JOURNAL OF BACTERIOLOGY, OCt., 1965Copyright © 1965 American Society for Microbiology

Vol. 90, No. 4Printed in U.S.A.

Numerical Taxonomy of Some Bacteria Isolated fromAntarctic and Tropical Seawaters'ROBERT M. PFISTER AND PAUL R. BURKHOLDER

Lamont Geological Observatory, Columbia University, Palisades, New York

Received for publication 23 April 1965

ABSTRACTPFISTER, ROBERT M. (Lamont Geological Observatory, Palisades, N.Y.), AND PAUL

R. BURKHOLDER. Numerical taxonomy of some bacteria isolated from Antarctic andtropical seawaters. J. Bacteriol. 90:863-872. 1965.-Microorganisms from Antarcticseas and from tropical waters near Puerto Rico were examined with a series of morpho-logical, physiological, and biochemical tests. The results of these analyses were codedon punch cards, and similarity matrices were computed with a program for an IBM1620 computer. When the matrix was reordered by use of the single-linkage technique,and the results were plotted with four symbols for different per cent similarity ranges,nine groups of microorganisms were revealed. The data suggest that organisms occur-ring in different areas of the open ocean may be profitably studied with standardizedcomputer techniques.

Increased interest in the use of the Adansonianprinciple of equal weight for characteristics usedto describe organisms (Sokal and Sneath, 1963;Rogers and Fleming, 1964; Quadling and Colwell,1964; Colwell and Mandel, 1965; Lockhart,1964) has led us to examine bacteria obtainedfrom the open ocean with similar techniques.Numerous physiological, morphological, and

biochemical characteristics were determined forcultures isolated from Antarctic seas and fromtropical waters in the vicinity of Puerto Rico.We have used this information and computertechniques to describe a system which appearsto be suitable for the study of small or infinitelylarge marine populations. There has been mucheffort devoted recently to the numerical approachin microbial taxonomy, especially among micro-biologists who recognize the need for improvedtechniques. The merits and faults of the numericalapproach to problems of bacterial classificationhave been well covered in recent papers (Sokaland Sneath, 1963; Quadling and Colwell, 1964).

MATERIALS AND METHODSOrganisms. The 151 microorganisms used in

this study were isolated from the Antarctic Seaand from waters in the vicinity of Puerto Rico.Cultures were purified by restreaking, and main-tained on a medium containing the following (perliter): Trypticase (BBL) or N-Z-Case, 2.0 g; Soy-tone (Difco), 2.0 g; yeast extract, 1.0 g; vitaminB12, 1.0 lg; aged seawater; agar (if desired), 16 g.

1 Contribution No. 840 from Lamont GeologicalObservatory.

The pH was adjusted to 7.0 prior to autoclaving.Isolation of Antarctic microorganisms was carriedout in two ways. (i) If the desired organisms werepsychrophilic and still able to withstand andgrow at warmer temperatures (e.g., 20 to 40 C),then a standard pour-plate technique of meltedagar at 47 C and incubation at 5 C was used. (ii)When obligate psychrophiles (microorganisms notable to grow above 10 to 15 C) were desired inaddition to the other types, their isolation wascarried out by a procedure which maintained theseawater sample below 10 C at all times. Obligatepsychrophiles were isolated by quantitatively de-positing the sample on the surface of a precooledseawater agar plate, spreading with a cold glassrod, and incubating at 5 C for periods up to 2weeks. Examination of the agar surface for thepresence of clones was made in the cold with astereomicroscope. Bacteria were isolated fromtropical waters by the common pour-plate tech-nique.Media and tests. To determine the carbohydrate

fermentation of these marine bacteria, a 1.0%carbohydrate medium (see Fig. 2 for specific car-bohydrates) was prepared according to Skerman(1959), with the seawater medium previously de-scribed, but with K2HPO4 added in the amount of0.05 g per liter and use of agar at the rate of 5.0g per liter. The solutions of carbohydrates weresterilized by Seitz filtration and added asepticallyto 1-liter batches of melted base medium. Allmedia were preincubated for a minimum of 72 hrto check for sterility. Each organism was inocu-lated in each test medium in duplicate with a buttstab. Samples were prepared for anaerobic fermen-tation by covering with sterile mineral oil. Withthe tropical bacteria, the test medium was covered

863

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

PFISTER AND BURKHOLDER

immediately after inoculation, because a few ofthe warm-water organisms utilized the carbohy-drates aerobically so rapidly that any anaerobictest was spoiled before leaving the inoculationarea. All cultures used for inoculum were 48-hrtransfers, actively growing on seawater agar. Incu-bation was at 25 C, except where otherwise speci-fied.

Indole production was determined (Skerman,1959) by use of 1.0% Tryptone (Difco) in the sea-

water medium.Starch hydrolysis was carried out by preparing

a 0.2% solution of soluble starch in the basal agar

medium. Cultures were streaked on the surfaceof the agar and incubated for 1 week, at whichtime a dilute iodine solution was poured over thesurface, and the cleared zones were recorded.

Gelatin hydrolysis was carried out by use ofseawater agar plates containing 0.4% gelatin.After incubation, the plates were flooded with a

15% acid HgC12 solution (Skerman, 1959) to makevisible the zones of hydrolysis.Methyl red and Voges-Proskauer tests were

tried according to Skerman (1959), with the sub-stitution of seawater as the diluent, but the highpercentage of K2HP04 formed an insoluble pre-cipitate and the phosphate had to be reduced to0.05 g per liter.

Citrate utilization as a source of carbon forgrowth was determined with a medium containingsynthetic seawater (Lyman and Fleming, 1940)containing the following (per liter): KH2P04, 50mg; Fe(NH4)2(S04)-6H20, 7 mg; N-Z-Case, 0.4 g;and sodium citrate, 3.0 g. To ensure that growthin the medium was due to citrate utilization andnot the substitution of amino acids as a carbonsource, the level of 0.4 g per liter of N-Z-Casewas determined experimentally to be suitable fordemonstrating increased growth with added cit-rate.Amino acids were tested as a carbon source by

inoculating a medium of synthetic seawater con-taining, per liter, KH2PO4, 0.025 g, and N-Z-Case,4.0 g.

Hydrolysis of triolein and tributyrin was de-tected by dispersing the fats in a melted 0.5%neutral agar solution (10% fat) and autoclaving.In preparation for use, the fat-agar emulsion was

melted and added aseptically to the melted basalseawater agar (1 ml of fat emulsion per 20 ml ofbasal medium), and then shaken and dispersed toplates. The organisms were streaked on the sur-

face, incubated up to 2 weeks and examined fora clearing zone (Tendler and Burkholder, 1960),and after staining with Sudan B soluition to helpintensify any visible zones.

Nitrate reduction was detected by the methodoutlined in the Manual of Microbiological Methods(Society of American Bacteriologists, 1957), byuse of seawater broth with 0.1% KNO.The Kovacs oxidase test was performed accord-

ing to the method of Kovacs (1956). The cyto-chrome oxidase test of Gaby and Free (1958) was

used as described, except for the addition of thereagents to an agar slant in preference to brothculture. Catalase activity was qualitatively de-tected by adding a drop of H202 to an agar slantculture.Litmus milk was prepared at 0.33% of the rec-

ommended concentration in seawater with theaddition of 1 ,ug per liter of vitamin B12. Steriliza-tion of this medium was done by a 3-day succes-sive tyndallization (5 min of free flowing steameach time), with incubation at 25 C between steamtreatments. Prepared medium was tested for ste-rility by preincubation for at least 72 hr prior touse.Ammonium nitrate was tested as a nitrogen

source by use of a medium of seawater with addi-tions of the following (per liter): KH2P04, 0.05 g;NH4NO3, 5.0 g; glucose, 4.0 g; fructose, 4.0 g; andsodium citrate, 4.0 g.H2S production was determined by a technique

modified from Skerman (1959), with the followingmedium (per liter): peptone, 10 g; cystine, 0.1 g;Na2SO4, 0.5 g; seawater; and vitamin B12, 1 ,ug.The pH was adjusted to 7.0. Sterile lead acetatepaper strips were placed at the top of the tubebetween the plug and the glass. Darkening of thepaper strips at any time during incubation wasconsidered a positive indication for the productionof H2S.

Sensitivity of the organisms to six antibiotics-bacitracin, novobiocin, oleandomycin, penicillin,tetracycline, and viomycin-was determined byuse of standard Difco sensitivity discs impreg-nated with "medium" levels of concentration foreach antibiotic. Cultures used to form the seedlayer for testing the sensitivity discs were grownin seawater broth on a rotary shaker at 25 C over-night, and then layered on the surface of pouredagar plates. The sizes of zones of inhibition werenot taken into account, but inhibition was re-corded as either present or absent. At the sametime, a solution of pteridine 0/129 (Shewan, Hodg-kiss, and Liston, 1954) was tested in moistenedpaper discs placed on the seeded plates, to aid inthe differentiation of Vibrio and Pseudomonas.Microscopic and macroscopic observations of

morphological and cultural characteristics weremade by use of seawater agar slants, isolatedclones on streak plates, and, where necessary, onseawater broth. Subjective decision was madewhere colony color was noted or diffusible pig-ments detected. Subjective decision was also usedin determining the amount of growth in any brothculture, based upon the development of relativeturbidity in a given time, which was arbitrarilyscaled 0, 3, or 5, with 0 equal to slight growth, 3equal to fair growth, and 5 good growth.Gram stains were tried according to several

methods, e.g., Hucker modification and Burke,Kopeloff-Beerman modification (Society of Ameri-can Bacteriologists, 1957), with the result that nopreference for a method could be made. Measure-ments of the cell dimensions were taken from

864 J. BACTERIOL.

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

NUMERICAL TAXONOMIY

Gram-stained preparations, whereas configurationof the cells (morphology) was determined from awet mount by use of phase-contrast optics.

Flagellation of bacteria was determined by themethod of Leifson (1960).

Numerical analyses. To comptute a per centsimilarity (%cS; Sokal and Sneath, 1963) for eachisolate (OTU) compared with all other isolates,two types of calculations were made. One methodwas that of Sneath (1957), in which negativematches were excluded, and the other was thetechniqute of Sokal and iMichener (1958), in whichnegative matches were included in the numerator.These methods are computed as follows. Accordingto Sneath (1957),

%S .A'sp X 100N'sp + Nd

OF MIARINE BACTERIA 865

and according to Sokal and Michener (1958),

-S Asp + NsnA'sp + N5sn + Nd

where %S similarity coefficient, AXsp = niumberof similar positive matches, Nsn = number ofsimilar negative matches, anid ANd = ntmber ofdissimilar matches. The raw data were punichedonto standard IBAI cards in a maniner describedby Colwell and Liston (1961).The program was written in Fortran II in con-

junction with an IBAI 1620 computer equippedwith an accessory disc memory storage uinit.Each per cent similarity was computed as three

digits, or to a 0.1% accuracy. Grouping of thevariouis OTU values was done with the method ofsingle-linkage cltustering (Sokal and Sneath, 1963),

rodscurved rodscoccobacillarycocci

soorespleonKrphicsirglesmic rococcdiploformsarcinaechains(4cells.)length -short (0.2- 0.6u)

i; -med(O.61,2)n - 0ong (>1.2u)

wicth- slim(< 0.5)-med.(O. 5 - l.Ou)-fat( > 1.0)

motilenon-motileunipo;ar flagellationbipolar flagellationperitrichous flagellationgram negative

l. positivevariable

acid fastpunctiform ( <.2mm)rred cooryl< Irrrn)large i> l.mm)entire edgeurdulate lobate erose (ule)curledf ilamentousopaqueflatra:sedconvex or pulvinateumboratefi! formechi nulatebeadedeff usewhitetransluc entpogmented colony- tan

- yellow- pink

Ii -orarge- brown-purple

1 - blac k

p:gment (dffusiblergrowvth 0

35

ePven turbidityflocculent turbidityr napellic.ef uorescent UVgrowth at 5°C

201Cu 35°C

sea H20 requiredgrowth in sea H20 & dist. H20

PERCEJA-G-- OF -OTAL ORGAN!SMSN) w.U Un

0 0 0 o 0

L

0) --u 0) (0)o o c O0 0 0 C

FIG. 1. Computed per cent occurrence of morphological characteristics, pigment production, and growthphenomena based upon the combined marine population of 150 microorganisms.

L

Nr0L. 90, 1965

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

PFISTER AND BURKHOLDER

which first clusters strains most closely relatedwith the highest similarity coefficients. Then, bylowering the level of inclusion by predeterminedamounts of equal magnitude, it permits the othermembers to join. In this manner, one can find theexact value at which all members of the matrixbecome numerically united.

RESULTSFigures 1, 2, 3, and 4 represent a compilation

of various taxonomic measures made upon apopulation of 151 marine microbes taken fromeither the cold Antarctic sea, or from warmtropical waters in the vicinity of Puerto Rico.

In Fig. 1, a bar graph shows the various cate-gories of morphology, pigment production, andgrowth observed during this study. Each bar onthe graph is the computed percentage of oc-currence of the character in question based uponthe total number of organisms examined for thatcharacter. The data in Fig. 1 were determined

for a total of 151 microbial isolates examined foreach character; 114 isolates were derived fromicold Antarctic waters and 37 from tropical watersnear Ptuerto Rico.The bar grap)h in Fig. 2 shows the results of

tests eml-ployed to help describe this marine popu-lation. The total numaber of organisms studied inthis series of tests was 151, except that cellobioseutilization was calculated from a total of 115organisms, and tributyrin, triolein, and pteridine0/129 reactions were based upon tests made witha total of 91 cultuies.A l)artial list of characters for the Antarctic

segment of the population is shown in Fig. 3.The abilitv of the 37 Puerto Rican isolates to usecarbohydrate can be seen in Fig. 4.The abilitv of the combined population (151

cultures) to use aerobically and anaerobicallyseven selected carbohy-drates in cumulative com-

bination is shown in Table 1.

0

oleandomycin sensitivity _novobiocin _penici., n _tetracyc:ne _bacitracir. 11

s.omrc n L_starch c!gestedelatin pqci eC

rtm-js m:k peptonized totalsurface

ac d red)co or essalkaline b:ije)accuC>alK

! aik'aciavoges - proskauermeth r-ed.*nico. +citrateatalase] Dos t vecytncnromrre o:,cdaseox,dase IKo>vaCS)gl!jcJse aeroboc

anaeroocfruLtose aerobic

anaerobicmarrc,s- aerotc.C

11 araeroD cga'actose aercDocanerr,cccxy.o7se aerc,b c

anaeroLcrharmnc aerob caraeracrc'ac rf,dcOC

anao-rocicsucrose aeroc -

anaerob cma tos, aerobic

ii anaerob c* c cc aercp c

I' araerobcr-aff inose aerobic U

I ananron coeXtrnr aero

anaerobicmann tol aerob c

anaeronicsorb to: aerobic

anaerobicammonujm nitraten-z- case -nitrates-- nitrites UH2Stri butyrirtrao esn0/129lip,c stan

PERCENTAGE OF TOTAL ORGANISMS5) w D CS a)0 0 0 0 0

I0 o 0 0

I-~~~~~~~~~~~~~

-m

U

FIG. 2. Computed per cent occurrence of antibiotic sensitivities, and enzymatic and biochemical activitiesbased upon the combined marine population of 150 microorganisms.

866

00

m

m

.l

J. BACTERIOL..

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

\ OL. 90, 1965

rodscJrve!d rodsspira!scoccobac' llarycOccIstorespleomorphtcmotileron-mo: leunipo ar fage at onperitrichous fiagel at ongram negat vegram 0o5ts egrarm var ablegrowth at 5°c

c20°c35°%

seaH20 requ recdgrowth n seaH2-2&dst.H20Mac trac nnovobioc .noleancomyc inpenic MlrlI-tracycl ne,omTyc inayar 0 qestedtarnch Cgested

geat n Iq ed.Jqes- proskaaermethy redintoIcitratecatalase 00os tcytochrome oxvdaseox idase Kovacs)gucose aerobic

n anaerotncfructose aerosic

1i anaerobicgaiactose aerobic

11 anaerobicsucrose aerob;c

anaerobicma tose aeroboc

11 anaerobicdextrn aerobic

I anaercDicmanntto' aerobic

1i anaerobicammonium nitraten-z-casernitrates -nitritesH2Stri buty rintr;oleinC-129

NUMNIERICAL TAXONOMY OF MARINE BACTERIA

PER'CENTA(G c-z: TOTAL ORGANISMS

FIG. 3. Computed per cent occurrence of a partial list of the characteristics in Fig. 1 and 2 based uponthe 91 microorganisms isolated froi? the Antarctic Sea.

A diagonal matrix of similarity coefficients(%-S) for 88 Antarctic bacteria, 6 tropical waterisolates, and 2 yeast cultures was computed ac-

cording to the formula of Sokal and Michener(1958), which includes negative matches and isbased upon the tests made of each isolate. Thematrix (M) generated consisted of 4,560 %'Svalues comparing each member OTE (N) withevely other [M = N(N - 1)/2]. This matrix isiandom, since no attempt was made to organizeor position any of the member OTIU -alues. Thematrix means little, and an attempt to rearrangethe order of the member isolates into meaningfulgroups was made by use of the single-linkagecluster technique (Sokal and Sneath, 1963).The rearranged matrix was partially plotted,

with four symbols to represent four numericalranges (Fig. 5). The total range of plotting, 80 to100 %S, was decided upon after sorting the ran-

dom matrix, because it was found that betweenthe 80 to 84.9c/%S level all the ml-embers of thetest became numerically% united. The four svm-

bols divide the range into 5%xy groups. The selec-tion of 5 %l groups was arbitrary and might beadjusted to smaller values to increase the resolu-tion of the technique.The single-linkage sorting procedure revealed

nine major groups (clusters) which have beennumbered from the smallest and farthest removedgroups (lower right hand corner of the diagonalin Fig. 5) to the largest single cluster (upper leftof Fig. 5), which represents the majority ofmicroorganisms included in the analysis.When the various characters used to calculate

the %S matrix are plotted according to thegroups shown in Fig. 5, it becomes obvious thateach cluster has a distinct pattern of characters.

Table 2 indicates the geographic areas of thethree U.S. Navy Eltanin (National ScienceFoundation research vessel) Antarctic cruisesfrom which the cold-water isolates were taken.The table shows the groups which were deter-mined by inspection of the sorted matrix (Fig. 5)

867

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

PFISTER AND BURKHOLDER J. BACTERIOL.

glucose11

fructose11

galactose11

suc rose11

maltose11

dextrin

mmanni tol11

aerobicanaerobicaerobicanaerobicaerobicanaerobicaerobicanaerobicaerobicanaerobicaerobicanaerobicaerobicanaerobic

PERCENTAGE OF TOTAL ORGANISMSb, Ln 0) -j OD

00 0 0 0 0 01,~ I1

--

-----

- .

0

0

FIG. 4. Computed per cent of the ability to use carbohydrate of 37 microbial isolates from the tropicalwater near Puerto Rico.

TABLE 1. Carbohydrate-utilizing ability of marine microorganisms

Computationt

Carbohydrate* 1 2 3

Aerobic Fermen- Aerobic Fermen- Aerobic Fermen-use tation use tation use tation

G.72.1 65.5 76.9 68.1 62.1 64.8G + F.60.2 52.9 61.5 49.4 62.1 64.8G + F + Ga.40.4 43.7 39.5 38.4 48.6 56.7G + F + Ga + S.39.0 39.7 38.4 35.1 48.6 54.0G + F + Ga + S + M.37.0 37.7 36.2 31.8 48.6 54.0G + F + Ga + S + M +D34.4 35.7 34.0 29.6 45.9 54.0G + F + Ga + S + M + D + Ma. 33.7 30.4 32.9 26.3 45.9 51.3

* G, glucose; F, fructose; Ga, galactose; S, sucrose; M, maltose; D, dextrin; Ma, mannitol.t Percentages computed for (1) combined Antarctic and tropical water populations of 151 micro-

organisms, (2) 91 Antarctic microorganisms, and (3) 37 tropical water isolates.

and the number of purified isolates from eachcruise in each group.The bacteria in each group were partially

identified in the classical manner (Bergey's Man-ual), and were found to be arranged in the fol-lowing way: Group I, a small cluster of bacteriabelonging to the family Micrococcaceae; GroupII, mainly gram-positive cocci or coccobacillaryforms which are nonmotile, nonsporeforming,and noncarbohydrate-using organisms; GroupIII, members of the family Pseudomonadaceaewhich do not attack carbohydrates and may bemembers of the Protoaminobacter, Alginomonas,or Mycoplana genera; Group IV, bacteria fromthe tropical waters in the vicinity of Puerto Rico,which are all gelatin liquifiers and methyl red-positive marine forms of the Pseudomonadaceae(appear to be similar to P. ichthyodermis as de-scribed in Bergey's Manual); Group V, membersof the family Achromobacteraceae and probablymembers of the genus Achromobacter; Group VI,members of the genus Flavobacterium; GroupVII, Pseudomonas types which liquify gelatin,

attack carbohydrates aerobically and anaero-bically, and hydrolyze starch; Group VIII,representatives of Pseudomonas species whichliquify gelatin but, except for one, do not hy-drolyze starch and do not use mannitol, lactose,and mannose; Group IX, also Pseudomonasspecies, but many of these digest starch and allshow good carbohydrate utilization, except for theaerobic use of mannose.The results of values computed according to

Sneath's (1957) method, in which negativematches were excluded, gave a matrix identical tothat plotted by including negative matches. Theonly difference noted was in the value of the %Scomputed.

DISCUSSIONThe purpose of this study was to establish a

foundation for studying a marine microbial popu-

lation on a large scale, with the eventual applica-tion of numerical taxonomy to act as a method oforganizing or separating members of a vast un-

known group. No attempt was made at this time

868

11

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

NUMERICAL TAXONOMY OF MARINE BACTERIA

a000

O0Do O0008| O88 Villa :9

oow0""- ~

0 ~ ~ ~:0

"90 *Vll

re 0F)@

KEY

95-100/.90-94 9-t. 085-89-9I 0

80-849. (D

0

00 VI

0

@0 v

000 v

Go

00

0@0006

QC)

00

FIG. 5. Matrix sorted and reordered by single-linkage clustering technique (Sneath, 1963). The range of80 to 100%S was divided into four 5% ranges and plotted as shown in the key. Nine major groups of orga-nisms are shown along the diagonal. Any value less than 80% was not plotted.

TABLE 2. Number of isolates per cruise on three Antarctic cruises, classified into groups as shown in Fig. 5

No. of cruise isolates in groupCruise Date Location (area)

I II III IV V VI VII VIII IX

Eltanin 9 September, 1963 South Georgia 3 6 4Eltanin 11 January, 1964 South Pacific 1 1 13 21Eltanin 12 April, 1964 South Orkney Islands 5 4 10 3 1

-* July, 1964 Puerto Rico 4

* For comparison, some Puerto Rican cultures are also included.

to identify explicitly any members of this groupof marine microbes, but only to gather as muchuseful information as possible for taxonomic pur-poses. From this kind of population survey, we

may learn what can be of taxonomic importance

within any given group, with the result that itmay be possible to weight accurately and legiti-mately the occurrence of characteristics used in a

numerical approach to classification. There hasbeen some discussion (Sokal and Sneath, 1963)

VOL. 90, 1965

I@04

0

1000.@os0000oo000000

oooSooo: oo 000300o0@@@0e @@ 00filQd beldbebddSCdb 2.d

! 0040@4>gW

869

186204

203193

199192

202187

206208

197175

188205

185190

183127

209176

182107

180222

212198

184196

194214

213189

211233

195178

181191

177201

200235

251236

253252

247232

241239

° 234 226Z 223w 219

0 216U 150

149172

171259

261262

260249

250242

217174

229170

161156

159207

179116

163166

258164

154173

221224

218169

160264

263167

1485

165152

151

:06COUSUCS00006066ocee aloeocooveI ea eadbdbiahafthn.

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

PFISTER AND BURKHOLDER

concerning the merits of this procedure, andsince little is known about the diagnostic prop-erties of large groups of marine microbes, it hasseemed worthwhile to make this study.The microbial population examined thus far

consisted of 114 isolates from Antarctic regionsand 37 isolates from tropical waters near PuertoRico. The methods of isolation varied, resultingin a collection of both obligate psychrophilicorganisms (not growing above 10 to 15 C) andorganisms capable of growth at higher tempera-tures. The majority of cultures tested from theAntarctic were of the latter type, and werecapable of growth in the temperature range from5 to 35 C. The obligate psychrophiles werehandled carefully with respect to their tempera-ture requirements and environment, since shortexposure to even 20 C caused death. The initialisolation of some of these psychrophiles was slow,and plates were examined in the cold up to a2-week incubation period. When pure cultureswere isolated, it was found that luxuriant growthoccurred in 48 to 72 hr at 5 C. Careful selectionof clones was imperative because many of thesmall colonies are translucent and can only beseen when using a stereomicroscope to observecolonies on the agar surface.The majority (66.2%) of the microbes ex-

amined were found to be gram-negative singlerods with one polar flagellum. This assemblageprobably belongs in the group of pseudomonads.Figure 1 shows the disposition of the variousforms found in the population, demonstrating apreponderance of cultures containing single cellsand diploforms rather than chains. Some of thecultures (8 7) were pleomorphic during activegrowth when examinations were made, whereasmany more became pleomorphic late in theirgrowth cycle.

There was some incidence of curved rods amongthe bacteria tested, and, since some of them weregram-negative with a single polar flagellum andsensitive to the pteridine (0/129) compound ofShewan et al. (1954), they are presumed to beVibrio types. Two of the Antarctic isolates wereyeast cultures, which were included in the testingprocedure with no further distinction or attemptat identification.

Figure 2 shows the various biochemical teststhat were made, and the resulting per cent ofreactions. The six antibiotic-sensitivity testsshowed a wide variance of these marine organismsin their susceptibility, with only one Antarcticand one tropical culture being susceptible to allsix compounds.

Antibiotics have not generally been consideredas a tool for taxonomic work because of microbialadaptability and for various other reasons. It is

doubtful, however, that the population of marinemicrobes had experienced prior contact with theantibiotic compounds, and perhaps the deter-mination of antibiotic sensitivity may be of somevalue in taxonomic studies. This is especially truewhen these tests are considered principally ascharacters to Le used in numerical taxonomicseparations. The inclusion of such data in anumerical scheme for taxonomy may be of sig-nificance in the ultimate assignment of any mem-bers of the population into appropriate phenons.

Examination of the results (Table 1) showingthe percentage of total organisms able to fermentcumulatively the seven carbohydrates in thetests suggests in general that both the Antarcticand tropical water species are able to fermentaerobically or anaerobically a variety of carbo-hydrates. In addition, it can be noted that oncecarbohydrate utilization is recorded (e.g., glucosefermented) the successive use of another carbo-hydrate (e.g., glucose plus fructose) may occurwith increasing probability. This is most easilyseen in the last column of Table 1 which shows allseven carbohydrates are being fermented by thesame organisms, and yet this group includesabout 30% of the population tested.

There were some distinct differences betweenthe Antarctic and Puerto Rican collections, otherthan expected temperature requirements. Noneof the 114 Antarctic specimens was able to syn-thesize indole, whereas 46% of the tropical iso-lates gave a positive test for the compound. Inaddition, the ability of the bacteria from tropicalwaters to ferment aerobically or anaerobicallythe various carbohydrates of the test was slightlygreater than that of the Antarctic specimens.There may be other undetected differences be-tween these populations, but it would be desir-able to increase the number of specimens, particu-larly the tropical isolates, before definitivestatements can be made. The significance of thedifference in indole production is not understoodat the present, but it may reflect a biochemicalalteration in tryptophan metabolism character-istic of the populations.That 83% of the organisms should require

seawater for growth and 17% show ability togrow also in distilled water media is of interest.The presence of nutritive materials in the shallowwaters near Puerto Rico must certainly influencethe microbial flora of the littoral and neritichabitats. Throughout the varied marine en-vironments of the world oceans, one expects tofind complex ecological relationships exertingprofound influences upon the types of micro-organisms which comprise the microfloral andfaunal communities.We are endeavoring in this program to develop

870 J. BACTERIOL.

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

NUMERICAL TAXONOMY OF MARINE BACTERIA

a system for storage and recall of quantities ofdata concerning large if not infinite populations.Because of the burdensome nature of this typeof research, modern methods of data storage andprocessing are being used. All data records arekept on IBM punched cards in a manner similarto that used by Colwell and Liston (1961) andon "Keysort" needle punch cards prepared forus by the Royal McBee Co., New York, N.Y.Using these tools, we can now survey populationsof microorganisms which heretofore could not bestudied.A computer technique with the principle that

every feature should have equal weight in de-termining its taxonomic position was used toanalyze a population of marine microorganisms.The population of microorganisms was composedof bacteria, with the exception of cultures 148and 167 which were yeast; these were includedso we could observe whether they would clusterwith any other known group. Examination ofFig. 5 shows the two yeasts with a fairly high %Svalue for each other, but not grouped with anyof the other nine major clusters.

There are a number of conclusions that can bederived from these studies. Most important isthe fact that this technique of computer analysisappears to be a practical method for studyingvery large collections of unidentified marine bac-teria.With a suitable data-storage program, the

computer technique can be used constantly tocalculate the %S value of organisms which areadded to the collection provided the requirednumber of definitive tests have been made, andin this way it becomes possible to place new en-tries in their appropriate clusters. These tech-niques carried out with a computer on board anoceanographic vessel would enable the investi-gator to remain at sea for long periods and makea systematic appraisal of marine microbial popu-lations from one geographic location to another.In this manner, much information about themicrobiology of the sea could be learned rapidly.There are probably many uses for this kind ofinformation. Certainly it would be significant tohave investigations about the food chain, roles ofmicroorganisms in the productivity, and thetransformation of organic and mineral matter inthe world oceans. It is possible that certain kindsof bacteria are indicators of temperature, pH, orsalinity values, but without some kind of sys-tematic study, their presence and usefulness instudies of the marine environment will not beappreciated.Examination of Table 2 shows clustering of

the bacterial isolates with the cruise from whichthey were taken, Eltanin 9, 11, or 12. This

clustering may be the result of either the geo-graphic origin of the isolates, since each cruisewas made in a different area, implying that theremay be different populations in different areas,or it may be related to the seasons, because eachcruise was made at a different time, implying sometemperature relationships responsible for thegrouping. There are other factors which mayinfluence this kind of study, such as local weatherconditions, proximity to land, pH, salinity,available substrates, and depth from which thecultures were obtained. Sampling proceduresmust be carefully regulated and many samplesneed be collected to obtain a representative popu-lation in any given area.A large-scale, well-planned, and coordinated

program is required to learn valuable informationabout the microbiology of the sea. It is believedthat the numerical approach to taxonomy (identi-fication) with the capability of massive datastorage will be of practical value for investigationsof large populations in the world oceans.

ACKNOWLEDGMENTSThis investigation was supported by research

grant GB-570 from the National Science Founda-tion to Lamont Geological Observatory of Colum-bia University. National Science Foundationgrants 19612 and GA15 for collecting on theEltanin are gratefully acknowledged.We are indebted to James Dorman for his skilled

assistance and helpful discussions concerningcomputer techniques.

LITERATURE CITEDCOLWELL, R. R., AND J. LISTON. 1961. Taxoniomic

relationships among the pseudomonads. J. Bac-teriol. 82:1-14.

COLWELL, R. R., AND AI. MANDEL. 1965. Adan-sonian analysis and deoxyribonucleic acid basecomposition of Serratia marcescens. J. Bacteriol.89:454-461.

GABY, W. L., AND E. FREE. 1958. Differentialdiagnosis of Pseudomonas-like microorganismsin the clinical laboratory. J. Bacteriol. 76:442-444.

KOVACS, N. 1956. Identification of Pseudomonaspyocyanea by the oxidase reaction. Nature 178:703.

LEIFSON, E. 1960. Atlas of bacterial flagellation.Academic Press, Inc., New York.

LOCKHART, W. R. 1964. Scoring of data and group-information in quantitative taxonomy. Develop.Ind. Microbiol. 5:162-168.

LYMAN, J., AND R. H. FLEMING. 1940. Compositionof sea water. J. Marine Res. 3:134-146.

QUADLING, C., AND R. R. COLWELL. 1964. Theuse of numerical methods in characterizing un-known isolates. Develop. Ind. Microbiol. 5:151-161.

ROGERS, D. J., AND H. FLEMING. 1964. A computer

VOL. 90, 1965 871

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

872 PFISTER AND BURKHOLDER

program for classifying plants. II. A numericalhandling of non-numerical data. Bioscience 14:15-28.

SHEWAN, J. M., W. HODGKISS, AND J. LISTON.1954. A method for the rapid differentiation ofcertain non-pathogenic, asporogenous bacilli.Nature 173:208.

SKERMAN, V. B. D. 1959. A guide to the identifica-tion of the genera of bacteria. The Williams &Wilkins Co., Baltimore.

SNEATH, P. H. A. 1957. The application of com-puters to taxonomy. J. Gen. Microbiol. 17:201-226.

J. BACTERIOL.

SOCIETY OF AMERICAN BACTERIOLOGISTS. 1957.Manual of microbiological methods. McGraw-Hill Book Co., Inc., New York.

SOKAL, R. R., AND C. D. MICHENER. 1958. A sta-tistical method for evaluating systematic rela-tionships. Univ. Kansas Sci. Bull. 38:1409-1438.

SOKAL, R. R., AND P. H. A. SNEATH. 1963. Prin-ciples of numerical taxonomy. W. H. Freemanand Co., San Francisco.

TENDLER, M. D., AND P. R. BURKHOLDER. 1960.Studies on the thermophilic actinomycetes. I.Methods of cultivation. Appl. Microbiol. 9:394-399.

on February 24, 2020 by guest

http://jb.asm.org/

Dow

nloaded from