55 0 1993 Elacvicr Scicncc Publishera B.V. All rights rcscrwd. 0378-l 119~‘~2’$~)5.00

GENE OhSSl

Sequence of the Kluyveromyces Zact is p -galactosidase: comparison with prokaryotic enzymes and secondary structure analysis ’

(Recombinant DNA; yeast; nucleotide sequence; full-length alignment; mean secondary structure predictions;

fl-glucuronidase)

Olivier Poch ‘, Hervt: L’HBte ‘*, Vincent Dallery b, Franqoise Debeaux b, Reinhard Fleer’ and Regis Sodoyerb

Rcccivcd b! J.-P. Lecocq~W. Szyhalski: 23 Dcccmbcr 1991; Rcvised:Acccptcd: 26 .4pril,27 .4pril 1992: Rcccivcd at puhlishcrs: 2’) .4pril 1992

The UC4 gene encoding the P-galactosidase (BGal) of the yeast, K~uJw~~~~Jw.s lucris, was cloned on a 7.2-kb fragment

by complementation of a lrrcZ-deficient Escherichia co/i strain. The nucleotide sequence of the structural gene, uith 42 bp

and 583 bp of the 5’- and 3’-flanking sequences, respectively, was determined. The deduced amino acid (aa) sequence of

the K. 1rcti.s /JGal predicts a 1025-aa polypeptide with a calculated M,. of 117618 and reveals extended sequence homolo-

gies with all the published prokaryotic jGa1 sequences. This suggests that the eukaryotic /{Gal is closely related, cvolu-

tionarily and structurally, to the prokaryotic BGal’s. In addition, sequence similarities were observed between the highly

conserved N-terminal two-thirds of the flGa1 and the entire length of the p-glucuronidase (BGlu) polypeptides, which suggests

that PGlu is clearly related, structurally and evolutionarily, to the N-terminal two-thirds of the fiGal. The structural anal-

ysis of the BGal alignment, performed by mean secondary structure prediction, revealed that most of the invariant residues

are located in turn or loop structures. The location of the invariant residues is discussed with respect to their accessibility

and their possible involvement in the catalytic process.

INTRODIICTION

The yeast K. 1acti.r is known to produce an inducible

intracellular BGal (EC 3.2.1.32) and can thus use lactose

as carbon and energy source. The mechanisms by which

lactose and galactose induce BGal activity have been ex-

tensively studied (Das et al., 1985; Dickson et al., 1990;

Kuger et al., 1990). Initially, Dickson and Markin (1978)

had shown that a K. k&s DNA fragment can complement

a IaF mutant of E. coli lacking the structural gene for

fiGal. This fragment was then shown to carry the k’. Iactis

LAC4 gene (Sheetz and Dickson, 1981). Subsequent stud-

C~)rvc,c~c”‘~~~,n[,r fry: Dr. 0. Poch, Institut de Biologic Mol&xlairc et Cel-

lulaire de CNRS. 15 rut Dcscartcs. 67084 Strasbourg Ccdex, France.

Tel. (33.88)4l 7040; Fax (U-88)61 0680.

’ This paper is dedicated to the memory of Dr. J.-P. Lecocq.

* Prcscnt address: Transgtne S.A., I1 rut de Molshcim, 67000 Stras-

bourg (France) Tel. (33.88)279100.

Abbreviations: aa, amino acld(s): B.. Brrcillur: /{Gal. [I-galactosidasc;

PGlu, P-glucuronidase: hp, base pair(s); C.. Cbstuic/irm~; K.. Klu~vwron~.v-

c’e~; kb. kilobase or 1000 hp; KI., KlehsirN~r: L.. Ltrcmhucill~r.~: L.4C4.

gene encoding K. ltrcfi.7 /IGal: nt, nucleotide(s); oligo. oligodcoqrihonu-

cleotidc; ORF, open reading frame; PCR, polynerasc cham reaction; S.,

Swpmc’oc’c’ur; SDS. sodium dodcql sulfate; SSC. Il. I5 M NaCl:‘O.Ol5 M

Na;citratc pH 7.6: St.. S/w,nrnn~~r~e.~.

56

ies revealed that regulation of LAG! transcription is com-

plcx and involves the galactose/lactose induction system as

well as the catabolite repression (Riley et al., 1986; Ruzzi

et al., 1986; Salmeron and Johnston, 1986; Dickson et al..

1990; Kuger et al., 1990; Giidecke et al., 1991). The se-

quence of the LAC4 promoter and the deduced N-terminal

part of the protein had been determined (Breunig et al.,

1984; Leonardo et al.. 1987). However, full-length sequence

data for this eukaryotic PGal were missing. Within the last

decade. the complete nt and the deduced aa sequences of

six diffcrcnt prokaryotic BGal have been reported, two from

E. coli (Kalnins ct al., 1983; Stokes et al., 1985) and one

each from KI. pneumoniw (Buvinger and Riley. 1985).

L. hulg~~ricus (Schmidt et al., 1989), C. ac~etohut~licum

(Hancock et al., 199 1) and C. thermosu(furogene.~. EM 1

(Burchhardt and Bahl, 1991). In addition, the nt sequences

encoding the N-terminal regions of the fiGal from St. lir-

ides (Eckhardt et al., 1987) and S. thermophilus (Poolman

et al., 1989) have been published. Previous sequence anal-

ysis of the prokaryotic enzymes revealed the existence of

extensive aa sequence homologies and allowed the delin-

eation of different highly conserved sequence regions

(Schmidt et al.. 1989). The most conserved region exhibits

two invariant residues (Glu3”’ and Tyrso3) shown to be

critical for the catalytic activity of the E. coli enzyme (Ring

et al., 1985: Edwards et al., 1990; Cupples ct al., 1990). In

addition, sequence similarities have been reported between

the fiGa from E. coli and the eukaryotic and prokaryotic

BGlu (Fowler and Zabin, 1978; Nishimura et al., 1986).

The aims of the present study were cloning and sequenc-

ing of the entire LAC4 gene of K. 1ucti.v CBS2359 strain and

comparison of the deduced aa sequence to all the available

/IGal and PGlu aa sequences.

RESULTS AND DISCUSSION

(a) Isolation of the Kluyvevomyces lactis CBS2359 LAC4

gene

In order to verify that the LAC4 gene of the K. kmtis

CBS2359 strain exhibits a restriction pattern similar to that

reported for the K. luctis isogenic strain Y 1140 (Sreekrishna

and Dickson, 1985). a Southern hybridization analysis of

restriction enzyme-digested CBS2359 genomic DNA was

done (Fig. 1). The probe was obtained by PCR amplifica-

tion on CBS2359 genomic DNA using two oligo primers

(5’ primer from nt + 1 to + 33 and 3 ’ primer complemen-

tary to nt + 322 to + 35 1 as derived from the sequence data

available on the Yl 140 strain; Leonardo et al., 1987). A

7.2-kb Xh~l1 fragment strongly hybridized with the PCR

probe (Fig. 1). This result agrees with the restriction map

published by Sreekrishna and Dickson (1985) for the Y 1140

strain and indicates that the CBS2359 LAC4 gene is in-

1 2 3 4

kb

d - 23.1

- 9.4

w 7.2

- 6.5

Fig. I. Southern nnnlysis of the K. /uri,v LACY4 gent. Gcnomic DNA of

CBS2359 strain was isolated by the method of Kaback and Davidson

(1979). DNA \%as digested with PvuII (lane 1). Strll (lane 2). XhoI (lane

3), .YhoI (lane 4) and clcctrophorcacd in a I”,, agarosc gel. Follou ing

Southern blotting onto nitroccllulosc, the filter was probed at 65 C as

described in Sambrook et al. (19X9), with it “P-radiolabelled PCR frag-

mcnt corresponding to the CBS2359 LAC4 ORF extending from positions

+ I to + 3.50 (Leonardo et al.. 1987). The filter wu subscquentl) washed

with 0.2 x SSCI “Cl SDS. Sires (kb) of a Hirzdlll-digested phagc i DNA

arc indicated to the right. The arrow indicatcb the unique 7.2.kb ,YhrrI

fragment containing the LACJ gene.

eluded in a 7.2-kb X&I fragment as shown for the Y 1140

strain (Sreekrishna and Dickson, 1985).

The genomic DNA from K. hctis CBS2359 was isolated

and digested with X&I. The fragments ranging from 6 to

8 kb were electro-eluted (6-8 X&I fraction). The 6-8 X~LII

fraction was cloned in a pSVL vector (Pharmacia LKB

Biotechnology AB, Uppsala, Sweden) and used to trans-

form E. co/i JMlOl strain devoid of BGal activity. Over

8000 transformants were obtained and screened for fiGal

activity. Plasmid DNA isolated from five blue colonies all

contained a 7.2-kb XbaI insert with identical restriction

maps (data not shown). Fig. 2 shows the 3703-nt sequence

containing the 3078-nt aGal coding sequence capable of

encoding a 1025-aa protein and 42 and 583 nt of the 5’ and

Fig. 2. Complete nt scqucnce of the LAC3 gene. Rankmg region, and Its

dcrivcd aa acqucnce. Putative -35 and IO promoter elemcnta arc in bold

type and the putative polyadcnylation signal is underlined. The noncoding

sequences arc in lower-case letters. The ocgative numerals rcfcr to the

5’-flanking region and positive numerals begin at the A of the start codon

(ATG). To construct the library. the 6-8 .YhuI fraction (see aectwn a) was

isolated from H 0.X”,, agarose gel b> clcctro-elution and cloned into the

.Y/uI site of plasmid pSVL (Pharmacia) by standard methods (Samhrooh

ct al.. 1989). The clone GL A IO which contained the L.4C.d pcnc was uxxl

as the vu-cc for nt sequencing. Scquencmg was done h! the dideorc>

MSCLIPENLRNPKKVHENRL AlGlXXTG

PTRAYYYDQDIFESLNGPWA. CC~QZXI

FALFDAPLDAPDA'KNLDWET TTl?xmT

A K K'W S T ; S V P'S H W'E L Q i D W K' G-m

Y G K'P I Y i! N V Q-Y P 1-P I D i P N P' T-p-

PTVNPTGVYARTF'ELDSKSI' CCCACIGTA?+ATCCCGAAATCGATI

ESFEHRLRFEGVDNCYELYV cLuxuzrT--m

N G Q'Y V G i N KG'S R N'G AE $ D I Q' APLPGC;PCAFLCATGn;GG--n

KYVSEGENLVVVKVFKWSDS A-

TYIEDQDQWWLSGIYRDVSL 200 AClT?U!&---m 600

LKLPKKAHIEDVRVTTTFVD 220 C!TAAAAT---w 660

SQYQDAELSVKVDVQGSSYD' TC-G?.XtXCAGXX~~

HINFTLYEPE'DGS'KVYIiASS' CcrAAAG-m

LLNEENGNTTFSTKEFISFS TT--

TKKNEETAFKINVKAPEHWT A--m-

AENPTLYKYQLDL'IGSDGSV' Gp

IQSIKHHVGFRQVELKDGNI ATPCAATCP

TVNGKDiLFR'GVNkHDHHPR' ACCAkGG

FGRAVPLDFV'VRDLILMKKF TT

NINAVRNSHY'PNHPKVYDLF' A

DKLGFWVIDE'ADL'ETHGVQE GATAAGclmxcrT~

PFNRHTNLEAEYPDTKNKLY c--lKZ?AC

D V N'A H Y i S D N'P E Y-E V A ; L D R' G-m-m-

ASQLVLRDVN-HPS'III!iSLG' TccpAAGp13FITGTcApLpmmn

NEACYGRNHKAMYKLIKQLD A-

PTRLVHYEGDLNA'LSADIFS CCl!ACCAG?+ClTG~---

FMYPTFEIMERWRKNHTDEN' TT

GKFEKPLILCEYGHAMGNGP GGmAGT--m-m

G S L'K E Y 6 E L F'Y K E'K F Y 6 G G F' GGcrcrcT~~Tl-ITAC~~

1WEWANHGIEFED;STADGK A-w-

LHKAYAYGGDFKEEVHDGVF 620 T-ATGClT--TGl,lC 1860

-1

20 60

40 120

60 180

00 240

100 300

M 360

140 420

160 480

100 540

240 720

260 780

280 840

300 900

320 960

340 1020

360 1080

380 1140

400 1200

420 1260

440 1320

460 1380

480 1440

500 1500

520 1560

540 1620

560 1680

5fJo 1740

600 1800

IMDGLCNSEHNPTPGLVEYK A ACPCCGGGCCrr-TATAAG

KVIEPViIKI'AHGSVTITNK A

HDFITTDHLLFIDKDTGKTI 600 C-c!AA!rc 2040

DVPSLKPEESVTIPSDTTYV 700 GAcxrrc~'-A!rGrI 2100

VAVLKDDAGVLKAGHEIAWG GTlGCLiG--TFGCCTGGGGC

Q A E'L P L i V P D'F V T'E T A i K A A' -mm-

720 2160

740 2220

KINDGKRYVSVESSGLHFIL A--m

DKLLGKIESLKVKGKEISSK

760 2280

780 2340

FEGSSITFWRPPTNNDEPRD 800 TGGAGaC~-w.%AC 2400

FKNWKKYNIDLMKQNIHGVS 820 TT 2460

VEKGSNGSLAVVTVNSRISP CCCCCA

VVFYYGFETVQKYTIFANKI 860 GTXT?XSTlTACT-ACXCZA- 2580

N L N'T S M i L T G'E Y Q'P P D 'i P R V' -w--m 2640

GYEFWLGDSYESFEWLGRGP 900 B---C 2700

GESYPDiKES'QRF'GLYiSKD' 920 --p-m 2760

VEEFVYDYPQENGNHTDTHF 940 G--pCCXCTFI 2820

L N I'K F E i A G K'L S I'F Q K i K P F' TTGAACATCAAPLPT 2880

NFKISDEYGVDEAAHACDVK 980 2940

RYGRHYLRLD'HAIHGVGSEA 1000 A-- 3000

CGPAVLDQYRLKAQDFNFEF 1020 TG 3060

57

640 1920

660 1980

840 2520

1025

3120

3180 3240 3300 3360 3420 3480 3540 3600 3661

chain-terminating method (Sang-z et al., 1977) for single-stranded or

double-stranded templates. Large fragments spanning the gcnc region

wcrc subcloned in M 13 and subjcctcd to cxonuclcasc digestion (Cyclone

system; Dale et al., 1985) giving about 50”,, of the sequcncc data. The

missing regions were obtained by ‘gene walking‘ usmg a set of 30 spccitic

primers synthesized on an Applied Biosyatcm 38lA DNA S>nthesi,w

and used without purification. The complete nt scquencc was dctcrmincd

independcntlg on both strands. Premixed scqucncing rcagcnts (Scqucnasc

11 kit) were obtained from IJnited States Biochemical Corporation (Clcvc-

land. OH). Sequencing reactions Mere carried out according to the sup-

plier’s specifications. The ‘SS-labelled nt used for the scqucncing reactions

~vcrc purchased from Amersham International (Bucka., UK). The acccs-

sion No. M84410 in the GenBank database has been assigned to this nt

sequence.

Prs

dic

#i

-f-

ir,rs

asss

- am

fw

mm

65

0 70

0

ME

RW

RK

NW

TiE

NG

KF

E

00

0

00

0

0

0

. . .

I

EF

EO

”ST

AO

GK

L ---

-PLM

NEFG

EYPH

LP

. .

“FY.

. .I.

. . .

KH

DC

i*

A(lo

tlW--

--.G

Y

PK

WS

!Krn

SLP

GE

T

*FR

____

___.

.Qyp

R

LIKY

DEN.

---

G

VPm

GlK

KWlS

LPO

EO

.,....

AF

*~--.

...-.E

YPR

!RKT

FdO

----.G

---

-KYI

EEYL

TNKP

A I”

_...

LE_“

“_._

._..K

”PH

LEK+

.*,---

-.-

----A

ilEEY

LNDN

PK

LED.

“__.

.““.

KYLM

LY

RKLP

D-.

.--G

NK

P~~K

~EVT

GLE

~N

. .

. .

. .

EHCL

------

..RHL

R 1.

. _.

_. . _

_G

EKVL

EKEL

LAM

QEK

LH

L)TL

*GL

AWLD

MYW

RVFD

..RYS

TS

O..~

--GIL

AY

OLO

LATO

FE%

VYKK

YQ

ETIA

GF

SLLE

QYH

LGLD

CIKR

RK

TE

Q”

-SPT

R”

OP*

LNSG

FEta

rfKT

M

DAfP

GI

AVLE

NYHS

VLDQ

KRKE

TN

R....

-SPL

RY

OLQ

LTSQ

FENW

YKM

YO

QAV

SGL

ALLE

NYHL

ILDE

KRKE

TN

*.~“.

-SPL

R”

++

+ +

+ t*

++

BB

Bi?B

~f

iliti~

J+f”

75

0

La04

Eb

QA

LaCZ

La

cL

Gal

3’ flanking regions, respectively. The 3’-noncoding region

contains a consensus polyadcnylation signal (5’-AAT-

AAA) located 280 bp downstream from the stop coding.

(b) Alignment of the fiGal aa sequences

Pairwise sequence comparisons were performed between

the deduced aa sequence of the K. 1wti.s /Gal and those

published for the E’. co/i (Kalnins et al., 1983: Stokes et al.,

1985), KI. ptzeuttwnicre (Buvinger and Riley, 19X5), L. hul-

gcwicu.s (Schmidt et al.. 1989). C. trwtohutylicut~t (Hancock

et al., 1991), C. rhc~rttto.sulfirrogette.s EM I (Burchhardt and

Bahl, 1991) and the N termini of Sr. 1ividcttl.s (Eckhardt

et al.. 1987) and S. thertmphihrs (Poolman et al., 1989). As

shown in Fig. 3 (upper sequence lines), strong homologies

exist over the entire length of the PGal, only falling off

markedly in the C-terminal regions (roughly from aa 760 to

the end in the arbitrary numbering of Fig. 3). In more than

80”,, of the aligned /IGal sequences. there are 385 conser-

vatively maintained na of which 150 are strictly invariant.

The gaps introduced in order to maximize the similarities

are of limited lengths, except for two regions (aa 330-351

and 502-524) where large insertions arc mainly due to

K. lacti.~ PGal.

A measure of the pair\visc relationships bct\veen the dif-

ferent complete /IGal proteins is given in Table I which

shows a compilation of the strictly and conservativelq

maintained aa according to the alignment of Fig. 3. With

respect to the numerous residues conserved bet\veen the

K. Otis cnzymc and the other bGa.1. it is clear that the

eukaryotic protein is closely evolutionarily related to the

prokarqotic ones. In addition, the data shown in Table 1

may suggest that K. /ctcti.v BGal is somcivhat more closclq

related to the E. co/i enzyme encoded by the ehgA gent (577

conservatively maintained aa among which 340 are invari-

ant). The two [IGal encoded by the 1rrc.Z genes from E. co/i

and KI. ptteutmt~iae arc so far the most closely related.

while the /~‘Gal from L. hu/garicu.s and C. ct~rtohu!\Gwttt

may define a third group. The thermostable /IGal from

59

C. thern?o.sulfirroRelle.v EM 1 is of special interest since it

appears to be extensively different from the six other PGal.

Based on the number of strictly maintained rcsiducs, there

is a slightly closer relationship bet\vccn C’. t/7erttto.~ul/lrro-

gettes EM I PGal and the ehgA gene product nhilc the lcast

related enzyme is KI. peutnc~ttirre. Howcvcr, thcsc data ma)

reflect structural analog), rather than evolutionary links

since these relationships are different ~vhen the conserva-

tive replacements are taken into account.

(c) Alignment of the fiGal and BGlu aa sequences

Additional searches for similaritics bctwecn the K. k/i.s

/IGal sequence and the proteins present in the Slviss-Prot

data bank allo\s,ed the detection. at statisticall\ significant

levels, of all reported PGlu. i.c., the E. co/i (JefTerson et al.,

1986). human (Oshima et al.. 1987), mouse (Gallaghci

et al.. 1988) and rat (Nishimura et al.. 1986) enzymes. In

Fig. 3, we propose an optimal alignment of the /iGal and

PGlu sequences in which I94 aa arc conservatively main-

tained of which 71 arc invariant among at least 80”,, of the

eleven aligned complete aa sequences. Analysis of the data

shoed in Table I reveals that none of the /jGal scqucnccs

stems more r&ted to one or another /IGlu scqucncc. In-

deed, the number of strict11 or conservativel) maintained

residues arc somewhat equi\alent. Special attention should

be paid to the C. thrntto.sltl/irro,~etrre.s EM 1 /<Gal which ap-

pears slightly mot-c closely related to the PGal than to the

/IGlu sequences. This ma!’ reflect an intermediate evolu-

tionary position of the thermostable enzyme between the

PGal and PGlu proteins. Such an intcrmediatc evolution-

ary position is further supported b) two observations: first,

the length of the C. IhL~rtrto.slrl/irrft~~~tt~,.~ EM 1 cn~\ mc which

is intcrmcdiate bct\\ecn the /jGal and [IGlu protein Icngths:

second. the presence. in the C’. tl7ertt7O.FItI/ilR)~~~,tte.~ EM I

BGal. of a region (aa 186-237) more similar to the /iGlu,

especially to the E. co/i enzynlc. than to the [IGal sequences.

This region encompasses stretches of aa scquenccs in the

C. thertttosu~~irro~pctrr.c ERI 1 /jGal (EH KGGYTPF. TVVV,

TABl,t I

Pairwiac aa squcnce cornparsons among elcvcn complete aa sequences aligned in Fig. 3

Scqucnce ’ Number of invarianl and conaer\ed (in parcnthcacs) aa”

K. la Lx-l E. co Ebg4 E. co LacZ Kl. p LacZ I.. bu Gal C. ac CBga C. th Lx% E. co Glc Hum Glc Mou Glc Rx Glc

K. la Lx-! 102 340 (577) 303 (554) 28X (524) 275 (530) 261 (474) 132 (307) I l-1 (243) 121 (265) 119 (1763) 1’7 (265)

E. co kbg4 1031 323 (570) 3 I I (556) 275 (547) 257 (501) 147 (304) 122 (259) 1X(262) 133 (269) 123 (264)

F. co LacZ 10’3 60 1 (757) 337 (558) 29x (503) I30 (384) 110 (250) 125 (259) 125 (162) 112 (260)

KI. p LacZ IO34 314 (54X) 265 (383) 108 (381) 105 (243) I I? (257) 109 (256) 109 (257)

I.. hu Gal I007 391 (582) 130 (301) lll(243) 114(246) lOS(234) II’)

c‘. aI‘ C‘Bga x97 132 (309) 121 (239) I I7 (236) Ii6 (234) 122 (23.1)

(‘. th Lx% 716 II-1 (250) I I? (XI) 10X (355) I14 (256)

I:. co Glc 602 177 (391) 275 (396) 775 (396)

Hum Glc 619 -177 (513) 4x0 (552)

hlou Glc 616 554 (593)

Rx1 Glc 6 2 6

’ Scqucncc names and abbrcvirttions arc given in Fig. 3 Icgcnd.

” Conscrvcd res~ducs corrcapond to strictly and conscrvalivcly mamtained rcsiducs. The aa families defining ux~servallvcly mamtaincd rcsiducb UC‘ ,g\‘en

111 Fig. 3 Iwcnd.

DIPP) which are almost identical to the sequences of E. cnli

/iGlu (EHQGGYTPF, TVCV. TIPP).

Previously published sequence comparisons between

E. co/i~Gal and rat PClu (Nishimura et al.. 19X6), revealed

similarities for a limited region corresponding to the ccn-

tral part of the BGlu (aa 144-591 in Fig. 3). In the present

alignment, \ve extended these similarities to the N-terminal

two-thirds of BGal and to the entire length of /3Glu (aa

l-77 1). This stronglq suggests that the flGlu sequences are

cvolutionarily strongly related over their entire length to the

N-terminal two-thirds of BGal. In addition, these results

support the view that the C-terminal regions of BGal (aa

772 to the end), which arc by far the less conserved regions

\vhcn the flGal are compared to one another. may consti-

tute individual domains distinct from the N-terminal parts

of the enzymes. Such a delineation of two main distinct

domains of homologies in all the /?Gal is in good agreerncnt

M ith the results obtained by limited proteolysis experiments

performed on the 15’. co/ijiGal (Edwards ct al.. 1988). These

cxperimcnts pointed out the existence of a major chymo-

tryptic cleavage site located between aa 585-586 (aa 723-

724 in Fig. 3) and to a major elastase cleavage site located

between aa 732-733 (aa 888-889). With respect to the

alignment, these two proteolytic cleavage sites clearly en-

compass the scquenccs which may correspond to the in-

terdomain region linking the conscrvcd N-terminal part

and the variable C-terminal part of /IGal. In this w’ay, it

should be noted that the PGal regions extending roughly

from aa 810 to 910 in Fig. 3 exhibit weak sequence sim-

laritics and contain numerous gaps, suggesting that these

regions are probably poorly structured. This observation

strongly supports the suggestion of Edwards ct al. (1988) that the region embedding the elastase site (aa 8X8-889 in

Fig. 3) is the major interdomain region of E. co/i BGal. In

addition, since the C-terminal regions of /IGal are peculiar

to these enzymes and do not have counterpart sequences

in /IGlu, it seems reasonable to assume that some features

specific to /JGal. such as substrate specificity, must bc

searched for in these C-terminal regions. This possibility is

further supported by genetic studies (Landridge, 1968:

Landridge and Campbell, 1968) and recent biochemical

experiments (Martinez-Bilbao ct al., 1991) showing that

the Gly’” (aa 952 in Fig. 3) is directly involved in substrate

recognition.

(d) Structural analysis of the alignment

The presence of extensive homologies between the BGaI

and /IGlu sequences suggests that the conserved regions

probably share some structural features and that most of

the rcsiducs involved in catalytic activities might bc found

in the homologous domains, especially among the invariant

residues. The conscrvativc and invariant au maintained

between the BGlu and the BGal sequences arc not ran-

domly distributed, but rather clustered into blocks of strong

conservation. These blocks are linked by variable regions

generally exhibiting gaps due to the insertion of residues in

one or another of the enzyme sequence. We compared the

positions of these variable regions to the positions where

linker insertions can be performed on the E. c,o/i/lGal Ivith-

out loss of activity (Breul et al., 1991). Strikingly, among

the tight linker insertions in the N-terminal part of the

E. co/i scquencc (aa l-771). seven were located within or

very near to the variable regions delineated in the alignment

(aa 90-97; 308-312; 346-347; 395-400; 602-606; 602-

608; 633-635) and the eighth insertion (aa 62-63) was

located upstream from the first set of invariant residues (XI

61

63-69). Considering the C-terminal part of the E. co/i en-

zyme, the linker insertions were also mostly located in var-

iable regions (aa 832-835; 957-958; 995-996; 1053-1054

and 1085-1088). This strong correlation observed between

the positions for which linker insertions do not induce loss

of activity and the positions of variable regions constitutes

a very powerful validation of this alignment. Moreover,

these observations allow us to predict with confidence that,

in the structural organisation of PGal and PGlu, the ma-

jority of the delineated variable regions may be involved in

loops allowing the insertion of various residues without

significant perturbation of the overall folding.

It should be pointed out that, among the potential glyc-

osylation sites (NXT/S, where X is any aa except Pro)

conserved in the three eukaryotic /~‘Glu sequences (bold

lettering in shaded boxes in Fig. 3), two are located in

variable regions with large insertions (aa 218-220; 502-

504), the third one being located at the C-terminal end of

the sequences where the sequence homologies fall off mark-

edly. With respect to the two variable regions, the presence

of potential glycosylation sites further reinforces the hypo-

thesis that these regions arc composed of accessible loops.

Consistent with this hypothesis is the presence of numer-

ous strong structure-breaker residues (Pro and Gly; see

Argos, 1987 and references therein) that are observed close

to the two glycosylation sites in the /JGlu sequences.

Among the distinct blocks of conservation, two regions,

from aa 361 to 501 and aa 53 1 to 567, are more conserved.

These regions contain most of the invariant aa (34 out of

72) as well as the longest stretches of conserved peptides

(motifs): GXN(R/K)HE (aa position 428); RXSHYP (aa

position 464); LCDXXG (aa position 477); RDXNHP (aa

position 545) and WSXXNE (aa position 555). Numerous

biochemical experiments pointed to the importance for cat-

alytic activity and Mg’+ binding of the invariant Glu (E)

present in the WSXXNE motif (Herrchen and Leglen,

1984; Badcr et al., 1988; Cupplcs and Miller, 1988; Cup-

pies et al., 1990; Edwards et al., 1990). The critical role of

the region encompassing the RDXNHP and WSXXNE

motifs is supported by the fact that these motifs represent

sequences for which local similarities can be observed bc-

tween BGal encoded by the B. s~euro~herrnophilus ebgA gene

and the PGal sequences (Schmidt et al., 1989).

A second residue, Tyr”” (aa 6 16 in Fig. 3) and the neigh-

bouring Met”” have been shown to be part of the catalytic

site and involved in the catalytic mechanisms of E. coli

PGal (Fowler et al., 1978; Ring et al., 1985; 1988; Ring and

Huber, 1990; Edwards et al., 1990). This residue is thought

to be the proton-donating species needed for catalysis. In-

spcction of the aligned sequences shows that this residue

is located close to the Mg2+ binding motif WSXXNE and

involved in an invariant dipeptide MY (aa 615) specific to

the PGal sequences. It is of special interest to note that

within the BGlu sequences Tyr”” and Tyr”“’ residues. lo-

cated in close proximity to the BGal-specific invariant MY

dipeptide, arc strictly conserved in the four fiGlu sequcnccs.

Therefore, it is tempting to propose that tither of thcsc

tyrosine residues may play, in the BGlu catalytic mecha-

nisms, a role similar to that attributed to the Ty? (aa 616

in Fig. 3) in the E. coli BGal activity.

Finally. to get additional insights into the three-

dimensional organisation of these two enzyme types. we

calculated the mean secondary structures over the entire

length of the eleven aligned sequences. WC took into ac-

count the similarly predicted structures (see legend to

Fig. 3) since their concordance over numerous sequences

lend more credence to the predictions (Zvclcbil ct al., 1987).

Eleven b-strands and two a-helices could be deduced from

this analysis. Strikingly, most of the invariant residues be-

longing to the aa known to be frequently involved in cat-

alytic processes (R, H, D and E) (Zvelebil and Sternberg,

1988), are located in positions bordering a b-strand or in

a predicted loop (random coiled) or turn structure. The

location of these invariant residues in turn or loop struc-

tures may reflect their accessibility and their possible in-

volvement in cation and substrate binding or in catalytic

processes. In this way, the two residues Glu’“’ and Ty?

(aa 561 and 616 in Fig. 3) shown to be directly involved in

the catalysis in the E. coli/?Gal are located at the C-terminal

part of the strongly predicted b-strands. Similarly, the long-

est conserved motif NXXRXSHYP (aa position 461)

seems to be part of an extended loop structure. Such a

location suggests that the invariant residues located in the

loop may be accessible for substrate interactions and/or

potential catalytic action.

ACKNOWLEDGEMENTS

We thank Dr. M. Zerbib and Dr. A Mouchaboeuf for

computer assistance. We are pleased to thank Dr. D. Lamy

for helpful discussions.

REFERENCES

Argos, P.: i\ sensitive procedure to compare amino acid scqucnccs. J.

Mol. Biol. 193 (19X7) 3X5-396.

Badcr. D.E., Ring, M. and Huber, R.E.: Sitedmxted mutagenic rcplace-

mcnt of Glu-461 with Gin in P-galactosldase (E. wli): cvidencc that

Glu-461 is important for activity. Biochem. Biophqs. Rcs. Commun.

153 ( 1988) 30 I-306.

Bairoch. A. and Boeckmann. B.: The SWISS-PROT protein sequence

databank. Nucleic Acids Res. 19 (1991) 2237-2219.

Breul. A., Kuchinkc, W., van Wilcken-Bergmann. B. and Miillcr-Hill, B.:

Linker mutagencsia in the IcrcZ gene of Esc~hrrirl~i~/ u>h yields variants

of active P-galactosidase. Eur. J. Biochem 195 (1991) 191-193.

Breunig, K.D.. Dahlcms. U., Das, S. and Hollcnbcrg. C.P.: 2nalysis 01

63

positlvc regulatory gcnc LACY reveals functional homology to. but

scquencc divergence from. the Sacch~ro~~~~ce.s crreaisirre GAL4 gene.

Nucleic 4clds Rca. I4 ( 1986) 7767-778 1. Sambrook. J.. Fritsch, E.F. and Maniatis. T.: Molecular Cloning. 4

Laboratory Manual, 2nd cd. Cold Spring Harbor Laboratory Press,

Cold Spring Harbor, NY, 1989.

Sangcr F.. Nicklen. S. and Coulson, .4.R.: DN.4 sequencing with chain-

terminating inhibitors. Proc. Nutl. .Acad. Sci. USA 74 (1977) 5463-

5467.

Schmidt. B.F.. Adams, R.M.. Rcquadt, C.. Poxver. S. and Mainzer. S.E.:

Expression and nucleotide sequence of the Lacrohrrcillus hulgcvicu.t

fl-galactosidasc gene cloned in the Ewherichitr co/i. J. Bacterial. 17 I (1989) 625-635.

Sheetr, R.M. and Dickson, R.C.: LAC4 is the structural gene for /i-

galactosidasc m K/uyveror~~.~ws k/cti.v. Gcnctics 98 ( I98 I ) 729-745.

Sreekrishna, K. and Dickson, R.C.: Construction of strains of

Strcchmm~~ces cerevititre that grow on lactose. Proc. Natl. Acad. SCI.

us.4 82 (1985) 7909-7913.

Stokes. H.W.. Betts. P.W. and Hall. B.C.: Sequence of the e/7x,4 gene of

E.cchrrichirr dr: comparison with the kwZ gent. Mol. Biol. Evol. 7

(1985) 469-477.

Zvclcbil. M.J. and Stcrnbcrg, M.J.E.: -\nalysia and prediction of the lo-

cation of catalytic residue in enzymes. Protcm Eng. 2 ( 1988) 127-13X.

Zvelcbll. M.J.. Barton. G.J.. Taylor. W.R. and Stcrnberg. M.J.E.: Prc-

diction of protein secondary structure and actwc sites using the align-

ment of homologous sequences. J. Mol. Biol. 195 (1987) 957-962.