Great Lakes: An Ecoepidemiological Study

The objectives of this study were to determine whether contaminant-associated immunosuppression occurs in prefledgling herring gulls and Caspian terns from the Great Lakes and to evaluate immunological biomarkers for monitoring health effects in wild birds. During 1992 to 1994, immunological responses and related variables were measured in prefledgling chicks at colonies distributed across a broad gradient of organochlorine contamination (primarily polychlorinated biphenyls), which was measured in eggs. The phytohemagglutinin skin test was used to assess T-lymphocyte function. In both species, there was a strong exposure-response


Introduction
Laboratory studies have shown that environmental contaminants can suppress immunological function and increase susceptibility to infectious diseases (1)(2)(3)(4)(5). Often these chemicals act at low doses and cause persistent effects, especially with perinatal exposure. These laboratory studies raise concerns about potential immunotoxic impacts on wildlife and humans. There have been few immunotoxicological investigations of free-living wildlife, especially birds. We investigated associations between contaminants and immune function in prefledgling fish-eating birds from the Great Lakes.
For ethical and financial reasons, manipulative field experiments often are not possible in ecotoxicology. Pollution patterns are determined by the locations of pollution sources and the movement of air, water, and sediments. In the absence of manipulative experiments, epidemiological criteria aid the elucidation of associations between contaminants and biological effects (72). For epidemiologists, establishing causation does not require that a factor be a necessary and sufficient condition to produce an effect. Rather, causal associations imply that a factor is part of a complex that increases the probability of an effect, and that reducing the factor reduces the probability of the effect. Epidemiological criteria for assessing causal associations include time order, strength of association, specificity, consistency upon replication, coherence, predictive performance, and probability (72). The criteria that support an association are weighed against those that detract from it.
The objectives of this ecoepidemiological study were to determine whether contaminant-associated immunosuppression occurs in prefledgling herring gulls and Caspian terns (Sterna caspia) from the Great Lakes and to evaluate immunological biomarkers for monitoring health effects in wild birds. We employed two in vivo tests of immune function: the phytohemagglutinin (PHA) skin test for T-cell-mediated immunity and the sheep red blood cell (SRBC) hemagglutination test for antibody-mediated immunity. We also measured white blood cell (WBC) numbers, plasma retinol, and thyroxine as general biomarkers relevant to immune function. Prefledgling birds were studied because the developing immune system is particularly sensitive to contaminants. The herring gull was chosen because it is the most frequently used and best understood avian bioindicator species in the Great Lakes. The Caspian tern was chosen because its elevated exposure and sensitivity to organochlorines have been associated with reproductive and population-level effects (12,19,73,74). Contaminant-associated immunosuppression in young Caspian terns is a possible mechanism for these population-level effects.

Sampling Desig
In 1992, herring gull chicks were sampled across a gradient of organochlorine contamination at four sites within the Great Lakes and one site outside the Great Lakes (Tables 1,2; Figure 1): a) Little Charity Island in Saginaw Bay, Lake Huron; b) Bird and Anchor Islands in the North Channel, Lake Huron; c) Monroe, on the western shore of Lake Erie; d) Hamilton Harbor on the western shore of Lake Ontario; and e) Pony Island in northern Monroe Figure 1. Sampling sites for immune function tests in herring gull and Caspian tern chicks. Bars represent total PCB concentrations found in pooled egg homogenates from each site. Numbers above bars indicate total PCB concentrations. PCB concentrations for the two species are graphed on different scales.  Figure 1): a) Gravelly Island in upper Green Bay, Lake Michigan; b) High Island in northern Lake Michigan; c) Elm Island in the North Channel, Lake Huron; d) the Confined Disposal Facility in southern Saginaw Bay, Lake Huron; and e) Pigeon Island in eastern Lake Ontario. The North Channel was chosen as a reference colony. In 1993 and 1994, Caspian terns were resampled at the reference site, the North Channel, and at one highly contaminated site, Saginaw Bay. Logistical difficulties prevented some variables from being measured at some sites (Table 1). At 8 of 10 sites, enclosures (1 x 2 cm plastic mesh and approximately 0.8 m high) were placed around groups of 10 to 20 herring gull nests or 30 to 40 Caspian tern nests during midincubation. Usually two or three enclosures were erected at each site. Chicks were confined until fledging or until the enclosures were removed. Chicks were banded with U.S. Fish and Wildlife Service leg bands for individual identification. At two sites, rocky ground prevented the construction of enclosures, so chicks were captured and released into thick vegetation that provided refuge and prevented chicks from fleeing too far from their nests. Immune function tests were initiated on 35 to 50 chicks at each site. WBC numbers, retinol concentrations, and thyroxine concentrations were assessed for 10 to 20 chicks per site.
Immune function tests were initiated on 3-week-old chicks. Age was determined by estimated hatch times and body size measurements. Target body size for herring gull chicks was a body mass of 400 to 700 g and a wing chord of 130 to 200 mm. Criteria for Caspian terns were a Environmental Health Perspectives -Vol 104, Supplement 4 * August 1996 body mass of 450 to 550 g and a wing chord of 130 to 200 mm. Body size measurements were made on the same chicks at the beginning and end of the functional tests.

Functional Tests for Immunocompetence
The PHA skin test for T-cell-mediated immunity was conducted in 3-week-old chicks following the procedures of Grasman and Scanlon (75) using a 0.1-ml dose of 1 mg/ml PHA-P (Sigma, St. Louis, MO) in phosphate-buffered saline (PBS). Feathers were plucked from both wing webs. One wing web was injected with PHA while the other received a placebo injection of PBS alone. The thickness of each wing web was measured to the nearest 0.05 mm immediately before and 24 ± 3 hr after the injections using a pressure-sensitive caliper with a low-tension spring that did not crush the skin (model 304-196, Dyer Co., Lancaster, PA). A stimulation index was calculated as the change in the thickness of the PHA-injected wing web minus the change in thickness of the PBSinjected wing web.
The SRBC hemagglutination test was initiated at the same time as the PHA skin test. Chicks were injected via the wing vein with 0.1 ml of a 1% SRBC suspension in sterile saline. Plasma samples were collected from chicks 5 to 7 days after SRBC injection because antibody titers peak in gulls and terns at approximately 6 days post-immunization (K Grasman, unpublished data). Total (IgM + IgG) and 2-mercaptoethanol-resistant (IgG) antibody activities were measured by the microtiter method of Gross and Siegel (76,77). Fifty microliters of normal saline were added to each well in 96-well microtiter plates with round-bottomed wells. Fifty microliters of plasma were added to the first well of each row, and serial 2-fold dilutions were performed across rows. Fifty microliters of a 0.25% SRBC suspension in normal saline were added to each well. The plates were incubated for 3 hr at 37°C. Titers were determined as the log2 of the reciprocal of the highest dilution showing agglutination. To measure IgG titers, plasma samples were incubated for 60 min with 0.2 M 2-mercaptoethanol before dilution. Red blood cells from one sheep (Colorado Serum Co, Denver, CO) were used for all injections and assays during 1992 and 1993. In 1994, SRBCs were obtained from another sheep of the same age and flock.

General Immunological and Hematological Biomarkers
One day after initiating the functional tests, a 4-ml blood sample was drawn from the wing vein of the same chicks using a 22gauge needle and Vacutainer tubes containing ethylenediaminetetraacetic acid (EDTA) (Beckton Dickinson, Rutherford, NJ). Two blood smears were made within 5 hr after blood collection. Blood was centrifuged at 2575xg for 5 min, and the plasma was stored in liquid nitrogen for retinol and thyroxine determinations. A second 4-ml blood sample was collected 6 days after SRBC immunization, and the plasma was stored in liquid nitrogen for antibody analysis.
Blood smears were fixed with methanol and stained with Wright stain (Accustain, Sigma, St. Louis, MO) using 100% stain for 30 sec followed by a 1:1 dilution with distilled water for 90 sec. Smears were rinsed with distilled water and allowed to air dry. Two hundred WBCs were counted and classified using oil immersion microscopy at 100Ox magnification.
Retinol was extracted from plasma after the addition of retinyl acetate as an internal standard. Retinol-protein complexes were dissociated by vigorous shaking after the addition of acetonitrile. The retinol was extracted with hexane, and the organic and aqueous phases were separated by centrifugation. The organic phase was dried under nitrogen, reconstituted with methanol, and filtered. The retinoids in the extract were separated by reverse-phase high-performance liquid chromatography (HPLC) using a 15-cm long, 5 pm octadecylsilane (ODS) analytical column and 100% methanol as a solvent. Either fluorescence (ex: 336 nm; em: 480 nm) or UV-visible (326 nm) was used to detect the retinoids. The detection limit for retinol in plasma was 5 pg/liter. Total plasma thyroxine was measured using a competitive binding enzyme immunoassay (veterinary modification of the EZ Bead T4 Test, Immunotech Corp, Boston, MA).

Organochlorine Analysis and Chick Hepatocyte Bioassay
The 12 eggs collected from each site were pooled for organochlorine analysis by the analytical services laboratory at the National Wildlife Research Centre of the Canadian Wildlife Service following the methods of Norstrom et al. (78). PCB residues are reported as the sum of the following 42  Non-ortho PCB congeners (IUPAC nos. 37, 77, 126, and 169) and all 2,3,7,8substituted polychlorinated dibenzo-pdioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) also were measured. Samples of the pooled egg homogenates were dried with anhydrous sodium sulfate and ground into a powder. An open chromatographic column wet-packed with multiple absorbents was used for the initial extraction and cleanup. After spiking with a 13C-PCDD mixture (Cambridge Isotope Laboratories) and a 13C-PCB 77, 126, and 169 mixture (Wellington Isotope Laboratories), the column was eluted using dichloromethane/hexane. A carbon column was used for further cleanup and trace enrichment. The concentrated eluent was cleaned up and separated on a deactivated Florisil column by first eluting with dichloromethane/hexane for non-ortho PCB analysis. The column was eluted with dichloromethane to produce a second fraction. This fraction was cleaned up on an activated basic alumina column by eluting with 1:50 dichloromethane:hexane to produce a fraction containing residual PCBs and other organochlorines. The column was eluted with dichloromethane:hexane to produce a fraction for PCDD/PCDF analysis. A Hewlett-Packard 5971A GC/MSD was used to separate and quantify non-ortho PCB, PCDD, and PCDF congeners. Detection limits were 75 pg/g for non-ortho PCB congeners and approximately 0.3 to 2.7 pg/g for PCDD and PCDF congeners.
Because different HAH congeners have different toxicities, the total biological activity of a mixture of congeners cannot be estimated by adding the concentrations of the individual congeners. The chicken embryo hepatocyte bioassay was used to measure the total TCDD-like activity in the pooled egg homogenates (79,80). HAHs were extracted using minor modifications of the procedures used for chemical residue analysis (81,82). Extracts added to chicken hepatocytes contained all PCDDs, PCDFs, PCBs, and structurally related nonpolar HAHs and chlorinated pesticides. Based on the in vitro induction of EROD in chicken hepatocytes, this bioassay compared the potency of a mixture of HAHs to that of a TCDD standard. The resulting measure of TCDD-like toxicity was designated as C-TEQs (chicken bioassay-derived TCDD-equivalents).

Statistica Analyses
The primary goal of this investigation was to determine whether there was an association between organochlorine exposure and immunosuppression based on intercolony differences in immunological variables. The purpose was not to show what percent of variability in immunological responses could be explained statistically by particular chemicals, but rather to determine the probability that the spatial patterns in response variables were associated with contaminants as opposed to other factors or random events. The strategy for statistical analysis was shaped by two factors: a) the importance of testing specific hypotheses regarding associations between immunological variables and organochlorine contaminants, and b) the great expense and tissue volume required for congener-specific PCB analysis. The expense and volume requirement prevented contaminant analyses on tissues from individual birds. Instead, sites were ranked in order of contamination for various chemicals based on analysis of the pooled egg homogenates.
The Jonckheere test for ordered alternatives was used to test specific hypotheses concerning contaminant-associated immunosuppression (84). It fit the purpose and design of the study and the constraints on chemical analysis. The null hypothesis for this nonparametric measure of exposure-response states that there is no difference among the central tendencies from different sites (HO: p1 = P2 = P3= ... =Pn). The alternative hypothesis is that there is a monotonic trend (not necessarily linear) based on a priori information (HA: p1 P2 P3 ... pn, where at least one of the inequalities is a strict inequality). We used data from laboratory and other field studies to predict the effects of various contaminants. Inverse relationships to contamination were predicted for the PHA skin response, antibody responses, plasma retinol, and thyroxine. The direction of the trends for WBC numbers could not be predicted from laboratory data, so a twoway test was performed by running the Jonckheere test in both directions and doubling the p-value for the most significant trend. Concentrations of total PCBs, DDE (1, I-dichloro-2,2-bis(p-chlorophenyl)ethylene), C-TEQs, and HG-TEQs in the egg homogenates were used to determine the relative order of the study sites. The Jonckheere test was performed using a custom-written FORTRAN program (B.Collins, Senior Statistician, Canadian Wildlife Service) using the algorithm provided by Gibbons (85). This program determined probability values based on 5000 random permutations of the data set being analyzed. These probability values were very similar to those found by the large sample normal approximation for the Jonckheere test (84).
Before the Jonckheere test, a preliminary ANOVA was used to determine whether the response variable was influenced by year or a year x site interaction for subsets containing multiple years of data. If there was no statistically significant year or a year x site interaction effect (p < O.O5), then data were pooled across years for the Jonckheere test. If the Jonckheere test did not show any statistically significant trends, then one-way ANOVA within years followed by Duncan's multiple range test was used to elucidate spatial differences in response variables. For ANOVA analyses, heterophil/lymphocyte ratios were transformed (loglo) to satisfy assumptions of homogenous variances and normality (86). This transformation also was used for the Jonckheere and correlation analyses for this variable. Pearson's correlation analysis was used to detect associations between biomarkers, especially to determine whether any biochemical biomarkers could serve as surrogates for measuring immune function. Statistically significant correlations (p< 0.05) were reported only if the absolute value of the correlation coefficient (r) was greater than 0.3.

Growth
In Saginaw Bay during 1992, both species experienced severe loss of body mass between 3 and 4 weeks of age (-11 g/day in herring gull and -5 g/day in Caspian tern chicks), which was much lower than the growth at other sites (14)(15)(16)(17)(18)(19)(20)

Organochlorine Concentrations
The range in organochlorine concentrations among sites was greater for pooled herring gull eggs than for Caspian tern eggs (Table 2; Figure 1). In herring gulls, total PCBs, C-TEQs, HG-TEQs, and DDE gave the same rank order, so the Jonckheere test was identical for these contaminants. In Caspian terns, the rank orders of contamination were different, requiring separate Jonckheere tests.

Functional Tests For lImmunocompetence
In both herring gull and Caspian tern chicks, several measures of organochlorine contamination showed strong inverse exposure-response associations with T-cell function as measured by the PHA skin test (Table 3; Figure 2). For herring gulls, a preliminary ANOVA indicated no evidence for a year or a year x site interaction effect (p> 0.45), allowing data to be pooled across years. There was strong evidence that the PHA response decreased as total PCBs, C-TEQs, HG-TEQs, and DDE increased (p=0.0002 for 1992 to 1994, p= 0.0009 for 1992  Organochlorine contaminants were measured in pooled egg samples. aWhen sites were ranked by different contaminants, their rank orders were identical for several chemicals. Therefore, the Jonckheere test was the same for these chemicals. (p = 0.0002), C-TEQ (p = 0.0002), and DDE (p= 0.0002) exposure increased. In 1992, there was strong evidence for an inverse relationship with total PCBs (p=0.0004) and DDE (p=0.0008) but not with C-TEQs (p = 0. 14). The most contaminated sites (Saginaw Bay and eastern Lake Ontario) were 30% lower than the least contaminated sites. In both species, there was no evidence for contaminant-associated suppression of the total antibody (IgM + IgG) and IgG responses following immunization with SRBC (p>0.50; Table 3; Figure 3). However, these biomarkers were influenced by site and (or) year. In herring gull chicks of Saginaw Bay, the only site with multiple years of data, there was marginal evidence that total antibody titers decreased from 1992 to 1994 (F2,84= 2.50, p=0.088) but stronger evidence for decreasing IgG titers (F2,84=4.13, p=0.020). During 1992 to 1994, there was little evidence for a difference among the four sites in total antibody (F3,135 = 1.74, p = 0. 16)  ing the differences in titers among Caspian tern colonies, a posteriori Jonckheere tests provided evidence for positive associations between the total antibody response and total PCBs, C-TEQs, and DDE in 1992 alone and 1992 to 1994 (p<0.02 for two-way tests).

Biochemical and Hematological Biomarkers
Several organochlorines showed inverse exposure-response relationships with plasma retinol, especially in Caspian tern chicks (Figure 4) There was no evidence that organochlorines influenced plasma thyroxine concentrations (p>0.15; Figure 5), although these biomarkers were affected by site and (or) year. In herring gull chicks from Saginaw Bay and the North Channel, the preliminary ANOVA indicated strong evidence that a year x site interaction influenced plasma thyroxine (F1,36 = 8.05, p= 0.007). In 1992, there was strong evidence that plasma thyroxine differed among the five sites (F4,44 =

Correlations among Functional Tests and More General Biomarkers
There was little evidence for any biologically significant relationships between T-cellmediated and antibody-mediated immunity (Table 4). For Caspian terns in 1992 and 1992 to 1994, there was moderate evidence for a weak negative correlation of the PHA response with the total antibody titer.
The PHA response showed few relationships to other biomarkers. For herring gull chicks in 1992, the PHA response was positively correlated with plasma thyroxine. This relationship was weaker during 1992 to 1993. The antibody responses were significantly correlated with WBC variables, but the nature of this relationship differed between species. During 1992 to 1993, the heterophil/lymphocyte ratio was negatively correlated with total antibody and IgG titers in herring gulls and positively correlated in Caspian terns. Plasma retinol was positively correlated with the total antibody and IgG responses in herring gull chicks in 1992. This relationship was weaker for 1992 to 1993. Several biomarkers were correlated with measures of body size or age. In 1992, body mass and wing chord length were positively correlated with plasma thyroxine and retinol in herring gull chicks but not in Caspian terns. In herring gulls, the heterophil/lymphocyte ratio was negatively correlated with body mass and wing chord length in 1992. Weaker relationships were observed during 1992 to 1993.

Discussion Epidemiological Evaluation
This field study provided strong epidemiological evidence for associations between perinatal exposure to organochlorines and suppression of T-cell-mediated immune function in herring gulls and Caspian terns at highly contaminated sites in the Great Lakes (Table 3; Figure 2). Suppression was most severe (30-45%) at colonies in Lake Ontario (1992) and Saginaw Bay (1992-1994) for both species and in western Lake Erie (1992) for herring gulls. Saginaw Bay, Hamilton Harbor, and the River Raisin, which enters Lake Erie near the Monroe study site, have been designated "areas of concern" by the International Joint Commission because of toxic pollution. In both species, TCDD concentrations in eggs were highest in Saginaw Bay and Lake Ontario. Support for associations between organochlorine exposure and suppression of T-cell-mediated immunity came from  Table 1 for site codes.
the following epidemiological criteria: probability, strength of association, specificity, consistency, coherence, predictive performance, and time order. Using the criterion of probability, the Jonckheere test provided strong evidence that the PHA skin response decreased as several measures of organochlorine contamination increased (p<0.001; Table 3). Beyond statistical significance, the strength of association criterion suggested that the magnitude of the suppression at the most contaminated sites was biol significant. Laboratory studies w have shown that elimination of T cyte function by irradiation or i suppressive drugs reduces ti response by 50 to 60% (75,87,88 the 30 to 45% suppression in gulls and Caspian terns at highly nated sites in the Great Lakes rej biologically significant impact. approach the maximal suppressioi in these species because the weig] Saginaw Bay in 1992 was not associated with further suppression. The criterion of replication strongly supported these relationships that were demonstrated a) in two species sharing a similar contaminated food supply (fish), b) at sites with similar magnitudes of contamination, and c) in multi-|45) ple years (2-3 years) of study at some sites.

35)
The criterion of coherence also supported an association between contaminants and suppression of T-cell-mediated immunity. Such associations are consistent with many laboratory experiments that have found severe impacts of HAHs on Tcell-mediated immunity in birds and mammals (22,26,32,33,(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47). In a parallel investigation that studied herring gull chicks at nine colonies in the Great Lakes and one colony on the Atlantic coast, thymic atrophy was associated with increasing liver EROD activity (89). Thymic mass was reduced 20 to 45% at S the sites with highest EROD activity. SagB Although thymic mass was not associated with any single organochlorine, the thymic 4) atrophy associated with high EROD activity strongly suggests that the complex mixtures of contaminants in the Great Lakes exert toxic effects on the immune systems 1445) of young herring gulls. Our findings are (34) consistent with the results of several marine mammal studies. Harbor seals fed HAHcontaminated herring from the Baltic Sea had reduced delayed-type hypersensitivity (T-cell-mediated) and reduced mitogeninduced proliferative T-cell responses (67,68). Mitogen-induced proliferative Tcell responses were inversely correlated with blood PCB and DDE concentrations in male bottlenose dolphins from the west SagB coast of Florida (69).
The criterion of predictive performance supported associations between contaminants and suppression of T-cell-mediated aspian tern ifieaspiang t Immunity. Published laboratory studies incNrumbesn and our own pilot field study were used successfully to predict suppression of Tcell-mediated function at colonies with high organochlorine contamination. In this logically pilot study during 1991, the PHA skin rith birds response was 60% lower in herring gull -lymphochicks from a highly contaminated site immuno-(Gull Island, Upper Green Bay) as comhe PHA pared to a reference site (Kent Island, '). Hence, Atlantic coast).
herring The specificity criterion refers both to contamia unique effect produced only by a single presents a cause and to a consistent effect that always fhis may accompanies a causal factor. One difficulty n possible in immunotoxicological studies is that [ht loss in immunosuppression is not specific to Environmental Health Perspectives -Vol 104, Supplement 4 * August 1996   pollutants; many other factors such as nutrition, stress, infections, and genetics influence immune function. However, suppression of T-cell-mediated immunity consistently occurs after developmental exposure to HAHs in laboratory animals, supporting this association in Great Lakes birds. The time order criterion supported this association because chicks were exposed to environmental contaminants throughout development and after hatch, before T-cell function was measured. Unfortunately, there are no Great Lakes I of organochlc determined immune fune mination. C residues are a than during t immunosupF and possibly Suppressi occurred ove Caspian terns than in herring gulls ( Figure  2). The sensitivity of Caspian terns to the immunosuppressive effects of organochlorines is not surprising considering their sensitivity to other developmental effects of these pollutants (12). The lower contaminant concentrations in Caspian tern eggs are probably related to the migratory habits of terns. They migrate to southern North, Central, and South America for 6 months of the year, where they presumably eat a less-contaminated food supply when inhab-SagB iting ocean beaches. When they return to TE \ (20) breed at highly contaminated Great Lakes WErie sites such as Saginaw Bay, the female terns (10) accumulate contaminants throughout the breeding season so that second clutch eggs have higher organochlorine concentrations I and lower rates of hatching than first clutch 20 25 30 eggs (12). In contrast, herring gulls are year-round residents of the Great Lakes, so they are chronically exposed to higher concentrations of organochlorines. Long-term banding studies have shown low recruitment into the breeding population of Caspian terns raised at highly contaminat-UGB ed colonies (19,73,74). Contaminant-asso-UGB ciated immunosuppression provides a (10) potential mechanism for explaining these population-level effects.
Because most organochlorines biomagnify up the food web, their concentrations Mich tend to be co-correlated. Hence, it is difficult to determine which organochlo-10) (10) rines were most closely associated with \E0nt suppression of T-cell-mediated immunity (10) in this study. In herring gull eggs, gradi-SagB ents in PCBs and DDE all occurred in the (15) same rank order. However, PCBs occurred at much higher concentrations than DDE, and suppression of T-cell-mediated i 6 7 8 immunity following perinatal exposure fetwt) is more characteristic of PCBs than DDE. In Caspian tern chicks, the PHA skin ination in herring gull (A) and Caspian response decreased as total PCBs, C-Closed circles indicate mean response TEQs, and DDE increased. However, the in parentheses indicate sample sizes. strongest association was with PCBs. PCBs had the highest concentrations of any organochlorines in Caspian tern eggs. data on immune function in While PCBs were most closely associated fish-eating birds before the era with immunosuppression in both species, Drine pollution, so it cannot be effects of, and interactions with, other vhether these spatial patterns in organochlorines cannot be ruled out.
ction existed before this conta-Our epidemiological evaluation did not Considering that contaminant support the hypothesis of contaminant-Lpproximately 80% lower today associated suppression of antibody-mediatthe early 1970s, it is likely that ed immunity (Table 3), even though the )ression was more widespread differences in total antibody and IgG titers more severe at that time.  Table 1 for site codes.
dose-response relationship, and prediction from laboratory to field investigations) detracted from the hypothesis of antibody suppression. Other criteria were indeterminate. A posteriori Jonckheere tests indicated an association between organochlorine contamination and higher antibody titers in Caspian terns, suggesting contaminant-associated deregulation of antibody-mediated immunity. Other factors suchnutrition, stress, and weather r influenced antibody titers. The suppression of T-cell immunity but not antibody immunity in this field study tent with laboratory studies c and (or) perinatal exposure PCBs most consistently suppress mediated immunity at high acute doses rather than at the chronic developmental exposures observed in this field study. In harbor seals fed HAH-contaminated herring, antibody responses to ovalbumin were reduced (68), but mitogen-induced proliferation of B cells was not affected (10) (67). B-cell proliferation showed no associations with organochlorines in male bottlenose dolphins from the west coast of Florida (69).
Although the anti-SRBC antibody response requires helper T lymphocytes, the suppression of T-cell-mediated immunity but not antibody-mediated immunity in this field study is consistent with current immunological theory. The contaminantassociated suppression of the PHA skin response may reflect suppression of a sub-gB set of helper T lymphocytes that boost inflammatory responses but not antibodymediated responses. In mice, TH1 cells promote inflammatory cytotoxic responses while TH2 cells promote antibody responses. Suppression of the PHA skin response in herring gulls and Caspian terns suggests suppression of TH1-like cells that would not participate in antibody responses. In laboratory animals, HAHs have been shown to suppress such inflammatory (delayed-type hypersensitivity) and cytotoxic responses (24,32,33,42,45). In many cases, these cell-mediated responses are suppressed without any effects on production of antibodies that depend on helper T-cell (THI) activity (33,43,45).

Evaluation of Inmunological and Biochemical Biomarkers
The PHA skin test was an extremely effective and sensitive biomarker for assessing T-cell function in wild birds. Low-withinsite variation allowed statistical differentiant tion among sites, especially when sample sizes were greater than 20 birds per site. At sites where the test was replicated for 2 or 3 years, the response was consistent. For both From the Great species in Saginaw Bay, immunosuppresicreasing total sion was similar during the year of severe In. Numbers in body mass loss and mortality (1992) compared to the following 2 years of better growth and fledging success, suggesting as genetics, that T-cell-mediated immune function is a might have more sensitive end point. In northern bobwhites (Colinus virginianus), the PHA skin [-mediated test was more sensitive to the immunosup--mediated pressive effects of a low protein diet as is consis-compared to in vitro proliferation of T :n chronic lymphocytes in response to PHA (90). The to HAHs. in vitro T-lymphocyte proliferation assay s antibodymeasures only very early events in the T-cell  response that are involved with cell division.
In contrast, the in vivo skin test incorporates numerous events in the T-cell response, including cell proliferation, differentiation, and cytokine production (90,91). The SRBC hemagglutination test was a good biomarker for measuring antibody-mediated immune function in wild fish-eating birds, allowing statistical differentiation among sites. The IgG titers usually gave smaller p-values in ANOVA analyses as compared to the total antibody titers, suggesting that IgG is a more sensitive biomarker. A disadvantage of this assay is the need to make two visits to each colony 6 days (or 5-7) apart to collect blood at the peak of the antibody response. Poor weather and other logistical problems can make it difficult to return to a colony during this narrow time window, especially if immunological tests are being conducted simultaneously at several sites.
Where the hemagglutination test was replicated for 3 years, the response often changed over time. Time trends in antibody titers were confounded by a change in the individual sheep that served as a source of SRBCs. The sheep used during 1992 to 1993 died before the 1994 field season, so a sheep of similar age from the same flock was substituted. In Saginaw Bay herring gull chicks, a nonsignificant decrease in total antibody titers occurred over 3 years. The IgG response decreased significantly from 1992 to 1993 but not from 1993 to 1994. In Caspian tern chicks from Saginaw Bay and the North Channel, the total antibody response dropped from 1992 to 1993 and again from 1993 to 1994. The IgG response showed a significant year x site interaction. North Channel IgG titers dropped significantly from 1992 to 1994, although the 1993 titers were not significantly different than the early or later years. Saginaw Bay IgG titers dropped significantly from 1992 to 1993, but not from 1993 to 1994. In both herring gull and Caspian tern chicks from Saginaw Bay, IgG titers dropped between 1992 and 1993 while the same source of SRBCs was used. This drop in IgG coincided with a doubling in total PCBs and DDE in herring gull eggs from 1992 to 1993 (DV Weseloh, personal communicataion).
Few general biomarkers were correlated with measures of T-cell-mediated or antibody-mediated immune function, suggesting that they were not mechanistically responsible for the immunosuppression observed (Table 4). Although retinol and thyroxine often influence immune function in laboratory studies, there was little evidence that these variables were good surrogate biomarkers for immune function in this field study. Plasma retinol was strongly correlated with antibody responses in herring gull chicks in 1992 but less so when 1993 data were added. Retinol did not appear to influence immune function in Caspian tern chicks, although retinol concentrations were one to two orders of magnitude lower in terns than in gulls. Apparently, physiological regulation and (or) dietary intake of vitamin A differs greatly in these two species. Nonetheless, in both species plasma retinol decreased as PCB contamination increased, exhibiting a similar association with contamination as the PHA skin test (Figure 4). However, in herring gull chicks this association was strongly influenced by the North Channel site, which had much higher retinol values than Lake Winnipeg, the other reference colony. Plasma thyroxine showed a strong relationship to the PHA skin response in herring gull chicks in 1992, although this relationship was weaker when 1993 data Environmental Health Perspectives * Vol 104, Supplement 4 -August 1996 were added. There was no evidence for such a relationship in terns.
Although the tests of immune function were not confounded by growth, several other biomarkers were related to body size and (or) to length of exposure to pollutants, which increases with growth. Larger herring gull chicks tended to have higher concentrations of plasma thyroxine and retinol and more lymphocytes than heterophils. These relationships were evident even in chicks that fit body mass and wing chord criteria for approximately 21 days of age.
The heterophil to lymphocyte ratio integrates differential counts of the most abundant WBCs in birds. With respect to disease resistance, this ratio quantifies the balance between the nonspecific, fast-acting defenses of heterophils and the antigen-specific, slower acting defenses of lymphocytes. Heterophil to lymphocyte ratios are increased by stress (92) and may be influenced by other factors. The two species displayed different relationships between differential WBC counts and antibody responses. In herring gulls, the antibody responses increased as number of lymphocytes, some of which produce antibodies, increased relative to the number of heterophils. Conversely, in Caspian terns the antibody responses increased as the relative number of heterophils to lymphocytes increased. Total lymphocytes counts might clarify the antibody relationships to differential counts, but total WBCs can be difficult to count under field conditions.

Conclusions
This study demonstrated contaminantassociated suppression of T-cell-mediated immunity in prefledgling chicks of two species of fish-eating birds from the Great Lakes. Suppression was most severe at colonies in Lake Ontario (1992) and Saginaw Bay (1992-1994) for both species and in western Lake Erie (1992) for herring gulls. The identity of the particular organochlorine(s) responsible for such suppression could not be determined since exposure to different organochlorines was correlated due to similarities in environmental chemistry and metabolism. However, PCBs were the most closely associated with immunosuppression. Additional research is needed to determine the relationship between suppression of functional assays and increased susceptibility to infectious diseases. Such research will be important for determining the consequences of T-cell-mediated immunosuppression on individual survival and population dynamics. Alterations in variables such as retinol and WBC counts suggest biologically significant differences in physiology and health among various Great Lakes colonies.