D. A., Prevo, M. E. & Ledger, P. W. (1988) inAlternative Methods in Toxicology, ed. Goldberg, A. M. (Leib-ert, New York), Vol. 6, pp. 265-269.

16. Draize, J. H., Woodard, G. & Calvery, H. 0. (1944) J. Phar-macol. Exp. Ther. 82, 377-390.

17. Goldberg, M. T., Tackaberry, L. E., Hardy, M. H. & Nose-worthy, J. H. (1990) Arch. Toxicol. 64, 116-121.

1912 Medical Sciences: Li et al.