NATIONAL STATUS AND TRENDS PROGRAM
NATIONAL BENTHIC SURVEILLANCE PROJECT: PACIFIC COAST
Fish Histopathology and Relationships Between Toxicopathic Lesions and Exposure to Chemical Contaminants for Cycles I to V (1984-88)
Mark S. Myers, Carla M. Stehr, O. Paul Olson, Lyndal L. Johnson, Bruce B. McCain, Sin-Lam Chan, and Usha VaranasiNational Marine Fisheries Service
Northwest Fisheries Science Center
Environmental Conservation Division
2725 Montlake Blvd. E.
Seattle WA 98112
U.S. DEPARTMENT OF COMMERCE
Ronald H. Brown, Secretary
National Oceanic and Atmospheric Administration
National Marine Fisheries Service
This report presents and interprets the results of histopathology studies conducted on 17 species of bottomfish captured at 45 sites between 1984 and 1988 as part of the Pacific Coast portion of the National Benthic Surveillance Program (NBSP) in conjunction with NOAA's National Status and Trends Program (NS&T). The species examined histopathologically for potentially contaminant-induced (toxicopathic) lesions in the liver and kidney were fourhorn sculpin (Myoxocephalus quadricornis), Arctic flounder (Pleuronectes glacialis), flathead sole (Hippoglossoides elassodon), yellowfin sole (Pleuronectes asper), Pacific staghorn sculpin (Leptocottus armatus), English sole (Pleuronectes vetulus), starry flounder (Platichthys stellatus), white croaker (Genyonemus lineatus), hornyhead turbot (Pleuronichthys verticalis), barred sand bass (Paralabrax nebulifer), spotted sand bass (Paralabrax maculatofasciatus), spotted turbot (Pleuronichthys ritteri), diamond turbot (Hysopsetta guttulata), California tonguefish (Symphurus atricauda), black croaker (Cheilotrema saturnum), and California halibut (Paralichthys californicus). In all, over 5,250 individual fish among these species were examined.
The location of sampling sites ranged from Prudhoe Bay in Alaska to San Diego Bay in California, and incorporated sites in the following embayments or geographic regions. In Alaska, two sites were sampled in Prudhoe Bay, two sites in embayments of the Aleutian Islands, one site in the Chukchi Sea, Kamishak Bay, one site in Prince William Sound, and four sites in southeast Alaska; in Washington, three sites in Puget Sound; in Oregon, one site each in the Columbia River Estuary, Youngs Bay, and Coos Bay; in California, Humbolt Bay, Bodega Bay, seven sites in San Francisco Bay, Moss Landing, Monterey Bay, seven sites in the Los Angeles vicinity, Dana Point, two sites in or outside Mission Bay, and six sites in or outside San Diego Bay. Of these sites, 23 were located in or near urban embayments, and the 22 remaining sites were in nonurban embayments, five of which served as comparison or reference sites on the basis of minimal levels of sediment contaminants detected (Kamishak Bay, AK; Nisqually Reach, WA; and Bodega Bay, Dana Point, and outer Mission Bay, CA). These comparison sites were selected so that the same fish species could be collected from minimally contaminated environments as those obtained at urban or more contaminated sites.
Six primary target species were identified from the above species for statistical analyses that examined the relationships between occurrence of particular toxicopathic diseases or lesions in the liver and kidney and levels of potential or actual exposure to contaminants, as represented by measures of multiple chemical contaminant classes or their metabolites in sediments, stomach contents, liver tissue and bile collected from subsets of the same individual fish examined for histopathology. These species were selected because they were broadly distributed geographically and were abundant among the sampling sites, they had previously documented significant prevalences of contaminant-associated diseases, or the geographic distribution of similar diseases among the sites suggested an association with contaminant exposure. These primary target species were flathead sole, English sole, starry flounder, hornyhead turbot, white croaker, and black croaker.
All specimens were examined for the presence of necrotic, sclerotic and proliferative lesions in the kidney, as well as lesions in the liver that could be placed into the following categories:
Liver lesions within all of the above categories have been used in previous field or laboratory studies as histopathologic biomarkers of contaminant-induced effects in fish. Kidney lesions within the above categories in feral fish show some promise as histopathological biomarkers of contaminant-induced effects, but have been regarded as requiring additional field and laboratory testing for verification of their utility.
For the six primary target species, simple statistical evaluations of overall site-specific prevalences for each lesion category determined in field samplings conducted from 1984-88, as compared to lesion prevalences in the same species at reference or comparison sites, were made and presented graphically. Additionally, the relative risk of occurrence of these lesions in individual fish was calculated using logistic regression, while controlling for the influence of fish age and gender. Logistic regression is a multivariate statistical method similar to stepwise regression that is commonly used in epidemiological and epizootiological studies; it offers significant advantages over simple intersite comparisons of lesion prevalences because the method allows for simultaneous incorporation of, and determination of the influence of, important biological variables such as age and gender in analyses that evaluate the relationships between hepatic and renal disease risk and site of capture, or other parameters associated with contaminant exposure in individual fish.
The influence of contaminant exposure on hepatic and renal disease risk was evaluated for the following chemical classes measured in separate compartments: high molecular weight aromatic hydrocarbons containing 4-6 benzene rings (HAHs), low molecular weight aromatic hydrocarbons containing 1-3 benzene rings (LAHs), and total AHs (HAHs+LAHs) in sediments and stomach contents; polychlorinated biphenyls (PCBs), DDTs (DDT and its derivatives), hexachlorobenzene, chlordanes, dieldrin, and summed levels of two separate classes of covarying trace elements (copper, zinc, lead and tin, or Metals 1; nickel, chromium, and selenium, or Metals 2) in sediments, stomach contents, and liver tissue; and two classes of fluorescent aromatic compounds (FACs) in bile, representing metabolites of high molecular weight (FACs-H) and low molecular weight (FACs-L) aromatic hydrocarbons. These classes of xenobiotics were selected and measured because 1) they are found at high concentrations in sediments or fish from many of the sites sampled in the Pacific Coast portion of the NBSP, especially in embayments and estuaries located near major metropolitan centers, and thus represent indices of urban-associated anthropogenic pollution, and 2) they represent broad classes of chemicals with documented toxic or carcinogenic potential in vertebrates.
The most important and meaningful analyses done in this study were those that directly related hepatic and renal lesion prevalences at the sampling sites to levels of potential or actual exposure to multiple classes of chemical contaminants or their metabolites measured in sediments (potential exposure), stomach contents (dietary uptake), liver (bioaccumulation) and bile (bioindicator of exposure to aromatic compounds). These analyses were also done by logistic regression methods that simultaneously accounted for mean age and gender ratio at the sampling sites. The relationships between concentrations of all contaminant classes measured in each compartment and disease risk were statistically evaluated by this method to examine the consistency of any observed relationships.
Risk factors related to contaminant exposure were less commonly associated with increased risk of nonspecific degenerative/necrotic lesions in any species, and this lesion category is considered a less consistently reliable biomarker of contaminant-induced effects than those discussed above.
The relationships between contaminant exposure and lesion risk were often species-specific, so that the same lesion type affecting several species was not always associated with the same contaminant class(es). These data suggest that not all teleosts respond similarly to toxicant exposure, and that not all of the hepatic lesions identified in this and other field studies can be universally applied as biomarkers of contaminant exposure effects in all feral fish species.
Hepatic lesion categories statistically linked in these species to exposure to specific classes of contaminants were as follows:
Although biomonitoring studies such as that reported here could benefit from the inclusion of a broader suite of measures of chemical contaminants and additional bioindicators, the existence of statistically significant, consistent and toxicologically relevant relationships between risk factors indicating exposure to, dietary uptake of, and hepatic bioaccumulation of chemical contaminants and the prevalence of toxicopathic hepatic and renal lesions provides strong epizootiological evidence supporting a role for certain anthropogenic contaminants in the etiology of these lesions. Moreover, these findings clearly indicate the utility of certain hepatic and renal lesions in some species as biomarkers of contaminant-induced effects in large-scale biomonitoring studies. This evidence provides a strong basis for direction in formulating a strategy for laboratory "cause-and-effect" studies that are essential to more conclusively establish the roles of specific toxicants or contaminant classes in the development of toxicopathic hepatic and renal lesions in marine bottomfish species residing in near-coastal regions.
The National Benthic Surveillance Project (NBSP) was initiated in 1984 by NOAA as a component of the National Status and Trends Program and was designed to assess and document the status of and long-term changes in the environmental quality of the Nation's coastal and estuarine waters. The NBSP is a cooperative effort between the National Marine Fisheries Service (NMFS) and the Coastal Monitoring and Bioeffects Assessment Division (CMBAD) of the Office of Ocean Resources Conservation and Assessment of NOAA's National Ocean Services. The specific objectives of the NBSP are to measure concentrations of chemical contaminants in sediment and in species of bottom-dwelling fish at selected sites in urban and nonurban embayments, to determine the prevalences of certain diseases in these same fish species, to relate chemical contaminant exposure to these diseases, and to evaluate temporal trends in the above-mentioned parameters. The NBSP on the East and Gulf Coasts currently involves a cooperative effort between two NMFS Centers (the Northwest Fisheries Science Center in Seattle, WA, and the Southeast Fisheries Science Center in Beaufort, NC) using similar protocols and analytical instrumentation, while the Northwest Fisheries Science Center is solely responsible for the NBSP on the Pacific Coast. Nationally, approximately 100 sites in embayments along the Atlantic, Gulf and Pacific Coasts, including Alaska, have been sampled from 1984 through 1988 (Cycles I-V, the period covered in this report), with many sites sampled on an annual basis and each annual sampling referred to as a "cycle" (e.g., 1984 = Cycle I).
Pathology data for the Pacific Coast portion of the NBSP have been previously summarized for Cycles I-III, Cycle IV, and Cycles I-IV for sites in San Diego Bay. The present report presents and summarizes all of the significant pathological findings in multiple bottomfish species (target species) captured and examined histopathologically in the NBSP for Cycles I through V (1984-88) at the 45 sites sampled over these years on the Pacific Coast (including Alaska). The prevalences of selected toxicopathic lesions in liver and kidney are summarized over this period, and intersite statistical comparisons among these lesion prevalences in these organs for six primary target species are performed. In these primary target species, prevalences of toxicopathic lesions are also statistically correlated with data reflecting potential and actual chemical exposure (based on levels of contaminants in sediments as well as stomach contents, liver tissue and bile), and disease risk at sites of capture is assessed using logistic regression models that examine the significance of biological and contaminant exposure-related risk factors. In addition, hepatic lesion prevalence data in five target species are analyzed for temporal trends at selected sites over Cycles I-V.
Only histopathologic data are reported here; actual data on sediment chemistry, liver tissue chemistry, stomach contents chemistry, and fluorescent aromatic compounds (FACs) in bile and trace metals in sediments, stomach contents and liver tissue for the Pacific Coast NBSP, Cycles I-V, are presented in separate reports and are reported here only as they relate to the statistical analyses and for determining the toxicologic relevance of contaminant data in comparisons with the pathology data. The present report is not meant to review the marine pollution literature comprehensively; however, pertinent references are included in discussions of the most significant findings.
The National Benthic Surveillance Project (NBSP) is a component of NOAA's National Status and Trends Program (NS&TP) which is designed to assess the status of and document any long-term changes in the environmental quality of the Nation's coastal and estuarine waters. Within this overall plan, more specific goals of the NBSP are to measure concentrations of selected chemical contaminants in sediments and in species of bottom-dwelling fish from urban and nonurban embayments and to determine the prevalences of contaminant-associated diseases in these same fish species. Temporal trends documenting changes in levels of chemical contamination and prevalences of fish disease are also evaluated in order to better manage our Nation's living marine resources.
Fish are known to bioaccumulate xenobiotic chemical contaminants from a number of sources, including bottom sediments (Eadie et al. 1982; Gossett et al. 1983; Varanasi and Stein 1991; Varanasi et al. 1985, 1987b, 1989c, 1992). Moreover, contaminant levels in bottomfish are more reflective of general exposure to chemical contaminants over a broader geographic region than are levels in sediments or sedentary biota, due to the mobility of fish and their consequent ability to act as integrators of chemicals from a variety of sources within a geographic region.
However, of greatest significance to the present subject is the existence of certain pathological conditions in wild fish that morphologically resemble lesions induced by experimental exposure of mammals and fish to a variety of toxicants (Frith and Ward 1980; Stewart et al.1980; Maronpot et al. 1986; Meyers and Hendricks 1982, 1985; Hinton and Lauren 1990a, 1990b; Hinton et al. 1992), and have been shown to be positively associated with exposure to xenobiotic chemical contaminants. For example, bottomfish species from a number of chemically contaminated marine coastal areas are affected with pathological conditions associated with chemical contaminant exposure, especially liver lesions such as neoplasms and other lesions involved in the histogenesis of hepatic neoplasia (McCain et al. 1977, 1983, 1988; Couch and Harshbarger 1985; Harshbarger and Clark 1990; Murchelano and Wolke 1985, 1991; Malins et al. 1987a, 1987b, 1988; Myers et al. 1987, 1990, 1991; Vogelbein et al. 1990), and kidney lesions (Rhodes et al. 1987). Cause-and-effect relationships between chemical contaminants and toxicopathic liver lesions have been inferred from field surveys (Malins et al. 1984, 1987b) and further substantiated by various statistical approaches (Landahl et al. 1990; Myers et al. 1990, 1991) and long-term laboratory exposure studies (Metcalfe et al. 1988, 1990; Schiewe et al. 1991; Varanasi et al. 1987a). Moreover, certain idiopathic hepatic lesions in English sole (Pleuronectes vetulus) have been associated with changes in serum chemistry parameters indicative of liver dysfunction (Casillas et al. 1983). Therefore, certain pathological conditions are currently regarded as having utility as biomarkers of contaminant exposure effects (Myers et al. 1987, 1990, 1991, 1992; Varanasi et al.1992a; Hinton and Lauren 1990a, 1990b; Hinton et al. 1992) and have become useful as indicators of environmental degradation.
In the present study, detected liver lesions of toxicologic significance were placed into six separate diagnostic categories, including 1) neoplasms, 2) putatively preneoplastic foci of hepatocellular alteration, 3) nonneoplastic proliferative lesions, 4) unique or specific degenerative/necrotic lesions, 5) nonspecific degenerative/necrotic lesions, and 6) hydropic vacuolation of biliary or hepatocellular epithelial cells, a unique lesion characterized by light microscopy as a severe hydropic ballooning of the cytoplasmic space, and characterized by transmission electron microscopy as massive distension of the perinuclear cisternae of the rough endoplasmic reticulum (Moore et al. 1989, Moore 1991, Stehr 1990, Stehr et al. 1991, Bodammer and Murchelano 1990). The microscopic anatomy of liver lesions in categories 1-5 in English sole as well as the geographic distribution (within Puget Sound) of sole affected with these lesions has been well documented (Myers et al. 1987, 1990, 1991; Malins et al. 1984, 1985a, 1985b).
As a result of several investigations on the co-occurrence of hepatic lesions in English sole from contaminated environments of Puget Sound, a useful model has been proposed for the histogenesis of hepatic neoplasia in this species, involving a probable progressive sequence of contaminant-related lesions leading towards the development of neoplasms (Myers et al. 1987, 1990, 1991). This proposed histogenetic sequence closely parallels that shown for models of experimentally induced chemical hepatocarcinogenesis in the rat (Farber and Cameron 1980; Farber and Sarma 1987), medaka (Oryzias latipes) (Hinton et al. 1988; Bunton 1990; Hawkins et al. 1990), guppy (Poecilia reticulata ) (Hawkins et al. 1990) and to a certain degree, in rainbow trout (Oncorhynchus mykiss) (Hendricks et al. 1984; Nunez et al. 1990). In addition, English sole exposed to an organic-solvent extract of sediment from a creosote-contaminated site in Puget Sound developed a spectrum of liver lesions (excluding neoplasms) involved in the proposed histogenesis of hepatic neoplasia in this species that were identical to lesions observed in wild English sole from the same and other contaminated sites (Schiewe et al. 1991). Similar hepatic lesions have also been reported for starry flounder (Platichthys stellatus) and rock sole (Pleuronectes bilineatus) from Puget Sound (Myers and Rhodes 1988; Myers et al. 1992), white croaker (Genyonemus lineatus) from the southern California coast (Malins et al. 1987a; Myers et al. 1991), and winter flounder (Pleuronectes americanus) from the Boston Harbor area (Murchelano and Wolke 1985, 1991; Moore 1991). Hepatic lesions in other species of wild fish and their general utility as biomarkers of contaminant exposure have been documented in several recent publications (Hinton and Lauren 1990b; Hinton et al. 1992; Varanasi et al. 1992a). The microscopic anatomy of the unique hydropically vacuolated lesion representing the sixth hepatic lesion category has also been well documented in several teleost species (Murchelano and Wolke 1985, 1991; Bodammer and Murchelano 1990; Moore et al. 1989, 1991; Stehr 1990; Camus and Wolke 1991), and the geographic distribution of this condition is described for subadult starry flounder and rock sole from Puget Sound (Myers et al. 1992) and winter flounder from Boston Harbor area and other sites on the eastern seaboard (Bodammer and Murchelano 1990; Moore et al. 1989, Moore 1991; Johnson et al. 1992). In summary, based on the above information, certain types of hepatic lesions are clearly useful as indicators of contaminant exposure effects in wild fish and were therefore used as the primary histopathologic biomarkers of contaminant exposure effects in the present study.
Kidney lesions of potential toxicologic significance were also detected and placed within the diagnostic categories of proliferative, necrotic or degenerative, and sclerotic or depositional disorders. The microscopic anatomy of some of the above lesions (mesangiolysis, mesangiosclerosis, hypermembranous glomeruli) has been previously described in English sole (Malins et al. 1980, 1982; McCain et al. 1982), and descriptions/discussions of most other lesions detected in this study can be found in basic pathology texts and reviews for both fish (Ferguson 1989; Roberts 1989; Meyers and Hendricks 1985 ) and mammals (Robbins et al. 1984), as well as in reviews of teleost (Pritchard and Renfro 1984; Hinton and Lauren 1990a) and mammalian (Hook and Hewitt 1986) nephrotoxicity, or specific articles on kidney lesions in teleosts (Hinton et al. 1976; Wester and Canton 1986; Lauren et al. 1989; Forlin et al. 1986). In general, the toxicologic bases for teleostean renal lesions other than degenerative/necrotic lesions of the various segments of renal tubules due to heavy metals toxicity (Pritchard and Renfro 1984) are poorly understood and there is a paucity of field studies on the utility of kidney lesions as biomarkers of contaminant exposure (Hinton and Lauren 1990a; Hinton et al. 1992). Therefore, the proliferative and sclerotic lesions are included as suspected toxicopathic lesions primarily on the basis of their heterogeneous distribution among the sampling sites in the present study and previously demonstrated associations between these lesions in English sole and chemically contaminated sites of capture within Puget Sound (Rhodes et al. 1987).
The prevalences of these contaminant-associated fish diseases are presented here for all target species captured and examined in Cycles I-V (1984-88) of the Pacific Coast portion of the NBSP. In six primary target species exhibiting significant prevalences of toxicopathic lesions (flathead sole, Hippoglossoides elassodon; English sole; starry flounder; white croaker; hornyhead turbot, Pleuronichthys verticalis; and black croaker, Cheilotrema saturnum) appropriate statistical comparisons of overall lesion prevalences among the sampling sites are performed. In these primary species, the concentrations of chemical contaminants (Landahl et al. in prep. Clark et al. in prep.) in sediments, liver tissue, and stomach contents, and also fluorescent aromatic compounds (FACs) in bile, are statistically related to the prevalences of certain toxicopathic lesions. In these analyses, the pathology data are subjected to several types of logistic regression analyses, with assessment of relative disease risk in individual fish from contaminated sites compared to disease risk in fish from relatively uncontaminated comparison sites (Fleiss 1981). The data is also used to construct models of disease risk which examine the significance of various risk factors related to contaminant exposure that were determined at the sites of capture each time a site was sampled. The primary advantage of utilizing a multivariate statistical method such as logistic regression is the capacity to assess the effect of, and control for, biological variables such as gender and age while simultaneously assessing the effects of contaminant exposure on disease risk (Breslow and Day 1980).
Also, for the first time in the NBSP, the results of temporal trend analyses on the hepatic histopathology data are presented for certain sites and species where sufficient data are available.
Sites were selected for inclusion in this study based on the following criteria:
A total of 45 Pacific Coast sites that met the above criteria were sampled during Cycles I-V of the NBSP. The locations of these sites, the types of samples collected, the frequency of sampling, and the site abbreviations and numbers are presented in Table 1. Figure 1 (Alaska sites) and Figure 2 (remaining Pacific Coast sites) illustrate the locations of the sampling sites. Twenty- three of these sites were located in or near urban embayments, and the 22 remaining sites were in nonurban embayments, five of which served as comparison sites on the basis of minimal levels of sediment contaminants detected (Kamishak Bay, AK; Nisqually Reach, WA; and Bodega Bay, Dana Point, and outer Mission Bay, CA). These comparison sites were selected so that the same fish species as those obtained at urban or more contaminated sites could be collected from minimally contaminated environments .
Pathologic observations in these "reference" fish from the comparison sites were used to aid interpretation as to the significance of pathological conditions observed in the respective species from contaminated urban sites. Lesion prevalences at sites other than the comparison sites were statistically related to those for the same lesion in the same species from the nearest comparison site.
Sites along the coasts of Washington, Oregon, and California were sampled using the NOAA ship McArthur and its ship-based launches, and the research vessels Harold W. Streeter or Sea Otter. The NOAA ship Miller Freeman and the U.S. Fish and Wildlife Service vessel Curlew were used in Alaska. To maximize comparability of data over space and time, standardized collection gear and sampling methods were used. All vessels were equipped with navigational equipment to determine the precise location (latitude/longitude) of all the sampling stations to ensure that the same stations at each site were sampled from year to year (Table 1).
Fish were collected at stations within a site with an otter trawl (7.5-m opening,
10.8-m total length, 3.8-cm mesh size) equipped with netting that was not chemically treated. Fish of a minimum size (generally greater than 15 cm, total length) were randomly selected from each haul and kept alive in flowing fresh seawater until necropsies were performed, generally within two hours of capture. A target number of 30 fish for full necropsy per target species captured among the three stations at each site (Table 2) was established for Cycles I-IV; up to three target species were sampled at a particular site and year, depending on fish availability. In Cycle V, this target number was increased to 60 fish/species/site. If the target number of fish per species could not be captured among the stations at a site, as many animals as possible were sampled. Only live fish were necropsied.
The necropsy procedure on individual fish involved a) assigning a unique field identification number, b) weighing (total weight in grams) and measuring (total length in millimeters), c) killing fish by cervical section, d) removing otoliths for age determination,
e) describing and documenting grossly visible anomalies, and f) excising/extracting the gall bladder, liver, kidney, gonad, and stomach contents. Tissue sections 3-4 mm thick from 1) the central portion of the liver along the longitudinal axis, 2) the posterior portion of the kidney, 3) a sagittal section of the gonad, and 4) other grossly visible lesions in internal organs, especially in the liver, were taken and preserved in Dietrich's fixative (Gray 1954) for routine histological processing and histopathological examination. Bile from the gall bladder (10-12 fish/species/site) and approximately one-half of the remaining liver tissue (30 fish/species/site) were placed in separate, solvent-rinsed glass vials and frozen at -20°C for later determination of levels of biliary FACs and organic chemicals, respectively. The remaining liver tissue was placed in a plastic vial and stored frozen until analyzed for selected metals (30 fish/species/site). The total stomach contents from at least 10 fish/species/site were removed and composited in a solvent-rinsed glass jar (up to three composites/species/site) and frozen for separate organic chemical and metals analyses. Surface sediments were collected as per protocols described in Varanasi et al. (1988, 1989a).
Encoding of Field Data
Each individual fish was assigned a unique specimen number in the field as the primary identifier, and all relevant biological data for each fish, including species, length, weight, gender, age or estimated age, year and site of capture, and information on grossly visible lesions were recorded and entered into a computer database.
Histologic Technique/Diagnostic Histopathology
Fish tissues were preserved in the field in Dietrich's fixative for a minimum of 48 hours, transferred and stored in 70% ethanol, routinely processed for paraffin embedment, embedded, sectioned at a 4-5- µm thickness, and stained with Mayer's or Gill's hematoxylin and alcoholic eosin-Y (Armed Forces Institute of Pathology 1968). For further characterization of specific lesions, additional sections were stained using various special staining methods such as Gomori's iron hematoxylin for hemosiderin, periodic acid-Schiff for glycoproteins and mucopolysaccharides, and Masson's trichrome for collagen (Thompson 1966, Armed Forces Institute of Pathology 1968, Preece 1972).
Prior to histopathologic examination, each fish with a unique field identification number was assigned a unique, computer-generated random number, or "pathology" number. Slides corresponding to each fish were labelled with this pathology number, ensuring that histopathologists were unaware of the site of origin for each specimen, thus eliminating potential diagnostic bias. These blind numbers were maintained until all specimens within a particular cycle had been examined, whereupon the matched field identification number was revealed for subsequent encoding of histopathologic diagnoses.
Lesion classification followed previously described, standardized diagnostic criteria in rodents and other mammals (Frith and Ward 1980, Jones and Butler 1975, Ward and Vlahakis 1978, Stewart et al. 1980, Maronpot et al. 1986, Robbins et al. 1984) and fish (Hendricks et al. 1984; Murchelano and Wolke 1985, 1991; Myers et al. 1987; Hinton and Lauren 1990a, 1990b). Consistency of diagnostic criteria among the several examining histopathologists was ensured by periodic meetings with the chief histopathologist (Mark S. Myers) to review and discuss unusual or problematic cases.
Histopathologic diagnoses were then encoded utilizing a lesion classification code modeled after the Systematized Nomenclature of Pathology (SNOP) code, and identified as File Type 13 by the National Oceanographic Data Center (NODC). These data were linked in the computer database to the unique field specimen number and other relevant biological, chemical, and geographical data. Site, year, and species-specific lesion prevalences were subsequently generated from this database and then linked in parallel to the chemical contaminant data determined for each year a particular site was sampled.
Determination and Estimation of Fish Age
The probability of occurrence of certain diseases in fish, such as hepatic neoplasms, has been shown to increase with age in a number of field studies (Rhodes et al. 1987; Myers et al. 1990, 1991; Baumann et al. 1991; Moore 1991; Vethaak and Rheinallt 1992). Because it was not always possible to capture specimens of the same length/age composition within a target species among the sampling sites, fish age was determined for a subset of the total number of specimens in each primary species that displayed significant toxicopathic lesions (i.e., white croaker, starry flounder, English sole, flathead sole, hornyhead turbot) collected over Cycles I-V. This subset consisted, at a minimum, of the entire sample for a particular species collected in at least one cycle within a particular geographic area (e.g., San Francisco Bay, Puget Sound, Southern California). Fish age was determined by counting the number of clearly defined opaque zones of whole otoliths under a binocular dissecting microscope (Chilton and Beamish 1982, Kimura et al. 1979) or in the case of white croaker, by more specialized techniques involving embedding and sectioning of otoliths (Love et al. 1984). Ages of black croaker have not yet been determined from collected otoliths, so length alone was used as an estimate of fish age for all collected specimens of this primary target species.
For all specimens in which actual age was not determined, individual ages were estimated as follows. For each species, both gender-specific and gender-combined age-length curves were established within a particular geographic area (e.g., starry flounder males, San Francisco Bay) by simple or polynomial (up to third-order) regression techniques. Resultant age-length curves and equations (Appendix Table A-1) were then chosen on the basis of best fit to the data (i.e., highest adjusted R-squared value). Fish age was then estimated, according to gender whenever possible, to the nearest year from these gender-specific age-length equations that were established for a particular species from a defined geographic area. When no gender data were available, or in cases where insufficient numbers of a particular gender existed to perform a meaningful age-length regression, estimated age was established by applying age-length equations for both genders combined. Ages for starry flounder from the Chukchi Sea site were estimated using the gender-specific age-length equations shown in Wolotira et al. (1977).
Chemical Analyses of Sediments, Stomach Contents,
Liver Tissue and Bile
Sediments and stomach contents were analyzed for a broad spectrum of aromatic hydrocarbons (AHs), chlorinated hydrocarbons (CHs), and selected trace elements (MacLeod et al. 1985; Varanasi et al. 1988, 1989a; Landahl et al. in prep.; Clark et al. in prep.). Liver tissues were analyzed for a broad spectrum of CHs and selected trace elements, but not for AHs due to the extensive metabolism of these labile compounds to more polar metabolic products (Varanasi and Gmur 1981; Varanasi et al. 1982, 1986, 1987b). Levels of FACs in bile were measured using the procedures of Krahn et al. (1984, 1986a, 1986b, 1987).
Although levels of many individual AHs (23), CHs (21) and trace elements (15) were routinely determined in sediments, stomach contents, and liver tissue (CHs and trace elements only), levels of broader groups of these compounds were also computed (Table 3). For biliary FACs, the data are reported as total levels of compounds fluorescing at two wavelength pairs. One wavelength pair (380/430 nm), also referred to as the benzo(a)pyrene (BaP) wavelength pair, was used to estimate the higher molecular weight range of FACs
(4 to 6 aromatic rings), which are referred to as FACs-H. Most FACs-H are derived from high molecular weight AHs, such as pyrene and BaP, that are combustion products (Krahn et al. 1987). The second wavelength pair (290/335 nm), also referred to as the naphthalene (NPH) wavelength pair, was used to estimate the lower molecular weight range of FACs
(2 to 3 aromatic rings), which are referred to as FACs-L. Most FACs-L are derived from fossil fuel components such as the naphthalenes (Krahn et al. 1987). For the present report, chemical compounds were placed into broader groups (Table 3) in order to provide a more reasonable number of variables or risk factors related to chemical contaminant exposure, against which logistic regression analyses with respect to lesion prevalences or lesion presence/absence could be performed. For the trace elements or metals, two groups were formed on the basis of their pattern of covariance in sediments for Cycles I-III (Varanasi et al. 1989a) as shown by principal components analysis (Wishart 1975): Metals 1 consisted of the summed individual levels of Cu, Zn, Pb, and Sn; and Metals 2 was composed of summed individual levels of Ni, Cr, and Se.
A variety of statistical methods was employed to assess the uniformity of lesion distribution in each primary target species among the sites of capture and to evaluate possible relationships between site- and year- specific lesion prevalences, levels of chemical contaminants in sediments at the capture sites, and actual exposure to chemical contaminants at these sites in species exhibiting significant prevalences of toxicopathic lesions. Potential relationships between the probability of lesion occurrence and degree of actual exposure to and metabolism of AHs in individual fish, as estimated by measurement of levels of FACs in bile, were also evaluated statistically.
Intersite Differences in Lesion Prevalence
To test the a priori null hypothesis of no statistical difference in lesion prevalences in the primary target species between the comparison or reference site and the individual test sites, and the alternate hypothesis of a significantly higher lesion prevalence at the test site, the G statistic (Sokal and Rohlf 1981) was computed. For this test, the lesion prevalence at the reference or comparison site of closest geographic proximity was designated as the expected value. Analyses were performed on site-specific lesion prevalence data combined for the 5 years of study. For each primary target species, the prevalence of a particular lesion category was considered to be higher than that at the reference site if the difference was significant at p < 0.05. Because fourhorn sculpin (Myoxocephalus quadricornis), Arctic flounder (Pleuronectes glacialis), spotted sand bass (Paralabrax maculatofasciatus), and diamond turbot (Hysopsetta guttulata ) examined for histopathological conditions were sampled at only one (Arctic flounder) or two sites each, intersite differences in lesion prevalences in these species could not be tested. Species not among the primary target species, and sites at which less than 15 fish of a given primary target species were collected over the 5-year period were excluded from this and other statistical analyses.
Logistic Regression and Multivariate Risk Factor Analyses
Stepwise logistic regression, also known as logit analysis (Breslow and Day 1980), was used to identify risk factors (e.g., age, gender, site of capture, PCBs in liver tissue) that might be associated with increased probability of lesion occurrence in the primary target fish species with significant prevalences of toxicopathic lesions. This method is especially suited to the analysis of data that is binomially distributed or expressed in terms of proportions or percentages (e.g., lesion prevalences) and allows for simultaneous adjustment for all risk factors included in the regression by iterative model fitting. Results can be expressed in terms of relative risk of disease associated with a particular risk factor and also as the proportion of variation in lesion prevalence attributable to discrete risk factors. This type of analysis of lesion occurrence has been used in retrospective epidemiological studies (Breslow and Day 1980) and more recently as specifically applied to the epizootiology of hepatic and renal lesions in English sole (Rhodes et al. 1987; Malins et al. 1988; Myers et al. 1990, 1991), rock sole and starry flounder (Myers et al. 1992), and winter flounder (Johnson et al.1992). The method represents a statistical, epidemiological approach to the examination of the influence of multiple risk factors on the probability of disease occurrence, as well as the investigation of dose-response relationships. In the logistic model, the probability of disease (p) is given by:
where x1, x2, ..., xp represent potential risk factors, and the ß's are coefficients that represent the effects of the x's on the risk (probability) of disease.
Three types of logistic regression analyses were conducted. First, an analysis was performed to examine the relative risk of disease, as estimated by the calculated odds ratio, in individual fish in relation to the variables of site of capture, gender, and age or estimated age. Second, an analysis was performed to examine the relative risk of disease, as estimated by the calculated odds ratio, in individual fish in relation to levels of FACs in bile, while adjusting for gender, and age or estimated age; this analysis was done because biliary FAC level was the only measure of contaminant exposure routinely determined in individual fish in this study. Third, an analysis was performed to assess the significance of the relationships between prevalences of lesions at the sampling sites and discrete risk factors, such as levels of contaminants in sediments and fish tissues, while simultaneously controlling for mean fish age and gender ratio (females:males); in this analysis each year's data for a species at a site was treated as an independent occurrence, so that if a species was examined annually at six sites over Cycles I-V, there would be 30 data points included within the logistic regression. These multivariate analyses allowed the evaluation of contaminant-lesion relationships while simultaneously adjusting for potentially confounding factors such as gender, gender ratio, age or mean age of fish sampled. For these analyses, models were initially evaluated for goodness of fit by standard stepwise multiple regression. Only those risk factors which significantly (p < 0.05) improved the fit of the models to the actual data were included in the logistic regression equations. The PECAN analysis module of the EGRET statistical package, version 111 (Statistics and Epidemiology Research Corporation, Seattle, WA) was employed to fit the logistic regression models, using a procedure similar to stepwise multiple regression. Independent analyses were performed for all major categories of liver and kidney lesions.
For the first type of analysis, odds ratios were determined from variable coefficients of the logistic regression equations, with the estimated odds ratio or relative risk for the occurrence of a lesion type representing a measure of the degree of association between a risk factor and lesion occurrence (Fleiss 1981). Odds ratios for particular lesion categories at a site were calculated and interpreted relative to the lesion occurrence at designated comparison or reference sites. Reference sites were designated as follows: English sole--Nisqually Reach and Bodega Bay; flathead sole--Kamishak Bay; starry flounder--Bodega Bay; white croaker--Bodega Bay and Dana Point; hornyhead turbot--Dana Point; and black croaker--outside Mission Bay. Increased probability of lesion occurrence was indicated by odds ratios greater than 1.000, and decreased probability of lesion occurrence was indicated by odds ratios less than 1.000.
For the second analysis relating lesion occurrence to levels of FACs in bile, the procedure was essentially the same as that described above, except that this analysis was performed independent of site of capture. Site was not included as a variable in this analysis because our intent was to directly relate lesion occurrence to actual AH exposure in individual fish, as estimated by levels of biliary FACs fluorescing at either B(a)P (FACs-H) or NPH (FACs-L) wavelengths, in all specimens of a species where biliary FACs were measured, regardless of where they were captured. These analyses were done separately for biliary FACs-H and FACs-L levels, and they simulaneously accounted for fish gender and age or estimated age in individual fish.
The third logistic regression analysis examined the significance of discrete risk factors, as determined at the sampling sites, to the risk of disease (i.e., lesion prevalence) while controlling for two biological risk factors: mean age and gender ratio. Data for each year's sampling of a species and sediments at a site were treated as independent occurrences. These analyses were performed for each separate contaminant exposure-related risk factor shown in Table 3, and the results were expressed as the proportion of variation in lesion prevalence (also referred to as the reduction in scaled deviance) attributable to these specific, significant (p < 0.05) risk factors. Because many of the contaminant exposure-related risk factors in this analysis were highly intercorrelated (Landahl et al. in prep.) and in many cases far exceeded the number of available data points, it was not mathematically possible in this method to include all risk factors simultaneously in a single multivariate analysis. Inclusion of intercorrelated and colinear risk factors in a single model invariably resulted in a singular data matrix and eventual elimination of less significant risk factors, which, when tested individually were identified as risk factors accounting for a significant proportion of variation in site- and year-specific lesion prevalences. Consequently, multiple models were necessary to fully describe all of the relationships between the individual risk factors and lesion prevalences.
Analysis of Temporal Trends in Hepatic Lesion Prevalences
Temporal (i.e., year-to-year) trends in prevalences of certain hepatic lesion categories with respect to year of capture were assessed in species at sites containing at least four years of data, which confined the analysis to the following major target species at the indicated sites: English sole (Bodega Bay), starry flounder (San Pablo Bay, Southampton Shoal , Hunters Point, Bodega Bay, Coos Bay), hornyhead turbot (Dana Point), barred sandbass (south San Diego Bay) and white croaker (Bodega Bay, Hunters Point, Southampton Shoal , Dana Point, San Pedro Outer Harbor). The significance of any differences in lesion category prevalences among years of capture at the individual sampling sites was determined by logistic regression, while accounting for mean age and gender ratio. In addition, hepatic lesion prevalence data were also related to sampling year by the Spearman rank correlation method (Zar 1984). This method is capable of detecting temporal trends, as defined by monotonic increases or decreases in lesion prevalences.
Table of Contents