| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Lipid Research, Vol. 48, 904-913, April 2007
Copyright © 2007 by American Society for Biochemistry and Molecular Biology
* Children's Hospital Oakland Research Institute, Oakland, CA
Department of Nutritional Sciences, Pennsylvania State University, University Park, PA
Mead Johnson Nutritionals, Evansville, IN
1 Part of this work was presented at the Federation of American Societies for Experimental Biology 2005, Washington, DC, by S.H.G. 
Published, JLR Papers in Press, January 18, 2007.
2 To whom correspondence should be addressed. e-mail: francisco.rosales{at}bms.com
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Supplementary key words hyporetinolemia • kinetic analysis • retinol • Windows version of Simulation, Analysis, and Modeling software
Abbreviations: AIC, Akaike's Information Criterion; fdose, fraction of the oral dose; IL-6, interleukin-6; LPS, lipopolysaccharide; RBP, retinol binding protein; rhIL-6, recombinant human interleukin-6; TMMP, trimethylmethoxyphenyl; VA, vitamin A; [3H]VA, [11,12-3H(N)]retinol; WinSAAM, Windows version of Simulation, Analysis, and Modeling software
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Several hypotheses have been proposed to explain inflammation-induced hyporetinolemia, including decreased intake, malabsorption, direct loss, increased requirement, and impaired utilization (5). The possible underlying mechanisms include the following: 1) impaired absorption or cleavage of dietary proformed carotenoids or preformed VA (6); 2) increased utilization of VA subsequent to its immune-potentiating role during infections (7); 3) increased glomerular filtration followed by an impaired reabsorption in the proximal tubule, leading to retinol and retinol binding protein (RBP) being excreted in urine (8); and 4) reduced hepatic synthesis of RBP and decreased secretion of retinol into plasma (9, 10). We have shown previously that inflammation induced with lipopolysaccharide (LPS) or recombinant human interleukin-6 (rhIL-6) decreases the availability of RBP and increases the concentrations of hepatic retinol or retinyl esters (9, 11). However, the effect(s) of this decrease in RBP and the increase in hepatic VA on the kinetics and dynamics of whole-body retinol flux during inflammation has not been studied previously.
In previous studies, model-based compartmental analysis has been fruitfully applied to describe and quantify the dynamics of whole-body VA metabolism and to understand the impact of various treatments on VA kinetics (12–17). A major observation from this type of analysis has been the extensive recycling of VA (i.e., retinol and retinyl esters) before it is irreversibly lost (18). Using these techniques, it was determined that the recycling of VA is affected by VA status (i.e., deficiency vs. sufficiency) (19) and by nutritional and chemical stressors, including iron deficiency (20), N-(4-hydroxyphenyl)retinamide treatment (21), and 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure (22). By comparing model parameters in different physiological, nutritional, or pathological states, the technique provides information on the sites of impact of various treatments. Our objective was to use mathematical modeling and model-based compartmental analysis to assess the potential mechanisms responsible for inflammation-induced hyporetinolemia.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Animals and sample collection
Male pathogen-free Sprague-Dawley rats were purchased from Charles River Breeding Laboratories (Kingston, NY). All experimental procedures were approved by the Institutional Animal Care and Use Committee at Pennsylvania State University. All animals were housed with a 12 h light/dark cycle at 22°C and provided diet and water ad libitum.
During kinetic studies, serial blood samples were collected from 2 h to 54 days after [3H]VA administration. Blood was collected from a caudal blood vessel (n = 34–36; 0.25 ml) into heparinized tubes (20 U heparin/ml blood). After centrifugation (Sorvall RT6000B, H-1000B rotor) at 1,300 g for 15 min at 4°C, plasma was collected, purged with nitrogen gas, and stored at –80°C until analysis. At the end of each experiment, rats were killed by carbon dioxide asphyxiation. Terminal blood was drawn via cardiac puncture into heparinized syringes. The animals were then perfused with PBS, and livers and carcasses were collected immediately by dissection, blotted, and frozen at –80°C and –20°C, respectively, until analysis.
Pilot experiment
A pilot study was initially conducted to establish parameters to use in the kinetic study, including the adequate oral dose of tracer ([3H]VA), the time for the tracer to equilibrate with body pools of VA, and the time the rats needed to recover from experimentally induced inflammation. At 2 months of age, VA-sufficient rats (n = 4) were given an oral dose of [3H]VA dissolved in 0.29 g of soybean oil (1.44 x 105 Bq/dose) using a polyethylene tube, a 21 gauge needle, and a 1 ml syringe. Twenty-one days later, VA kinetics was perturbed by continuous administration of either LPS (n = 2; 500 µg/kg/day) for 3 days or rhIL-6 (n = 2; 65 µg/kg/day) for 7 days via osmotic minipumps implanted subcutaneously. Sixteen to 19 d later, each rat was euthanized by carbon dioxide asphyxiation, and terminal blood and livers were collected and frozen until analyses.
Kinetic study
Upon arrival, weanling rats (n = 22) were fed 8.02 µmol retinyl palmitate/kg diet for 7 days. They were then fed 2.44 µmol retinyl palmitate/kg diet for the remainder of the study. The first diet was intended to increase their liver stores to 
200 µg (700 nmol), and the second diet provided the rats with 14 µg retinol equivalents/day assuming consumption of 20 g diet/day. These levels of VA were based on estimates from a previous study (23). The diet was made in the laboratory, and its composition (except VA) was based on the American Institute of Nutrition-93 rodent diet formula for growth, pregnancy, and lactation (24).
At 7 weeks of age, 14 rats received an oral dose of [3H]VA in soybean oil (9.3 x 105 Bq/dose) (Fig. 1 ). After equilibration of tracer with whole-body VA pools, inflammation was induced on day 21 with LPS (n = 5; 500 µg/kg/day) for 3 days or rhIL-6 (n = 5; 65 µg/kg/day) for 7 days via osmotic minipumps implanted subcutaneously. PBS was similarly given to controls for 3 days (n = 2) or 7 days (n = 2). At 0 and 21 days, additional rats were euthanized to determine baseline and perturbation liver VA concentrations (n = 4/day). We selected LPS and rhIL-6 at these doses because in our previous experiments (9, 11) we demonstrated that these doses induced inflammation without septicemia or renal failure, and they allowed us to examine the effects of the acute-phase response of inflammation on the distribution of VA and RBP in various tissues (9, 11). Additionally, rhIL-6 induced and maintained the acute-phase response of inflammation for up to 7 days without causing metabolic alterations or reducing food intake (11).
|
error level of 1% after a counting efficiency calibration was performed using a quench curve, and dpm was determined. Retinol concentrations in plasma for 17 of the samples collected over the experimental period for each rat were determined by HPLC (Hewlett-Packard 1100) using TMMP-retinol as an internal standard (25) after extraction with hexane. The reverse-phase system included a Zorbax Eclipse XDB-C8 (5 µm, 4.6 x 150 mm) with a C8 guard column (Agilent Technologies, Wilmington, DE) and a multiple wavelength ultraviolet light detector. Retinol was detected at 325 nm.
Frozen livers were lyophilized and analyzed for VA mass and/or radioactivity. Aliquots of freeze-dried livers were saponified using a modified method described previously (16). Triplicate weighed aliquots of liver (0.15 g) were combined with 1.0 ml of ethanol containing 0.1% pyrogallol as an antioxidant. TMMP-retinol (internal standard) and 0.5 ml of 60% aqueous KOH were then added. Samples were purged with nitrogen and allowed to incubate at 60°C for 60 min. After saponification, the lipid fraction was extracted three times using 5 ml of hexane containing 5 µg butylated hydroxytoluene/ml and 1 ml of distilled water. A portion of the extract was dried under a stream of nitrogen in liquid scintillation vials, 10 ml of scintillation solution was added, and its radioactivity was determined. A separate aliquot of the extract was dried under nitrogen, reconstituted in methanol, and analyzed for retinol by HPLC as described above.
Kinetic analysis
Model-based compartmental analysis was performed to develop the simplest compartmental model compatible with tracer and tracee data for plasma, liver, and carcass. The Windows version of the Simulation, Analysis, and Modeling software (WinSAAM) (26) was used for this purpose. First, the fraction of the oral dose (fdose) in plasma at each time was estimated as dpm/ml in plasma divided by administered dpm/estimated plasma volume (ml). Plasma volume was estimated from body weight on day 25 of the study (g) x 0.038 ml plasma/g body weight (27).
To determine a model of VA metabolism before the perturbation caused by inflammation, normalized plasma data before 21 days were used. Values in each group (PBS, LPS, and rhIL-6) were averaged to develop a preperturbation control model. Because the tracer was administered orally, an absorption/processing model first needed to be developed to account for this (28). After the absorption model was developed, a three-compartment model was developed, similar to a previously published conceptual model (23). Fractional transfer coefficients [L(I,J)s, or the fraction of compartment J that is transferred to compartment I per unit time] were then determined.
To account for the administration of the perturbing doses of PBS, LPS, or rhIL-6, a time interrupt was introduced into plasma data at day 21. Moreover, a second independent variable, theta, was introduced to allow for the adjustment of parameters during the perturbation. Fractional transfer coefficients before day 21 were fixed. After day 21, the fractional transfer coefficients were adjusted one at a time to find the minimal change necessary to fit the observed data obtained after the perturbations.
The LPS data were used to develop an initial model. Four potential causes explaining LPS-induced hyporetinolemia were investigated (increased urinary excretion, increased irreversible loss, increased movement into interstitial fluid, and decreased hepatic mobilization of retinol) by adjustment of kinetic parameters believed to represent these changes. Best-fit models were determined based on a statistically close fit of the model to the observed data (sum of squares). A two-part difference equation was used to represent the changes in one parameter. This equation used the second independent variable feature of WinSAAM to represent the time since the beginning of the perturbations. Changes to parameters in the model were as follows: 1) increased urinary excretion, added L(10,6) (see Table 2, Fig. 5 below); 2) increased irreversible loss, adjusted L(10,7); 3) increased movement into interstitial fluid, adjusted L(6,5) and L(5,6); and 4) decreased hepatic mobilization of retinol, adjusted L(5,7).
|
|
0.05 considered significant. The data for each treatment group (PBS, LPS, and rhIL-6) were averaged for the kinetic analyses using WinSAAM (26). Because compartmental modeling was done using group mean data at each time, it was not possible to compare kinetic parameters statistically between groups. With the present study design (preperturbation, perturbation, and postperturbation portions of the model), each group of rats served as its own control (i.e., data before the perturbation). For weighting purposes, model development based on observed data was set at a fractional standard deviation of 5% (SD/mean). The weighted sum of squares represents the sum of the squared differences between observed and model-predicted data points and was used to evaluate the closeness of the predicted model to the observed data. Akaike's Information Criterion (AIC) was calculated for each model to determine which had the lowest number (the best fit) for the observed data (29). |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
21 days. When the inflammation-inducing agents were administered beginning on day 21, there was an 63% reduction in plasma tracer levels over 6 days in the LPS-treated rats, whereas rhIL-6-treated rats had an 89% reduction over 10 days. Tracer in plasma for both treatment groups returned to preperturbation levels after inflammation. However, it was apparent that the postperturbation time after rhIL-6 treatment was not long enough to clearly see the postinflammation and terminal data. Therefore, the subsequent experiment was extended an additional 2 weeks.
Kinetic study: dynamics of retinol (tracee)
Hyporetinolemia was observed in rats treat with LPS (–42% plasma retinol concentrations relative to controls; P = 0.0004) or rhIL-6 (–57% plasma retinol concentrations relative to controls; P = 0.0002). Note that retinol concentrations in PBS controls decreased by 16% at 3 days after surgery (P = 0.14) (Fig. 2
). The fraction of [3H]VA in plasma decreased during inflammation and returned to control levels after 6 days in LPS-treated rats and after 16 days in rhIL-6-treated rats. In contrast to effects related to inflammation, plasma concentrations of retinol before and after perturbation were the same within each group, as indicated by the slope of the line describing this change (P > 0.20).
|
10 days after oral tracer administration. This terminal slope remained unchanged until day 21, at which point each rat was perturbed with PBS, LPS, or rhIL-6. In PBS control rats, plasma tracer was reduced slightly for the duration of the treatment period (7 days). In rats treated with LPS, there was a rapid reduction in plasma tracer that reached a nadir by 2 days and then increased to preperturbation levels by 6 days. In rats treated with rhIL-6, there was an even more rapid and larger reduction than in LPS-treated rats. This reduction in tracer persisted for 8 days, when it started to return to preperturbation levels. After the perturbation, the terminal slope was less shallow compared with the preperturbation terminal slope.
Specific activity of tracer and tracee
The specific activity of tracer in plasma (the ratio of the fdose to the concentration of plasma retinol) is shown in Fig. 3
. Before inflammation, changes in plasma retinol specific activity were similar for the three groups. The specific activity in both LPS- and rhIL-6-treated rats decreased at the beginning of inflammation and returned to PBS control levels after inflammation.
|
23–27%). Compartment 3 is a delay built into the model to represent the biological processes of chylomicron and hepatic retinyl ester metabolism. This processing or delay lasted 1 h. After this delay, the tracer entered compartment 4 and then immediately entered the postabsorption/processing plasma pool (compartment 5).
|
|
Kinetic analysis: potential causes of hyporetinolemia
To model urinary excretion, irreversible loss was added from the fast turning-over extravascular pool [compartment 6; L(10,6) (Fig. 5A)]. Even if the model was adjusted so that all VA entering the fast turning-over pool was irreversibly excreted, this did not adequately fit the observed VA kinetics and dynamics. The same was true when an increase in irreversible loss from the slow turning-over pool was modeled (Fig. 5B). It may also have been possible that VA moved into interstitial fluid during inflammation. To model this phenomenon, a temporary net input of VA into the fast turning-over extravascular pool was applied (i.e., increased movement of VA out of the blood into the fast turning-over pool and/or decreased movement of VA out of the fast turning-over pool into blood) (Fig. 5C). This model fit better the observed plasma VA kinetics and dynamics than the previous two models. However, biologically, this model of inflammation-induced hyporetinolemia was unlikely, because an elimination of irreversible loss of VA is required for this model to fit the observed data. In contrast, a 79% reduced mobilization of hepatic retinol (from 4.3 to 0.9 nmol/h) by 15 h best accounted for the observed changes in plasma VA kinetics in the LPS group (Fig. 5D), which was modeled using a two-part multiexponential difference equation. Model parameter changes for these potential causes are represented in Table 2
. Changes to parameters in the model were as follows: increased urinary excretion, add L(10,6); increased irreversible loss, adjust L(10,7); increased movement into interstitial fluid, adjust L(6,5) and L(5,6); and decreased hepatic mobilization of retinol, adjust L(5,7). The sum of squares and the AIC number indicated that the model that best fit the data was decreased hepatic mobilization.
The model of decreased hepatic mobilization of retinol was then applied to kinetic data from rhIL-6-perturbed rats. Similar to LPS treatment, the adjustment of one parameter provided an adequate fit of the model to the observed changes in tracer (Fig. 6
). At the beginning of the perturbation, an immediate reduction in L(5,7) accounted for the first 8 days since the perturbation. This parameter decreased by 75% (from 0.24 to 0.06 pools/day) with rhIL-6 treatment compared with a reduction of 76% (from 0.29 to 0.07 pools/day) with LPS treatment (Fig. 7
). The maximum reduction caused by LPS took 17.5 h versus 5 h with rhIL-6 treatment. After 8 days, the increasing portion of the exponential equation was applied to account for the increase in observed plasma tracer. This increase occurred at a slightly faster rate in the rhIL-6 model compared with the rate predicted by the LPS model.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In the kinetic study, rats received sufficient VA in the diet to maintain growth and other normal functions yet avoided the accumulation of total whole-body VA stores, as described previously (23). The absorption of tracer was estimated to range from 23% to 27%, lower than expected. Once the tracer was absorbed, the calculated time for enterocyte and hepatic processing of chylomicrons before it was secreted into plasma as the RBP-retinol complex averaged 40.2 min, and four kinetically different compartments were necessary to model this process (Fig. 4), as shown previously (12). Once the tracer entered the plasma, VA kinetics fit a three-compartment model proposed previously (23). In this model (Fig. 4), the central compartment represents plasma retinol bound to RBP- Transthyretin. Although in all animals, tracer had equilibrated with body pools of VA within 10 days after dose administration (Fig. 3A), inflammation was not induced until 21 days to ensure that the dose was highly equilibrated with whole-body pools; then, a control model could be identified. As expected, the parameters representing the preperturbation models for PBS, LPS, and rhIL-6 treatments were similar, because at this point all animals had been treated similarly. Irreversible utilization of VA averaged 36 nmol/day, which was greater than the 27 nmol/day that we estimated rats were consuming, suggesting that these animals were in a negative VA balance (e.g., marginal VA deficiency).
Whole-body VA metabolism was perturbed by surgery (implantation of osmotic minipumps) and then continuous administration of LPS or rhIL-6. The reduction of plasma retinol in PBS control rats was likely attributable to the surgery and implantation of the minipump and was self-limiting and not as substantial as that caused by either LPS or rhIL-6 (Fig. 2). It is interesting that although the duration of LPS treatment was 3 days, the plasma retinol nadir occurred after 1 day of treatment and then increased rapidly thereafter. In contrast, 7 days of rhIL-6 treatment caused a large reduction in plasma retinol concentration that was sustained to 10 days (3 days beyond the end of treatment) before increasing rapidly. These differences reflected the tolerance to endotoxin induced by repeated or continuous LPS dosing (38) and confirmed the appropriateness of using rhIL-6 to maintain the acute-phase response of inflammation in a model of prolonged hyporetinolemia (11). It is noteworthy that a long-lasting effect of IL-6 in inducing hyporetinolemia has been observed in clinical studies of infections caused by respiratory syncytial virus or measles virus (39, 40). Plasma retinol concentrations before and after hyporetinolemia were not significantly different in either of the treatment groups, suggesting that plasma retinol rebounds from inflammation-induced hyporetinolemia when the inflammatory stimulus is no longer present. In this regard, the study design allowed for the preperturbation and postperturbation comparison within each group.
Similarly, plasma tracer concentrations were affected more by LPS or rhIL-6 treatments than by surgery. Mathematical modeling using WinSAAM revealed that the effect of inflammation was acute and transitory. Although the reductions in plasma tracer paralleled the observed hyporetinolemia, the specific activity (the ratio of tracer to tracee) decreased during the perturbation with LPS or rhIL-6 (Fig. 3). This observation has significant implications for the use of a stable isotope dilution technique to measure VA status (28): if inflammation reduces plasma retinol specific activity, the stable isotope dilution technique will be less accurate.
As indicated above, several mechanisms have been proposed to explain inflammation-induced hyporetinolemia. These include an increased catabolism of VA, an increased transcapillary escape rate, and an increased urinary excretion of retinol (41). To examine these possibilities, we used the mean observed data from LPS-treated rats, because the shape of the tracer curve during the perturbation was less complex. The approach was to determine the smallest necessary change in the model that would account for the observed reduction in plasma tracer and the hyporetinolemia. Four possible models and their associated alteration(s) in parameters were tested (Table 2). The model based on a reduction in the mobilization of retinol from the liver into plasma best fit statistically the observed data (Fig. 5–7, Table 2).
The calculated reduced mobilization was based on a user-defined two-component multiexponential difference equation (Fig. 7). The first part of the equation describes the decrease in the parameter, and the second part represents the increase. The concept and basis for this equation are from a kinetic analysis performed previously (12), and an exponential function was chosen, because the components might have biological meaning. From a biological perspective, hyporetinolemia caused by a decreased hepatic mobilization of VA coincides with our previous findings of a decrease in both protein and mRNA levels of hepatic RBP and the accumulation of retinol and retinyl esters in the liver in response to LPS or rhIL-6 (9, 11). The prolonged hyporetinolemia caused by continuous rhIL-6 infusion was modeled by prolonging the reduction in the rate of mobilization of VA from the liver into plasma. Compared with LPS treatment, rhIL-6 caused a similar but earlier reduction in the rate of mobilization (–79% by 15 h for LPS vs. –75% by 5.6 h for IL-6). This reduction was maintained for 8 days before retinol concentrations increased to preperturbation levels in a manner similar to the modeled increase in LPS-treated rats (Fig. 7). This corroborates the concept that the effect of LPS on VA metabolism is through its induction of IL-6 synthesis by various immune cells, hence LPS's delayed onset in the reduction rate of mobilization. Subsequently, IL-6 presumably affects the VA system by decreasing hepatic RBP production (11).
In our previous experiments with LPS, we observed a decrease in food intake, which was associated with a mild and limiting reduction in plasma retinol (9). In the rhIL-6 model (11), we found that the continuous administration of rhIL-6 for 7 days did not affect food intake significantly. In the present experiments, we observed that LPS and rhIL-6 decreased plasma retinol (Fig. 2), and in both cases the compartmental analysis predicted a reduced mobilization in hepatic retinol (Fig. 7). These results suggest that regardless of changes in food intake (LPS-induced anorexia vs. rhIL-6's no effect on food intake), there was an accumulation of VA in the liver.
In contrast, the increased irreversible utilization and urinary excretion models were inadequate to describe tracer kinetics during the perturbation (Fig. 5, Table 2). It was impossible to account for the dramatic decrease in plasma retinol and its tracer by altering the parameters that represented these changes or to adequately increase plasma tracer concentrations once the perturbation was complete. For these postulated mechanisms, both the time (initiation, termination, and duration) and magnitude of the parameters were altered to attempt to account for the observed changes; however, these modifications did not improve the fit of the observed to the calculated data (Fig. 5). Comparatively, the hypothesis of increased movement into interstitial fluid provided a closer fit to the observed data (Fig. 5). Here, the size of the pool of VA in the fast turning-over extravascular compartment may have provided a site where VA could transiently move during inflammation, possibly as a result of an associated increase in vascular permeability. If this were the case, however, we would likely also have observed decreases in other similarly sized blood constituents in response to inflammation. Nonetheless, this model did not perfectly fit the observed data and therefore was judged not to be a major contributor to inflammation-induced hyporetinolemia (Table 2; AIC number –404 vs. –450 for decreased hepatic mobilization).
In summary, these findings and their interpretation indicate that inflammation-induced hyporetinolemia can be caused by trauma (i.e., surgery) or by bacterial or viral infections (i.e., LPS or rhIL-6). The duration and extent of the hyporetinolemia depend on the type of immune stressor. Based on our data, we conclude that the most likely underlying mechanism is a reduced synthesis of hepatic RBP (9, 11). Consequently, there is a reduced mobilization or an accumulation of hepatic VA, in part attributable to a reduced recycling of retinol from liver stellate cell retinyl ester stores back into plasma and from impairing the hepatic secretion of recently absorbed dietary VA. Both hyporetinolemia and the accumulation of hepatic VA reduce the availability of VA to extrahepatic tissues, especially those that depend on retinol bound to RBP. In this regard, these findings help to clarify the severity of measles infection and its amelioration with large doses of VA (i.e., 200,000 IU/day on 2 consecutive days) in VA-deficient children.
|
ACKNOWLEDGMENTS |
|---|
Manuscript received December 12, 2006 and in revised form January 17, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  C. J. Cifelli, J. B. Green, Z. Wang, S. Yin, R. M. Russell, G. Tang, and M. H. Green Kinetic Analysis Shows that Vitamin A Disposal Rate in Humans Is Positively Correlated with Vitamin A Stores J. Nutr., May 1, 2008; 138(5): 971 - 977. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS |
