Airborne Environmental Endotoxin: a Cross-Validation of Sampling and Analysis Techniques

Transcription

1 APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Mar. 1994, p Vol. 6, No /94/$4.+ Copyright C 1994, Amerian Soiety for Mirobiology Airborne Environmental Endotoxin: a Cross-Validation of Sampling and Analysis Tehniques M. WALTERS,' D. MILTON,"* L. LARSSON,2 AND T. FORD1 Department of Environmental Health, Harvard Shool of Publi Health, Boston, Massahusetts 2115,1 and Department of Medial Mirobiology, University of Lund, Lund, Sweden2 Reeived 28 April 1993/Aepted 29 November 1993 A standard method for measurement of airborne environmental endotoxin was developed and field tested in a fiberglass insulation-manufaturing faility. This method involved sampling with a apillary-pore membrane filter, extration in buffer using a soniation bath, and analysis by the kineti Limulus assay with resistant-parallel-line estimation (KLARE). Cross-validation of the extration and assay method was performed by omparison with methanolysis of samples followed by 3-hydroxy fatty aid (3-OHFA) analysis by gas hromatography-mass spetrometry. Diret methanolysis of filter samples and methanolysis of buffer extrats of the filters yielded similar 3-OHFA ontent (P =.72); the average differene was 2.1%. Analysis of buffer extrats for endotoxin ontent by the KLARE method and by gas hromatography-mass spetrometry for 3-OHFA ontent produed similar results (P =.23); the average differene was.88%. The soure of endotoxin was gram-negative bateria growing in reyled washwater used to lean the insulation-manufaturing equipment. The endotoxin and bateria beome airborne during spray leaning operations. The types of 3-OHFAs in bateria ultured from the washwater, present in the washwater and in the air, were similar. Virtually all of the bateria ultured from air and water were gram negative omposed mostly of two speies, Deleya aesta and Ainetobaterjohnsonii. Airborne ountable bateria orrelated well with endotoxin (r2 =.64). Repliate sampling showed that results with the standard sampling, extration, and Limulus assay by the KLARE method were highly reproduible (95% onfidene interval for endotoxin measurement +.28 log1o). These results demonstrate the auray, preision, and sensitivity of the standard proedure proposed for airborne environmental endotoxin. Although airborne endotoxin is pervasive in the environment, efforts to study health effets of inhaled endotoxin and to regulate oupational exposures are hampered by a lak of standardized and validated methods for sample olletion and analysis. Elevated levels of endotoxin have been measured in agriulture (6), the biotehnology industry (25), offie buildings (37), and swimming pools (21). Work environments with high endotoxin levels inlude otton mills (14, 33), grain storage and proessing buildings (7, 26), farm animal barns (6, 24, 4), sewage treatment plants (15), industrial water reyling systems (8), and humidified work areas (36). Experimental studies in humans exposed to otton dust aerosols found aute effets on lung funtion orrelated with endotoxin but not with dust onentration (2, 36). Inhalation of endotoxin at elevated onentrations has been assoiated with mill fever (23), byssinosis (3), humidifier fever (35), aute airway obstrution in agriultural workers (6), hypersensitivity pneumonitis (9, 32), and redued baseline lung funtion (14). Therefore, reommendations have been made to limit exposure to this agent (13, 25, 33). At least one study of very high exposures, however, failed to find adverse effets (29), and several epidemiologi studies did not find an exposure-response relationship for endotoxin and aute airflow obstrution (3, 14, 34). Consistent, standardized methods to measure and quantify exposure to environmental endotoxin have not been established. Also, standard proedures for the sampling of environmental endotoxin have not been fully developed. Filter media have not been speified, and little * Corresponding author. Mailing address: Oupational Health Program, Bldg. 1, Rm. 146, Harvard Shool of Publi Health, 665 Huntington Ave., Boston, MA Eletroni mail address: dmilton@hohp.harvard.edu. 996 researh has been undertaken to ompare sampling and analysis methods in the same environment (13). These fators may explain some of the inonsistenies among previous studies. Analysis and quantifiation of strutural omponents of lipopolysaharide (LPS), suh as 3-hydroxy fatty aids (3- OHFAs), an provide useful information onerning the total amount of LPS or endotoxin in environmental samples (16, 22, 38). Other benefits inlude the partial identifiation of gramnegative baterial speies and estimates of the amount of baterial mass present (17). It is possible, however, that LPS may be present in partially degraded or denatured form in some environments, for example, detergent-water mixtures. Under these onditions, analysis of 3-OHFA ontent may detet LPS with redued or altered biologial ativity. In the Limulus assay, poteny of an endotoxin sample is highly orrelated with pyrogeniity (27); Limulus ameboyte lysate (LAL) will not reat with denatured LPS. However, some lysates an give false-positive results, suggesting the presene of endotoxin even in the absene of LPS, as onfirmed by the absene of 3-OHFAs (16). A reent omparison of airborne endotoxin ontent of agriultural samples estimated by gas hromatography-mass spetrometry (GC-MS) analysis of 3-OHFA and with Limulus methods suggested that the Limulus method underestimated endotoxin by many orders of magnitude (38). Thus, further ross-validation of methods using improved Limulus assay methods is important. This paper presents a omprehensive method for measuring environmental endotoxin in aerosols that inludes sample olletion, extration, and Limulus-based analysis. The rossvalidation study was a omponent of an aute respiratory health effets and exposure assessment study performed in a fiberglass insulation-manufaturing plant. Endotoxin estimates from the aqueous buffer extration and Downloaded from on April 22, 218 by guest

2 VOL. 6(), 1994 modified Litmulus method used for the study, the kineti Limol/us assay with resistant-parallel-line estimation (KLARE) method (2), are ompared with estimates from an alternative extration and GC-MS method for 3-OHFA (38). Endotoxin measurements are also ompared with ulturable and total baterial ounts. Additionally, the soure of the baterial and endotoxin ontamination, the reyled washwater, was analyzed and ompared with air samples taken from the proximity of the washwater leaning operations or high-mist areas and, ultimately, with air in the manufaturing plant to whih workers are exposed. MATERIALS AND METHODS Limulus assay. Control standard endotoxin (Esherihia oli 113, PPE-E-4349 lot 46) and LAL (lot T) were obtained from Assoiates of Cape Cod, Woods Hole, Mass. USP referene standard endotoxin (EC5) was obtained from the U.S. Pharmaopeia, In. (Rokville, Md.). Co-lyophilized LAL with hromogeni substrate (BioWhittaker, Walkersville, Md.) was used in one experiment for omparison with turbidimetri lysate. The Assoiates of Cape Cod ontrol standard endotoxin poteny was 1.1 endotoxin units/ng when assayed with Assoiates of Cape Cod lysate and 8.2 endotoxin units/ng when assayed with hromogeni lysate. Endotoxin-free water was obtained from Assoiates of Cape Cod for reonstitution of LAL reagents. For all other reagents, endotoxin-free water-(sterile water for irrigation [USP])- was obtained in sealed srew-top plasti ontainers from Baxter Health Corp. (Deerfield, Ill.). Phosphate salts and triethylamine (Fisher Sientifi, Fairlawn, N.J.) were used to prepare endotoxin-free standard buffer for extrations and dilutions,.5 M potassium phosphate-.1 % triethylamine (ph 7.5), as previously desribed (2). Serial dilutions were prepared with disposable polypropylene tips from Rainin (Woburn, Mass.). Miroplates were provided by Linbro/Titertek, ICN Biohemials, In. (Horsham, Pa.). Glassware was baked at 27 C for a minimum of 3 min before use. Dilution tubes were new, disposable glass tubes (12 by 75 mm; VWR Sientifi, Boston, Mass.) that were inspeted for absene of debris and baked before use. A model 32 bath soniator manufatured by Branson (Danbury, Conn.) was used for all aqueous buffer extrations. Liml/luls reations were monitored in a kineti miroplate reader (Vmax; Moleular Devies Corp., Menlo Park, Calif.) operated inside an inubator (37 C) or in a BioWhittaker kineti miroplate reader with internal 37 C temperature ontrol. The miroplate readers were ontrolled, and data were reorded by IBM-PC-XT omputers. The KLARE method, reported in detail by Milton and olleagues (2), was applied with the following modifiations. Control standard endotoxin and field samples were serially diluted for endotoxin analysis in the standard buffer. Dupliate 5-pA aliquots of five serial dilutions of sample and standard were plaed in a 96-well, flat-bottomed polystyrene miroplate and agitated. The optial density (OD) of eah well was reorded at 45 nm every 15 s for 9 min without subsequent agitation. The response parameter for eah well of the miroplate was the maximum rate of OD hange omputed as the maximum of the first derivative of a smoothing spline (4) fit to the OD data, similar to the rate method of Ditter et al. (5). The rate of turbidimetri lysate was saled to the maximum OD from the highest standard endotoxin onentration to ompensate for signifiant variation of maximum OD among wells. The rate response was hosen beause it gave improved sensitivity to AIRBORNE ENVIRONMENTAL ENDOTOXIN 997 low endotoxin onentrations and appears to be less affeted by interfering ompounds than the previously reported onset method (3). The log poteny and its variane were omputed by using resistant regression. Interferene was deteted by analysis of ovariane, and a standard algorithm sueeded in eliminating suh effets (2). No data were disarded as invalid due to interferene. All results are reported as equivalent amounts of EC5 referene standard endotoxin (U.S. Pharmaopia, In.). Figure 1 summarizes the standard proedures for sample olletion, extration, and Limit/us assay used in this study. 3-OHFA assay. Trideanoi aid hydroxylated in position 3 (3-OH-13:) was purhased from Larodan Lipids (Malmo, Sweden). Bis(trimethylsilyl)trifluoroaetamide (98%), pentafluorobenzoyl hloride (98%), dihloromethane (p.a. stabilized with 5 ppm of amylene), diethyl ether (p.a.), and aetonitrile (99%) were from Janssen Chimia (Geel, Belgium). Methanol (p.a.) and n-hexane (99%) were from Lab San (Dublin, Ireland), aetyl hloride (p.a.) and pyridine (p.a.) were from Merk (Darmstadt, Germany), and triethylamine was from Sigma (St. Louis, Mo.). All hemials were used without further purifiation. The glass test tubes (equipped with polytetrafluoroethylene-lined srew aps) were heated at 35 C overnight before use. Details of the GC-MS analysis were previously desribed (19). A VG Trio-IS GC-MS system (Manhester, United Kingdom) was used. The GC was a Hewlett-Pakard model 589 (Hewlett-Pakard, Avondale, Pa.) equipped with a fusedsilia apillary olumn (25 m by.25-mm inside diameter) ontaining ross-linked OV-1 as the stationary phase. Injetions were made with a Hewlett-Pakard model 7673 autosampler in the splitless mode. Helium was used as the arrier gas, at an inlet pressure of 7 lb/in2, and the temperature of the olumn was programmed from 12 to 26 C at 2 C/min. Both the injetor and the interfae (between GC and MS) were kept at 26 C. Trimethylsilyl derivatives were analyzed in the El mode (ion soure temperature, 22 C), and pentafluorobenzoyl derivatives were analyzed in the NICI mode with an ion soure temperature of 15 C. Isobutane (1 lb/in2) was used as the reagent gas; ionization was performed at 7 ev (18). Results are presented as the relative areas of ion responses (perent GC-MS response) for 3-OHFAs with 1, 12, 14, 16, or 18 arbons. The mass of endotoxin in eah sample was estimated from the GC-MS analysis of 3-OHFA ontent by assuming that eah mole of LPS ontained 4 mol of 3-OHFA (39) and that the average moleular weight of the LPS was 8, (28, 31, 41). Aerosol sampling. Polyarbonate membrane filters,.4-pm pore size in 37-mm-diameter polystyrene assettes with AP4 glass fiber baking pads (Millipore Corp., Bedford, Mass., or Nuleopore Corp., Pleasanton, Calif.), were used for endotoxin sampling. Filter assette sampling was performed with Gilian air sampling pumps alibrated before and after use with a Gilibrator bubble flow meter (Gilian Instrument Corp., Wayne, N.J.). Beause of the high mist and heavy manufaturing operations in the plant, all sampling was performed losed faed, through the assette inlet opening. The assettes were positioned so that the inlets were faing down. The pumping rate was a nominal 1.5 liters/min for most sampling. Desiation units were used to preserve eah sample after olletion, during storage at 4 C, until extration. Air samples for total baterial ounts were olleted on.4-p.m polyarbonate membranes with the same assette and baking pad used for endotoxin sampling. Sample flow rates were adjusted (.2 to 4 liters/min) to give appropriate baterial ount loading in eah area while maintaining onstant sam- Downloaded from on April 22, 218 by guest

3 998 WALTERS ET AL. Filter Media Cassette Seletion Sampling Proedure:.4 pm polyarbonate filter membrane 37 mm polystyrene assette AP4 glass fiber baking pad Laboratory, Field Blanks & Bakground Sampling APPL. ENVIRON. MICROBIOL. Storage Preservation Transportation Sample Extration IMMEDIATE ANALYSIS desiator unit for eah sampling assette refrigerated storage at approx. 4C KLARE Method: Kineti Limulus Assay with Resistant-parallel-line Estimation.5M potassium phosphate,.1% triethylamine, ph 7.5 buffer bath soniation for 6 min at 2 C internal standard with known poteny (EC5 Endotoxin Units) kineti turbidimetri or hromogeni lysate 96 well kineti miroplate reader dupliate assay of at least 5 1:8 serial dilutions response = maximum rate of OD inrease over 9 minutes log poteny and variane from parallel line analysis using robust and resistant regression Downloaded from on April 22, 218 by guest Results: "endotoxin poteny" with onfidene intervals FIG. 1. Standard proedure for environmental endotoxin sampling and analysis. pling times of approximately 4 h. These samples were desiated and stored until ounted. The total baterial ounts were made by using standard aridine orange staining proedures and epifluoresene mirosopy (12). Airborne ulturable baterial samples were olleted with AGI-3 glass impingers (Ae Glass In., Vineland, N.J.) ontaining 2 ml of sterile filtered washwater to minimize osmoti shok to the bateria during olletion. Samples were olleted for 3 min with a arbon vane vauum pump; a flow rate in exess of 12.5 liters/min was applied to the impinger.

4 Voi.. 6, 1994 The impingers were autolaved prior to eah use to ensure sterility. Samples for ulturable bateria were diluted and plated immediately after olletion. Serial dilutions were made with filter-sterilized washwater to avoid osmoti shok. Washwater samples were plated at final dilutions of 1-6, 1 7, and 1 8? and impinger samples were plated at final dilutions of 1'2 1 3, and 1 4. Nutrient agar (BBL, Baltimore, Md.) plates were inubated at room temperature for 1 week prior to ounting and Gram staining. An attempt to use the highly olored reyled washwater as a growth medium in the agar plates was unsuessful beause olonies ould not be learly seen against the olored plates. The olor in the reyled washwater results from a dye whih is added during the manufaturing proess, making the washwater somewhat opaque to light. The plates were ultured and ounted at the OCF Biologial Laboratory (Newark, Ohio) and at Harvard University Laboratory of Mirobial Eology (Cambridge, Mass.). Gram-negative isolates were subultured on trypti soy agar (Difo, Detroit, Mih.) for identifiation, using the Mirolog Manual Bateria Identifiation System (Biolog, In., Hayward, Calif.). Two speies were predominant in most of the ultured samples in both the washwater and the mist. These two speies were isolated and further ultured separately in nutrient broth. After suffiient growth, the bateria were harvested by entrifugation followed by lyophilization. Standard proedure inluded the use of blank samples. Laboratory blanks, samples (filter and assette holders) that remained in the laboratory after assembly, were analyzed to assure that only lean filters were used for sampling. Field blanks were proessed by using the same alibration and storage tehniques used on regular samples. Appropriate bakground sampling (a number of outside, upwind samples within 1, feet [a. 35 m] of the manufaturing plant were taken for bakground) was also performed to ensure that external air was not the soure of ontamination. Bulk sampling of liquids. Fluids and liquids, suh as reyled washwater and impinger fluids, were plaed in new, sterile plasti jars, and formaldehyde was added to make a final onentration of 2C (vol/vol). An empty jar was sent with eah bath of bulk fluid samples to test for sample ontainer ontamination. Endotoxin-free water was used to rinse and extrat the inside surfaes of the empty jar. The extrat water was tested for Linuullis ativity by the KLARE method. Empty sample jars had endotoxin onentrations more than 4 log units lower than the washwater sampled. To ensure that samples of reyled washwater were representative, samples were olleted from the main reirulation pipeline. Bulk washwater samples were taken at approximately the same time airborne mist samples were olleted to allow omparison. The bulk washwater samples were analyzed for endotoxin and viable ulturable bateria. Endotoxin onentrations (determined by the KLARE method) were stable for several weeks in the presene of 2% formaldehyde. Extration methods. Endotoxin was extrated from filter samples for Linuluts assay by inubation in 5 ml of standard buffer for 6 min at 2 C in a soniation bath. Analysis by the KLARE method immediately followed eah extration. A l-ml aliquot of eah aqueous extrat was lyophilized and shipped to the University of Lund for GC-MS analysis. Endotoxin was extrated from assettes and baking pads, using the standard buffer and soniation method. The baking pads were removed from the assette, extrated, and assayed by the same proedure used for the filters. The inside surfaes of the polystyrene assettes were extrated by plaing 8 ml of AIRBORNE ENVIRONMENTAL ENDOTOXIN 999 buffer into the assette inlet, plugging the openings, and soniating for 6 min in a 2 C bath. Extrats for GC-MS analysis of fatty aids were made by two methods. First, buffer extrats prepared for the Limulus assay were lyophilized and then subjeted to methanolysis. Alternatively, polyarbonate filters were diretly subjeted to methanolysis without prior buffer extration. The proedure onsisted of adding 5 ng of the internal standard 3-OH-13: to the sample (lyophilized extrat or filter) in a glass reation vessel. One milliliter of 3.6 M methanoli HCl, prepared by adding aetyl hloride dropwise to methanol (7.5 ml), was added, and the sample was heated to 1(C for 18 h. The hydroxy aid methyl esters formed were extrated and purified by silia gel olumn hromatography as desribed before (19). Pentafluorobenzoyl derivatives of the 3-OHFA methyl esters were formed by adding 3.1I of 35C pentafluorobenzoyl hloride and 2.I1 of triethylamine in aetonitrile to the solvent-dried samples and then heating at 1 C for 1 h. Trimethylsilyl derivatives were similarly formed by adding 5 pl of bis(trimethylsilyl)trifluoroaetamide and 5 pl. of pyridine and heating to 8 C for 15 min. Experimental design. Multiple simultaneous air samples were olleted on two oasions, the first in November 1991 and the seond in Marh On the first oasion, two of the filters were extrated in buffer and analyzed by the KLARE method at the Harvard Shool of Publi Health. One buffer extrat and one filter were sent to the University of Lund, Lund, Sweden, for GC-MS analysis. Diret methanolysis served as the only extration for the filter sample. On the seond oasion, two filters per loation were extrated and analyzed by eah tehnique, and both buffer extrats were analyzed for 3-OHFA ontent. Sample blanks were part of all analysis programs. Washwater samples were olleted and analyzed on both oasions. Cultures of the most frequent baterial isolates, Deleya aesta and Ainetobater johnsoniii, were lyophilized, weighed, and suspended in buffer extrat for assay, using the KLARE method. These ultures were dried, and the dried bateria were subjeted to diret methanolysis prior to GC-MS analysis for 3-OHFA ontent. Multiple simultaneous samples of washwater and mist were taken on seven oasions at five loations in the high-mist areas during the 6-week sampling program. Washwaters were analyzed for ulturable bateria and for endotoxin by the KLARE method. Four-hour air samples were olleted for endotoxin (KLARE assay) at 2 liters/min and for total ountable bateria at 2 m3/min. Two 3-min onseutive impinger samples were taken during eah high-mist sampling event for ulturable bateria. To asertain reproduibility of the aqueous buffer extration and KLARE method, repliate endotoxin samples were taken on 24 randomly hosen oasions during a 6-week exposure assessment study. Either two, four, or six samples were olleted simultaneously on eah oasion over a wide range of endotoxin onentrations. As muh as 8 weeks elapsed between sample olletion and extration for some repliate samples in the repliate sets. In addition, paired samples were olleted for omparison of total baterial ounts with endotoxin measurements. Seventy-seven paired samples for total baterial ounts and endotoxin analysis were olleted at various sites in the plant, ranging from high-mist to internal offie areas and external bakground onditions, over 6 weeks. Statistial analyses were made by using regression, a t test for paired data, and analysis of variane in SAS version 6.7 (SAS Institute, Cary, N.C.), Pro Means and GLM. Lotus 123R23 Downloaded from on April 22, 218 by guest

5 1(( WALTERS ET AL. TABLE 1. Mirobiologial sampling results" Parameter (units) Mean Range Washwater Culturable bateria 1.8 x x 1"-4.9 x 18 (CFU/ml) Endotoxin poteny (,ug/ml)" Mist Culturable bateria 1,2 1-7,5 (CFU/m3) Endotoxin poteny (p.g/m3) Total ountable bateria 2.3 x x x 1(7 (ounts/m3) For seven sampling events witlh dupliate sampling over 6 weeks. KLARE method; results are mirograms of referene standard endotoxin, EC5. (Lotus Development In., Cambridge, Mass.) was used for data analysis and data management. RESULTS Charateristis of soure washwater. Essentially, all bateria in the washwater were gram negative. A total of six different genera were identifiable over the ourse of the study (Ainetobater, Deleya, Methylobaterium, Comamonas, Pseudomonas, and Gluonobater). However, A. johnsonii and D. aesta were the most frequently identified baterial speies from both washwater and mist in the proximity of the washwater spray. Other speies were only identified one. A. johnsonii is ommonly found in soil, water, and sewage and on the skin (1). D. aesta was originally isolated from seawater, and members of the genus are thought to have an absolute sodium requirement; the predominane of this speies may reflet the relatively high ioni strength of the washwater. Table 1 shows total CFU and endotoxin onentrations in the washwater. Charaterization of mist area air ontaminants. Bateriologial measurements in the high-mist area were made to more fully haraterize the aerosol. Table 1 also shows total ountable and ulturable bateria and endotoxin (KLARE method) results for the mist. Comparison of the onentrations of endotoxin and ulturable bateria in the washwater with that in the mist indiates a muh lower onentration of ulturable bateria in the mist than would have been predited by endotoxin onentrations in the mist. Figure 2 plots multiple parameters from seven sets of repliate samples taken of washwater and air in a high-mist area over a 6-week period. Generally, the mirobial parameters of the washwater and mist were onsistent over the 6-week study. These parameters varied only within I to 2 orders of magnitude during the study. Beause these parameters displayed so little variability, the ratio of endotoxin to ultured bateria in the washwater was also reasonably onstant. 3-OHFA assay. Figure 3A shows the relative GC-MS response for eah of the 3-OHFAs in washwater samples from three loations. The predominant moiety was 3-OH-12:. The 3-OH-14: moiety was not deteted in two area samples and was present at a low perentage in the third. Figure 3B displays the GC-MS response for eah 3-OHFA present in D. aesta and A. johnsonii, isolated from washwater samples. The baterial speies have similar fatty aid ompositions. Two moieties make up the majority of the 3-OHFAs; CZ a) C ) -j C C C t C C t t t t Loation: B2 61 B3 B1 B2 Sampling Day: Sampling Loation & Day FIG. 2. Analysis of simultaneous washwater and airborne mist samples. The results of seven simultaneous sampling events are plotted against sampling oasion. C, ulturable bateria (CFU per milliliter), and E, endotoxin poteny (nanograms per milliliter), in reyled washwater., ulturable bateria (CFU per ubi meter); e, endotoxin poteny (nanograms per ubi meter); and t, total ountable bateria (ounts per ubi meter), in the mist in the immediate proximity of the washwater spray soure. over 5% of the total was 3-OH-1: and over 35% was 3-OH-I 2:. The 3-OHFA omposition of the mist in the immediate proximity of the washwater spraying operations is shown in Fig. 3C. The fatty aid omposition of the mist is similar to that of the washwater (Fig. 3A) and the isolated bateria (Fig. 3B). The 3-OH-12: and 3-OH-1: moieties also predominate in the airborne mist. No signifiant amount of 3-OH-14: was present in either isolated bateria or any of the washwater or airborne mist samples. Extration effiieny. The effetiveness of endotoxin extration by an aqueous buffer was ompared with diret extration of 3-OHFAs by methanolysis of samples olleted on polyarbonate membrane filters. Figure 4 allows omparison of airborne endotoxin onentrations estimated from the 3-OHFA ontent of both the aqueous extrats and the filters subjeted to diret methanolysis. The two extration methods are in lose agreement when analyzed by GC-MS for 3-OHFA ontent. In three loations, the diret methanolysis produed slightly higher estimates, and in two loations (a total of three samples) the aqueous extrats were slightly greater. The differene between the two extration methods was not signifiant (mean = 2.1%; P =.72). Comparison of the aqueous extrat with the diret methanolysis indiates that the aqueous buffer extration of polyarbonate membrane filters is highly effiient. Comparison of analysis methods. Figure 5 shows the endotoxin ontent of aerosol samples estimated with the GC-MS and KLARE methods. For this omparison, endotoxin is expressed as nanograms of LPS as desribed above. The two methods agree in most sets, and no onsistent pattern of C APPL. ENVIRON. MICROBIOL. e e e e E Ee 6 E E E e e Downloaded from on April 22, 218 by guest

6 VOL. 6, 1994 AIRBORNE ENVIRONMENTAL ENDOTOXIN 11 A 4 o en 2 e CL e 6 CL Io M S 6 B C Area 1 Deleya aesta Area 2 Area 3 Ainetobater johnsonil Area IN Area 2N Area 3N Aea I M Area 3M Area 4M FIG. 4. Effet of extration method on endotoxin onentration estimated from 3-OHFA ontent in aerosol samples. Paired samples were olleted in November 1991 (N) and quadrupled samples were olleted in Marh 1992 (M). Half of the samples were extrated by soniation in buffer (X). The remaining half were extrated by diret methanolysis (u). All samples were analyzed by GC-MS for 3-OHFA. Results are represented as nanograms of LPS with a nominal moleular weight of 8,. The error bars on the Marh samples represent the standard deviation of two samples. differene has been noted between them. The differene between the two analytial methods was not signifiant (mean =.88%; P =.23). The Marh sampling program also inluded sampling in some loations with lower ambient air endotoxin onentrations. These loations had onentrations of less than 15 ng/m3, below the limit of detetion by GC-MS. Therefore, only omparisons based on high-mist areas are shown. The estimated endotoxin ontent of the two isolated baterial speies by the GC-MS and KLARE methods is displayed in Fig. 6. Again, the two methods show lose agreement for the isolated bateria. Figure 7 is a satter plot of log endotoxin (nanograms per C m.s T T_ Downloaded from on April 22, 218 by guest Area 1 Area 2 Area 3 FIG. 3. Relative ontent of 3-OHFAs. (A) Washwater. Data are shown for three representative washwater samples from different areas and were olleted in November 1991 and analyzed for 3-OHFA ontent. The perent GC-MS response ompares eah 3-OHFA moiety with the total 3-OHFA present in the sample. Carbon hain lengths of the different 3-OHFA moieties: *, 1:; U, 12:; E, 14:; X, 16:; [, 18:. (B) Isolated bateria. Data are shown for a representative analysis of the 3-OHFA omposition of the two most frequently isolated gram-negative baterial speies, D. aesta and A. johnsonii, from washwater and mist samples. The perent GC-MS response and symbols are as desribed above. (C) Mist. Data are shown for three representative mist samples from areas orresponding to washwater samples in panel A, olleted in November The perent GC-MS response and symbols are as desribed above. UJ.j 1- Area 1 N Area 2N Area IM Area 3M Area 4M FIG. 5. Comparison of endotoxin onentration of aerosol samples estimated from 3-OHFA ontent and Limulus assay. Buffer extrats were analyzed by the KLARE method (U). The 3-OHFA results (1) are from diret methanolysis only in Marh 1992 samples and from both diret methanolysis and buffer extrats in November 1991 samples. KLARE assay results are reported as nanograms of referene standard endotoxin (EC5), and 3-OHFA assay results are in nanograms of LPS with a nominal moleular weight of 8,. The error bars represent the standard deviation of dupliate samples.

7 12 WALTERS ET AL. E -S.x w Deleya aesta Ainetobater Johnsonhl FIG. 6. Comparison of endotoxin onentration of isolated bateria estimated from 3-OHFA ontent and Limulus assay. The onentrations of the gram-negative baterial speies shown in Fig. 3B were estimated from methanolysis and GC-MS analysis for 3-OHFA ontent (E) and by soniation in buffer and Limulus assay by the KLARE method ( ). KLARE assay results are reported as nanograms of referene standard endotoxin (EC5), and 3-OHFA assay results are in nanograms of LPS with a nominal moleular weight of 8,. ubi meter) and log total ountable bateria (ounts per ubi meter) for 77 paired area samples taken over the full range of endotoxin measured at this faility. Log endotoxin orrelated with log total ountable bateria (r2 =.64). The KLARE method an be performed with either turbidimetri or hromogeni lysates. To test the effet of lysate, 13 assays were performed with both turbidimetri and hromogeni lysates. Results were similar for both methods (r2 =.8); the mean differene was.1 endotoxin unit/ml, with a standard deviation of 7.2 endotoxin units/ml. Repliate analysis. The mean and standard deviation of 24 sets of repliate aerosol samples analyzed for endotoxin by the KLARE method are shown in Fig. 8. The number of samples E I-, m 4- - w m -J 5S * * ~~~~* ***** ** * ** * It *f 46 *** * * * * * * * * 4I 5- _ Log Countable Bateria (ounts/m3) FIG. 7. Satterplot of airborne endotoxin and total ountable bateria. Results of 77 paired aerosol samples from multiple sampling loations analyzed for endotoxin by the KLARE method and for total ountable bateria by epifluoresene mirosopy (r2 =.64). for eah repliate set is also shown. The agreement between all repliate samples is exellent. Also, some repliate samples were analyzed within days of sampling, while other repliates were analyzed over larger intervals (up to 8 weeks after sampling). Results, however, were uniformly onsistent over a wide range of time intervals between sample olletion as well as over a wide range of environmental onentrations. Mean square error from an analysis of variane of the 24 repliate sets was.14. Thus, the average 95% onfidene interval for individual measurements was of ±.28 loglo. Laboratory and field blanks were taken frequently, often in repliate, and were assayed blindly with other samples. The analysis of 42 blanks for endotoxin produed a geometri mean of.72 ng/m3 and an arithmeti mean of.24 ng/m3, using a nominal sample volume (.36 m3), typial to the study. Cassette holder and baking pad loss. Table 2 quantifies the endotoxin levels olleted on the inside of the polystyrene sampling assette and the perentage of reovered endotoxin that penetrated the filter and reahed the baking pad. The amount of endotoxin deposited on the inside walls of the assette represents a signifiant portion of the total sampled, partiularly in the dry areas. It appeared that high humidity levels may greatly redue the quantity of endotoxin deposited on the assette walls. The baking pad did not ontain signifiant endotoxin. DISCUSSION APPL. ENVIRON. MICROBIOL. Of the biologial methods presented in this ross-validation paper, only diret measurement of endotoxin ould be readily and reliably employed for personal air sampling to assess worker exposure. The standard Limulus proedure produed highly reproduible results, using 4-h samples to make measurements over 5 orders of magnitude, from.35 ng/m' (five times the geometri mean blank) to 3.6 p.g/m3. This was aomplished without having to adjust the sampling rate or duration for antiipated filter loading, an advantage whih is of onsiderable benefit when sampling highly mobile workers exposed to highly variable airborne endotoxin onentrations. Other biologial or baterial parameters are diffiult to adapt for personal sampling due to their sensitivity to over- and underloading of samples. Measurement of endotoxin is also attrative for assessing exposure to gram-negative bioaerosols beause it is representative of intat, biologially ative toxin that may ause health effets. Other mirobiologial methods require more indiret inferenes about exposure to the important toxin. Alternatively, a variety of distint health effets are assoiated with endotoxin inhalation, inluding fever, aute and hroni airflow limitation, and hypersensitivity pneumonitis. Identifiation of bateria and the LPS struture in environments when endotoxin is assoiated with health effets is needed to further our understanding. In this ontext, bateriologial investigation and hemial analysis of the LPS strutures in these environments may lead us to reognize a variety of biologial ativities that are important in environmental endotoxin exposure. The motivation for the ross-validation sampling and analysis was to provide additional assurane that the endotoxin results by the KLARE method were relevant to the exposures of people working in the plant. The lose agreement in the quantity of endotoxin present, as measured by both the KLARE method and GC-MS analysis for 3-OHFA, may be one of the more unusual findings of this study. The endotoxin poteny determined by Limulus assay an be affeted by many parameters, inluding the LPS aggregation state (lamellar or Downloaded from on April 22, 218 by guest

8 VOL. 6, 1994 AIRBORNE ENVIRONMENTAL ENDOTOXIN ) - o C m w ) S a n ua U a WIa -J Number of Repliates: I I I I I I I I I I I I I I I I I I I I I I Sampling FIG. 8. Repliate airborne environmental endotoxin sampling. A total of 68 samples were olleted during 24 sampling events at a variety of loations in the plant. All samples were analyzed by soniation in buffer and analyzed by the KLARE method. The mean and standard deviation are shown. miellar struture and the size distribution of the mielles), the size of the polysaharide, the quantity and types of 3-OHFA present, the arrangement of the 3-OHFAs in the lipid A moleule, and the presene of phosphates on the digluosamine bakbone struture (1, 39, 41). The KLARE method was designed to minimize these effets to the extent possible by ontrol of ph, ioni strength, and soniation with triethylamine. Yet, while orrelation was expeted, the degree of quantitative agreement may depend on the speifi organisms or on the struture of the LPS, in partiular, the number and types of 3-OHFA present. The soure of the endotoxin was a washwater that ontained essentially only gram-negative bateria. The 3-OHFA omposition of the washwater, the mist, and the bateria isolated from the washwater and mist were essentially idential. Thus, the high orrelation between airborne total ountable bateria and endotoxin was antiipated. TABLE 2. Analysis of sampling assettes and baking pads No. of Mean (%) Range Sampling loation samples of sample (%) mass (SD) Inside surfae of assette filter holder High mist (humid onditions) (3.4) Dry loations (18.4) Baking pad (behind filter) High mist (humid onditions) 5.4 (.4) LQa-1.2 Dry loations (4.8) Combined assette and baking pad loss High mist (humid onditions) Dry loations alq, limit of quantifiation. Event Researh on measurement tehniques for environmental endotoxin is motivated by a desire to determine whether inhaled endotoxin has an assoiation with adverse health effets. Consistent assoiations between inhaled environmental endotoxin and health effets have not been established. The diffiulty that may be enountered in determining and generalizing suh relationships an be appreiated in the literature on the relation of the Limulus assay with the rabbit pyrogen test for environmental endotoxin and the relationship of the rabbit pyrogen test to pyrogeniity in humans. Pearson et al. (27) found that the Limulus assay overestimates the pyrogeniity of environmental endotoxin for rabbits. And, earlier work by Greisman and Hornik (11) found that humans and rabbits shared a similar threshold for pyrogeniity for pure LPS but exhibited very different dose-response relationships above the threshold. Given these diffiulties, generalizing the high degree of ross-validation that ourred in this investigation to other environmental endotoxins would not be wise. Therefore, future work is needed to provide omprehensive haraterization of endotoxin exposures. These efforts an be direted at resolving unertainties in endotoxin measurement and ultimately in its assoiation with health effets. ACKNOWLEDGMENTS This work was funded in part by the Owens-Corning Fiberglas Corp., NIEHS Oupational and Environmental Health Center grant 2P3 ES 2, the Swedish Work Environment Fund, and the Center for Indoor Air Researh. A miroplate reader and hromogeni LAL were provided by BioWhittaker (Walkersville, Md.). We speifially aknowledge the assistane of T. Mullady, T. Fuhs, D. Phillips, T. Gartner, and J. Skavarka of the Owens-Corning Fiberglas Corp.; Olga Yulikova and Helen Reihel of the Endotoxin Laboratory, Harvard University, Shool of Publi Health; G. Patel of the Laboratory of Mirobial Eology, Harvard University; and Zbigniew and Eugenia Mielnizuk during their fellowship visits at the Department of Medial Mirobiology, University of Lund. Downloaded from on April 22, 218 by guest

9 14 WALTERS ET AL. Additionally, we aknowledge Polaroid Corporation for use of industrial hygiene sampling equipment and support of M. Walters during partial fulfillment of requirements for the Dotor of Siene degree. REFERENCES 1. Balows, A., H. G. Truper, M. Dworkin, W. Harder, and K. H. Shleifer The prokaryotes. Springer Verlag, New York. 2. Castellan, R. M., S. A. Olenhok, K. B. Kinsley, and J. L. Hankinson Inhaled endotoxin and dereased spirometri values, an exposure-response relation for otton dust. N. Engl. J. Med. 317: Cinkotai, F. F., M. G. Lokwood, and R. Rylander Airborne miro-organisms and prevalene of byssinoti symptoms in otton mills. Am. Ind. Hyg. Asso. J. 38: de Boor, C A pratial guide to splines. Applied mathematial sienes. Springer-Verlag, New York. 5. Ditter, B., K.-P. Beker, R. Ursbashek, and B. Ursbahek Detetion of endotoxin in blood and other speimens by evaluation of photometrially registered LAL-reation-kinetis in miroliter plates, p In S. W. Watson and J. Levin (ed.), Endotoxins and their detetion with the Limulus ameboyte lysate test. Alan R. Liss, New York. 6. Donham, K., P. Haglind, Y. Peterson, R. Rylander, and L. Belin Environmental and health studies of farm workers in Swedish swine onfinement buildings. Br. J. Ind. Med. 46: Dutkiewiz, J., S. A. Olenhok, W. G. Sorenson, V. F. Gerenser, J. J. May, D. S. Pratt, and V. A. Robinson Levels of bateria, fungi, and endotoxin in bulk and aerosolized orn silage. Appl. Environ. Mirobiol. 55: Flaherty, D. K., F. H. Dek, J. Cooper, K. Bishop, P. A. Winzenburger, L. R. Smith, L. Bynum, and W. B. Witmer Baterial endotoxin isolated from a water spray air humidifiation system as a putative agent of oupation-related lung disease. Infet. Immun. 43: Flaherty, D. K., F. H. Dek, M. A. Hood, C. Liebert, F. Singleton, P. Winzenburger, K. Bishop, L. R. Smith, L. M. Bynum, and W. B. Witmer A Cytophaga speies endotoxin as a putative agent of oupation-related lung disease. Infet. Immun. 43: Galanos, C., and. Luderitz Lipopolysaharide: properties of an amphipathi moleule, p In E. Rietshel (ed.), Chemistry of endotoxin. Elsevier, New York. 11. Greisman, S. E., and R. B. Hornik Comparative pyrogeni reativity of rabbit and man to baterial endotoxin. Pro. So. Exp. Biol. Med. 131: Hobbie, J. E., R. J. Daley, and S. Jaspers Use of Nulepore filters for ounting bateria by epifluoresene mirosopy. Appl. Environ. Mirobiol. 33: Jaobs, R. R Airborne endotoxins: an assoiation with oupational lung disease. Appl. Ind. Hyg. 4: Kennedy, S. M., D. C. Christiani, E. A. Eisen, D. H. Wegman, I. A. Greaves, S. A. Olenhok, T. Ye, and P. Lu Cotton dust and endotoxin exposure-response relationships in otton textile workers. Am. Rev. Respir. Dis. 135: Laitinen, S., A. Nevalainen, M. Kotimaa, J. Liesivuori, and P. J. Martikainen Relationship between baterial ounts and endotoxin onentration in the air of wastewater treatment plants. Appl. Environ. Mirobiol. 58: Maitra, S. K., R. Nahum, and F. C. Pearson Establishment of beta-hydroxy fatty aids as hemial marker moleules for baterial endotoxin by gas hromatography-mass spetrometry. Appl. Environ. Mirobiol. 52: Mattsby-Baltzer, I., M. Sandin, B. Ahlstrom, S. Allenmark, M. Edebo, E. Falsen, K. Pedersen, N. Rodin, R. A. Thompson, and L. Edebo Mirobial growth and aumulation in industrial metal-working fluids. Appl. Environ. Mirobiol. 55: Mielnizuk, Z., S. Alugupalli, E. Mielnizuk, and L. Larsson Gas hromatography-mass spetrometry of lipopolysaharide 3-hydroxy fatty aids: omparison of pentafluorobenzoyl and trimethylsilyl methyl ester derivatives. J. Chromatogr. 623: Mielnizuk, Z., E. Mielnizuk, and L. Larsson Gas hro- APPL. ENVIRON. MICROBIOL. matography-mass spetrometry methods for analysis of 2- and 3-hydroxylated fatty aids: appliation for endotoxin measurement. J. Mirobiol. Methods 17: Milton, D. K., H. A. Feldman, D. S. Neuberg, R. J. Brukner, and I. A. Greaves Environmental endotoxin measurement: the kineti limulus assay with resistant-parallel-line estimation. Environ. Res. 57: Milton, D. K., C. S. Rose, J. W. Martyny, and K. Kreiss Airborne endotoxin assoiated with hypersensitivity pneumonitis outbreak among lifeguards. Am. Rev. Respir. Dis. 141:A Morris, N. M., E. A. Catalano, and R. J. Berni Hydroxymyristi aid as a measure of endotoxin in otton lint and dust. Am. Ind. Hyg. Asso. J. 49: Neal, P. A., R. Shneiter, and B. H. Caminita Report on aute illness among rural mattress makers using low grade stained otton. JAMA 119: Olenhok, S. A., J. C. Mull, and W. G. Jones Endotoxins in otton: washing effets and size distribution. Am. J. Ind. Med. 4: Palhak, R. B., R. Cohen, M. Ainslie, and C. L. Hoerner Airborne endotoxin assoiated with industrial-sale prodution of protein produts in gram-negative bateria. Am. Ind. Hyg. Asso. J. 49: Palmgren, M. S Mirobial and toxi onstituents of grain dust and their health impliations, p In J. Laey (ed.), Trihotheenes and other myotoxins: Proeedings of the International Myotoxin Symposium, Sydney, Australia. John Wiley & Sons, Ltd., London. 27. Pearson, F. C., M. E. Weary, J. Bohon, and R. Dabbah Relative poteny of "environmental" endotoxin as measured by the Limulus ameboyte lysate test and the USP rabbit pyrogen test, p In S. W. Watson and J. Levin (ed.), Endotoxins and their detetion with the Limulus ameboyte lysate test. Alan R. Liss, In., New York. 28. Peterson, A. A., and E. J. MGroarty High-moleularweight omponents in lipopolysaharides of Salmonella typhimurium, Salmonella minnesota, and Esherihia oli. J. Bateriol. 162: Rask-Anderson, A., P. Malmberg, and M. Lundholm Endotoxin levels in farming: absene of symptoms despite high exposure levels. Br. J. Ind. Med. 46: Reynolds, S., and D. K. Milton Comparison of methods for analysis of airborne endotoxin. Appl. Oup. Environ. Hyg. 8: Rivera, M., L. E. Byran, R. E. W. Hanok, and E. J. MGroarty Heterogeneity of lipopolysaharides from Pseudomonas aeruginosa: analysis of lipopolysaharide hain length. J. Bateriol. 17: Rose, C. S., L. S. Newman, J. W. Martyny, D. Weiner, K. Kreiss, D. K. Milton, and T. E. King Outbreak of hypersensitivity pneumonitis in an indoor swimming pool: linial, pathophysiologi, radiographi, pathologi, lavage and environmental findings. Am. Rev. Respir. Dis. 141:A Rylander, R Health effets of otton dust exposures. Am. J. Ind. Med. 17: Rylander, R., P. Haglind, and B. T. Buther Reations during work shift among otton mill workers. Chest 84: Rylander, R., P. Haglind, M. Lundholm, I. Mattsby, and K. Stenqvist Humidifier fever and endotoxin exposure. Clin. Allergy. 8: Rylander, R., P. Haglind, and M. Lundholm Endotoxin in otton dust and respiratory funtion derement among otton workers in an experimental ardroom. Am. Rev. Respir. Dis. 131: Rylander, R., S. Soerensen, H. Goto, K. Yuasa, and S. Tanaka The importane of endotoxin and gluan for symptoms in sik buildings, p In C. J. Biera, Y. Courtois, and M. Govaerts (ed.), Present and future of indoor air quality: proeedings of the Brussels onferene. Exerpta Media, New York. 38. Sonesson, A., L. Larsson, A. Shutz, L. Hagmar, and T. Hallberg Comparison of the Limulus ameboyte lysate test and gas hromatography-mass spetrometry for measuring lipopolysaha- Downloaded from on April 22, 218 by guest

10 VOL. 6, 1994 AIRBORNE ENVIRONMENTAL ENDOTOXIN 15 rides (endotoxins) in airborne dust from poultry proessing industries. Appl. Environ. Mirobiol. 56: Takada, H., and S. Kotani Strutural requirements of lipid A for endotoxiity and other biologial ativities. Crit. Rev. Mirobiol. 16: Thedell, T. D., J. C. Mull, and S. A. Olenhok A brief report of gram-negative baterial endotoxin levels in airborne and settled dusts in animal onfinement buildings. Am. J. Ind. Med. 1: Vukajlovih, S. W., and D. C. Morrison Conversion of lipopolysaharides to moleular aggregates with redued subunit heterogeneity: demonstration of LPS-responsiveness in "endotoxin-unresponsive' C3H/HeJ splenoytes. J. Immunol. 13: Downloaded from on April 22, 218 by guest