ANTiMICROBIAL AGENTs AND CHEMOTHERAPY, OCt. 1973, p. 402-409Copyright 0 1973 American Society for Microbiology

Vol. 4, No. 4Printed in U.S.A.

Structure-Activity Relationships Among theAminoglycoside Antibiotics: Role of Hydroxyl

and Amino GroupsR. BENVENISTE' AND J. DAVIES

Department of Biochemistry, College of Agricultural and Life Sciences, University of Wisconsin-Madison,Madison, Wisconsin 53706

Received for publication 10 April 1973

The close structural similarity of several of the deoxystreptamine-containingaminoglycoside antibiotics (gentamicins, neomycins, kanamycins, tobramycin)and the recent isolation of enzymatically N-acetylated aminoglycosides havepermitted a systematic comparison of the structure-activity relationships in thisgroup of antibiotics. The number and location of amino groups on the hexosesand the site of attachment of the other rings to deoxystreptamine have beenshown to exert a profound effect on the ability of these compounds to inhibit or tocause misreading of polypeptide synthesis in vitro. The conclusions allow certainpredictions to be made concerning the interaction of these compounds with thebacterial ribosome.

Most studies on structure-activity relation-ships among antibiotic analogues have beencarried out on whole cells. Although this ap-proach is important in finding new antibiotics,it does not always give exact structure-activityinformation since many compounds may notenter the cell. Studies of antibiotics at their siteof action have been somewhat limited, althoughcell-free systems which carry out the accuratesynthesis of macromolecules are available. Onesuch system is protein synthesis, where a vari-ety of antibiotics have been useful in mechanis-tic studies.The aminoglycoside-aminocyclitol antibiotics

(streptomycin, spectinomycin, neomycins, pa-romomycin, kanamycins, gentamicins, to-bramycin) inhibit protein synthesis through aninteraction with the 30S ribosomal subunit (12).Structure-activity studies with these antibioticsmight provide useful information on the chem-istry of the ribosome, as well as suggestingroutes for synthetic or semisynthetic productionof antibiotics.

Past studies have shown marked differencesin activity between different classes of theseantibiotics, but recently a number of new deoxy-streptamine antibiotics and some enzymati-cally modified derivatives of gentamicin, kana-mycin, and neomycin have become available. A

' Present address: Viral Leukemia and Lymphoma Branch,National Cancer Institute, Bethesda, Maryland 20014.

study of the effects of these compounds onprotein synthesis in vitro has provided moreclues as to the functional groups necessary formaximal biological activity.

MATERIALS AND METHODSIn vitro R17 phage ribonucleic acid (RNA)-

directed polypeptide synthesis. Reaction mixtures(100 Mliters) contained 66 mM tris(hydroxymethyl)-aminomethane-hydrochloride (pH 7.8 at 5 C), 9 mMmagnesium acetate, 50 mM NH4Cl, 2.6 mM adeno-sine triphosphate, 60 MM guanosine triphosphate,10 mM phosphoenolpyruvate, 2 Ag of pyruvatekinase, 10 mM 2-mercaptoethanol, 3.3 mM dithio-threitol, 0.5 mM of each of the 20 amino acids(excluding valine), 50 ug of transfer RNA, 4 nmol of[14C]valine (25 uCi/Mmol), and 1.6 OD,.. (opticaldensity at 260 nm) units of a preincubated S-30 ex-tract (8) of Escherichia coli MRE-600 (prototroph,RNase-). Incorporation was started by the additionof 1 OD,., unit of R17 phage RNA. Incubation wascarried out for 40 min at 37 C; then 1.5 ml of 10% tri-chloroacetic acid was added and the tubes wereheated at 90 C for 20 min, cooled, filtered on glass fi-ber disks, and counted in a liquid scintillation counter.Misreading of poly U. Reaction mixtures (100

Mliters) were the same as above except 13 mMmagnesium acetate was used and 0.53 nmol of [14C]-isoleucine (128 MCi/Mmol), 0.2 nmol of [14C]tyrosine(330 MCi/Mgmol), and 0.43 nmol of ['4C]serine (156MCi/Mgmol) were added instead of [14C]valine. Incor-poration was started by the addition of 10 gg ofpolyuridylic acid (poly U). Incubation was carried outfor 15 min at 37 C, then 1.5 ml of 10% trichloroacetic

402

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

RELATIONSHIPS AMONG AMINOGLYCOSIDE ANTIBIOTICS 403

acid was added and the tubes were heated at 90 C for20 min, cooled, filtered, and counted.

Sources of antibiotics. Gentamicin A, C1, C12, C2,the gentamines, and sisomicin were provided byDavid Cooper and Peter Daniels of the ScheringCorp., tobramycin by Kay Koch of Lilly ResearchLaboratories, kanamycins A and B by the late A.Gourevitch of Bristol Laboratories, kanamycin C byH. Umezawa, neomycin and neamine by G. B. Whit-field of the Upjohn Co., paromomycin and butirosinby T. H. Haskell of Parke-Davis, methyl neobiosami-nide B by Kenneth Rinehart, and ribostamycin by T.Wakazawa. Synthetic substituted deoxystreptamineswere provided by Ray Lemieux. 2'-N- and 6'-N-acetylgentamicin Ci. (1), 3-N-acetylgentamicin Cl1(3), 6'-N-acetylkanamycins A and B (1), 6'-N-acetyl-neomycin B (1), and gentamicin C,2-adenylylate (2)were prepared by using enzymes obtained from re-sistant bacterial strains as previously described.

RESULTSThe deoxystreptamine-containing amino-

glycoside antibiotics include the neomycins andparomomycins (Fig. 1), kanamycins (Fig. 2),gentamicins (Fig.In order to obta

HO0HO

b'

HO

HO H2N

R4

NEOMYCIN B

NEOMYCIN C

HYBRIMYCIN Al

HYBRIMYCIN A2

HYBRIMYCIN BI

HYBRIMYCIN B2

PAROMOMYCIN

FIG. 1. Structurebostamycin, and par(mine) consists of rind3 + 4, and ribostam:cins contain a strepA2) or epistreptamof 2-deoxystreptamir4-amino-2-hydroxybLdeoxystreptamine. Atibiotics are N-ac'transferase to yieldo-phosphorylated bytransferase (b) andase (c).

CH2X' 0

HO

HO NH2b2

O 6 IIHO NH2

0

C i\OCH2H20HHO OH

H2N

X

KANAMYCINS A NH2y

OH

B NH2 NH2

C OH NH2

FIG. 2. Structure of kanamycins A, B, and C.Arrows indicate the sites of N-acetylation by kanamy-cin acetyl transferase to yield 6'-N-acetyl antibiotics(a), the site of o-phosphorylation by neomycin-kanamycin phosphotransferase (b), and the site ofo-adenylylation by gentamicin adenylyl transferase(c) .

3), and tobramycin (Fig. 4). structure-activity relationships in this group oftin a clearer picture of the antibiotics, we decided to test these com-

pounds, certain derivatives, and their degrada-tion products for inhibition of phage RNA-

<H2 __ directed protein synthesis and for misreading ofI poly U.,>H2N H N Effects of the antibiotics and their enzy-0Hc 0 H2N

R2 matically modified derivatives and degrada-CH2°H OH H2 tion products on R17 RNA-directed proteinkCmUl synthesis. The primary action of aminoglyco-

sides is inhibition of protein synthesis, and R17o H RNA-directed synthesis is sensitive to these

RRiRR3 R~ compounds at low concentrations. In Table 1

R Rl R2 R3 R4 are listed typical inhibition data for the com-NH2 H H H CH2NH2 pounds tested at various concentrations. It isNH2 H H CH2NH2 H clear that there is a considerable variation inNH2 OH H H CH2NH2 the activities of these compounds; the greatestNH2 OH H CH2NH2 H levels of inhibition (88 to 96% at the highestNH2 H OH H CH2NH2 concentration tested) are seen with the neomy-NH2 H OH CH2NH2 H cin-type antibiotics (neomycins B and C, ribo-OH H H jCH2NH2 H stamycin, butirosin, and paromomycin [Fig. 1]).

lH CH2NH2J This inhibition has been noted previously (5);of the neomycins, butirosin, ri- neomycin B is the only deoxystreptamine anti-omomycin. Neamine (or paroma- biotic which gives close to 100% inhibition atgs 1 + 2, neobiosaminide of rings relatively low concentrations. Figures 5 and 6ycin of i + 2 + 3. The hybrimy- show representative inhibition curves whichtamine ring (hybrimycin A and indicate that neomycin B inhibits polypeptidemne ring (B1 and B2) instead synthesis to a greater extent than do the kana-ne. Butirosin (1 + 2 + 3) has a mycin or gentamicin antibiotics. It can also beutyryl substituent on N-i of 2- seen that the various acetylated derivatives oftrrows indicate where these an-..r i w t a gentamicin (Fig. 6), kanamycin (Fig. 5), andetylated by kanamycin acetyl6'-N-acetyl antibiotics (a), and neomycin (Table 1) are less potent than theirneomycin-kanamycin phospho- nonacetylated parent compounds. Several of

by lividomycin phosphotransfer- the antibiotics show biphasic curves; the rea-sons for this are not known. One possibility

VOL. 4, 1973

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

404 BENVENISTE AND DAVIES

-- a

d

RI R2 R3 R4 R5 R6

GENTAMICINS A H OH OH OH H OH

CIO H NH2 H H OH CH3

C2 CH3 NH2 H H OH CH3Cl CH3 NHCH3 H H OH CH3

FIG. 3. Structure of the gentamicins. Ring I ispurpurosamine, II is 2-deoxystreptamine, and III isgentosamine (gentamicin A) or garosamine (gen-tamicin C's). Sisomycin is 4,5-dehydrogentamicinCia (the purpurosamine ring is reduced). In dihy-drosisomicin, the carbon #6 group of that ring isinverted (L-sugar). Arrows indicate where this groupof antibiotics can be N-acetylated by kanamycinacetyl transferase to yield 6'-N-acetyl antibiotics (a),by gentamicin acetyl transferase II to yield 2'-N-acetyl antibiotic (d), by gentamicin acetyl transferaseI to yield 3-N-acetyl antibiotic (c), o-phosphorylatedby neomycin-kanamycin phosphotransferase (gen-tamicin A only) (b); and o-adenylylated by gen-tamicin adenylyl transferase (e).

CH2NHp o

HO0

H2N NH?

HO NH2

0

0 l CH20HHO H

H2N

FIG. 4. Structure of tobramycin (nebramycin fac-tor 6). The arrow indicates where this antibiotic canbe enzymatically N-acetylated by kanamycin acetyltransferase to yield 6'-N-acetyl tobramycin (a), bygentamicin acetyl transferase I to yield 3-N-acetyltobramycin (b), and o-adenylylated by gentamicinadenylyl transferase (c).

which has been suggested (4, 5) is that theantibiotics can interact with ribosomes at morethan one site depending on the drug concentra-tion.

Effects of the aminoglycosides and theirderivatives on poly U-directed misreading.The activity of the aminoglycoside antibiotics

on polypeptide-synthesizing systems from bac-teria can be demonstrated in two ways. On onehand the drugs inhibit "natural" messengerRNA-directed synthesis (see above), and on theother the drugs induce translational errors (mis-reading) in synthetic polymer-directed synthe-sis. The latter system can be assayed by meas-uring the extent of incorporation of the aminoacids serine, isoleucine, and tyrosine in thepresence of poly U and antibiotic. The signifi-cance of this effect on the mode of action of thedrugs is not clear, but it does provide a verysimple and sensitive assay for biological activity(5). Drug-induced misreading often occurs atconcentrations of antibiotic which are too low toinhibit protein synthesis, and it is an accurateprobe for structure-activity relationships.

All of the aminoglycosides and their deriva-

TABLE 1. Inhibition of in vitro R17 phageRNA-directed polypeptide synthesis by variousaminoglycoside antibiotics and their degradation

products

Drug concn (Mgg/mi)aAntibiotic

0.1 1.0 10 50

Gentamicin Cia 26 45 55 782'-N-acetyl gentamicin Cia 0 7 40 556'-N-acetyl gentamicin Cia 0 0 22 413-N-acetyl gentamicin Cia 0 0 9 27Gentamicin C.i-adenylylate 0 0 0 10Gentamicin C1 28 37 37 60Gentamicin A 8 8 30 50Sisomicin 23 41 56 75Dihydrosisomicin 0 0 0 2Tobramycin 22 45 55 80Kanamycin A 11 21 50 55

B 22 48 58 82C 0 9 30 50

6'-N-acetyl kanamycin A 0 0 0 46'-N-acetyl kanamycin B 0 5 14 354, 6-di-0-(a-D-glucopyranosyl)- 0 0 0 0

deoxystreptamine4,6-di-O-(6-amino-6-deoxy-a-D- 0 0 10 20

glucopyranosyl)-deoxystreptamine

Neomycin B 28 55 80 96C 10 35 72 88

6'-N-acetyl neomycin B 0 6 52 76Ribostamycin 5 45 65 92Butirosin 30 60 72 88Methyl neobiosaminide B 0 0 0 10Paromomycin 18 40 65 80Gentamine A° 0 9 9 32Neamine 2 6 37 75Gentamine Cia 0 0 32 72Tobramine 0 3 30 65

aExpressed as percent inhibitionsynthesis at four concentrations.

° Same as paromamine (Fig. 1).

of polypeptide

ALNTiMICROB. AG. CHEMOTHER.

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

RELATIONSHIPS AMONG AMINOGLYCOSIDE ANTIBIOTICS 405

la are the 1,2 substituted compounds which in-10 clude the neomycins, paromomycins, ribo-

E~.. \\\ <stamycin, neamine, butirosins, and lividomy-o1 \ \ s \ tcins. The 1,3 substituted group includes the0

820 kanamycins, gentamicins, and tobramycin. It is

zuj:D ~~~~~40'~t1X CD\L~~~~~~~~Ls _5\G\ee6t-\\XaX1 m k-

o°U4t \ X\ 607 001°0- o 0

0 Q:2x80 /O_ \ \ _ ~~~~ ~ ~~~~<x 3 8

100 a

° 0.1 0.2 10 50 150 O° 2DRUG CONCENTRATION (,g/ml) Z 2 /

FIG. 5. Inhibition of in vitro R17 phage RNA- L°directed polypeptide synthesis by acetylated and a_ -

unacetylated antibiotics. The incubation conditions/are described in Materials and Methods. Symbols: A,6'-N-acetyl kanamycin A; 0, 6'-N-acetyl kanamycinB; 0, kanamycin C; E, kanamycin A; A, kanamycin 0 0l 10 100 1000B; 0, neomycin B. DRUG CONCENTRATION (yg/ml)

FIG. 7. Misreading of poly U induced by variouskanamycins. Incubation conditions are described in

-E '-41 =o Materials and Methods. Symbols: 0, kanamycin A;0<>b \ A, kanamycin B; U, kanamycin C; 0, 6'-N-acetyl0ix20 kanamycin A; 0, 6'-N-acetyl kanamycin B; A, 4,6-di-0~~~~~~~~~~~~26 0-(6-amino-6-deoxy-a-D-glucopyranosyl)-deoxystrept-cr ~~~~~~~0amine.

°2 4

X

+ 0O04 zXz ~~~~~~~~~60

cr 80

0 005 01 02 05 10 O L- /DRUG CONCENTRATION (,ug/mI) 2/

FIG. 6. Inhibition of in vitro R17 phage RNA-3/\directed polypeptide synthesis by gentamicin C1a and I-its enzymatically modified derivatives. Symbols: 0, a. xgentamicin C,adenylylate; A, 3-N-acetylgentamicin °

0

C10; A, 6'-N-acetylgentamicin C10; 0, 2'-N-acetyl- 0 zgentamicin C10; 0, gentamicin Cl8. 0 _

tives were tested in this system. Some repre- a. a._sentative results are shown in Fig. 7 through 9.Again, it is clear that there is a wide variation in ,Idrug activity. 0 01 1015 100 200 1000

DISCUSSION DRUG CONCENTRATION (g/mi)FIG. 8. Misreading of poly U induced by gen-Rinehart (10) has noted that the deoxystrep- tamicin C10 and its enzymatically modified deriva-

tamine-containing aminoglycoside antibiotics tives. Symbols: 0, gentamicin C10; 0, 2'-N-acetylcan be divided into two classes based on the gentamicin C10; A, 6'-N-acetyl gentamicin Ca; *,positions of substitutions of sugar moieties on 3-N-acetyl gentamicin C10; A, gentamicin C10-the deoxystreptamine ring. These two classes adenylylate.

VOL. 4, 1973

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

406 BENVENISTE AND DAVIES

L1JD

<-F-

Q- z

05

0.1 10 100 1000DRUG CONCENTRATION (,ug/ml)

FIG. 9. Misreading of poly U induced by the neo-

mycin-type antibiotics. Symbols: 0, neomycin B; *,neamine; A, 6'-N-acetylneomycin B; A, paromomy-

cin; 0, ribostamycin.

clear from the data presented in Table 1 thatthe 1,2 substituted deoxystreptamines are themore potent antibiotics. Since several of theaminoglycosides have close structural similari-ties, certain conclusions can be drawn about theeffects of various amino and hydroxyl moietieson the biological activity of these compounds.

Effects of amino groups in the hexoselinked to the 4-position of deoxystreptamineon the biological activity of theaminoglycosides. Kanamycins A, B, and Cdiffer only in the number and location of theamino groups in the hexoses in the 4-position(Fig. 2). Kanamycin A has a 6'-amino-6'-deoxy-D-glucose ring, kanamycin C has a 2'-amino-2'-deoxy-D-glucose, and kanamycin Bhas a 2', 6'-diamino-2'6'-dideoxy-D-glucose ring.The extent of inhibition of R17 RNA-directedpolypeptide synthesis by these antibiotics isshown in Fig. 5. As can be seen, kanamycin B isa more potent antibiotic than either kanamy-cins A or C. The presence of a diamino hexose,therefore, results in a compound that is a betterinhibitor of protein synthesis than one contain-ing only one amino group. This same conclusionis obtained by comparing neomycin and paro-

momycin. The data in Table 1 indicate thatneomycin is a more potent antibiotic thanparomomycin. The only difference betweenthese compounds is that the former contains a

2', 6'-diamino-2'6'-dideoxy-D-glucose ring, andthe latter a 2'-amino-2-deoxy-D-glucose moiety,at the 4-position of deoxystreptamine (Fig. 1).Kanamycin A is a better inhibitor of protein

synthesis than kanamycin C (Fig. 5). Therefore,when only one amino group is present, anantibiotic that contains a 6-amino substituentis more active than one containing a 2-aminosubstituent.One of the ways that the aminoglycosides can

be modified is by acetylation of various aminogroups by enzymes that are present in bacterialstrains resistant to these antibiotics (1). It hasbeen observed that the acetylation of an amino-glycoside does not always eliminate biologicalactivity (1). The only aminoglycoside that isinactivated by enzymatic N-acetylation of theaminohexose ring in the 6-position is kanamy-cin A; 6'-N-acetylkanamycin A is not an antibi-otic (Fig. 5). However, 6'-N-acetylkanamycin Bstill retains considerable biological activity.Since 6'-N-acetylkanamycin B very closely re-sembles kanamycin C in the extent of inhibitionof protein synthesis, it seems that acetylationsimply neutralizes the contribution of aminogroups. As would be expected, 6'-N-acetyl-neomycin B is very similar to paromomycin inbiological activity. It can thus be concludedthat at least one amino group is required in theaminohexose at the 4-position for the compoundto retain biological activity.This same pattern of relative activity is

observed with the gentamicins and their acety-lated derivatives (Fig. 6). Gentamicin C18,which contains two amino groups in the pur-purosamine ring (Fig. 3), is the most potentantibiotic, followed by 2'-N-acetylgentamicinC18 (which has a free 6'-amino group) and by6'-N-acetylgentamicin C1l (which has a free2'-amino group). To summarize: antibiotic ac-tivity can be related to the number and locationof amino groups in the hexose moiety glycosidi-cally linked to the 4-position of deoxystrepta-mine as follows (in decreasing order of potency):

2', 6'-diamino > 6'-amino > 2'-amino > noamino

The patterns of misreading activity are alsostrongly dependent on the number of aminogroups. Those antibiotics with a 2,6-diaminosugar (tobramycin, kanamycin B, gentamicinsCla, Cl, or C2, neomycins B or C) show a uniquepattern of misreading. There is a progressiveincrease in the level of misreading over a broadconcentration range (0.1 to 15 Ag/ml, about 10-7to 10' M), followed by a marked inhibition ofmisreading at higher concentrations (Fig. 7-9).The compounds with one amino group (ka-namycins A and C, paromomycin, 6'-N-acetyl-neomycin B, 6'-N-acetylkanamycin B, 6'-N-acetylgentamicin C1l, and 2' -N-acetylgen-tamicin C18) do not show this inhibition of

ANTIMICROB. AG. CHEMOTHER.

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

RELATIONSHIPS AMONG AMINOGLYCOSIDE ANTIBIOTICS 407

misreading between 15 and 1,000 lsg/ml. Com-pounds with no amino groups (6'-N-acetyl-kanamycin A) do not cause any misreading atconcentrations up to 200 gg/ml (Fig. 7).

Streptomycin shows only a very gradual in-crease in misreading of poly U as the antibioticconcentration is increased from 10-6 to 10-i M(5). Bacterial mutants with altered ribosomesthat are resistant to either streptomycin orparomomycin have been obtained, but no ribo-somal resistant mutants have been isolated forgentamicins C16, Cl, C2, kanamycin B, orneomycin. Perhaps these latter antibiotics havemore than one binding site on the ribosome; thesecond site may be the one responsible for theinhibition of misreading at high concentrations(5). Only those antibiotics with a 2,6-aminohexose show biphasic inhibition and mis-reading curves. Two mutations might thereforebe needed to obtain a strain containing ribo-somes resistant to these antibiotics. It may bepossible to select for a strain that is resistant toneomycin by starting with one that is resistantto paromomycin. Similarly, kanamycin B- orgentamicin Cia-resistant strains might be ob-tained by selecting first for ribosomal resistanceto kanamycins A or C or to gentamicin A (or2'-N-acetylgentamicin C16), respectively. It isnot possible to state what role the amino groupsplay in the interaction of the antibiotic with theribosome; they could be involved in hydrogenbonding or ionic interactions.

Effect of hydroxylation of the hexose gly-cosidically linked to the 4-position of 2-deoxystreptamine. Kanamycin B and tobra-mycin are identical in structure except thatthe latter lacks a 3-hydroxyl moiety (compareFig. 2 and 4). Both are equally effective ininhibiting protein synthesis (Table 1). There-fore, the presence or absence of a hydroxylgroup at the 3'-position of this hexose does nothave a great effect on the activity of theseantibiotics. Gentamicin C16 contains pur-purosamine (a 3', 4'-dideoxy sugar) (Fig. 3) andis as good an antibiotic as kanamycin B. Thesethree antibiotics differ only slightly in misread-ing activity (Fig. 7 and 8), they all producegreater misreading with increasing concentra-tions up to approximately 10 gg/ml, and afurther increase in concentration is thenstrongly inhibitory. It is apparent that thedegree of hydroxylation of the sugar in the4-position of deoxystreptamine has no effect onthe activity of these antibiotics.That the presence or absence of hydroxyl

groups does not have a marked effect on thebiological activity of these aminoglycosides canalso be seen by comparing the activities of

neamine, tobramine (3-deoxy), and gentamineCla (3',4'-dideoxy) (see Table 1). These threecompounds, which consist of only one diamino-hexose linked to deoxystreptamine, have essen-tially identical biological activity in vitro.The aminoglycosides lacking the 3-hydroxyl

group are not substrates for neomycin-kanamycin phosphotransferase, an enzymefound in bacteria resistant to neomycin andkanamycin, since this enzyme phosphorylatesthat hydroxyl group in the aminoglycosides.Since the 3'-deoxy compounds are potent an-tibiotics, they are currently being used to treatinfections involving bacteria that are resistantto kanamycin and neomycin. Phosphorylationof the 3-hydroxyl leads to complete inactiva-tion since the phosphorylated antibiotic can nolonger bind to the ribosome (13). The studiesdescribed here suggest that inactivation occursbecause the hydroxyl group is replaced by abulky charged group and not as a result ofblocking a functionally important hydroxylgroup.

Effect of other substituents on the deoxy-streptamine ring on the biological activity ofthe aminoglycosides. Kanamycin B, to-bramycin, and gentamicin C16 also differ in thenature of the substituent at the 6-position ofdeoxystreptamine. The former two compoundshave a kanosamine substituent (Fig. 2 and 4),and the latter has garosamine. Since there isvery little difference in the in vitro biologicalactivity of tobramycin, kanamycin B, and gen-tamicin C16, and between gentamicin A andkanamycin C, we conclude that althoughkanosamine and garosamine have substantialroles in determining antibiotic potency (i.e.,compare tobramycin with tobramine and genta-micin C16 with gentamine C16), they are verysimilar in this respect.These substituents are important, however,

in the recognition of these compounds by thevarious aminoglycoside inactivating enzymes.For example, the enzyme which acetylates the3-amino group in the deoxystreptamine ring ofgentamicin C16 (Fig. 3) does not recognize thissame group in kanamycin B or tobramycin.This specificity is presumably determined bydifferences in the substituents on the deoxy-streptamine ring (3).The neomycins are structurally different from

the kanamycins and gentamicins in that sub-stituents are on adjacent hydroxyls (4, 5) ofdeoxystreptamine (Fig. 1). Various degradationproducts of the neomycins allow an assessmentof the requirements for biological activity in thisseries. Table 1 shows that the order of decreas-ing potency in inhibiting protein synthesis is

VOL. 4, 1973

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

408 BENVENISTE AND DAVIES

neomycin B, neomycin C, butirosin, ribosta-mycin, neamine, and methyl neobiosaminideB. Therefore, the presence of all four rings yieldsmaximal activity, followed by three rings, andthen two. Neamine is a good antibiotic, andmethyl neobiosaminide B is not; the lattercompound inhibits phage RNA-directed proteinsynthesis by only 10% at 50 .ug/ml. The lack ofactivity of methyl neobiosaminide B is possiblydue to the fact that the C6 substituent isinverted (dihydrosisomicin is inactive) ratherthan to the replacement of deoxystreptamine byribose. In this regard it would be interesting toexamine the antibacterial activity of methylneobiosaminide C.Role of 2-deoxystreptamine in determining

the biological activity of the aminoglyco-sides. It has been proposed that the deoxy-streptamine ring is the primary determinantof the biological activity of the aminoglyco-side antibiotics (7, 11). However, it has beenshown that deoxystreptamine can be replacedby streptamine (4), and it is apparent from ourresults that the sugar substituents on thedeoxystreptamine ring play the most impor-tant role. We have confirmed this suppositionby studies with synthetic compounds providedby R. Lemieux.These compounds contain a 2-deoxystrepta-

mine ring substituted with various sugars andamino sugars. 4, 6-Di-0-(a-D-glucopyranosyl)-deoxystreptamine, which consists of two D-glucose moieties glycosidically linked to 2-deoxy-streptamine, is completely inactive in inhibit-ing protein synthesis (Table 1). 4, 6-Di-O-(6-amino - 6 - deoxy - a - D - ,glucopyranosyl) - de-oxystreptamine is only weakly active. Therefore,the presence of a deoxystreptamine moiety isnot enough to guarantee antibacterial activity;the deoxystreptamine has to be linked to theappropriate amino sugars. Thus the locationof the amino group in the 3-amino-D-glucosemoiety (kanosamine) of kanamycin A is impor-tant for biological activity. When that sugaris changed to 6-amino-6-deoxy-D-glucose, thecompound loses almost all biological activity.2-Deoxystreptamine has two amino groups.

When one of them is enzymatically acetylatedby an enzyme found in a gentamicin-resistantstrain of Pseudomonas aeruginosa (3) to yield3-N-acetylgentamicin Cia, the compound is nolonger an antibiotic (Fig. 6 and 8). In butirosin,the 1-amino group of deoxystreptamine is sub-stituted with 4-amino-2-hydroxybutyric acid,and this does not substantially reduce biologicalactivity in our in vitro assays. This suggests thatthe 1-amino group of 2-deoxystreptamine maybe unimportant for biological activity; however,

studies of other compounds with substituents atthis position will be necessary to establish this.Similar substitution of kanamycin results in anantibiotic (BBK-8) which differs little in activ-ity from kanamycin. The importance of thissubstitution is that it increases anti-Pseudomonas activity and makes the com-pound inert to various forms of inactivation (9).General conclusions. We have shown that

the number and location of amino groups in thesugars attached to deoxystreptamine pro-foundly affect the biological activity of theaminoglycoside antibiotics. The presence of atleast one amino group in these sugars is neededfor biological activity. The addition of extrasubstituents to the deoxystreptamine-aminosugar combination (even a ribose moietyas in ribostamycin) can further increase thebiological activity of these compounds.As new aminoglycosides and additional enzy-

matically modified derivatives become availa-ble, it may be possible to derive additionalconclusions as to the biological activity of theseantibiotics. The structural requirements forsubstrate activity for the nine aminoglycosideinactivating enzymes that have so far beendiscovered have been reviewed (R. Benvenisteand J. Davies, Annu. Rev. Biochem., in press).Fortunately, the requirements for activity as anenzyme substrate are somewhat different fromthose for antibiotic activity, or the design of newaminoglycosides that could inhibit resistantbacterial strains would not be possible.The ribosome is a very complicated mac-

romolecule, and genetic studies with ribosomesfrom antibiotic-resistant strains of bacteriahave proved useful in characterizing certain ofthe protein components (6). Among the amino-glycoside-aminocyclitol antibiotics there aredistinct binding sites for the streptomycin-typecompounds, spectinomycin, and kasugamycin,which have been established by studies ofresistant mutants. In the case of the antibioticgroups typified by kanamycin and neomycin, itis not known if these two groups share the samebinding site or have separate sites. The bindingof neomycin to ribosomes affects the binding ofstreptomycin (13); kanamycin has no effect.The effects of both neomycin and kanamycin Bon protein synthesis appear to be the result ofmultiple binding. The binding of aminoglyco-side-aminocyclitols to the ribosome would seemto be the role predominantly of amino groups inthese antibiotics.We have completely ignored one important

aspect of the aminoglycoside antibiotics-theirtoxicity. There is a need for a simple and directassay which would relate structural require-

ANTIMICROB. AG. CHEMOTHER.

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from

RELATIONSHIPS AMONG AMINOGLYCOSIDE ANTIBIOTICS 409

ments for antibacterial activity to those respon-sible for the various toxic effects of aminoglyco-sides (e.g., nephro- and ototoxicity). If thiscould be done, the rational design of new anduseful aminoglycoside antibiotics would begreatly facilitated.

LITERATURE CITED

1. Benveniste, R., and J. Davies. 1971. Enzymatic acetyla-tion of aminoglycoside antibiotics by Escherichia colicarrying an R-factor. Biochemistry 10:1787-1796.

2. Benveniste, R., and J. Davies. 1971. R-factor mediatedgentamicin resistance: a new enzyme which modifiedaminoglycoside antibiotics. FEBS Lett. 14:293-296.

3. Brzezinska, M., R. Benveniste, J. Davies, P. J. L.Daniels, and J. Weinstein. 1972. Gentamicin resistancein strains of Pseudomonas aeruginosa mediated byenzymatic N-acetylation of the deoxystreptaminemoiety. Biochemistry 11:761-766.

4. Davies, J. 1970. Structure-activity relationship amongthe aminoglycoside antibiotics: comparison of theneomycins and hybrimycins. Biochim. Biophys. Acta222:674-676.

5. Davies, J., and B. Davis. 1968. Misreading of ribonucleic

acid code words induced by aminoglycoside antibiotics.J. Biol. Chem. 243:3312-3316.

6. Davies, J., and M. Nomura. 1972. The genetics ofbacterial ribosomes. Ann. Rev. Genet. 6:203-234.

7. Masukawa, H., and N. Tanaka. 1968. Miscoding activityof amino sugars. J. Antibiot. (Tokyo) 21:70-72.

8. Nirenberg, M. W. 1964. Cell-free protein synthesis di-rected by messenger RNA, p. 17-23. In S. P. Kolowickand N. 0. Kaplan (ed.), Methods in enzymology, vol. 6.Academic Press Inc., New York.

9. Price, K. E., D. R. Chisholm, M. Misiek, F. Leitner, andY. H. Tsai. 1972. Microbiological evaluation of BBK-8,a new semisynthetic aminoglycoside. J. Antibiot. (To-kyo) 25:709-731.

10. Rinehart, K. L., Jr. 1969. Comparative chemistry of theaminoglycoside and aminocyclitol antibiotics. J. In-fect. Dis. 119:345-350.

11. Tanaka, N., H. Masukawa, and H. Umezawa. 1967.Structural basis of kanamycin for miscoding activity.Biochem. Biophys. Res. Commun. 26:544-549.

12. Weisblum, B., and J. Davies. 1968. Antibiotic inhibitorsof the bacterial ribosome. Bacteriol. Rev. 32:493-528.

13. Yamada, T., K. Kvitek, and J. Davies. 1970. The bindingof streptomycin and R-factor inactivated streptomycinto ribosomes, p. 562-566. Progr. Antimicrob. Antican-cer Chemother. International Congress of Chemother-apy (Tokyo) B2.

VOL. 4, 1973

on June 26, 2018 by guesthttp://aac.asm

.org/D

ownloaded from