Compartmental analysis of plasma and liver n-3 essential fatty acids in alcohol-dependent men during withdrawal.

The mechanism by which chronic ethanol consumption reduces concentrations of long chain polyunsaturated (LCP) fatty acids (FA) in tissues of humans was investigated in alcohol-dependent (AD) men during early withdrawal and to a well-matched control group by fitting the concentration-time curves of d(5)-labeled n-3 FA from plasma and liver, which originated from an oral dose of d(5)-linolenic acid (d(5)-18:3n-3) ethyl ester to a compartmental model. Blood sampled over 168 h and a liver specimen obtained 96 h after isotope administration were analyzed for d(5)-18:3n-3, d(5)-20:5n-3, d(5)-22:5n-3, and d(5)-22:6n-3. Plasma 20:5n-3 and 22:5n-3 were lower in AD subjects, compared with controls (20:5n-3: -50%, 22:5n-3: -34%). Increased amounts of d(5)-18:3n-3 were directed toward synthesis of d(5)-20:5n-3 in AD subjects (P < .05). However, this effect was offset by larger amounts of 20:5n-3 lost from plasma (control: 2.0 vs. AD: 4.2 mg d(-1)). In livers of AD subjects, more d(5)-18:3n-3 and d(5)-22:5n-3 were utilized for synthesis of d(5)-20:5n-3 (+200%) and d(5)-22:6n-3 (+210%), respectively, than was predicted from plasma kinetics. Although, the potential to utilize linolenic acid for synthesis of LCP FA was greater in AD subjects compared with controls, heightened disappearance rates of 20:5n-3 reduced overall plasma concentrations of several endogenous n-3 LCP FA.

a-Linolenic acid, 18:3n-3, the most abundant n-3 essential fatty acid (EFA) in the North American diet, represents only a small percentage of the polyunsaturated FA intake (15). The LCP FA, eicosapentaenoic acid, (20:5n-3), docosapentaenoic acid, (22:5n-3), and 22:6n-3 are synthesized from 18:3n-3 through sequential steps of elongation and desaturation in hepatocytes (16), and stable isotope studies have indicated that men have a lower capacity to biosynthesize 22:6n-3 than women (17,18). Observational studies in humans that have examined the effects of alcohol abuse on EFA status without controlling for tobacco smoking can be confounded by the direct effect that smoking has on dietary EFA intake and metabolism (12,13,19,20).
Few animal studies have examined the direct effects of ethanol consumption on EFA metabolism using quantitative tracer techniques (6,7) and studies, which use tissue lipid compositional data as a means of comparison, are insufficient for determining mechanistic effects of ethanol on FA metabolism.
Previously, kinetic parameters relative to LCP FA biosynthesis in humans were determined using a quantitative tracer technique together with a compartmental modeling procedure (21). Recently, the complex effects of habitual smoking on plasma n-3 EFA status and metabolism has been reported in men and women consuming a metabolically controlled diet (22). These investigations have been further extended to an examination of plasma and liver kinetics of n-3 EFA metabolism in alcohol-dependent (AD) males paired with a group of chronic smokers. Plasma kinetic parameters were determined directly from the concentration-time curves of the d 5 -labeled FA 18:3n-3, 20:5n-3, 22:5n-3, and 22:6n-3 for each subject. This study also offered for the first time the unique opportunity to compare the plasma kinetic n-3 fatty acid profile to n-3 FA metabolism in the liver of humans directly. In vivo liver n-3-FA kinetics were extrapolated from the fatty acyl composition of specimens obtained 96 h after the d 5 -18:3n-3 ethyl ester had been consumed.

Subjects
Subjects provided informed consent (protocol # 92-AA-0194) and were evaluated at the clinical research unit of the National Institute on Alcohol Abuse and Alcoholism at the Clinical Center in Bethesda, Maryland. Subjects were evaluated by physical examination and clinical laboratory testing. Tests included hepatitis A, B, and C, and electrocardiogram, chest X-ray, nurse and social worker interviews, and psychiatric diagnoses by structured interview. Recent and chronic alcohol consumption was characterized using the Michigan Alcoholism Screening Test (MAST) (23) and a structured research questionnaire (24). Age of onset of alcohol abuse was calculated by subtracting years of excessive alcohol consumption from current age. Subjects were included in the alcohol dependency group if they smoked at least 20 cigarettes d 21 and met DSM-III-R criteria for current alcohol dependence as evaluated by the Structured Clinical Interview for Diagnosis (SCID) (25). Socioeconomic status was determined using the Hollingshead scale (26). AD subjects reported having their last drink within 7 days of admission and received benzodiazapines to treat alcohol withdrawal but were otherwise medication free during the time of the study. Control subjects (18-65 yrs) were included if they smoked at least 20 cigarettes d 21 (range: .
Exclusion criteria for the control group included: any major medical problems, including hepatic, endocrine, and metabolic disorders, a history of head trauma, seizures, prolonged loss of consciousness, a lifetime history of a major psychiatric diagnosis, abnormal clinical laboratory findings, or the consumption of more than the equivalent of two glasses of beer/wine d 21 . AD subjects were excluded if they had major medical problems or major psychiatric disorders that were unrelated to alcoholism. AD subjects with a history of pancreatitis or conditions interfering with fat absorption were excluded. Subjects were excluded from the study if they used prescription medications within the last month or if they persistently used over-the-counter medications including aspirin, ibuprofen, acetaminophen, antihistamines, topical steroids, vitamins E and C, multivitamins, herbal and home remedies if they had unusual dietary habits, or if they donated blood within the last 3 months. Absence of illicit drug use was confirmed by random urine testing; abstinence from alcohol consumption was verified by breath testing for ethanol.
AD subjects were inpatients and received the isotope an average of 5.2 d (range: 2-14 d; SD 0.93) after their last drink. Smoking controls consumed no alcohol for at least the prior 21 d and were admitted overnight prior to receiving the isotopes. All subjects followed an ad librium diet. After training by a registered clinical nutritionist in general features of keeping accurate food intake records, subjects recorded the types and quantities of all foods consumed for two wk. Food records were used to calculate energy and macro-and micronutrient intake including an estimate of dietary EFA.
After admission for an overnight fast, subjects received 1 g of the deuterated FA ethyl ester blended into 12 oz of lowfat yogurt prior to their breakfast. The isotope used was d 5 -17,17,18,18,18-18:3n-3 (Cambridge Isotope Labs, Woburn, MA). Blood was drawn under fasting conditions, at baseline, and at the following intervals over the following wk: 8, 24, 48, 72, 96, and 168 h and processed as previously described (22).

Liver biopsies and grading
Nine AD subjects underwent percutaneous liver biopsy using a 19 gauge, modified Menghini biopsy needle following alcohol withdrawal (mean: 6.4 6 0.7 d from the time subjects entered the study). Biopsy material was divided and immediately processed separately for FA analysis and histopathologic examination. Liver tissue slides were prepared in standard fashion, including staining with hematoxylin and eosin, and Massonʼs trichrome reagents. Biopsies were assessed and scored according to the Clinical Research Network (CRN) for nonalcoholic steatohepatitis (NASH): parenchymal inflammation (0-3), ballooning cellular injury (0-2), steatosis (0-3), and fibrosis (0-4) (27). The pathologist was blinded to patient identity and sequence of biopsies. The activity of the liver disease (NASH Activity Scores, NAS) was calculated as the sum of parenchymal inflammation, ballooning cellular injury, and steatosis (0-8) (27).

Plasma and liver lipid analysis
Analytical procedures have been described previously (8). Briefly, plasma (0.2 ml) and liver (80-100 mg) lipid extraction was carried out using a 1 ml solution of chloroform:methanol (2:1) and the FA methyl esters were prepared using a 14% solution of boron trifluoride in methanol. After extraction into hexane, the methyl esters were analyzed on a model HP-5890 gas chromatograph with flame ionization detection (Agilent, Wilmington, DE). Concentrations of individual FAs were calculated using the peak area counts in comparison with an internal standard.

Mass spectral analysis
From a 0.1 ml portion of the lipid extract, chloroform was evaporated and the lipids were hydrolyzed (5% potassium hydroxide in methanol) as described (28). The FA pentafluorobenzyl (PFB) esters were made and analyzed by gas chromatographymass spectrometry (GC-MS) according to conditions previously described (28). Data were acquired in the selected ion mode, monitoring the M-PFB anion and converted to the absolute quantity of the d 5 -FA by reference to the concentration of an internal standard.

Compartmental models
The hepatocyte is a main site for the biosynthesis of the 20-and 22-carbon LCP FA from 18:3n-3. Since the compartmental model for n-3 FA metabolism has been described previously (21) only a brief description follows. The model was developed from the plasma concentration-time curves of the labeled and endogenous n-3 FA using WinSAAM (Windows Simulation and Analysis Modeling) software (http://www.winsaam.com). The liver d-5 FA concentration-time curve values were extrapolated from the tissue fatty acyl concentrations of specimens obtained 96 h after isotope administration.

Liver n-3 FA concentrations
Total liver volume was estimated as 2.5% of each subjectʼs body weight, and liver blood volumes were estimated as 30% of the total liver volume where 70% of the total blood volume was assumed to be of portal circulation (29). To determine amounts of d 5 -labeled and endogenous n-3 FA within the liver proper, raw GC and GC-MS concentration data were adjusted to account for residual amounts of labeled-FA present in arterial blood volume (systemic circulation). Subsequent time point entrees for liver d 5 -n-3 FA concentrations were estimated using a modeling paradigm appropriate for coupled multicompartmental systems that synchronizes plasma d 5 -FA time course data to the liver compartment (30). Final d 5 -FA concentrations were adjusted for isotope fluctuations originating from arterial blood flow. Liver d 5 -FA rate constants were determined by fitting extrapolated values to modelderived time-course curves. Final kinetic values were determined using the programʼs iterative curve-fitting subroutines.
Fractional transfer rates, flow rates, percents, turnover, and errors The fractional rate constant coefficient, L (I,J) , represents the fraction of substrate, which is transferred from substrate-compartment J to product-compartment I. The units are in h 21 . The rate of flow, R (I, J) , from J to I is obtained by multiplying the mass M ( J) of unlabeled FA in compartment J by L (I, J) and is given in mg h 21 . The percentage of isotope transferred from J to I is given as P (I, J) and is the fractional synthetic rate (fsr) (i.e., the fraction of isotope remaining within the metabolic pathway as opposed to isotope taken up by tissues or irreversibly lost from the compartment). Initial fsr estimates were derived from the concentrationtime curves generated from the experimental data. Values assigned to kinetic parameters were then adjusted to compensate for each subjectʼs individual variances in the data (e.g., variances in isotope dilution due to differences in body mass) until the model prediction gave the best fit to the experimental determinants. Final values were determined using the programʼs iterative nonlinear least squares routine. Data points were weighted by assigning a fractional standard deviation of 0.1 to each measurement, which is consistent with the precision of the methods (21). The error model included the assumptions of independence, constant variance, and normal distribution about zero. Variances for the determined parameters are reported as standard deviation or coefficient of variance (cv) where appropriate.

Model and rate equations
Both liver and plasma models consisted of five compartments for which isotope data was obtained (Fig. 1). Compartment 1 represents the d 5 -18:3n-3 dosage and absorption through the gastrointestinal tract. Compartments 2 through 5 denote individual pools of the n-3 FA (18:3n-3, 20:5n-3, 22:5n-3, and 22:6n-3). Rate equations corresponding to the flux of the d 5 -FA substrates through their respective compartments are defined by a set of differential equations.

Limits, constraints, and statistics
Daily energy, macronutrient, and EFA intake for each subject were estimated using values determined from an analysis of each subjectʼs responses to the food frequency questionnaire. Upper and lower n-3 FA intake limits were assigned to each subject, which constrained 18:3n-3 daily intake values to within known limits.
Throughout the168 h, variances of all the endogenous plasma FA concentrations were less than 15% (range: 9 6 2% to 14 6 2%). Therefore, mean values of the plasma n-3 FA concentrations over the sampling period were used to approximate the steady state mass of the endogenous substrate M ( J) available for biosynthesis and these values were held constant (Table 4).
To determine differences in plasma FA concentrations and model-derived rate parameters between groups, an unpaired t-test analysis was performed. A P value of .05 or lower was considered significant.

Subject characterizations
Seven control and 12 AD subjects were recruited into the protocol. Of these, five control and nine AD subjects completed the study. Smokers were well matched to AD subjects in age, body mass index, cigarette smoking, and socioeconomic status ( Table 1). Medium age of smoking subjects was 36.5 yr (6 3.1) and that of AD subjects was 41.0 yr (6 2.3). Mean body weight for smoking subjects  (Table 1). AD subjects exhibited marked increases in serum bilirubin, lactate dehydrogenase, and GGTP relative to the smoking group, but decreases in urea and creatinine ( Table 2). Although, AD subjects had elevated plasma ALT/GPT and AST/GOT levels, values were widely distributed and differences between groups had not reached significance (P values: .09 and .07 for ALT/GPT and AST/GOT, respectively) ( Table 2). No differences were observed in the following serum measures: albumin, amylase, aldolase, ferritin, electrolytes, glucose, calcium, magnesium, zinc, phosphorus, uric acid, folate, vitamin E, vitamin B12, T3, T4, and thyroid stimulating hormone between the groups. AD subjects had a higher erythrocyte count and corpuscular volume but a decrease in prothrombin time compared with smoking controls ( Table 2).

Nutrient intake
Nutrient compositions were estimated for each subject based on their responses to the food frequency questionnaire using values obtained from the Minnesota Food and Nutrient Database (Table 1). Calculated caloric intake for control and AD subjects were 2,614 6 270 and 3,084 6 233 kcal d 21 , respectively. There were no differences in protein, fat, and carbohydrate intake between groups (smokers: 104, 96, and 342; AD subjects: 85, 102, and 240 g d 21 ) and no differences in estimated 18:2n-6 and 18:3n-3 FA intake values. None of the subjects reported eating fish during the 2 week study period, and intake of LC n-3 FA from other food sources was not significant.
Plasma FA and d 5 -18:3n-3 AUC Mean concentrations of plasma FAs obtained from blood samples taken over 168 h, and concentrations of liver FA from AD subjects are given in Table 3. AD subjects had significantly lower plasma concentrations of 20:5n-3 and 22:5n-3 and tended to have lower concentrations of 22:6n-3 (P , .1) compared with smokers. There were no differences in saturated, monounsaturated or other polyunsaturated FA between groups.  Values are expressed as mean 6 standard deviation. All clinical chemistries were from blood samples.
a Value is significantly different from smoking control group at P , .05 using Studentʼs paired test.
The model carried the assumption that fat absorption was essentially complete for all subjects (98%). Therefore, variances in the amount of labeled-precursor appearing in plasma represent differences in bioavailability among subjects. Bioavailability of d 5 -18:3n-3 from plasma was calculated as percent of dose (parameter, P (2,1) , Table 4) using isotope values obtained from compartments 1 and 2 ( Fig. 1). There were no differences in the percent of d 5 -18:3n-3 appearing in plasma between smokers (4.7%) and AD subjects (4.2%) and hence no apparent differences in bioavailability of the precursor. Plasma d 5 -18:3n-3 values may also be expressed in terms of area under the concentration-time curve (AUC) over the 168 h period (expressed here as ug ml 21 · h 6 SEM). There were no differences in AUC values for d 5 -18:3n-3 between these groups (smokers were: 51 6 17; AD subjects: 58 6 16) ( Table 3). Data in Table 3 may be used to calculate a tracer/trace ratio by dividing AUC by the appropriate unlabeled fatty acid. Disappearance rates of d 5 -20:5n-3 in plasma, quantified as fractional rate constant coefficients, L (0, 3) , are given in Table 4 (smokers: 0.0014 h 21 ; AD subjects: 0.0061 h 21 , P , .05). The 5-fold difference in the rate of disappearance of 20:5n-3 from the plasma in AD subjects emphasizes a greatly accelerated turnover rate and consequently a shorter plasma half-life of this FA. A somewhat lower turnover rate of 22:6n-3, L (0,5) , in AD subjects compared with smokers was another difference noted between groups (Table 4).
Among AD subjects the fractional synthetic rate (fsr) for d 5 -20:5n-3, P (3,2) , was twice that of smokers (smokers: 0.32; AD subjects: 0.72, Table 4). A larger fsr indicates that a higher percentage of 18:3n-3 is utilized for synthesis of 20:5n-3. As a consequence of the spike in the 20:5n-3 fsr occurring at this stage in the biosynthetic sequence, then both d 5 -22:5n-3 and d 5 -22:6n-3 also carried higher amounts of the deuterium label in the plasma of AD subjects relative to the controls (Fig. 2). However, there were no differences in the fsr for formation of either 22:5n-3 or 22:6n-3 (P (4,3) and P (5,4) ) indicating that chronic alcohol consumption had not negatively influenced the synthesis of either of these FA ( Table 4).

Liver FA and rate constants
The livers of AD subjects contained greater amounts of n-3 FA compared with plasma whether expressed in units of concentration (mg g 21 , Table 3) or in absolute amounts (M ( j) , Table 4). Also, the livers contained greater amounts of d 5 -n-3 FA compared with the plasma at 96 h ( Table 3). The mean fsr for formation of 20:5n-3, P (3,2) , in the liver was 1.5 (6 1.9) indicating a greater capacity for utilizing 18:3n-3 for synthesis in the liver than that which was predicted from the plasma kinetics (0.72 6 0.2) (Table 4). Also, the fsr for 22:6n-3 formation, P (5,4) , in the liver was greater (85 6 8) than that predicted from subjectsʼ plasma kinetic profiles (41 6 9). The higher fsr values together with the greater endogenous mass of n-3 FA indicates a potentially greater synthetic output of LCP n-3 FA from liver than was predicted from the plasma kinetic profile for subjects ( Table 4). The mean daily liver production of 20:5n-3, 22:5n-3, and 22:6n-3 were estimated as 17 (6 7), 71 (6 17), and 104 (6 15) mg, respectively, and these production rates were 3-, 5-, and 5-fold, respectively, above those predicted from the plasma kinetic profiles.

DISCUSSION
To our knowledge, this is the first detailed study examining the effects of chronic alcohol dependency on plasma and liver kinetics of essential n-3 FA metabolism in a group of habitual smokers during early alcohol withdrawal. A compartmental model, previously validated (21), was used to compare plasma kinetic rate profiles from AD subjects  going through withdrawal to those of a group of habitual smokers. We extended this analysis to an evaluation of liver n-3 FA kinetics in AD subjects using tissue fatty acyl profiles obtained from biopsied specimens. Compartmental mod-els based on total plasma values are often limited as they do not account for kinetics within individual lipoprotein fractions and a more detailed model of lipoprotein kinetics is desirable.   All but one AD subject presented with some degree of liver steatosis and inflammation, and one third of these subjects had initial stage fibrosis accompanied with other alcohol-related liver injury. There was no correlation in the amount of liver fat accumulation with any kinetic parameter of n-3 FA metabolism. Liver synthetic output was determined using mean values of the daily dietary intake of linolenic acid (averaged over 2 wk) for each subject as the principle n-3 FA for the biosynthesis of all other LCP FA. The predicted liver synthetic output of 20:5n-3, 22:5n-3, and 22:6n-3 from the livers of AD subjects was potentially greater than that determined from the plasma kinetic profile (Table 4). A greater synthetic output of LCP n-3 FA is consistent with the larger amounts of liver n-3 FA available for biosynthesis relative to the plasma and possibly the higher turnover rate of 22:6n-3 in the liver ( Table 4). The fsr for formation of 20:5n-3 and 22:6n-3 were also greater in the liver compared with the values from the total plasma profile, suggesting that perhaps an analysis of specific plasma lipoproteins may give a more accurate reflection of liver metabolism (Table 4).
Smoking (31,32) and ethanol metabolism (33)(34)(35)(36)(37) are well known to increase formation of oxygenated LCP FA. Since highly unsaturated FA are more likely to be oxidized, this may lead to decreases in the percentage and/or total amounts of the LCP FA in tissues. Also, 20:5n-3, an analog of 20:4n-6 may also be recruited as a potential cycoloxygenase substrate in cytokine-mediated inflammatory response processes. AD subjects who smoked had significantly lower plasma amounts of 20:5n-3 and 22:5n-3 compared with a well-matched group of habitual smokers having similar availability to dietary n-3FA.
A persistent question has been what effect ethanol has on LCP FA formation, desaturation, and catabolism. The desaturases (D-9, D-6, and D-5) dehydrogenate FA between specific carbon positions and in vitro studies across different species have suggested that an ethanol-induced decrease in liver LCP FA may result from an inhibition of either D-5 or D-6 desaturase (38-40). Others have found little or no change in desaturase activity (41), and findings from a cell culture study reported a positive effect of ethanol on desaturase activity across a wide range of concentrations (42). In vivo investigations of EFA metabolism in alcohol-consuming monkeys found no difference in the plasma uptake of labeled-18:2n-6 or -18:3n-3 and AUC for labeled-20:4n-6 and -22:6n-3 resulting from the metabolism of these precursors tended to be greater in the alcohol group than in the controls (7). A study in felines (6) also suggested that ethanol consumption may have had a positive influence on the production of labeled LCP n-3 FA based on AUC determinations.
Evidence obtained from the fatty acyl composition of livers suggested that desaturation was not affected in humans with alcoholic liver disease since concentrations of LCP n-6 FA, 20:4n-6, 22:4n-6, and 22:5n-6 (products of D-5 and D-6 desaturation) were no different than normal livers (43). However, lower amounts of LCP n-3 FA in these livers may have been the result of a low dietary supply of n-3 FA or perhaps greater catabolic rates. These find-ings were consistent with evidence from alcohol studies in micropigs (4), felines (9), and rhesus monkeys (7). An n-3 FA compartmental model was used to determine the efficiency of individual biosynthetic steps by estimating the fsr for LCP FA formation along the synthetic pathway. A greater percentage of 18:3n-3 was used for formation of 20:5n-3 in AD subjects who smoked compared with controls. Previously, we reported that men who smoked had a higher fsr for formation of 22:5n-3 compared with nonsmokers (smokers: 97; nonsmokers: 62), and in the present study both smokers and AD subjects had similarly high 22:5n-3 fsr values (smokers: 98; AD subjects: 91) (26). Male smokers subsisting on a low 18:3n-3 diet (0.59 g d 21 ) were also shown to have higher plasma turnover rates of 18:3n-3 (1330 mg d 21 ) compared with nonsmokers (432 mg d 21 ). In the present study turnover rates for 18:3n-3 in subjects on ad-librium diets were more moderate (smokers: 656 mg d 21 ; AD subjects: 421 mg d 21 ) reflecting perhaps the higher 18:3n-3 content of their diet (? 1.1 g d 21 ).
AD subjects had both a greater disappearance and turnover rate of plasma 20:5n-3 compared with smokers (L (0, 3) and R (0,3) , Table 4). An alcohol-induced utilization or catabolism of 20:5n-3, causing a deficit of this fatty acid, may have had an indirect effect on increasing D-6 desaturase activity to compensate for losses of 20:5n-3 (thereby also extending isotope accumulation to both 22:5n-3 and 22:6n-3) (44). However, the fsr values for formation of 22:5n-3 and 22:6n-3 were no different in AD subjects compared with controls ( Table 4), suggesting that enzyme systems leading to their production may be saturated or are perhaps more refractory to the ethanol effects. The FA kinetic parameters in AD subjects were consistent with observations of lower plasma concentrations of 20:5n-3, 22:5n-3, and 22:6n-3 compared with a group of smoking subjects and support the view that chronic alcohol consumption enhances specific anabolic processes in the metabolism of EFA while also enhancing the catabolism of 20:5n-3 in the plasma. These findings also appear to be consistent with observations of lower concentrations of n-3 FA in livers of humans with alcoholic liver disease (43). This study provided no evidence to support the view that chronic ethanol consumption inhibits desaturation of LCP FA in vivo, because production rates of 22:5n-3 and 22:6n-3 were not diminished and production of 20:5n-3 was greater in AD subjects during early stages of alcohol withdrawal.