Proc. NatL Acad. Sci. USAVol. 78, No. 12, pp. 7652-7656, December 1981Genetics

Mutations in the lacY gene of Escherichia coli define functionalorganization of lactose permease

(membrane transport/membrane protein/quaternary structure/negative dominance/active site)

M. MIESCHENDAHL, D. BUCHEL, H. BOCKLAGE, AND B. MULLER-HILLInstitut ffir Genetik der Universitit zu Koin, Cologne, Federal Republic of Germany

Communicated by Eugene P. Kennedy, August 24, 1981

ABSTRACT Mutations in the lacY gene of Escherichia colihave been used to analyze the functional organization of lactosepermease. Deletions suggest that the NH2 terminus oflactose per-mease is not essential and can be replaced by residues of the cy-toplasmic enzyme (3-galactosidase. Negative dominant mutationsin the lacY gene can be explained by the assumption that mem-brane-associated lactose permease is active as a dimer or oligo-mer. The map positions of these mutations and other point mu-tations that lower or alter the sugar specificity define regions oflactose permease involved in sugar or proton binding andtransport.

Lactose permease transports and accumulates galactosidesthrough the inner membrane into the cytoplasm ofEscherichiacoli (1). Genetic analysis of the lacY gene coding for lactose per-mease has shown that the permease is made ofone polypeptidechain (2). The sequence ofthe Y gene and its reading frame havebeen determined (3). The NH2-terminal protein sequences oflactose permease synthesized in vivo and in vitro are in agree-ment with the DNA sequence and define the start codon (4).According to these analyses, lactose permease is a protein com-posed of 417 residues.

Lactose permease is well analyzed functionally. Mitchell'schemosmotic theory applies to it. It transports one proton withevery molecule of galactoside (5). Vesicles containing it neednot be energized to bind galactosides (6). It is symmetricallyaccessible in the membrane to its substrates (7). The insolubilityof lactose permease, however, has rendered its biochemicalanalysis difficult. It has at least one essential SH group whichcan be protected by various galactosides (8). This property hasbeen used as assay for its partial purification (8). It has beenlabeled by a photoaffinity label (9). But so far, nothing is knownabout the functional organization ofthe domains a protein ofthissize is expected to have.We used genetic methods to elucidate the molecular mode

of action of lactose permease. A similar analysis has been per-formed previously with the lac repressor producing I gene inorder to define the functional domains of lac repressor (10, 11).As in the case of the I gene we isolated negative dominant mu-tants ofthe Y gene. They indicate a dimeric or oligomeric struc-ture for lactose permease. We isolated mutants of the Y genewith decreased affinity for lactose and increased affinity formaltose (which is normally not transported by lactose permease)to define the sugar-binding site. We also isolated a deletionmutant that replaces the intercistronic DNA in front of the ri-bosomal binding site of the Y gene and others that replace theextreme NH2 terminus of lactose permease with the NH2 ter-minus of f3-galactosidase. Random deletions (2) and defineddeletions with known end points produced in vitro by partial

digestion with Hha I and subsequent ligation were used to mapthe mutations.

MATERIALS AND METHODSEnzymes and Chemicals. Hinfi and Hha I were purchased

from New England BioLabs. Hpa II, EcoRI, and the Klenowfragment of DNA polymerase I were from Boehringer Mann-heim. Chemicals for electrophoresis were obtained from BDHBiochemicals (Poole, England). [32P]dATP was purchased fromAmersham Buehler. Dimethylsulfate and hydrazine werekindly provided by E. Vogel (Institut fur Organische Chemie,Cologne). 5-Bromo-4-chloro-3-indolyl (8-D-galactoside (X-Gal)and isopropyl thiogalactoside were obtained from Bachem FineChemicals (Torrance, CA). Plates and media were as described(2). Other chemicals were purchased from Sigma (Munich) orMerck (Darmstadt).

Production and Analysis of the Hha I Deletions in the lacYGene. The 2100-base-pair EcoRI fragment containing the Ygene (3) was inserted into the plasmid pUR2 (12). Recombinantplasmids were selected for growth on melibiose. plates at 420Cin a bacterial strain carrying a lacpro deletion. The DNA of oneof the recombinant plasmids was partially cleaved by Hha I.Linear DNA molecules not more than 500 base pairs shorterthan the original linearized plasmid were purified by electro-phoresis and subsequent isolation from a low-melting agarosegel and circularized by ligation. Transformants of a recA lacY-strain (ara- [lacpro], thi strA recA rif, F'lacIQZ'Y-, MAA13A' pro') were tested for growth on lactose plates. Only thosetransformants that could not grow on lactose plates containedplasmids with internal deletions in theY gene, which was shownby the examination of the EcoRI-cleaved plasmids. The dele-tions were mapped by restriction analysis with Hpa II.

Isolation and Mapping of the y-d Mutations. A logarithmic-phase culture ofthe heterogenote lac+ L8 recA/F'lacZ- M330bY'pro' growing in rich medium was treated with the mutagenN-methyl-N'-nitro-N-nitrosoguanidine for 3 min at a final con-centration of 40 Ztg of mutagen per ml. The recA mutation didnot abolish mutagenesis as control experiments'showed. To stopfurther mutagen activity the culture was diluted 1:200 in liquidrich. medium and grown overnight at 30'C. Cells were platedapproximately 100 cells per plate on minimal glucose platescontaining streptomycin and incubated for 48 hr at 30'C. Col-onies were then replica plated-on minimal lactose and minimalglucose plates. Colonies that grew on minimal glucose platesbut not on minimal lactose plates after incubation at 30'C for48 hr were purified and the episome was crossed into a nalidixicacid-resistant strain carrying a [lacpro],, deletion. Colonies thatwere unable to grow in this nalidixic acid-resistant backgroundon minimal melibiose plates at 42C and which therefore hada mutation in the lacY gene of the F'lac Z-AM330b Y' pro'episome were further analyzed. The F'lac Z-AM330b Y- pro'episomes were crossed back again into a lac+L8 recA strain and

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

7652

Proc. Nati. Acad. Sci. USA 78 (1981) 7653

the negative dominant character of the episomal lacYd muta-tion was verified by the failure of the lac+L8 recAlF'lacZ-AM330 y-d pro+ heterogenote to grow on minimal lac-tose plates. For mapping, the episomes were crossed into a setofchromosomal lacY- deletions. The lacY-"/F'lacZ-M 33ObY dpro+ heterogenotes were grown overnight in liquid rich me-dium, and drops of these overnight cultures were spotted onminimal melibiose plates; recombination was analyzed after 48-hr incubation at 420C.

RESULTSDeletion of the Intercistronic Region Between lacY and

lacZ. The intercistronic region between the lac Z and lac Ygenes (3) includes a typical Shine-Dalgarno (13) sequence, T-A-A-G-G-A, which precedes the ATG start codon for the en-zyme by eight base pairs. A stem and loop structure can be con-structed from the intercistronic region beginning with the threeochre codons at the end of the Z gene and the region coding forthe NH2 terminus of lactose permease (3). This structure hasno apparent function because all of the intercistronic DNA pre-ceding the Shine-Dalgarno sequence can be deleted withoutdecreasing the expression of lactose permease. We concludethis from the DNA sequence of the internal Z deletion M330bwhich lacks the DNA from codon 38 of the Z gene to the firstbase pair of the Shine-Dalgarno sequence (Fig. 1). M330b wasisolated as a revertant of the strongly polar ochre mutation U118in codon 18 of the Z gene on melibiose at 420C (17). The activityof lactose permease is the same in mutant and wild type. Thus,the Shine-Dalgarno sequence is sufficient to initiate normaltranslation of the lactose permease.

Fusions of the lacY Gene. We wondered whether the wholeintercistronic region between the Z and Y genes and maybeeven the first codons of the Y gene could be replaced by othersuitable DNA sequences without impairing the activity of lac-tose permease. We used Z-Y' revertants of the Y- mutantMAB16 (2) to study this. MAB16 is a nonsuppressible (2), non-polar Y- mutant. It maps very close to the Z gene. Some of itsZ+Y+ revertants are temperature sensitive, suggesting thatMAB16 lies in the structural part (i.e., in one of the first codonsof the Y gene).

In order to isolate ZdelY+ revertants from Z+Y-MAB16 weplated 0.1 ml of 1000 independent overnight cultures on me-libiose plates containing the 13-galactosidase indicator 5-bromo-4-chloro-3-indolyl ,t-D-galactoside (18). The plates were incu-bated for 2 days at 420C and screened for white revertant col-onies. We found six Z-Y+ revertants. Fig. 2 shows the mappositions of these mutations. One mutant arose by a doubleevent. The five others carried deletions that extended from theY proximal end of the Z gene into or beyond the Z gene. Onedeletion ended in the middle of the Z gene and presumablyfused a Z restart to lactose permease. Two others ended closeto the lac operator in the Z gene. One of them (MAB16RZ-36)still contained codon 23 of the Z gene [i.e., it recombined withZ-1 (B. Gronenborn, personal communication; ref. 19)]. Theother deletion (R26) lacked all known Z markers including U118but left the I gene functionally intact. Three deletions (R2, R22,R23) destroyed I gene activity. They presumably ended in theI gene because similar q--Z+ fusions have not been found toextend beyond the start codon of the I gene (20).

Revertant MAB16RZ-36 containing codon 23 of the Z genewas used to test whether the NH2 terminus of B-galactosidasemay be fused to functional lactose permease. We crossed theochre mutation Z-U118 which lies in codon 18 of the Z gene(21) into the revertant and thereby destroyed the lactose per-mease activity. For this purpose we plated the heterogenoteZ-U118/F'lacproMAB16RZ-36 on lactose plates and isolated

revertants. In some ofthe revertants the F'lacproMAB16RZ-36had become permease negative by reciprocal recombination asshown by crossing into a (lacpro)&,1 background. The simulta-neous presence of the Z deletion MAB16RZ-36 and the U118mutation on the F'lacpro episome was verified by recombi-nation analysis with Z- point mutations (Fig. 2).

Negative Dominant Mutations in the lacY Gene. The exis-tence of negative dominant constitutive mutations (I-d) in thelacI (repressor) gene could be interpreted by assuming lac re-pressor to be an oligomer (22). In this interpretation, one in-active subunit is sufficient to alter or even abolish the functionof the tetramer. Mapping (23) and sequence determination (24)of the -d mutants suggested that only the NH2 terminus of lacrepressor was involved in operator binding and recognition.This was fully confirmed later by in vitro experiments using theNH2-terminal headpiece of lac repressor in lac operator shield-ing experiments (25)..The symmetrical behavior (7) of lactose permease in kinetic

experiments led us to suspect lactose permease to be dimericor oligomeric. Before induction, only a few molecules of lactosepermease are present in the membrane of each cell. Thus, un-der suitable conditions, negative dominant y-d mutants may beisolated. By definition, a y-d mutant permease would have lostthe capacity to bind or transport galactosides or protons withoutlosing the capacity to enter the membrane and to aggregate. Themap positions of such mutants therefore outline the regions ordomains of sugar and proton binding and transport.

In order to decrease the concentration of active lactose per-mease as much as possible, we used the lac promoter mutantL8 (26) as the Y' wild type allele in the construction of the het-erogenote. The promoter mutation L8 reduces the expressionof the lac operon to 8% without abolishing growth on lactose(26). We isolated four y-d mutants on an F'lacproZ-M330bepisome. They inhibit, in a recA background, growth of lac+L8on lactose and even on lactose plus isopropyl thiogalactoside(1 mM). We also constructed controls carrying other Y- mu-tations of the F'lacproZ-M330b episome. They were not dom-inant, indicating that dominance could not be produced withjust any Y- mutation.One of the mutants (y-d) mapped in deletion group V (Fig.

3). The nonsense mutation NE4 had been shown to have anexchange in codon 33 (3) and to map in deletion group IV (2).Thus Y-di has a defect somewhere between codons 33 and 43,assuming a statistical distribution of the deletions. The DNAbetween codons 33 and 45 of lactose permease codes for a smallhydrophilic region between two long lipophilic sequenceswhich presumably anchor lactose permease in the membrane(Fig. 3).One other yd mutation maps in deletion group XXI which

corresponds to the region between codons 152 and 191 (Fig. 3).This is not far from the cysteine which is shielded by thiodi-galactoside against alkylation with N-ethylmaleimide (K. Bey-reuther, personal communication). The last two y-d mutationsmap in deletion group XXVIII somewhere between codons 191and 279. If we assume a random distribution of the deletions,the exchanges are to be found somewhere between residues 250and 279.

Mutations in the lacY Gene Which Change Affinity andSpecificity of Lactose Permease. Another way to look for mu-tations that make lactose permease specifically defective insugar or proton binding and transport is to look for mutationsthat decrease or alter the specificity of lactose permease. Wescreened our collection of "leaky" (i.e., easily reverting) Y-mutants (2) for mutants that grew normally on a high concen-tration oflactose. Usual lactose plates contain 5mM lactose. Forour screening we used plates containing 0.1 M lactose. We

Genetics: Mieschendahl et al.

7654 Genetics: Mieschendahl et aL

A A

__-

start of I Y:laccY A'

r. X

e -

0-

etw

Io

U,

-ninwild type

lacZ1021

TyrGlnLeuValTrpCysGlnLysTerTerTer

lacY

MetTyrTyrLeuLysAsn-lThrAsnPheTro

MetTyrTyrLeuLysAsnThrAsnPheTrp

20 30 38

lacZ lacy

M330b

FIG. 1. Nucleotide sequence of the area surrounding the lacZ deletion M330b. (Upper) Autoradiogramof a sequence-determination gel. AHinfI/Hpa II fragment of plasmid pGM 21 (14) was isolated. This fragment contained sequences coding for NH2-terminal amino acids of P-galactosidasebeginning at codon 5 of the lacZ gene and the genetic material coding for the 81 NH2-terminal amino acids of lactose permease. The Hinfl-end atcodon 5 of lacZ was labeled at the 3' position by filling in with [a-32P]dATP and Klenow polymerase I (15). The nucleotide sequence was determinedaccording to the method of Maxam and Gilbert (16). The complementary sequences corresponding to the U118 ochre codon in lacZ and theShine-Dalgarno box and the ATG start codon of lacY are indicated. (Lower) Nucleotide sequence of the mutant M330b (below) compared to thenucleotide sequence of the wild type. (above). The amino acids or codons are numbered according to their position in the protein or gene. Boxes,Shine-Dalgarno box of the Y gene; brackets, nontranslated codons of the Z gene; Ter, termination codon.

found 18 mutants that grew on these plates but not on the usuallactose plates. We called these mutants y-K to indicate that theymight be altered in their binding constants (K) for galactosidesor protons.One of the YK mutations (MUB7) mapped in deletion group

III. This implies that the mutation lies somewhere before codon33 (Fig. 3). The other y-K mutations mapped in the region cod-

ing for the COOH-terminal part of lactose permease-i.e., indeletion groups XXV to XXXV (Fig. 3). Thus, ally-K mutationsbut one are due to exchanges between codons 191 and 360 oflactose permease.A mutant lactose permease that could transport maltose has

been described (24). Because the mutant has not lost the ca-pacity to transport lactose and melibiose it could not be

4 Ulla

Proc. Nad Acad. Sci. USA 78 (1981)

.*

Proc. Natl. Acad. Sci. USA 78 (1981) 7655

I IPO1 Z3

U118Z6

AM330 bz

l IZ1O Z15

R36R2, 22, 23, 26

y

Z20

IAZ27

MAB16

R24

FIG. 2. Isolation and mapping of lacZ-Y' revertants of lacY-MAB16. Samples of 1000 independent overnight cultures of the nonsuppressibleY- mutation MAB16 mapping closest to Z (2) were tested. Z-Y' revertants were isolated after 48 hr of incubation at 42TC on melibiose indicatorplates. The episomes were crossed into a set of strains containing chromosomal Z- point mutations (18) and recombination was analyzed after 48-hr incubation at 37TC on lactose plates.

mapped. We looked for other such mutants. A deletion of themaltose transport system (maiB 107) (28) was crossed in a (lac-pro),,, strain into which an F'lacproI+Z+Y' episome was intro-duced. In this background, a strain having the genotypeF'IZ+Y' can grow on 0.1 M maltose plus 1 mM isopropyl thio-galactoside but not on 5 mM maltose plus inducer. Thus, weplated 0.1 ml of 50 independent overnight cultures on platescontaining 5 mM maltose and 1 mM isopropyl thiogalactoside.

From each plate one revertant was picked and purified. All re-vertants could still grow on lactose and melibiose at 42°C andbe transferred with the episome the capacity to make a (lac-pro)djmalB 107) strain grow on plates containing 5 mM mal-tose plus inducer.

If growth on 5 mM maltose can be acquired relatively easilythrough mutation in the lac operon, one may ask whether someof the yK mutants have acquired the capacity to grow on plates

26 22 24 11 11 9 9 9 11 15a)JEJEDO El CO QEI --

150 200

i .. I ... .,... .I amino acid+ i- i I * i 1 1 1 * . 1 f 4 residues250 300 350 400

279 312 380 .H ca ~~~~1;, i __Hho I cuts

5,33,35.45,6510,15.39 , i13,55

---,a 27 deletions

40,48,59,67,69

51,62

K XXXXXXX X X*

== »>> >>-Kx x xKK KK KK k x x x x x ^

I I""' " "" 'I" 'I"y-dl y-d2 Y-d3,4

K >> >K K KKxKcK xK KKK*-x

_=f_ ll> X X _=1> >>iIXXX X X X X X XJ XX *

,

Xx x xx xx Y Y Y _wx YYxx

AE43I IAC43AV38 AJ33AE41 AE10AD47 AF34

4AA36 NP4AN14 AV40AA22MAB20AG47

AJ36AG38

FIG. 3. Schematic illustration of the genetic and physical map of the Ygene and of lactose permease. (a) Hydrophobic regions of lactose permease(3) are shown in boxes. The numbers above the boxes indicate the number of residues comprising each hydrophobic region. The numbers shouldbe regarded as indicating minimal length; particularly the fourth and fifth hydrophobic regions can be seen to be substantially larger. (b) Scalerepresenting amino acid positions in lactose permease. (c) Restriction map of the Y gene for Hha I. The numbers above the arrows indicate thecorresponding codon positions of the Y gene. Internal deletions with known end points constructed by in vitro mutagenesis are shown below therestriction map. The deletions are numbered according to the numbers of different isolates. (d) Distribution of the 36 deletion groups of the Y geneand y-d mutations. A correlation between the genetic and the physical map was obtained by recombination experiments with point mutations ofthe Y gene (2) and the internal deletions with known end points. One or two point mutants of each deletion group of anF'lacpro episome were crossedinto Filacprodj strains containing plasmids with internal Y gene deletions. Recombination between point mutations and deletions that were indifferent regions of the gene led to an Y+ phenotype that could be tested by examination for growth on lactose plates. In regions defined by the endpoints of deletions that comprise more than one deletion group, the distribution of the deletions groups was assumed to be statistical. In the regionfrom codon 389 to 417 only one mutation of the XXXV deletion group was found (*). For details see ref. 27. (e) y-K mutations (see text).

0b)

c )

50 100

. 25,78

7, 34,50,6 1.

-~-N---Kr

d )

e)

MUB7

Genetics: Mieschendahl et aL

77 If ly? IT

= >5),._xX-= -= > > ; 5: ; X

- = > 5:3xx;;c- x x x x x x X

7656 Genetics: Mieschendahl et aL

containing 5 mM maltose and isopropyl thiogalactoside. Wechecked all the yK mutants for growth on such plates. Wildtype and all but one (AJ33) of the Y` mutants did not grow.

AJ33 mapped in deletion group XXVIII, corresponding to a lo-cation of the mutation close to codon 265. Five of the yK mu-

tants (MUB7, AV38, AJ36, MAA36, and AN14) mapping indeletion groups XXXI to XXXII (Fig. 3) had kept the capacityto grow on 0.1 M maltose plus inducer, thus showing a possibledifferential destruction of lactose versus maltose transportcapacity.

DISCUSSIONStudy of the revertants of the Y- mutation MAB16 allows theconclusion that the NH2 terminus of f-galactosidase can replacethe most NH2 terminal residues of lactose permease. Thus,there are no specialized ribosomes for starting translation of theintegral membrane protein lactose permease. The NH2 ter-minus ofthis mutant permease is long enough to be probed withantibodies against f3-galactosidase in order to decide whetherit is present on both sides of the membrane.The existence of negative dominant mutations suggests that

lactose permease is a dimer or oligomer in its active state in themembrane. All other explanations for the negative dominanceseem less plausible to us. Why, when normally active as a mono-

mer in the membrane, should lactose permease be inactivatedby the y-d products? It does not make sense either to assume

a limited number of sites in the membrane which the yd prod-uct would occupy and thereby prohibit the proper positioningof native lactose permease in the membrane. The membranescan support a 10-fold excess of active lactose permease over thewild-type level (14). The simplest explanation for negative dom-inance seems to be that normally occurring dimers or oligomersoflactose permease that may be formed in the cytoplasm or, lesslikely, in the membrane (see below) are inactive when one ofthe subunits is damaged, as in the y-d products.A dimeric or oligomeric structure would provide the basis

of symmetry of the sugar-binding sites on both sides of themembranes (7). A symmetrical structure, however, creates theproblem ofhow lactose permease enters the membrane. Eitherone ofthe subunits rotates in the membrane at least once beforeaggregation, or lactose permease aggregates first in the cyto-plasm and enters the membrane as dimer or oligomer. Neitherpossibility is excluded. However, we tend to prefer aggregationin the cytoplasm.The sequence of lactose permease and the existence of the

y-d mutations favor two models in our opinion. In the firstmodel, the symmetrically dimeric (or possibly tetrameric) lac-tose permease forms a channel with some of its helical lipophilicsequences. Three helices (the first three lipophilic regions?; see

Fig. 3) from each subunit would be sufficient to form a channelof the necessary diameter for sugar transport. This would placethe machinery of the sugar and proton recognition sites includ-ing the allosteric mechanisms of concerted proton/sugar trans-port on the outside of the membrane. Specificity would begained by keeping all unwanted substrates from entering thechannel. Compatible with such a model is the ease by whichspecificity can be altered. The other model would place all thesugar and proton recognition sites into the interior of a largechannel formed by at least a dozen lipophilic helices ofthe dimeror tetramer. This interior structure would then recognize sugar

and proton and move them in a concerted fashion from one sideof the membrane to the other.

The map positions ofthe y-d, y-K, and Y'"" mutations defineregions oflactose permease involved in sugar or proton bindingand transport. The y-d mutations and the y-K mutations which

have not gained the capcity to bind and transport maltose couldbe defective in the capacity to bind and transport galactosidesor protons. We think that the y-d and yK mutations, whichmap in deletion groups V and III, respectively, are possiblydamaged in proton binding and transport because the hydro-philic sequences involved seem to be too short to play a rolein sugar recognition. The y-d and y-K mutations mapping inregions downstream from codon 150 could be damaged in sugarbinding. The yK mutant which has gained the capacity to trans-port maltose is certainly altered in its sugar binding site. An-other explanation may be that the y-d and y-K mutations affectsubstrate binding indirectly rather than directly by causing con-formational changes of the carrier.

We thank Caecilia Ruth for technical help. This work was supportedby Deutsche Forschungsgemeinschaft through Grant SFB74.

1. Rickenberg, H. V., Cohen, G. H., Buttin, G. & Monod, J. (1956)Ann. Inst. Pasteur 91, 829-857.

2. Hobson, A. C., Gho, D. & Muller-Hill, B. (1977) J. Bacteriol.131, 830-838.

3. Buchel, D. E., Gronenborn, B. & Muller-Hill, B. (1980) Nature(London) 283, 541-545.

4. Ehring, R., Beyreuther, K., Wright, J. K. & Overath, P. (1980)Nature (London) 283, 537-540.

5. West, I. C. & Mitchell, P. (1973) Biochem. J. 132, 587-592.6. Wright, J. K., Teather, R. M. & Overath, P. (1979) in Function

and Molecular Aspects of Biomembrane Transport, eds. Quag-liariello, E., Palmiero, F., Papas, S. & Klingenberg, M. (Elsevier/North-Holland, Amsterdam), pp. 239-248.

7. Teather, R. W., Hamelin, O., Schwarz, H. & Overath, P. (1977)Biochim. Biophys. Acta 467, 386-395.

8. Fox, C. F. & Kennedy, E. P. (1965) Proc. Natl Acad. Sci. USA54, 891-899.

9. Kaczorowski, G. J., Leblanc, G. & Kaback, H. R. (1980) Proc.Natl Acad. Sci. USA 77, 6319-6323.

10. Muller-Hill, B. (1975) Prog. Biophys. Mol Biol 30, 227-252.11. Miller, J. H. (1978) in The Operon, eds. Miller, J. H. & Rezni-

koff, W. S. (Cold Spring Harbor Laboratory, Cold Spring Har-bor, NY), pp. 31-88.

12. Ruther, U. (1980) Mol Gen. Genet. 178, 475-477.13. Shine, J. & Dalgarno, L. (1974) Proc. Natl. Acad. Sci. USA 71,

1342-1346.14. Teather, R. M., Bramhall, J., Riede, I., Wright, J. K., Furst, M.,

Aichele, G., Wilhelm, W. & Overath, P. (1980) Eur. J. Biochem.108, 223-231.

15. Klenow, H. & Henningsen, I. (1970) Proc. Natl Acad. Sci. USA65, 168-175.

16. Maxam, A. & Gilbert, W. (1977) Proc. Natl Acad. Sci. USA 74,560-564.

17. Teather, R. W., Muller-Hill, B., Abrutsch, U., Aichele, G. &Overath, P. (1978) Mol. Gen. Genet. 159, 239-248.

18. Miller, J. H. (1972) Experiments in Molecular Genetics (ColdSpring Harbor Laboratory, Cold Spring Harbor, NY).

19. Gho, D. & Miller, J. H. (1974) Mol Gen. Genet. 131, 137-146.20. Muller-Hill, B. & Kania, J. (1974) Nature (London) 249, 561-563.21. Zabin, J., Fowler, A. V. & Beckwith, J. R. (1978) J. Bacteriot

133, 437-438.22. Muiler-Hill, B., Crapo, L. & Gilbert, W. (1968) Proc. Natl. Acad.

Sci. USA 59, 1259-1264.23. Adler, K., Beyreuther, K., Fanning, E., Geisler, N., Gronen-

born, B., Klemm, A., Muller-Hill, B., Pfahl, M. & Schmitz, A.(1972) Nature (London) 237, 322-327.

24. Beyreuther, K. (1978) in The Operon, eds. Miller, J. H. & Rez-nikoff, W. S. (Cold Spring Harbor Laboratory, Cold Spring Har-bor, NY), pp. 123-154.

25. Ogata, R. T. & Gilbert, W. (1978) Proc. Natl. Acad. Sci. USA 75,5851-5854.

26. Ippen, K., Miller, J. H., Scaife, J. G. & Beckwith, J. R. (1968)Nature (London) 217, 825-827.

27. Buchel, D. (1981) Dissertation (Univ. Cologne, Cologne, FederalRepublic of Germany).

28. Shuman, H. A. & Beckwith, J. (1979)J. BacterioL 137, 365-373.

Proc. Nad Acad. Sci. USA 78 (1981)