letter

670 nature genetics • volume 32 • december 2002

Fig. 1 Lentiviral vector–mediatedgene transfer of type VII colla-gen to RDEB cells. a, The mini-mal lentiviral expression vectorfor human type VII collagen.Expression of the human type VIIcollagen cDNA was driven by amodified murine leukemiavirus–derived promoter (MND),designed to circumvent theproblem of viral promotermethylation. The minimallentiviral transfer vector alsocontained (i) an HIV central poly-purine tract (cPPT) and a centraltermination sequence (CTS), (ii)a self-inactivating (SIN) deletionin the U3 region of the 3′ LTR(∆U3), (iii) an SV40-derivedpolyadenylation enhancer ele-ment (USE3) inserted into the 3′LTR U3 and (iv) a reduction ofHIV-derived sequences to only774 nt. b, Immunoblot analysisof proteins from conditionedmedia of normal human ker-atinocytes (lane 1), parentalRDEB keratinocytes (lane 2) andRDEB keratinocytes transducedwith the lentiviral expressionvector for type VII collagen (lane3). The position of the band corresponding to full-length 290-kD type VII collagen (C7) is indicated. Matrix metalloproteinase 2 (MMP2) was included as an internalcontrol of protein quality and loading. c, Immunoblot analysis of proteins from conditioned media of normal human fibroblasts (lane 1), parental RDEB fibroblasts(lane 2) and RDEB fibroblasts transduced with the lentiviral expression vector for type VII collagen (lane 3). d, Immunohistochemistry with an affinity-purified poly-clonal antibody to the NC1 domain of type VII collagen. We achieved a very high efficiency of type VII collagen gene transfer to RDEB cells. RDEB/KC, parental RDEBkeratinocytes; RDEB/FB, parental RDEB fibroblasts; RDEB/KC/C7, RDEB keratinocytes transduced with the lentiviral expression vector for type VII collagen;RDEB/FB/C7, RDEB fibroblasts transduced with lentiviral expression vector for type VII collagen. e, Sustained expression of type VII collagen after lentiviralvector–mediated transduction of RDEB cells. Immunoblot analysis of proteins from conditioned media of RDEB keratinocytes (KC) and fibroblasts (FB) transducedwith the lentiviral vector containing type VII collagen after 1 wk, 1 mo, 2 mo, 4 mo and 5 mo (lanes 2–6, respectively) and of RDEB cells transduced with the controllentiviral vector expressing green fluorescent protein (lane 1). The position of the band corresponding to full-length type VII collagen (C7; 290 kD) is indicated.

Restoration of type VII collagen expression and functionin dystrophic epidermolysis bullosa

Mei Chen1, Noriyuki Kasahara2, Douglas R. Keene3, Lawrence Chan4, Warren K. Hoeffler5, Deborah Finlay5,Maria Barcova2, Paula M. Cannon2, Constance Mazurek2 & David T. Woodley1

1Department of Medicine, Division of Dermatology & 2Departments of Pathology and Biochemistry, University of Southern California, CRL 204, 1303 MissionRoad, Los Angeles, California 90033, USA. 3Shriners Hospital for Children, Portland, Oregon, USA. 4Department of Dermatology, Northwestern University,Chicago, Illinois, USA. 5Xgene Corporation, San Carlos, California, USA. Correspondence should be addressed to M.C. (e-mail: chenm@hsc.usc.edu).

Dystrophic epidermolysis bullosa (DEB) is a family of inher-ited mechano-bullous disorders caused by mutations in thehuman type VII collagen gene (COL7A1). Individuals with DEBlack type VII collagen and anchoring fibrils, structures thatattach epidermis and dermis. The current lack of treatmentfor DEB is an impetus to develop gene therapy strategies thatefficiently transfer and stably express genes delivered to skincells in vivo. In this study, we delivered and expressed full-length type VII collagen using a self-inactivating minimallentivirus-based vector. Transduction of lentiviral vectorscontaining the COL7A1 transgene into recessive DEB (RDEB)

keratinocytes and fibroblasts (in which type VII collagen wasabsent) resulted in persistent synthesis and secretion of typeVII collagen. Unlike RDEB parent cells, the gene-correctedcells had normal morphology, proliferative potential, matrixattachment and motility. We used these gene-corrected cellsto regenerate human skin on immune-deficient mice. Humanskin regenerated by gene-corrected RDEB cells had restoredexpression of type VII collagen and formation of anchoringfibrils at the dermal–epidermal junction in vivo. These stud-ies demonstrate that it is possible to restore type VII collagengene expression in RDEB skin in vivo.

Published online 11 November 2002; doi:10.1038/ng1041

R/U5CMV cPPT/CTS

minimal lentiviral transfer vector SMPU-C7

MND+ R/U5

U3 SIN

USE35 ́LTR 3´ LTR

SD/ gag

C7

205

C7 C7

kD 1 2 3 1 2 3

MMP2

200290

1 2 43 5 61 2 43 5 6kD

KC FB

C7

a cb

d

e

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

letter

nature genetics • volume 32 • december 2002 671

Fig. 3 Expression of type VII collagen reverted RDEB hypermotility. In the colloidal gold migration assay, cover slips were coated with colloidal gold, and ker-atinocytes (a) or fibroblasts (b) were plated on type I collagen (15 µg ml–1) and incubated for 18 h. The top panels show representative fields photographed at×40 under dark field optics. IKC, immortalized normal human keratinocytes; NHF, normal human fibroblasts; RDEB, RDEB keratinocytes or fibroblasts; RDEB/C7,RDEB keratinocytes or fibroblasts transduced with the lentiviral expression vector for full-length type VII collagen. The bottom panels are computer-generatedmigration indices (MIs). The migration index is the percentage of the total field area occupied by migration tracks. Error bars represent s.e. of three differentexperiments.

Individuals with DEB have defects in COL7A1, the gene thatencodes type VII collagen, which is the major component ofanchoring fibrils at the basement membrane zone (BMZ)1–3. Thefull-length cDNA sequence for type VII collagen contains 8,833nucleotides encoding 2,944 amino acids4. Over 100 distinctmutations in COL7A1 have been identified in individuals withDEB5. The most severe form of DEB is RDEB, characterized bymutilating scarring, joint contractures, strictures of the esopha-gus and aggressive squamous cell carcinomas. The BMZ of indi-viduals with DEB is characterized by a paucity or diminutive sizeof anchoring fibrils6.

Retroviral vectors are the most frequently utilized vehicles inhuman gene therapy trials7, but because they have insert-sizelimitations of 7–8 kb, they cannot accommodate the 9-kb typeVII collagen cDNA8. The recent development of replication-defective and multiply attenuated lentiviral vectors holdspromise for gene transfer in skin diseases owing to (i) their abil-ity to infect and integrate into both dividing and non-dividingcells both in vitro and in vivo9–11, (ii) their high transduction effi-

ciency, (iii) their long-term, sustained expression of the trans-gene, (iv) the absence of sequences encoding viral proteins thatmay evoke an immune response and (v) their maximum cloningcapacity of more than 9 kb, which can accommodate most trans-genes including full-length type VII collagen cDNA12.

DEB is an incurable, potentially fatal skin disease that mayrespond to gene therapy. In an effort to achieve efficient, sus-tained, corrective gene expression in DEB skin cells, we devel-oped a minimal lentiviral transfer vector to express thefull-length type VII collagen cDNA (Fig. 1a). This minimallentivirus vector was used to transduce immortalized ker-atinocytes and fibroblasts from two individuals with RDEB. Theindividuals in this study lacked expression of type VII collagenbut expressed other BMZ proteins including BP230, BP180, typeIV collagen and laminin 5 (data not shown). Western-blot analy-sis of the secreted proteins detected the expression of type VIIcollagen of 290 kD in conditioned media from normal ker-atinocytes and fibroblasts (Fig. 1b,c). This band was completelyabsent in RDEB keratinocytes and fibroblasts (Fig. 1b,c). Trans-

NHF RDEB/FB RDEB/FB/C7

NHK RDEB/KC RDEB/KC/C7Fig. 2 Expression of type VII colla-gen reversed RDEB morphology.Photographs of monolayer cul-tures were taken using phase-contrast microscopy. NHK,normal human keratinocytes;RDEB/KC, parental RDEB ker-atinocytes; RDEB/KC/C7, RDEBkeratinocytes transduced withthe lentiviral expression vectorfor type VII collagen; NHF, nor-mal human dermal fibroblasts;RDEB/FB, parental RDEB fibrob-lasts; RDEB/FB/C7, RDEB fibrob-lasts transduced with thelentiviral expression vector fortype VII collagen.

NHF RDEB RDEB/C7

mig

ratio

n in

dex

(%)

0

5

10

15

20

25

30

35

40

NHF RDEB RDEB/C7

IKC RDEB RDEB/C7

mig

ratio

n in

dex

(%)

05

10152025303540

IKC RDEB RDEB/C7

a b

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

letter

672 nature genetics • volume 32 • december 2002

duction with the lentiviral vector containing full-length type VIIcollagen (SMPU-C7) resulted in expression of the 290-kD proteinin both RDEB keratinocytes (Fig. 1b) and fibroblasts (Fig. 1c).Gene transfer done at multiplicities of infection (MOI) of 25achieved >95% efficiency in unselected RDEB cells with noimmunostaining detected in the parent RDEB cells (Fig. 1d). Theexpression of type VII collagen was sustained for at least 5 monthsin vitro in the gene-corrected RDEB cells (Fig. 1e).

RDEB keratinocytes and fibroblasts are morphologically dif-ferent from normal keratinocytes and fibroblasts13. RDEB cellswere polymorphic, elongated and enlarged compared with nor-mal human keratinocytes and fibroblasts (Fig. 2). Expression oftype VII collagen in RDEB keratinocytes and fibroblasts causedRDEB cells to return to normal morphology.

To determine if type VII collagen gene transfer exerts any func-tional effects on RDEB cells, we evaluated the cells’ migrationcapacity using migration assays on a collagen matrix. Comparedwith RDEB keratinocytes and fibroblasts, which had enhancedmotility (migration indices (MIs) 36.9 and 33.6), gene-correctedRDEB keratinocytes and fibroblasts had lower motility (MIs 22.5and 22), similar to that of normal cells (MIs 21.9 and 21; Fig. 3).

Compared with normal human keratinocytes (Fig. 4a) andnormal human fibroblasts (Fig. 4b), RDEB cells had lowerattachment to fibronectin and collagens type I, IV and VII. Incontrast, gene-corrected RDEB keratinocytes and fibroblastsattached to the matrices at levels identical to those observed innormal cells. Proliferation potentials of RDEB keratinocytes(Fig. 5a) and fibroblasts (Fig. 5b) were also restored to normallevels after gene correction.

After restoring type VII collagen expression to RDEB cells, wewished to determine if such gene-corrected RDEB cells couldrestore type VII collagen and anchoring fibrils in vivo. To do this,we reconstituted human skin under in vivo conditions by infus-ing a mixed-cell slurry of keratinocytes and dermal fibroblastsinto an inert silicone chamber grafted onto SCID mice14. Weused gene-corrected and parental RDEB cells and normal controlcells to regenerate human skin tissue on the mice. Immunostain-ing of the regenerated human skin with monoclonal antibodyLH 7.2 (which only recognizes human type VII collagen) showedthat skin equivalents composed of gene-corrected RDEB ker-atinocytes and fibroblasts expressed type VII collagen at the BMZlocation, identical to normal control cells (Fig. 6b,c). In contrast,as expected, the parental RDEB cells regenerated human skinthat entirely lacked type VII collagen expression (Fig. 6a). Weconfirmed the human-specific origin and the expression ofanother BMZ component by immunostaining with antibodiesspecific for human laminin 5 (Fig. 6e–g).

Keratinocytes are believed to be the main source of type VIIcollagen in vivo2. To examine whether gene-corrected ker-atinocytes alone could produce type VII collagen at the BMZ, weconstructed skin equivalents composed of gene-corrected RDEBkeratinocytes and parental RDEB fibroblasts. Type VII collagenstaining was detected at the BMZ as a tight linear band (Fig. 6d),similar to the staining pattern observed for laminin 5 (Fig. 6h).This indicates that type VII collagen derived solely from ker-atinocytes can restore type VII collagen at the BMZ.

To determine whether type VII collagen gene transfer could leadto restoration of anchoring-fibril formation at the BMZ in vivo, we

C1 C4 FIN C7 LAM10

0.2

0.4

0.6

0.8

1

A54

0

NHFRDEBRDEB/C7

0

0.2

0.4

0.6

0.8

1

1.2

C1 C4 FIN C7 LAM 1

A54

0NHKRDEBRDEB/C7

Fig. 4 Expression of type VII collagen enhanced RDEB adhesion. Keratinocytes (a) or fibroblasts (b) were plated into wells coated with laminin 1 (LAM 1),fibronectin (FIN) or collagen types I (C1), IV (C4) or VII (C7) at a concentration of 20 µg ml–1 overnight at 4 °C. Attached cells were fixed and stained, and theabsorbance at 540 nm was measured as an index of cell adhesion. NHK, normal human keratinocytes; NHF, normal human fibroblasts; RDEB, parent RDEB ker-atinocytes or fibroblasts; DRDEB/C7, RDEB keratinocytes or fibroblasts transduced with the lentiviral expression vector for full-length type VII collagen. Data rep-resent the mean ± s.d. of triplicate determinations in one representative experiment. Similar results were obtained in two other independent experiments.

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3

time (d)

num

ber

of c

ells

× 1

05

NHKRDEB/C7RDEB

1 2 3

time (d)

0

0.5

1

1.5

2

2.5

3

3.5

num

ber

of c

ells

× 1

05

NHFRDEB/C7RDEB

Fig. 5 Gene-corrected RDEB cellsshowed enhanced growth potential.Keratinocytes (a) or fibroblasts (b)were incubated for 1, 2 or 3 d. At theend of the incubation periods, cellswere trypsinized, resuspended andcounted with a hemacytometer intriplicate. The experiment wasrepeated three times with similarresults. NHK, normal human ker-atinocytes; NHF, normal humanfibroblasts; RDEB, parent RDEB ker-atinocytes or fibroblasts; RDEB/C7,RDEB cells transduced with thelentiviral expression vector for full-length type VII collagen.

a b

a b

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

letter

nature genetics • volume 32 • december 2002 673

carried out transmission electron microscopy on the skin regen-erated from gene-corrected RDEB keratinocytes and fibroblasts.The regenerated skin had an epidermis and dermis separated bya well formed BMZ (Fig. 6i). This neo-BMZ exhibitedhemidesmosomes, a laminin lucida, a laminin densa andnumerous anchoring fibrils. The anchoring fibrils were wheatstack–shaped structures of 200 nm or larger that emanateddown perpendicularly from the laminin densa into the dermis.In contrast, skin regenerated from parent RDEB cells entirelylacked anchoring fibrils (data not shown). These data indicatethat type VII collagen gene transfer can correct RDEB cells andinduce the in vivo expression of type VII collagen and anchoringfibrils at the BMZ.

The levels of type VII collagen expression in gene-correctedRDEB cells were similar to those observed in normal cells and weresustained for at least five months in vitro and two months in vivo.This contrasts with some studies using other human genes inwhich gene expression mediated by long terminal repeat (LTR)sequences in keratinocytes could not be maintained in vivo forlonger than four weeks15,16. In these systems, the loss of expres-sion of the transgene was due to promoter inactivation ratherthan death of the genetically engineered cells. One explanationfor these previously reported instances of promoter inactivationmay be silencing of LTR sequences owing to methylation or othermechanisms. In this study, we used a modified promoter derivedfrom murine leukemia virus, designated MND, to drive theexpression of type VII collagen. The MND promoter has beenengineered with three major modifications: (i) inclusion ofmyeloproliferative sarcoma virus LTR sequences, (ii) substitu-tion of the primer binding site and (iii) deletion of a negativecontrol region at the 5′ end of the LTR. These multiple modifica-tions have been shown to decrease methylation and allowincreased expression in transduced mouse embryonic stem cellsand hematopoietic stem cells17,18.

Before applying this gene therapy to humans, it is necessary toevaluate whether the transgene would continue to be expressedin vivo. In this regard, there may be an advantage to working withcollagen as a transgene product, because collagens have slowturnover times and are long-lived molecules2. Applying genetherapy that is highly active for a sufficient period of time to forma considerable quantity of anchoring fibrils could cause thera-peutic improvement in the skin of individuals with RDEB. Evenif expression of the transgene was later silenced, these stablestructures would persist, perhaps for years.

For gene therapy of skin, which is a renewable tissue undergo-ing constant turnover, gene targeting to keratinocyte stem cellswill be required. Because stem cells are postulated to constituteabout 5% of keratinocytes isolated in vitro, we believe that ourhigh-efficiency lentiviral vector (>95% transduction efficiency)can also successfully target stem cells. Lentiviral vectors offerimproved stem cell transduction because they do not require tar-get cell replication. In this regard, a recent study has shown thatlentiviral vectors are capable of introducing genes into ker-atinocyte stem cells in vitro10. Using primary RDEB ker-atinocytes, we found that the high transduction efficiency bylentiviral vectors was the same in both primary and immortal-ized cells (unpublished observation).

The data presented in this paper may provide a basis for thera-peutic gene transfer in human DEB, and offers a rational andgenerally applicable approach for developing future ex vivogenetic therapies to correct dermatological disorders associatedwith defects in large structural proteins.

MethodsCell culture. We cultured the human embryonic kidney cell line 293T inDulbecco’s modified essential medium (DMEM) supplemented with 10%fetal bovine serum and primary and immortalized RDEB keratinocytes(gift from S. Herron, Stanford University) in low calcium, serum-free ker-

200 µm

BMZ

E

D

AFLD

LL

AF

AF

AF

HD

200 nm

Fig. 6 Restoration of type VII collagen and anchoring fibrils in reconstitutedDEB skin in vivo. Immunofluorescence staining of reconstituted DEB and con-trol skin grafts was done with antibodies specific for human type VII collagen(a–d) and laminin 5 (e–h). a,e, Tissue reconstituted from RDEB keratinocytesand RDEB fibroblasts without type VII collagen gene transfer. b,f, Tissuereconstituted from RDEB keratinocytes and RDEB fibroblasts after type VIIcollagen gene transfer in vitro. c,g, Normal control tissue reconstituted fromnormal human epidermal keratinocytes and normal human dermal fibrob-lasts. d,h, Tissue reconstituted from gene-corrected RDEB keratinocytes com-bined with parental RDEB fibroblasts. Scale bar, 200 µm. i, Transmissionelectron microscopy was done on skin regenerated from gene-correctedRDEB keratinocytes and fibroblasts. Restored anchoring fibrils (AF) measur-ing 200 nm or more are labeled with arrows in the BMZ. The BMZ also had awell formed lamina lucida (LL), lamina densa (LD) and hemidesmosome (HD).Scale bar, 200 nm.

a b c d

e f g h

i

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

atinocyte growth medium supplemented with bovine pituitary extract andepidermal growth factor (SFM; GIBCO), as previously described19 andmodified20. We immortalized RDEB keratinocytes using the E6 and E7genes of human papillomavirus type 6 as described21. Primary RDEBfibroblasts (gift from S. Herron, Stanford University) were cultured inDMEM/Ham’s F12 (1:1) supplemented with 10% fetal bovine serum.

Individuals affected with RDEB. We established clinical, immunohisto-logical and ultrastructural criteria to define RDEB diagnosis in affectedindividuals. We selected individuals with RDEB from the National Epider-molysis Bullosa Registry site of Stanford University.

Expression vector construction. The lentivirus-based minimal gene trans-fer vector SMPU (M.B. and P.M. Cannon, manuscript in preparation; seealso ref. 22) contained (i) a modified 5′ LTR sequence in which the U3 regionof the HIV-1 LTR is replaced by the cytomegalovirus promoter, (ii) a splicedonor sequence that has been inactivated by two point mutations combinedwith a minimal packaging signal that extends only 39 bp into the gag codingsequence23, (iii) the central polypurine tract and central termination motif24

and (iv) a self-inactivating (SIN) 3′ LTR25 in which the majority of the U3region has been replaced with USE3, a polyadenylation enhancer sequencederived from the SV40 late polyadenylation signal (Fig. 1a). HIV-derived cis-acting sequences required for nuclear export and viral genome packagingwere substantially reduced (from 1–2 kb of the wildtype HIV genome pre-sent in standard third-generation lentiviral vectors25 to only 774 nucleotidespresent in the minimal vector), thereby improving the safety profile byreducing overlap homology with HIV sequences present in the packagingconstruct and allowing accommodation of larger cDNA inserts (up to 9 kb).Just downstream of the packaging signal, the vector also contained a modi-fied murine leukemia virus LTR–based internal promoter, designated MND(gift from D. Kohn, Children’s Hospital of Los Angeles). This promoter con-tained altered methylation target sites and negative control elements withinthe promoter that have been shown to be resistant to silencing in mouse stemcells in culture and in vivo17,18.

Lentiviral vector production and gene transfer. We subcloned a 8,850-bpfull-length type VII collagen cDNA (excluding its polyadenylation signaland 3′ untranslated sequences) into the polylinker sequence just down-stream of the MND internal promoter in the SMPU minimal lentiviraltransfer vector construct described above. The resultant transfer vector wasdesignated SMPU-C7. We produced lentivirus vector preparations bythree-plasmid transient co-transfection of human 293T cells as describedpreviously11. Typical titers for the minimal lentiviral vectors were between1 × 106 and 5 × 106 transducing units per ml (TU ml–1) without concentra-tion, and between 5 × 107 to 5 × 108 TU ml–1 after ultracentrifugation toconcentrate the virus vectors, using a green fluorescent protein markergene and measured directly by fluorescence-activated cell sorting. Forlentiviral infection, keratinocyte or fibroblast cultures were trypsinized,seeded onto 60-mm plates and incubated for 24 h. We supplemented theviral supernatant (MOI 25) with 8 µg ml–1 polybrene, added it to culturesthat were 30–40% confluent and then incubated the plates at 37 °C. After 6 h,the viral supernatant was removed, and the cells were washed twice. Cul-tures were then incubated with SFM (keratinocytes) or DMEM/F-12(fibroblasts). Approximately 72 h after transduction, we stained the cellswith antibody to type VII collagen as described below. Untransduced cellswere used as controls for background staining. We calculated gene-transferefficiency as the total number of cells positive for type VII collagen (asdetermined by immunofluorescence staining using an affinity-purifiedpolyclonal antibody to type VII collagen) divided by the total number ofcells per field (as determined by phase-contrast microscopy). The actualtransduction titers of the SMPU-C7 vector on RDEB cells, as determinedby immunofluorescence detection of type VII collagen expression, were3.5 ± 0.4 × 105 TU ml–1 for unconcentrated vector preparations and 7.5 ±0.8 × 107 TU ml–1 for concentrated vector preparations.

Protein analysis. For immunoblot analysis, we grew lentivirally transducedfibroblasts to confluency, changed the medium to serum-free medium con-taining 150 µm ascorbic acid and then maintained the cultures for an addi-tional 24 h. The media were collected, equilibrated to 5 mM EDTA, 50 µM N-ethylmaleimide and 50 µM phenylmethylsulfonyl fluoride, concentrated 10-to 15-fold (Centricon-100, Amicon) and subjected to 6% SDS–PAGE. We

detected the presence of type VII collagen by western-blot analysis using apolyclonal antibody to the NC1 domain of type VII collagen.

Cell migration assay. We measured keratinocyte and fibroblast migration bythe method of Albrecht-Buehler26 as modified by Woodley et al.27. Confirma-tion of a difference in migration as statistically significant required rejectionof the null hypothesis of no difference between mean migration indicesobtained from replicate sets at the P = 0.05 level with a Student’s t test.

Cell adhesion assay. We evaluated cell attachment to ECM proteins as pre-viously described28.

Grafting the cell-sorted skin equivalent (CeSSE) onto SCID mice. Wereconstituted full-thickness human skin on SCID mice by CeSSE as pre-viously described14, with all studies conducted following protocolsapproved by the University of Southern California Institutional AnimalCare and Use Committee.

Immunofluorescence staining. We carried out immunofluorescencestaining of lentivirally transduced RDEB cells as previously described29.

Immunofluorescence microscopy. We cut 5-µm sections from tissuesembedded in optimal cutting temperature compound in a cryostat, fixedthem for 5 min in cold acetone and then air-dried them. CeSSE sectionswere incubated with a monoclonal antibody against human type VII colla-gen (clone LH 7.2; Sigma) and then with a goat antibody against mouseIgG conjugated with fluorescein isothiocyanate. Working dilutions were1:10 for the primary antibody and 1:50 for the secondary antibody. Weimmunolabeled tissue using standard immunofluorescence methods asdescribed previously30.

AcknowledgmentsThis work was supported by grants from the US National Institutes of Health toM.C. and D.T.W. M.C. was supported by a Dermatology Foundation CareerDevelopment Award and a Dermatology Foundation Research Grant. C.M. andN.K. are also funded in part by a grant from the US National Institutes ofHealth through the Molecular Biology Core/Virus Vector Subcore of theUniversity of Southern California Research Center for Liver Diseases.

Competing interests statementThe authors declare that they have no competing financial interests.

Received 5 September; accepted 11 October 2002.

1. Uitto, J. & Christiano, A.M. Molecular basis for the dystrophic forms ofepidermolysis bullosa: mutations in the type VII collagen gene. Arch. Dermatol.Res. 287, 16–22 (1994).

2. Burgeson, R.E. Type VII collagen, anchoring fibrils, and epidermolysis bullosa. J.Invest. Dermatol. 101, 252–255 (1993).

3. Sakai, L.Y., Keene, D.R., Morris, N.P. & Burgeson, R.E. Type VII collagen is a majorstructural component of anchoring fibrils. J. Cell Biol. 103, 1577–1586 (1986).

4. Christiano, A.M., Greenspan, D.S., Lee, S. & Uitto, J. Cloning of human type VIIcollagen: complete primary sequence of the α1(VIII) chain and identification ofintragenic polymorphisms. J. Biol. Chem. 26, 20256–20262 (1994).

5. Christiano, A.M. & Uitto, J. Impact of molecular genetic diagnosis on dystrophicepidermolysis bullosa. Curr. Opin. Dermatol. 3, 225–232 (1996).

6. Briggaman, R.A. & Wheeler, C.E. Jr. The epidermal–dermal junction. J. Invest.Dermatol. 65, 71–84 (1975).

7. Morgan, R.A. & Anderson, W.F. Human gene therapy. Ann. Rev. Biochem. 62,191–217 (1993).

8. Anderson, W.F. Human gene therapy. Nature 392, 25–30 (1998).9. Naldini, L., Blomer, U., Gage, F.H., Trono, D. & Verma, I.M. Efficient transfer,

integration, and sustained long-term expression of the transgene in adult ratbrains injected with a lentiviral vector. Proc. Natl Acad. Sci. USA 93, 11382–11388(1996).

10. Kuhn, U., Terunuma, A., Pfutzner, W., Foster, R.A. & Vogel, C. In vivo assessmentof gene delivery to keratinocytes by lentiviral vectors. J. Virol. 76, 1496–1504(2002).

11. Sakoda, T., Kasahara, N., Hamamori, Y. & Kedes, L. A high-titer lentiviralproduction system mediates efficient transduction of differentiated cellsincluding beating cardiac myocytes. J. Mol. Cell Cardiol. 31, 2037–2047 (1999).

12. Naldini, L. Lentivirus as gene transfer agents for delivery to non-dividing cells.Curr. Opin. Biotechnol. 9, 457–463 (1998).

13. Chen, M. et al. Development and characterization of a recombinant truncatedtype VII collagen “minigene”: implication for gene therapy of dystrophicepidermolysis bullosa. J. Biol. Chem. 275, 24429–24435 (2000).

14. Wang, C., Nelson, C.F., Brinkman, A.M., Miller, A.C. & Hoeffler, W.K. Spontaneouscell sorting of fibroblasts and keratinocytes creates an organotypic human sinequivalent. J. Invest. Dermatol. 114, 674–680 (2000).

letter

674 nature genetics • volume 32 • december 2002

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

letter

nature genetics • volume 32 • december 2002 675

15. Fenjves, E.S., Yao, S.N., Kurachi, K. & Taichman, L.B. Loss of expression of aretrovirus-transduced gene in human keratinocytes. J. Invest. Dermatol. 106,576–581 (1996).

16. Choate, K. & Khavari, P.A. Sustainability of keratinocyte gene transfer and cellsurvival in vivo. Hum. Gene Ther. 8, 895–901 (1997).

17. Robbins, P.B. et al. Increased probability of expression from modified retroviralvectors in embryonal stem cells and embryonal carcinoma cells. J. Virol. 71,9466–9474 (1997).

18. Robbins, P.B. et al. Consistent, persistent expression from modified retroviralvectors in murine hematopoietic stem cells. Proc. Natl Acad. Sci. USA 95,10182–10187 (1998).

19. Boyce, S.T. & Ham, R.G. Calcium-regulated differentiation of normal humanepidermal keratinocytes in chemically defined clonal culture and serum-freeserial culture. J. Invest. Dermatol. 81 Suppl., 33S–44S (1983).

20. O’Keefe, E.J. & Chiu, M.L. Stimulation of thymidine incorporation inkeratinocytes by insulin, epidermal growth factor, and placental extract:comparison with cell number to assess growth. J. Invest. Dermatol. 90, 2–7 (1988).

21. Halbert, C.L., Demers, G.W. & Galloway, D.A. The E6 and E7 genes of humanpapillomavirus type 6 have weak immortalizing activity in human epithelial cells.J. Virol. 66, 2125–2134 (1992).

22. Sastry, L., Johnson, T., Hobson, M.J., Smucker, B. & Cornetta, K. Titering lentiviralvectors: comparison of DNA, RNA and marker expression methods. Gene Ther. 9,1155–1162 (2002).

23. Cui, Y., Iwakuma, T. & Chang, L.J. Contributions of viral splice sites and cis-regulatory elements to lentivirus vector function. J. Virol. 73, 6171–6176 (1999).

24. Sirven, A. et al. The human immunodeficiency virus type-1 central DNA flap is acrucial determinant for lentiviral vector nuclear import and gene transduction ofhuman hematopoietic stem cells. Blood 96, 4103–4110 (2000).

25. Zufferey, R. et al. Self-inactivating lentivirus vector for safe and efficient in vivogene delivery. J. Virol. 72, 9873–9880 (1998).

26. Albrecht-Buehler, G. The phagokinetic tracks of 3T3 cells. Cell 11, 395–404 (1977).27. Woodley, D.T., Bachmann, P.M. & O’Keefe, E.J. Laminin inhibits human

keratinocyte migration. J. Cell Physiol. 136, 140–146 (1988).28. Chen, M., O’Toole, E.A., Li, Y.-Y. & Woodley, D.T. α2β1 integrin mediates dermal

fibroblast attachment to type VII collagen via a 158-amino-acid segment of theNC1 domain. Exp. Cell Res. 249, 231–239 (1999).

29. Chen, M., Costa, F.K., Lindvay, C.R., Han, Y.P. & Woodley, D.T. The recombinantexpression of full-length type VII collagen and characterization of molecularmechanisms underlying dystrophic epidermolysis bullosa. J. Biol. Chem. 277,2118–2124 (2002).

30. Gammon, W.R., Briggaman, R.A., Inman, A.Q. III, Queen, L.L. & Wheeler, C.E.Differentiating anti-lamina lucida and anti-sublamina densa anti-BMZ antibodiesby indirect immunofluorescence on 1.0 M sodium chloride-separated skin. J.Invest. Dermatol. 82, 139–144 (1984).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics