A Human Model of Selenium that Integrates Metabolism from Selenite and Selenomethionine

Abstract

Selenium (Se) metabolism is affected by its chemical form in foods and by its incorporation (specific vs. nonspecific) into multiple proteins. Modeling Se kinetics may clarify the impact of form on metabolism. Although the kinetics of Se forms have been compared in different participants, or the same participants at different times, direct comparisons of their respective metabolism in the same participants have not been made. The aim of this study was to simultaneously compare kinetics of absorbed Se from inorganic selenite (Sel) and organic selenomethionine (SeMet) in healthy participants (n = 31). After oral administration of stable isotopic tracers of each form, urine and feces were collected for 12 d and blood was sampled over 4 mo. Tracer enrichment was determined by isotope-dilution-GC-MS. Using WinSAAM, a compartmental model was fitted to the data. Within 30 min of ingestion, Se from both forms entered a common pool, and metabolism was similar for several days before diverging. Slowly turning-over pools were required in tissues and plasma for Se derived from SeMet to account for its 3-times–higher incorporation into RBC compared with Se from Sel; these presumably represent nonspecific incorporation of SeMet into proteins. Pool sizes and transport rates were determined and compared by form and gender. The final model consisted of 11 plasma pools, 2 pools and a delay in RBC, and extravascular pools for recycling of Se back into plasma. This model will be used to evaluate changes in Se metabolism following long-term (2 y) Se supplementation.

Introduction

A factor contributing to the complexity of selenium (Se) metabolism is the form of Se ingested (1); selenomethionine (SeMet),¹² the dominant form in plant foods (2), is better absorbed than inorganic forms, is incorporated nonspecifically into proteins in place of methionine, and is recycled (3–6). A second factor is that Se from both forms can be incorporated as selenocysteine (SeCys) into as many as 25 selenoproteins (7) encoded by genes whose expression differs by tissue, nutritional regimen, and genotype (8). Thirdly, Se is excreted in the urine in several forms (9, 10). Finally, there also appear to be significant species and gender differences in how Se is metabolized (11). These factors confound the comparison of studies of Se metabolism.

Modeling may reduce the complexity by providing insight into the metabolism of Se when it is given in different forms, specifically the metabolic pools and their relationships, size, and turnover rates. Studies in humans (5, 6, 12, 13) have compared the kinetics of Se derived from 2 forms: SeMet and selenite (Sel), an inorganic form widely used in animal production and experimentation. Two multi-compartmental models, one for Sel (12, 14) and one for SeMet (13), described human Se metabolism. The models showed some differences in metabolism between the forms. However, because they were based on studies in 2 separate groups, it is unclear whether participant variability may have contributed to these differences. Other studies where participants were given both forms of Se but at different times (5, 6) also do not provide a direct comparison of the metabolism of those forms.

Because diet is composed of organic and inorganic forms, we wanted to determine the fate of the 2 forms in the same participants to establish whether they follow similar or different pathways. The aim of this study was to develop a single model to describe the absorption, distribution, and excretion of Se, when ingested as both Sel and SeMet, using stable isotopes as tracers.

Materials and Methods

Participants.

Participant recruitment and eligibility have been described elsewhere (11). Briefly, the participants were nonsmoking, healthy men (n = 16) and women (n = 15) 20–60 y old, within 20% of their ideal weight, consuming regular diets, not taking Se supplements of >25 μg/d (to convert μg to μmol, multiply by 0.0127) who ranked high on a Health Consciousness Scale (15), and whose plasma Se concentrations were between 80 and 160 μg/L. The study protocol was approved by the National Cancer Institute Special Studies Institutional Review Board, the Cornell University Committee on Human Subjects, the Johns Hopkins University Human Subjects Committee (for the Beltsville Human Nutrition Research Center), and the University of North Dakota Human Subjects Committee (for the Grand Forks Human Nutrition Research Center). This study was considered as a baseline study and was followed by a 2-y Se supplementation period in the same participants and a repeat kinetic study, described elsewhere (BH Patterson, GF Combs, Jr.,WK Canfield, PR Taylor, KY Patterson, AD Hill, JE Moler, and ME Wastney, unpublished data).

The reported sample consists of 7 males and 13 females. The other participants are not reported for the following reasons: 1 male was excluded during the study for underlying disease (Hashimoto’s disease), 3 males excreted very low levels of the ⁷⁶Se (from Sel), which was best explained by incomplete stool collection, as near total absorption of this form is counter-indicated by previous results (12, 13); 1 male had missing urine and fecal samples for 4 consecutive d; 1 male had extremely high urinary excretion of natural Se (i.e. ⁸⁰Se, 351 μg/d), which suggested that he may have been taking a Se supplement; and 1 female had very low urine (37 μg/d) and fecal Se excretion (23 μg/d), suggesting low intake and absorption, in direct conflict with her tracer kinetics. Another male completed only the baseline study; his results were excluded so that the participants reported here and for the supplemented portion of the study (BH Patterson, GF Combs, Jr.,WK Canfield, PR Taylor, KY Patterson, AD Hill, JE Moler, and ME Wastney, unpublished data) were the same. Preliminary kinetic analysis of the participants enrolled early in the study indicated the presence of multiple plasma pools. Because of the length of the baseline and supplementation studies (>2 y), we decided to retain (unanalyzed) the samples from 3 participants (2 males, 1 female) so that once methods were developed, they could be analyzed biochemically to identify the pools.

Se tracers.

The tracer solution was prepared at the BHNRC. It contained 150 μg Se each of enriched ⁷⁴Se as L-SeMet (⁷⁴SeMet) (Amersham Laboratories) and ⁷⁶Se as sodium selenite (Na₂⁷⁶SeO₃) (Oak Ridge National Laboratory). The amounts of tracer in the respective stock solutions were determined by isotope dilution and the amount for each dose was aliquoted into individual glass tubes.

Isotope administration and sampling.

Fasting participants were given the tracers orally, in water, on d 0 and 10. The second dose was given to ensure detection of tracers throughout the 4-mo period of observation, based on previous results (12, 13).

Sample collection.

Participants resided in the Cornell Metabolic Research Unit for the first 38 h of the study, during which time the tracer doses were administered and the early specimen samples were collected. An indwelling venous canula was installed and maintained for 38 h. Blood (10 mL) was drawn immediately prior to and at 30 min after the initial dose, then hourly from 1 to 14 h, at 2-h intervals from 14 to 20 h, and at 4-h intervals from 20 to 32 h. All subsequent samples were drawn by venipuncture after an overnight fast on d 3–12, 14, 21, 30, 45, 60, 75, 90, and 105 after initial dosing. Urine collections were made for 1–2, 4–6, 6–8, and 8–24 h periods followed by 24-h collections for d 2–12 after initial dosing. Fecal collections (24 h) were made the day preceding the initial dose day, then for the next 12 d. A blood sample (25 mL) was drawn from fasting participants at the end of 4 mo to reconfirm the health status of the participants by clinical chemistry.

Chemical analysis.

Urine, fecal, RBC, and plasma samples were analyzed for both Se tracers as well as for total Se. These samples were spiked with an internal standard, digested using a method that has been previously described (16, 17), treated with 4-trifluoromethyl-o-phenylenediamine as the chelating agent, and assayed by isotopic-dilution GC-MS (model PQ5050A, Shimadzu) with single ion monitoring (17, 18). The enriched ⁷⁴Se, enriched ⁷⁶Se, and total Se contents of the samples were determined by triple isotope dilution using enriched ⁸²Se as the internal standard. Certified reference materials were analyzed along with the samples; National Institute of Standards and Technology standard reference materials included Wheat Flour 1567a, Bovine Liver 1577a, and Bovine Serum 1598. The control material for the urine samples was Seronorm Trace Elements Urine, lot 108 (Nycomed Pharma).

We measured the enriched isotopes in plasma, RBC, urine, and fecal samples, but not the chemical forms of those isotopes in the samples. Although we gave ⁷⁶Se as SeMet, we can only say that the measured species was derived from SeMet; therefore, we have noted the species measured in italics, namely SeMet refers to a Se-containing compound originating as SeMet and Sel refers to a species originating as Sel. We considered natural Se to consist of either SeMet or Sel-exchangeable Se (meaning that Sel can be changed chemically into other Se-containing forms, denoted ‘Sel-like’).

Concentrations (ng/g) of total Se, ⁷⁴Se, and ⁷⁶Se in plasma were converted to total masses by correcting for plasma-specific gravity [assumed to be 1.026 (19)] and plasma volume [assumed to be 4% of body weight (20)]. The same measurements in RBC were converted by correcting for RBC-specific gravity [assumed to be 1.09 (19)] and RBC volume [assumed to be 3% of body weight (20)]. Urine and fecal data were fitted as the cumulative, or daily, amounts (μg or μg/d) excreted over the collection period.

Kinetic analysis.

The time-course data for both tracers in plasma, RBC, urine, and feces were analyzed by compartmental modeling using the WinSAAM software (21) as described in the Supplemental Material. Briefly, once the final model was developed, we individually adjusted the parameters for each participant, as in previous studies (12, 13), to obtain a best fit of the data. This fit was judged both visually and by the sum of squared deviations of the model-calculated values from the observed data. There is no accepted test for goodness of fit: here, the model fit was judged acceptable when there were no consistent deviations between the data and calculated values. When the model had been fitted to all participants for ⁷⁶Se (from Sel) and ⁷⁴Se (from SeMet) separately, it was used to estimate the transfer coefficients, L(i,j) (fraction/h), masses of pools of Se in the body (μg), transport rates between pools (μg/h), delays (h), and the turnover times of each pool as the reciprocal of the sum of all loss pathways from each compartment (h) .

Statistical analyses.

Parameter values were averaged across participants and were expressed as mean ± SEM. Parameter values for each form were compared using paired t tests for all participants, males and females, and using t tests for females compared with males. Statistical software used for the analyses was SAS version 9.1.3 (SAS Institute). All P-values shown are nominal and unadjusted for multiple comparisons; P < 0.05 was considered significant.

Model description and calculations

The frequent plasma sampling between 8 and 24 h in the current study (Fig. 1A) showed a more complex picture than our previous study (13) and the initial model containing 4 plasma pools was not able to fit the new data (Supplemental Fig. 1). It was therefore extended (Supplemental Fig. 2) by adding the minimum number of pools and pathways necessary to fit the observed data (Figs. 1 and 2). Goodness of fit of the model to observed data for each participant is shown by the sum of squares (Supplemental Table 1).

FIGURE 1.

Tracer data for a female participant in plasma 35 h (A), 250 h (B), and 2880 h (C) after oral dosing, showing the similar timing of the peaks between Sel and SeMet, for RBC (D) cumulative urine (E), daily urine (F), and feces (G). Symbols are observed values; lines are values calculated using the model (Fig. 4). Plasma data from a male participant 0–260 h on semi-log scale showing divergence between the forms after ~40 h (H).

FIGURE 2.

Fit of the model (Fig. 4) to the multiple peaks in plasma SeMet data with plasma pools (identified by name) over 30 h (A), for 250 h (B), and for the whole study (4 mo or 2880 h) (C).

The model schema (Fig. 3) shows the major pathways. Note that pools are labeled (rather than numbered), but these are only tentative descriptors. Se moves from the intestinal tract into enterocytes from which it travels in 1 of 3 directions; into plasma without being taken up by liver; via the hepatopancreatic subsystem, or, alternatively, the lymphatic system into plasma; or taken up by liver. From there, some is returned to the intestine via pancreatic secretions or bile, some enters plasma, and the rest enters a pool considered to be tissues. From this pool, Se moves through delay compartments into plasma or into other tissue pools, including the RBC. From plasma, Se returns to liver, tissues, or urine. A more detailed version of the model (Fig. 4) shows the multiple pools in the intestine, liver, RBC, and plasma. Specifically, Se moves from the first tissue pool, Tissue-1, through delays that range from 4 to 175 h into 10 plasma pools that have unique appearances and turnover times (Fig. 2). An additional plasma pool, Pl-11 (Fig. 4), was only required for SeMet data and was considered to account for nonspecific incorporation into proteins. From each plasma pool, Se returns to Liver-1 or is excreted, except for Plasma-3 and Plasma-9, which return Se to Tissue-1 and Tissue-2, respectively. From Tissue-2, Se moves into RBC (which consist of a delay and a pool, RBC-1, or for SeMet also RBC-2) or Tissue-3. All of Sel from Tissue-3 was excreted, whereas some SeMet was released into Plasma-11. Note that the category labeled “kidney” could be the bladder; it is the pool from which tracer was irreversibly lost into urine. The distribution of Se among the categories is expressed as percentages of Se leaving each pool (Fig. 4).

FIGURE 3.

Model schematic for Se metabolism in humans showing compartments (Supplemental Fig. 2) grouped into categories with putative physiological labels. Dotted arrows out of plasma indicate pathways that existed for only certain plasma pools (Fig. 4).

FIGURE 4.

Model with compartments grouped into categories, shown by the dotted boxes and putative labels shown in quotes. (The full model with compartment numbers is provided in Supplemental Fig. 2 and with compartments grouped into categories with numbers and labels in Supplemental Fig. 3.) The numbers in italics next to the arrows are the distribution (%) from 1 category to another [a single value represents both forms; if the forms differ, the ⁷⁴Se (from SeMet) value is the lower value]. Delay times (h) were the same for both forms except Delay-175 h was 143 h for Sel and the RBC delay 1908 h was 1387 h for Sel.

Model calculations.

The model was used to estimate the amounts of SeMet and Sel-like Se (which would include SeCys) consumed by these participants (22). We assumed that all Se in the diet behaved as 1 of these 2 forms. Specifically, the unknown fractions of the 2 forms in the diet were calculated using 2 equations: 1) by assuming steady-state conditions, the diet Se intake (consisting of the 2 forms) was equated to the total Se excreted in urine and feces; and 2) from the kinetics, total Se absorbed was calculated as well as the percent absorption of each form. Solving these equations algebraically yielded the amounts of Sel-exchangeable Se and SeMet-Se in the diet (11), termed Diet-Sel and Diet-SeMet, respectively. To calculate the actual pool sizes in the body, we combined the calculated mass for each pool (from fitting each tracer separately) in the same ratio as that of Diet-Sel and Diet-SeMet, which we termed the pool Diet mass. The transport rates between pools were calculated in the same manner. All Diet-pool masses were summed to provide the total body mass of Se (mg).

Results

Participants were similar in age, weight, and plasma and RBC Se concentration (Table 1). Urinary and fecal Se excretion were both higher in males than in females (Table 1). The levels of ⁷⁴Se (from SeMet) exceeded those of ⁷⁶Se (from Sel) in plasma, RBC, and urine, but the reverse was observed in feces, as shown for a female (Fig. 1). Over the first 24 h after dosing, a series of peaks was seen for both tracers with similar timing but differing magnitudes ( ). Plasma curves from a male participant showed apparent divergence after ~40 h (Fig. 1H); this divergence occurred in both genders but was more pronounced in males.

TABLE 1.

Age, weight, and Se measurements of participants¹

	All, n = 20	Males, n = 7	Females, n = 13
Age, y	40 ± 3	39 ± 6	40 ± 3
Weight, kg	70 ± 3	77 ± 7	66 ± 3
Plasma Se,2μg/L	134 ± 3	141 ± 6	131 ± 4
RBC Se, μg/L	231 ± 7	236 ± 13	227 ± 8
Urine Se, μg/d	71 ± 4	84 ± 9	64 ± 4*
Fecal Se, μg/d	36 ± 2	44 ± 4	31 ± 2*

¹Values are means ± SEM. *Different from males, < 0.05.

²To convert g to μmol, multiply by 0.0127.

By calculation (22), the diet contained more Sel-like Se than SeMet (Table 2). Se that behaved like Sel was possibly SeCys, which, like Sel, is metabolized to hydrogen selenide (H₂Se) (9) and in rats has similar kinetics to Sel (23). However, because the fractional absorption of Sel was lower than SeMet, more Se as SeMet than Sel Se was actually absorbed from the diet (Table 2). There were no gender differences in percent absorption, but because males were estimated to have higher Se intakes than females, they absorbed more Se than did females (Table 2).

TABLE 2.

Calculated values for Se intake; absorption of Se, Sel, and SeMet; daily intake of Se, Sel, and SeMet; and amount of Se in the whole body¹

	All, n = 20	Males, n = 7	Females, n = 13
Se intake,2μg/d	107 ± 6	128 ± 11	96 ± 5*
Se absorption, %	73 ± 1	72 ± 3	73 ± 2
Se absorbed, μg/d	79 ± 5	92 ± 10	70 ± 4*
Sel:SeMet intake3	60:40	58:42	61:39
Sel intake, μg/d	64 ± 5	74 ± 12	59 ± 4
Sel absorption, %	57 ± 2	54 ± 2	58 ± 3
Sel absorbed,4μg/d	36	40	34
SeMet intake, μg/d	43 ± 5	54 ± 11	37 ± 4
SeMet absorption, %	97 ± 0.2	98 ± 0.2	97 ± 0.3
SeMet absorbed, 4μg/d	42	52	36
Total body Se, mg	21 ± 1	25 ± 3	18 ± 1*

¹Values are means ± SEM of those calculated for each participant by the model (Fig. 4). *Different from males, < 0.05.

²To convert g to mol, multiply by 0.0127.

³Ratio in intake of Sel-exchangeable-Se:SeMet-Se.

⁴Calculated as % absorption × SeMet (or Sel) intake.

Modeling results are compared by Se form in terms of fractional transfer rates, pool sizes, body distribution, turnover times, and contributions to urine for all participants and then by gender. Interpretation of the model is provided in the Discussion. There were relatively few differences between parameter values calculated for SeMet and Sel from fitting the model (Fig. 4) to tracer data in plasma, RBC, urine, and feces for all participants (Supplemental Table 2).

Significant differences (Table 3) affecting body distribution of SeMet compared with Sel included higher movement to enterocytes from the gastrointestinal tract (GI) (Fig. 4), namely absorption (Table 1). Common to both forms, following absorption, the liver appeared to extract 50% of the Se, with the remainder staying in plasma or entering the lymphatics (Fig. 4). Distribution from Liver-2 differed between the forms, with higher endogenous excretion of SeMet and less going to Tissue-1 (Fig. 4). From Tissue-1, 70% of both forms was released through various delays into plasma. More SeMet went through Delay-19 into Plasma-8 (Fig. 4). RBC uptake from Tissue-2 was 2-fold higher for SeMet than for Sel (Fig. 4). Together with the higher absorption of SeMet, this greater uptake explained the 6-fold higher SeMet than Sel tracer in RBC (Fig. 1D). The RBC delay was longer for SeMet (Table 4). This longer delay affected the length of the plateau of the RBC curve but not the height (Fig. 1D). Urinary excretion from 4 plasma pools was lower for SeMet than for Sel (Table 3). For 2 of the pools (Plasma-5 and Plasma-7), this was a small fraction of the pool. By contrast, the difference between excretion of the forms from Plasma-9 was large (Fig. 4). This difference largely explained the divergence in plasma curves at ~40 h between forms, which was more pronounced in males (Fig. 1H) than in females (urine to Plasma-9) (Table 3).

TABLE 3.

Comparison of parameter values between Sel and SeMet in all participants and in each gender, and gender differences in the metabolism of each Se form¹

Parameter		SeMet vs. Sel2			Sel3	SeMet3
To	From	All, n = 20	Males, n = 7	Females, n = 13	Females vs. males	Females vs. males
	Ratio			% difference
Enterocytes	GI-1	2.42		2.85
Enterocytes	GI-2	1.63		1.54
Tissue-1	Liver-2	0.53	0.52	0.53		55
Pancr/bile	Liver-2	1.88	3.22	1.57	138
Plasma-3	Liver-2	0.74	0.65	0.79
Delay-9	Tissue-1				29
Delay-13	Tissue-1				22
Delay-19	Tissue-1	1.54
Delay-30	Tissue-1					−67
Liver-1	Plasma-2		1.36
Liver-1	Plasma-3					−83
Liver-1	Plasma-6				−51
RBC delay		1.38	1.44	1.34
Tissue-2	Plasma-9					−90
Urine	Plasma-2	0.39	0.10			408
Urine	Plasma-7	0.03		0.02
Urine	Plasma-5	0.05		0.06
Urine	Plasma-9	0.07	0.01	0.23	−79
Urine	Tissue-3	0.42

¹Parameters refer to the model (Fig. 4). Only L(i,j) or compartment with a delay time that differ significantly (P < 0.05) for Sel vs. SeMet or males vs. females are shown (see Supplemental Table 2 for all parameter values). Additionally, movement down the GI was slower for SeMet).

²Ratio is L(i,j) value for SeMet/value for Sel.

³Differences are calculated as [100 × (value for females − value for males)/(value for males)]. Negative values mean that values for female were lower than values for male.

TABLE 4.

Calculated mass of Se in body pools and differences by gender, whole body distribution of body Se, and turnover times of Sel and SeMet in each pool¹

	Diet mass23	Female vs. male4	Whole body distribution	Turnover time
Pool				Sel	SeMet
	μg	% difference	%	h
Enterocytes	0.12 ± 0.01	-30	0.0006	0.03	0.03
Lymphatics	0.33 ± 0.04	-48	0.002	0.62	0.52
Delay-7	0.76 ± 0.09	ns5	0.004	7.6	7.4
Tissue-1	0.99 ± 0.19	ns	0.005	0.09	0.07
Liver-1	1.43 ± 0.26	−52	0.007	0.10	0.10
Liver-2	2.80 ± 0.63	−60	0.013	0.12	0.18
Kidney	10.2 ± 1.71	ns	0.05	3.15	0.85
Delay-4	11.5 ± 2.56	ns	0.06	4.2	4.4
Delay-175	12.4 ± 1.30	ns	0.06	143.	175
Delay-30	12.8 ± 2.06	−56	0.06	34.1	30.2
Delay-13	17.8 ± 3.05	ns	0.09	13.8	12.7
Delay-9	23.2 ± 6.60	ns	0.11	9.12	8.70
Delay-19	27.5 ± 5.67	ns	0.13	18.9	18.8
Liver-Delay	38.2 ± 11.4	−71	0.18	2.13	2.08
Pancr-bile	67.1 ± 7.50	ns	0.32	37.7	37.3
RBC-1	142 ± 14.8	ns	0.68	1378.	342
RBC-delay	151 ± 14.4	ns	0.72	1387.	1908
RBC-2	234 ± 27.8	ns	1.12	0	4796
Tissue-2	266 ± 32.3	ns	1.28	53.9	56.9
Plasma6	375		1.80
Tissue-3	19,434 ± 1371	−27	93.3	5115	10,325
Total	20,830		100

¹Pools refer to the model (Fig. 4) and are compartments or delays.

²Diet mass is the mass calculated from a diet containing organic and inorganic Se. See text for details. Values are means ± SEM.

³To convert from g to μmol, multiply by 0.0127.

⁴Differences are calculated as [100 × (value for females − value for males)/(value for males)].

⁵ > 0.05; otherwise, P ≤ 0.05.

⁶Plasma pools are listed in Table 5. RBC consists of pools RBC-1, RBC-delay, and RBC-2 and has a mass of 527 g or 2.5% of whole body Se.

Pool masses calculated by the model (Fig. 4) for all participants as if they were consuming only inorganic (Sel-exchangeable Se) or only organic SeMet-Se (Supplemental Table 3) were used to calculate the actual or diet pool mass based on each participant’s relative dietary intake of Sel-exchangeable Se to SeMet-Se (22) (Table 4). The bulk of body Se was in Tissue-3 (Table 4). Within plasma, 5 of the pools contained ≤1% of Se in plasma, whereas 1 pool (Plasma-11) contained over one-half of the plasma Se (Table 5). This pool was labeled only by SeMet and was considered nonspecific labeling.

TABLE 5.

Calculated mass of Se in each plasma pool, gender differences in the mass, distribution of Se within plasma, and turnover time of each Se form¹

	Diet mass23	Females vs. males4	Distribution within plasma				Turnover time
Pool			All, n = 20	Males, n = 7	Females, n = 13	Females vs. males	Sel	SeMet
	μg	% Difference	%			% Difference	h
Plasma-4	0.5 ± 0.1	ns5	0.1	0.2	0.1	−41	2.1	1.2
Plasma-1	2.0 ± 0.4	−54	0.5	0.7	0.4	−42	0.5	0.6
Plasma-6	3.0 ± 0.7	ns	0.8	0.8	0.8	1	0.8	0.7
Plasma-7	3.6 ± 0.5	ns	1.0	0.9	1.0	18	1.7	2.5
Plasma-5	3.7 ± 1.0	ns	1.0	1.0	1.0	1	0.7	0.9
Plasma-3	8.0 ± 1.7	ns	2.1	2.0	2.2	11	1.2	1.4
Plasma-8	11.2 ± 1.5	ns	3.0	2.9	3.0	4	5.9	7.0
Plasma-2	12.5 ± 1.5	−48	3.5	4.0	3.0	−34	4.3	4.6
Plasma-9	44 ± 4	ns	11.8	10.7	12.5	16	37	86
Plasma-10	85 ± 10	ns	22.6	24.9	21.1	−15	803	674
Plasma-11	202 ± 13	ns	53.7	51.8	55.1	6	0	706
Total	375		100	100	100

¹Pools refer to the model (Fig. 4).

²Diet mass is the mass calculated from a diet containing both Sel-exchangeable Se and SeMet. See text for details. Values are means ± SEM.

³To convert g to μmol, multiply by 0.0127.

⁴Differences are calculated as [100 × (value for females − value for males)/(value for males)].

⁵ > 0.05; otherwise, P ≤ 0.05.

Pool turnover times ranged from 0.03 h (or 2 min) to 10,325 h (or 430 d) (Table 4). The turnover times were similar for Sel and SeMet with the notable exception of Tissue-3, which turned over faster for Sel, and the RBC compartments (Table 4). Most of the plasma pools turned over in <7 h, but 2 pools turned over in ~700 h (or 30 d) (Table 5).

Several pools contributed to the urine Se: over one-half from the largest pool in the body (Tissue-3), some from the pool that did not go through liver (Plasma-1), and also RBC (Table 6).

TABLE 6.

Source of urinary Se for all participants and by each gender¹

Pool	All, n = 20	Males, n = 7	Females, n = 13
	%
Tissue-3	54.0	50.9	56.2
Plasma-1	24.9	26.1	24.1
Plasma-11	8.0	10.9	6.0
Plasma-9	2.8	3.5	2.3
RBC-1	2.7	2.6	2.8
Plasma-2	1.9	1.7	2.0
Plasma-3	1.9	1.3	2.3
Plasma-10	1.3	1.2	1.3
Plasma-4	1.0	0.9	1.1
Plasma-8	0.6	0.5	0.7
Plasma-5	0.5	0.3	0.6
Plasma-6	0.3	0.3	0.2
Plasma-7	0.3	0.02	0.4
Total	100	100	100

¹Pools refer to the model (Fig. 4). The contribution was calculated as: 100 × [diet Se (i.e. sum of Sel-exchangeable Se and SeMet-Se) excreted from each pool per d/total Se excreted in urine per d].

The same trends were observed in both genders when comparing forms, except that males had higher recycling of SeMet to Liver-1 from plasma pool Plasma-2 (Table 3). The differences between genders for Sel and SeMet were not seen in the same pathways. The main differences in SeMet related to Plasma-2, which was excreted more in females, and Plasma-3 and Plasma-9, which were recycled less to liver and tissues in females than in males (Table 3). Total body Se was 27% less in females than males (Tables 2) due to Tissue 3 (Table 4). Only 2 plasma pools (Plasma-1 and Plasma-2) were smaller in females (Table 5). The distribution of Se among the plasma pools was similar between genders, although females had a smaller percentage of Se in the faster turning-over pools (Plasma-1, -2, and -4) (,Table 5). The main difference between genders in excretion was that in females a higher percentage of urine Se originated from 2 plasma pools (Plasma-5 and Plasma-7) and a smaller percentage from the nonspecifically labeled plasma pool (Plasma-11) (Table 6).

In summary, plasma curves were explained by 70% higher absorption of SeMet than of Sel and RBC curves by 2-fold higher uptake of SeMet than of Sel. Urine curves were explained by differences in absorption and excretion of the 2 forms. Tracer appearance in urine was dominated by Plasma-1, although the major source of natural Se was the large pool, Tissue-3. Fecal curves were explained by differences in absorption and, because SeMet was almost totally absorbed, even endogenously excreted tracer was reabsorbed with very little being excreted. The larger divergence between forms in males at ~40 h was explained by higher urinary excretion from one of the plasma pools, Plasma-9.

Discussion

The proposed model represents the kinetics of both Sel and SeMet and was developed by combining separate models for each form and adding plasma, RBC, and extravascular pools, as required, to fit data from studies extending over 4 mo. Strengths of the model are that it integrated the metabolism of organic and inorganic Se forms when studied in the same participants and it was based on urine and fecal collections for 12 d and frequent blood sampling for 120 d. Our approach used the simplest model (i.e. fewest pools and parameters) to fit the observed data. As a result, it likely combined transformations (i.e. a pathway could represent several reactions) and the pools, which are kinetically distinct species, will require additional research to metabolically and/or physiologically define. An alternative approach to the modeling, involving the representation of all known metabolic transformations, would have required an even larger model with many parameters having unknown values. A possible limitation of the study was the size of the tracer doses, although these were within physiological intakes. Thomson et al. (24) reported ~40 μg Se/brazil nut, and so our dose represented ∼4 nuts. Because the turnover of Tissue-3 was >213 d, a study longer than 120 d would be required to fully explore the nature of this pool. A 3rd limitation is the small number of males.

Our estimates for Se intake were similar to those reported for U.S. adults: 153 ± 78 μg/d (mean ± SD) for males and 109 ± 37 μg/d for females (25). Similarly, our values for absorption of Sel and SeMet (57 and 97%, respectively) matched previous reports by us and others (5, 6, 12, 13). Rayman et al. (26) recently summarized their progress in characterizing the major dietary Se species. At least 17 forms of Se have been described in plants, with SeMet the predominant form in grains and selenate the main form in leaves (27). Selenite is thought to be rare in foods; however, like selenate, it is reduced metabolically to selenide, an intermediate that is also produced by SeCys. Accordingly, these inorganic Se species can act like SeCys to supply Se for selenoprotein synthesis. Our studies of the simultaneous administration of SeMet and Sel tracers revealed that, apart from absorption, the metabolic fates of these Se moieties are quite similar, likely reflecting their common conversion to selenide and subsequent incorporation into selenoproteins as SeCys. Similar metabolism of these forms was also reported in rats (28).

The large number of pools is consistent with the fact that the human selenoproteome appears to contain at least 25 selenoproteins (7), which may be expected to have different turnover rates. Selenoproteins as well as other proteins with differing turnover rates can also contain nonspecifically incorporated SeMet. It is likely that the large number of pools required to fit the observed data reflects this level of complexity. Here, we speculate on the interpretation of the pools and pathways in the model from known physiology and metabolism (i.e. what does metabolism say about the model) and then discuss new insights into Se metabolism provided by the model.

To identify extravascular pools, we used data from Zachara et al. (29) on tissue weight and Se concentration to calculate Se mass in various tissues for a U.S. participant (Supplemental Table 4). The mean whole body mass of Se predicted by the model (20.8 mg) agrees with the value calculated from U.S. tissue concentrations (19.2 mg) (Supplemental Table 4). The largest pool in our model (Tissue-3) contained more body Se (93%) than the value (58%) estimated for muscle using actual tissue measurements (29), although an additional 16% of body Se was unaccounted for, using the latter method (Supplemental Table 4). The amount in muscle, however, is affected by Se intake; e.g. U.S. adults had 7-fold higher Se in muscle than did Polish adults whose Se intake was lower (Supplemental Table 4). Alternatively, this pool (Tissue-3) may represent a form of Se that is present in muscle but also in other tissues. For example, the reported mass of Se in liver (1.116 mg) was larger than the sum of compartments we considered to be in liver. Indeed, adding the mass of Se in the compartments designated as liver and the delays as well as the slowly-turning of pool, Tissue-2 resulted in less than one-half the amount reported in liver. This suggests that there is probably a nonspecifically-labeled pool of Se in liver. Therefore, the model’s largest pool may consist mainly of SeMet nonspecifically incorporated into proteins as a mimic of its sulfur-analogue methionine but may also contain some selenoproteins. All Sel in this large pool (Tissue-3) was excreted, whereas 16% of SeMet returned to plasma. The published tissue distribution data therefore provide limited insight into the location of the pools, suggesting that the pools may represent metabolic species rather than actual physical locations.

The model predicted that 50% of absorbed Se was taken up by liver during the first pass. This agrees with animal studies, where Kato et al. (30) measured a ratio of 2:1 in portal compared with aorta plasma following oral tracer administration in rats, indicating that liver uptake was 50%. Within liver in humans, most (76%) of the Se was associated with macromolecules (>2 kDa) and was distributed among 8 selenoproteins that ranged from 8.5 to 335 kDa (31). But, because several proteins had similar masses and could not be separated, the authors speculated that more proteins could exist. Based on the differences in subcellular distribution of the various proteins, they suggested that Se could be incorporated via different metabolic pathways. Our results, which imply multiple species originating in the liver, are consistent with these results.

From the liver, 71% of Sel and 56% of SeMet entered a pool (Tissue-1) that turned over rapidly (~5 min). This pool may represent selenide, a common intermediary for Sel and SeMet during conversion to SeCys, in which form Se is incorporated into selenoproteins (9) and would likely correspond to the central Se pool in Burk’s schema (32). About one-third of this rapidly turning over pool (Tissue-1) went to a slower pool (Tissue-2) with a turnover of 57 h (just over 2 d). Neither the location nor the composition of this slower pool is known, but we used it to distribute Se to RBC and to the largest pool in the body.

Numerous plasma peaks after tracer administration have been reported in the literature (33), and we summarized some reported values for the distribution of Se among plasma and serum proteins in humans (4, 34–39) for comparison with our calculated pool masses (Table 7). The percent distribution of Se in plasma is affected by the amount and form of Se in the diet (4). Gao et al. (36) identified 5 Se-containing protein bands in Chinese adults, whereas other investigators reported 3 major forms. We note that the albumin fraction is often calculated as total Se minus the amount in selenoprotein P (SeP) and glutathione peroxidase (GPx)3, which is likely an overestimate, and may contain more than one Se species.

TABLE 7.

Literature values for concentration and distribution of Se in human plasma or serum¹

						Distribution
	Intake Se	Concentration				68 kDa	57.5 kDa	47 kDa	41 kDa	21.5 kDa
Tissue		Se	Albumin	SeP	GPx	"Albumin"	"Se-P"	"Se-P isoform"	"GPX dimer"	"GPX"	Country (Ref)
	μg/d	μg/L	g/L	mg/L	units/L	μg/L (%)
Serum		66				28 (42)	20 (30)			18 (27)	Poland (34)
Plasma		77				12 (16)	47 (61)			18 (23)	Germany (35)
Serum		94				7 (8)	30 (32)	21 (23)	14 (15)	22 (23)	China (36)
Plasma	35	84	44		391	37 (44)	28 (33)			20 (23)	France (37)
Plasma		122	39	5.3	159	44 (36)	64 (52)			17 (14)	US (4)
Serum		126				35 (28)	67 (53)			24 (19)	Japan (38)
Serum		140				78 (56)	22 (16)			40 (29)	Poland (34)
Plasma	138	146									US (39)

¹To convert g to mol, multiply by 0.0127.

In the proposed model, plasma pools that turn over in ~4 h and together represent 3% of plasma Se (Plasma-2) match the lipoprotein fraction reported by others (40). The calculated Se content of apo B in LDL was reported to be more than 3 times that expected (1.7 vs. 0.5%) from random substitution of SeMet for methionine (40). Burk (41) directly showed that labeled selenite binds to various lipoprotein fractions for a prolonged period (>24 h) in humans. We defined the largest plasma pool in the model (Plasma-11) as nonspecifically incorporated Se, presumed to be albumin. SeP appears to be expressed in all tissues, but the greatest contributor of SeP to plasma is the liver (42). When Schweizer et al. (43) used liver-specific SeP knockout mice, there was a marked decrease in plasma SeP, confirming that liver is the main source of this protein. However, they found that the brain retained its Se level, apparently by synthesizing its own SeP, which served as a Se storage pool. Plasma GPx3, by contrast, is synthesized and released from the kidney (44). We speculate that, based on their masses, 2 plasma pools (Plasma-9 and Plasma-10) may be SeP and GPx3, respectively. However, in rats, SeP turnover was faster than GPx3 [T_1/2 of 3–4 h vs. 12 h, respectively (42, 45)]. If the same relationship occurs in humans, then Plasma-9 (turnover of 80 h) would likely be SeP and Plasma-10 (turnover of 1000 h) would be GPx3.

In humans, RBC Se consists mainly of SeMet nonspecifically incorporated into hemoglobin (46). In the present study, 44% of RBC Se was in RBC-2, considered a nonspecifically labeled pool. Consistent with other studies (5, 6, 17), we report that the percentage of Se from SeMet appearing in RBC is twice that of Se from Sel. The delay (~80 d) is close to the mean age (51 d, range 38–60 d) of RBC reported for 6 healthy adults (47).

Three chemical species comprise most of the Se in human urine: selenosugars, trimethylselenonium, and selenite (10). Trimethylselenonium has been reported as 2.2% of urine Se (48) in 1 study and 11% in another (49). We included pathways into urine from all plasma pools. About 25% of urinary Se originated from a pool (Plasma-1) that did not appear to pass through the liver and 50% originated from the largest body pool (Tissue-3, probably muscle). Burk et al. (50) proposed a competition between the synthesis of SeP and urinary metabolites in the mouse liver, because Se excretion increased when the SeP gene was absent. Setting 1 pathway of the model (Liver-2 to Tissue-1) to zero caused urine excretion of tracer to double (data not shown), suggesting that this pathway is involved in Se incorporation into SeP.

In summary, modeling revealed: 1) that there were 11 pools of Se in plasma with unique turnover times; 2) different turnover times for Sel compared with SeMet in the largest body pool (213 vs. 430 d, respectively); 3) over 50% of plasma (and 44% of RBC) Se was in pools labeled only by SeMet; 4) that a pool that contained 93% of body Se contributed to 50% of urine Se and a single plasma pool (Plasma-1) contributed 25%; and 5) that RBC Se came from a pool, Tissue-2, that had a turnover time of ~2 d.

There remain some anomalies with respect to Se kinetics from this modeling. How can Sel-Se and SeMet-Se have differential incorporation into RBC once they have mixed in a common pool? There may be a small fraction of SeMet incorporated before the mixing occurs, but our data were unable to define such a pathway. A more sensitive chemical technique may detect if this pathway exists. What do “liver/tissue” pools represent and how can Se be taken up into RBC from an apparent “liver” pool without going through plasma? Either the liver/tissue pools represent common or parallel pathways that occur in several tissues (including bone, the source of hemoglobin), or there is a compound that passes through plasma with such small pool size that it is not detectable from these data. Determining the compounds labeled in RBC would reveal this. What do delays in the model represent? They may represent different synthesis times required for various selenoproteins in liver, or tissues with different expressions for proteins (this would assume the tissues secrete the proteins into plasma). Determining the time to incorporate Se into different selenoproteins may reveal the meaning of these delay compartments.

In conclusion, a human model of whole body Se metabolism has been developed that integrates Sel and SeMet kinetics. Previous models were unable to fit the additional plasma peaks measured in this study, the long-term distribution of Se in the body pools, or the differences between forms that were measured in the same participants. The model could be used to explain differences in Se metabolism in participants ingesting different forms of Se. A follow-up study in the same participants examines how ingestion of a Se supplement affects metabolism of each form (BH Patterson, GF Combs, Jr.,WK Canfield, PR Taylor, KY Patterson, AD Hill, JE Moler, and ME Wastney, unpublished data). The challenge now is to put real names on the compartments by speciation of plasma and tissue proteins and also perform kinetic studies of the individual selenoproteins. As pools are identified biochemically, their metabolism can be determined in vivo by using knowledge from the model. The model, although rudimentary, is a step toward integrating cellular and tissue metabolism of Se in humans into a whole body framework.