|Home | About | Journals | Submit | Contact Us | Français|
There is considerable evidence that consuming fish has numerous health benefits, including a reduced risk of cardiovascular disease. However, fish is also the primary source of human exposure to mercury (Hg). In a cross-sectional study of 9–11 year old children (N = 100), we measured fish consumption, blood lipids, total blood Hg, diurnal salivary cortisol (4 samples collected throughout the day), and performed a proteomic analysis of serum proteins using spectral count shotgun proteomics. Children that consumed fish had a significantly more atheroprotective lipid profile but higher levels of blood Hg relative to children that did not consume fish. Although the levels of blood Hg were very low in these children (M = 0.77 μg/L; all but 1 participant had levels below 3.27 μg/L), increasing blood Hg was significantly associated with blunted diurnal cortisol levels. Blood Hg was also significantly associated with acute-phase proteins suggesting systemic inflammation, and several of these proteins were found to significantly reduce the association between Hg and diminished cortisol when included in the model. This study of a pediatric population is the first to document an association between blood Hg, systemic inflammation, and endocrine disruption in humans, in a pediatric sample. Without a better understanding of the long-term consequences of an atheroprotective lipid profile relative to blunted diurnal cortisol and systemic inflammation, a determination of the risk-benefit ratio for fish consumption by children is not possible.
There is considerable evidence that consuming fish is associated with a reduced risk of cardiovascular disease (CVD; (Mozaffarian and Rimm 2006). This is likely derived from fish being a source of long chain omega-3 polyunsaturated fatty acids, which have been shown to decrease CVD risk (Dolecek and Grandits 1991) and promote an atheroprotective lipid profile (Singer and Wirth 2004). These benefits to cardiovascular health have led to calls for greater fish consumption (Kris-Etherton et al. 2003; Organization 2010). However, there are also risks associated with exposure to environmental toxicants as a consequence of fish consumption. For example, the primary source of mercury (Hg) exposure in humans is through consumption of contaminated fish (US EPA 2006). Ionic Hg is a highly reactive heavy metal that is quickly converted by microorganisms in the environment into organic methylmercury (MeHg), which bioaccumulates in the food chain and is most highly concentrated in larger, predatory fish. As such, fish consumption recommendations are typically accompanied by warnings regarding how much and what kind of fish should be consumed, particularly for pregnant women and children (US FDA 2004). Nonetheless, pervasive chronic low-level Hg exposure is widespread. A recent study of New York City residents estimate that almost one quarter of NYC residents and nearly half of Asian New Yorkers have levels of blood Hg above 5.8 μg/L (New York City Department of Health and Hygiene 2007), the Environmental Protection Agency’s established level for potential health risks (US EPA 2011). Given that nonessential metals can have adverse effects at levels well below the current thresholds for identifying ‘elevated’ levels (Bellinger et al. 1992; Gump et al. 2011; Freire et al. 2010), it is important for us to understand the pathophysiological changes that may occur during chronic low-level metal exposure.
Chronic low-level Hg exposure appears to disrupt a number of physiological systems. For example, inorganic Hg (iHg) increases the release of pro-inflammatory cytokines from human immune cells in vitro (Gardner et al. 2009; Kempuraj et al. 2010). A cross-sectional study demonstrated elevated pro-inflammatory cytokines in miners exposed to elemental and iHg relative to levels in miners without occupational Hg exposure (Gardner et al. 2010). Unlike ethyl Hg, both iHg and MeHg were found to increase pro-inflammatory cytokine release in vitro in human peripheral blood mononuclear cells (Gardner et al. 2010). In addition, several types of Hg has been shown to be associated with endocrine dysfunction, as shown in catfish with reduced cortisol levels (Kirubagaran and Joy 1991). Similarly, the cortisol levels in yellow perch and northern pike following capture stress were reduced in populations exposed to a set of environmental pollutants including MeHg (Hontela et al. 1992). While human adult populations have not revealed a significant association between elemental Hg exposure and endocrine functioning (Langworth et al. 1990; Erfurth et al. 1990), research in this field is limited. As children are presumed to be more sensitive than adults to the effects of environmental toxicants (Faustman et al. 2000), such effects might be evident in children in the absence of any observable effects in adults.
The aim of this study was to elucidate the potential risks and benefits of fish consumption by children. Dietary and family histories, anthropomorphic measurements, blood, and saliva were collected from 100 children of ages 9–11. We determined the potential benefits of fish consumption to children’s lipid profile and the potential adverse consequences of low-level Hg exposure in children, including changes to acute-phase proteins and disruption of adrenocortical function.
Participants (N = 100) were recruited as part of an ongoing study designed to address the effects of nonessential metals on cardiovascular responses to acute stress. Using a direct mailing list, we mailed invitations to homes in Oswego County, NY, containing a child within our target age group of 9–11 year olds. This recruitment method elicits participation from a sample that closely resembles an eligible population and is cost effective (Hinshaw et al. 2007). Further inclusion criteria included: 1) reporting no use on the day of testing of medication that might affect cardiovascular functioning (e.g., Ritalin), and 2) having no significant developmental disorders that might affect task performance (a component of our broader study). A blood draw for measuring nonessential toxic metal levels was followed within 2 weeks by a laboratory visit. Children were paid $100 for their participation in the current study.
Parents were asked to report their child’s fish consumption using the food frequency questionnaire developed for the Oswego Children’s Study (Lonky et al. 1996). This dietary record included an item regarding fish (specifically, ‘Fish (including tuna)’). Responses were made using checkboxes for consumption frequency on a 9-point scale from none to 4+ servings/day. An additional survey requested data on lifetime consumption of 26 varieties of sport fish using an 8-point scale that ranged from never to 5/week. Because of the distribution of maternal reports of their child’s fish consumption (53 reported no fish consumption on the first survey and 73 reported no sport fish consumption), we created a single dichotomous variable for fish consumption that included no fish consumption in either category (N = 45) vs. some fish consumption in at least one category (N = 50). This variable was coded as 0 (no fish) or 1 (fish consumed) for use in regression models.
Fasting blood samples were collected in the morning. Serum was collected in a 4 mL Griener Vacuette® Serum Gel Evacuated Tube (Greiner Bio One North America, Inc., Monroe NC) and immediately shipped to the Oswego Hospital Laboratory (Oswego, NY). Total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and triglycerides (TG) were all analyzed on the Siemens Advia 1800 chemistry analyzer (Siemens Healthcare Diagnostics Inc., Deerfield, IL) per their protocol. Low-density lipoprotein cholesterol (LDL-C) was calculated using the Friedewald equation: LDL-C = (TC - HDL-C) - (TG × 0.20). As a measure of atherogenic dyslipidemia, we used the TG/HDL-C ratio (Grundy et al. 2005).
Whole blood specimens (2 mL) were collected into Royal Blue Top Vacutainer® (BD Franklin Lakes, NJ, USA) tubes that had been pre-certified by the analyzing laboratory for low-level measurements of total Hg and Pb (a potential confound, see below). Blood specimens were refrigerated pending shipment to the Trace Elements section of the Laboratory of Inorganic and Nuclear Chemistry at the New York State Department of Health’s Wadsworth Center, Albany, NY. The analysis for Pb and Hg in whole blood was carried out using a Perkin Elmer Sciex Model ELAN DRC Plus inductively coupled plasma-mass spectrometer (Shelton, CT, USA) equipped with a dynamic reaction cell (DRC-ICP-MS) (Palmer et al. 2006). For this study, the instrument was operated in the ‘standard’ rather than ‘DNC’ mode. The method detection limits (MDL) were 0.24 μg/L (Hg) and 0.34 μg/dL (Pb). Given the lower levels of blood Hg that were expected, the superior detection limits of ICP-MS were appropriate for this study and have been used in several biomonitoring studies (McKelvey et al. 2007). Four levels of internal quality control materials were analyzed before, during and after each run. The method has been validated against NIST SRM 966 (Toxic Metals in Bovine Blood), as well as a new standard reference material SRM 955c (Toxic Metals in Caprine Blood) that has been certified for Pb and Hg (Murphy et al. 2009). Laboratory values below the MDL were used in data analyses with the assumption that such values constitute the best available estimate of the true value and are preferable to assigning a zero or an arbitrary constant such as ½ the MDL (Fitzgerald et al. 2004; Stewart et al. 2008).
For those samples with sufficient serum (N = 95), proteomic analysis was performed as described earlier (Birdsall et al. 2010), with some changes. Briefly, blood was collected using Vacutainers (BD Vacutainer K2, REF 367856), centrifuged to remove any cellular debris, and stored in Protein LoBind microcentrifuge tubes (Brinkmann Instruments) at −80°C until further analysis. One mL of serum samples were packed in dry ice and shipped overnight to METACyt Biochemical Analysis Center (MBAC) located in the department of Chemistry at Indiana University, Bloomington, IN where seven highly abundant proteins (albumin, IgG, IgA, transferrin, haptoglobin, anti-trypsin, and fibrinogen) were depleted using Agilent Multi Affinity Removal System column. The remaining proteins underwent buffer exchange into 50 mM ammonium bicarbonate and were pre-concentrated to ca. 0.5 μg/μL using a 5 kDa MWCO spin concentrators.
After thermal denaturation, reduction, and alkylation, samples were treated with trypsin and LC-MS/MS analyses of the digests were performed in triplicate using a Dionex 3000 Ultimate nano-LC system (Dionex, Sunnyvale, CA) interfaced to LTQ Orbitrap hybrid mass spectrometer (Thermo Scientific, San Jose, CA). Prior to separation, a 1 μg-aliquot of protein equivalent was loaded onto a PepMap300 C18 cartridge (5 μm, 300 Å, Dionex) and eluted through the column (150 mm × 100 μm i.d, 200 Å pores) packed with C18(Michrom Bioresources, Auburn, CA). Peptides originating from tryptic digests were separated using a reversed-phase gradient from 10–55% of 99.9% acetonitrile with 0.1% formic acid over 3 hrs, at 350 nL/min flow rate. The mass spectrometer was operated in an automated data-dependent mode that was switching between MS scan and CID-MS. In this mode, eluted LC products undergo an initial full-spectrum MS scan from m/z 300 to 2000 in the Orbitrap at 15,000 mass resolutions, and subsequently CID-MS (at 35% normalized collision energy) was performed in the ion trap. The precursor ion was isolated using the data-dependent acquisition mode with a 2 m/z isolation width to select automatically and sequentially five most intense ions (starting with the most intense) from the survey scan. The total cycle (6 scans) was continuously repeated for the entire LC-MS run under data-dependent conditions with dynamic exclusion set to 60 sec. Performing MS scanning in the Orbitrap offers high mass accuracy and accurate charge state assignment of the selected precursor ions.
Mascot version 2.1.3 was used to search against Homo sapiens in the Swiss-Prot database using appropriate parameters (Birdsall et al. 2010). The quantitative analysis of proteins was carried out using the spectral count method (Liu et al. 2004; Lundgren et al. 2010). Using this method, 109 proteins were quantified and identified. However, proteins with very low abundance had progressively worse reliability across triplicate runs. Therefore, we reduced the set of reliably identified proteins to the 58 most abundant proteins. We tested distribution of spectral counts using the Kolmogorov-Smirnov test in Prism 4.03 (GraphPad Software Inc., La Jolla, CA) and log-transformed those demonstrating significant skewness.
After the blood draw, participants were sent home with a saliva collection kit containing 4 Sorbette saliva collection tubes (Salimetrics, State College, PA) and instructions regarding collection procedures and collection times (upon awakening, 30 min after awakening, 12 noon, and 6 pm). Participants were instructed on and practiced the procedures during the visit for the blood draw. Parents were also informed of the procedures and times to improve adherence. Participants were instructed not to collect samples within 60 minutes of eating a major meal and to rinse their mouth thoroughly 10 minutes before collection as dairy products and high acidic or high sugar foods can compromise the assay. After sample collection, participants were asked to store the collection tubes in their home freezer until their scheduled laboratory visit. Saliva was recovered from the collection tube via centrifugation and was stored at −80°C until used.
Salivary cortisol was measured using an enzyme immunoassay kit (Salimetrics) per the manufacturer’s instructions. Briefly, samples were thawed completely at room temperature and centrifuged at 1,500 × g to remove precipitated mucins. Standards (in duplicate), samples (in triplicate), and controls (in triplicate) were added to a 96-well microplate, a 1:1600 dilution of conjugated antibody was added to each well, the plate was placed on a shaker for 5 minutes, and then incubated at room temperature for 55 minutes. The plate was washed four times using a Bio-Tek ELx50 plate washer (Winooski, VT), 3, 3′, 5, 5′-tetramethylbenzidine solution was added, mixed on a shaker for 5 minutes, and then incubated at room temperature in the dark for 25 minutes. Stop solution was added and the absorbance at 450 nm for each well was measured using a Bio-Tek PowerWave XS plate reader. The absorbance of each sample, standard, and control was used to calculate the percent cortisol bound and the concentrations of the samples and controls were determined from the standards using a 4-parameter sigmoid minus curve fit calculated using KC Junior software (Bio-Tek). Pooled saliva was used as an internal control for all analyses. The intra- and inter-assay coefficients of variance were 4.7% and 6.8%, respectively.
Following standards in the field (e.g., (Cohen et al. 2006), measures of diurnal cortisol levels included the following: 1) log-transformed values (to approximately normalize the distribution) at each time point, 2) diurnal slope and intercept by separately fitting a linear regression line for each participant that predicted the log-transformed cortisol values from the actual collection times reported, 3) peak cortisol (maximum for the 4 samples for each participant), 4) mean cortisol (average across 4 samples), and 5) ‘area under the curve with respect to ground’ (AUCg). AUCg is calculated as the average log-transformed cortisol for two consecutive samples divided by the length of time between these samples, and these values are then summed across multiple time intervals representing the multiple collection times (Pruessner et al. 2003). Two participants did not return 2 or more saliva samples and were treated as missing all data for this measure. Four participants had 1 missing sample and linear extrapolation based on the remaining 3 samples was used to assign a cortisol level for this missing sample.
Potential confounders were chosen based on the National Institutes of Health guidelines for the inclusion of covariates in multivariate models, which recommends a prior selection of a limited set of variables shown in prior literature to relate to the outcome (Ewout and Harrell 2009). This approach avoids over-fitting a model that occurs when ‘cherry picking’ covariates from a larger set of potential confounds (Babyak 2004). The following covariates were included: gender, age, race, BMI percentile standing (age and gender adjusted using a SAS program developed by the Centers for Disease Control and Prevention (Prevention), family history of high cholesterol (yes vs. no for parents or grandparents), and socioeconomic status (SES). SES was a single measure derived from normalizing (using z-scores) and averaging the parents’ education (using an 8 level item), occupation (using a 9 level occupational classification developed by Hollingshead (Hollingshead 1975), and income (using a 9-level item). Because of the potential effects of Pb exposure on the adrenocortical system (Gump et al. 2008) this metal was measured as a potential confound and included as a covariate in all analyses of Hg associations.
All statistical analyses were performed using SAS software (SAS, Cary, NC). Hg was log transformed because the distribution of blood Hg levels was not normally distributed and showed a significant positive skew (skewness = 6.77, standard error of skewness = 0.24, z = 28.04, p < 0.05). Log Hg was analyzed in regression models (using SAS PROC REG). In addition to this approach, we analyzed nontransformed Hg by creating quartiles that contained a roughly equal number of participants and corresponded to the following blood levels: < 0.24 μg/L (1st quartile; N = 25), 0.25 – 0.45 μg/L (2nd quartile; N = 25), 0.46 – 0.79 μg/L (3rd quartile; N = 24), and 0.80 – 11.82 μg/L (4th quartile; N = 26). For the analysis of quartiles, SAS PROC GLM was used with a linear contrast to test the effects of increasing blood Hg levels. Partial correlations (using SAS PROC CORR) were used to analyze associations between serum proteins, Hg, and cortisol. In all analytic models, the seven covariates outlined above were entered first. Sample size varied slightly across analyses due to missing data (as outlined for each measure above). Analysis of salivary cortisol involved separate analytic models for each outcome (e.g., awaking and noon cortisol were analyzed in separate models).
For the proteomics data analysis, we increased our protection from Type I error by focusing on groups of proteins generated from a principal components factor analysis (PCA) with varimax rotation, a common approach to reducing the complexity of intercorrelated proteomics data (Lancashire et al. 2005; Meyrick et al. 2008). Because inflammatory processes are integrally related to cortisol (Miller et al. 2002), we tested proteins as potential mediators of the associations between Hg and diurnal cortisol as well as diurnal cortisol as a potential mediator for the association between Hg and proteins. Meditational modeling was performed using the Sobel test with bootstrapping (Preacher and Hayes 2004).
Table 1 presents characteristics of children in our sample. Our sample included 9, 10, and 11 year-olds (Ns = 47, 50, and 3, respectively), a roughly equal number of males and females (57 and 43, respectively), and was predominantly white (87%). The parents of these children had, on average, ‘some college’ (scale score = 5.4), a family income between $45,000 and $65,000, and an occupational status between ‘Clerical and Sales’ (Hollingshead score = 5) and ‘Technician or Semiprofessional’ (Hollingshead score = 6). The mean BMI percentile for our sample was 75.1% and 55.0% of the children had a family history of high cholesterol (at least 1 parent and grandparent with known high cholesterol). Based on parental reports, 50 children (52.6 %) consume either commercial fish (e.g., tuna) or sport fish at least once/week. Mean blood Pb in our study population was 1.00 μg/dL (90th percentile = 1.53 μg/dL). Mean blood Hg was 0.77 μg/L and with the exception of one participant at 11.82 μg/L, all participants had Hg levels below 3.27 μg/L. The 90th percentile for Hg was 1.55 μg/L. The average for our study population is well below the Environmental Protection Agency’s established level 5.8 μg/L for potential health risks (US EPA 2011).
This cross-sectional study of 9–11 year old children demonstrated some associations that are well established in adults. In these children, fish consumption was associated with a significantly more atheroprotective lipid profile. In unadjusted models (no covariates included), children consuming fish had significantly higher HDL-C (F (1, 93) = 4.44, p < 0.05), less dyslipidemia (F (1, 93) = 5.89, p < 0.05). However, they also had greater blood Pb (F (1, 93) = 7.20, p < 0.01) and Hg levels (F (1, 93) = 4.30, p < 0.05). As shown in Table 2, these associations remained after inclusion of covariates. In these multivariate models, children consuming fish had significantly higher HDL-C (F (1, 85) = 4.17, p < 0.05; full model R2 = 0.16), a better ratio of TC:HDL-C (F (1, 85) = 7.12, p < 0.01; full model R2 = 0.23), lower triglycerides (F (1, 85) = 6.76, p < 0.05;; full model R2 = 0.23), less dyslipidemia (F (1, 85) = 10.03, p < 0.01; full model R2 = 0.28), greater blood Pb (F (1, 87) = 8.05, p < 0.01; full model R2 = 0.25), and greater blood Hg (F (1, 87) = 14.48, p < 0.05; full model R2 = 0.23). While studies have demonstrated the benefits of fish consumption for lipid profiles in adults (Bulliyya 2002; Gunnarsdottir et al. 2008), particularly by increasing HDL-C (Smith et al. 2009), we are not aware of this being shown before in children. Fish consumption in children was also associated with significantly greater blood Hg, which is not surprising given that fish consumption represents the primary exposure route for Hg in humans. Therefore, the data indicates that while fish consumption in children leads to a more atheroprotective lipid profile, there is also a significant increase in nonessential toxic heavy metal exposure.
Some research suggests Hg-induced endocrine disruption in animals (Hontela et al. 1992) and humans (Langworth et al. 1990; Erfurth et al. 1990); however, we are not aware of research on the potential effects of Hg on cortisol levels. To determine if Hg exposure disrupts endocrine function in children, saliva was collected at four points in the day and cortisol levels were measured. We observed an association between increasing blood Hg and blunted diurnal cortisol. In unadjusted models, increasing blood Hg in children was associated with lower levels of cortisol as reflected by numerous significant associations (see Table 3), including: awakening cortisol (β = −0.27 p < 0.05; full model R2 = 0.16), awakening + 30 min cortisol (β = −0.24 p < 0.05; full model R2 = 0.10), 6 pm cortisol (β = −0.29, p < 0.01; full model R2 = 0.17), mean cortisol (β = −0.37, p < 0.001; full model R2 = 0.16), peak cortisol (β = −0.31, p < 0.01; full model R2 = 0.16), and AUCg (β = −0.31, p < 0.01; full model R2 = 0.16). Neither fish group nor blood HDL-C was significantly associated with diurnal cortisol in multivariate models (p values > 0.10). As such, the addition of these variables as covariates did not alter the association between Hg and diurnal cortisol. When testing for linear contrast (Table 3), increasing Hg quartiles were associated with significantly less noon cortisol (F (1, 76) = 4.53, p < 0.05; full model R2 = 0.14), 6 pm cortisol (F (1, 76) = 12.74, p < 0.001; full model R2 = 0.21), mean cortisol (F (1, 76) = 12.32, p < 0.001; full model R2 = 0.16), peak cortisol (F (1, 76) = 8.29, p < 0.01; full model R2 = 0.21), and AUCg (F (1, 76) = 7.27 p < 0.01; full model R2 = 0.15). Figure 1 shows the mean diurnal cortisol for the top and bottom blood Hg quartile. These findings clearly show that blood Hg levels are associated with blunted diurnal cortisol level.
In contrast to the limited research on Hg-induced endocrine disruption in humans, the effects of psychological stress on cortisol have been extensively studied and may offer a model on how chronic Hg exposure could blunt cortisol levels. The onset of a chronic stress produces high cortisol output; however, as time passes and the chronic stressor remains, the body is unable to mount a normal adrenocortical response and cortisol output declines to below normal (Miller et al. 2007; Dallman 1993). It is possible that Hg is acting as an ‘environmental stressor’ (Yunus et al. 2000) that blunts diurnal adrenocortical functioning. If so, this would be an important finding as a persistent lack of cortisol (whether it be induced by environmental or psychosocial stressors) may result in the development of stress-related physical and psychological disorders (Chikanza et al. 1992; Yehuda et al. 1993; Dreger et al. 2010).
We used a powerful exploratory tool, proteomics, to see if serum protein levels are altered due to Hg exposure. Shotgun spectral count proteomics provides a method to simultaneously identify and quantify numerous proteins in a heterogeneous sample. We used PCA (with rotation) to identify intercorrelated proteins that were grouped into ‘Factors’. A Scree plot from PCA suggested eight distinct factors, which was confirmed with a Minimum Average Partial (MAP) test (O’Connor 2000). Factor 1 (Eigenvalue = 16.96) explained the greatest variance relative to the other factors (all other Eigenvalues < 4.14). The remaining seven Factors revealed by the PCA contained fewer proteins loading at 0.60 or greater (3 or less for all factors) and were not readily interpretable. Fourteen proteins (Table 4) loaded strongly at 0.60 or greater on Factor 1 and of these, eight are clearly identified as acute-phase proteins (Gabay and Kushner 1999). In addition, the remaining proteins (e.g., alpha-1B-glycoprotein, gelsolin, and apolipoprotein A-II) may be implicated in the inflammatory process (Hashim et al. 2001; Bucki et al. 2005; Thompson et al. 2008). Many of the proteins had a positive association with Hg and a negative association with cortisol, except for apolipoprotein A-II, which had a marginally significant negative association with Hg and a significant positive association with cortisol (Table 4). Figure 2 illustrates the significant association between Hg and the combined Factor 1 score (r = 0.36, p < 0.001).
These findings suggest that increasing blood Hg levels are associated with inflammatory processes in children as indicated by significant associations with acute-phase proteins. This is important because these effects are being observed in children whose Hg levels are well below the levels presumed to result in negative health effects and pro-inflammatory cytokine activity and the acute-phase response have been implicated in the development atherosclerosis and diabetes (Ross 1999; Pickup and Crook 1998). Fish consumption was not associated with significantly different Factor 1 levels after controlling for our standard covariates (data not shown). Since these proteins (as well as diurnal cortisol) were not significantly associated with fish consumption, the observed proximal pathways (fish → Hg and Hg → proteins/cortisol) were presumably weakened somewhat by either consumption of fish with varying levels of Hg or alternative Hg exposure routes.
Sobel tests for mediation were used to test whether the association between Hg and cortisol can be explained by the acute-phase serum proteins. Six of the proteins were found to significantly reduce the previously significant association between Hg and mean salivary cortisol (Table 4). This cluster of proteins (particularly hemopexin and complement factor C3) may represent cortisol-mediated promotion of the class I genes involved in the acute phase of inflammation (Poli 1998). In addition, the combined Factor 1 score was found to significantly reduce the previously significant Hg-cortisol association (p < 0.05).
The immune and neuroendocrine systems are integrally connected, suggesting an ‘immunoneuroendocrine axis’ (Beishuizen and Thijs 2004). As such, we were unable to statistically differentiate the two potential meditational models: ‘Hg → inflammatory proteins → cortisol’ and ‘Hg → cortisol → inflammatory proteins.’ It is possible that Hg induces proinflammatory cytokine activity (Gardner et al. 2009; Kempuraj et al. 2010) and thereby increases acute-phase proteins (a proximal effect) and eventually produces a blunted diurnal cortisol (a more distal effect resulting from chronic elevations in acute-phase proteins). However, glucocorticoids (e.g., cortisol) also inhibit the inflammatory response and suppress the production and release of pro-inflammatory cytokines (Franchimont et al. 2002), which is seen as compensatory mechanism that restrains the inflammatory reaction and thereby prevents tissue damage (McEwen et al. 1997). Therefore, it is also possible that Hg-induced hypocortisolemia removes this restraint, resulting in greater systemic inflammation. The particular causal model linking Hg, cortisol, and acute-phase proteins remains to be determined.
First, the cross-sectional design has known weaknesses. For example, it is possible that diurnal cortisol or acute-phase proteins affect the toxicokinetics of this nonessential environmental metal. Second, we measured total Hg in our samples and therefore cannot address the relationship between the various forms of Hg (elemental, ionic, organometallic) and elevated acute-phase proteins and adrenocortical dysfunction. While most of the Hg present in human originates from fish in the food supply in the form of MeHg (Ravichandran 200) we cannot exclude the possible contributions of other Hg species such as elemental Hg exposure through dental amalgams, especially in gum chewers (Hansen et al. 2004). Third, if we consider that most of the total blood Hg is found as MeHg, we must also take into account that the whole-body half-life of MeHg is estimated to be 70–80 days (Aberg et al. 1969; US EPA 1997) and that health behaviors and dietary characteristics likely vary as these children age. Therefore, it is difficult to assess the potential clinical implications for the current Hg-associated changes in cortisol and systemic inflammation. Without repeated measurement of Hg in their blood, we cannot determine if our current associations represent a short-term response to acute exposure or a chronic state associated with long-term exposure. This is an important distinction; if systemic inflammation is associated with an increased risk for CVD (Kaptoge et al. 2010), it likely involves chronic inflammation in order to contribute to a disease that develops over many years. Fourth, our measure of fish consumption relied on parental retrospective recall, and studies investigating the correlation between fish consumption recall reports and Hg biomarkers have been mixed (Lincoln et al. 2011; Hertz-Picciotto et al. 2010). However, our observed association between the fish consumption reports and lipids and blood Hg provides some validation for the parental recall measure used in this study. Finally, we did not measure proinflammatory cytokines that are presumed to produce the increase in acute-phase proteins. In the absence of direct measurement of proinflammatory cytokines, the hypothesized meditational pathway (Hg → proinflammatory cytokines → acute-phase proteins) remains to be tested.
This study confirmed the well-known association between fish consumption and atheroprotective lipid profiles as well as increased blood Hg. It also demonstrated a significant association between blood Hg levels, acute-phase proteins, and reduced diurnal cortisol in children. Although some research has suggested an association between proinflammatory markers and Hg, this is the first study that we are aware of that has demonstrated this association in children. Similarly, while the association between blunted cortisol and Hg has been shown in animals, this is the first demonstration of such an association in children. Most importantly, the associations demonstrated here were significant at Hg levels well below the potential health risk level of 5.8 μg/L that has been established by the Environmental Protection Agency (2011). The present finding of elevated acute-phase proteins and adrenocortical dysfunction at levels of Hg typical for many US children and produced by fish consumption suggests potentially broad public health implications.
We are grateful for the assistance of the Oswego Hospital Laboratory and Dr. Robert Morgan (Oswego Family Physicians) for their help with blood specimen collection. The authors have no conflict of interests associated with this paper. The Human Research Committee of SUNY Oswego approved this study. Written informed assent and consent was obtained from all participants and their parents, respectively. This work was supported by National Institutes of Health grants ES15619 and ES15619-1S1 (to K.B., B.B.G., and J.A.M).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.