Vol. 51, No. 4APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Apr. 1986, p. 813-8200099-2240/86/040813-08$02.00/0Copyright X) 1986, American Society for Microbiology

Removal of Endotoxin from Water by Microfiltration through a

Microporous Polyethylene Hollow-Fiber MembraneYOSUKE SAWADA,1t REIKO FUJII,1 IKUO IGAMI,1 ATSUSHI KAWAI,1 TERUO KAMIKI,2

AND MAKOTO NIWA3*Product and Development Center, Mitsubishi Rayon Co., Ltd., Nagoya 461,1 Public Health Research Institute of Kobe

City, Kobe 650,2 Department of Bacteriology, Osaka City-University Medical School, Osaka 545,3 Japan

Received 26 June 1985/Accepted 21 January 1986

The microporous polyethylene hollow-fiber membrane has a unique microfibrile structure throughout itsdepth and has been found to possess the functions of filtration and adsorption of endotoxin in water. Themembrane has a maximum pore diameter of approximately 0.04 ,um, a diameter which is within the range ofmicrofiltration. Approximately 10 and 20% of the endotoxin in tap water and subterranean water, respectively,was smaller than 0.025 ,um. Endotoxin in these water sources was efficiently removed by the microporouspolyethylene hollow-fiber membrane. Escherichia coli 0113 culture broth contained 26.4% of endotoxinsmaller than 0.025 ,um which was also removed. Endotoxin was leaked into the filtrate only when endotoxinsamples were successively passed through the membrane. These results indicate that endotoxin smaller than thepore size of the membrane was adsorbed and then leaked into the filtrate because of a reduction in binding sites.Dissociation of 3H-labeled endotoxin from the membrane was performed, resulting in the removal of endotoxinassociated with the membrane by alcoholic alkali at 78% efficiency.

The removal of endotoxin (lipopolysaccharide [LPS]), amajor pyrogen, from water or parenteral substances is ofmajor importance for manufacturing processes and in qualitycontrol. Although many methods (4, 8, 10, 14-19, 21, 29, 31,32; N. S. Harris and R. Feinstein, U.S. patent 3,944,391,March 1976) for removing endotoxin have been used, idealand reliable methods are still needed. The recent progress ofmembrane technology led to an examination of the potentialuse of molecular filtration membranes such as those used inreverse osmosis (18) or ultrafiltration (3, 20, 29, 32) for watersamples. However, these methods encountered the difficultyof low water flow rates.The LPS monomer is an amphiphile with a large

hydrophilic polysaccharide chain and a hydrophobic fattyacid tail. The LPS monomer can be found in various aggre-gate forms (7, 35). LPS in large forms can be converted intosmaller forms under various conditions (13, 22, 27). Theconverted forms may then be conjugated to other com-pounds to form hybrids (23, 26). Such aggregation anddisaggregation of LPS brings about bewildering and incon-sistent results in the removal of endotoxin by filtration.Therefore, we focused on the removal of endotoxin smallerthan the pore size of the membrane by adsorption.The microporous polyethylene hollow fiber (EHF;

Mitsubishi Rayon Co., Tokyo, Japan) (M. Shindo, T.Yamamoto, 0. Fukunaga, and H. Yamamori, U.S. patent4,401,567, August 1983) was developed into a water purifi-cation device (11) which was found to remove variousbacterial cells as well as authentic LPS (Escherichia coliO111:B4; Difco Laboratories, Detroit, Mich.) from water. Inaddition, it was previously found that an adsorbent preparedfrom this EHF had the ability to adsorb LPS in water (12).Therefore, the EHF membrane appears to utilize a combi-nation of adsorption and filtration of endotoxin in waterwhen used as a filter membrane. To clarify the mechanisms

* Corresponding author.t Present address: Bristol-Myers Research Institute, Tokyo 153,

Japan.

of endotoxin removal, we examined the contributions of thefunctions of adsorption and filtration of endotoxin. This wasdone by examining the following: (i) the maximum pore sizeof the EHF membrane; (ii) the removal of endotoxin fromvarious water samples by the EHF membrane; (iii) theresults of an analysis of the size distribution of endotoxinactivity; and (iv) the results of the dissociation of 3H-labeledLPS from the EHF membrane.

MATERIALS AND METHODS

EHF and module. The EHF was produced from high-density polyethylene as described previously (Shindo et al.,patent 4,401,567, August 1983). Figure 1 shows a scanningelectron micrograph (JSM 35C electron microscope; NihonDenshi Co., Tokyo, Japan) of the inner surface of the EHFmembrane (code EHF 390 C). The outer surface was virtuallythe same (data not shown). The EHF membrane has goodmechanical strength and chemical stability and does notcontain any fillers or plasticizers. Its specifications are asfollows: outer diameter, 390 ,um; membrane wall thickness,55 ,xm; porosity, 63% (34); bubble point, 4.85 + 0.13 kg/cm2in 99% ethanol (1); water flux, 3.6 liters/min per m2 ofmembrane surface area at a pressure of 0.5 kg/cm2 at 25°C;and nitrogen-adsorbing surface area, 31.7 m2/g (2). A modulewas formed with 128 filaments of the EHF membrane, apolycarbonate jacket, and a potting agent of polyurethane(Nippon Polyurethane Co., Tokyo, Japan) (Fig. 2). The totalouter surface area of the EHF membrane was approximately70 cm2/116 mg of the module. The module was confirmed tohave no sealing insufficiency by feeding water at a pressureof 1 kg/cm2 for 1 min. Sealing was confirmed to be sufficientunless water leaked through. To make the module permeableto water, 30 ml of 99% ethanol was passed through themembrane, which was then rinsed with 50 ml ofnonpyrogenicwater (Otsuka Pharmaceutical Co., Naruto, Japan). Theprewashed module was used without being dried.

Pore characteristics of the EHF membrane. A suspensionof beads in 40 ml of water (0.1% [wt/vol]) was applied to theEHF membrane in the module with a Mini-Pump (TMP 1OH;

813

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

814 SAWADA ET AL.

FIG. 1. Scanning electron micrograph of the inner surface of theEHF membrane. Bar, 1 ,um.

Toyo Kagaku Sangyo Co., Tokyo, Japan) at a flow rate of 8ml/min (Fig. 2). Air in the module was vented out throughtwo air outlets. A 2-ml fraction was collected. The maximumconcentration of the beads eluted in the filtrate was inter-preted as a rejection of the EHF membrane and expressed asa percentage of the total concentration. Each value repre-sented the mean of three or four determinations. Uniformbeads, colloidal silica, and polystyrene in monodispersedform were obtained from Maruwa Bussan Co., Tokyo, Japan(originally from E. I. du Pont de Nemours & Co., Inc.,Wilmington, Del.), Dow Chemical Co., Indianapolis, Ind.,and Shokubai Chemical Co., Tokyo, Japan, respectively.For assessment of the concentration of beads being used, theA210 to A290 (different wavelengths were used depending onthe type and size of the beads) was measured with aShimadzu UV-250 spectrophotometer attached to an OPI-2recording unit (Kyoto, Japan). A standard concentrationcurve was prepared by diluting the beads with distilledwater.Treatment of water samples. Water samples obtained in

Japan were immediately used either for determination ofendotoxin activity or for filtration treatment. A sample fromPhiladelphia, Pa., was frozen and airmailed. The water fluxin the module was adjusted to 0.5 ml/min at 25°C. The filtrateoutlet of the module was wiped with cotton saturated in 70%ethanol before sample collection.

Analysis of size distribution of endotoxin activity. Variousfilter membranes with nominal pores were purchased from

Millipore Corp., Bedford, Mass., and Nuclepore Corp.,Pleasanton, Calif. The membranes were extensively washedbefore use with nonpyrogenic water, and water drops on themembrane surface were removed by shaking the membrane.The membrane was set in an adaptor for an injection syringe.The water sample in the syringe was forced to pass throughthe membrane. The first drops were discarded. A 2-mlfiltrate sample was collected, and the endotoxin activity wasdetermined. Preheating of the water sample was done in awater bath for 30 min.Determination of endotoxin activity. The Limulus

amoebocyte lysate-based assay kits Pyrodick (assay A) andToxicolor (assay B), purchased from Seikagaku Kogyo Co.,Tokyo, Japan, were used. These assays have been describedpreviously (5, 6, 9). Incubation was done at 37°C for 30 min.The wavelengths were read at 405 nm for assay A and at 545nm for assay B. All data represent the mean of duplicatedeterminations and were expressed in amounts equivalent topurified LPS of E. coli O111:B4 (36). All glassware wasdepyrogenated by being heated at 250°C for 2 h.Nonpyrogenic water was used both for the dilution of watersamples, if necessary, and for the controls.Pyrogen tests on rabbits. Pyrogen tests were carried out in

accordance with the Japan Pharmacopeia X. Three rabbitswere used for each sample, and their body temperatureswere measured at 1 h after intravenous injection. Concen-tration of the endotoxin in tap water was done with a rotaryevaporator in vacuo at 30 to 35°C.Column chromatography with Sephadex G-75. A column

(1.6 by 20 cm) of Sephadex G-75 (Pharmacia Japan, Tokyo)was washed extensively with nonpyrogenic water. The flowrate was adjusted to 1 mllmin. A 2-ml fraction was collected.

Bacterial strains and their cultivation. E. coli and Salmo-nella minnesota were from our stock and were grown in

*- -A~

1 cm

Air outlet

Hollow fiber6 cm

Air outlet

Sample out

Modu 1 e Mi n i-Pump SampleFIG. 2. Schematic diagram of the filtration process.

APPL. ENVIRON. MICROBIOL.

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

NEW METHOD FOR REMOVING ENDOTOXIN FROM WATER

1 00

o 0

(U

ce

0

0.01 cSi2

FIG. 3. Plot of rejection versus I

Trypticase soy broth (BBL Microbiology Systems, Cockeys-ville, Md.) at 37°C for 18 h. For determination of viable cells,the cultures were diluted with sterilized physiological saline(Otsuka Pharmaceutical Co.). Viable cells were counted inheart infusion agar (Eiken Kagaku Co., Tokyo, Japan) aftercultivation at 37°C for 48 h. All values represent the mean oftriplicate determinations.

Preparation of [3H]LPS. [3H]LPS was prepared from theLPS isolated from S. minnesota R595 by 3H-labeled gasexposure (New England Nuclear Corp., Boston, Mass.) (30).The specific radioactivity was 1.14 mCi/mg (3.0 mCi/ml ofwater). A mixture of cold LPS (1.0 mg) and [3H]LPS solution(30 ,ul) in water (10 ml) was dispersed with a cell disruptor(Sonifier cell disruptor 200; Branson Sonic Power Co.,Danbury, Conn.) at 20 kc and 20 W for 3 min on an ice bath.The sonicated solution (2-ml aliquots) was applied to acolumn of Sephadex G-75. The column was eluted withnonpyrogenic water. A 2-ml fraction was collected. The 3Hin 100-pI samples of the filtrate was measured with a counter(1215 Rackbeta; LKB-Wallac, Turku, Finland) and an aque-ous counting scintillant (ACS II; Amersham Corp.,Arlington Heights, Ill.). Fractions 6 and 7 were combined,LPS (76%) was recovered by measuring the A214, and the[3H]LPS (26%) was recovered from the total 3H. Radioac-tivity in the latter fractions from the column was discarded.The specific radioactivity of the [3H]LPS was 11.84 ,uCi/mg.

Preparation of the 3H-labeled EHF membrane. The purified[3H]LPS solutions were combined, and 10 ml of this solutionwas loaded onto the EHF membrane in the module andsubsequently washed with distilled water (10 ml) at a flowrate of 1 ml/min. The module was broken, and the EHFmembrane was removed for further study.

).

ze(parti

)2 0.03 0.04 0.05 0.1

,Pm)ticle size for the EHF membrane.

Dissociation of 3H from the [3H]LPS-adsorbed EHF mem-brane. The [3H]LPS-adsorbed EHF membrane was cut into1-cm-long pieces. The cut fiber pieces (60 mg) were sus-pended in 10 ml of solvent and gently stirred at 35°C. A200-p,u sample of the solvent was removed, and the 3H wasmeasured. The 3H remaining in the fibers after 120 min wasalso determined by measuring the whole fiber pieces.

RESULTSMaximum pore size of the EHF membrane. A determina-

tion of the maximum pore size of the EHF membrane iscrucial in perceiving the removal of contaminant materialsby filtration. Because the EHF membrane used for themodule (Fig. 2) has lattice-like pores on its surfaces (Fig. 1)

TABLE 1. Endotoxin activity in various water samples beforeand after treatment with the EHF membranea

Endotoxin concn (ng/ml)b

Water Location Before filtration, After filtrationassay A Assay A Assay B

Tap Tokyo, Japan 1.98 <0.01 0.004Tap Osaka, Japan 9.00 <0.01 0.003Tap Nagoya, Japan 5.20 <0.01 0.004Tap Philadelphia, Pa. 7.29 <0.01 0.004Subterranean Nagoya, Japan 2.19 <0.01 0.003Distilledc 0.30 <0.01 0.003

a Sampling and assaying were performed in the summer of 1983.b A 50-m filtrate was collected, and endotoxin activity was assayed.' Freshly distilled water made in a glass apparatus was stored in a plastic

bottle for a few days at room temperature.

VOL. 51, 1986 815

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

APPL. ENVIRON. MICROBIOL.

1 00

50.4J

'4-)

C

0100

tol

5-

0. 025 0. 05 0. 1 0. 2 0. 8Pore si ze of membrane( Jm)

FIG. 4. Size distribution of endotoxin activity in tap water (A) and subterranean water (B). The endotoxin concentration in tap water was3.57 ng/ml, and that in subterranean water was 0.42 ng/ml. Symbols: *0, 15°C; x, 38°C; O, 50°C; A, 70°C.

and a tortuous fiber maze throughout its depth, like theCelgard membrane (24), the pore size was studied. Filtrationof spherical particles of a uniform diameter has been wellstudied (25, 28). The EHF membrane was estimated to havea maximum pore size (100% rejection by extrapolation) ofapproximately 0.04 ,um (Fig. 3). Therefore, the maximumpore size of the EHF membrane is of microfiltration grade,i.e., a pore size ranging from 0.02 to 10 ,um (20).Removal of endotoxin from various water samples. Munic-

ipal tap water contains a mixture of various sources ofendotoxin. Although tap water samples in both Japan andthe United States contain nanogram levels of endotoxinactivity, the filtrate from the EHF membrane had an activityof less than 10 pg/ml, as determined by assay A. Todetermine low levels of endotoxin activity, we used amodified assay system, assay B. Because p-nitroaniline

hydrolyzed from a substrate is converted to a red diazoderivative, assay B is 10-fold more sensitive than assay A.All filtrates contained endotoxin activity at 3 to 4 pg/ml.Even subterranean water and single-distilled water for labo-ratory use contained high endotoxin activity. Filtrates ofthese water samples contained endotoxin activity at 3 pglml(Table 1).

Elimination of factors in water filtrates which affect theendotoxin assay system. To eliminate possible contaminationfactors in filtrates which affect the endotoxin assay system,a filtrate from the module containing nonpyrogenic waterwas added to various concentrations of a reference LPS (E.coli O111:B4). Virtually no effect was observed on thestandard curve of the reference LPS in assay A for thefiltrate (data not shown).

Analysis of the size distribution of endotoxin activity in two

816 SAWADA ET AL.

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

NEW METHOD FOR REMOVING ENDOTOXIN FROM WATER

TABLE 2. Concentrations of endotoxin in filtrates from tap andsubterranean water

Endotoxin (ng/ml)a in:Filtrate vol (liters)

Tap water Subterranean water

Unfiltered sample 3.57 0.420.1 <0.01 <0.011 <0.01 <0.015 0.02 <0.0110 0.08 <0.0113 0.10 <0.01a The total organic concentrations in tap water and subterranean water

were 1.7 and <0.1 mg/liter, respectively.

water sources. Tap water from Nagoya City was filteredthrough various membranes with nominal pore sizes atseveral temperatures. Four filtrates at various temperatureshad similar profiles of size distribution. Endotoxin smallerthan 0.025 ,um (designated here as small-sized endotoxin)represented approximately 10% of the total (Fig. 4A). An-other sample, subterranean water from Nagoya City, had aprofile of size distribution at 15°C similar to that of the tapwater, with a small-sized endotoxin concentration of approx-imately 20%o (Fig. 4B).Removal of the endotoxin activity from two water sources.

To remove the endotoxin activity from the tap water andsubterranean water samples, 13 liters of water from eachsource was passed through each EHF membrane module,

ng/ml0.31

4-

t 0.2._-

x4 0. 1to~0

C

0

P 0.2

*_

x04- 0. 10

0

A

and the endotoxin activity in the filtrates was determined.For tap water, the endotoxin activity appeared gradually inthe last few fractions. For subterranean water, almost all ofthe endotoxin activity was removed from 13 liters (Table 2).

Size of endotoxin in tap water samples. The endotoxinactivity in tap water was analyzed by column chromatogra-phy with Sephadex G-75. Two peaks appeared, with a totalrecovery of 86% (Fig. 5A). The first peak was at 68% of theeluate. Analysis of the size distribution of both peaks wasperformed by using a membrane with a nominal pore size of0.025 ,um. Approximately 5% of the endotoxin of the firstpeak and approximately 95% of that of the second peakpassed through the membrane. In addition, the endotoxinactivity in the 13-liter tap water filtrate (Table 2) wasanalyzed. The peak fraction of activity coincided with thesecond peak fraction of activity in tap water (Fig. 5B).The 13-liter tap water filtrate (Table 2) was filtered through

various membranes with different nominal pore sizes, andapproximately 90% of the endotoxin in the filtrate was foundto be smaller than 0.05 ,um (data not shown).Pyrogen tests on rabbits. To determine whether the

endotoxin activity in tap water causes a rise in body temper-ature, pyrogen tests on rabbits were carried out. To obtain ahigh concentration of pyrogen activity, a tap water samplecontaining low levels of endotoxin activity was chosen. Thetap water of Nagoya City (endotoxin activity, 0.6 ng/ml) wasfirst evaporated to about 1/70th its original volume. Theconcentrated sample had an endotoxin activity of 43.4 ng/ml.As a result, the concentrations of and increases in endotoxinactivity were in good correlation. A part of the concentrated

0.3t B

10Fraction Number

20 30

FIG. 5. Chromatographic profile of endotoxin activity in tap water. (A) Tap water (endotoxin activity, 2.4 ng/2 ml) analyzed on a SephadexG-75 column. (B) Analysis of the fractions corresponding to 0.1-liter filtrates (x) and 13-liter filtrates (0) in Table 2.

817VOL. 51, 1986

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

818 SAWADA ET AL.

41

0

S-to

S-a)

+3

E

0o

39 40

a -am a a 22 3 - 1 0 1 2 3 - 1 0 1 2 3

Time after injection(hr)FIG. 6. Changes in the body temperature of rabbits after intravenous sample injection. Bars indicate the standard deviation. Arrows

indicate the point of sample injection. Samples: 1, tap water (0.6 ng/ml); 2, cQncentrated endotoxin (43.4 ng/ml); 3, E. coli 0111:B4 LPS (56ng/ml); and 4, filtrate of sample 2 (<10 pg/ml).

sample was then treated with the EHF membrane. Thefiltrate contained an endotoxin activity of less than 10 pg/ml.Although the tap water showed no pyrogen activity inrabbits, the concentrated endotoxin caused an increase in

TABLE 3. Simultaneous removal of cultured bacterial cells andtheir endotoxin activitya

No. of viable cells per Endotoxin concnml (106) (ng/ml)

BacteriumBefore After Before After

filtration filtration filtration filtration

E. coli 0113 3.0 0 18.63 <0.01E. coli 0111:B4 4.5 0 8.53 <0.01S. minnesota R595 2.0 0 0.18 <0.01

a A 50-mi filtrate was collected, and endotoxin activity and viable cells weredetermined.

body temperature of 1.5°C at 2 h after intravenous injection.However, filtration of the concentrated endotoxin elimninatedall the pyrogenic response from the rabbits. In a parallelexperiment, reference LPS from E. coli 0111:B4 at 56 ng/mlraised the body temperattire by 1.65°C (Fig. 6). Thus,endotoxin activity in a tap water sample and increasedtemperature showed a good correlation.

Simultaneous removal of cultured bacterial cells and theirendotoxin activity. Gram-negative bacteria produce LPS ontheir cell surfaces and liberate it into the surrounding me-dium under certain environmental conditions. To remove thebacterial cells and LPS, two strains of E. coli and one strainof S. minnesota were grown. Diluted cultures were filteredthrough the EHF membrane module. All filtrates containedno viable cells and had undetectable endotoxin activity(Table 3). A profile of the endotoxin activity in the dilutedcultures was obtained by column chromatography withSephadex G-75. Three diluted cultures showed only a peakcorresponding to the first peak shown in Fig. 5A; no peak

0--

APPL. ENVIRON. MICROBIOL.

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

NEW METHOD FOR REMOVING ENDOTOXIN FROM WATER

corresponding to the second peak was observed (data notshown).Next, all cultures were filtered through membranes with a

pore size of 0.025 ,um. The filtrate from the E. coli 0113culture contained the highest small-sized endotoxin activity(26.4%), whereas the filtrates from the E. coli 0111:B4 andS. minnesota R595 cultures contained only 0.2 and 1.0% ofthe small-sized endotoxin activity, respectively (Table 4).

Adsorption and dissociation of the [3H]LPS. To determinethe dissociation of endotoxin from the EHF membrane,[3H]LPS prepared from LPS isolated from S. minnesotaR595 was passed through the EHF membrane. An attemptwas made to regenerate the [3H]LPS-adsorbed membrane byrinsing it with certain solvents. Water as a solvent dissoci-ated 5.6% of the 3H. Ethanol at a 70% concentrationremoved 43% of the 3H from the EHF membrane, andethanolic alkali removed 78% of the 3H (Table 5).

DISCUSSIONMost filter membranes have been designed to remove

contaminant materials by having a pore size that preventsthe contaminant materials from flowing through the mem-brane. The removal efficiency and the accuracy of the filtermembrane are directly influenced by the following: (i) themembrane structure, e.g., a top skin layer and a poroussublayer; (ii) the water quality; (iii) the flow rate; and (iv)other variables. The EHF membrane has lattice-like poreson its surface which form a tortuous fiber maze throughoutthe depth of the membrane. This structure is similar to thatof Celgard membrane, which is made from polypropylene(24). Although the pore size of the EHF membrane seemslarge (Fig. 1), the membrane has been experimentally fouidto have a maximum pore size of 0.04 ,um. Therefore, theEHF membrane can remove contaminant materials largerthan 0.04 ,um by the filtration process. The endotoxinactivity in tap water was analyzed by column chromatogra-phy with Sephadex G-75 and showed two peaks. The pro-portions of the activity in the first peak and the second peakwere 68 and 32%, respectively (Fig. 5A). Approximately95% of the second peak contained endotoxin smaller than0.025 p,m. The small-sized endotoxin corresponding to thesecond peak was experimentally found to be removed fromwater samples by the EHF membrane (Table 2). Further-more, endotoxin activity gradually appeared in the filtratesafter the successive feeding of endotoxin, indicating thatsmall-sized endotoxin passed through the membrane be-cause of a reduction in adsorption sites. We have alreadydemonstrated the adsorption characteristics of EHF forendotoxin and tetrachloroethylene (12). The organic com-pounds adsorbed were, in turn, well dissociated from themembrane after an ethanol rinse (11). To evaluate the

TABLE 4. Size distribution of endotoxin activity in culturesaEndotoxin activity (ng/ml) in:

Bacterium FiltrateDiluted culture (from 0.025-,um

membrane)

E. coli 0113 252.2 66.7E. coli O111:B4 309.8 0.6S. minnesota R595 20.3 0.2

a Bacterial cells were centrifuged (1,000 x g, 10 min), and the supernatantwas removed. The supernatant was diluted 10 times with nonpyrogenic water.Endotoxin activity in either diluted cultures or filtrates was determined byassay A.

TABLE 5. Dissociation of 3H from the [3H]LPS-adsorbed EHFmembrane

Dura- Radioactivity (cpm) in:tion Total %

Rinsing agent of radio- Dissoci-treat- Solution Fiber activity ationment (cpm)(min)

Water 0 2.0 x 103 0.260 4.0 x 104120 5.0 x 104 8.4 x 105 8.9 x 105 5.6

Ethanol (70%) 60 2.9 x 105120 3.1 x 105 4.1 x 105 7.2 x 105 43

Ethanol (70%)- 60 5.5 x 105NaOH 120 6.0 x 105 7.7 x 105 1.37 x 106 78(0.1 N)

dissociation of endotoxin from the EHF membrane,[3H]LPS was used. Alkali is known to destroy the endotoxinmolecule (15). Ethanol releases the hydrophobic interaction.Therefore, alcoholic alkali is an ideal solvent for regenera-tion of the EHF membrane. During use of the EHF mem-brane for tap water, rinsing of the membrane with 70%ethanol-0.1 N NaOH caused rejuvenation of the membrane.As for water quality, not only the concentration of endotoxinbut also the concentrations of other organic compounds inwater have an effect. The efficiency of removal of endotoxinmay be affected because other organic compounds can beadsorbed on the same binding sites as is endotoxin. The totalconcentrations of organic compounds in tap water andsubterranean water were 1.7 mg/liter and less than 0.1mg/liter, respectively. The greater efficiency in removingendotoxin from subterranean water seemed to be attribut-able to the quantity of the contaminated solids (Table 2). Thehydrophobic compounds were more efficiently bound to themembrane (unpublished data). Further comparative adsorp-tion studies are currently under way. Since the adsorbingcapacity of an adsorbent for endotoxin made from this EHFresulted in an insignificant difference at a pH range of 5.1 to9.1 (unpublished data), the effect of pH on adsorption in thefiltration process may not exist in this pH range.A confident analysis of endotoxin activity is of utmost

importance in solving the problems associated withendotoxin. Based on the clarification of the biochemicalmechanism of gelation of Limulus amoebocyte lysate,Iwanaga and his collaborators (9) recently developed asensitive and quantitative method for assaying endotoxin.This method, the synthetic chromogenic substrate method,is recommended for its potential application to water, par-enteral substances, and blood samples. It must be noted thatthis assay system only detects the active forms of endotoxin(not the amount of it) and other reactive substances, such as,-glucan and its analogs. Recently, various water sampleswere tested by using this assay system, and the reactivesubstance in tap water samples was interpreted to beendotoxin (33, 37).Because the EHF membrane is hydrophobic, it has been

found to possess interesting characteristics, e.g., the adsorp-tion of organic compounds, including endotoxin. The EHFmembrane has good mechanical strength in addition to someadvantages of hollow-fiber-type filters over other membraneconfigurations (32). Therefore, the EHF membrane can beused to remove endotoxin from water used in laboratories,

VOL. 51, 1986 819

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from

APPL. ENVIRON. MICROBIOL.

pharmaceutical houses, and certain health care environ-ments such as dialysis centers.

ACKNOWLEDGMENTThis work was supported in part by a Grant-in-Aid from the

Ministry of Science, Education and Culture of Japan.

LITERATURE CITED1. American National Standards Institute. 1970. Standard method

of testing for pore size characteristics of membrane filters foruse with aerospace fluids. ASTM designation F316-70. Annualbook of ASTM standards, part 23. American National Stan-dards Institute, New York.

2. Bohra, J. N. 1984. Characterization of the pore structure ofcarbonized viscose rayon yarn: analysis of adsorption isothermsof nitrogen and carbon dioxide. Fibre Sci. Technol. 21:135-147.

3. Cradock, J. C., L. A. Guder, D. L. Francis, and S. L. Morgan.1978. Reduction of pyrogens-application of molecular filtra-tion. J. Pharm. Pharmacol. 30:198-199.

4. Ditter, B., R. Urbaschek, and B. Urbaschek. 1983. Ability ofvarious adsorbents to bind endotoxins in vitro and to preventorally induced endotoxemia in mice. Gastroenterology84:1547-1552.

5. Fujita, Y., and C. Nakahara. 1982. Preparation and applicationof a new endotoxin determination kit, Pyrodick®, using achromogenic substrate. Prog. Clin. Biol. Res. 93:173-182.

6. Hakogi, E., Y. Shimada, T. Kume, and K. Tabuchi. 1984/1985.Perchloric acid treatment and use of chromogenic substrate inthe Limulus test: application to veterinary diagnosis. Vet.Microbiol. 10:33-42.

7. Hannecart-Pokorni, E., D. Dekegel, and F. Dupuydt. 1973.Macromolecular structure of lipopolysaccharides from gram-negative bacteria. Eur. J. Biochem. 38:6-13.

8. Hou, K., C. P. Gerba, S. M. Goyal, and K. S. Zerba. 1980.Capture of latex beads, bacteria, endotoxin, and viruses bycharge-modified filters. Appl. Environ. Microbiol. 40:892-896.

9. Iwanaga, S., T. Morita, T. Harada, S. Nakamura, M. Niwa, K.Takada, T. Kimura, and S. Sakakibara. 1978. Chromogenicsubstrates for horseshoe crab clotting enzyme. Its applicationfor the assay of bacterial endotoxins. Haemostasis 7:183-188.

10. Kaden, H. 1975. The use of asbestos filter beds in the productionof sterile and pyrogen-free solutions. Pharmazie 29:752-763.

11. Kamiki, T., A. Kawai, I. Igami, and R. Fujii. 1982. On a newmedical sterile-water equipment utilizing hollow fibers. J.Antibact. Antifung. Agents (Osaka) 10:239-247.

12. Kamiki, T., Y. Sawada, A. Kawai, R. Fujii, M. Inoue, and I.Igami. 1985. A new water-purification device, Sterapore® IIPKC, for multiple use. M. A. Journal (Tokyo) 6:8-13.

13. Leive, L., V. K. Schvlin, and S. E. Mergenhagen. 1968. Physical,chemical and immunological properties of lipopolysaccharidereleased from Escherichia coli by ethylenediamine tetraaceticacid. J. Biol. Chem. 243:6384-6391.

14. Maitra, S. K., T. T. Yoshikawa, L. B. Guze, and M. C. Schotz.1981. Properties of binding of Escherichia coli endotoxin tovarious matrices. J. Clin. Microbiol. 13:49-53.

15. Niwa, M., K. C. Milner, E. Ribi, and J. A. Rudbach. 1969.Alteration of physical, chemical, and biological properties ofendotoxin by treatment with mild alkali. J. Bacteriol.97:1069-1077.

16. Niwa, M., M. Umeda, and K. Ohashi. 1982. Inactivation andimmobilization of endotoxin. A novel endotoxin binding sub-stance, polymyxin-Sepharose. Jpn. J. Med. Sci. Biol.35:114-115.

17. Nolan, J. P., J. J. McDevitt, and G. S. Goldman. 1975.

Endotoxin binding by charged and uncharged resins. Proc. Soc.Exp. Biol. Med. 149:766-770.

18. Parkinson, G. 1983. Reverse osmosis: trying for wider applica-tions. Chem. Eng. 90:26-31.

19. Pegues, A. S., S. S. Sofer, R. E. McCallum, and L. B. Hinshaw.1979. The removal of '4C labelled; endotoxin by activatedcharcoal. Int. J. Artif. Organs 2:153-158.

20. Porter, M. C. 1975. Selecting the right membrane. Chem. Eng.Prog. 71:55-61.

21. Reichelderfer, P. S., J. F. Manischewitz, M. A. Wells, H. D.Hochstein, and F. A. Ennis. 1975. Reduction of endotoxin levelsin influenza virus vaccines by barium sulfate adsorption-elution.Appl. Microbiol. 30:333-334.

22. Ribi, E., R. L. Anacker, R. Brown, W. T. Haskins, B.Malmgren, K. C. Milner, and J. A. Rudbach. 1966. Reaction ofendotoxin and surfactants. I. Physical and biological propertiesof endotoxin treated with sodium deoxycholate. J. Bacteriol.92:1493-1509.

23. Rudbach, J. A. 1970. Some requisites in the system leading tohybrid formation between bacterial endotoxins. J. Infect. Dis.122:139-145.

24. Sarada, T., and L. C. Sawyer. 1983. Three dimensional structureof Celgard® microporous membranes. J. Membr. Sci.15:97-113.

25. Sato, H., P. S. Malchesky, S. Matsubara, and Y. Nose. 1983.Characterization of polyethylene membrane for plasma separa-tion. Artif. Organs (Cleveland) 7:213-220.

26. Seid, R. C., and J. C. Sadoft. 1981. Preparation and character-ization of detoxified lipopolysaccharide-protein conjugate. J.Biol. Chem. 256:7305-7310.

27. Shands, J., Jr., and P. W. Chun. 1980. The dispersion ofgram-negative lipopolysaccharide by deoxycholate. J. Biol.Chem. 255:1221-1226.

28. Sueoka, A., P. S. Malchesky, and Y. Nose. 1984. Pore charac-teristics of microporous membranes for plasma separation. Jpn.J. Artif. Organs 13:903-906.

29. Sweadner, K. J., M. Forte, and L. L. Nelsen. 1977. Filtrationremoval of endotoxin (pyrogens) in solution in different stagesof aggregation. Appl. Environ. Microbiol. 34:382-385.

30. Tomasulo, P. A., S. C. March, and J. Levin. 197M. Tritiation ofendotoxin. Biochem. Biophys. Acta 444:399-406.

31. Tsuji, K., and S. J. Harrison. 1979. Limulus amoebocyte ly-sate-a means to monitor inactivation of lipopolysaccharide.Prog. Clin. Biol. Res. 29:367-378.

32. Tutunjian, R. S. 1982. Pyrogen removal by ultrafiltration. Prog.Clin. Biol. Res. 93:319-327.

33. Usami, H., and A. Shimohira. 1981. Study of microassay ofpyrogen in vitro. 1. Application of chromogenic assay method(synthetic substrate method). Annu. Rep. Tokyo Metr. Lab.Public Health 32:83-87.

34. Washburn, E. W. 1921. Note on a method of determining thedistribution of pore sizes in a porous material. Proc. Natl. Acad.Sci. USA 7:115-116.

35. Weiser, M., and L. Rothfield. 1968. The reassociation of lipo-polysaccharide, phospholipid and transferase enzyme of thebacterial cell envelope. Isolation of binary and ternary com-plexes. J. Biol. Chem. 243:1320-1328.

36. Westphal, O., 0. Luderitz, and F. Bister. 1952. Uber dieExtraktion von Bacterien mit Phenol/Wasser. Z. Naturforsch.76:148-155.

37. Yamagami, S., H. Yoshihara, S. Iritani, M. Umeda, M. Senju, T.Kishimoto, M. Maekawa, M. Niwa, S. Iwanaga, and S. Tanaka.1985. Is an LAL reactive substance contained in the washingsolution of unused dialyzer endotoxin? Jpn. J. Artif. Organs14:30-33.

820 SAWADA ET AL.

on July 24, 2018 by guesthttp://aem

.asm.org/

Dow

nloaded from