Toxicity of Intratracheally Instilled Cotton Dust, Cellulose, and Endotoxin 1- 4
DONALD K. MILTON,s JOHN J. GODLESKI, HENRY A. FELDMAN, and IAN A. GREAVES
Introduction
Cotton dust consists of inhalable (~ 15 urn in diameter) and respirable (~3Ilm) particles, with as much as 170/0 of the total particulate mass having diameters ~ 2.5 urn (1). Aerosols of cotton dust are composed of plant constituents, including cellulose, and particles derived from microbial flora of the cotton plant that include endotoxins and proteinases. Recent attention has focused on the airway effects of endotoxin in cotton dust (2, 3); however, because cotton dust contains considerable amounts of fine respirable material, parenchymal effects should also be considered. Chronic airflow obstruction may result from prolonged cotton dust exposure in some persons (4). Such obstruction could theoretically result from intrinsic airway narrowing or from emphysema or from a combination of both conditions (5). Epidemiologic studies of airflow obstruction among cotton mill workerscould not distinguish between these possible mechanisms (6), and postmortem studies (7-9) have had serious flaws in study design, data, and analysis (10-12) that may have obscured the mechanisms of airflow obstruction attributable to cotton dust. Two of the potentially toxic agents in cotton dust, elastase and endotoxin, have been linked to emphysema. Elastase, recently reported in cotton dust (13), is thought to playa central role in the development of emphysema (14). Endotoxin, long recognized as a constituent of cotton dust, also has been linked to emphysema in animal models of intravascular leukocyte sequestration (15, 16). Animal models of cotton dust toxicity (2, 17-20) have concentrated on guinea pig airway responses, but changes also occur at the alveolar level. Morphometric analysis of guinea pig lungs after the inhalation of high levels of respirable cotton dust for 1 yr has demonstrated both airway and parenchymal changes (21). The thickened alveolar septa were consistent also with the earlier findings of 184
SUMMARY Cotton dust Includes respirable particles containing endotoxin and elastase, agents associated with emphysema. Toexamine whether a respirable fraction of cotton dust could produce emphysema in an animal model, we Intratracheally Instilled hamsters with respirable cotton dust particles (0.75 mgl100-g animal), mass median aerodynamic diameter ~ 4.8 11m, twice weekly for 6 wk. We also examined whether Instilled endotoxin (225 I1gI1OO-g animal) could produce emphysema In hamsters and whether cellulose (0.75 mg/100-g animal) Is an appropriate Inert comparison dust. A saline-Instilled group was the control. Hamsters were killed 8 wk after the last Instillation. Static pressure-volume deflation curves of air-filled excised lungs were analyzed to measure lung distensibility. Lungs were fixed In Inflation using glutaraldehyde and were examined morphometrlcally to obtain surface area and numbers of granulomata. Endotoxin-treated animals had Increased distensibility, reduced surface-to-volume (SN) ratio, and morphologically apparent mild centrllobular emphysema. Cellulose-treated animals had decreased distensibility, normal SN ratio, and significant numbers of granulomata with patchy areas of thickened interalveolar septa. Cotton-dustinstilled animals had normal distensibility, reduced S/V ratio, significant numbers of granulomata, and mild centrllobular emphysema. These data suggest that cotton dust produces a significant parenchymal lesion with elements similar to both the emphysematous response to endotoxin and the fibrotic nodular response to cellulose. AM REV RESPIR DIS 1990; 142:184-192
Prausnitz (22) who concluded that "cotton dust has a great power ofpenetrating into the deepest parts of the lungs and ofproducing in them very extensive irritative changes." There is no evidence that the character of the human response to cotton dust is similar to the alveolar thickening in guinea pigs. And guinea pigs may be inappropriate as a model for parenchymal effects in humans because of their high proteinase inhibitor levels (23). Yet, the presence of parenchymal changes substantiates the need to consider the effects of alveolar deposition of respirable cotton dust particles. Cellulose constitutes a small and variable fraction of the mass of cotton dust. The American Conference of Governmental and Industrial Hygienists (ACGIH) (24) asserts that airborne cellulose is nontoxic. Aerosolized, respirable cellulose was used as an "inert" control in a study of acute ventilatory responses of guinea pigs to cotton dust (2). Although cellulose had little acute ventilatory effect, no morphologic or chronic exposure data were presented to substantiate that cellulose was merely a nuisance dust. This background raisesthree questions that we will address here: Can cotton dust produce a destructive parenchymal le-
sion? If so, can the lesion be attributed to endotoxin or to elastase in the dust? Is cellulose an appropriate control dust? We studied whether a subchronic exposure to cotton dust via intratracheal instillation could produce emphysema in hamsters. Wechose to study hamsters be-
(Received in original form June 26, 1989 and in revised form November 27, 1989) 1 From the Departments of Environmental Science and Physiology, Harvard School of Public Health, and the Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts. 2 Supported by Occupational and Environmental Health Center Grant 2 P30 ES-OOOO2 from the National Institutes of Environmental Health Sciences. 3 A preliminary report of the data in this article was published in Cotton Dust: Proceedings of the Twelfth Cotton Dust Research Conference, published by the National Cotton Council, Memphis, Tennessee, 1988. 4 Correspondence and requests for reprints should be addressed to Dr. Donald K. Milton, Occupational Health Program, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115. 5 Recipient of National Institute of Environmental Health Sciences National Research Service Award 5T32 ES07069 from the Harvard School of Public Health. This work was completed in partial fulfillment of the degree of Doctor of Public Health.
185
TOXICITY OF INTRATRACHEALLY INSTILLED COTTON DUST, CELWLOSE, AND ENDOTOXIN
cause their blood n-l-proteinase inhibitor levels more closely approximate human levels (23) and they are known to be sensitive to instilled elastase (25). Additionally, hamsters have been reported to be insensitive to agents expected to produce emphysema by neutrophil recruitment, possibly because of low leukocyte elastase levels (26). Thus, we expected that hamsters would be an appropriate species to test for the effects of exogenous elastase as distinct from endogenous elastase released in response to inflammatory agents. We tested whether an inflammatory agent in the absence of exogenous elastase could produce emphysema, and thus whether a variety of inflammatory agents in cotton dust might contribute to emphysema by instilling high doses of endotoxin. We also tested whether cellulose is an appropriate nontoxic comparison material for cotton dust by instilling equal doses of cellulose and cotton dust. The dust doses chosen approximate the expected deposition for hamsters exposed to 16mg/rn" for 5 days a week or to 40 mg/m" for 6 h twice a week. Methods
Respirable Particulates and Endotoxin The USDA Cotton Quality Research Station in Clemson, SC, supplied dust generated during carding of low-middling grade cotton (from the Memphis, TN area), which is representative of the average cotton grown in the United States (MQ170). Total dust was collected on Zefluor 3.0-llm filters, stored at - 20° C, and shipped on dry ice. Cellulose dust was Whatman CC41 microgranular cellulose (Whatman, Clifton, NJ), a thin-layer chromatography adsorbent produced from cotton fiber and provided as a powder. Respirable samples of cotton dust and cellulose (mass median aerodynamic diameters, 4.2 and 4.8 urn, respectively) were obtained by generating an aerosol from the filter cake or bulk powder with a fluidized bed generator patterned after the Pitt 3 generator (27). Large particles were removed with a vertical elutriator (diameter, 10inches; length, 9 feet). Particle sizedistribution of the elutriated aerosol was evaluated with an Andersen Mark II cascade impactor and light microscope. Polytetrafluoroethylene filters (pore size, 3.0 urn) wereused for gravimetric measurements. Bulk respirable samples were collected on 0.45-llm cellulose mixed-ester membrane filters and stored at - 20° C. The endotoxin for these experiments was from Escherichiacoli 0111 :B4 (Sigma Chemical, St. Louis, MO). Instillation Protocol Male Syrian Golden hamsters were exposed to endotoxin, saline, suspensions of respira-
ble cotton dust, or cellulose by intratracheal instillation according to the method of Brain and coworkers (28). Bulk samples of respirable celluloseand cotton dust, described above, weresuspended in pyrogen-free saline by bath sonication and examined with light microscopy to verify dispersion. Hamsters were obtained from Harlan Sprague-Dawley(Indianapolis, IN) in two lots and allowed to acclimatize for 2 wk. Eight animals, four from each lot, were assigned to each of four exposure groups and were instilled twice per week for 6 wk. Starting and ending dates for the two lots were separated by 1 wk. Surviving animals were killed 8 wk after the last instillation. A hiatus between exposure and examination of the lungs was introduced to allow any acute pathologic changes to subside. The doses of cellulose and cotton dust (0.75 mg/100-g animal) were derived from preliminary experiments to ensure survival and absence of major fungal infections while being at the upper end of the expected dust dose per day in aerosol inhalation studies. The dose of endotoxin (225 Ilg/100-g animal) was chosen to be similar to those in experimental studies of repeated intravenous injection in rats (16), rather than to approximate the unknown total amount of endotoxin in 0.75 mg of cotton dust. High dose endotoxin, because it recruits neutrophils and causes emphysema in some species, was included as a test of whether hamsters can develop emphysema from endogenous elastase alone. Respirable cellulose particles were included to test whether cellulose is an appropriate inert comparison material for studies of cotton dust's toxicity. Cellulose constitutes about 7 to 20% of cotton dust and has been thought an inert substance without respiratory effects (2,24). A group instilled with saline served as the negative control.
Physiologic Measurements Hamsters wereanesthetized with sodium pentobarbital (0.7 mg/g). When the animal became unresponsive, the trachea was cannulated, and the lungs were excised and floated on a saline bath at room temperature. Each animal's lungs were connected to a controlled volume system for inflation with air and recording lung pressure-volume (P-V) relationships. The system was based on the design described by Smith and Mitzner (29) modified for small animal lungs by decreasing the water column diameter and using handheld syringes for calibration and inflation. Pressure and volume signals were recorded on an x-y plotter and digitized with a Bit Pad One (Summagraphics Corp., Fairfield, CT) for computer analysis. Lungs were consistently inflated to 30 cm H 2 0 over 10 s and held at this maximal inflation volume (Vmax) for 30 s, by which time the elastic recoil pressure had reached a steady value of 23 to 25 em H 2O. The lung was then deflated in 1 to 2 cm H 2 0 steps with a 5-s interval between the end of each deflation and the recording of
volume and static recoil pressure. Inflationdeflation cycles were repeated until two or three reproducible curves were obtained. The gas volume remaining in the lungs after the final deflation was measured by displacement of saline. All lung gas volumes were corrected for the compressibility of gas in the system and in the lung. Static deflation of hamster lungs produced an inflection in the P-V curves as described by Smith and Stamenovic (30). Presence of this inflection prevented analysis of the P-V curves with a simple exponential function that has been described by Haber and coworkers (31). Toanalyze further these data, wedevised an empirical mathematical model to fit the deflation curve, and for consistency with the approach of Haber and coworkers, we fitted this model over the upper 50% of lung volume. Our model has exponential and sigmoid terms of the form: V = RV + Vt [Vtf (11(1
-
+
[Vt (1-f)e- kP] (p/Po)b»]
-
where V is the volume observed at pressure P, RV + Vt is an asymptotic maximal volume that does not correspond directly to the measured Vmax, Vt is a volume of deflation extrapolated to P = 0, f describes the proportion of the curve fit by the sigmoid component, k and b describe the shape or steepness of the curve, and Po represents the location of the center of the sigmoid component. Parameter estimates were obtained by the method of least squares (32). The error about the curve was extremely small (typical standard error of estimate ~ 0.02 mllcm H 2 0 ). Therefore, mean curves for each group were based on a summary of calculated volumes for each one em H 2 0 step from 23 to 5 em H 2 0 computed for each P-V curve from the fitted parameters. Mean curves were computed taking into account the number of curves per animal and the number of animals per group using the LSMEANS option of the SASGLM procedure. Lung distensibility (oV/OP)/Vt 1c was calculated from the chord compliance between 60 and 70070 ofTLC and the exact distensibility at these volumes from the first derivative of the fitted equation for each curve. Lung distensibility rather than compliance (oV/oP) was computed to adjust for effects of different lung sizeson the measurement of compliance.
Morphologic Measurements After measurement of the P-V curves, the lungs were inflation-fixed with 2.5% glutaraldehyde in 0.03 M phosphate buffer at pH 7.4. Fixed lung volume was measured by saline displacement after removal of the attached mediastinal structures and diaphragm. Two parasagittal sections from the left lung and two from the right basal lobe were mounted in paraffin and stained with hematoxylineosin. Morphometric measurements of parenchymal tissue density and surface density were made at a magnification x 110using a
186
MILTON, GODLESKI, FELDMAN, AND GREAVES
Zeiss microscope equipped with a projection screen and a grid with 42 intersections. Morphometric densities were converted to absolute volumes and surface area by reference to fixed lung volumes obtained before dehydration (33). Although light microscopy underestimates surface area and overestimates tissue density relative to electron microscopy, it is adequate for comparisons with compliance data (34). Glutaraldehyde inflation fixes lungs at approximately 70 to 80% of TLC (35). This was confirmed in the present study by comparison of Vmax (after inflation with air to 30 em H 20) with the total fixed lung volume and with the morphometrically determined air space. However, lung inflation with aqueous media removes the surface forces that otherwise minimize surface area by reversible folding of the alveolar septa below TLC (36). The surface area obtained from our morphometric measurements truly reflects total surface area and is greater than the surface exposed in equally inflated air-filled lungs. Therefore, a surface-to-volume (S/V) ratio using the fixed lung volumes is an overestimate of the physiologic S/V ratio in air-filled lungs. To compensate for this in our analyses, we calculated the SIV ratio from the physiologically determined TLC (Vmax) rather than from the volume of the fixed lung.
Statistical Analyses Group comparisons of distensibility, surface density, surface area, tissue density, and tissue volume were made by analysis of variance (ANOVA) for a nested design (each animal belonged to one batch and each batch was divided among the four exposures) with unbalanced data (unequal numbers of survivors) using the SAS-PC GLM procedure (version 6.03 SAS Institute, 1988). For physiologic parameters with more than one observation per animal, the appropriate combination of the mean square error estimates of the model and the intra-animal variance was used as the denominator for F-tests of effects. All reported means were adjusted for the possible effects of different numbers of curves per animal, animals per group, and for possible batch effects by the least-squares means method. When the exposure effect was significant, the three experimental groups were individually contrasted with the saline-exposed control group by t tests for unpaired samples. We controlled for multiple comparisons by only examining the prior hypothesis of difference from saline control, and by only performing t tests when an overall exposure effect was significant. Calculations were performed on an IBM-compatible computer (pC's Limited model 286 12 ; Dell Computers, Austin, TX).
Results
There were no significant differences between the initial mean weights of animals in the four exposure groups. Endotoxin-
7.0
Fig. 1. Group mean static deflation curves for exposure groups. Curves were based on a summary of volumes calculated for each P-V curve at 1 cm H20 intervals from 23 to 5 cm H2 0 using the fitted equation (typical standard error of estimate, 0.02ml/cm H2 0 ) for each curve. Group summary curves were constructed from the least squares means of the calculated volumes, allowing for different numbers of curves per animal and animals per group. Error bars at Vmax are the SEM for highest and lowest groups.
6.0 VOLUME 50 (ml) .
4.0
3.0 5
10
15
25
20
STATIC RECOIL PRESSURE (em H
20)
exposed animals weighed significantly less than did animals in the other groups at the end of the exposure period. By the end of the study, however, mean weights were not significantly different among the exposure groups. Animals were acquired in two lots, which were different in their initial weights but not in their final weights or in the effects of exposures. Eight animals wereassigned to each of the saline and cellulose groups, and nine were assigned to each of the endotoxin and cotton dust groups. All spontaneous deaths occurred while the animals were anesthetized; two each in the saline (doses 1, 11), endotoxin (doses 7, 11), and cellulose (doses 2, 3) groups and one in the cotton dust group (dose 1). Instillation was not performed if an animal lost more than 10 g in weight after the previous instillation or if it appeared to have respiratory difficulty. Three endotoxin (two for one animal), one cellulose, and one cotton dust dose were withheld. All four of the animals with missing doses survived and were included in the subsequent analyses. Of the sur-
viving animals, one from each of the saline control, cellulose, and cotton dust groups were excluded from analysis of the physiologic data because of excessive leakage or poor quality of the P-V tracings. One additional cellulose-exposed animal was excluded from the physiologic analysis because of gross air trapping. One cotton-dust-exposed animal was excluded from morphometric analysis because of extensive atelectasis. Each animal included in the physiologic analyses, after the exclusions described above, had two or three acceptable static deflation curves. The group mean static deflation curves (figure 1), adjusted for the number of animals and number of curves per animal, demonstrate a small shift to higher volumes of the curves for cotton dust and endotoxin groups, with cotton dust approaching the curves for saline control animals at low pressures. The curve for the celluloseexposed group is very close to that for the saline control near Vmax, but shifted downward at lower pressures. There were no differences among the groups in
TABLE 1 PHYSIOLOGIC MEASUREMENTS· Cotton Dust
Saline Number Hamsters P-V curves (total) Volume Vmax, ml Vmin, ml Distensibility (x 10-2 cm H2O-1) Vmax, 60% Chord,60-70% Vmax, 70%
5 13
8 20
Cellulose
4 11
Endotoxin
7 17
5.9 (0.38) 0.74 (0.07)
6.6 (0.34) 0.64 (0.06)
5.9 (0.42) 0.62 (0.08)
6.3 (0.34) 0.62 (0.06)
0.77 0.95
7.5 (0.43) 5.5 (0.35) 3.9 (0.30)
7.3 (0.35) 5.5 (0.29) 4.1 (0.24)
6.1 (0.48)§ 4.5 (0.40) 3.3 (0.33)
8.1 (0.39) 6.3 (0.32) 4.9 (0.27)§
4.50:f: 3.50:f: 5.14:f:
Definition of abbreviations: Vmax = total lung capacity from physiologic measurements; Vmin = trapped air volume measured by displacement; chord = distensibility from chord compliance between 60 and 70% of Vmax; distensibility = compliance (slope of P-V curve) divided by Vmax. • Values are adjusted mean with SEM shown in parentheses. t F statistic from ANOVA for homogeneity of means. :j: Significant differences found in test for homogeneity of exposure group means: p < 0.05. § Significantly different in test for equality with mean of saline control group: p < 0.05.
TOXICITY OF INTRATRACHEALLY INSTILLED COTTON DUST, CELWLOSE, AND ENDOTOXIN
the maximum or the minimum (trapped) air volumes (table 1). Distensibilities calculated at 60 and 70070 of Vmax and from the chord compliance between 60 and 70% of Vmax all had significant differences among the exposure group means (table 1). Comparison of lung distensibilities of the exposure groups with those of the saline control group shows that the lungs of the cellulose group were consistently and significantly less distensible than were the saline group's lungs. The lungs of the endotoxin group were consistently and significantly more distensible than were the salinegroup's lungs. The cotton dust group was not significantly different from the saline control group on any of the distensibility measures. Statistically significant differences in distensibility, when present, occurred at 60 or 70% Vmax and not at higher lung volumes. Histologic examination (figure 2A-D) showed that lungs of cellulose- and cotton-dust-exposed animals contained noncaseating granulomata. Cellulose could be demonstrated in the granulomatous areas by polarized light. No evidence of caseous necrosis was present in the granulomata. Endotoxin-exposed animals had thickened pleura with chronic inflammatory infiltrates and areas of centrilobular emphysema. Cotton-dust-exposed lungs also had centrilobular emphysema as well as granulomata similar to those seen in cellulose-exposed animals. Point and intercept counting for morphometry was performed on all acceptable specimens with the exclusions described above. Conversion of volume and surface density measurements to volumes and surface area required exclusion of a saline-exposed hamster for which the fixed volume could not be estimated because of ligation of a leaking lobe. There were no differences in the total points, intercepts, or areas counted among the groups (table 2). Volume densities of the nonparenchyrna in lung and of the cellular and connective tissue elements in the parenchyma weresignificantly different among exposure groups. All of the experimental groups had higher volume densities of nonparenchyma than did the saline control group, but only the endotoxin group was significantly higher. The density of cellular and tissue elements was significantly elevated in the cellulose group but equal to saline control in both the cotton dust and endotoxin groups. Volume densities were converted to absolute vol-
urnes by multiplying with the displacement volume of the fixed lung (33). There wereno significant differences among the mean volumes for air space or cellular and tissue elements in the parenchyma. The volume of nonparenchyma was increased in all experimental groups, but the increase was significant only in the endotoxin-exposed hamsters. Surface density of alveoli in the parenchyma was significantly different among the exposure groups. All of the experimental groups had significantly lower surface density than did the saline control group (table 2). Total alveolar surface area was slightly lower in the experimental groups than in the control group; however, there were no significant exposure effects. As noted above, fixation resulted in under inflation. There was no effect of exposure on the ratio of displacement volume of the fixed lung to Vmax in either group. The ratio of parenchymal air space to physiologic Vmax suggests that glutaraldehyde inflation was 66 to 71070 complete in accordance with the results of Hayatdavoudi and coworkers (35). We therefore calculated the ratio of total surface area to physiologic Vmax and observed a significant exposure effect on the S/V ratio; both the endotoxin and cotton dust groups had significantly lower S/V ratios than did the saline control group, whereas the cellulose group was similar to the saline control group (table 2). In summary, endotoxin-instilled hamsters had increased lung distensibility, nonparenchymal tissue volume, and decreased S/V ratio; these results are consistent with the histologic appearance of mild centrilobular emphysema and chronic inflammatory changes in the pleura. The cellulose-treated animals had decreased lung distensibility, noncaseating granulomata, and increased volume density of parenchymal tissue elements (hallmarks of the histologically apparent lung fibrosis); the S/V ratio was normal. Cotton dust produced an intermediate lesion between those seen for endotoxin and cellulose:measurements of lung distensibility and the volume of parenchymal tissue elements among the cotton-dust-exposed hamsters wereunchanged from the control, but the S/V ratio was significantly decreased, and noncaseating granulomata were present. Discussion
The present findings show that repeated
187
intratracheal instillations of endotoxin produced functional and morphologic evidence of a mild emphysematous lesion in hamsters. Boudier and Bieth (26) reported that Syrian Golden hamsters had low leukocyte elastase levels and offered this as an explanation for earlier experiments, which had not been able to produce emphysema with a chemotactic peptide. Our success with this protocol may be due to the secretory activity of neutrophils recruited by endotoxin, recruitment or activation of macrophages, inhibition of repair processes by endotoxin, or to differences in the timing and duration of the exposures (16). These data suggest that hamsters have a delicate balance between proteolytic and antiproteolytic factors in the lung. Hamsters are thus well suited for studying the effects of elastolysis by endogenous as well as exogenous proteinases. In contrast to the findings in rats (16), and more in keeping with the results from dogs (15), we demonstrated both a decrease in surface to volume ratio and mild physiologic impairment from repeated doses of endotoxin without manipulation of u-l-proteinase inhibitor levels. This difference between studies may be due to species differences because hamsters and dogs have lower levels of u-l-proteinase than do rats (23). Alternatively, the difference may derive from the use of more sensitivephysiologic measurements by Guenter's group (15) and in the present study. Long-term experimental exposures to aerosols of endotoxin in rabbits (37) and guinea pigs (38) resulted in pathologic changes in airways after brief daily exposures to high concentrations of endotoxin. Cavagna and colleagues (37) found airway disease and alveolar thickening but no change in baseline compliance among four animals exposed for 5 wk. Rylander (38) reported airway changes but did not comment on alveolar pathology. A study of short-term inhalation exposure of hamsters at levels thought to represent "realistic" environmental levels of endotoxin demonstrated significant changes in the alveoli (39). The present study suggests that over a longer period the acute parenchymal effects of endotoxin may lead to emphysema. Further studies are required to determine what chronic lesion may be produced in hamsters after long-term inhalation of lower concentrations of endotoxin. Hamsters and rabbits offer an advantage for models of exposure to endotoxin in that they have
188
MILTON, GODLESKI , FELDMAN, AND GREAVES
/> ;
•
,.~ . . j
, "'"
.-. .
•
..
~
;
ID
J
.( .......
.' I
JI'
~
•
V
DONALD K. MILTON,s JOHN J. GODLESKI, HENRY A. FELDMAN, and IAN A. GREAVES
Introduction
Cotton dust consists of inhalable (~ 15 urn in diameter) and respirable (~3Ilm) particles, with as much as 170/0 of the total particulate mass having diameters ~ 2.5 urn (1). Aerosols of cotton dust are composed of plant constituents, including cellulose, and particles derived from microbial flora of the cotton plant that include endotoxins and proteinases. Recent attention has focused on the airway effects of endotoxin in cotton dust (2, 3); however, because cotton dust contains considerable amounts of fine respirable material, parenchymal effects should also be considered. Chronic airflow obstruction may result from prolonged cotton dust exposure in some persons (4). Such obstruction could theoretically result from intrinsic airway narrowing or from emphysema or from a combination of both conditions (5). Epidemiologic studies of airflow obstruction among cotton mill workerscould not distinguish between these possible mechanisms (6), and postmortem studies (7-9) have had serious flaws in study design, data, and analysis (10-12) that may have obscured the mechanisms of airflow obstruction attributable to cotton dust. Two of the potentially toxic agents in cotton dust, elastase and endotoxin, have been linked to emphysema. Elastase, recently reported in cotton dust (13), is thought to playa central role in the development of emphysema (14). Endotoxin, long recognized as a constituent of cotton dust, also has been linked to emphysema in animal models of intravascular leukocyte sequestration (15, 16). Animal models of cotton dust toxicity (2, 17-20) have concentrated on guinea pig airway responses, but changes also occur at the alveolar level. Morphometric analysis of guinea pig lungs after the inhalation of high levels of respirable cotton dust for 1 yr has demonstrated both airway and parenchymal changes (21). The thickened alveolar septa were consistent also with the earlier findings of 184
SUMMARY Cotton dust Includes respirable particles containing endotoxin and elastase, agents associated with emphysema. Toexamine whether a respirable fraction of cotton dust could produce emphysema in an animal model, we Intratracheally Instilled hamsters with respirable cotton dust particles (0.75 mgl100-g animal), mass median aerodynamic diameter ~ 4.8 11m, twice weekly for 6 wk. We also examined whether Instilled endotoxin (225 I1gI1OO-g animal) could produce emphysema In hamsters and whether cellulose (0.75 mg/100-g animal) Is an appropriate Inert comparison dust. A saline-Instilled group was the control. Hamsters were killed 8 wk after the last Instillation. Static pressure-volume deflation curves of air-filled excised lungs were analyzed to measure lung distensibility. Lungs were fixed In Inflation using glutaraldehyde and were examined morphometrlcally to obtain surface area and numbers of granulomata. Endotoxin-treated animals had Increased distensibility, reduced surface-to-volume (SN) ratio, and morphologically apparent mild centrllobular emphysema. Cellulose-treated animals had decreased distensibility, normal SN ratio, and significant numbers of granulomata with patchy areas of thickened interalveolar septa. Cotton-dustinstilled animals had normal distensibility, reduced S/V ratio, significant numbers of granulomata, and mild centrllobular emphysema. These data suggest that cotton dust produces a significant parenchymal lesion with elements similar to both the emphysematous response to endotoxin and the fibrotic nodular response to cellulose. AM REV RESPIR DIS 1990; 142:184-192
Prausnitz (22) who concluded that "cotton dust has a great power ofpenetrating into the deepest parts of the lungs and ofproducing in them very extensive irritative changes." There is no evidence that the character of the human response to cotton dust is similar to the alveolar thickening in guinea pigs. And guinea pigs may be inappropriate as a model for parenchymal effects in humans because of their high proteinase inhibitor levels (23). Yet, the presence of parenchymal changes substantiates the need to consider the effects of alveolar deposition of respirable cotton dust particles. Cellulose constitutes a small and variable fraction of the mass of cotton dust. The American Conference of Governmental and Industrial Hygienists (ACGIH) (24) asserts that airborne cellulose is nontoxic. Aerosolized, respirable cellulose was used as an "inert" control in a study of acute ventilatory responses of guinea pigs to cotton dust (2). Although cellulose had little acute ventilatory effect, no morphologic or chronic exposure data were presented to substantiate that cellulose was merely a nuisance dust. This background raisesthree questions that we will address here: Can cotton dust produce a destructive parenchymal le-
sion? If so, can the lesion be attributed to endotoxin or to elastase in the dust? Is cellulose an appropriate control dust? We studied whether a subchronic exposure to cotton dust via intratracheal instillation could produce emphysema in hamsters. Wechose to study hamsters be-
(Received in original form June 26, 1989 and in revised form November 27, 1989) 1 From the Departments of Environmental Science and Physiology, Harvard School of Public Health, and the Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts. 2 Supported by Occupational and Environmental Health Center Grant 2 P30 ES-OOOO2 from the National Institutes of Environmental Health Sciences. 3 A preliminary report of the data in this article was published in Cotton Dust: Proceedings of the Twelfth Cotton Dust Research Conference, published by the National Cotton Council, Memphis, Tennessee, 1988. 4 Correspondence and requests for reprints should be addressed to Dr. Donald K. Milton, Occupational Health Program, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115. 5 Recipient of National Institute of Environmental Health Sciences National Research Service Award 5T32 ES07069 from the Harvard School of Public Health. This work was completed in partial fulfillment of the degree of Doctor of Public Health.
185
TOXICITY OF INTRATRACHEALLY INSTILLED COTTON DUST, CELWLOSE, AND ENDOTOXIN
cause their blood n-l-proteinase inhibitor levels more closely approximate human levels (23) and they are known to be sensitive to instilled elastase (25). Additionally, hamsters have been reported to be insensitive to agents expected to produce emphysema by neutrophil recruitment, possibly because of low leukocyte elastase levels (26). Thus, we expected that hamsters would be an appropriate species to test for the effects of exogenous elastase as distinct from endogenous elastase released in response to inflammatory agents. We tested whether an inflammatory agent in the absence of exogenous elastase could produce emphysema, and thus whether a variety of inflammatory agents in cotton dust might contribute to emphysema by instilling high doses of endotoxin. We also tested whether cellulose is an appropriate nontoxic comparison material for cotton dust by instilling equal doses of cellulose and cotton dust. The dust doses chosen approximate the expected deposition for hamsters exposed to 16mg/rn" for 5 days a week or to 40 mg/m" for 6 h twice a week. Methods
Respirable Particulates and Endotoxin The USDA Cotton Quality Research Station in Clemson, SC, supplied dust generated during carding of low-middling grade cotton (from the Memphis, TN area), which is representative of the average cotton grown in the United States (MQ170). Total dust was collected on Zefluor 3.0-llm filters, stored at - 20° C, and shipped on dry ice. Cellulose dust was Whatman CC41 microgranular cellulose (Whatman, Clifton, NJ), a thin-layer chromatography adsorbent produced from cotton fiber and provided as a powder. Respirable samples of cotton dust and cellulose (mass median aerodynamic diameters, 4.2 and 4.8 urn, respectively) were obtained by generating an aerosol from the filter cake or bulk powder with a fluidized bed generator patterned after the Pitt 3 generator (27). Large particles were removed with a vertical elutriator (diameter, 10inches; length, 9 feet). Particle sizedistribution of the elutriated aerosol was evaluated with an Andersen Mark II cascade impactor and light microscope. Polytetrafluoroethylene filters (pore size, 3.0 urn) wereused for gravimetric measurements. Bulk respirable samples were collected on 0.45-llm cellulose mixed-ester membrane filters and stored at - 20° C. The endotoxin for these experiments was from Escherichiacoli 0111 :B4 (Sigma Chemical, St. Louis, MO). Instillation Protocol Male Syrian Golden hamsters were exposed to endotoxin, saline, suspensions of respira-
ble cotton dust, or cellulose by intratracheal instillation according to the method of Brain and coworkers (28). Bulk samples of respirable celluloseand cotton dust, described above, weresuspended in pyrogen-free saline by bath sonication and examined with light microscopy to verify dispersion. Hamsters were obtained from Harlan Sprague-Dawley(Indianapolis, IN) in two lots and allowed to acclimatize for 2 wk. Eight animals, four from each lot, were assigned to each of four exposure groups and were instilled twice per week for 6 wk. Starting and ending dates for the two lots were separated by 1 wk. Surviving animals were killed 8 wk after the last instillation. A hiatus between exposure and examination of the lungs was introduced to allow any acute pathologic changes to subside. The doses of cellulose and cotton dust (0.75 mg/100-g animal) were derived from preliminary experiments to ensure survival and absence of major fungal infections while being at the upper end of the expected dust dose per day in aerosol inhalation studies. The dose of endotoxin (225 Ilg/100-g animal) was chosen to be similar to those in experimental studies of repeated intravenous injection in rats (16), rather than to approximate the unknown total amount of endotoxin in 0.75 mg of cotton dust. High dose endotoxin, because it recruits neutrophils and causes emphysema in some species, was included as a test of whether hamsters can develop emphysema from endogenous elastase alone. Respirable cellulose particles were included to test whether cellulose is an appropriate inert comparison material for studies of cotton dust's toxicity. Cellulose constitutes about 7 to 20% of cotton dust and has been thought an inert substance without respiratory effects (2,24). A group instilled with saline served as the negative control.
Physiologic Measurements Hamsters wereanesthetized with sodium pentobarbital (0.7 mg/g). When the animal became unresponsive, the trachea was cannulated, and the lungs were excised and floated on a saline bath at room temperature. Each animal's lungs were connected to a controlled volume system for inflation with air and recording lung pressure-volume (P-V) relationships. The system was based on the design described by Smith and Mitzner (29) modified for small animal lungs by decreasing the water column diameter and using handheld syringes for calibration and inflation. Pressure and volume signals were recorded on an x-y plotter and digitized with a Bit Pad One (Summagraphics Corp., Fairfield, CT) for computer analysis. Lungs were consistently inflated to 30 cm H 2 0 over 10 s and held at this maximal inflation volume (Vmax) for 30 s, by which time the elastic recoil pressure had reached a steady value of 23 to 25 em H 2O. The lung was then deflated in 1 to 2 cm H 2 0 steps with a 5-s interval between the end of each deflation and the recording of
volume and static recoil pressure. Inflationdeflation cycles were repeated until two or three reproducible curves were obtained. The gas volume remaining in the lungs after the final deflation was measured by displacement of saline. All lung gas volumes were corrected for the compressibility of gas in the system and in the lung. Static deflation of hamster lungs produced an inflection in the P-V curves as described by Smith and Stamenovic (30). Presence of this inflection prevented analysis of the P-V curves with a simple exponential function that has been described by Haber and coworkers (31). Toanalyze further these data, wedevised an empirical mathematical model to fit the deflation curve, and for consistency with the approach of Haber and coworkers, we fitted this model over the upper 50% of lung volume. Our model has exponential and sigmoid terms of the form: V = RV + Vt [Vtf (11(1
-
+
[Vt (1-f)e- kP] (p/Po)b»]
-
where V is the volume observed at pressure P, RV + Vt is an asymptotic maximal volume that does not correspond directly to the measured Vmax, Vt is a volume of deflation extrapolated to P = 0, f describes the proportion of the curve fit by the sigmoid component, k and b describe the shape or steepness of the curve, and Po represents the location of the center of the sigmoid component. Parameter estimates were obtained by the method of least squares (32). The error about the curve was extremely small (typical standard error of estimate ~ 0.02 mllcm H 2 0 ). Therefore, mean curves for each group were based on a summary of calculated volumes for each one em H 2 0 step from 23 to 5 em H 2 0 computed for each P-V curve from the fitted parameters. Mean curves were computed taking into account the number of curves per animal and the number of animals per group using the LSMEANS option of the SASGLM procedure. Lung distensibility (oV/OP)/Vt 1c was calculated from the chord compliance between 60 and 70070 ofTLC and the exact distensibility at these volumes from the first derivative of the fitted equation for each curve. Lung distensibility rather than compliance (oV/oP) was computed to adjust for effects of different lung sizeson the measurement of compliance.
Morphologic Measurements After measurement of the P-V curves, the lungs were inflation-fixed with 2.5% glutaraldehyde in 0.03 M phosphate buffer at pH 7.4. Fixed lung volume was measured by saline displacement after removal of the attached mediastinal structures and diaphragm. Two parasagittal sections from the left lung and two from the right basal lobe were mounted in paraffin and stained with hematoxylineosin. Morphometric measurements of parenchymal tissue density and surface density were made at a magnification x 110using a
186
MILTON, GODLESKI, FELDMAN, AND GREAVES
Zeiss microscope equipped with a projection screen and a grid with 42 intersections. Morphometric densities were converted to absolute volumes and surface area by reference to fixed lung volumes obtained before dehydration (33). Although light microscopy underestimates surface area and overestimates tissue density relative to electron microscopy, it is adequate for comparisons with compliance data (34). Glutaraldehyde inflation fixes lungs at approximately 70 to 80% of TLC (35). This was confirmed in the present study by comparison of Vmax (after inflation with air to 30 em H 20) with the total fixed lung volume and with the morphometrically determined air space. However, lung inflation with aqueous media removes the surface forces that otherwise minimize surface area by reversible folding of the alveolar septa below TLC (36). The surface area obtained from our morphometric measurements truly reflects total surface area and is greater than the surface exposed in equally inflated air-filled lungs. Therefore, a surface-to-volume (S/V) ratio using the fixed lung volumes is an overestimate of the physiologic S/V ratio in air-filled lungs. To compensate for this in our analyses, we calculated the SIV ratio from the physiologically determined TLC (Vmax) rather than from the volume of the fixed lung.
Statistical Analyses Group comparisons of distensibility, surface density, surface area, tissue density, and tissue volume were made by analysis of variance (ANOVA) for a nested design (each animal belonged to one batch and each batch was divided among the four exposures) with unbalanced data (unequal numbers of survivors) using the SAS-PC GLM procedure (version 6.03 SAS Institute, 1988). For physiologic parameters with more than one observation per animal, the appropriate combination of the mean square error estimates of the model and the intra-animal variance was used as the denominator for F-tests of effects. All reported means were adjusted for the possible effects of different numbers of curves per animal, animals per group, and for possible batch effects by the least-squares means method. When the exposure effect was significant, the three experimental groups were individually contrasted with the saline-exposed control group by t tests for unpaired samples. We controlled for multiple comparisons by only examining the prior hypothesis of difference from saline control, and by only performing t tests when an overall exposure effect was significant. Calculations were performed on an IBM-compatible computer (pC's Limited model 286 12 ; Dell Computers, Austin, TX).
Results
There were no significant differences between the initial mean weights of animals in the four exposure groups. Endotoxin-
7.0
Fig. 1. Group mean static deflation curves for exposure groups. Curves were based on a summary of volumes calculated for each P-V curve at 1 cm H20 intervals from 23 to 5 cm H2 0 using the fitted equation (typical standard error of estimate, 0.02ml/cm H2 0 ) for each curve. Group summary curves were constructed from the least squares means of the calculated volumes, allowing for different numbers of curves per animal and animals per group. Error bars at Vmax are the SEM for highest and lowest groups.
6.0 VOLUME 50 (ml) .
4.0
3.0 5
10
15
25
20
STATIC RECOIL PRESSURE (em H
20)
exposed animals weighed significantly less than did animals in the other groups at the end of the exposure period. By the end of the study, however, mean weights were not significantly different among the exposure groups. Animals were acquired in two lots, which were different in their initial weights but not in their final weights or in the effects of exposures. Eight animals wereassigned to each of the saline and cellulose groups, and nine were assigned to each of the endotoxin and cotton dust groups. All spontaneous deaths occurred while the animals were anesthetized; two each in the saline (doses 1, 11), endotoxin (doses 7, 11), and cellulose (doses 2, 3) groups and one in the cotton dust group (dose 1). Instillation was not performed if an animal lost more than 10 g in weight after the previous instillation or if it appeared to have respiratory difficulty. Three endotoxin (two for one animal), one cellulose, and one cotton dust dose were withheld. All four of the animals with missing doses survived and were included in the subsequent analyses. Of the sur-
viving animals, one from each of the saline control, cellulose, and cotton dust groups were excluded from analysis of the physiologic data because of excessive leakage or poor quality of the P-V tracings. One additional cellulose-exposed animal was excluded from the physiologic analysis because of gross air trapping. One cotton-dust-exposed animal was excluded from morphometric analysis because of extensive atelectasis. Each animal included in the physiologic analyses, after the exclusions described above, had two or three acceptable static deflation curves. The group mean static deflation curves (figure 1), adjusted for the number of animals and number of curves per animal, demonstrate a small shift to higher volumes of the curves for cotton dust and endotoxin groups, with cotton dust approaching the curves for saline control animals at low pressures. The curve for the celluloseexposed group is very close to that for the saline control near Vmax, but shifted downward at lower pressures. There were no differences among the groups in
TABLE 1 PHYSIOLOGIC MEASUREMENTS· Cotton Dust
Saline Number Hamsters P-V curves (total) Volume Vmax, ml Vmin, ml Distensibility (x 10-2 cm H2O-1) Vmax, 60% Chord,60-70% Vmax, 70%
5 13
8 20
Cellulose
4 11
Endotoxin
7 17
5.9 (0.38) 0.74 (0.07)
6.6 (0.34) 0.64 (0.06)
5.9 (0.42) 0.62 (0.08)
6.3 (0.34) 0.62 (0.06)
0.77 0.95
7.5 (0.43) 5.5 (0.35) 3.9 (0.30)
7.3 (0.35) 5.5 (0.29) 4.1 (0.24)
6.1 (0.48)§ 4.5 (0.40) 3.3 (0.33)
8.1 (0.39) 6.3 (0.32) 4.9 (0.27)§
4.50:f: 3.50:f: 5.14:f:
Definition of abbreviations: Vmax = total lung capacity from physiologic measurements; Vmin = trapped air volume measured by displacement; chord = distensibility from chord compliance between 60 and 70% of Vmax; distensibility = compliance (slope of P-V curve) divided by Vmax. • Values are adjusted mean with SEM shown in parentheses. t F statistic from ANOVA for homogeneity of means. :j: Significant differences found in test for homogeneity of exposure group means: p < 0.05. § Significantly different in test for equality with mean of saline control group: p < 0.05.
TOXICITY OF INTRATRACHEALLY INSTILLED COTTON DUST, CELWLOSE, AND ENDOTOXIN
the maximum or the minimum (trapped) air volumes (table 1). Distensibilities calculated at 60 and 70070 of Vmax and from the chord compliance between 60 and 70% of Vmax all had significant differences among the exposure group means (table 1). Comparison of lung distensibilities of the exposure groups with those of the saline control group shows that the lungs of the cellulose group were consistently and significantly less distensible than were the saline group's lungs. The lungs of the endotoxin group were consistently and significantly more distensible than were the salinegroup's lungs. The cotton dust group was not significantly different from the saline control group on any of the distensibility measures. Statistically significant differences in distensibility, when present, occurred at 60 or 70% Vmax and not at higher lung volumes. Histologic examination (figure 2A-D) showed that lungs of cellulose- and cotton-dust-exposed animals contained noncaseating granulomata. Cellulose could be demonstrated in the granulomatous areas by polarized light. No evidence of caseous necrosis was present in the granulomata. Endotoxin-exposed animals had thickened pleura with chronic inflammatory infiltrates and areas of centrilobular emphysema. Cotton-dust-exposed lungs also had centrilobular emphysema as well as granulomata similar to those seen in cellulose-exposed animals. Point and intercept counting for morphometry was performed on all acceptable specimens with the exclusions described above. Conversion of volume and surface density measurements to volumes and surface area required exclusion of a saline-exposed hamster for which the fixed volume could not be estimated because of ligation of a leaking lobe. There were no differences in the total points, intercepts, or areas counted among the groups (table 2). Volume densities of the nonparenchyrna in lung and of the cellular and connective tissue elements in the parenchyma weresignificantly different among exposure groups. All of the experimental groups had higher volume densities of nonparenchyma than did the saline control group, but only the endotoxin group was significantly higher. The density of cellular and tissue elements was significantly elevated in the cellulose group but equal to saline control in both the cotton dust and endotoxin groups. Volume densities were converted to absolute vol-
urnes by multiplying with the displacement volume of the fixed lung (33). There wereno significant differences among the mean volumes for air space or cellular and tissue elements in the parenchyma. The volume of nonparenchyma was increased in all experimental groups, but the increase was significant only in the endotoxin-exposed hamsters. Surface density of alveoli in the parenchyma was significantly different among the exposure groups. All of the experimental groups had significantly lower surface density than did the saline control group (table 2). Total alveolar surface area was slightly lower in the experimental groups than in the control group; however, there were no significant exposure effects. As noted above, fixation resulted in under inflation. There was no effect of exposure on the ratio of displacement volume of the fixed lung to Vmax in either group. The ratio of parenchymal air space to physiologic Vmax suggests that glutaraldehyde inflation was 66 to 71070 complete in accordance with the results of Hayatdavoudi and coworkers (35). We therefore calculated the ratio of total surface area to physiologic Vmax and observed a significant exposure effect on the S/V ratio; both the endotoxin and cotton dust groups had significantly lower S/V ratios than did the saline control group, whereas the cellulose group was similar to the saline control group (table 2). In summary, endotoxin-instilled hamsters had increased lung distensibility, nonparenchymal tissue volume, and decreased S/V ratio; these results are consistent with the histologic appearance of mild centrilobular emphysema and chronic inflammatory changes in the pleura. The cellulose-treated animals had decreased lung distensibility, noncaseating granulomata, and increased volume density of parenchymal tissue elements (hallmarks of the histologically apparent lung fibrosis); the S/V ratio was normal. Cotton dust produced an intermediate lesion between those seen for endotoxin and cellulose:measurements of lung distensibility and the volume of parenchymal tissue elements among the cotton-dust-exposed hamsters wereunchanged from the control, but the S/V ratio was significantly decreased, and noncaseating granulomata were present. Discussion
The present findings show that repeated
187
intratracheal instillations of endotoxin produced functional and morphologic evidence of a mild emphysematous lesion in hamsters. Boudier and Bieth (26) reported that Syrian Golden hamsters had low leukocyte elastase levels and offered this as an explanation for earlier experiments, which had not been able to produce emphysema with a chemotactic peptide. Our success with this protocol may be due to the secretory activity of neutrophils recruited by endotoxin, recruitment or activation of macrophages, inhibition of repair processes by endotoxin, or to differences in the timing and duration of the exposures (16). These data suggest that hamsters have a delicate balance between proteolytic and antiproteolytic factors in the lung. Hamsters are thus well suited for studying the effects of elastolysis by endogenous as well as exogenous proteinases. In contrast to the findings in rats (16), and more in keeping with the results from dogs (15), we demonstrated both a decrease in surface to volume ratio and mild physiologic impairment from repeated doses of endotoxin without manipulation of u-l-proteinase inhibitor levels. This difference between studies may be due to species differences because hamsters and dogs have lower levels of u-l-proteinase than do rats (23). Alternatively, the difference may derive from the use of more sensitivephysiologic measurements by Guenter's group (15) and in the present study. Long-term experimental exposures to aerosols of endotoxin in rabbits (37) and guinea pigs (38) resulted in pathologic changes in airways after brief daily exposures to high concentrations of endotoxin. Cavagna and colleagues (37) found airway disease and alveolar thickening but no change in baseline compliance among four animals exposed for 5 wk. Rylander (38) reported airway changes but did not comment on alveolar pathology. A study of short-term inhalation exposure of hamsters at levels thought to represent "realistic" environmental levels of endotoxin demonstrated significant changes in the alveoli (39). The present study suggests that over a longer period the acute parenchymal effects of endotoxin may lead to emphysema. Further studies are required to determine what chronic lesion may be produced in hamsters after long-term inhalation of lower concentrations of endotoxin. Hamsters and rabbits offer an advantage for models of exposure to endotoxin in that they have
188
MILTON, GODLESKI , FELDMAN, AND GREAVES
/> ;
•
,.~ . . j
, "'"
.-. .
•
..
~
;
ID
J
.( .......
.' I
JI'
~
•
V
