Resistance of Gram-negative Bacteria to

Purified Bactericidal Leukocyte Proteins

RELATION TO BINDING ANDBACTERIAL LIPOPOLYSACCHARIDE

STRUCTURE

JERROLDWEISS, SUSANBECKERDITE-QUAGLIATA, and PETERELSBACH,Department of Medicine, New York University School of Medicine,Netw York 10016

A B S T R A C T The sensitivitv or resistance of gram-negative bacteria to antibacterial systemis appears to berelated to the length of the saccharide chain of thebacterial envelope lipopolysaccharides (LPS). To ex-plore this relationship further, we miade use of twobactericidal, miembrane-active cationic proteins, re-cently purified to near homiogeneity, one from humanand one fromi rabbit polymnorphonuclear leukocytes(PNIN). We have studied the effects of these twoclosely simiilar proteins on strains of Salmiionella typhi-nuriutrm and Escherichia coli, each separate strain

differing in the saccharide chain length of its outermemlbrane LPS. Binding of these proteins to the bac-terial outer membrane is required for killing, and isaccompianied by an almost immnediate increase in outermemnbrane permeability to normally impermieant actin-omycin D. Sensitivity to the bactericidal and perme-ability-increasing activities of the human and rabbitproteins increases with decreasing LPS-saccharidechain length (chemiotype: [S < Ra < Rb3 < Rc < Rd1]).S. typhimurium G-30 and E. coli J5, mutant strainslacking UDP-galactose-4-epimnerase, synthesize in-complete LPS (chemotype Rc) whein grown withoutgalactose, and are then as sensitive to both PMIN pro-teins as the S. typhimurium strains 395 RIO (Rd1) andR5 (Rb3). However, when these imutants are grownwith galactose, they synthesize complete LPS (chemio-type S) and exhibit nearly the same relative in-sensitivity as the smooth strains S. typhimurium395 NIS aind E. coli 0111:B4.

Th-e differences amiong strains in sensitivity to theeffects of the proteins on bacterial viability andpermeability correspond to differences in bacterialbinding of these PMIN proteins. Thus, at protein con-

Receivecl for ptublication 1 Auiguist 1979.

centrations that produce maximal antibacterial activitytoward the rough bacteria, but little or no activitytoward the smiooth strains, rough bacteria bind from3- to 10-fold more protein (S. typhirnuriumit 395RIO; S. typhitmurium G-30, and E. coli J5 [grownwithout galactose]) than do the smooth bacteria (S.typhimuritum 395 MIS; E. coli Olll:B4; S. typhimnuriumiG-30 and E. coli J5 [grown with galactose]). Thesefindings suggest that bacterial sensitivity or resistanceto these purified bactericidal P.IN proteins is de-termined by the binding properties of the outer memn-brane, which in turn depends upon the LPS-saccharidechain length.

INTRODUCTION

Wehave recently purified, to near homogeneity, twovery simiilar cationic, membrane-active proteins fromhtumian and rabbit polymorphonuclear leukocytes(PXIN)l that are potently bactericidal toward severalstrains of the gram-negative bacterial species, Esche-richia coli and Salmiiotnella typhimurium (1, 2). Rapidkilling of these bacteria is accompanied by subtleenvelope alterations that appear restricted to the outermemubrane. These alterations include an almost im-miediate, discrete increase in outer membrane perme-ability (1-4; unpublished observations). To producethese antibacterial effects, the bactericidal/perme-ability-increasing (PI) proteins apparently must bind tothe outer memibrane of the gram-negative bacterialenvelope (3-5).

The degree of effectiveness against gram-negativebacteria of many antibacterial systemns, including

1Abbreviationts used in this Paper: ActD, actinoinycin D;LPS, lipopolysaccharides; PI, permeability increasing; PMN,poal Imorplhonuileatr leukocytes.

619J. Clin. Invest. (© The American Society for Clinical Investigation, Inc. * 0021-9738180/0310619/10 $1.00Volume 65 March 1980 619-628

cellutlar and humoral host defenses and antibiotics,depends largely on the physicochemical properties ofthis outer membrane (6-10). These properties imainlyappear determined by the length of the polysaccharidechain of the outer menmbrane lipopolysaccharides(LPS) (11-13). Smnooth strains (containing long-chainLPS) are more resistant than rouigh strains (containingshort-chain LPS) to host defenses in vivo (6) and tokilling in vitro by intact PMN(7), crude and partiallypurified extracts of PMN(14-17) and serum (18).

None of these earlier studies was carried outwith defined antibacterial substances. The availabilityof nearly pure and well-characterize(d leukocyte pro-teins (1, 2), with apparent antibacterial specificity forgramn-negative bacteria, prompted uis, therefore, to fur-ther examiiine the relationship between bacterial LPSstructuire and the senisitivity of S. typhimurium andE. coli to the action of these proteins.

We find that rough strains of S. typ1irnuriuwii1 andE. coli, with short-chain LPS, avidlly bind both pro-teins and are exquisitely senisitive to their antibacterialactivity. In contrast, smnooth strains, withl long-chainiLPS, show mullch less binding and a correspondinglyreduced sensitivity, suiggestinig that the LPS-poly-saclcharide chain length (letermiines bacterial sensitivityor resistance to these bactericidal PMNproteins byaffecting the binding properties of the bacterial outermembrane.

METHOD)S

BactericidallPI proteinsThe bactericidal/PI proteins were isolated from hulman andral)l)it PMNas previously described (1, 2). The protein frac-tions uised in this stui(ly were at least 95-98% pure asdetermiined by gel filtration and electrophoresis in three ormore different btuffer systems.

BacteriaThe smnooth, mouise-virulent straini S. typhimnuriurm MS 395and rouigh inutants derived from it (R2, R5, and RIO)

Antigenicpolysaccharide

I

vere kindly donated by Dr. Olle Stendahl (Departmient ofMedical Microbiology, University of Link6ping, Sweden).S. typhimurium G-30 and E. coli J5, imtitants that lack theenzymiie UDP-galactose-4-epiinerase, and E. coli 0111:B4(smooth parent strain of J5) were gifts of Dr. Mary J. Oshorn(Departmiient of Microbiology, University of ConnlecticutHealth Center, Farmington, Conn.) and Dr. Loretta Leive(Laboratory of Biochemical Pharmacology, National In stituteof Arthritis, Metabolic, and Digestive lDiseases, NationalInstituites of Healtlh, Bethesda, Md.). E. coli s15 (a K12strain) was donated by Professor S. Nojima, University ofTokyo, Japan.

The struieture of the LPS of these bacterial strainis isschemiiatically shown in Fig. 1. S. typhimurinin strainis weregrown in nultrient broth (pH 7.3) and E. coli strainis in atriethanolamine-buffered (pl1 7.75-7.9) minimiial salts mediumin(22). Bacteriall culltuires grown overniight to stationary phasewere transferred to fresh mediumiii (diluite(d 1:10) and( thesubcultures were grown to micl-late logarithmic i )hase(-3-4 h at 370C; 6-110 x 108 bacteria/mIl). Where in(licated,the groxvth media of both overnight and suhbcuiltures weresupplemiiented with 5 mM1 D-galactose. Bacterial conicen-trations vere (letermiiine(d by measuiring absorbance at 550nmn with a Colemnan junior spectrophotometer Colemncan Sys-tems, Irvine, Calif. The bacteria were sedimented bycenitrif'uigationi at 6,000 g for 10 mmiin and( resuispenide(d insterile physiological saline to provide the desire(d con-cenitrationi.

Assays for bactericidal and PI activitiesTypical incubation mixtuires contained from 5 x 106 to5 x 101 bacteria in a total volumile of 0.4 ml of sterilephysiological saline that also containe(d 10 j.cmol Tris-HClbuffer (pH 7.0), 25 tcI of Hanks' solution (flanks' balancedsalt solution [withouit phenol red], Microbiological Asso-ciates, Inc., Bethesda, Md.), 250 ,ug of vitamin-free casamiiinoaci(d (Difco Laboratories, Detroit, Mich.) and the purified bac-tericidal/PI protein in the indicated amouint. Incli)ationswere carried ouit for 30 min at 37°C.

Bactericidal activitql. After 30 min incubation, 10-ulsamiples were taken from the incubation imixtures, seriallydiluted in sterile isotontic saline and plated on nutrientagar. Bacterial viability \vas measuired as the nu-mber ofcolony forming uinits on the plates after incubation at37°C overnight.

PI activity. An effect of the bactericidal/PI proteins onthe envelope permeability of S. typhiimurium and E. coliwas measurecl by determining the susceptibility of thesebacteria to actinomycin D (ActD). The use of ActD for assay

Outer core Backbone Lipid AI S.typhimurium E.coli ChemotyPe

hep R 10

lc-hep + - | G 30 -gal

I-]R 5

F -I----] R 2

Rd1

J-5 -gal Rc

Rb3

Ra

- 1 395 MS 0 lll:B4| G 30 +gal J-5 +gal

FIGURE 1 Schemnatic struetuires of the LPS of the variouis bacterial strains used in this study. Thedletails of these struettures have been previously described (19-21).

620 J. Weiss, S. Beckerdite-Quagliata, and P. Elsbach

of envelope permeability is based on the following ob-servations:

(a) The envelope of gram-negative bacteria such as S.typhimuriun2 and E. coli is normally impermeable to ActD(23). Hence, this antibiotic normally does not affect RNAandprotein synthesis by these bacteria.

(b) Bactericidal concentrations of the bactericidal/PI pro-teins produce little or no inhibition of bacterial macro-molecular synthesis for at least 2 h (1-3). Therefore, in-hibition of RNAand protein synthesis by bacteria exposedto ActD and bactericidal/PI protein reflects an increase inenvelope permeability to the antibiotic. Assays of PI activitywere carried out in the incubation imixture described aboveby determining the bacterial incorporation of [14C]leucine(0.063 ACi; 0.13 mM) (New England Nuclear, Boston,Mass.) into cold trichloroacetic acid-precipitable materialin the presence or absence of 12.5 ,ug ActD (MerckSharp and Dohmne Div., Merck and Co., Inc. West Point,Pa.) (5, 24). Simnilar resuilts were obtained using ['4C]-turacil as precursor (24).

Assay of binding of bactericidallPI proteinsto bacteria

Because mnultiple attemiipts at radiolabeling the bactericidal/PI proteins, withloult loss of bactericidal/PI activity, haveso far remained unsuccessful, we have continued to rely on abiological assay for imieasuiring bacterial binding of the pro-teins (5). The flow-sheet shown in scheine I (Fig. 2)serves to aid the reader in following the proceduire.

After preincubation for 15 min at 37°C of a given bacterialstraini (6 x 108 or 1.2 x 109 bacteria) with 50,ug of bactericidal/PI protein in 0.75 mnl of the standard incubation mixtture,the bacteria are pelleted by centrifugationi at 6,000g for 10inin at 4°C. The amiiount of protein bound by the bacteriais estimated by imieasuirinig the reduiction in bactericidal andPI activities recovered in the suiperniataint (I; Fig. 2) fractioll.'Fhe bactericidlal and PI activities of the puirified hulman and

Bactericidal/PI protein + bacteria(50 pg) (6 x 108 or 1.2 x 109)

Preincubation, 37', 15 min

6000; g, 4', 10 min

bacterial pellet (I) Supernatant (I)(bound protein) (unbound protein)

1 1~~~~~~~~~~~~~~~~~~~~~~~~~~~IResuspension in standard assay: 1) bactericidal and/or PI

incubation mixture containing activities vs. 5 x 10740 mm MgC1] (sensitive) E.coli S15

incubation, 37', 15 min 2) double immunodiffusion

I6000 g, 4', 10 min

pellet supernatant (II)

assay: double immunodiffusion

FIGURE 2 Flowsheet of steps in assay of binding of bacterici-dal/PI proteins to bacteria.

rabbit leukocyte proteins are dose dependent (1, 2). Boundprotein can therefore be quantitatively determined by com-paring the bactericidal and PI activities of supernatant I ob-tained from preincubation mixtures with or without (con-trol) bacteria. Each supernate is tested at various concen-trations in the standard incubation mediuim against 5 x 107(sensitive) E. coli S15. Under these experimental conditions,loss of activity from samples of protein preincubated alone,presumably because of adsorption to the glass tube, was<20%.

Binding of the human PMNprotein to the bacteria was alsodemonstrated by the loss of immunochemically reactive pro-tein from supernatant I fraction and its recovery in supernatantII fraction after release from the bacterial pellet I (see Fig. 2)upon resuspension in 0.75 ml of the standard incubation mix-ture supplemented with 40 mMMgCl2 and incubation at 370Cfor 15 min. The supernatant II fraction was separated from thebacteria by centrifugation at 6,000 g for 10 min at 4°C.

Immnunnocheinical analysis of the two suipernatant fractionswas carried out by double immunodiffusion against rabbitaniti-humnan bactericidal/PI protein-serum, obtained as pre-viotisly described (1), in 1% agarose gels according to themethod of Ouichterlony (25).

RESULTS

Inifluetnce of saccharide chaini length of LPS onsensitivittl of S. ti phimnuriumti 395 to bactericidallPlproteins. Three strains of S. typhimnurium 395, dif-fering in the polysaccharide chain length of their outermeml)rane LPS (see Fig. 1), were compared for theirsensitivity to the bactericidal and PI activities of thehtumian and rabbit bactericidal/PI proteins. Fig. 3 showsthat the sensitivity of these strains to the bactericidalactivity of both proteins is inversely related to theLPS-saccharide chain length. Thus, ROcontaining theshortest-chain LPS is mnost senisitive and MS, thesimooth parent strain witlh complete LPS, is least sensi-tive. The bactericidal activities of the huiman andrabbit leukocyte proteins are nearly identical.

As with all other sensitive gram-niegative bac-terial strains previously examiniied (1, 2), killing of S.tyjphinurimn 395 RIO and( R5, by either the humanor rabbit bactericidal/PI protein, is accomiipanied by analmost imminediate increase in ouiter inemibrane perme-ab)ility to normially imnpermeanit ActD (Fig. 3). Bothbactericidal proteins exert their effect oni bacterialviability without substantially inhibiting bacterialmacromiioleclllar synthesis (1, 2). Thus, incorporation of['4C]leuicine into cold trichloroacetic aci(l-precipitablemiaterial by strains R10 and R5 is onily slightly in-hibited during incubation with bactericidlal concentra-tions of either protein in the absence of ActD(Fig. 3). However, in the presence of ActD, 1oth pro-teins cauise a dose-dependenit inhibition of bacterialproteini synthesis that paralllels their l)actericidalaction.

Neither the humnani nor the rabbit protein pro-dtices detectable killing of S. typhimurlium 395 MSat the concentrations of protein and bacteria tused in

Bacterial Resistance to Bactericidal Letnkocyte Proteins 621

/Human

I I-F5 10

Rabbit

I--

0 5 10

I T

0 5 10±g Protein

FIGURE 3 Effect of LPS structure on sensitivity of S. ty-phimurium 395 to human and rabbit PMNbactericidal/PIproteins. Of three strains of S. typhimurium 395 (MS [chemo-type S], 0; R5 [chemotype Rb3], C; R10 [chemotype Rd], 0),5 x 107 bacteria were incubated for 30 min at 37°C with in-creasing concentrations of either human or rabbit bacteri-cidal/PT protein (abscissa). The upper panels show bacterialviability expressed as percent (ordinate) of the number ofcolony forming units of a given bacterial strain incubatedalone. The lower panels show effects on bacterial permea-bility, as determined by measuring bacterial incorporationof ['4C]leucine (['4C]leu) into trichloroacetic acid-precipi-table material in the absence ( ) and presence (--- ) ofActD as described in Methods. Leucine incorporation isexpressed as percent (ordinate) of incorporation by bacteriaincubated alone (>2,000 epm; 1-2 nmol). The results repre-sent the means of three closely similar experiments.

the experiments shown in Fig. 3. However, killing isreadily apparent upon exposure of fewer bacteria(MS) to higher concentrations of the two proteins(Table T). Table I further shows that a fourth strain ofS. typhimurium 395 R2 (chemotype Ra), with an LPSpolysaccharide chain length intermediate between thatof MSand R5, exhibits a correspondingly intermediatesensitivity to the two proteins. These findings suggestthat the resistance of strains of S. typhimurium 395to the leukocyte bactericidal/PI proteins is related tothe length of their LPS-polysaccharide chain. How-ever, other differences in the envelope structure ofthese strains might also contribute to these differ-ences in sensitivity.

UDP-galactose-4-epimeraseless mutants: sensitivityto bactericidallPI proteins after growth with or with-out galactose. The influence of LPS-polysaccharide

TABLE ISensitivity of Salmonella typhimurium 395 to the Bactericidal

Activity of Purified BactericidallPI Protein from Humanor Rabbit PMN: Effect of Polysaccharide Chain

Length of Outer Membrane LPS

Add bactericidal/PI protein.

Strain Hunian Rabbit

l.gl/ml

R10 (Rd) 1-2 1-2R5 (Rb3) 4-5 3-4R2 (Ra) 20-30 6-9MS >30 30

Bactericidal activity was determined as described in Methods.Results shown represent the amount of protein (microgramsper milliliter) required to produce >90% killing of 107bacteria/ml and are the mean of at least three experiments.

chain length on gram-negative bacterial resistance tothe bactericidal/PI proteins can be more specificallyexamined by using mutants of S. typhimurium (G-30)and E. coli (J5) that lack the enzyme UDP-galactose-4-epimerase. These mutants are conditionally defectivein LPS biosynthesis, synthesizing galactose-deficient,incomplete LPS (chemotype Re, see Fig. 1) whengrown without added galactose. However, when grownin media supplemented with galactose, these bacteriaproduce complete LPS (chemotype S) that is indis-tinguishable from LPS of its smooth parent strain.In contrast to their differences in LPS structure, thebacteria grown with or without galactose manifest nodetectable differences in outer membrane protein andphospholipid content (26, 27).

Figs. 4 and 5 show that the synthesis of completeLPS by S. typhimurium G-30 or E. coli J5 is accom-panied by increased resistance to the human andrabbit bactericidal/PI proteins. Whereas 4-5 gg ofeither the human or the rabbit protein producemaximnal bactericidal and PI effects toward bothmutant strains grown without galactose, >10 jig ofeither PMN protein is needed to produce similareffects on the mutant strains after their growth withgalactose. Supplementation of growth media with con-centrations of galactose as low as 0.25 mMelicitsthis increase in resistance. In contrast, resuspensionof the mutants in 20 mMgalactose just before incuba-tion with human or rabbit leukocyte protein does notincrease bacterial resistance. The resistance of E. coliJ5 grown with galactose is practically identical to thatof its smooth parent strain, E. coli 0111:B4.

The bactericidal and PI activities of the human andrabbit PMNproteins toward all S. typhimurium andE. coli strains that we have studied are remarkablysimilar. However, whereas these activities of the twoproteins are nearly identical toward bacteria with short-

622 J. Weiss, S. Beckerdite-Quagliata, and P. Elsbach

0-a0_

~3.5 (.3= 0

00

e-

0 (P

(.3

b.o

0-

o qJ._.c

_.

=._.a)vi

0

Rabbit Human

_3'b sm-------- --.

_ - - Act D

%\ 1--- = + Act 0

b-----o

0 5 10

b

b-- eI

0 5 10

,Ug Protein

FIGURE 4 Effect of LPS struetture on sensitivity of S. typhi-mu1ilritlmn G-30 to hulnan ancl rab)lit PMN bacterici(lal/PI

proteins. Viability (uipper panels) andl permeability (lowerpanels) of S. typhinmtriumn G-30 (5 x 1O(7 bacteria) incuihatedwith increasing concentrations of the humtian and rabbit pro-

teins after bacterial growth in the presence (M) or absence

(Z) of 5 miN1 D-galactose were determined as described inMethods and the legend to Fig. 3. The resuilts represent themean of three closely similar experimenits.

chain LPS, the rabbit protein is consistently slightlyimore active than the human protein toward bacteriawith longer-chain LPS.

Binding of bactericidallPI proteins to S. ttyphi-mnuriumn and E. coli: effect of LPS-saccha ride cliaini

lenigth. To prodtuce their antibacterial effects, thebactericidal/PI proteins apparently must bind to thebacterial outer memnbrane (3, 5). Longer-chaiin LPSmight enhance bacterial resistance by redtucing theaffinity of the ouiter membrane for the bactericidal/PIproteins. Therefore, we compared binding of thehtuman and rabbit proteins to bacteria containinig shortor long chain LPS. As described in Miethods, we have

been compelled to measure binding indirectly: that is,as the loss of antibacterial (bactericidal and PI) activityfrom the soluble phase during incubation of thebactericidal/PI proteins with different bacterial strains.

Figs. 6-8 show that much less of the bactericidal andPI activities are recovered in the supernatant fractionafter preincubation of either protein with rouighI)acterial strains than in the supernatant fraction afterpreincubation with smooth strains. This prestumablyreflects greater binding of the two proteins to bacteriawith short-chain LPS. The extent of apparent protein

a1)0

- 0D._

b-C.), 0

0

o O)

crZ

Cl

o

C.0 .C

C-)* 0

a),

.13 ,C

10 0,1g Protein

FIGURE 5 Effect of LPS strulctture on1 sensitivitv of E. colito hulmlani aind rabbit bactericidal/P1 proteins. Viability (uipperpanels) and permiieability (lowver panels) of E. coli O111:B4aind of E. coli J5, grown in the presence (e) or absence (O) of5 mXMD-galactose and incuibated with increasing concentra-tions of the huianii and rabbit proteins, were determnined as

descril)ed in Methods and the legend to Fig. 3. The resiultsrepresent the mean of three closelv similar experiments.

binding to bacteria with either short- or long-chain LPSdepends on the number of bacteria present in the pre-

incubation mixture (Figs. 6-8). Wecalcutlate fromi theseexperimiients that during incutbation of 4.5 Ag of eitherprotein1 with 4 x 107 bacteria --3 uLg is bound to eachof the rouighl bacterial strains, but only 0.3-1.1 utg isl)ounid to the smnooth strainis (Table II).

The extent of apparent binding of the bactericidal/PIproteins generally correspon-ds to their biological ef-fects. Thus, the lower antibacterial potency of thebactericidal/PI proteins toward the smooth bacteriaparallels their reduced blinding. This correlation ismost clearly illustrated by comnparing the apparentbinding anid biological activity of either the htumanor the rabbit PXIN protein toward the UDP-galactose-4-epimeraseless mutants S. typhirnuriulm G-30 andE. coli J5. When grown withouit galactose, these twomuttants exhibit a similar sensitivity to that of the Rdmutant of S. typhimurium 395 RIO and bind approxi-mately the same amouniit of protein. However, aftergrowth in the presence of galactose, both the bindingand sensitivity of these two mutant strains becomegreatly redtuced and indistingtuishable fromn that ofsmnooth strainis (S. typliphmrimn 395 XIS; E. coli 0111:

Bacterial Resistatnce to Bactericidal Leukocyte Proteitns

Human Rabbit

- u-a)C0

o ._

-. 50-C.>0

0

I-e

a 100-O w

o a)C3o -oC.. OO ._

.' ,, 50-C-)0,6 C

Jz_

a)-,Z3'a.

I

623

Human

a)

_cC:>*0

c 00

0 =.:" .

I.. 50-

_ wcJ

).O-

0

c 100-0 _

0

C3o0 =

.' 0 50-. -

_ 18

-

0

> " ; ~~~~~^ 1.2x109

''$ K- 6X108'% -' 1. 2 x 109

I --F x0

5 75 150O±I Supernotant

FIGURE 6 Binding of humnan and rabbit bactericidal/PIproteins to S. typhimurium 395: influence of LPS structure.Supernatant fractions were obtained as described in Methodsafter preincubation of 50 ,g of the human or the rabbitbactericidal/PI protein alone (A), or with 6 x 108 (circles),or with 1.2 x 109 (squares) S. typhimurium 395 (MS, closedsymbols; R10, open symbols). These fractions were assayedfor bactericidal and PI activities toward 5 x 107 E. coli S15as described in Methods. For simpler presentation, only[14C]leucine (['4C]leu) incorporation in the presence ofActD (--- ) is shown. The results represent the mean of twoclosely similar experiments carried out in duplicate.

100-10 -

_ 0

o C

C.

C ,,, 50-*. ui

-i

0

Human100- tr,~-Z:-z7z --------0

5X " 1.2X1O9 X'

"\\ \\ 6X0\ \\

_____ £_I-r1 75 150 0

ci Sunernatant

Rabbit

Pre -/ncubation:a Protein alone *,- +G-30(+golactose) 0,0 +G-30 (-galoctose)

FIGURE 7 Binding of human and rabbit PMNbactericidal/PIproteins to S. typhimnuriumn G-30: influence of LPS structure.PI activity of the supernatant fractions, recovered afterpreincubation of 50 ,ug of the humnan or the rabbit bactericidal/PI protein alone (A), or with 6 x 108 (circles), or with 1.2 x 109(squares) S. typhimurium G-30 grown in the presence (closedsymbols) or absence of galactose (open symbols) was deter-mined as described in Methods and the legend to Fig. 6. Theresults represent the mean of two closely similar experimentscarried out in duplicate.

Rabbit

0O 0±I Supernatant

A Protein alone 0 + 0:111: B4 * + J5 (+ galactose)(1.2 X109) (I.2 X109)

75 150

+ J5 (- galoctose)O (6XI08)O (1.2 X109)

FIGURE 8 Binding of human PMNbactericidal/PI proteinto E. coli: influence of LPS structure. Supernatant fractionswere obtained as described in Methods after preincubationof 50 ,ug of human PMNprotein alone (A), or with E. coli01 11:B4 (1.2 x 109; *), or with E. coli J5 grown with galactose(1.2 x 199, *), or without galactose (6 x 108, 0; 1.2 x 109, O).These fractions were assayed for bactericidal and PI activitiesas described in the legend to Fig. 6. The results representthe mean of two closely similar experiments carried out induplicate.

B4). Furthermore, the human and rabbit proteins exhibitvery similar activity and bind in nearly identicalamounts to rough bacteria, whereas the somewhatgreater activity of the rabbit protein toward smoothbacteria is matched by its greater apparent binding.

Immunochemical analysis of the supernatant frac-tions obtained by centrifugation of the preincubationmixtures provides evidence that the loss of bactericidaland permeability increasing activities from the solublephase is indeed due to binding of the bactericidal pro-teins to the bacterial pellet and not to inactivation.

TABLE IIComparison of Binding of Humanand Rabbit BactericidallPI

Proteins to Various Strains of Gram-negative Bacteria

Bacterial strain Human Rabbit

jg protein boundl5 x 107bacteria

RoughS. typhimurium 395-RlO 2.8-3.4 3.0-3.3S. typhimurium G-30 (-galactose) 2.7-3.1 2.5-2.9E. coli J5 (-galactose) 2.6-3.4 NT*

SmoothS. typhimurium 395-MS 0.2-0.5 0.6-0.9S. typhimurium G-30 (+galactose) 0.4-0.6 0.8-1.1E. coli 011l:B4 0.3-0.5 NT*E. coli J5 (+galactose) 0.3-0.7 NT*

Protein binding was calculated as described in Methods,using the data shown in Figs. 5-7. Results shown are fromtwo or more experiments and represent the minimum andmaximum values obtained.* Not tested.

624 J. Weiss, S. Beckerdite-Quagliata, and P. Elsbach

In this experiment we mnade use of our previous ob-servation (3, 4) showing that addition of Mg2+ (or Ca2+) tosuspensions of sensitive bacteria exposed to human orrabbit PMNprotein initiates repair of the permeabilitybarrier and other envelope alterations.

After incubation of the human protein with E. coli J5(grown without galactose), no bactericidal/PI proteinis evident, immnunochemiically, in the supernatantfraction (Fig. 2). However, after resuspension and in-cubation of the bacterial pellet in Tris-buffered salineplus 40 mMMgCl2, the supernatant fraction II (seeMethods and Fig. 2), obtained after centrifugation,produces an iminunoprecipitation reaction with rabbitantiseruin against human bactericidal/PI protein. Thedensity of this immunoprecipitate, comnpared with thatof precipitates generated by known amounts of humilanPMNprotein in control wells, is consistent with re-lease from the bacteria of >50% of the proteininitially added to the bacterial suspension (Fig. 9).Sodium dodecyl sulfate-polyacrylamnide gel electro-phoresis of supernatant II fraction reveals one miajorprotein component that migrates exactly like thebactericidal/PI protein (not shown). Similar analysisof supernatant fractions I and II obtained from bacterialsuspensions incubated without protein shows no re-

2 R *

,--t ,5)

N .

/

..I

s.v . .._,, ,

...PI

FIGURE 9 Inmmunochemnical evidence of binding of humanPMN bactericidal/PI protein to E. coli J5. Supernatantfractions I and II, obtained as described in Methods andFig. 2, were removed from bacterial suspensions after pre-incubation with (top left) or without (top right) 50 gg of humanPMNprotein and were assayed by double immunocliffusion(25). Aliquots of 20, 40 and 60 ,l of supernatant fractionsI and II were added to peripheral wells numbered 1, 2, and 3and wells 4, 5, and 6, respectively. Aliquots (20,40, and 60 ,ul)of a supernatant fraction derived from 50 ug protein pre-incubated in 0.75 ml saline containing 40 mMMgCl2 wereadded to peripheral wells 2, 4 and 6, respectively (lower left).Rabbit antiserum against human bactericidal/PI protein wasadded to each center well.

lease of such mnaterial as assessed by immnunoreactivityor by gel electrophoresis.

DISCUSSION

Numerous studies have shown that the surface prop-erties of smnooth gram-negative bacterial strainisprovide imore protection against antibiotics and thehost's humioral and cellular defenses than do those ofrough bacteria (6, 7, 14-18). Although experimientswith S. typhimurium and E. coli strains differing inthe polysacclharide chaini length of the outer membraneLPS have suggested that LPS structure is a majordetermninant of gramn-negative bacterial virulence (6), ithas so far not become clear what the basis is for thisapparent relationship. In this study we have shown thatbacterial sensitivity or resistance to the effects ofpurified bactericidal proteins fromn humilan anid rabbitPMNon viability and envelope permneability is a funic-tion of the binding properties of the outer membranethat vary with the length of the polysaccharide chainof the LPS.

Our previous observations, showing that the bacteri-cidal/PT protein is remnoved from the inediumn duringincubation with sensitive E. coli but can be fullyrecovered fromn the imediumn when incubated with re-sistant microorganismns such as Serratia inarcescens orStaphylococcus aureus (5; unpublished observations),strongly suggested that binding to the microbial en-velope is an absolute requisite for expression of theantibacterial activities of these proteins. Althoughthese earlier studies did not reveal which of the mianyenvelope constituents determined whether or notbinding takes place, several findings focused our at-tention on the surface layers of the gram-negative en-velope: (a) all of the microbial species tested lackingthe gram-negative bacterial outer memnbrane (includingseveral gramn-positive bacterial species and twostrains of the fungus Candida) show no sensitivity tohigh concentrations of the bactericidal/PI proteins(1, 2), and (b) sensitive gram-negative bacteria exposedto bactericidal concentrations of the bactericidal/PIproteins exhibit almost unimpaired macromnolecularsynthesis and K+ transport (1-3; unpublished ob-servations), indicating that the structure and functionof the cytoplasmic membrane remain singularly un-affected. In contrast, the two envelope alterations thataccompany the bactericidal action of the purifiedleukocyte proteins appear to be confined to the outermembrane, namely, the breakdown of the normal per-meability barrier in the outer membrane to ActD (23,28) and the activation of bacterial phospholipid-splitting activity that resides predominantly in theouter memnbrane (29-31). That the bactericidal/PIproteins bind to sites on or near the surface is furthersuggested by the reversal of these envelope effects,

Bacterial Resistance to Bactericidal Leukocyte Proteins 625

either by addition of Mg2+ (or Ca2+) with concomitantrelease of bound leukocyte protein (Fig. 9) (but notof appreciable bacterial protein) or by trypsin treat-ment (unpublished observations) that only destroysprotein at the bacterial surface (32). Binding to surfacesites, the relatively large size of the bactericidal humanand rabbit leukocyte proteins (Mr 60,000 and 50,000,respectively), and the apparently discrete lesions pro-duced, all together presumably prevent these proteinsfrom traversing beyond the outer membrane.

A common feature of the antimicrobial proteins ofPMNis their positive charge (33-35). Polyanions, suchas LPS and heparin, complex and inactivate thesecationic proteins, and it has been proposed in the pastthat their action follows binding to negative chargesin or on the envelope (5, 33, 34). At the surface of thegram-negative envelope the LPS are the most domi-nant negatively charged components (36). Currentviews of the molecular organization of the outer mnem-brane hold that the phospholipids mnake up the innerleaflet of the hydrophobic bilayer and the LPS theouter leaflet (12). According to this concept, imost orall LPS molecules in the outer imembranie turn theirnegative charges toward the exterior environmient andthus becomiie available for interaction with opposingcharges. However, the negative charges of the LPS(36) are located near the base of the polysaccharidechain, where this is attached to the lipid A moiety em-bedded in the bilayer (12, 36). Conse(quently, as thepolysaccharide chain lengthens, the negative chargesare more buried, and chalrge interaction witlh mnoleculesin the environmient may becomie hindered. It is in thislight that we interpret our resiults of experimiients withE. coli and S. typhimuriumrl strains as differing in thepolysaccharide chain length of their LPS (Fig. 10). Wepropose that approximiately the samne numinber of bac-tericidal protein molecules produce their effects onviability and permeability of rough and smiiooth strainsof E. coli and S. typhimluriumn. However, because theapparent binding affinity of the bactericidal proteinsvaries with the polysaccharide chain length of the LPS,these effects are elicited at different protein coIn-centrations. The number of sites per bacteriumn, esti-miated by the miiaximal aimount of bound protein and itsmolecular weight (1, 2), is roughly 5 x 105. The numberof LPS inolecules per E. coli, regardless of polysac-charide chain length, has been calculated to be -2.5x 106 (12). Thus, saturation of binding sites wouldcorrespond stattistically to one bactericidal proteininolecule bounid per five LPS imolecules. Mlore precisedetermnination of the binding properties of differentl)acterial strains with respect to these bactericidalproteins must await their successful radiolabeling.

How these apparently noincatalytic proteins affectthe bacterial envelope and imultiplication by inter-acting with the surface layer is not clear. The studies

Cationic Protein

GL)

a,0

Q

C.)

0

m

OMf

Pep. Glyc.

IM 0_

IOM

_R )

FIGURE 10 Model describing relationship between LPS-saccharide chaiin length and action of bactericidal/PI proteinon gramn-negative bacteria. The gram-negative bacterial en-velope consists of three distinct layers (28,36): the cytoplasmicor inner membrane (IM), the peptidoglyean (Pep. Glyc.)layer, and the ouiter membrane (OM). The LPS of the outermemnbrane mtainly accouint for the net negative charge typicalof intact gram-negative bacteria. In the envelopes of S.typhinmuriumn acnd E. coli with short-chain saccharide-LPS(Rc), these negative charges are suifficienitly exposed to allowready interaction with the positively charged bactericidalproteins. The longer the polysaccharide chains of the LPS,the more the negative charges are shielded, thereby reducingthe likelihood of charge-charge interaction (S). However, athigher acimbient concentrations of bactericidal proteins,binding does occur and, when reaching satturation (5), theeffects oni viability and permnieability are the samiie as on roughstrains.

of Nikaido (13) and Leive (28) suggest that exogenoushydrophobic stibstances, such as ActD, cannot per-ineate the intact outer memnbrane becauise of a hydro-philic barrier external to the hydrophobic region ofthe ouiter membrane. Promnpt passage of ActD ocscurs,however, wheni conditions are created that exposethe hydrophobic bilayer: for examiple, when EDTAtreatmiienit chelates divalent cations cross-linking outermemnbrane comnponents (36).

Apparenitly, the interactioni of the bactericidal pro-teins witlh the outer mnembrane also genierates hydro-phobic channiels allowinig penietrationi of ActD. Attach-imient of the cationic bactericidal protein to the negativesuirface charges may produce such anl effect throughconforimnational changes in outer membrane coimi-ponents. It is also possible that the tightly membrane-bound bactericidal leukocyte proteins (1, 5) possesshydrophobic domiains that, after the initial ionic inter-actioni of proteini and bacteriumil, insert themiselvesinto the hydrophobic bilayer acting as detergents.

As shown in this and previouis sttudies in our labora-tory, the remarkable siinilarity of the biological proper-ties of the humnan and rabbit PMNbactericidal/PI pro-teins (1, 2) is miatched by the close resemnb-lance of theirmiiolecular clharacteristics, includ(ling their molectular

626 J. Weiss, S. Beckerdite-Quiagliata, and P. Elsbach

weight, amino acid composition, net charge, andthermal sensitivity (1, 2). In addition, the rabbit PMNprotein shows cross-immunoreactivity with antiserumgenerated against the human PMNprotein (unpub-lished observations). It further appears that the role ofthese proteins in the cell, derived from the human andfrom the rabbit, is also similar. Thus, the effects ofintact human and rabbit PMN, whole PMNhomoge-nates, and progressively purified PMN fractions, onviability and permeability of rough gram-negativebacterial strains, are essentially the same as of theisolated bactericidal/PI protein (1, 2, 5, 24). Other PMNfractions obtained during purification of the bacterici-dal/PI protein, however, are devoid of these activities.Such findings suggest that the bactericidal/PI proteinis the principal, 02-independent antibacterial agentdirected at these gram-negative bacteria.

ACKNOWLEDGMENT

This work was supported by grant AM05417 from the U. S.Public Health Services.

REFERENCES

1. Weiss, J., P. Elsbach, I. Olsson, and H. Odeberg. 1978.Purification and characterization of a potent bactericidaland membrane active protein from the granules of humanpolymorphonuclear leukocytes.J. Biol. Chem. 253: 2664-2672.

2. Elsbach, P., J. Weiss, R. C. Franson, S. Beckerdite-Quagliata, A. Schneider, and L. Harris. 1979. Separationand purification of a potent bactericidal/permeability-increasing protein and a closely associated phospholipaseA2 from rabbit polymorphonuclear leukocytes. Obser-vations on their relationship. J. Biol. Chein 254: 1 1000-11009.

3. Weiss, J., K. Schmeidler, S. Beckerdite-Quagliata, R.Franson, and P. Elsbach. 1976. Reversible envelopeeffects during and after killing by a highly pturifiedlgranulocyte preparation. Biochimn. Biophys. Acta. 436:154-169.

4. Weiss, J., and P. Elsbach. 1977. The use of a phospholipaseA-less Escherichia coli Inutant to establish the action ofgranulocyte phospholipase A on bacterial phospholipidsduring killing by a highly purified granulocyte fraction.Biochim. Biophys. Acta. 466: 23-33.

5. Weiss, J., R. Franson, S. Beekerdite, K. Schmeidler, andclP. Elsbach. 1975. Partial characterization and purificationlof a rabbit granulocyte factor that increases permeabilityof Escherichia coli. J. Clin. Invest. 55: 33-42.

6. Medearis, D. N., B. M. Camitta, and E. C. Heath. 1968.Cell wall composition and virulence in Escherichiacoli. J. Exp. Med. 128: 399-414.

7. Stendahl, O., and L. Edebo. 1972. Phagocytosis of muttantsof Salmonella tyxphimnuriunm by rabbit polymorphonuclearcells. Acta Pathol. Microbiol. Scand. Sect. B Microbiol.Immunol. 80: 481-488.

8. Van Oss, C. J., and C. F. Gillman. 1972. Phagocytosisas a surface phenomenon. I. Contact angles and phago-cytosis of non-opsonized bacteria. J. Reticuloendothel.Soc. 12: 283-292.

9. Cunningham, R. K., T. 0. Soderstrom, C. F. Gillman,

and C. J. van Oss. 1975. Phagocytosis as a surface phenom-enon. Contact angles and phagocytosis of rough andsmooth strains of Saltzoniella tlphintinrinmn and the in-fluence of specific antiserumiii. Im mtu nol. Cotrn mlun. 4:429-442.

10. Sanderson, K. E., T. MacAlister, J. W. Costertoni, ain(I K-J.Cheng. 1974. Permeabiltiy of lipopolysaccharide-defi-cient (rough) mutants of Salmtioniella typhimnurinmn toantibiotics, lysozvme and other agents. Can. J. Mlicrobiol.20: 1135-1145.

11. -Magnusson, K. E., 0. Stendahl, C. Tagessoni, L. Edebo,and G. Johanssen. 1977. The tendency of smilootlh anidrough Salmloniella tijphinim rin m bacteria and lipopoly-saccharide to hydrophobic and ionic interactioni, asstudied in aqueous polymer two-phase systemns. ActaPathol. Microbiol. Scatnd. Sect. B Microbiol. Imilinulnol.85: 212-218.

12. Smit, J., Y. Kamio, and H. Nikaido. 1975. Ouiter membraneof Salmonella typhimutriumn: cemleiiical anialysis andl freeze-fracture sttudies with lipopolvsacccharide muttants. J.Bacteriol. 124: 942-958.

13. Nikaido, H. 1976. Outer imemiibranie of Salmiionellatyphimuritumni: transmnemiibranie diffusion of soimie hydro-phobic substances. Biochim. Biophys. Acta. 433: 118-132.

14. Friedberg, D., I. Friedberg, anid Mi. Shilo. 1970. Inter-action of gram-negative bacteria with the lysosomnal frac-tion of polymorphontuclear leuikocytes. II. Changes in thecell envelope of Escherichia cOli. Infect. Immun. 1:311-318.

15. Tagesson, C., and 0. Stendahl. 1973. Influence of cellsurface lipopolysaccharide structture of Salmiontella typh i-inn ri n in on resistanice to intracellular bactericidalsystems.Acta Pathol- Microbiol. Scatnd. Sect. B .ficrobiol.Immnluno. 81: 473-481.

16. Rest, R. F., NI. H. Cooney, anid J. K. Spitzniagel. 1978.Bactericidal activity of specific andl azturophil granutilesfrom human neutrophils: stu(lies with outer imiembraniemutants of Salmontella tmlplhii imriumirl LT-2. Intfect. I ninum.19: 131-137.

17. Modrzakowski, NI. C., MI. H. Coonev, L. E. Martin, andlJ. K. Spitznagel. 1979. Bactericidal activity of fractioniatedlgranule contents from humilan polymorphonuclear leuko-cytes. Infect. I mmun. 23: 587-591.

18. Reynolds, B. L., and H. Prtutil. 1971. Protective role of.smooth lipopolysaccharide in the serumii bactericidalreaction. Infect. Immrut1n. 4: 753-771.

19. Holme, T., A. A. Lindberg, P. J. Garegg, and T. Oinl. 1968.Chemical composition of cell wall polysaccharide inrouigh imtitants of Salm)}on)ella tphimnurimni. .1. Gen).Microbiol. 52: 45-54.

20. Osborn, NI. J., S. NI. Roseni, L. Rothfield, anid B. L.Horecker. 1962. Biosvnthesis of bacterial hipopolv-saccharide. I. Envelope incorporation of galactose inl amultant straini of Salmtioniella. Proc. Natl. Acad. Sci. U.S.A.48: 1831-1838.

21. Elbein, A. D., and E. C. Heath. 1965. The b)iosvnthesis ofcell wall lipopolysaccharide in Escherichia coli. J. Biol.Chemn. 240: 1919-1925.

22. Simon, E. J., and D. Van Praag. 1964. Inhibition of RNAsynthesis in Escheric hia coli by levorphanol. Proc. Natl.Acad. Sci. U.S.A. 51: 877-883.

23. Hturwitz, J., J. J. Fuirth, MI. NMalainy, anid NI. Alexanider.1962. The role of DNA in RNA synthesis. III. The in-hilition of the enzymatic synthesis of RNA anid DNAby actinomycin and proflavin. Proc. Natl. Acad. Sci. U.S.A.48: 1222-1230.

24. Beckerdite, S., C. Xloonev, J. Weiss, R. Fransoin, K.

Bacterial Resistance to Bactericidal Letukocyte Proteins 627

Schmeidler and P. Elsbach. 1974. Early and discretechanges in permeability of Escherichia coli and certainother gram-negative microorganismns during killing 1hgranuilocytes.J. Exp. Med. 140: 396-409.

25. Ouchterlony, 0. 1949. Antigen-antibody reactions in gels.Acta Pathol. Microbiol. Scand. 26: 507-515.

26. Schniaitinan, C. A. 1974. Outer membrane proteinis ofEscherichia coli. IV. Difference in outer imemnbraneprotein due to strain anid cultural (lifferences.j. Bacteriol.118: 454-464.

27. Osborn, M. J., J. E. Gander, E. Parisi, and J. Carson. 1972.Mechanism of assemibly of the outer membrane of Sal-nionella tijphimurium. Isolation and characterization ofcytoplasmnic and ouiter membrane. J. Biol. Chein. 247:3962-3972.

28. Leive, L. 1974. The barrier fuinction of the gramii-niegativebacterial envelope. Ann. N. Y. Acad. Sci. 235: 109-129.

29. Bell, R. M., R. D. Mavis, M. J. Osborn, anid P. R. V'agelos.1971. Enzymnes of phospholipid metabolismn: localizatioilin the cytoplasmic and ouiter ineinbrane of the cell en-velope of Escherichia coli and Salmontella tiyphinm urium.Biochim. Biophys. Acta. 249: 628-635.

30. White, D. A., F. R. Albright, W. J. Lenniiarz, and C. A.Schnaitm-an. 1971. Distribution of phospholipid-syn-thesizing enzyvmles in the wall andl imemibranie sub-

fractions of the envelope of Escherichia coli. Biochi in.Biophys. Acta. 249: 636-647.

31. Vos, M. XI., J. A. F. Op dcen Kamp, S. Beckerdite-Quiagliata, and P. Elsbaclh. 1978. Acylation of mono-acylglycerophosphoethanolam-ine in the inner and outermemnbrane of the envelope of an Escherichia coli K12strain and its phospholipase A-deficient miiutant. Biochimii.Biophys. Acta. 508: 165-173.

32. Maede, A., anid M. Noomura. 1966. Interaction of colicinswith bacterial cells. I. Studies with radioactive colicinis.

J. Bacteriol. 91: 685-694.33. Hirsch, J. G. 1960. Further stutdies on preparation anid

properties of phagocytin. J. Exp. AMed. 111: 323-337.34. Zeya, H. I., and J. K. Spitznagel. 1966. Cationic p)roteins

of polymorphonticlear leukocyte lysosomiies. II. Comllposi-tioIn, properties and mnechlanismii of antibacterial actioni.

J. Bacteriol. 91: 755-762.35. Odeberg, X{., I. Olssoni, and P. Venge. 1975. Cationiic

proteins of human granulocytes. J. Cliti. Itnvest. 56:1118-1124.

36. Ghtuysen, J. MI. 1977. Biosynthesis andcl assembly ofbacterial cell walls. In The Synthesis, Assembly andTturnover of Cell Surface Components. G. Poste and G. L.Nicolson, editors. Elsevier/North-Holland, BiomnedicalPress., New York. IV: 463-595.

628 J. Weiss, S. Beckerdite-Quagliata, anid P. Elsbach