Journal of NeurochemistryRaven Press, Ltd . . New York((, 1995 International Society for Neurochernistry

The 3' Flanking Region of the Human Tyrosine HydroxylaseGene Directs Reporter Gene Expression in Peripheral

Neuroendocrine Tissues

Shou C . Wong, Mark A. Moffat, George T. Coker, *John P. Merlie, and Karen L. O'Malley

Departments of Anatomy and Neurobiology and *Pharmacology, Washington University Medical School,St. Louis, Missouri, U.S.A .

Abstract : Cell type-specific expression of the catechola-mine synthetic enzyme, tyrosine hydroxylase (TH), ap-pears to be mediated in part by cis-acting elements lo-cated at the 3' end of the human gene . Further delinea-tion of this region indicated sequences corresponding toa CACGTG motif significantly stimulated transcription ofa heterologous promoter in various cell types . Mutationof this site led to a complete loss of activity. DNase foot-printing, gel retardation, and UV cross-linking experi-ments indicated that a 74-kDa cellular factor(s) boundspecifically to the CACGTG motif in the pheochromocy-toma cell line PC12 . The size of this protein and its patternof expression are compatible with those of the CACGTGbinding protein TFE3 . Transgenic animals were createdusing a 261-bp human TH 3' fragment encompassing theCACGTG motif in front of a thymidine kinase promoter/chloramphenicol acetyltransferase reporter gene . In threelines of mice this fragment was sufficient to direct a pat-tern of mRNA expression in peripheral neuroendocrinetissues that mimicked TH mRNA distribution . However,these sequences were not sufficient for CNS-specificpatterns of expression . Thus, multiple cell type-specificenhancers may regulate TH gene expression in theCNS and periphery . Key Words : Cell type specificity-Helix-loop-helix leucine zipper-Tyrosine hydroxy-lase-Transgenic mice-Gene regulation-3' enhancer .J . Neurochem. 65, 23-31 (1995) .

Tyrosine hydroxylase (TH) is the rate-limiting en-zyme in the biosynthesis of catecholamines . The ex-pression of the TH gene is restricted to dopaminergic,noradrenergic, and adrenergic cells in the CNS as wellas in sympathetic ganglia and adrenal medullary cellsin the periphery . Such a restricted tissue distributionsuggests the possible involvement of cis-acting se-quences whose association with positive or negativeregulatory factors results in the tissue-specific expres-sion of the TH gene . Little is known about the molecu-lar mechanisms underlying either the CNS or periph-eral expression of this gene . However, several recentstudies have begun to focus on identifying tissue-spe-

23

cific elements and transcriptional factors responsiblefor the observed distribution . For example, we andothers (Cambi et al ., 1989 ; Gandelman et al ., 1990 ;Yoon and Chikaraishi, 1992 ; Wong et al ., 1994) havefound elements within the 5' promoter region of therat TH gene that appear to confer tissue-specific ex-pression in transient transfection paradigms .

In contrast, the human TH (hTH) 5' region failedto exhibit a cell type-specific response (Gandelman etal ., 1990) . Because other cell type-specific enhancershave been localized downstream from their transcrip-tional start sites, e.g ., immunoglobulins (Dariavach etal ., 1991 ; Pongubala and Atchison, 1991), the globingene (Purucker et al ., 1990), histone genes (Zhao etal ., 1990), the myosin light chain gene (Rosenthal etal ., 1990), and the hypoxia gene (Beck et al ., 1991),we tested fragments spanning the human TH loci fortheir efficacy in conferring "tissue specificity" ona heterologous promoter . Previously, we reported thata 759-bp region immediately 3' of exon 13 is cap-able of activating a heterologous promoter in an orien-tation-independent and neuroendocrine-specific man-ner (Gandelman et al ., 1990) . Further truncation ofthese sequences suggested an E box motif (CACGTG)---230 by downstream of exon 13 was important forthe in vitro response (Gandelman et al ., 1990) . Subse-quently, Kaneda et al . (1991 ) reported that constructscontaining 2,500 by of human TH 5' sequences as

Received October 17, 1994; revised manuscript received Decem-ber 23, 1994 ; accepted January 12, 1995 .

Address correspondence and reprint requests to Dr. K . L .O'Malley at Box 8108, Department of Anatomy and Neurobiology,Washington University Medical School, 660 South Euclid Avenue,St . Louis, MO 63110, U.S.A .The present address of Dr. S . C . Wong is Cytotherapeutics, Inc .,

4 Richmond Square, Providence, RI 02906, U.S .A .The present address of Dr . G . T . Coker is Gliatech . Inc ., 23420

Commerce Park Road, Cleveland, OH 44122, U.S .A .Abbreviations used: CAT, chloramphenicol acetyltransferase ;

hINS, human insulin promoter region ; HLH, helix-loop-helix ;PCR, polymerise chain reaction ; TH, tyrosine hydroxylase ; TK,thymidine kinase .

24

well as the primary transcript and 3'-flanking (500-bp) sequences were sufficient for tissue-specific ex-pression in transgenic animals . In contrast, when 5 .0,2.5, and 0.2 kb of the 5'-flanking promoter region ofthe human TH gene were tested in transgenic mice, nocell type-specific gene expression was observed (Sa-saoka et al ., 1992) . These results suggested that se-quences downstream of the human TH promoter in-cluding 3'-untranslated sequences are important inachieving high-level, cell type-specific gene expressionin vivo . Because the full-length tissue-specific con-struct described by Kaneda et al . (1991 ) encompassedthe 3' domain we had previously defined, the in vivoresults are consistent with our transient assays . Con-versely, it is conceivable that the tissue culture para-digm used to define human TH cell type-specific ele-ments does not accurately reflect endogenous behavior,or that a combination of both 5' and 3' elements willbe important for accurate transcriptional regulation . Totest these hypotheses, we have further characterizedthe 3' elements in the human TH gene by both func-tional assays and in vivo paradigms. In this report wepresent evidence that the 3' enhancer region of thehuman TH gene contains sufficient regulatory informa-tion to direct an appropriate pattern of expressionin peripheral neuroendocrine cell types but not inthe CNS.

MATERIALS AND METHODS

Plasmid constructsThe vector pTK81-Luc vector obtained from Dr . S . K.

Nordeen (University of Colorado) contains the herpes sim-plex virus type I thymidine kinase (TK) promoter in frontof the firefly luciferase reporter gene . The human 3' cle-ment-TK-luciferase clone J was constructed as previouslydescribed (Gandelman et al ., 1990) . Mutations within the Ebox were generated by the recombinant polymerasc chainreaction (PCR) method (Higuchi, 1990) using internal oli-gonucleotides 634/635. The human insulin promoter region(hINS) encompassing nucleotides -365 to --36 (Bell et al .,1982) was also cloned by PCR using the oligonucleotides393/395. All PCR constructs were sequenced to verily theaccuracy of the amplification process. To create the ,I-TK-chloramphenicol acetyltransferase (CAT) plasmid, the anal-ogous luciferase construct was digested to completion byBantHI and HindIll, filled in, size-selected in low-tneltinoagarose, and ligated into ptkCAT-342 . Because the TK pro-moter in ptk-CAT-342 contains the promoter sequence -- 108to +50, we truncated it to -81 to be consistent with theminimal promoter used in the previous studies (Gandelmanet al ., 1990) . The final plasmid was named ptk-CAT-.I-380 .The human TH 3' flanking sequence was previously Submit-ted to GcnBank under accession no . M93281 .

Cell culturesThe rat pheochromocytoma cell line PC 12 and the human

hepatotna cell line HepG2 were cultured as described (Gan-delman et al ., 1990) . 8TC-1 cells were derived froth theislets of transgenic animals expressing SV40 large T antigenunder the control of the insulin gene pronwter (Efrat et al .,1988) . These cells express large amounts of insulin and

J- Neuroc'dAem., Vol. 65, No . /, /995

S. C. WONG 117 AL .

were obtained from Dr . R . Stein (Vanderhilt University ) ./3TC-I cells were 11rown in Dulhecco', Modified Ea~cle'smedium supplemented with 2.5Sc fetal call scrum and 10'/,horse scrum_

Transient transfectionAll cell lines were translected u,in`a the calcium phosphate

method (Chen tuid ()kayama, 1987) . PC 12 ( I x 10 ( ') andßTC-I ( 2 x 10' ) cells were tran,fected with 3 pgo of plasmidDNA, whereas HepG2 cell, ( 3 x 10' ) were transfected with2 Fig of pl,ismid DNA. All cells were incubated with thecalcium phosphate DNA precipitate lit 5 h . The PC'I 2 cell,were subsequently exposed to 20(7r glycerol for I min .HepG2 cells to 10'4 glycerol for 3 inin, and fITC-I cells to20 1/c glycerol Ior 2 min . At 4S h later, the cells were har-vested, and cell Ipsates were used for the luciferase assay Ll ."

described by De Wet et al . ( 1987 ) . Lucilerase activitieswere normali/ed using RSV-/)gal a,, an internal control a,previously described (Gandelman et al ., 1990 : Won, et al .,1994) . Protein content was ntca .stn- ed by the assay of Brad-ford ( 1976) wing bovine serum alhumin as the standard .

OligonucleotidesOligcmucleotidcs were ,yn(hcsi/ed by the Protein Chemis-

try Core Facility at Washington University Medical School .For gel retardation assays, complenrenlary oligonucleotideswere synthesi/ed and then anucnlcd to frnm double-strandedoligonucleotides . Oli`aonuclccriides inclu(IC(I the 1ollowittg :338 (E box: 30-mcr) . 5'-AACi(i"IGGCiÇiAGC'CAC'GTCi-ACAGTGGGAGG,G-3' : 1104 (E box ; 20-mcr) . 5'-GGG-GAGC'CACGTCiAC'AG"fGG-3' ; 363 (mutan( E box ; 20-mcr), S'-CiCiGGACiC'AC'CiAG'fAC'A(iT(iG-3' : 360 (ca-nonical myc/USF/TFL CAC( TG motif (Blackwell et al . .1993)1, 5'-A l'AGGTCiTAGGC'('ACCi fGACCGGGTC~T-TC'C-3' ; 344 1 c,uuniic,tl MyoD C'ACC'T('j motif (Weintraut)et al ., 1991)1, 5'-GA'TC`C'CC'CCAACACCTGC'TG('CT-GA-3' : 342 (random) . 5'-AGGAAGCCGTGC'TA000Ci-TAGCC'TCTG f(1A-3' ; 441 IhTH (this study) I, 5'-CCCAAGC'I"fCTGAG(7"I"l'CCC'C`"F"hGG'fTC'GC'-3' ; 443 I h - I'H(this study)( . 5'-CC'C'GGA"f('C'CiC'f1GCTCC'AGAGGC'f-CGGGCA-3' : 393 I hINS (Bell et al . . 1982)1 . 5'-CCCA-AGCTTACAC;CAGC'GC'AAACiACK'CCC'G-3' : 395 I hINS(Bell et al ., 1982 )I, 5'-CC'C"AAÇiC"l'TCTC'AGyG000'A't-CTCCCCTAC'-3' :

634 111TH

(tlli, s(udy) I .

5'-AAGCi-1'GGGAGCTG('CICAACAG"IC'CiGAGGCi-3' ; 291 (CAI'(Alton and Vapuck, 1979)1, 5'-AGCTAAGGAAGC'hA-AAATGG-3' ;

292

I CA]

(Alton

and

Vapnck,

1979)1,5'-AC'(~ I"I"f('ACi"pT"hGC "! C' .4 hGG-3' ;

739

I myc (Kerk-hoff et al ., 1991 ) I, 5'-CiACiCiAC',~l'C' -l'CiGAACiAAl1T"I-(.'-GAGC'"fG-3' ; 741 I inyc ( KerkholI et al . . 1991 ) I . 5'-('-CGC'"f('C:ACA'I-AC'AGTCC°h(iGA"fCiA

805 1 TI h(Carr and Sharp. 1990) I . 5'-(1GAGATTCjA'I'C ;ATCi fCA'I'-TGA-3' ; 806 I TFE (Canr and Sharp. 1990)1 . 5'-GATG(i f-CC'C'C'I"l'G'TTCCAG-3' : 807 I l!SF ( Gregor et al . . 1990) 1 .5'-CTCAGCATAATGAAGF(i(i-3' : and 808 11 !SF (Grc-gor et al ., 1990)1, 5'-AGA'TC'"1 . lC . C . ACCTG'CTGT-3'.

DNA binding assayElectrophorelic mobility shill assays and UV cross-linking

were performed exactly a, dcscrihcd ( Won`_ et al ., 1994) .I)Nasel footprint analysis ~,,-as performed according to (hel'romega Core I "oothrinting Protocol (no. 137) using I0 Imolof probe and 50 /tg of (otal protein . The 299-bp probe wasgenerated by PCR using oligonucleotides 4-13 and 441 . Theresulting PCR product vas purilied front a 5'/r polyacryl-

amide gel, and the purified DNA fragment was subsequentlyradiolabeled using T4 polynucleotide kinasc ( Maniatis et al .,1982) . The sense strand probe was generated by digestingthe labeled DNA fragment with Hindlll .

mRNA analysis by reverse transcription-MRTotal RNA was isolated from PC12 and HepG2 cell lines

(Chomczynski and Sacchi, 1987), reverse-transcribed usingohgonucleotides 741, 806, and 808 for c-Myc, TFE3, andUSF, respectively (Krug and Berger, 1987), and MINIM'amplified according to the manufacturer's protocol ( l'crkinHmer Cetus, Alameda, CA, U.S .A .) . TFE oliponucleotideswere chosen such that TFE3 and TFEB would be ccruTnpli-fied . Amplification temperatures for the oligonucleotide set739/741 (c-Myc) were denaturation at 94°C for I min, an-nealing at 63°C Ior I min, and extension at 72°C for I min.Amplification temperatures for oligonucleotide sets 805/806('TFE3) and 807/808 (USF) were 94 (1 min), 46 ( I min),and 72°C ( I min) .To increase the sensitivity of the reverse transcription-

PCR experiments, the 5' oligonucleotides were radiolabeledand included in the reaction mixture at a Specific activity of5 X 10' cpm/pmol . Typically, radioactive PCR productswere elect rophoresed on 5% polyacrylamidc gels andscanned with the Phosphorlmager SF ( Molecular Dynamics,Sunnyvale, CA, U.S .A .) . To confirm the identity of ampli-fied products, PCR Fragments were excised from the gel,extracted as described by Maniatis et al . ( 1982), and cyclesequenced using the AmpliTaq Cycle Sequencing Kit (Per-kin Elrner Cetus) .

Production of transgenic miceThe 1,900-bp DNA fragment containing the J-TK-CAT

sequence was generated by K/ml and Noel digestion andsize-selected by agarose electrophoresis (Maniatis et al .,1982) . DNA was purified using the EIuQuik kit accordingto the manufacturer's instruction (Schleicher and Schnell ) .The DNA was dissolved in in buffer 110 mM Tris-HC1 (pH 7.5), 10 mM NaCI . and 0.1 mM EDTA I at aconcentration o1 I N.g/ml and dialyzed against the samehul'ler . The DNA was microinjected into a single premuclcusof 12 h (C57BL/6J X CBA/J Fl hybrid) mouse crnhryosby standard methods (Hogan et al ., 1986) . Identification ofthe integration of the ,1=TK-CAT transgene in the mousegenorne was done by PCR using unpurified DNA extractsfrom tail digests according to the method of Hanley andMerlie ( 1991 ) .

RESULTS

Our previous studies indicated tissue-specific ele-ments involved in hTH gene expression were down-stream of the last exon (Fig . I A) (Gandelman et al .,1990) . In particular, aCANNTG motif or E box withinthe maximal response region of fragment J (Gandel-man et al ., 1990) appeared to be important for elicitinghigh-level activity over basal level controls . Such se-quences constitute binding sites for a family of tran-scription factors containing helix-loop-helix (FILM)structural domains, such as Myc ( Kerkhoff et al .,1991 ), MyoD (Weintraub et al ., 1991 ), USF (Gregoret al ., 1990), and AP-4 (Hu et al ., 1990) . However,because the 3' end of the human TH gene is only 2.7kb 5' of the insulin gene (O'Malley and Rotwein,

ANALYSIS OF HUMAN TH GENE EXPRESSION 25

1988) (Fib . IA), the question arose as to whetherthe 3' elements previously identified might in fact beinvolved in the regulation of the human insulin gene .We tested this possibility by cloning the insulin pro-moter and determining whether the downstream THfragments enhanced the expression of an insulin pro-Inoter/luciferase construct in the mouse pancreatic 0-cell line /3TC-1 .

Several Studies have indicated that cell type-specificregulation of the insulin gene is also controlled by CAN-NTG-binding transcription factors clustered within 300by of the transcription start site (Ohlsson et al ., 1988 ;Cordle et al ., 1991a,h) . Therefore, we subcloned thehINS (Fig . IA ; -365 to -36) (Bell et al ., 1982) infront of the TK prornoter/lucifcrase gene construct. Theresulting construct, hINS/TK/Luc, showed a relativeluciferase activity of 47-fold in f3TC-1 cells versus 1 .3-and 2 .6-fold in HepG2 and PC 12 cells, respectively(Fig . I B ) . Therefore, as previously shown, the 5' hINSdirects pancreatic tissue-specific expression .

OTC- I cells also express TH mRNA and have en-zyme activity at approximately half the level of PC 12cells (S .C.W ., M.A.M ., and K.I ..O'M ., unpublisheddata) . This is analogous to the in vivo endocrine pan-creas, which also cocxpresses insulin and TH (Teitel-man et al ., 1981 ) . Therefore, we expected to see someexpression of the TH constructs in this cell line . Tran-sient transfection of the human TH 3' flanking regionJ fragment in the 0TC-I cells showed a 30-fold en-hanced luciferase activity (Fig . I B ) .To test whether the 3' flanking region of the human

TH gene contributed to insulin gene expression, the Jfragment was subcloned in front of the insulin/TK/luciferase construct. As shown in Fig . 113, inclusionof J did not significantly alter insulin promoter expres-sion ; J/hINS/TK/I.uc and hINS/TK/Luc constructsshowed approximately, the same relative luciferase ac-tivity in /3TC-1 . It is interesting that the juxtapositionof the J and insulin promoter sequences (J-hINS) re-duced luciferase activity in PC 12 cells by fourfold( 127% vs . 35~Y ), suggesting that although sequenceswithin the J fragment did not significantly enhanceinsulin gene expression, factors interacting with insulinenhancer elements could repress J expression .

The TH 3' E box is necessarv for functionalactivity

In our previous Study, we demonstrated that removalof another 23 nucleotides from the J fragment led toa dramatic loss of activity in PC 12 cells [ K fragment(Gandelman et al ., 1990) 1 . Because this deletion was2 by 3' to the E box noted above, it seemed possiblethat truncation of the E box flanking sequences resultedin a loss of factor recognition . To test this directly,we mutated the E box core sequence present in the Jfragment from CACGTG to TGCGCA . As shown inFig. 113, Inutagenized E box constructs resulted in acomplete loss of activity . Thus, the CACGTG DNArecognition element is necessary for cell type-specific

J. Ncmv.hcm ., Vol. 0 .5 . No. /, /995

26

expression of the TH 3' luciferase reporter gene con-struct in PC12 cells .

Nuclear extracts bind the E box motifTo assess whether the E box motif bound nuclear

factors in a cell type-specific fashion, we preparedcomplementary oligonucleotides or DNA fragmentsspanning the E box domain . When a 299-bp 3' frag-ment was incubated with nuclear extracts from eithercell line, a clear footprint over the E box motif wasobserved (Fig . 2A, lanes 2-7) . In contrast, no specificpattern of protection was observed elsewhere in thisfragment . Despite the 13-fold increased level of re-porter gene activity exhibited by the J fragment inPC 12 versus HepG2 cells, no obvious cell type-specificprotection was observed in the footprinting assay .HLH proteins that bind to DNA recognize a core

DNA sequence, CANNTG. Subfamilies of these fac-tors have been distinguished by their ability to recog-nize target sequences with specific bases at the two

J . Neurochent., Vol. 65, No. 1, 1995

S . C. WONG ET AL.

FIG . 1 . A : Schematic representation of thehuman TH and insulin (INS) loci depictsfragments used in transient transfection ex-periments . The exons are depicted by solidboxes . Introns and flanking regions are rep-resented by the thin line . The E box motif isnoted by the single solid box . J and WINSrepresent subcloned fragments used fortransient transfection experiments . B : TheTH 3' flanking region does not contributeto INS gene expression . PC12, /3TC-1, andHepG2 cells were transfected with the indi-cated TK/luciferase (Luc) fusion constructs .Units represent relative Luc activities deter-mined after defining the activity of TK/Lucas 1 .0 . Data are mean + SEM (bars) values .

central positions . The J E box sequence, CACGTG, isidentical to the recognition site for the HLH subfamilyof proteins, which include the c-Myc family (Black-well et al ., 1993), N-Myc (Ma et al ., 1993), L-Myc(Ma et al ., 1993), USF (Gregor et al ., 1990), andTFE3 and TFEB (Fisher et al ., 1991 ) . Conceivablyany one of these factors or heterodimers thereof couldbe binding to the TH 3' E box . These proteins notonly share a common core recognition sequence butalso constitute a subgroup of HLH proteins (HLH-LZ) owing to the presence of a leucine zipper domainadjacent to the HLH motif (Fisher et al ., 1991 ) .We used gel shift analysis to assess which subfamily

of HLH-LZ proteins might be binding to this site.When radiolabeled, double-stranded probes were incu-bated with nuclear extracts prepared from TH-express-ing versus -nonexpressing cell types, probes encom-passing the E box motif formed multiple protein com-plexes with extracts from both cell lines (Fig . 213, lanes2 and 9) . The major complex could be competed for

FIG . 2. DNA binding analysis of the TH 3'flanking region . A: The DNasel footprint is ob-served in the E box region . A 299-bp senseprobe encompassing the J E box was usedin DNasel footprint assays as described inMaterials and Methods . The protected regionis indicated by an arrow on the right . A sche-matic representation of the position of the Ebox is shown on the left . Positions were deter-mined by a juxtaposed sequencing ladder .Lane 1, DNA without added nuclear proteins ;lanes 2-4, DNA plus increasing amounts ofPC12 nuclear proteins ; lanes 5-7, DNA plusincreasing amounts of HepG2 nuclear pro-teins . B : Sequence specificity of E box bindingproteins . A 208-bp Styl fragment encom-passing the E box response element was usedin gel retardation assays . Nucleotide se-quences of the competitor double-strandedoligonucleotides used in 75-molar excess aredescribed in the text . The Styl fragment wasincubated with the indicated nuclear extractswith or without the indicated E box competi-tors . Oligonucleotides and their complementsused in lanes 3 and 10 were 364, lanes 4 and11 were 338, lanes 5 and 12 were 363, lanes6 and 13 were 360, lanes 7 and 14 were 344,and lanes 8 and 15 were 342 (see Materialsand Methods) .

with bona fide TH E box sequences (Fig . 213, lanes 3,4, 10, and 11 ) but not with mutated sites (lanes 5 and12) . The TH E box could also be successfully com-peted for by CACGTG core sequences with variableflanking nucleotides present in Myc, USF, or TFE rec-ognition sites (Fig . 2B, lanes 6 and 13) but not withthe closely related myoD binding sequence (CAC-CTG, Fig . 213, lanes 7 and 14) . Moreover, oligonucleo-tides corresponding to CACGTG sites [AP-4 (Hu etal ., 1990) ] could not block binding to the 3' E boxsite (data not shown) . Collectively, these data confirmand extend the footprinting results, which indicatedtranscription factors present in either cell line recog-nize the TH E box site . However, these results cannotaccount for the differences in functional activity ob-served between the two cell lines .

Identification of HLH-LZ transcripts in PC12 andHepG2 cellsThe restricted expression of HLH-LZ proteins them-

selves might contribute to the observed cell type-spe-cific response . Sequences flanking the E box mightalso influence the selectivity of binding . To determinewhether different proteins were binding to the E boxdomain, we used two types of assays-UV cross-link-ing and amplification of expressed sequences-to de-termine which of the known HLH-LZ transcripts werepresent .

ANALYSIS OF HUMAN TH GENE EXPRESSION 27

Figure 3 shows the results of the latter analysis inwhich oligonucleotides specific for c-Myc familymembers, USF or TFE3, and TFEB were used to primereverse transcription of RNA isolated from HepG2 andPC12 cells . To identify unequivocally PCR products,cell line-specific bands were isolated and cycle se-quenced . The results from this analysis indicated thatc-Myc is present in HepG2 cells but not in PC 12,whereas only B-Myc is expressed in the latter cell line .Neither N- nor L-Myc is expressed in either cell type

FIG . 3 . Identification of CACGTG binding protein transcripts inPC12 and HepG2 cells . Reverse transcription-PCR in conjunc-tion with the transcript-specific oligonucleotides indicated aboveeach panel was used to identify the CACGTG binding proteins .PCR products generated from either HepG2 or PC12 RNA wereseparated on polyacrylamide gels, and the resulting fragmentswere excised and used as a template for cycle sequencing . Theidentity of the amplified cDNAs is indicated below .

J. Nemoc'hem . . Vol. 65, No . 1, 1995

2S

(data not shown) . Similarly, TFEB is only expressedin HepG2 cells, whereas TFE3 is present in PC12 .Finally, USF transcripts are only expressed in HepG2cells . Ol'the three known families of CACGTG bindingproteins, only B-Myc and TFE3 are expressed in undif-fercntiated PC12 cells . However, whereas B-Myc isrelated to c-Myc with an overall 40% identity at theamino acid level (Resar et al ., 1993), it lacks the spe-cific DNA-binding motif found in other Myc familymembers . 'therefore, of the known irans-acting factorscapable of binding to a CACGTG motif, TFE3-likefactors are the most likely candidates mediating theVI'K-lucifcrase response .

UV cross-linking demonstrates cell type-specificproteins bind to the TH 3' E boxAs a further aid in determining whether cell line

differences existed in the factors binding to the TH 3'E box, UV cross-linking was used to estimate thesize (s) of the CACGTG binding protein(s) . PC 12 andI -ICI)(;'nuclear extracts were incubated with the 30-rner E box oligonucleotide (338 and its complement)and irradiated with UV light to link the DNA cova-lently to its binding protein(s) (Ausuhel et al ., 1992) .Analysis on sodium dodccyl sulfate gels revealed celltype-specific differences in the DNA binding patterns .In PC 12 cells, a major band with an apparent molecularmass of 74 kDa was observed (Fig . 4) . In contrast,bands of 90, 80, and 62 kDa were observed withHepG2 extracts (Fig . 4) . Because the cross-linked Ebox oligonucleotide has a molecular mass of 18 kDa,we estimate that the TH 3' E box binding proteinhas a molecular mass of -56 kDa, whereas the majorHepG2 peptide would be 43 kDa . These values aresimilar to those reported for the TFE3 159 kDa ( Beck-mann et al ., 1990) ] and USF 143 kDa (Sawadogo etal ., 1988)1 rr -crrr,s-actin, factors, respectively . Collec-

.l. Ncmvrhem ., Vol. 05 , No. / . /995

FIG. 4. UV cross-linking of E boxDNA binding protein(s) . The E box30-mer probe was UV cross-linkedto PC12 and HepG2 nuclear ex-tracts as described in Materials andMethods. The amount of UV irradia-tion energy is indicated at the topin units of J/cm' . Protein molecularweight markers are indicated to theleft .

S. C. WONG ET AL..

tively, these data demonstrate that different transcrip-tion factors do bind to the TH 3' CACGTG box andthat they do so in a cell type-specific fashion . Theseresults, together with the cycle sequencing data, sug-gest that either bona fide TFE3 or a family membermight be involved in the regulation of the J-TK-lucifer-ase construct in PC12 cells . This is in agreement withexperiments in which we found that cotransfection ofa c-Myc expression vector could not trans-activate theJ construct (data not shown) . Moreover, antibodiesagainst the c-Myc N-terminal amino acids (obtainedfrom Dr. R . Eisenman, University of Washington),failed to supershift the DNA/protein fragments shownin Fig . 2B (data not shown) .

Fragment ,1 in transgenic mice directstissue-specific expression of the reporter genein peripheral neuroendocrine tissues butnot in the CNS

The functional importance of the TH 3' E box (Gan-delman et al ., 1990 : present study) combined with thewell-documented role(s) of HLH proteins in develop-ment and differentiation prompted us to test the in vivorole of these sequences in a transgenic paradigm . Nineindependent lines of transgenic mice were generatedwith a vector containing fragment J cloned in frontof a TK promoter driving a CAT reporter gene . Thepresence of the transgene was determined by PCR oftail DNA for the CAT reporter gene as described byHanley and Merhe ( 1991 ) . Three lines of transgenicmice were analyzed for the expression of CAT mRNAin various CNS and peripheral tissues . In all three lines,CAT rnRNA was primarily expressed in adrenal glandsfollowed by pancreas ( 19% of adrenal gland expres-sion), cerebellum (15°lc ), olfactory bulb (8°Ic), lung(_5%), tnidbrain (5%), pons (3%), and heart (2°l0)(Fig . 5) . CAT mRNA could not be detected in cerebralcortex, liver, kidney, spleen, and thymus (Fig . 5) .

DISCUSSION

We have drawn four major conclusions from theresults presented above . First, sequences involved incell type-specific expression of the human TH geneare located within 237 by downstream of the last exon .Second, these sequences are sufficient to direct in vivoCAT expression in appropriate peripheral neuroendo-crine tissues but are probably not sufficient for correctCNS expression . Third, a CACGTG E box appears tomediate this effect. Fourth, the TH 3' E box does notappear to be involved in the regulation of the contigu-ous human insulin gene .The importance of the E box within the J fragment

is demonstrated by the fact that constructs truncatingthis sequence (Gandelman et al ., 1990) or with muta-tions in the hexameric recognition site exhibited noactivity in the cell lines tested (Fig . 1 B) . Inclusion ofthe E box resulted in 13-fold higher levels of luciferaseactivity in PC12 versus HepG2 cell lines (Fig . 1B),

FIG . 5 . Tissue distribution of the CAT reporter gene in transgenicanimals . RNA from tissues indicated above were analyzed usingreverse transcription-PCR as described in Materials and Meth-ods . Radiolabeled samples were resolved on polyacrylamidegels, dried down, and exposed to x-ray film . The relative levelsof the CAT mRNA were determined by densitometry . Units repre-sent relative levels of CAT mRNA determined after defining thedensitometric value of adrenal (Ad .) gland as 1 .0 . Data are meanSEM (bars) values . Cbl ., cerebral ; Olf ., olfactory . Mid . Brain,

midbrain .

suggesting that traps-activation occurs via specific nu-clear factors . Based on our previous study and the datapresented in Fig . l B, our working hypothesis was thattraps-acting factors, possibly interacting with a CAC-GTG motif 230 by downstream of the last exon, wereinvolved in the observed functional response . BothDNasel footprint analysis and gel shift experimentsdemonstrated nuclear proteins could recognize the coreE box hexamcr (Fig . 2) . Mutational analysis suggestedthat a subset of proteins with a preference for a CGinternal dinucleotide was the most likely binding can-didate. To date, several known enhancer binding pro-teins recognize the CACGTG core sequence, includingthe USF family of transcription factors (Pognonec andRoeder, 1991 ), the Myc family (Sawadogo et al .,1988), and the itntnunoblobulin enhancers TFEB andTFE3 (Fisher et al ., 1991) .

Because neither footprinting nor gel shift assayswould be able to distinguish among these proteins, weused two more definitive assays to determine which,if any, of these E box binding proteins was interactingwith the TH 3' enhancers in a cell type-specificfashion . Using HLH-LZ transcription factor-specificprobes, reverse transcription-PCR experiments re-vealed B-Myc and TFE3 to be the only known CAGGTG binding proteins transcripts present in basal,undifferentiated PC12 mRNA (Fig . 3) . However, itwould seem unlikely that B-Myc was involved in thefunctional response because it lacks a DNA bindingdomain and is thought to function via the inhibition of

ANALYSIS OF HUMAN TH GENE EXPRESSION 29

transcriptional

activation

by

c-Myc

( Resar

et

al .,1993) . In contrast, although it was originally definedas an imtnunoglobulin N,E3-activating protein (Beck-mann et al ., 1990), TFE3 does bind to DNA . More-over, it is widely expressed in many peripheral tissuesand, to a lesser extent, in the brain .An involvement of TFE3 binding proteins is

strengthened by the UV cross-linking experiments sug-gesting the formation of a protein/DNA complex of74 kDa with the TH 3' E box . Although secondarystructure within the E box oligonucleotide might sig-nificantly alter the observed protein/DNA complexmobility, the size of the product is compatible withbinding to a TFE3-like factor . However, we cannotrule out the possibility that unknown HLH proteinswith a molecular weight similar to that of TFE3 arebinding to this site . Nevertheless, the molecular weightof the cross-linked species would rule out other HLH-LZ proteins Such as c-Myc or USE It is interestingthat in some experiments the 74-kDa band is clearlyresolved into a doublet in which each band has equalintensity (Fig . 4) . This might suggest the presence ofanother CACGTG binding protein forming a hetero-dimeric complex structure . Like other HLH-LZ pro-teins, TFE3 can form homo- and heterodimers (Fisheret al ., 1991 ) . Therefore, by analogy to myoD and otherproteins determining myogenesis, dimerization is amechanism by which TFE3 function could be modu-lated and thus achieve tissue-specific and/or develop-mentally regulated gene expression . Other potentialmechanisms by which specificity could be achievedinclude cell-specific posttranslational modifications orprotein interactions among different classes of trans-activators that are expressed in a tissue-specific fash-ion . Whether the TFE3-like protein binding to the TH3' E box site is interacting with any other proteins oris further modified is not yet clear and awaits furtherexperimentation .The most common models to explain distinct spatial

and temporal patterns of gene expression propose com-binatorial interactions among developmental irons-ac-tivators that define specific cell phenotypes . A simplemodel to explain the delimited expression of TH inboth the brain and PNS and neuroendocrine tissueswould be that factors leading to its expression in theperiphery would be the same as those in the CNS .It is more likely, however, that multiple factors areinvolved . For example, in this Study the high level ofexpression of J/TK/CAT in adrenal gland correlateswith TH gene expression in this tissue . In addition,many studies have documented the occurrence of cate-cholamines and their biosynthetic enzymes in the endo-crine pancreas (reviewed by Teitelman, 1991 ) . Pre-viously, we used the sensitivity and specificity of re-verse transcription-PCR techniques to determine thatisolated human islets expressed levels of TH that werecomparable to levels detected in human adrenal (Cokeret al ., 1990) . We have also detected TH gene expres-sion in rat spleen using northern blot analysis

J. N"m," ~cm ., Vnl . 05, No . /, /995

30

(K.L.O'M ., unpublished data) . However, the low levelof expression in other tissue such as lung and heart aswell as expression in CNS regions, such as cerebellumand pons, is clearly ectopic .

Based on our analysis of transgenic tissues express-ing the J/TK/CAT construct (Fig. 5), we proposethat, at the very least, factors contributing to cell type-specific expression of TH in peripheral neuroendocrinetissues are not sufficient for delimited CNS expression .Conversely, 5 kb of human TH 5' sequence by itselfwas insufficient to achieve correct peripheral and CNSexpression (Sasaoka et al ., 1992) . Because the mostfaithful transgenic reproduction of endogenous TH ex-pression was achieved with only 2.5 kb of 5' sequenceas well as 0.5 kb of 3' sequence (Kaneda et al ., 1991 ),we suggest that elements located at both ends of thegene are necessary for spatial fidelity . Our in vitrostudies suggest that a HLH-LZ recognition site located230 by downstream of the last exon is critical inachieving high levels of cell type-specific expression .Conceivably, this is due to the interaction of a TFE3-like factor. To test this hypothesis we are currentlyconstructing a mutated TH 3' E box transgene . Datafrom these animals will aid in defining mechanismscontributing to the delimited expression of both sympa-thoadrenal differentiation but also pivotal catechol-aminergic CNS nuclei .

Acknowledgment : This work was supported by fundsfrom National Institute of Mental Health (grant MH45530) .S.C.W . was supported by Research Training in Cellular andMolecular Neurobiology grant 5-T32-NS07071 from the Na-tional Institute of Neurological Disorders and Stroke .

J . Ncurnc'hem_ Vol. 65, No . /, 1995

REFERENCES

Allot N . K . and Vapnek D . ( 1979) Nucleotide sequence analysisof the chloramphenicol resistance transposon Tn9 . Nature 282,864-869 .

AUSUbel F . M ., Brent R ., Kingston R . E., Moore D . D ., SeidmanJ . G ., Smith J . A ., and Struhl K . (1992) Current Protocols inMolecular Biology . John Wiley & Sons, New York .

Beck L, Ramirez S ., Weinmann R ., and Caro J . ( 1991 ) Enhancerelement at the 3'-flanking region controls transcriptional re-sponse to hypoxia in the human erythropoietin gene . J. Biol.Chem . 266, 15563-15566 .

Beckmann H ., Su L . K ., and Kadesch T. ( 1990) TFE3 : a helix-loop-helix protein that activates transcription through the im-munoglobulin cnhancer muE3 motif . Genes Dcv . 4, 167-179 .

Bell G . I ., Selby M . J ., and Rutter W . .I . ( 1982) The highly polymor-phic region near the human insulin gene is composed of simplytandemly repeating sequences . Nature 295, 31-35 .

Blackwell T . K ., Huang J ., Ma A ., Kretzner L ., Alt F. W ., EisenmanR . N ., and Weintraub H . ( 1993) Binding of myc proteins tocanonical and noncanonical DNA sequences . Mol. Cell. Biol.13, 5216-5224 .

Bradford M . M . ( 1976) A rapid and sensitive method for the quanti-tation of microgram quantities of protein utilizing the principleof protein-dye binding . Anal. Biochem . 72, 248-254 .

Cambi F ., Fung B ., and Chikaraishi D. (1989) 5' Flanking DNAsequences direct cell-specific expression of rat tyrosine hydrox-ylasc . .1 . Neurochem . 53, 1656-1659 .

Carr C . S . and Sharp P . A . (1990) A helix-loop-helix proteinrelated to the immunoglobulin E box-binding proteins . Mol.Cell. Biol. 10, 4384-4388 .

S. C. WONG ET AL.

Chen C . and Okayama H . ( 1987) High-efficient transformation ofmammalian cells by plasmid DNA . Mol. Cell. Biol . 7, 2745-2752 .

Chomczynski P . and Sacchi N . ( 1987) Single-step method of RNAisolation by acid guanidinitan thiocyanatc-phenol-chloroformextraction . Anal. Biochem . 162, 156-159 .

Coker G . T . 111, Studelska D ., Hannon S ., Burke W., and O'MallcvK . L . ( 1990) Analysis of tyrosine hydroxylase and insulin tran-scripts in human ncurocndocrine tissues . Mol . Brain Res. 8, 93-98 .

Cordle S . R ., Henderson E ., Masuoka H ., Weil P. A ., and Stein R .(1991a) Pancreatic beta-cell-type-specific transcription of theinsulin gene is mediated by basic helix-loop-helix DNA-bind-ing proteins . Mol . Cell . Biol. 11, 1734-1738 .

Cordle S . R ., Whelan J ., Henderson E., Masuoka H ., Weil P. A .,and Stein R . ( 19916) Insulin gene expression in nonexpressingcells appears to be regulated by multiple distinct negative-actingcontrol elements . Mol . Cell. Biol. 11, 2881-2886 .

Dariavach P ., Williams G . T ., Campbell K ., Pettcrsson S ., and Neu-berger M . S . ( 1991 ) The mouse IgH 3'-cnhancer. Eur. ,1 . lnunu-nol. 21, 1499- 1504 .

De Wet .1 . R ., Wood K . V ., DeLuca M ., Helinski D. R ., and Suhra-mani S . ( 1987) Firefly luciferasc gene : structure and expressionin mammalian cells . Mol. Cell . Biol. 7, 725-737 .

Efrat S ., Linde S ., Kofod H ., Spector D ., Delannoy M ., Grant S .,Hanahan D ., and Baekkeskov S . ( 1988) Beta-cell lines derivedfrom transgenic mice expressing a hybrid insulin gene-onco-gene. Proc . Nall. Acad. Sci. USA 85, 9037-9041 .

Fisher D . E ., Carr C . S ., Parent L . A ., and Sharp P . A . ( 1991 ) TFEBhas DNA-binding and oligomerization properties of a uniquehelix-loop-helix/leucine-zipper family . Genes Dev. 5, 2342-2352 .

Gandelrnan K.-Y ., Coker G . T . Ill, Moffat M ., and O'Malley K . L .( 1990) Species aid regional differences in the expression ofcell-type specific elements at the human and rat tyrosine hydrox-ylase gene loci . .1. Neurochem. 55, 2149-2152 .

Gregor P . D ., Sawadogo M ., <aid Roeder R . G . ( 1990) The adenovi-rus major late transcription factor USF is a member of thehelix-loop-helix group of regulatory proteins and binds toDNA as a llimer . Genes Des . 4, 1730-1740 .

Hanky T . and Merlie J . P. ( 1991 ) Transgcne detection in unpurifiedmouse tail DNA by polymerase chain reaction . Biotechnique10, 56 .

Higuchi R . ( 1990) Recombinant PCR, in PCR Protocols : A Guideto Methods and Applications (firms M . A ., Gelfand D . H .,Sninsky J . J ., and White T . J ., eds), pp . 177-183 . AcademicPress, San Diego, California .

Hogan B ., Costantini F ., and Lacy t: . ( 1986) Manipulating theMouse Embryo : A Laboratory Manual . Cold Spring HarborLaboratory . Cold Spring Harbor, New York .

Hu Y.-F ., Luescher B ., Admon A ., Mermod N., and Tjian R . ( 1990)Transcription factor AP-4 contains multiple dimerization do-mains that regulate dimer specificity . Genes De v . 4, 1741 -1752 .

Kaneda N ., Sasaoka T ., Kobayashi K ., Kiuchi K ., Nagatsu 1 ., Kuro-sawa Y ., Fujita K ., Yokoyama M ., Nomura T ., KatSUki M ., andNagatsu T . ( 1991 ) 'tissue-specific and high-level expression ofthe human tyrosine hydroxylase gene in transgenic mice. Neu-ron 6, 583-594 .

Kerkhoff E., Bister K ., and Klempnauer K . H . ( 1991 ) Sequence-specific DNA binding by Myc proteins . Pro( . Nod. Acad. Sci .USA 88, 4323-4327 .

Krug M . S . and Bcrgcr S . L . ( 1987) Determination of the molarconcentration of messenger RNA, in Methods in En:vmolog),Vol . 152 (Berger S . L . and Kimmel A . R ., cds), pp . 262-266 .Academic Press . San Diego, California .

Ma A ., Moroy T ., Collum R ., Weintraub H ., Alt F . W ., and Black-well T. K . ( 1993) DNA binding by N- and L-Myc proteins .Oncogene 8, 1093-1098 .

Maniatis T ., Fritsch E . F., and Samhrook J . ( 1982) Molecular Clon-ing : A Laboratory Manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, New York .

Ohlsson H ., Karlsson O., and Edlund T . ( 1988) A beta-cell-specific

protein binds to the two major regulatory sequences of theinsulin gene enhancer . Proc. Natl . Acad. .Sci. USA 85, 4228-4231 .

O'Malley K . L . and Rotwein P . (1988) Human tyrosine hydroxylaseand insulin genes are contiguous on chromosome 11 . NucleicAcids Res. 16, 4437-4445 .

Pognonec P . and Roeder R . G . ( 1991 ) Recombinant 43-kDa USFbinds to DNA and activates transcription in a manner indistin-guishable from that of natural 43/44-kDa USE Mot. Cell . Biol.11, 5125-5136 .

Pongubala J . M . and Atchison M . L . ( 1991 ) Functional characteriza-tion of the developmentally controlled immunoglobulin kappa3' enhancer : regulation by Id, a repressor of helix-loop-helixtranscription factors . Mot. Cell. Biol . 11, 1040-1047 .

Purucker M ., Bodine D ., Lin H ., McDonagh K ., and Nienhuis A . W .( 1990) Structure and function of the enhancer 3' to the humanA gamma globin gene. Nucleic Acids Res. 18, 7407-7415 .

Resar L . M ., Dolde C ., Barrett J . F ., and Dang C . V . (1993) B-mycinhibits neoplastic transformation and transcriptional activationby c-myc . Mot. Cell. Biol . 13, 1130-1136 .

Rosenthal N ., Berglund E . B ., Wentworth B . M., Donoghue M.,Winter B ., Bober E ., Braun T ., and Arnold H . H . (1990) Ahighly conserved enhancer downstream of the human MLC I /3 locus is a target for multiple myogenic determination factors .Nucleic Acids Res. 18, 6239-6246 .

Sasaoka T ., Kobayashi K ., Nagatsu I ., Takahashi R ., Kimura M .,Yokoyama M ., Nomura T ., Katsuki M ., and Nagatsu T. ( 1992)Analysis of the human tyrosine hydroxylase promoter-chlor-

ANALYSIS OF HUMAN TH GENE EXPRESSION 31

amphenicol acetyltransferase chimeric gene expression intransgenic mice . Brain Res . 16, 274-286 .

Sawadogo M., Van Dyke M. W ., Gregor P . D ., and Roeder R . G .(1988) Multiple forms of the human gene-specific transcriptionfactor USF : complete purification and identification of USFfrom HeLa cell nuclei . J . Biol. Chem. 263, 11985-11993 .

Teitelman G . ( 1991 ) Cellular and molecular analysis of pancreaticislet cell lineage and differentiation . Recent Prog . Horm. Res .47, 259-294 .

Teitelman G ., Job T. H ., and Reis D . J . (1981 ) Transformation ofcatecholaminergic precursors into glucagon (A) cells in mouseembryonic pancreas . Proc. Natt. Acad. Sci. USA 78, 5225-5229 .

Weintraub H ., Davis R ., Tapscott S ., Thayer M., Krause M ., BenezraR ., Blackwell T . K ., Turner D ., Rupp R ., Hollenberg S ., et al .(1991) The myoD gene family : nodal point during specificationof the muscle cell lineage. Science 251, 761-766 .

Wong S . C ., Moffat M . A ., and O'Malley K . L. ( 1994) Sequencesdistal to the API/E box motif are involved in the cell type-specific expression of the rat tyrosine hydroxylase gene. .l. Newrochenz. 62, 1691-1697 .

Yoon S . O . and Chikaraishi D . M . (1992) Tissue-specific transcrip-tion of the rat tyrosine hydroxylase gene requires synergy be-tween an AP- I motifand an overlapping E box-containing dyad .Neuron 9, 55-67 .

Zhao A . Z ., Colin A . M ., Bell J ., Baker M., Char B . R ., and MaxsonR . ( 1990) Activation of a late H2B histone gene in blastula-stage sea urchin embryos by an unusual enhancer element lo-cated 3' of the gene . Mot. Cell . Biol. 10, 6730-6741 .

J. Neurochem., Vol. 65, Nu. 1, 1995