Development 109, 157-165(1990)Printed in Great Britain © T h e Company of Biologists Limited 1990

157

KTF-1, a transcriptional activator of Xenopus embryonic keratin expression

A. M. SNAPE*, E. A. JONAS and T. D. SARGENT

Laboratory of Molecular Genetics, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD20892, USA

• Corresponding author

Summary

Nuclear extracts from embryos of Xenopus laevis wereshown to contain a protein activity, KTF-1, which bindsin vitro to the promoter of the embryonic, epidermis-specific keratin gene, XK81A1. Mobility shift assays,methylation interference and footprinting analysis wereused to show that the KTF-1 binding site contains animperfect, palindromic sequence, ACCCTGAGGCT.This sequence occurs once in the XK81A1 promoter,152-162 base pairs upstream of the transcription startsite. A construct of the keratin gene in which thissequence was altered so that it no longer binds KTF-1 invitro showed severely reduced transcription levels upon

injection into Xenopus embryos, but retained epidermalspecificity. Addition of KTF-1 binding sites alsoenhanced epidermal and non-epidermal activity of aheterologous promoter, Xenopus /J-globin, in embryos.These results suggest that KTF-1 is a general activator ofembryonic keratin transcription, which acts in concertwith other factors to produce high levels of epidermis-specific expression.

Key words: DNA binding protein, embryonic keratin,transcriptional activator, Xenopus.

Introduction

The XK81A1 embryonic epidermal keratin gene ofXenopus laevis is an example of a tissue-specific genethat is regulated in response to early developmentalevents. At mid-blastula transition it undergoes rapid,cell autonomous activation in the outer ectoderm (Sar-gent et al. 1986; Jamrich et al. 1987), but it also showsresponse to induction; in vivo it is turned off by neuralinduction (Jamrich et al. 1987), and in vitro can bedeactivated by growth factors, such as XTC-MIF, whichconvert ectoderm to mesoderm (Symes et al. 1988).Thus, a knowledge of the transcription factors thatcontrol XK81A1 expression should lead to greaterunderstanding of both mosaic (cell autonomous) andregulative (cell interactive) developmental decisions.

Previously, we used injection of the cloned keratingene into fertilised Xenopus eggs to identify regions ofthe gene that are necessary for tissue-specific expression(Jonas et al. 1989). We showed that the coding and 3'sequences of the gene, downstream of +26, could besubstituted by a reporter gene (human ^-globin), andthat 5' flanking sequence could be deleted to -487,without loss of epidermis-specific expression. Further 5'deletions resulted in progressive loss of regulation, sothat expression was both reduced in epidermis andincreased in non-epidermal tissues. This suggests thatboth positive and negative transcription factors suf-ficient for correct regulation of this gene bind sequence

elements between —487 and +26. Our next aim is toidentify these elements and the embryonic proteins thatrecognise and bind them.

In this paper, we report on the identification of aDNA-binding activity, KTF-1, from tailbud stageXenopus embryos, which recognises a specific sequencecentred around position —157 in the XK81A1 pro-moter. In vitro mutagenesis was used to create a gene inwhich the KTF-1 binding site is non-functional, and thiswas injected into embryos. The results of such exper-iments support a role for KTF-1 as a transcriptionalactivator which increases XK81A1 expression in epider-mis. However, the mutated gene, which lacks the KTF-1 binding site, is still expressed, to a lower level,specifically in epidermis. This suggests that, like otherregulated genes, XK81A1 is controlled by a number oftranscription factors, including KTF-1.

Materials and methods

Nuclear extractsNuclear extracts from 300 to 3000 tailbud Xenopus embryos,stages 17-22 (Nieuwkoop and Faber, 1967), were preparedaccording to the method of Mohun et al. (1989), withmodifications: the cell homogenisation buffer contained 2.2 Msucrose, and was supplemented with 1 % no-fat, dry milk;only a single centrifugation was used to purify nuclei. Afterlysis of the nuclei in 0.55 M KC1, the chromatin was pelleted asdescribed (Mohun et al. 1989), and the supernatant concen-

158 A. M. Snape, E. A. Jonas and T. D. Sargent

trated in Centricon 10 microconcentrators (Amicon, W. R.Grace and Co., Danvers, MA). It was then diluted withpotassium-free buffer (26mM Hepes pH7.6, 0.1 mM EDTA,10% glycerol, lmM DTT, 0.1 mM PMSF), to a final KC1concentration of 40 mM. The diluted supernatant was thenreconcentrated to give approximately 4 embryo equivalentsper f<l of the final extract.

DNA mobility shift assaysEach 15^1 binding reaction contained 50 mM NaCl, 0.1 mMEDTA, 20 mM Hepes pH8, 5% glycerol, lOj/gmP1 BSA,4% Ficoll-400, 0.004% bromophenol blue, 0.004% xylenecyanol, with DTT added to 1 mM just before use, 5 /<g double-stranded poly(dI.dC).poly(dI.dC) (Pharmacia) as non-specific competitor, 0.1 ng end-labelled, double-strandedDNA probe (approximately 104ctsmin~', see below),specific competitor DNA, as appropriate, and 0.5-5 [A nu-clear extract, added last. Specific competitors were isolatedrestriction fragments of plasmid DNA, their concentrationestimated by ethidium bromide staining in agarose gels, ordouble-stranded synthetic oligo-deoxyribonucleotides.Double-stranded oligo competitors were prepared by anneal-ing complementary single-stranded oligos at O.lmgml"1, in200 mM NaCl, for 30min, at 37°C. After precipitation andresuspension, the DNA concentration was checked bymeasuring the OD26o- The binding reactions were incubatedat room temperature for 45min, then loaded directly onto a3.5% acrylamide gel (acrylamide: bis-acrylamide ratio=19:l)which was run in 50 mM Tris, 20 mM boric acid, 10 mM EDTApH8.8. Gels were dried and exposed to X-ray film.

ProbesTo make probe 1, the -423 to -140 fragment of XK81A1 wassubcloned intopBS+ (Stratagene). For labelling of the codingstrand, the subclone was linearised using BgUl, at —274, andend-labelled using polynucleotide T4 kinase (Baker andHowley, 1987). The probe fragment was then released bycutting with Rsal, which cuts in the polylinker, and isolatedfrom a 5 % acrylamide gel (Ausubel etal. 1987). An aliquot ofthe probe was counted, and its concentration and specificactivity estimated from the known starting concentration ofthe plasmid, and the specific activity of the y-[32P]ATP usedfor labelling. For labelling the non-coding strand, the pro-cedure was essentially the same, except that the Bgfll endswere labelled by filling in, using the Klenow fragment of DNApolymerase I and [32P]dGTP (Ausubel et al. 1987).

DMS interference assayEnd-labelled probe 1 was partially methylated by DMS(Ausubel et al. 1987), and used in scaled up binding reactions(1 ng probe, 30/.ig poly(dl.dC), 6-10 jd nuclear extract). Freeand bound probe were separated on an acrylamide gel, as formobility shift assays, and then electroblotted onto NA45DEAE-cellulose paper (Schleicher and Schuell). The bandswere located by autoradiography, and the DNA eluted andpiperidine cleaved as described (Ausubel et al. 1987) beforeanalysing on 6% sequencing gels. A standard Maxam andGilbert (1980) DMS-sequencing analysis of probe 1 was run inparallel for comparison (Ausubel et al. 1987).

FootprintingFor 1,10-phenanthroline-copper ion (OP-Cu) footprinting,scaled up binding reactions were run as for DMS interference,except that no prior treatment of the probe was necessary.The free and bound probe were digested with OP-Cu, withinthe acrylamide matrix (Kuwabara and Sigman, 1987). Thecleaved DNA was recovered from NA45 paper, as above, and

analysed on a 6% sequencing gel. A Maxam and Gilbert(1980) A/G sequencing analysis of probe 1 was run in parallel(Ausubel et al. 1987).

Plasm idsK487 has been described previously (Jonas et al. 1989).M157E was derived by replacing base pairs -157 through— 150, as shown in Fig. 5, using the in vitro mutagenesisprocedure of Norris et al. (1983). KG487 was derived byremoval of upstream sequence from KG5900 (Jonas et al.1989). Derivatives of 64-X/SG (the gift of D. Melton) weremade by ligating kinased, double-stranded 36-mers, (c6 or c7,Fig. 2B) into the Klenow filled-in 5' EcoR I site of 64-X/3G.The number and orientation of the inserts were confirmed bydouble-strand sequencing using a Sequenase Kit from USB.

Embryo injection and analysisPlasmid DNA stocks for injection were prepared as described(Jonas et al. 1989). All constructs were linearised at the 3' endof the insert, by digestion with an appropriate restrictionenzyme, before injection. Embryo injections at the 2-cellstage, and total RNA and DNA preparation from injectedembryos, were carried out as before (Jonas et al. 1989).Keratin and human /J-globin transcripts from the injectedgenes were analysed by RNase protection (Jonas et al. 1989).Xenopus /J-globin transcripts from injected genes, and keratinand actin transcripts from endogenous genes (used as stan-dards) were analysed by Northern hybridisation (Sargent etal. 1986), using, as probes, nick-translated, isolated insertsfrom cDNA clones encoding Xenopus /J-globin (pXGD82;Kay et al. 1980), keratin XK81A1 (pC8128; Jonas et al. 1985),or n^actin (SPAC 9; Sargent et al. 1986). Labelling by nicktranslation, using [a-32P] dCTP, was carried out using a kitfrom BRL. DNA from injected embryos was analysed usingdot blots (Jonas et al. 1989), hybridised against nick-trans-lated, labelled inserts from cDNA clones encoding keratin(see above), Xenopus /3-globin (see above), or human /S-globin (the gift of S. Karlsson), as appropriate.

Results

DNA-protein binding studiesDNA mobility shift assays (Fried and Crothers, 1981)were used to look for evidence of proteins binding theXK81A1 promoter. Nuclear extracts prepared fromearly tailbud Xenopus embryos, which are activelyexpressing the gene, were found to contain an activitythat specifically binds a Bglll-HinQ fragment, from—274 through —140 (probe 1; Fig. 1). The binding iscompeted by 500x molar excess of unlabelled probe,but not by a non-specific competitor, the 676 base pair(bp) Rsal fragment of pUC19. The shift was abolishedby preincubation of the nuclear extract with proteinaseK (not shown), indicating that the DNA-binding ac-tivity depends on a protein, or proteins. When theprobe fragment was divided into two competitor frag-ments, of 59 and 80 bp, the upstream fragment (2) didnot compete, while the downstream fragment (3) com-peted as efficiently as the whole probe sequence,indicating that the protein-binding site lies within frag-ment 3.

In order to identify the binding site more precisely,we made a series of overlapping, 30 bp oligodeoxyribo-

Transcriptional activator of Xenopus keratin 159

probe 1-274 135bp -140

-215

H R4bp

competitors

„ B9

competitor

H R

pUC

Fig. 1. Nuclear extract from Xenopus embryos contains anactivity that binds the keratin promoter. Diagrams show therestriction fragment probe 1, and competitor sub-fragments2 and 3 (see text). Solid lines represent keratin promotersequence; the broken line represents 4bp of vectorsequence. Relevant restriction sites are shown, with theirpositions in the XK81A1 promoter: Bg, flg/11; H, HiniV(not regenerated in subcloning); R, Rsa\ (in vector); X,Xmnl. DNA mobility shift assays: each lane contains anequal amount of end-labelled probe 1, poly(dl.dC)competitor, and nuclear extract from tailbud embryos (seeMaterials and methods). In the first lane, no specificcompetitor was added; in subsequent lanes, unlabelledcompetitors were used in 5, 50 and 500-fold molar excess,as shown. pUC(676R) is a 676 bp, Rsal fragment of pUC19(unrelated competitor). In each lane, the upper band isprotein bound, retarded probe and the lower band isunbound probe.

nucleotides covering the sequence of fragment 3, andused them as unlabelled competitors of probe 1 in themobility shift assay. These data are summarised inFig. 2A. Only oligo c (-165 to -136) gives compe-tition, whereas the oligos which overlap c (f and g) donot, suggesting that a binding site lies within c, aroundthe junction of f and g. This was confirmed by making aset of variants on oligo c, each one having a smallnumber of altered base pairs (Fig. 2B). Only alterationsin the first sixteen base pairs abolished the ability of theoligo to compete for binding (Fig. 2C). This regionincludes an imperfect palindromic sequence of elevenbases, ACCCTGAGGCT, centred around the G resi-due at position -157. Transcription factors often bindto imperfect palindromic sequences (Curran and

Franza, 1988); however, this sequence does not appearto correspond to the binding site of any previouslyidentified factor (see Discussion). We were able to showthat this sequence can act as a positive regulatoryelement for keratin gene transcription (see below). Wetherefore named the embryonic DNA-binding activityKTF-1, for Keratin Transcription Factor-1.

DMS-methylation interference and footprintinganalysis were used to obtain more information on theDNA-KTF-1 interaction. DMS-interference identifiesG residues, methylation of which interferes with bind-ing of a protein to DNA (Siebenlist and Gilbert, 1980).As shown in Fig. 3, KTF-1 binding to probe 1 is reducedby DMS methylation of either of the GG pair at -155and -156 on the coding strand, or any of the three Gs at-159 through -161 on the non-coding strand, sugges-ting that all these residues are directly involved inprotein binding. However, methylation of the central Gat -157, or of the G at —153 (which constitutes animperfection in the palindromic repeat), does notinterfere with KTF-1 binding, implying that theseresidues are not directly associated with the protein.

In 1,10-phenanthroline-copper ion (OP-Cu) foot-printing (Kuwabara and Sigman, 1987), protein bindingprotects the DNA from cleavage by this reagent. Asshown in Fig. 4, KTF-1 binding to probe 1 protectsresidues -166 through -148 on the coding strand, and-166 through -148 on the non-coding strand. On thenon-coding strand, residues -166 through —169 arealso partially protected, while on the coding strand,KTF-1 binding appears to create a site, at residue -147,which is hypersensitive to OP-Cu digestion. The infor-mation from methylation interference and footprintingis summarised in Fig. 4B. These results are in goodagreement with the data from mobility shifts.

Functional analysis of KTF-1 bindingThe role of KTF-1 in the control of keratin transcriptionwas investigated by combining in vitro mutagenesis withan embryo injection assay. Our previous analysis (Jonaset al. 1989) had shown that 487 bp of the upstreampromoter is sufficient to confer epidermis-specific ex-pression on an injected XK81A1 gene (K487, seeFig. 5). We therefore constructed a mutated version ofK487, M157E (Fig. 5), replacing nucleotides -165 to-136 with oligo c5 (Fig. 2B), so that the KTF-1 bindingsite is altered, and no longer competes for binding in themobility shift assay (Fig. 2C). We compared the pro-moter activity and tissue specificity of K487 and M157E,by injecting them into fertilised Xenopus embryos,allowing the embryos to reach tailbud stage, and thenanalysing expression from the injected genes by RNaseprotection assay, as previously described (Jonas et al.1989).

For the analysis of promoter strength, the twoconstructs were injected separately into sibling em-bryos. A reference construct, KG487 (Fig. 5), which isa fusion of the unaltered keratin promoter (from -487to +26) with the coding sequence of human /3-globin,was mixed and coinjected with each of the test genes. Ineach experiment, filter hybridisation of DNA from the

160 A. M. Snape, E. A. Jonas and T. D. Sargent

oligo comp.

abc +defg

Boligo comp.

AACACCCTGAGGCTACGTAACTGAATCAAG CGGTGTTTCGAGGCTACGTAACTGAATCAAG C1AACACCCTAGCCTCGTGTAACTGAATCAA^ C2AACACCCTGAGGCTAC&CG.G1CAGATCAAG C3AACACCCTGAOGCTACGTAACTGAGiXGjIA C4

ACAAACACCCTA£A1AICIGTAACTGAATC C5GATCCAACAAACACCCTGAGGCTACGTAACTGGATC C6GATCCAACAAACATTTCAGAATTACGTAACTGGATC C7

Q.Eoo o

Xin

X

om

X

ooin

C1 C2

*#»

C3 C4 C5

C6

X

m

XinCM

XOin

CMC7

Fig. 2. Mobility shift competition analysis allows location of the protein binding site within probe 1. (A) Overlapping 30-mers a-g cover the sequence of competitor fragment 3 (see Fig. 1). Competitors a, b and c cover the XK81A1 promotersequence from -225 through -136, while e, f and g are offset by lObp, and cover -215 through -126. Competitor d isvector sequence, including the 4bp in probe 1 and fragment 3. In the mobility shift analysis (not shown) only c competeswith probe 1 for binding at 50-500x molar excess. (B) Localisation of the protein binding site within c. Competitors cl-c7contain altered base pairs (underlined). Competitors c6 and c7 have additional GATC linker sequences (bold type). Wherethe mutation alters the protein binding site (cl, c2, c5, c7), competition with probe 1 is abolished. (C) Results of mobilityshift competition analysis. Mobility shift on probe 1 was carried out as in Fig. 1. Excess unlabelled oligonucleotides wereused as competitors (comp.) at the molar ratios indicated.

injected embryos against probes specific for eitherglobin or keratin showed that the control and testinjected DNA survived and replicated to a similarextent. Thus it is valid to normalise expression of thetest genes against KG487 expression, which was assayedby a globin-specific RNase protection probe (Karlssonet al. 1988). By this method, we found that mutation ofthe KTF-1 binding site significantly lowers the level oftranscription from the injected keratin construct. Asample set of results is shown in Fig. 6A.

The experiment was repeated several times, usingthree separate preparations of M157E and two prep-arations of K487. The results were quantified by densi-tometric scanning of the RNase protection gels, andexpressed as the ratio of intensities of the band arisingfrom correct initiation of the injected keratin gene (i inFig. 6A) and the smallest of the four protected globinbands (the intensities of the four globin bands beingsimilar to each other). The results are tabulated inTable 1 (where experiment 6 is the one illustrated inFig. 6A), and show that, on average, mutation of theKTF-1 binding site confers an eight-fold reduction intranscriptional activity on the keratin promoter. Corn-

Table 1. Comparison of transcriptional activity of thekeratin gene, XK81A1, with (K487), and without

(M157E), a functional KTF-1 binding site

Experiment

Ratio oftranscription

levelsK487:M157E

Ratio ofDNA levels

K487:M157E

Transcriptionlevel relativeto DNA level

9.411.812,45.24.56.18.2

1.00.91.31.31.40.91.1

9.413.19.54.03.26.87.5

123456Average

Transcription and DNA levels were measured by densitometry ofautoradiographs from RNase protection assays and DNA dot blots.In each experiment, transcription and DNA levels were normalisedwith respect to a co-injected, reference gene (see text). Experiment6 is illustrated in Fig. 6A.

parison with the co-injected reference gene, KG487,showed that each set of embryos received, and wascapable of expressing, injected DNA. DNA dot blots

Transcriptional activator of Xenopus keratin 161

CO

r

en

f

+ i

f

Fig. 3. DMS methylation interference analysis of KTF-1 binding to the keratin promoter. Binding reactions were carried outon partially methylated, end-labelled probe 1, using nuclear extracts from tailbud embryos. After separation and recovery ofbound (b), and free (f) probe (see Materials and methods) the DNA was cleaved at the methylated G residues, andelectrophoresed. Each band represents a G residue in the probe sequence, as shown at the left of the gel. The probe waslabelled on the coding strand (+) in A, and on the non-coding strand (—) in B. Where methylation of a G residue interfereswith protein binding, that residue is under-represented in the bound fraction relative to the free, as seen in the densitometricscans. Two G residues on the coding (+) strand (—155 and —154) and three on the non-coding (-) strand (—159, —160 and— 161) are suppressed 2- to 3-fold in the bound fraction.

were scanned for each experiment. The ratio of tran-scription level to amount of template DNA is given inTable 1 (but see Discussion), and indicates that thereduced expression of M157E cannot be attributed tothe injection or persistence of lower levels of M157EDNA. Thus, the binding site for KTF-1 serves toactivate XK81A1 transcription, and KTF-1 is suggestedto be a transcriptional activator of keratin expression inthe Xenopus embryo.

Besides band i, representing correct initiation of theinjected keratin gene, the keratin probe protects otherbands specifically in injected embryos. We interpretthese as representing transcription from cryptic startsites in the gene or vector, because deletion of theTATA box causes disappearance of band i, but not ofthese bands (data not shown). The mismatch betweenthe RNase protection probe, which includes 5' flankingsequence, and the mutation at -157 in M157E, causesthe longest of these bands to be truncated compared toits equivalent in K487-injected embryos. We observedthat the additional bands are often stronger in M157E-injected embryos, as in Fig. 6A. The reason for this isnot known, but reduced use of the correct initiation sitein M157E may allow the cryptic start sites to be moreavailable for transcription, for instance, if they liewithin the XK81A1 transcription unit.

To investigate whether the KTF-1 site is necessary forepidermis-specific expression of XK81A1, M157E-injected embryos were dissected into epidermis and

carcass (non-epidermis) fractions before RNase protec-tion analysis (Fig. 6A(iv)). The results of four separateexperiments show that although M157E expression isdrastically reduced, it is still epidermis specific.

To investigate whether KTF-1 can act as an activatorof promoters other than the XK81A1 gene, we testedthe ability of the KTF-1 binding site to increase ex-pression of the Xenopus adult /J-globin gene, which isnormally inefficiently transcribed on injection into em-bryos (Krieg and Melton, 1985; Bendig and Williams,1984). We constructed synthetic oligos containing eitherthe sequence of the KTF-1 binding site (oligo c6,Fig. 2B), or a sequence altered so that it did notcompete for KTF-1 binding in mobility shift assays(oligo c7, Fig. 2B,C). Either c6 or c7 was cloned intoplasmid 64-X/JG, 500 bp upstream from the initiationsite (Fig. 5). 64-X/3G contains the entire Xenopus adult/3-globin gene, including approximately 500 bp ofupstream flanking sequence. The constructs wereinjected into Xenopus embryos, and transcription wasanalysed by Northern hybridisation (Fig. 6B). To cor-rect for variations in RNA loading, globin expressionwas normalised to that of an endogenous gene, either a-actin or keratin XK81A1.

No consistent effect on transcription was observed inconstructs containing only one copy of the KTF-1binding site (data not shown). However, in 4 out of 5experiments, constructs containing two binding sitesshowed much higher levels of transcription than both

162 A. M. Snape, E. A. Jonas and T. D. Sargent

A/G b f A/G b f

K487/M157E

-487EZD- SX-

K487 ACCCTGAGGCTACG (wild type-162 to-149)M157E ACCCTAGATATCTG (mutant -162 to -149)

64-X(3G

B

-178 -140TT

+ TTTTTGTTTAACAAACACCCTGAGGCTACGTAACTGAAT

- AAAAACAAATTGTTT.^TGGGACTCCGATGCATTGACTTA

1 I *** I

Fig. 4. (A) Footprinting analysis of KTF-1 binding to thekeratin promoter. Binding reactions were carried out onprobe 1, as described for DMS interference analysis, andanalysed by mobility shifting. OP-Cu ions were used tocleave the bound (b) and free (f) fractions, and theproducts were recovered, and electrophoresed alongside anA/G sequencing reaction of probe 1. KTF-1 binding toprobe 1 gives a single footprint, covering sequences from-148 to -166 on the coding (+) strand, and -151 to -169on the non-coding (—) strand. (B) Diagram summarisingthe footprinting and DMS interference analysis of KTF-1binding to the keratin promoter. Sequence represents theXK81A1 upstream region from residue —178 to -140. '+ 'is the coding, and '—' the non-coding, strand. G residuesshown by DMS interference to be important in KTF-1binding are marked by arrows. The extent of the OP-Cufootprint on each strand is delimited by a square bracket.The broken line marks three residues on the non-codingstrand, where KTF-1 binding results in partial blockage ofOP-Cu digestion. The asterisk marks an OP-Cu hyper-sensitive site in the coding strand, apparently created byKTF-1 binding (Fig. 4A).

/ \/ \

:—• *- XpG-K2X " X XPG-M2

Fig. 5. Gene constructs used in functional analysis of KTF-1 binding. Solid lines represent eukaryotic DNA. Brokenlines are vector sequence. Stippled boxes are XK81A1exons. Cross-hatched boxes are human /S-globin exons.Solid boxes are Xenopus /3-globin exons. The restriction sitein the 3' flanking DNA, used to linearise each constructbefore injection, is shown: H, Hindlll; K, Kpnl. K487/M157E: K487 contains the coding sequence of the XK81A1keratin gene, with 487 bp of upstream sequence, cloned intopUC18. The gene is modified by insertion of 36 bp into thefirst exon (marked by a vertical line), which allowsdetection of transcripts by RNase protection (Jonas et al.1989). M157E is identical to K487, except that 7bp of thepromoter sequence, within the KTF-1 binding site, werealtered to give an Eco RV binding site, as shown. Thissequence does not compete for KTF-1 binding (see Fig. 2).The position of the mutation is marked by a vertical arrow.KG487: Fusion of XK81A1 sequence, from -487 to +26(upstream promoter sequence and 5' untranslated sequenceof the first exon), to the human /3 globin gene at thetranslation initiation codon; vector is pBS+ (Jonas et al.1989). 64-XfiG: The Xenopus /S-globin gene, withapproximately 500bp of upstream sequence, cloned intopSP64. Derivatives of 64-X/SG, containing either functionalor mutated, non-functional KTF-1 binding sites cloned intothe 5' Eco RI site at -500, are shown below. X/SG-K2contains 2 copies of oligo c6 (wild-type KTF-1 site, seeFig. 2), both in the correct orientation. X/3G-M2 contains 2copies of oligo c7 (mutant KTF-1 site, see Fig. 2), one inthe correct, and one in the incorrect, orientation.

64-X/SG, and 64-X/3G with two copies of the mutatedbinding site (Table 2). The results of a sample exper-iment (experiment 3 in Table 2) are shown in Fig. 6B.Thus it appears that the KTF-1 binding site can act as astrong enhancer of Xenopus /3-globin transcription.

However, the requirement for two copies of the sitemay be due to some aspect of its environment which isless favourable in the globin promoter than in XK81A1.Dissection of the injected embryos shows that theKTF-1 site enhances globin expression in both epider-mal and non-epidermal tissues (Fig. 6B(iii)).

Transcriptional activator of Xenopus keratin 163

A KM B

(i)

M K2M K M K Ep Ca X K2 M2 Ep Ca

D N A • • • • ' • # # # # #

RNAkeratin1 actin

r ' 9 globin

C X K2M2 EpCaCMK CMK a «

(ii) (iii) (iv) (ii) (jjj)

Fig. 6. Functional analysis of KTF-1 binding. (A) Mutation of the KTF-1 binding site in the keratin promoter, (i): Agarosegel showing DNA preparations for embryo injection, illustrating that similar concentrations of each construct were injected.Upper bands: K, K487 (XK81A1 gene with wild-type KTF-1 binding site); M, M157E (XK81A1 gene with mutated KTF-1binding site). Lower band: KG487 (reference gene) in both lanes. Each DNA preparation was injected into fertilisedXenopus embryos, and expression from the injected genes analysed at tail-bud stage, (ii) and (iii): DNA dot blot and RNaseprotection analysis of injected embryos. Each dot, or each lane, contains DNA or total RNA from the equivalent of oneembryo. The probes were specific for keratin XK81A1 (ii), or human /S-globin (('/'/'). Dot blots show the amount of injectedDNA remaining at tailbud stage, e, RNA from endogenous keratin gene; i, RNA from injected keratin or globin gene; C,control, uninjected embryos; M, M157E/KG487 injected embryos; K, K4S7/KG487 injected embryos. The samples analysedin (ii) and (iii) are from the same experiment. M157E-injected embryos express keratin from the injected gene to asignificantly lower level than K487-injected embryos, even though they contain similar amounts of injected DNA, andexpress the reference injected gene, KG487, to a similar extent. Longer protected bands, presumably representingtranscription from upstream cryptic start sites in the injected plasmid, are also equally strong in M157E- and K487-injectedembryos, (iv): Epidermal specificity of M157E expression. M157E-injected embryos were dissected into epidermal (Ep) andcarcass (Ca) fractions, and subjected to DNA and RNA analysis. Each lane contains material from the equivalent of oneembryo. Most of the injected DNA is in the larger, carcass fraction, but keratin expression from both the injected and theendogenous gene is mainly in the epidermis. (B) KTF-1 binding can confer high levels of expression on a heterologouspromoter. (/): Agarose gel showing DNA preparations for injection. X, 64-X/3G; K2, X/3G-K2 (two wild-type KTF-1 bindingsites); M2, X/3G-M2 (two mutated KTF-1 binding sites). DNA was injected into fertilised Xenopus embryos, and expressionanalysed at the tailbud stage. (ii): DNA dot blot and RNA Northern blot analysis of injected embryos. Dot blots werehybridised with Xenopus /S-globin specific probe, to show the amount of injected DNA remaining. Duplicate RNA gels werehybridised with probes specific for Xenopus /3-globin (injected, i), and for keratin XK81A1 and oactin (endogenous, e) toallow standardisation. Each dot, or each lane, contains DNA or total RNA from the equivalent of one embryo. C,uninjected, control embryos. X/3G-K2 injected embryos (K2) express globin from the injected gene to a higher level thaneither 64-X£SG injected embryos (X), or X/3G-M2 injected embryos (M2), containing similar amounts of injected DNA.(iii): KTF-1 binding does not confer epidermal specificity on a heterologous promoter. X/3G-K2 injected embryos weredissected into epidermal (Ep) and carcass (Ca) fractions, and subjected to DNA and RNA analysis. Each lane containsmaterial from the equivalent of one embryo. Hybridisation of a duplicate RNA gel with probes for the endogenoustranscripts keratin XK81A1 (epidermis specific), and o--actin (mesoderm specific), shows purity of the dissection. Globintranscripts from the injected gene are detected in both tissue fractions, reflecting the distribution of the injected DNA.

Discussion embryonic keratin expression. We show that the sitebinds a nuclear protein from Xenopus embryos, KTF-1,

In this paper, we identify a protein-binding sequence in in vitro, and that its presence in the XK81A1 promoterthe promoter of the Xenopus keratin gene XK81A1, strongly enhances expression of this gene in embryonicand demonstrate the role of this site in the control of epidermis. It is, therefore, likely that KTF-1 functions

164 A. M. Snape, E. A. Jonas and T. D. Sargent

Table 2. Effect of the addition of functional ornonfunctional KTF-1 binding sites on transcriptional

activity of the Xenopus fi-globin gene

Experiment

1

2

3

4

5

Average

Plasmid

X/3G-K2X/3G-M2

X/3G-K2X/3G-M2

X/3G-K2X0G-M2

X/3G-K2X/3G-K2X/3G-M2X/3G-K2X/SG-M2X/3G-K2X/3G-M2

Relativetranscription

level

7.712.19.01.3

21.01.6

37.87.24.0

16.22.216.54.2

RelativeDNA level

1.25.20.71.61.30.93.47.93.69.09.93.94.2

Transcriptionlevel relativeto DNA level

6.42.3

12.90.8

16.21.8

11.20.91.11.80.24.21.0

Transcription and DNA levels were measured by densitometry ofautoradiographs from Northern blots and DNA dot blots.Transcription levels were corrected for differences in RNA loadingby comparison with signal from an endogenous gene in eachsample. Transcription and DNA levels of the test plasmids, whichhave either functional (X/3G-K2) or non-functional (X/3G-M2)KTF-1 binding sites, are expressed as a ratio with respect to 64-X/K3, which was injected into sibling embryos as a standard in eachexperiment. Experiment 3 is illustrated in Fig. 6B.

in the embryo as a transcriptional activator of keratinexpression, although we cannot formally exclude thepossibility that another protein, with similar DNA-binding properties, acts at the KTF-1 site in vivo.

As shown in Tables 1 and 2, there was some varia-bility between experiments in the level of transcrip-tional activation conferred by a functional KTF-1 site orsites, especially when the sites are transferred to aheterologous promoter (Table 2). Some of the varia-bility may be attributed to differences in the level ofplasmid DNA persisting at the time of analysis; how-ever, since we do not know what fraction of injectedDNA is transcribed, and whether any of the measuredexpression arises from transiently high levels of plasmidDNA, it may not be strictly valid to express transcrip-tion relative to the amount of template DNA. Anotherexplanation might be heterogeneity between clutches ofembryos, in the level of KTF-1.

Although the absence of a functional KTF-1 bindingsite reduces expression of XK81A1, it does not alter itstissue specificity (at least within the limits of our assay).Similarly, insertion of KTF-1 binding sites increasestranscription from a heterologous promoter but doesnot confer epidermal specificity. Thus, it appears thatKTF-1 may be an example of a more general transcrip-tion factor contributing to the expression of a tissue-specific gene. Consistent with this interpretation, nu-clear extracts made from either epidermal or carcassfractions of tailbud embryos were found to containKTF-1 activity (assayed by DNA mobility shift),although KTF-1 appeared to be especially abundant inepidermis (data not shown). Dynan (1989) suggests that

such non-specific transcription factors, which are fre-quently found to bind the promoters of regulated genes,may create an environment where transcription ishighly responsive to inducible or tissue-specific factors.

The KTF-1 binding site may act in concert with anumber of other regulatory elements to give epidermis-specific keratin expression. Many transcriptional con-trol regions have a modular structure, which mayinclude some degree of redundancy. For instance, Herrand Clarke (1986) have shown that a functional SV40enhancer can be built either by duplication of homolo-gous sequence motifs, or by combinations of heter-ologous pairs of motifs. In the keratin promoter, theremight be multiple control elements, including the KTF-1 site, combinations of which would give tissue-specificexpression. The results of deleting the XK81A1 pro-moter support such a model; progressive 5'—»3' de-letions from -487 to -100 show gradual loss of tissuespecificity (Jonas et al. 1989), but a series of individual,internal 50bp deletions covering the same region, in abackground of 1310 bp of upstream sequence, were alltissue specific (E. Jonas, unpublished observations).Similar results have been obtained by Mohun et al.(1989), working on another tissue-specific Xenopusembryonic gene, cardiac actin. Here, deletion of asingle protein-binding motif, CArG box 1, from thepromoter, abolishes expression of the injected gene,but transfer of one or more copies of the CArG box failsto confer muscle-specific expression on a heterologouspromoter. This implies that the actin promoter mustcontain other control elements, besides the CArG box,but no evidence of them can be found by mutation ofsmall regions of the promoter.

Keratin expression in Xenopus laevis is both spatiallyand temporally regulated. Different sets of keratingenes are expressed in eggs, embryos and adult tissues(Franz et al. 1983; Jamrich et al. 1987; Ellison et al.1985). XK81A1 is one of a number of Xenopus type Ikeratins, including other members of the XK81 family(Miyatani et al. 1986) and the distantly related type Igene, XK70A (Winkles et al. 1985), which show similarbut not identical expression patterns in the early em-bryo. Upstream sequence information is available fortwo other XK81 genes, XK81B1 and XK81B2 (Miya-tani et al. 1986), and for XK70A (Krasner et al. 1988).No exact match to the XK81A1 KTF-1 site is found inany of these genes. The closest match is in XK81B2,which contains the sequence GCCCTGAAGGT (read-ing 3'—»5') 180bp upstream of the initiation site. Thismatches the KTF-1 binding site at 8 out of 11 nucleo-tides, but no protein-binding data are yet available forthe XK81B2 gene. There is also some sequence simi-larity between the KTF-1 site and an 8bp element(consensus GCCTGPuPuG), previously noted as lyingbetween -250 and -270 in the three sequenced XK81genes, including 81 Al and XK70A (Jonas et al. 1989),and having the same sequence as the recognitionelement for AP-2, a cell-type-specific transcriptionfactor of humans (Williams et al. 1988). However, wehave so far been unable to detect any nuclear factorbinding to this site in the XK81A1 promoter, so it is

Transcriptional activator of Xenopus keratin 165

unclear whether it plays a role in the gene's regulation,and its relationship, if any, to KTF-1 remains unknown.Thus the question of whether KTF-1 also regulatesother embryonic keratins or is specific for XK81A1remains open.

There is some information about sequences regulat-ing keratin gene expression in organisms other thanXenopus (see the review by Steinert and Roop, 1988).Blessing et al. (1989) identified a 5' upstream region,between nucleotides -180 and -605, of the bovinecytokeratin gene, CKIV*, which shows tissue-specificenhancer activity. We note that this region contains thesequence ACCCTCAGAGA, which matches the KTF-1 binding site in 8 out of 11 nucleotides. In addition, ahuman 50X103 Mr type I epidermal cytokeratin gene(Marchuk et al. 1985) contains the sequence ACCCA-CAGGCT (3'-»5') at approximately -100, whichmatches the KTF-1 site in 9 out of 11 nucleotides.

Apart from its limited similarity to the potential AP-2site, we have not found homology of the KTF-1 bindingsite to that of any published transcription factor. ThusKTF-1 may represent a previously unknown transcrip-tional activator. However, given the extreme degener-acy of binding sequences for many transcription factors(see Johnson and McKnight, 1989, for review) this issuemay only be resolved by purification and/or cloning ofKTF-1.

Special thanks to Tim Mohun and Mike Taylor for provid-ing their Xenopus nuclear extract protocol. We are grateful toIgor Dawid, Der-Hwa Huang and Alan Wolffe for helpfulcomments and discussion. Igor Dawid, Stefan Karlsson, DougMelton and Alan Wolffe kindly provided experimental ma-terials. A.M.S. was supported in part by a grant from theWellcome Trust.

References

AUSUBEL, F. M., BRENT, R., KINGSTON, R. E., MOORE, D. D.,SEIDMAN, J. G., SMITH, J. A. AND STRUHL, K. (eds.), (1987).Current Protocols in Molecular Biology, John Wiley and Sons,New York.

BAKER, C. C. AND HOWLEY, P. M. (1987). Differential promoterutilization by the bovine papillomavirus in transformed cells andproductively infected wart tissues. EMBOJ. 6, 1027-1035.

BENDIG, M. AND WILLIAMS, J. (1984). Fidelity of transcription ofXenopus laevis globin genes injected into Xenopus taevis oocytesand unfertilized eggs. Mol. Cell. Biol. 4, 567-570.

BLESSING, M., JORCANO, J. L. AND FRANKE, W. W. (1989).Enhancer elements directing cell-type-specific expression ofcytokeratin genes and changes of the epithelial cytoskeleton bytransfections of hybrid cytokeratin genes. EMBO J. 8, 117-126.

CURRAN, T. AND FRANZA, B. F. JR. (1988). Fos and Jun: the AP-1connection. Cell 55, 395-397.

DYNAN, W. S. (1989). Modularity in promoters and enhancers. Cell58, 1-4.

ELLISON, T. R., MATHISEN, P. M. AND MILLER, L. (1985).Developmental changes in keratin patterns during epidermalmaturation. Devi Biol. 112, 329-337.

FRANZ, J. K., GALL, L., WILLIAMS, M. A., PICHERAL, B. ANDFRANKE, W. B. (1983). Intermediate-size filaments in a germ cell:expression of cytokeratins in oocytes and eggs of the frogXenopus. Proc. natn. Acad. Sci. U.S.A. 80, 6254-6258.

FRIED, M. AND CROTHERS, D. M. (1981). Equilibria and kinetics oflac repressor-operator interactions by polyacrylamide gelelectrophoresis. Nuc. Acids Res. 9, 6505-6525.

HERR, W. AND CLARKE, J. (1986). The SV40 enhancer is composed

of multiple functional elements that can compensate for oneanother. Cell 45, 461-470.

JAMRICH, M., SARGENT, T. D. AND DAWID, I. B. (1987). Cell-type-specific expression of epidermal cytokeratin genes duringgastrulation of Xenopus laevis. Genes and Dev. 1, 124-132.

JOHNSON, P. F. AND MCKNIGHT, S. L. (1989). Eukaryotictranscriptional regulatory proteins. Ann. Rev. Biochem. 58,799-839.

JONAS, E., SARGENT, T. D. AND DAWID, I. B. (1985). Epidermalkeratin gene expressed in embryos of Xenopus laevis. Proc. natn.Acad. Sci. U.S.A. 82, 5413-5417.

JONAS, E. A., SNAPE, A. M. AND SARGENT, T. D. (1989).Transcriptional regulation of a Xenopus embryonic keratin gene.Development 106, 399-405.

KARLSSON, S., BODINE, D. M., PERRY, L., PAPAYANNOPOULOU, T.AND NIENHUIS, A. W. (1988). Expression of the human /J-globingene following retroviral-mediated transfer into multipotentialhematopoietic progenitors of mice. Proc. natn. Acad. Sci. U.S.A.85, 6062-6066.

KAY, R. M., HARRIS, R., PATIENT, R. K. AND WILLIAMS, J. G.(1980). Molecular cloning of cDNA sequences coding for themajor a- and ^-globin polypeptides of adult Xenopus laevis. Nuc.Acids Res. 8, 2691-2707.

KRASNER, A. S., CHENG, A. K., DAWID, I. B. AND SARGENT, T. D.(1988). The gene encoding Xenopus embryonic epidermal keratinXK70A exhibits a hybrid type I-type II intron pattern. Nuc.Acids Res. 16, 8728.

KRIEG, P. A. AND MELTON, D. A. (1985). Developmentalregulation of a gastrula-specific gene injected into fertilizedXenopus eggs. EMBO J. 4, 3463-3471.

KUWABARA, M. D. AND SIGMAN, D. S. (1987). FootprintingDNA-protein complexes in situ following gel retardation assaysusing 1,10-phenanthroline-copper ion: Escherichia coli RNApolymerase-/ac promoter complexes. Biochemistry 26, 7234-7238.

MARCHUK, D., MCCROHON, S. AND FUCHS, E. (1985). Completesequence of a gene encoding a human type I keratin: sequenceshomologous to enhancer elements in the regulatory region of thegene. Proc. natn. Acad. Sci. U.S.A. 82, 1609-1613.

MAXAM, A. M. AND GILBERT, W. (1980). Sequencing end-labelledDNA with base-specific chemical cleavages. Methods Enzymol.65, 499-560.

MIYATANI, S., WINKLES, J. A., SARGENT, T. D. AND DAWID, I. B.(1986). Stage-specific keratins in Xenopus laevis embryos andtadpoles: The XK81 gene family. J. Cell Biol. 103, 1957-1965.

MOHUN, T. J., TAYLOR, M. V., GARRETT, N. AND GURDON, J. B.(1989). The CArG promoter sequence is necessary for muscle-specific transcription of the cardiac actin gene in Xenopusembryos. EMBOJ. 8, 1153-1161.

NIEUWKOOP, P. D. AND FABER, J. (1967). Normal table of Xenopuslaevis (Daudin). North-Holland Publishing Co., Amsterdam.

NORRIS, K., NORRIS, F., CHRISTIANSEN, L. AND FIIL, N. (1983).Efficient site-directed mutagenesis by simultaneous use of twoprimers. Nuc. Acids Res. 15, 5103-5112.

SARGENT, T. D., JAMRICH, M. AND DAWID, I. B. (1986). Cellinteractions and the control of gene activity during earlydevelopment. Devi Biol. 114, 238-246.

SIEBENLIST, U. AND GILBERT, W. (1980). Contacts betweenEscherichia coli RNA polymerase and an early promoter ofphage T7. Proc. natn. Acad. Sci. U.S.A. 77, 122-126.

STEINERT, P. M. AND ROOP, D. R. (1988). Molecular and cellularbiology of intermediate filaments. Ann. Rev. Biochem. 57,593-625.

SYMES, K., YAQOOB, M. AND SMITH, J. C. (1988). Mesoderminduction in Xenopus laevis: responding cells must be in contactfor mesoderm formation but suppression of epidermaldifferentiation can occur in single cells. Development 104,609-618.

WILUAMS, T., ADMON, A., LOSCHER, B. AND TJIAN, R. (1988).Cloning and expression of AP-2, a cell-type-specific transcriptionfactor that activates inducible enhancer elements. Genes andDev. 2, 1557-1569.

WINKLES, J. A., SARGENT, T. D., PARRY, D. A. D., JONAS, E. ANDDAWID, I. B. (1985). Developmentally regulated cytokeratingene in Xenopus laevis. Mol. Cell. Biol. 5, 2575-2581.

(Accepted 25 January 1990)