The available biomarkers of Se status have been developed for use in individuals suboptimally nourished with respect to the element. Indeed, they have been found generally useful in such cases, as differences in their values tend to correlate with differences in Se intake. More importantly, two of those biomarkers have direct functional significance: GPX3 participates in antioxidant function; while SEPP1 is involved in Se transport and also has antioxidant properties. However, these biomarkers are not informative of Se status under conditions of Se intake exceeding the level necessary to support their maximal expression, which coincides with the level apparently effective in reducing cancer risk.
This study demonstrates the problem of characterizing the Se status of a cohort that is not deficient in the element. It shows the discordance of biomarkers of Se status in a cohort of healthy Americans and the lack of utility of biomarkers with functional significance to assess Se status in a Se-adequate population. The most commonly used biomarker of Se status, plasma Se content, is presumed to reflect both the amounts of Se in various body pools and the level of Se intake . At 142.0 ± 23.5 ng/ml, it showed that this cohort was of relatively high Se status. Half of subjects had plasma Se levels comparable to those of the upper quintile of the NHANES 2003-2004 cohort , and 6% had plasma Se levels in the range achieved only by Se-supplementation in the NPC trial . No subject had a plasma Se level <70 ng/ml, above which Nève  noted subjects do not show GPX3 responses to Se-supplementation. Accordingly, plasma Se levels, all of which were above that threshold, were not associated with differences in GPX3 activity or SEPP1 level. This indicates maximal expression of those selenoproteins, rendering their measurement uninformative as biomarkers of Se status in this cohort. Differences in SeMet intake affect only the non-specific component of plasma Se, is also likely applicable to other such populations of relatively high Se status
Hill et al  estimated that GPX3 and SEPP1, when maximally expressed, should account for about 80 ng/ml of the Se in human plasma. Their estimate was based on the amount of Se associated with SEPP1 in one apparently Se-adequate individual, and the amount of Se reported by Avissar et al  to be associated with GPX3. The GPX3 activities and SEPP1 levels of our cohort indicate a slightly lower value, approximately 73 ng/ml, which may reflect weaknesses in our assumptions, particularly those concerning SEPP1. From the reports of Steinbrenner et al  and Méplan et al , we assumed the presence of multiple SEPP1 variants with an average molecular weight of 60 kD. We also assumed that SEPP1 contained an average of 9.9 g-atoms Se as SeCys per mole based on the findings of Saito et al  with human SEPP1; although the genetic coding of the protein would suggest the possibility of as many as 10 SeCys residues per mole . Our estimate suggests that the amount of Se present as SeCys in GPX3 and SEPP1 comprised about 54% of the total amount of Se in the plasma of these non-deficient subjects, a lower percentage than observed (80%) by Deagen et al  for Se-deficient men in China. The differences between these different cohorts may reflect the preferential response of the non-specific plasma pool to SeMet. This phenomenon is indicated by nearly half (47%) of plasma Se occurring in the non-specific fraction in the present cohort. Indeed, the variation we observed in plasma Se level was almost exclusively limited to variation in the non-specific component. That this non-specific fraction was not confounded by such variables such as age, gender, metabolic body weight or methylation status (serum folate, vitamin B12, Hcy) suggests that it may be a useful biomarker of Se status in non-deficient populations; since differences in SeMet intake affect only the non-specific component of plasma Se, these findings are also likely applicable to other such populations of relatively high Se status.
We are not aware of buccal cells, which offer the advantage of sampling a metabolically active tissue, having been used previously to assess somatic cell Se. We found buccal cells to contain appreciable amounts of Se; however, their Se showed no significant correlation with plasma Se level. The levels of Se in buccal cells and urine were each positively associated with the use of nutritional supplements. In this regard, these biomarkers may be more useful than plasma Se, which did not show such an effect.
Urinary Se, which consists mostly of methylated selenosugars [[40–42]], was the only Se biomarker not in dynamic equilibrium with other pools of Se in the body, although it would be expected to show first-order relationships with such pools. That urinary Se was also positively associated with serum folate and vitamin B12, but negatively associated with serum hCys, is consistent with its metabolic production being dependent on the availability of methyl groups. This is supported by the findings of Gonzalez et al.  that serum Se level was positively associated with serum folate level and negatively associated with serum hCys level, the latter explaining nearly 6% of the variance of serum Se. The lack of a relationship between plasma Se and urinary Se suggests that these two pools are not in a first-order relationship. Because the non-specific Se was the only variable component of plasma Se, it's apparent that very little of that protein-bound pool turns over into urine. The dimorphic Se excretion between men and women adds to the list of sex-specific differences in (hepatic) Se metabolism, which has been observed in both rodents and humans .
The relationship of serum SEPP1 and BMI (r2 = 0.054; p < 0.001), with greatest values among subjects with BMIs of 25-30 (figure 4), was unexpected as Kimmons et al  noted low plasma Se levels (<100 ng/ml) somewhat more frequently among women in this BMI class in the NHANES III cohort. That this may reflect the dysregulation of gluconeogenesis in obesity is suggested by studies in cultured cells that have shown SEPP1 to be regulated as a gluconeogenic enzyme [[46–48]].
It is also clear that genetic variability contributes to variance of Se biomarkers. Of the four allelic selenoprotein variants studied three, GPX1, SEPP1 and SEP15, were significantly related to the values of a Se biomarker (Table 4). That individuals with GPX1 679T/T alleles showed significantly lower plasma Se levels than those of the C/C alleles is of particular interest, as the former genotype has been associated with increased risk to cancers of the lung  and breast . That some SEPP1 genotypes have less (11%) SEPP1 expression than others, and that low BMI individuals had significantly lower SEPP1 than those with mid-level and high BMI, it is possible that SEPP1 in some individuals in this cohort may not be maximally expressed, even at these apparently adequate levels of Se intake.
The estimated intakes of macronutrients were similar to those reported for other cohorts of Americans . However, estimated Se intake did not significantly correlate with any Se-biomarker. This is not surprising, considering the inherent errors in determining nutrient intakes from an FFQ and in estimating the amount of Se in particular foods, which can vary considerably depending on the location and/or means of food production . The results of the FFQ method suggested that a core group of 22 foods provided 80% of the Se consumed by this cohort, a number similar to that estimated by Schubert et al  for the American population. This core included pork, beef and wheat products, the Se contents of which Finley et al [52, 53] showed can vary enormously (by 3-, 11- and 57-fold, respectively, for items purchased in the upper Midwest). Such great uncertainty severely compromises the value of Se intake estimated in this way as a useful indicator of Se status.