INFECTlON AND IMMUNITy, Mar. 1974, p. 550-558Copyright 0 1974 American Society for Microbiology

Vol. 9, No. 3Printed in U.S.A.

Hemodialysis Culture of Serratia marcescens in aGoat-Artificial Kidney-Fermentor System'

JOHN M. QUARLES, RALPH C. BELDING, TEOFILA CABRERA BEAMAN,AND PHILIPP GERHARDT

Department of Microbiology and Public Health, Michigan State University, East Lansing, Michigan 48824

Received for publication 27 July 1973

Hemodialysis was employed to simulate in vivo conditions for growth inmammalian blood, but without phagocytosis, by using the goat and Serratiamarcescens as a host-parasite model. The blood stream was shunted surgicallyvia prosthetic tubing from a carotid artery through the hollow-fiber membranesin an artificial kidney hemodialyzer and back into a jugular vein. The dialysatesolution concurrently was pumped from a modular fermentor through thehemodialyzer jacket outside of the membranes and back into the fermentor.Hemodialysis between the two circuits was maintained continuously. Whenequilibrium was attained, bacteria inoculated into the dialysate circuit multi-plied first exponentially at the maximal rate and then arithmetically at a lesserrate equally well under aerobic or anaerobic conditions. When a population ofabout 109 viable bacteria/ml was exceeded, the goat reacted acutely with signs ofgeneral toxemia, pyrexia, and leukopenia, apparently because of dialyzable toxicmaterial produced by the culture. The maximal molecular size of the toxicmaterial was defined relative to a rigid globular protein of 15,000 in molecularweight and 1.9 nm in hydrodynamic radius or to a flexible fibrous polyglycol of5,500 in molecular weight and 2.6 nm in hydrodynamic radius, based ondeterminations of the membrane porosity threshold for dialysis.

We have devised a new way to grow organismsby in vivo dialysis, with the experimental ra-tionale derived from the fact that the interiormilieu in animals is maintained essentially bycirculation of the blood. Dialyzers connecteddirectly with the circulatory system are incommon clinical use as artificial kidneys forhumans. It appeared that such a hemodialyzercould be used to establish continuous communi-cation between the blood stream of an animaland a fixed volume of dialysate solution inocu-lated with an organism. In this way, dialyzablemolecular constituents of the blood would dif-fuse into the culture and so feed it, yet the bloodcells and macromolecules would be too large todiffuse through the membrane so that phagocy-tosis and immunological reactions against theculture would be prevented. Conversely, meta-bolic products of small molecular size from theculture would diffuse through the membranebarrier into the blood and thus relieve thefeedback inhibition that often limits growth in aclosed culture system. The dialyzable cultureproducts, including toxins, could also exert

'Journal article no. 6446 of the Michigan AgriculturalExperiment Station.

effects on the host. Membranes of differentporosity could be employed in the system todistinguish between molecules of different sizein the blood components and culture products.The system would be analogous to the batchfermentor-continuous reservoir mode of operat-ing an in vitro dialysis culture (see Fig. 8 inreference 11), but with the blood supply of ananimal used as the nutrient reservoir.

In vivo dialysis culture was first attempted in1896 by Metchnikoff, who grew cholera bacteriain a collodion membrane sac implanted withinthe peritoneal cavity and thereby demonstratedthe production of a diffusible toxin (6). Subse-quently, dialysis chambers were implanted inthe peritoneal cavity and elsewhere for a num-ber of purposes by using a number of differentorganisms, including animal cells and tissues(see review by Schultz and Gerhardt [11]).However, the use of implanted chambers islimited by the environment in the peritonealfluid being different from that in the blood, bythe occluding growth of macrophages on themembrane surface, by the restricted size, andby the difficulty in sampling.We conceived the use of hemodialysis to offset

55()

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

HEMODIALYSIS CULTURE OF S. MARCESCENS

these limitations and essentially to transposethe blood milieu extracorporeally. The feasibil-ity of hemodialysis culture was tested with thedomestic goat and Serratia marcescens as thehost-parasite model. Described below are themethodology, the unusual growth characteris-tics of the bacterial culture, the acute toxemicresponse of the host, and the maximal molecu-lar size of the dialyzable toxic material.

MATERIALS AND METHODSThe hemodialysis culture system is depicted and

diagramed in Fig. 1. Blood from an experimentalanimal was circulated by the heart through one side ofan artificial kidney hemodialyzer (the blood circuit),and dialysate culture from a modular fermentor wascirculated by a pump through the opposite side (thedialysate-culture circuit).The goat was selected as the experimental animal

because of its convenient size, long neck for accessibil-ity in vascular surgery, placid disposition, and hardi-ness. The goat has disadvantages in its relativesensitivity to anaesthesia and hemolysis (2). Short-haired domestic goats of mixed breed were used, atotal of 15 in this study. All were females, 1 to 3 yearsof age and 50 to 100 lb (22 to 45 kg) in weight. Theywere maintained indoors in stalls, unrestrained ex-cept for a rear-leg hobble. During dialysis trials thegoats were partially harnessed into a caged platformbut were free to stand, lie, eat, drink, and excretewastes. The body temperature was measured with arectal Fahrenheit thermometer as frequently asneeded.A permanent external prosthetic shunt was estab-

lished between a carotid artery and jugular vein in thelower neck by vascular surgery, based on the tech-niques of Quinton et al. (8, 9). Vessel tips, connectorsand infusion T's were made of medical-grade Teflon,and the tubing was made of Silastic (Cobe Laborato-ries, Inc., Denver, Colo.).

Sodium heparin (Upjohn Co., Kalamazoo, Mich.)was used to prevent the blood from clotting, withsubcutaneous injections of 15,000 to 20,000 USPunits/24 h. During dialysis, about 1,500 USP units/hwere continuously introduced into the arterial shuntby means of an infusion pump (model 240, SageInstruments, Inc., White Plains, N.Y.). Ethyl isobu-trazine hydrochloride was administered as a tranquil-izer in early experiments but later was obviated bytraining the animal. Neither drug affected growth ofthe test bacterium. The possibility of blood clotting inthe hemodialyzer was monitored visually and by athermometer in the effluent tubing, where the tem-perature went down if the flow was reduced byclotting.A given goat became usable for hemodialysis ex-

perimentation about a week after surgery and re-mained so for about 1 to 3 months, when accidentalexsanguination or the formation of a clot in an arteryor vein (despite heparinization) caused termination.A Cordis-Dow hollow-fiber artificial kidney was

selected as the hemodialyzer (model 2 or 3, CordisCorp., Miami, Fla.). The jacketed cylindrical unit

measures 7.0 cm in diameter and 21.6 cm in heightand contains a bundle of 11,000 hollow-fiber mem-branes fabricated from regenerated cellulose, throughwhich the blood is circulated. Each fiber is 13.5 cm inlength and 225 Am in inside diameter, and the bundleprovides about 1 m2 of total membrane surface area.The dialyzer was sterilized with 1.5% formaldehydeand rinsed thoroughly with water before use. Thedialysate culture was circulated outside the hollow-fiber membranes, through the dialyzer jacket.

Beaker chemical dialyzers with a looped bundle ofhollow-fiber membranes (Dow Chemical Co., Mid-land, Mich.) also were employed as hemodialyzers.The beaker dialyzers measure 7 cm in diameter and 14cm in height and provide a total membrane surfacearea of about 0.1 M2. They were used with threedifferent porosities identified in terms of the nominalmolecular weight (MW) of retained molecules: MW300 (Osmolyzer, model b/HFO-1); MW 10,000(Dialyzer, model b/HFD-1, with the same porosity asthe artificial-kidney dialyzer); and MW 30,000 (Ul-trafilter, model b/HFU-1). The beaker dialyzers thatwere first tested caused hemolysis and toxemia andtherefore were unsatisfactory for use as hemodialyz-

PUMP

FERMENTOR-

GOAT ARTERIALSHUNTRL HE PAR ININFUSER

THERMOMETER

iHEMODIALYZER.

VENOUSK

BLOOD CIRCUIT O A LYSATE - CULTURECIRCUIT

FIG. 1. Hemodialysis culture system. The majorcomponents of the experimental system shown in thephotograph (top) are identified in the tracing (bot-tom). An artificial kidney is shown, but beakerdialyzers also were employed as the hemodialyzer.

VOL. 9, 1974 551

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

QUARLES ET AL.

ers. Howevrer, the embedding material used to securethe fiber bundles subsequently was changed in manu-f'acture, and more recent products proved satisf'actoryas hemodialyzers. In differential dialysis trials, threebeaker dialyzers were inserted in series into thedialysate-culture circuit in addition to the artif'icialkidney as the hemodialyzer. Consequently, threeseparate samples of' dialyzable culture product wereobtained simultaneously with hemodialysis.A modular 1-liter glass fermentor with control of

temperature, agitation, and aeration (Microf'ermmodel 102, New Brunswick Scientific Co., Inc., NewBrunswick, N.J.) was used to contain the dialysate-culture suspension. Anaerobic conditions were

achieved by continuously sparging the fermentorcontents with sterile argon, which was scrubbed freeof' oxygen by passage over hot copper wire keptreduced with a stream of hydrogen gas. Aerobicconditions were achieved by continuously spargingthe fermentor contents with f'ilter-sterilized air at a

rate of 2.4 liters/min and by driving the impellor at a

rate of' 300 rpm. Unless otherwise stipulated, theculture in the fermentor was aerated and stirred.A Maisch sterilizable gear pump (model 53DFC,

Tuthill Pump Co., Chicago, Ill.) was used to circulatethe dialysate culture from the fermentor throughtubing to the dialyzer and back at a rate of 400 ml/min.Thick-walled rubber tubing was used for the dialy-sate-culture circuit. Medical-grade Tygon or Silastictubing was used for the blood circuit.The dialysate-culture circuit was charged with 800

ml of a glucose-salts solution approximately balancedin makeup to that of goat blood (2). I'he solution was

constituted as follows (per 100 ml): glucose, 50 mg;Na+ (as NaCl and sodium acetate), 310 mg; K+ (asKCl), 15 mg; Ca2+ (as CaCl2), 11 mg; and Mg2+ (asMgCl2), 3.7 mg. The solution volume representedabout 40%c of the total blood volume of a goat.

S. marcescens strain 8UK was selected as the testorganism because of its previous use in developmentalstudies on in vitro dialysis culture (3. 4). Thisgram-negative species occurs mostly as single cells.grows rapidly on synthetic or natural media undereither aerobic or anaerobic conditions, and is redpigmented when grown on most media at about 30 C.S. marcescens is quite capable of parasitic growth,produces endotoxin, and may cause serious diseases inman. The inoculum was prepared f'rom an aeratedexponential culture at 35 C in Trypticase soy broth(Baltimore Biological Laboratory, Cockeysville, Md.),which was sedimented, and the cells were resus-

pended in the dialysate solution to an appropriateconcentration.

Hemodialysis was initiated, and the system was

allowed to equilibrate for 1 h prior to introducing theinoculum. The culture was maintained at 39 C (ap-proximately the normal temperature of' goats) andconsequently was not pigmented. Samples of' theculture were removed periodically by syringe andneedle through a self-sealing rubber diaphragm posi-tioned in the circuit tubing. Cell populations were

measured by optical density and by viable cell counts,which were performed by surface plating on Trypti-case soy agar.

Cell-free samples of the culture for biochemi-cal analyses were obtained by membrane filtration.Samples of' the blood were collected in ethylenedi-aminetetraacetic acid for cell counts or were allowedto clot for serum, which was immediately stored at-20 or -70 C until assayed. Assays for the followingroutinely were made on the serum samples with aflame photometer and autoanalyzer (model SMA 12,Technicon Corp., Tarrytown, N.Y.) by use of stan-dard procedures: sodium, potassium. calcium, mag-nesium, inorganic phosphate, glucose. urea nitrogen,uric acid, total protein, albumin, bilirubin, cholesterol.alkaline phosphatase, lactic dehydrogenase, andglutamic-oxaloacetic transaminase. Leukocyte anderythrocyte counts of the blood samples were madeby use of microscope counting chambers, and differ-ential cell counts were made by use of Wright-stainedsmears.

The membrane porosity threshold of' the artificialkidney was determined with the distribution analysismethod of Scherrer and Gerhardt (10). A polyethyleneglvcol (PEG) sample of number average molecularweight (M,,) :3.350 was dialyzed against a PEG sampleof Mn 17,500, with the concentrations adjusted tomaintain osmotic balance across the membrane. Dial-ysis was continued o)r 4.5 h at 23 C. The change inconcentration of the PEG sample of' Mn 3,350 wasmeasured by ref'ractometry and dry weight, and itsmolecular weight distribution before and af'ter dial-vsis was determined by column chromatography withBio-Gel P-10 (Bio-Rad Laboratories, Richmond,Calif.). The mass of polymer in each elution fraction(1.5 ml) was measured by dry weight.

RESULTSHemodialysis without culture. During, con-

trol hemodialasis with the artificial kidney, thedialyzable materials that were measured insamples of the blood and the dialysate solution(sodium, potassium, calcium, magnesium, inor-ganic phosphate, glucose, and urea nitrogen)reached equilibrium within 1 h and then re-mained within normal ranges of concentrationfor goat blood. Large molecules in the blood(total protein, albumin, bilirubin, cholesterol,alkaline phosphatase, lactic dehydrogenase,and glutamic-oxaloacetic transaminase) did notmeasurably diffuse into the dialysate solutionand remained at normal concentrations. Bodytemperature of the goat (Fig. 2A) and blooderythrocytes (Fig. 2C) also were unaffected.However, total leukocytes in the blood in-creased in number during the first few hours ofcontrol hemodialysis and then remained essen-tially constant at the higher level (Fig. 2B). Thetwo most numerous types of leukocytes (neutro-phils and lymphocytes) were not affected selec-tively.Growth characteristics of hemodialysis

culture. The glucose-salts solution originallycharged into the dialysate circuit was shown to

552 INFECT. IMMU.NITY

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

HEMODIALYSIS CULTU:

WITHOUTCULTURE

to0

0m

(Dto

0

0

WITHCULTURE

O 2 4 6 8 10 12 0 2 4 6 8 10 12TIME (HOURS)

FIG. 2. Representative time courses of host re-sponses to hemodialysis without culture (left) andwith culture of S. marcescens (right). A and E, bodytemperature; B and F, white blood cells; C and G, redblood cells. In D there is shown the correspondingbacterial growth curve in the dialysate-culture circuit.

be inadequate to support growth of S. marces-cens, primarily because of the absence of anitrogen source. Even after attainment of equi-librium with blood, the dialysate solution sup-ported only a slight amount of bacterial multi-plication when samples of the dialysate werewithdrawn into separate culture tubes and inoc-ulated.When hemodialysis was maintained continu-

ously for an extended period with the artificialkidney, an inoculum of S. marcescens in thedialysate circuit multiplied extensively and in acharacteristic bimodal pattern, with a short lagphase (Fig. 2D, 3A, 3C, 4A, and 5A) or withoutone (Fig. 5D). The multiplication at first pro-gressed in the usual way as an exponentialfunction of time. The rate of exponential multi-plication in hemodialysis culture, about 2.0generations/h, was essentially as great as themaximum that had been attained with this

RE OF S. MARCESCENS 553

bacterium by other means of culture or types ofmedium.When a population density of about 109

viable cells/ml was reached, the multiplicationchanged from an exponential to an arithmetic(linear) function and progressed at a rate aboutone-sixth of that before (Fig. 2D, 3A, 3C, 4A,5A, and 5D). This second mode of multiplica-tion was verif'ied by plotting multiples (ratherthan logarithms) of' bacterial numbers versustime, with a resulting straight-line progression(Fig. 3B and 3D). The shift in growth mode wasexplained by the bacterial population reachinga density for which the nutrient demand ex-ceeded the steady-state supply of' nutrientsfrom hemodialysis. Because diffusion proceedsas a direct function of time, the bacterialmultiplication became limited similarly. Thebimodal pattern of multiplication fulf'illed theo-retical predictions (see Fig. 25 in reference 11).When an inoculum of 109 viable cells/ml was

employed (near the critical density for change

SMALLER INOCULUM LARGER INOCULUMI I .

A

-J

U)-J-J

C.)

w-J

4

0-J

C

10-

9-

7+40

B D

1, 30-30L)-J-Jw

,,20.m4

ao 10-

O _r

0 2 4 6 8 10 12 0 2 4 6 8 10 12HOURS OF CULTURE

FIG. 3. Representative growth curves of S. mar-cescens in hemodialysis culture, from a smaller inocu-lum (A and B) and a larger one (C and D). The resultsare plotted with both logarithms (A and C) andmultiples (B and D) of the bacterial numbers on theordinates to show exponential and arithmetic multi-plication modes, respectively.

VOL. 9, 1974

(

Z: I

a- 1(mw H

H

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

554 QUARLES ET AL. INFECT. IMMUNITY

I I l l l l l l highly aerobic conditions in the fermentor be-A cause of knowledge that oxygen diffuses slowly

-j through a cellulose dialysis membrane (3, 11).5 10 - - However, strictly anaerobic conditions in the1% . fermentor resulted in the same bimodal growth

pattern and unexpectedly supported nearly theW9Jg _ / _ same exponential and arithmetic rates of multi-O plication as with aerobic conditions (Fig. 5DW / versus 5A). Apparently the limitation on growthco d rate of S. marcescens by oxygen supply (3, 4, 12)< 8 _ / _ occurs only in artificial media or at population> densities in excess of those attained with hemo-

o / dialysis culture.27 - l _ The artificial kidney was available with only

one membrane porosity, which retains mole-cules larger in nominal molecular weight than

6 - _ about 10,000. However, chemical beaker dialyz-ers constructed similarly with a bundle of hol-

I I I I I I low-membrane fibers and available with threel00 B t. different porosities of membrane were found90 _ - satisfactory for use as a hemodialyzer. The

_j 80 _/ \C7 BLOODA0AEROBIC ANAEROBIC

10

60cscn-AonAlcscocnrtoDnthil- 2w 18 1

50-w 9--Jw 8(n 40Q

o DiIALYTem > 7

(D20O 107-1u6 B E

o. 105-01024681012 ~~~~~~~w104-HOURS OF CULTURE 103

102FIG. 4. Effects of goat hemodialysis culture of S. 22

marcescens (A) on glucose concentration in the dialy- 20 Fsate and in the blood (B). c

,n- 18

in multiplication mode with a smaller inocu- 16lum), the multiplication soon became arith- T14metic (Fig. 3D) and progressed in this mode for 0 1several hours. A maximum was reachedatterabout 10`16 viable cells/ml (Fig. 3C), followed 0 10by a slight decline in the population (Fig. 3D)...0 8The effects of the bacterial multiplication on 6-

glucose concentrations in the dialysate and in 4 ITthe blood are shown in Fig. 4. After an initial 0 2 4 6 8 10 12 0 2 4 6 8 10 12rise, the dialysate glucose decreased rapidly and TIME (HOURS)eventually became undetectable. The pattern FIG. 5. Comparison of aerobic (left) with anaerobicwas similar for glucose in the blood, with greater (right) hemodialysis culture conditions on the bacte-hyperglycemia initially and lesser hypo- rial growth curves (A and D) and on the host bodyglycemia subsequently than in the dialysate. temperature (B and E) and blood leukocytes (C andThe cultures were managed usually with F).

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

HEMODIALYSIS CULTURE OF S. MARCESCENS

growth responses of 'beaker dialyzers areMW-300 beaker dial)significant change in Ian extended time pe:dialysate glucose decr4trial, indicating metalture. In vitro dialysis i

when using Trypticasereservoir also failed totial nutrient moleculvented from passageThe MW-10,000 beakwith the same membkidney, and the MW-ported a bimodal patte6G, respectively) similcial kidney. The menregulate the passage ofof small molecules,proteins. Representati'tiplication during egrowth, respectively, i

tions/h with the MW-10.3 generations/h withHost responses to

When an inoculum ofduced into the dialsequilibration period uthe goat seemed quitehours, depending on 1The body temperatu

-a MW 300s ,,,, ...

A D,, 10 .

g

w~~~~~~~5; 87-°8 6-a°

106E 105> 104-103 OIVa~102

18 , , F

a 16

0 2

°2 12CD3Wa

T

FIG. 6. Time courses

host responses with bporosity used as the henretained molecules with300 (left), 10,000 (centey

S. marcescens with the *erythrocyte count (Fig. 2G), and the serumshown in Fig. 6. The biochemistry for the most part remained withinyzer did not support a normal ranges. The blood glucose concentration)acterial population over (Fig. 4B) and the white blood cell count (Fig.riod (Fig. 6A) although 2F) both increased in the initial hours of hemo-eased slightly during the dialysis culture, although the latter reactionbolic activity of the cul- also occurred in hemodialysis without culturewith the MW-300 beaker (Fig. 2B).soy broth medium in the After this uneventful incubation period, ansupport growth, so essen- acute episode of illness developed in the goat[es apparently are pre- with dramatic suddenness, apparently becausethrough this membrane. of dialyzable toxic materials produced by theer, which is constructed culture. The animal evidenced general signs oforane as in the artificial toxemia with violent shivering and lethargy-30,000 beaker both sup- that persisted for 2 h or more. In some casesirn of growth (Fig. 6D and bloody urine was passed; in others urine outputlar to that with the artifi- ceased. Feeding and drinking also ceased duringnbranes in both beakers an episode.much the same size class The time courses of some host responsesand both retain serum measured in a representative trial are shown inve rates of bacterial mul- Fig. 2, together with control responses duringbxponential and linear hemodialysis without culture. The body tem-were 1.7 and 0.2 genera- perature (Fig. 2E) started to rise after about 7 h10,000 beaker and 1.8 and of hemodialysis culture, which coincided withthe MW-30,000 beaker. the time when a population of 109 viablehemodialysis culture. cells/ml was exceeded (also see Fig. 7A) and

S. marcescens was intro- when multiplication changed from an exponen-ysate solution after the tial to a linear rate in the culture (Fig. 2D). Anvith the artificial kidney, acute reduction in the number of white bloodnormal for a number of cells was a second main sign of host toxemia

the size of the inoculum. (Fig. 2F). The number of red blood cells re-tre (Fig. 2E), the blood mained within the normal range (Fig. 2G).

Among the other host parameters monitored,MW 10.000 MW 30.000 only the concentration of blood glucose changed

G significantly, with the development of moderatehypoglycemia (Fig. 4B).

General pictures of the host pyrexia andleukopenia responses are shown in Fig. 7, inwhich the average changes in six trials were

H plotted as functions of the number of viablebacterial cells in the hemodialysis culture. Itwas apparent that both responses ensued after apopulation of about 109 viable cells/ml was

...attained.The fever and leukopenia observed with the

aerated hemodialysis cultures were less appar-ent with cultures maintained under strictlyanaerobic conditions, although the culturesmultiplied similarly in the two situations (Fig.5). The reason for the difference in host re-

0 . .......,,,sponses was not apparent.4 6 8 10 12 0 4 6 8 1012 Substitution of each of the three beaker

'IME (HOURS) dialyzers in place of the artificial kidney in theof hemodialysis cultures and hemodialysis culture system resulted in predict-eaker dialyzers of different able host responses (Fig. 6). Because thenodialyzer, each identified by>nominal molecular weight of MW-300 beaker as a hemodialyzer faled tor). and 30.000 (right). support appreciable multiplication of the inocu-

555VOL. 9, 1974

I

Jst*&%A "W-ow/k -

,-.

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

QUARLES ET AL.

+4

LL

LUJ

z

LUL(Dz

43

1

0

0

o

z

LL(5

z

I

U

0

1

- 2

LOG VIABLE CELLS/MLFIG. 7. Normalized change in body temperature

(A) and in white blood cell count (B) as a function ofthe numbers of viable S. marcescens cells in hemodi-alysis cultures, averaged from six cultures with fourgoats. At the reference point of origin, the meantemperature was 102.8 F (39.3 C) and the meancount was 12.2 x 103 white blood cells/mm3.

lum (Fig. 6A), no significant responses were

observed in the host (Fig. 6B and 6C). TheMW-10,000 or MW-30,000 beaker as a hemodi-alyzer supported similar patterns of culturegrowth (Fig. 6D and 6G) and host responses(Fig. 6E, 6F and 6H, 61) as did the artificialkidney (cf. Fig. 2), although the fever andleukopenia responses with the two beakers oc-

curred earlier in the time course of growth.The three beaker dialyzers were then inserted

in series into the dialysate-culture circuit tocollect three different size classes of cultureproduct simultaneously with artif icial kidneyhemodialysis. When the goat evidenced acutetoxemia, the three differential dialysis sampleswere removed. Injection of the MW-10,000beaker-dialysate sample into another goat (3ml, intravenously) evoked pyrexia but not mea-

surable leukopenia, thus partially duplicatingthe results obtained by direct hemodialysisculture with either the artificial kidney or thesame beaker dialyzer even though the amountof dialyzable toxic material was much less than

that transferred over a period of time in contin-uous hemodialysis. Similar results were ob-tained with the M-W-30,000 beaker sample.However, the MW-300 beaker sample (and alsothe original salts solution, employed as a con-trol) did not evoke any measurable host re-sponse.

It might be thought that the host responseswere caused by some nonspecific effect of bacte-rial growth, such as the depletion of bloodglucose. This was shown unlikely by the obser-vation that goat hemodialysis culture of Bacil-lus anthracis (the avirulent Sterne vaccinestrain) did not result in a toxemic host responseeven though massive bacterial growth and hosthypoglycemia occurred with either aerobic oranaerobic conditions in the fermentor. Theanthrax cells grew in long chains but did notform spores.The argument also might be advanced that

the results were caused by intact bacterial cellspassing through an undetected opening in themembrane of the hemodialyzer. In the controlhemodialysis trials, smaller particles (bloodproteins) did not pass into the dialysate circuit.Furthermore, in dialysis culture in vitro withthe artificial kidney and a reservoir of nutrientmedium, the reservoir remained sterile for atleast 24 h. When the differential dialysis systemwas employed, the dialysate solutions in thebeaker dialyzers also remained free of the bacte-ria. Moreover, during the hemodialysis culturetrials bacteria were not detected by routineaerobic and anaerobic culturing of blood sam-ples.Molecular size of the dialyzable toxic

material. The results with the beaker dialyzersenabled an estimation of the maximal molecu-lar size of the dialyzable toxic material relativeto the molecular weight of proteins that report-edly just diffuse through or are just excluded bythe membrane in the artificial kidney andMW-10,000 beaker dialyzer (Bulletin no.175-1187-71, Dow Chemical Co.). Of the pro-

teins reported, cytochrome c (presumably ofbovine heart origin, molecular weight 13,370[13 ]) lies closest within the dialysis threshold ofthe membrane. An equivalent Einstein-Stokeshydrodynamic radius (rES) of 1.88 nm was

calculated, based on a diffusion coefficient of11.4 x 107 (13). Myoglobin (presumably of horseheart origin, molecular weight 16,890 [13]) liesjust outside the dialysis threshold of the mem-brane, from which an rER of 1.90 nm was

calculated based on a diffusion coefficient of11.3 x 107 (13). Consequently, the size of a

molecule just able to dialyze is intermediate inequivalent size between these rigid globular

A

B

6 7 8 9 10 11

INFECT. IMMUNITY556

2

1

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

HEMODIALYSIS CULTURE OF S. MARCESCENS

proteins, i.e., a molecular weight of 15,000 andan rEs of 1.89 nm.The dialyzable toxic material from S.

marcescens might be more fibrous and flexiblethan globular and rigid in its molecular configu-ration. Consequently, its size might be betterrelated to the dialysis of a molecule like that of aPEG which in solution coils randomly andloosely into a form that is large relative to itsmolecular weight. Scherrer and Gerhardt (10),by using polyglycol samples, have devised adistribution analysis method to determine theporosity threshold of bacterial cell walls. By useof this method, the dialysis threshold of theartificial kidney membrane was successfullydetermined with a polydisperse PEG sample ofMn 3,350, whereas smaller (Mn 1,540) and larger(Mn 9,500) polyglycol samples proved unsatis-factory. The results (Fig. 8) showed that the sizeof a molecule just able to dialyze, determinedfrom the intersection point of the two distribu-tion curves, was equivalent to a quasi-monodis-perse PEG with a Mn of 5,500 and an rEs of 2.6nm.Thus, the maximal molecular size of the

dialyzable toxic material from S. marcescenswas defined relative to two different types ofmolecules. If its molecular configuration waslike a rigid globular protein, the toxic materialwas less than a molecular weight of 15,000 andan rES of 1.9 nm. If it was like a flexible fibrouspolyglycol, it was less than a molecular weightof 5,500 and an rEs of 2.6 nm.

DISCUSSIONThe results demonstrated the feasibility of

employing hemodialysis and a goat-artificialkidney-fermentor system to grow organisms.Selection of the goat and S. marcescens as thehost-parasite model was made less for intrinsic

103 MOLECULAR WEIGHT, Mn1 2 3 4 5 6 7 S 9 lo

0-3E

, 2

g,

60 55 50 45 40 35 30 25

ELUTION FRACTION NO.

FIG. 8. Molecular weight distribution of PEG Mn3,350 before (0) and after (0) equilibrium dialysisthrough the artificial-kidney membrane. The inter-section point at Mn 5,500 represents the dialysisthreshold. VO indicates the void volume of the column.

interest and more for convenience in developingthe new system and studying its operation. Inits potential, this prototype hemodialysis cul-ture system appears to have broader applicabil-ity, but within certain limitations. Animalsother than the goat appear adaptable, providedthat they are sufficiently large and can be heldquietly. Organisms other than S. marcescens,including ones having more complex nutritionalrequirements, appear capable of culture pro-vided that they have a comparably short gener-ation time so as to multiply adequately withinthe period (e.g., about 15 h) that an animal canbe held quietly and attended constantly. Themilieu in the dialysate appeared to be much thesame as that in the blood to the extent thatsmall dialyzable molecules are involved, butmacromolecules such as serum proteins do notpermeate the most porous membrane used, andit is questionable whether dialysis exchange ofCO2 and 02 is sufficiently rapid. Consequently,the in vivo milieu of the blood was only partiallyduplicated. This was evidenced when B.anthracis grew in the system asporogenously asin vivo, but with the formation of long chains asin vitro. The hemodialysis culture system ap-pears especially useful for examining the effectsof small microbial products on a host, and ofsmall host molecules on a parasite population,in a situation where the microbial cells canattain a mass comparable to that in severeinfections and produce relatively large amountsof products. The system also appears promisingfor studying the effects of antibiotic drugs onmicrobial cultures.The results also demonstrated methods for

sampling culture products by differential dial-ysis and for accurately determining the maxi-mal size of a dialyzable product relative todifferent molecular configurations, withoutprior chemical extraction or treatment. Thesemethods probably could be extended to mem-branes of coarser porosity and molecules oflarger size and appear useful especially for thestudy of polydisperse products.The nature of the dialyzable toxic material

demonstrated by goat hemodialysis culture ofS. marcescens is unknown. One possibility wasthat of an endotoxin, which is a known productof the organism and was suggested by the fever,leukopenia, and general physiological responseof the animal. Native endotoxin usually isconsidered to consist of a polydisperse mac-romolecular complex of lipid, polysaccharide,and protein with a nominal molecular weight ofabout 105 to 10', and molecules of this size couldnot permeate the membranes of any of thedialyzers used in this study. The smallest frag-

c Z, . . . I ..

rl"%I

10I VO0I

1 1111

? tI

%

ell It 11 .11 . 11 .(% 'A 11 'A () 2 11

VOL. 9, 1974 557

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

QUARLES ET AL.

ment of endotoxin that yields some or all of thetypical biological reactions is undefined (14,15), but the state of dispersion of endotoxinfractions is known to affect certain biologicalreactions (7, 15). Continuous hemodialysis con-ceivably might allow the passage of small activefragments or of small inactive ones which thencould aggregate or become adsorbed to bloodconstituents and thereby acquire endotoxic-likeactivity.The dialyzable toxic material from S.

marcescens might also be attributed to anordinary metabolic product that is deleteriousto the host. An example of this situation occursamong the complex of toxins produced byVibrio cholerae, in which the dialyzable andheat-stable type 3 toxin eventually was identi-fied as ammonia (1). However, a similar situa-tion with the dialyzable material from S.marcescens seems unlikely because of the rangeand severity of host reactions and because theeffects were less marked under anaerobic cul-ture conditions where the accumulation of in-termediary metabolic products would be great-est.

Finally, a toxin different from any of theabove might account for the dialyzable toxicmaterial from S. marcescens. This possibilitydeserves consideration in view of emerging clini-cal and experimental evidence that factorsother than endotoxin may be more critical thanendotoxin in determining lethal outcome ofsevere sepsis due to gram-negative organisms

ACKNOWLEDGMENTSWe gratefully thank Josephine Belding for surgical assist-

ance and T. C. Belding for surgical anaesthesia.This investigation was supported by Public Health Service

research grant AI-09760 from the National Institute of Allergyand Infectious Diseases, a general research support grant, andtraining grant GM-01911 from the National Institute ofGeneral Medical Sciences.

LITERATURE CITED1. Burrows, W. 1968. Cholera toxins. Annu. Rev. Microbiol.

22:245-268.

2. Fletcher, W. S., A. L. Rogers, and S. S. Donaldson. 1964.The use of the goat as an experimental animal. Lab.Animal Care 14:65-90.

3. Gallup, D. M., and P. Gerhardt. 1963. Dialysis fermentorsystem for concentrated culture of microorganisms.Appl. Microbiol. 11:506-512.

4. Gerhardt, P., and D. M. Gallup. 1963. Dialysis flask forconcentrated culture of microorganisms. J. Bacteriol.86:919-929.

5. Kass, E. H., P. J. Porter, M. W. McGill, and E. Vivaldi.1973. Clinical and experimental observations on thesignificance of endotoxemia. J. Infect. Dis.128:S299-S302.

6. Metchnikoff, E., E. Roux, and T. (Aurelli-)Salimbeni.1896. Toxine et antitoxine cholerique. Ann. Inst. Pas-teur 10:257-282.

7. Milner, K. C., and R. A. Finkelstein. 1966. Bioassay ofendotoxin: correlation between pyrogenicity for rabbitsand lethality for chick embryo. J. Infect. Dis.116:529-536.

8. Quinton, W., D. Dillard, J. J. Cole, and B. H. Scribner.1961. Possible improvements in the technique of longterm cannulation of blood vessels. Trans. Amer. Soc.Artif. Internal Organs 7:60-63.

9. Quinton, W. D., D. Dillard, and B. H. Scribner. 1960.Cannulation of blood vessels for prolonged hemodial-ysis. Trans. Amer. Soc. Artif. Internal Organs6:104-113.

10. Scherrer, R., and P. Gerhardt. 1971. Molecular sieving bythe Bacillus megaterium cell wall and protoplast. J.Bacteriol. 107:718-735.

11. Schultz, J. S., and P. Gerhardt. 1969. Dialysis culture ofmicroorganisms: design, theory, and results. Bacteriol.Rev. 33:1-47.

12. Smith, C. G., and M. J. Johnson. 1954. Aeration require-ments for growth of aerobic microorganisms. J. Bacte-riol. 68:346-350.

13. Smith, M. H. 1970. Molecular weights of proteins andsome other materials including sedimentation, diffu-sion and frictional coefficients and partial specificvolumes. In H. A. Sober (ed.), Handbook of biochemis-try. Selected data for molecular biology. ChemicalRubber Co., Cleveland, Ohio.

14. Wober, W., and P. Alaupovic. 1971. Studies on theprotein moiety of endotoxin from gram-negative bacte-ria. Characterization of the protein moiety isolated byphenol treatment of endotoxin from Serratiamarcescens 08 and Escherichia coli 0 141: K85(B).Eur. J. Biochem. 19:340-356.

15. Wober, W., and P. Alaupovic 1971. Studies on theprotein moiety of endotoxin from gram-negative bacte-ria. Characterization of the protein moiety isolated byacetic acid hydrolysis of endotoxin from Serratiamarcescens 08. Eur. J. Biochem. 19:357-367.

558 INFECT. IMMUNITY

on January 4, 2021 by guesthttp://iai.asm

.org/D

ownloaded from