Proc. Nati. Acad. Sci. USAVol. 89, pp. 5809-5813, July 1992Biochemistry

In vivo protein-DNA interactions at hypersensitive site 3 of thehuman 8-globin locus control region

(chromatin structure/dimethyl sulfate/erythroleukemia cells/transcription factor GATA-1)

ERICH C. STRAUSS*t AND STUART H. ORKIN*t§*Division of Hematology/Oncology, Children's Hospital and the Dana-Farber Cancer Institute, Department of Pediatrics, Harvard Medical School,tHarvard-Massachusetts Institute of Technology Division of Health Sciences and Technology, and THoward Hughes Medical Institute, Boston, MA 02115

Contributed by Stuart H. Orkin, March 27, 1992

ABSTRACT The expression of 13-globin genes in develop-ing erythroid cells is dependent on distant, upstream regulatorysequences, known as the locus control region (LCR), which aremarked in chromatin by DNase I hypersensitive sites (HS-1 toHS-4). Linkage of the ,B-globin gene complex LCR or frag-ments surrounding core regions of 200-300 base pairs to thehuman (3-globin gene permits consistent, high-level expressionof the transgene in mice. To define the array of nuclear factorsinteracting with ,B-LCR HS-3, we have performed in vivodimethyl sulfate footprinting of the active HS-3 core in eryth-roid cells by a modified procedure that permits assessment ofprotein-DNA contacts at adenine, as well as guanine, residues.In vivo protein occupancy differs considerably from that pre-dicted from previous in vitro binding analyses. In vivo foot-printing detects protein binding at four sites recognized by theerythroid transcription factor GATA-1, at two CACC/GTmotifs, and at a single AP-1/NF-E2 site. The regulatoryclements occupied in vivo in HS-3 appear similar to thosedescribed previously in globin gene promoters and 3' enhanc-ers. These rindings suggest that the distinctive properties of theHS-3 region may be attributable to the organization of theseoccupied motifs and the consequent protein interactions, ratherthan to the binding of unique LCR regulatory factors.

Distant, upstream regulatory sequences termed locus controlregions (LCRs) (1) are necessary for expression of the humana- and p-like globin genes in developing erythroid cells (2, 3).LCRs serve to maintain chromatin in an open configurationand override inhibitory influences on gene expression (4).Current models suggest that high-level, developmentallyappropriate transcription of the individual globin genes isachieved through an interaction between their promoters orenhancers and LCR sequences (1, 5, 6).

In the human ,3-globin gene cluster the LCR is marked byerythroid-specific, DNase I-hypersensitive sites (HS-1 toHS-4) that lie 6-18 kilobases (kb) upstream of the embryonice-globin gene (7, 8). Linkage of the f3-LCR to the humanf3-globin gene permits consistent expression of the transgenein mice in a manner dependent on gene copy number, butindependent of integration site (2). The functional activity ofthe ,-LCR has been further subdivided into discrete coreregions of 200-300 base pairs (bp), each coincident with ahypersensitive site in native chromatin (9, 10). Most of theactivity of the ,8-LCR resides within the HS-2 and HS-3 coreelements. In transient transfection assays the HS-2 coreconfers erythroid-specific enhancer activity that is dependenton the integrity ofa dimeric AP-1 motif (11-13). However, nosimilar intrinsic enhancer activity is associated with HS-3.The f3-LCR core elements each contain one or more potentialbinding sites for the essential erythroid transcription factorGATA-1 (1, 14, 15), as well as AP-1/NF-E2 and CACC/GT

motifs, which bind a variety of erythroid and ubiquitousproteins in nuclear extracts in vitro (9, 10).As protein binding in vitro may not accurately reflect

binding-site occupancy in situ, the nature of in vivo proteinbinding to the core LCR elements has recently been ad-dressed in an effort to further understand the interactionsresponsible for their distinctive properties. Using ligation-mediated polymerase chain reaction (LMPCR) dimethyl sul-fate (DMS) footprinting (16), Reddy and Shen (17) and Ikutaand Kan (18) recently reported in vivo protein interactions atHS-2. To increase the analytical potential and informative-ness of in vivo footprinting, we have developed a procedure,termed GA-LMPCR in vivo footprinting, which permits de-tection of protein-DNA contacts at adenine, as well asguanine, residues (19). With this method, we previouslyexamined the 300-bp core of the upstream regulatory regionof the human a-globin locus in erythroid cells (19). In thisstudy, we describe the results of GA-LMPCR in vivo foot-printing off-LCR HS-3. These findings provide a frameworkfor understanding the functional properties of this element.

MATERIALS AND METHODSCell Lines and Culture. The interspecies human-mouse

somatic cell hybrid line Hull (20) was cultured in RPMI 1640medium supplemented with 10% fetal bovine serum. Thepresence ofhuman chromosome 11 ,-globin locus sequenceswas maintained by selection in hypoxanthine/aminopterin/thymidine medium in which 10 ,uM methotrexate was used inplace of aminopterin. HepG2 hepatoma and K562 erythro-leukemia cells were cultured as described (19).

Methylation and Isolation of Genomic DNA. Induced Hullcells were prepared by treatment with 1.5% dimethyl sulfox-ide (DMSO) for 48 hr prior to in vivo methylation with DMSby the procedure of Becker and Schutz (21). In vivo meth-ylated and control protein-free genomic DNAs were preparedas described (19).

Base-Specific DNA Cleavage and in Vivo Footprinting. Base-specific cleavage of methylated DNA at guanine and adenineresidues followed the procedure of Maxam and Gilbert (22),as modified (19). LMPCR in vivo footprinting was performedas described by Mueller and Wold (16) with the followingexceptions. To provide an appropriate founder population foranalysis of adenine residues, 5 ,ug of guanine/adenine-cleaved genomic DNA was used in the annealing reaction.Moreover, to facilitate compatibility with the melting tem-peratures of the PCR amplification primers used in thisanalysis, the sequence of the linker primer was modified to5'-GCAGTGACTCGAGAGATCTGAATTC-3'. Finally, as

Abbreviations: LCR, locus control region; LMPCR, ligation-mediated polymerase chain reaction; DMS, dimethyl sulfate;DMSO, dimethyl sulfoxide.§To whom reprint requests should be addressed at: Division ofHematology/Oncology, Children's Hospital, 300 Longwood Ave-nue, Boston, MA 02115.

5809

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

5810 Biochemistry: Strauss and Orkin

a b c d

1. 2 3 4

0)72)

____

e

1 2 3 4

-, p-,.,.._ -i (73)-717, -~ -0 m(72) GATA

GATA imomo i I-I.." ~-0o(64)

f

1 2 34

ammwmwt

mum,

p- I 4

*- 20'. W " " I"Hi-,om

___.

g

1 2 3 4

.,: "* h. . . -0(56)

-usmmn_n _ am__"04

_I_

1 2 3 4

urnWom mm

m0111 m8 -

M. _nmM

_~ -m*M m- _ __W-.0'

I 3 I

---- Vf220)

-Ia.G124AlA

(158) GATA a__m _

i .-. X:

---mm

_ __-'.-

_00=='""'

FIG. 1. In vivo DMS footprinting of the GATA sites in human HS-3. Top- and bottom-strand analyses are shown in a, c, e, and g and inb, d, f, and h, respectively. Lanes: 1, in vitro methylated protein-free K562 DNA; 2, in vivo methylated HepG2 DNA; 3, in vivo methylatedHull DNA (uninduced cells); 4, in vivo methylated Hull DNA (DMSO-induced cells). Protections and enhancements are represented by openand filled circles, respectively. Numbers in parentheses refer to nucleotides as shown in Fig. 5.

detailed previously (19), the amplification and labeling stepswere separated to optimize GA-LMPCR in vivo footprinting.This modification to the Mueller and Wold procedure (16)appears to be important in obtaining appropriate band inten-sities at adenine residues. Three cycles were used in thelabeling reaction.

Oligonucleotide Primers. All oligonucleotides were synthe-sized on an Applied Biosystems 380B DNA synthesizer andpurified by gel electrophoresis prior to use. For top- andbottom-strand LMPCR in vivo footprinting analysis, specificprimer sets were used for the Sequenase extension reaction(primer 1), the PCR amplification reaction (primer 2), and thelabeling reaction (primer 3).Primer set for top-strand analysis (sense):

1. GATGATCCTGAAAACATAGG2. AAACATAGGAGTCAAGGCACTTGCC3. GAGTCAAGGCACTTGCCCCTAGCTG

Primer set for bottom-strand analysis (antisense):1. GCCATGAAGAAGTCTATGACTG2. CTGTAAATTGGGAGCAGGAGTCTC3. GGGAGCAGGAGTCTCTAAGGACTTGGA

RESULTSIn Vivo GA-LMPCR DMS Footprinting of the Human

13-LCR HS-3. The copy number-dependent, integration site-

1 2 3 4

*0 (94)t- 0-(95)Icxcc

___ (98) C

w o(99)

_,_ __ oxloo00)

independent activity of HS-3 has been localized to a 225-bpDNA fragment (10). We performed GA-LMPCR in vivofootprinting of 300 bp encompassing this core. In noneryth-roid HepG2 cells, no detectable footprints were observed(Figs. 1-4). In addition, no protein occupancy was seen in thehuman erythroleukemia cell line K562 (data not shown),consistent with the absence of formation of HS-3 in thesecells (7). To examine human HS-3 in an adult erythroidenvironment, we employed mouse erythroleukemia cellscontaining a portion of human chromosome 11 (line Hull;ref. 20). A prominent, erythroid-specific hypersensitive sitecoincident with the HS-3 region has been demonstrated inthese cells (20). Upon induced maturation with DMSO, thesecells coexpress mouse and human 8-globins. In vivo proteinoccupancy within HS-3 in these cells is displayed in Figs.1-4, summarized in Fig. 5, and compared with in vitrofindings in Fig. 6.GATA Elements. HS-3 contains four consensus GATA

sites [(T/A)GATA(A/G)]. Three of these, as well as anonconsensus GATA site (AGATGG), bind the transcriptionfactor GATA-1 in vitro (10). Potential binding to one of theconsensus sites was obscured in vitro by protein binding to anoverlapping CACC/GT element (10). In vivo footprintinganalysis detected protein occupancy at all four consensusGATA motifs in both uninduced and DMSO-induced Hullcells (Fig. 1). Following DMSO-induced maturation of Hullcells, additional protections were detected at two of these

1 2 3 4

-CG Is8-O C; 188) CACA

AR (191'9

-_ -_ _ -*

FIG. 2. In vivo DMS footprinting of the top strand of the two bound CACC/GT sites. The upstream and downstream elements are depictedin a and b, respectively. Lanes are as indicated in Fig. 1.

--o M1211 GoAI'--igiq

Proc. Natl. Acad. Sci. USA 89 (1992)

_ A:WAw

am..,mi: ;qW. 'A W-.-,-1f.1.

goft4mm",wool 000

Proc. Natl. Acad. Sci. USA 89 (1992) 5811

a

1 2 3 4

Hfl~fl -

"*__-*~~~( 4)

-MS-S SUE ! 0P~- 0(13) AP-1|11 -Add o(13) NF-E2;I'aa-uAO - (18)

- C-am >

b1 2 3 4

-m e- -.. ~-

emsnew 4tm..

.in earn m -6(9)WP-*WO-0(10)

FIG. 3. In vivo DMS footprinting of the top (a) and bottom (b) strands of the upstream AP-1/NF-E2 site. Lanes are as indicated in Fig. 1.

GATA elements (Fig. id, lane 4, A-119, A-121, and G-122;Fig. le, lane 4, G-156). No in vivo footprints were observedat the nonconsensus GATA site (data not shown).CACC/GT Elements. Seven CACC/GT motifs are con-

tained in HS-3; six of these elements correspond to in vitroprotein binding sites as demonstrated by DNase I footprints(10). Two of the sites are conserved between humans andmice (T. Ley, personal communication); however, only one

of these elements is conserved among humans, mice, andgoats (23). In uninduced Hull cells, only single in vivocontacts were detected at two CACC/GT elements (Fig. 2a,lane 3, G-100; Fig. 2b, lane 3, G-191). Following DMSOinduction, both of these sites revealed additional protein-DNA interactions (Fig. 2 a and b, lane 4; Figs. 5 and 6). Ofthese two occupied sites, one (site 187-191) corresponds tothe highly conserved motif. The other is unique to the humanf-LCR.AP-1/NF-E2 Element. Upstream, just outside the HS-3

core defined by Philipsen et al. (10), is a single AP-1/NF-E2binding site (24) that is recognized in vitro by murine NF-E2(N. C. Andrews and S.H.O., unpublished data). In vivofootprinting at the AP-1/NF-E2 motif, reflected by hyper-sensitivity at A-9 and A-10 (Fig. 3b), was weak in uninducedHull cells. Following DMSO induction, additional contactswere evident, particularly a protection within the core bind-ing sequence at A-13 (Fig. 3a). The absence of in vivofootprints at this site in HepG2 cells is consistent with ouranalysis ofAP-1/NF-E2 motifs in the human and mouse HS-2(data not shown) and the a-LCR (19). In these instances, wehave not detected in vivo footprints in nonerythroid cells.These findings contrast with previous studies that show invivo protein interactions in both erythroid and nonerythroid(e.g., HeLa) cells at multiple sites in HS-2, including theAP-1/NF-E2 elements (17, 18). Our results are most consis-tent with a chromatin structure that renders these motifsinaccessible in nonerythroid, non-globin-expressing cells.A+T-Rich Sequence. In vivo footprinting additionally re-

vealed erythroid-specific effects on several nucleotides lo-calized to a discrete A+T-rich sequence in the central regionof HS-3 (Fig. 4). This region does not contain recognizableDNA-binding-protein motifs. Gel shift experiments using

a

1 2 3 4

-0(139)

.n.ir sC}-O0(141)

_ _

suma S0

b1 2 3 4

o(140)_ 0(138)mm~s'-"

.z____~~FIG. 4. In vivo DMS footprinting of the top (a) and bottom (b)

strands of the central A+T-rich region. Lanes are as indicated inFig. 1.

MEL cell nuclear extracts have failed to reveal specific invitro binding of proteins to this region (data not shown). Inprevious studies of the a-LCR element (19) and upstream ofan occupied GATA motif in the human HS-2 core (Fig. 7), wehave detected similar patterns of altered DMS reactivity invivo at distinct sequences that do not resemble DNA-binding-protein motifs. As in HS-3, these altered regions show nocorresponding protein binding in vitro (data not shown).Hence, the apparent contacts in the A+T-rich sequence ofHS-3, and similarly the contacts in HS-2 and a-LCR, mayreflect a local chromatin structure consequent to interactionsof proteins assembled on their respective core element.

DISCUSSIONThe in vivo footprinting analysis of HS-3 described herecomplements other studies of LCR elements (17, 18) andprovides evidence for changes in protein-DNA interactionswithin cells that are not apparent in binding assays in vitro.As a first approximation the HS-3 core is simple in itsorganization and comprises alternating motifs of two types:GATA and CACC/GT (10). Only a subset of the potentialbinding sites are occupied in vivo in the mouse erythroleu-kemia cell environment (Fig. 6). These bound sites arearranged symmetrically about a central A+T-rich region (seeFigs. 5 and 6). Flanking this area on each side are two GATAsites surrounding a CACC/GT motif. As the integrationsite-independent activity of HS-3 has been localized to the225 bp downstream of an AP-1/NF-E2 site (10), we speculatethat this configuration of bound GATA-1 and CACC/GTprotein(s) (perhaps Spl) is uniquely suited to endow theelement with its LCR-like properties. This conclusion isconsistent with functional, in vitro protein binding and within vivo footprinting analyses of HS-2 and the a-LCR (9, 10,17-19, 25, 26). In these LCR elements the active cores display

AP-1/NF-E2+1 * 0

CAAGGAATTT rGACTCAPCAGTTCCTTAAAMkCTGAG*GT

GATA0 0

GATA IT

00 CACC

00 0@0

C

GT

GACATGGGAA ap~akGGCTGTACCCTT *CAXCCT

A&CACAAG&C CCTCACGGTG

TTGTGTTCTG GGAGTGCCAC

CACC CACGT GT

00 @00

GGTCCCCC

ACTTTGCGAG TC 60

TGAAACGCTC GC

CACCGT GATA

0 0

TCA=GGCTGG 120AGTCCCG&CC

; GATA~~GT GT

AGGCTGGCAG Is8TCCGACCG;TC GCTdC AlATA G a:

GATA

CTCCCAGGAG C GTAGAGGGCA 260

GAGGGTCCTC TACT CATCTCCCGT0 0

GOCACASCAT AGCAGCATTT 5TCATTCTAC TACTACATGG 300

CCGTGTCGTA TCGTCGTAAA AAGTAAGATG ATGATGTACC

FIG. 5. Summary of altered DMS reactivities in HS-3. Protec-tions and enhancements are represented by open and filled circles,respectively. Arrows denote protections that do not correspond to a

known or identified protein binding site.

AP-1NF-E2

Biochemistry: Strauss and Orkin

5812 Biochemistry: Strauss and Orkin

GATACACQ

CACtlGATA GT

FIG. 6. Schematic comparison of in vitro (10) and in vivo footprinting analysis of HS-3. Potential binding sites, where no footprints were

observed, are indicated by open boxes. Protein occupancy of regulatory motifs is indicated by filled boxes. The hatched box depicting theAP-1/NF-E2 motif indicates that in vitro binding was not assessed by Philipsen et al. (10).

multiple binding sites of only three types [GATA, AP-1/NF-E2, and CACC/GT], many of which exhibit protein occu-

pancy in vivo. Furthermore, the integration site-independentactivity of HS-2 has been localized downstream of an AP-1/NF-E2 dimer, to a region including a single GATA andCACC/GT motif (25).Our in vivo footprinting analysis demonstrates that protein

complexes formed at HS-3 are developmentally regulatedboth with respect to cell type and with respect to stage oferythroid maturation. First, HS-3 exhibits no in vivo foot-prints in a nonerythroid cell environment, such as HepG2(Figs. 1-4). Second, HS-3 is not formed in human K562 cells,a line of mixed hematopoietic lineages expressing embryonicand fetal hemoglobins (refs. 7 and 18; and unpublished data).Third, and most important, we observe a limited in vivofootprint pattern in HS-3 in uninduced Hull cells, whichbecomes more complex upon DMSO-induced maturation. Inuninduced cells, which do not express appreciable ,B-globin,contacts are evident on at least one strand at the GATAmotifs. In vivo footprints at the CACC/GT motifs, as well asthe upstream AP-1/NF-E2 site, are less evident in uninducedcells but are readily discernable following DMSO-inducedmaturation (Figs. 2 and 3). Thus, complexes formed at HS-3are altered during induced cellular maturation. As the inter-action of LCRs with promoters or enhancers of the down-stream individual globin genes is thought to be involved inestablishing their transcriptional competence (1), dynamicinteractions of this type may influence the complex devel-opmental pattern ofglobin gene expression. As a corollary tothis hypothesis, it follows that seemingly subtle changes inthe levels (or availability) or properties of a few proteins,such as GATA-1, AP-1/NF-E2, or CACC/GT factors, maysubstantially influence LCR-gene interactions so as to con-tribute to switching of hemoglobins during development.

'I I t 4

-MGM =o C;A1-A

_ _ _9= lw--a

so 4Oms

1 2 3 4__II

......<<>s>wa_

_-- off

_- ""N__0!

l

t 0CTAGAGTGATGACTc GTCCCCGATCTCACTACTGA CCAGGGG

4 00

FIG. 7. In vivo footprinting of the top (Right) and bottom (Left)strands of a discrete region of hypersensitivity adjacent to a GATAelement downstream of the dimer AP-1/NF-E2 motifs in humanP-LCR HS-2. Lanes are as indicated in Fig. 1. Primer set fortop-strand analysis: 1, CCATTTTCTTTATGATGCCG; 2, GC-CGTTTGAGGTGGAGTTTTAGTCAG; 3, GGTGGAGTTT-TAGTCAGGTGGTCAGC. Primer set for bottom-strand analysis: 1,ATTTCTGTGTGTCTCCATTAG;2,CATTAGTGACCTCCCAT-AGTCCAAG; 3, GACCTCCCATAGTCCAAGCATGAGC.

Remarkably, evidence is emerging that all active globinLCR cores contain a limited array of protein binding sites,which bind a limited group of regulatory factors. Moreover,it appears that the same three factors, GATA-1, AP-1/NF-E2, and CACC/GT-binding protein(s), assemble on DNA inthe promoters and 3' enhancers of the individual globingenes. It is not yet evident which aspects of the configurationof binding motifs and interacting bound proteins uniquelyendow LCRs with their dominant properties for chromatinorganization. In our in vivo footprinting analyses ofHS-3, thea-LCR region (19), and HS-2, we have consistently detectedone set of protections or enhancements that do not localizeto known protein binding sites or to sequences bound in vitroby nuclear extract proteins (see Figs. 4 and 7). Although wecannot exclude the possibility that in vitro assays fail todetect low-abundance or unstable proteins that interact withthese sequences, we favor the view that these in vivo alter-ations reflect local chromatin structure induced by the bind-ing and interaction of neighboring proteins. In both thea-LCR and HS-3 these interesting regions are surrounded bysymmetrically positioned, occupied GATA sites.

We are grateful to Doug Higgs and Bill Wood for providing theHull cells for this analysis and to Nancy Andrews for discussion andcomments. This work was supported in part by a grant from theNational Institutes of Health to S.H.O. and a grant from the Johnsonand Johnson Research Awards to E.C.S. and S.H.O. through theHarvard-Massachusetts Institute of Technology Division of HealthSciences and Technology Program. S.H.O. is an Investigator of theHoward Hughes Medical Institute.

1. Orkin, S. H. (1990) Cell 63, 665-672.2. Grosveld, F., van Assendelft, G. B., Greaves, D. R. & Kollias,

B. (1987) Cell 51, 975-985.3. Higgs, D. R., Wood, W. G., Jarman, A. P., Sharpe, J., Lida,

J., Pretorius, I.-M. & Ayyub, H. (1990) Genes Dev. 4, 1588-1601.

4. Felsenfeld, G. (1992) Nature (London) 355, 219-224.5. Behringer, R. R., Ryan, T. M., Palmiter, R. D., Brinster, R. L.

& Townes, T. M. (1990) Genes Dev. 4, 380-389.6. Enver, T., Raich, N., Ebens, A. J., Papayannopoulou, T.,

Costantini, F. & Stamatoyannopoulos, G. (1990) Nature (Lon-don) 344, 309-313.

7. Tuan, D., Solomon, W., Li, Q. & London, I. M. (1985) Proc.Natl. Acad. Sci. USA 82, 6384-6388.

8. Forrester, W. C., Takegawa, S., Papayannopoulou, T., Stai-atoyannopoulos, G. & Groudine, M. (1987) Nucleic Acids Res.15, 10159-10177.

9. Talbot, D., Philipsen, S., Fraser, P. & Grosveld, F. (1990)EMBO J. 9, 2169-2178.

10. Philipsen, S., Talbot, D., Fraser, P. & Grosveld, F. (1990)EMBO J. 9, 2159-2167.

11. Ney, P. A., Sorrentino, B. P., McDonagh, K. T. & Nienhuis,A. W. (1990) Genes Dev. 4, 993-1006.

12. Ney, P. A., Sorrentino, B. P., Lowrey, C. H. & Nienhuis,A. W. (1990) Nucleic Acids Res. 18, 6011-6017.

13. Moi, P. & Kan, Y. W. (1990) Proc. Nati. Acad. Sci. USA 87,9000-9004.

14. Evans, T. & Felsenfeld, G. (1989) Cell 58, 877-885.15. Tsai, S.-F., Martin, D. I. K., Zon, L. I., D'Andrea, A. D.,

AP-1NF-E2

.m>.1E1 liN )

CACCGT

Proc. Natl. Acad. Sci. USA 89 (1992)

ZI--

GATA I U H --I-I .-m

GT

IT!;i

Biochemistry: Strauss and Orkin

Wong, G. G. & Orkin, S. H. (1989) Nature (London) 339,446-451.

16. Mueller, P. R. & Wold, B. (1989) Science 246, 780-786.17. Reddy, P. M. S. & Shen, C.-K. J. (1991) Proc. Nat!. Acad. Sci.

USA 88, 8676-8680.18. Ikuta, T. & Kan, Y. W. (1991) Proc. Natl. Acad. Sci. USA 88,

10188-10192.19. Strauss, E. C., Andrews, N. C., Higgs, D. R. & Orkin, S. H.

(1992) Mol. Cell. Biol. 12, 2135-2142.20. Dhar, V., Nandi, A., Schildkraut, C. L. & Skoultchi, A. I.

(1990) Mol. Cell. Biol. 10, 4324-4333.

Proc. Nat!. Acad. Sci. USA 89 (1992) 5813

21. Becker, P. B. & Schutz, G. (1988) Genet. Eng. (NY) 10, 1-19.22. Maxam, A. M. & Gilbert, W. (1977) Proc. Natl. Acad. Sci.

USA 74, 560-564.23. Li, Q., Zhou, B., Powers, P., Enver, T. & Stamatoyanno-

poulos, G. (1990) Proc. Natl. Acad. Sci. USA 87, 8207-8211.24. Mignotte, V., Wall, L., deBoer, E., Grosveld, F. & Romeo,

P.-H. (1989) Nucleic Acids Res. 17, 37-54.25. Talbot, D. & Grosveld, F. (1991) EMBO J. 10, 1391-1398.26. Jarman, A. P., Wood, W. G., Sharpe, J. A., Gourdon, G.,

Ayyub, H. & Higgs, D. R. (1991) Mol. Cell. Biol. 11, 4679-4689.