Highlight Article: Retinol isotope dilution accurately predicts liver reserves in piglets but overestimates reserves in lactating sows

Short abstract

Retinol isotope dilution (RID) is used to estimate total body vitamin A (VA) stores in groups to assess VA status. Metabolic differences during lactation may affect RID calculations as currently applied. We evaluated the time required for isotopic equilibration between serum and liver retinol in piglets, and the utility of milk retinol isotopic enrichment as a proxy for serum in lactating sows. Piglets (n = 24) and sows (n = 6) were fed 1.75 or 20 µmol ¹³C₂-retinyl acetate, respectively. Piglets (n = 5 or 7) were killed on d 0, 4, 7, or 14. Blood and milk were collected at d 0, 0.5, 1, 2, 4, 7, 10, 14, and 21 before the sows were killed to collect liver. Retinol ¹³C-enrichment was determined by gas chromatography-combustion-isotope ratio mass spectrometry. Equilibration time and RID-predicted liver VA reserves were calculated. In piglets, serum and liver retinol ¹³C-enrichment differed significantly in individuals at d 4 and 7 (P = 0.008, 0.03) but not d 14 (P = 0.06); however, mean values were not different by d 4 (P = 0.62). Current RID equations accurately predicted VA deficiency (means ≤0.027 µmol/g liver) in the piglets. In sows, milk and serum retinol ¹³C-enrichment reached equilibrium between 2 and 7 d post-dose. After correcting for dose lost to milk, RID equations predicted higher liver stores than measured values even though the serum to liver atom % was 1.00 ± 0.01 at kill. In VA deficient infants, a shorter period may be accurate in population-level RID studies when using appropriate assumptions. In lactating women, the RID may have decreased accuracy due to variable losses of tracer in milk. Furthermore, assumptions about storage and loss of the dose in milk must be evaluated in lactating women considering the observed discrepancy between predicted and measured stores.

Impact statement

Vitamin A (VA) deficiency and hypervitaminosis A have been reported in groups of people worldwide. Conventional biomarkers of VA deficiency (e.g. serum retinol concentration, dose response tests) are not able to distinguish between sufficiency and hypervitaminosis A. Retinol isotope dilution (RID) predictions of VA status have been validated in humans and animal models from deficiency through toxicity; however, RID during life stages with unique issues related to isotopic tracing, such as infancy and lactation, requires further evaluation. This study investigated RID in piglets and lactating sows as models for human infants and women. In piglets, RID successfully determined VA deficiency (confirmed with liver analysis), and that the tracer mixes quickly. Conversely, in lactating sows, although serum and milk enrichments were similar, traditional RID equations overestimated VA stores, likely due to losses of tracer and higher extrahepatic VA storage than predictions. These data inform researchers about the challenges of using RID during lactation.

Introduction

Vitamin A (VA) deficiency is a major problem worldwide, especially in Southeast Asia and sub-Saharan Africa.¹ The World Health Organization (WHO) currently defines a country as having a severe risk of VA deficiency if the prevalence of low serum retinol concentrations (<0.7 μmol/L) in children less than 6 y is greater than 20%²; however, this biomarker suffers from low sensitivity when compared with retinol isotope dilution (RID) in children.^3,4 Liver biopsy is considered the gold standard for VA status,³ but collecting liver samples is difficult and unethical in most situations in humans. To avoid this, the RID test has been used to indirectly predict total body VA stores (TBS) and estimate total liver VA reserves (TLR) using appropriate assumptions.⁵

In the RID test, a dose of isotopically labeled VA (usually retinyl acetate labeled with deuterium or ¹³C) is consumed and allowed to mix with TBS before a blood sample is taken. The time period for equilibration is less for rats or humans with VA deficiency than VA adequacy,^6,7 and efforts have been made to sample at time points as early as 3 d.^8,9 Following the mixing period, the ratio of labeled retinol to endogenous retinol (tracer-to-tracee ratio; TTR) can be determined from baseline ¹³C-natural abundance, enrichment of the dose, and final ¹³C enrichment of retinol in the tissue as described by Cobelli.¹⁰ From this and the initial dose size of labeled compound, which is corrected for absorption efficiency and catabolism over time, the principle of mass balance can be used to determine the size of the turning-over pool of endogenous retinol.¹¹

RID-estimated TLR have been validated against liver samples using rats,⁶ rhesus monkeys,¹² and two human trials in which liver was accessible for other reasons, such as an unrelated surgery^7,13; however, these trials examined healthy, non-lactating adults. Lactation increases demand for VA in women because they supply VA to their infant through breast milk.¹⁴ Whether from breast milk or elsewhere, maintaining adequate VA status during infancy is especially important for satisfactory growth and immune function.³ Therefore, it should be considered a priority to verify the accuracy of VA status measurement in infants and women. During lactation, RID is affected by irreversible loss of the tracer to milk and likely altered storage and mobilization of the vitamin. The RID is also affected by differences in VA metabolism in infants and children with different VA status; for example, the half-life of VA was determined to be 32 d in Peruvian children with low VA stores¹⁵ compared with 136 d in Zambian children with high TBS¹⁶ and 135–140 d in healthy adults.¹⁷ Researchers have used the 32-d-estimated half-life in Ghanaian infants with adequate VA status¹⁸ and Thai children with VA deficiency.¹⁹

Lactating sows and weanling piglets were chosen for this study because of the anatomical and physiological similarity of their gastrointestinal tracts to humans, similar VA metabolism, and access to liver samples.²⁰ In addition, this model was previously used in VA studies.^20–23 In particular, piglets are a good model for human infants because of their similar body weights and low VA stores at birth.²⁴ In these experiments, we evaluated the ¹³C-RID test in weanling piglets as a model for infants, and in lactating sows as a model for nursing mothers, to gain insight into VA metabolism during these life stages.

Methods

Animals and diets

Approval for these studies was obtained from the University of Wisconsin-Madison Research Animal Resources Center. A total of 15 sows (crossbreeds of Large White and Landrace) were housed at the Swine Research and Teaching Center in Arlington, WI. Sows were fed standard feed until the time of their first or second pregnancy, and then maintained on a maize-based, preformed VA-free feed for multiple gestation-lactation cycles for the purpose of depleting their liver VA stores.²⁰ Weanling piglets from the second parity after the sows were switched to the preformed VA-free feed (n = 24, 2 from each of 12 sows, Table 1) were placed on a maize-based feed,²⁵ for the duration of the study, including a 10-d weaning period prior to dosing (Figure 1(a)). The type of yellow field maize available for feed mixing typically contains a small amount of the provitamin A carotenoid β-cryptoxanthin but was otherwise devoid of VA activity. After two parities consuming the maize-based VA-free feed, 6 sows (Table 1) were chosen for the lactation study and placed on a wheat-based VA-free feed²⁰ for their third or fourth gestation-lactation cycle (this study, Figure 1(b)). They consumed this feed until the end of the study when they were killed. All sows were given 2 kg feed/d during gestation and allowed ad libitum access during lactation.

Table 1.

Characteristics of weanling piglets and lactating sows dosed with ¹³C-retinyl acetate.^a

Characteristic	Weanling piglets (n = 24)	Lactating sows (n = 6)
Age at dosing	17 d	3 y
Prior parities of sows, n	2	3.3 ± 0.5
Body weight at time of death, kg	7.1 ± 2.6	271 ± 21
Liver weight, kg	0.22 ± 0.07	2.7 ± 0.2
Baseline serum retinol, µmol/L	0.83 ± 0.41	1.28 ± 0.38

^aValues are means ± SD.

Figure 1.

Timeline of work performed in (a) weanling piglets and (b) lactating sows.

Dose preparation, administration, and sample collection

[14, 15]-¹³C₂-retinyl acetate was synthesized as previously described.²⁶ After an overnight fast, piglets were orally dosed with 1.75 μmol ¹³C₂-retinyl acetate in corn oil. Blood samples (7 mL) were collected on d 1, 2, and 4 by cervical venipuncture. Piglets were killed on d 0 (n = 5, predose), d 4 (n = 5), d 7 (n = 7) and d 14 (n = 7) by electrical stunning and exsanguination, and blood and whole liver were collected (Figure 1(a)).

Sows were fasted overnight 1–2 d after farrowing and orally dosed with 20 μmol ¹³C₂-retinyl acetate dissolved in 4.7 mL corn oil by syringe. The sows were observed to ascertain that the entire dose was consumed. Sows were not anesthetized during dose administration. Blood (10 mL, as described above) and milk (40 mL) samples were collected at d 0 (predose), 0.5, 1, 2, 4, 7, 10, 14, and 21 (Figure 1(b)). Before milk collection, piglets were removed to a heated creep to maximize available milk. Sows were injected with 40 IU oxytoxin and fore- and hindmilk were collected from 1 or more teats to obtain the desired volume. On d 28, sows were killed and blood and whole liver were collected. All samples were stored on dry ice before transfer to a −80°C freezer.

Serum, milk, and liver retinol analysis

Serum, milk, and liver retinol were determined by HPLC as previously described.²⁵ Serum (1 to 1.5 mL) was treated with ethanol to precipitate proteins, and retinol was extracted with three hexanes, pooled, and evaporated under argon.²⁷ Milk (1 mL) was mixed with ethanol (1.5 mL) and saponified at 45°C for 2 h using 800 μL 50:50 potassium hydroxide:water (mass:vol). Retinol was extracted with 3 × 1 mL hexanes. Liver samples (∼1.5 g using at least three sampling sites) were weighed, mixed with 4–5 g NaSO₄, ground with a mortar and pestle, and extracted with 25 mL dichloromethane, from which 5 mL was dried, saponified, and extracted. The organic extract was evaporated under argon and resuspended in 100 μL 50:50 methanol:dichloroethane (vol:vol) and 10–100 μL (depending on concentration) was injected on the first of two HPLC systems as published.²⁷ Retinol concentration was determined at 325 nm using an external standard curve; C23-β-apo-carotenol²⁸ was used as the internal standard for all analyses.

Analysis of ¹³C-enrichment of retinol

Determination of ¹³C atom percent of retinol, calculated as ¹³C/(¹²C + ¹³C) X 100, was performed by gas chromatography-combustion-isotope ratio mass spectrometry (GC-C-IRMS) as previously described.^6,27 The purified retinol was dried by centrifugation in a ThermoSavant SpeedVac system (Thermo Scientific, Waltham, MA), resuspended in hexanes, and injected on the GC. As retinol elutes from the GC column, it is completely combusted to CO₂ and H₂O and the ratio of produced ¹³CO₂ to ¹²CO₂ was determined by a magnetic-sector mass spectrometer.

Determination of distribution of labeled and endogenous retinol

The TTR, TBS, and TLR were calculated as previously described¹¹

$$\mathit{TTR}=\frac{F_c-F_b}{F_a-F_c}$$ (1)

where F_a is the proportion of ¹³C in the isotopic tracer (experimentally determined to be 0.11), F_b and F_c are the GC-C-IRMS-determined ¹³C-enrichment of serum retinol at baseline and post-dose, respectively.

$$\mathit{TBS}=a\times \frac{1}{\mathit{TTR}}\times \left(\mathit{factors}\mathit{for}\mathit{absoption}\mathit{and}\mathit{storage}\right)$$ (2)

where a is the amount of isotopic dose given (in μmol). Absorption is typically assumed to be 75–100% in RID^16,29 based on studies in rats,³⁰ non-human primates,³¹ and humans;³² 75% was selected here due to the larger dose sizes used in comparison with other studies using ¹³C-labeled retinol and GC-C-IRMS analysis. Irreversible degradation of the remaining dose was assumed to occur according to a half-life of 140 d in adults for sows^13,17,33 or 32 d in children with low VA stores for piglets.^15,19 Specific activity between serum and liver was assumed to be 1 by d 14 because they were receiving VA-deficient feed¹¹ (an assumption that was examined in the piglets).

$$\mathit{TLR}=\frac{\mathit{TBS}}{\mathit{Liver}\mathit{mass}\left(g\right)}\times \mathit{fraction}\mathit{of}\mathit{TBS}\mathit{in}\mathit{liver}$$ (3)

where liver mass was measured directly. Fifty percent of VA stores were assumed to be in the liver due to the low TLR expected in these swine.¹¹

Further correction for dose lost to milk in lactating sows by day x was performed by the following formula using a trapezoidal approximation of daily tracer in milk

$$\begin{array}{l}\text{\% }\mathit{of}\mathit{dose}\mathit{lost}\mathit{to}\mathit{milk}=\mathit{milk}\mathit{production}\times \sum _{t=0.5}^x\\ (d\times \frac{(\frac{\mathit{TT}{R}_{t-d}}{1+\mathit{TT}{R}_{t-d}}\times \mathit{milk}\mathit{total}{R}_{t-d})+(\frac{\mathit{TT}{R}_t}{1+\mathit{TT}{R}_t}\times \mathit{milk}\mathit{total}{R}_t)}{2})\\ \times 100\text{\%}\end{array}$$ (4)

where t is the timepoint in days, and d is the number of days since the preceding timepoint. TTR_t is the GC-C-IRMS-determined tracer-to-tracee ratio in milk at time t, therefore “TTR_t/(1+TTR_t)” yields the proportion of retinol that is tracer at time t, milk total R_t is the concentration of retinol (tracer + tracee) in milk at time t determined by HPLC (in μmol/L), the initial dose was 20 μmol and milk production was assumed to be 10 L/d.²⁰

Dose lost to milk reduces retention in addition to the non-lactating human-adult dose degradation assumed by a half-life of 140 d; therefore, the retention factor in TBS (2) at any given timepoint x that is d days since the last timepoint is estimated by

$$\begin{array}{l}R\mathit{etentio}{n}_x={\displaystyle \sum }_{t=0.5}^x\\ (\mathit{Retentio}{n}_{t-d}\times e^{-\frac{d\times \ln \left(2\right)}{140}}-\frac{\left(\text{\%}\mathit{dose}\mathit{lost}\mathit{to}\mathit{mil}{k}_t-\text{\%}\mathit{dose}\mathit{lost}\mathit{to}\mathit{mil}{k}_{t-d}\right)}{100\text{\%}})\end{array}$$ (5)

where retention at d 0 is 1.

Statistical analysis

Statistical analyses were performed in Microsoft Excel (Microsoft, Redmond, WA). Group values are mean ± SD. Comparisons between serum and liver retinol ¹³C-enrichment in piglets were performed using Student’s t-test, paired and unpaired as noted, while liver and serum retinol concentrations were compared using least-squares linear regression. Individual sow enrichment data and predicted/measured TLR values are presented and compared by paired Student’s t-test, and group means are used when appropriate. Significance was defined as P ≤ 0.05.

Results

Isotopic equilibration of labeled retinol, and predicted and actual liver stores in piglets

The ratio of ¹³C-enrichment in serum and liver in the weanling piglets was measured at d 4, 7, and 14 after dosing with 1.75 μmol ¹³C₂-retinyl acetate using GC-C-IRMS (Table 2, Supplemental Table 1). When comparing paired liver and serum samples from each piglet, the difference in ¹³C-enrichment between the two VA pools was significant at d 4 (P = 0.008) and 7 (P = 0.03), and a trend occurred at d 14 (P = 0.062). Comparing mean enrichment of the two pools at each time point (Figure 2(a)), similar to determining the mean VA status in a large group, showed no difference at any time point after dosing. Serum retinol concentrations did not correlate with liver retinol concentration in these piglets (Figure 2(b)).

Table 2.

¹³C-enrichment in matched serum and liver retinol and total liver reserves (TLR) of vitamin A in piglets determined by retinol isotope dilution (RID) or liver sample, by day since dosing with ¹³C₂-retinyl acetate.^a

Time since dose (d)	n	Serum retinol 13C-enrichment (At%)	Liver retinol 13C-enrichment (At%)	P b	Serum-to-liver ratio of retinol 13C-enrichment	Predicted TLR (µmol/g)	Measured TLR (µmol/g)
4	5	2.43 ± 0.65	2.22 ± 0.59	0.008	1.09 ± 0.04	0.016 ± 0.010	0.020 ± 0.011c
7	7	1.97 ± 0.16	2.12 ± 0.26	0.03	0.93 ± 0.05	0.026 ± 0.029	0.027 ± 0.008c
14	7	1.64 ± 0.28	1.81 ± 0.36	0.06	0.91 ± 0.08	0.021 ± 0.021	0.013 ± 0.005c

At%: atom percent; TLR: total liver reserves of vitamin A.

^aValues are means ± SD. ¹³C enrichment values were determined by gas chromatography-combustion-isotope ratio mass spectrometry as previously described.⁶ Predicted total liver reserves were calculated using serum retinol ¹³C enrichment by retinol isotope dilution calculations (see text). Measured total liver reserves were determined by high pressure liquid chromatography.

^bPaired-sample Student’s t-test.

^cPredicted and measured TLR values were not significantly different, determined by paired-sample Student’s t-test.

Figure 2.

(a) ¹³C-enrichment in weanling piglet serum retinol and liver (mean ± SD) expressed as At% by days since dose, determined by GC-C-IRMS. For the four timepoints with both liver and serum data (d 0, liver n = 5, serum n = 19; d 4, liver n = 5, serum n = 19; d 7, liver n = 7, serum n = 10; d 14, n = 7), mean values are not significantly different by two-sample student’s t-test. Liver samples were all collected from different piglets at death, while serum was taken at death and from blood draws in remaining piglets. At%, atom percent of ¹³C in total carbon. (b) Liver and serum retinol concentrations of piglets. Retinol concentrations were determined by HPLC. Liver total retinol was determined by summing retinol and retinyl esters. The slope of the line is not significantly different from zero using Least Squares Regression.

Although the main focus of the piglet study was isotopic equilibration, TLR were estimated by RID at each time point and compared to HPLC-determined liver VA concentration (Table 2). Calculated values were not different from measured liver VA reserves at any time point (paired Student’s t-test), and RID accurately determined the piglets to be VA deficient [<0.1 μmol VA/g liver³] in all cases.

Isotopic equilibration of labeled retinol, and predicted and actual liver stores in sows

¹³C-enrichment of serum and milk retinol from the sows was determined at each time point (d 0, 0.5, 1, 2, 4, 7, 10, 14, 21, and 28) and the ratio of the two was plotted over time (Figure 3(a)). Retinol concentrations in sow milk were initially high (1.76 ± 0.75 μmol/L) but decreased after d 2 (1.00 ± 0.24 μmol/L) (Figure 3(b)). Because the quantity of ¹³C-labeled dose present in milk was determined to be high before equilibrating with whole body stores, and milk is produced at a rate of approximately 10 L/d in lactating sows, the cumulative proportion of the dose lost to milk over time was determined and plotted (Figure 3(c)). Final serum and liver ¹³C-enrichments were compared after kill at d 28, showing that complete equilibrium had occurred with a final ratio of ¹³C-enrichment of serum to that in liver of 1.00 ± 0.01 (Table 3).

Figure 3.

(a) Ratio of ¹³C-enrichment in milk and serum retinol, determined by gas chromatography-combustion-isotope ratio mass spectrometry (GC-C-IRMS), of lactating sows (n = 6) after a 20 μmol dose of ¹³C₂-retinyl acetate. At%, atom percent of ¹³C in total carbon. (b) Milk concentration at each sampling timepoint following dosing. Error bars represent SD. (c) Estimated cumulative percent of the ¹³C₂-retinol dose lost through the milk of lactating sows, determined by GC-C-IRMS, which was calculated as the area under the curve, determined by trapezoidal approximation, of the product of the retinol concentration and tracer-tracee ratio of ¹³C₂ enrichment of milk retinol over time, multiplied by average milk production of a sow [10 L/d]²⁰.

Table 3.

Predicted and measured total liver vitamin A reserves (TLR), and ¹³C enrichment of lactating sow serum and liver retinol after 21 and 28 days of isotopic equilibration of a 20 μmol ¹³C₂-retinyl acetate dose.

	21-d timepoint		28-d timepoint (measured at death)
Sow	Serum-predicted TLR (µmol/g)	Milk-predicted TLR (µmol/g)	Measured TLRa (µmol/g)	Serum At%b	Liver At%b		Ratio
1	0.50	0.41	0.14	1.110	1.103		0.99
2c	0.67	0.52	–	1.114	–		–
3	0.49	0.46	0.34	1.112	1.111		1.00
4	3.81	2.99	0.18	1.087	1.083		1.00
5	0.25	0.23	0.10	1.124	1.116		0.99
6	1.89	0.83	0.37	1.093	1.088		1.00
Meand	1.27 ± 1.37e	0.91 ± 1.04e	0.23 ± 0.12	1.105 ± 0.015f	1.100 ± 0.014f		1.00 ± 0.01

At%: atom percent; TLR: total liver reserves of vitamin A.

^aDetermined by high pressure liquid chromatography.

^bDetermined by gas chromatography-combustion-isotope ratio mass spectrometry.

^cThe liver from sow 2 was missing at the point of supply and could not be analyzed.

^dValues are mean ± SD.

^eValues were not significantly different by paired sample t-test.

^fValues were not significantly different by paired sample t-test. Neither mean includes sow 2.

Total liver VA stores were determined by RID with a 21-d equilibration period using either milk or serum values as the accessible pool and compared with actual liver values (Table 3). No difference was noted in the calculated TLR using milk or serum retinol ¹³C enrichment. In this case, the retention of dose used in the RID calculations incorporated loss of dose to milk and irreversible degradation from one timepoint to the next. Sows 4 and 6 had particularly large discrepancies between calculated and measured TLR (5–20 times more predicted liver VA than measured and likely reflects loss of tracer during dose administration).

Discussion

In this study, we examined the equilibration of a tracer dose of ¹³C₂-retinyl acetate in weanling pigs as a model for infants. Similar to adult rats,⁶ these piglets had a significantly different isotopic enrichment in liver and serum VA before 14 d as the tracer enters serum from absorption and is then taken up by storage organs. In this case, however, there was a crossover point at approximately 4 d (Figure 2(a)), after which the serum retinol ¹³C-enrichment was maintained below the liver value, although it is no longer significantly different at d 14. This pseudo-equilibrium, in which the dose is preferentially retained in the liver due to new, incoming unlabeled VA in the diet and/or sequestration of a large dose, has been seen in earlier RID studies, leading to the inclusion of a specific activity factor of 0.65–0.80 in RID equations when these factors are present.^7,11,34 In the absence of either of those two factors, it has been predicted^11,35 and shown in rats and monkeys^6,31 that complete equilibrium (specific activity of 1) can be achieved. In this study, feed was used that contained a small amount of β-cryptoxanthin but was otherwise VA-deficient, which resulted in deficiency (liver VA stores ≤0.043 μmol/g) in these piglets. Furthermore, the dose used was relatively larger than other studies in children (1.75 versus 1.0 μmol currently being used in children), potentially leading to its preferential storage in the liver, and equilibrium specific activities below 1 at 0.93 and 0.91 on d 7 and 14, respectively. However, this discrepancy did not affect the calculated versus actual measured TLR values.

Differences in liver and serum retinol ¹³C-enrichment values indicate that a specific activity factor is still needed in studies with large doses, but the inaccessibility of liver in infants and children means that determination of this factor is unlikely to progress in humans due to ethical issues. The natural conclusion is therefore to use small doses that are not sequestered in the liver whenever possible; however, these smaller doses can push the limits of detection of most current mass spectrometers used with RID, e.g., GCMS and liquid chromatography-MS/MS.³⁶ Data are needed comparing dose equilibration and sequestration using different doses of ¹³C-retinol before undertaking large-scale studies. Regardless, assuming a specific activity factor of 1, the mean RID-predicted TLRs in the piglets did not differ from measured TLR at any time point. Given that serum and liver retinol ¹³C-enrichment crossover had occurred by d 4 and were not different when using the unpaired mean ¹³C-enrichment in each pool, it follows that mean TBS and therefore TLR would be accurate. This finding gives support to the use of the RID in population-level studies with many subjects, where the impact of random error and the inter-individual variation in VA kinetics, would be minimized.^37,38

Adequate VA status of lactating women is vital to providing infants with VA-adequate breast milk. Two aspects of lactation physiology are important to the RID. First, if isotopic equilibration between serum and milk retinol occurs in a similar timeframe as that between liver and serum retinol, milk could be used as an accessible pool to sample ¹³C-enrichment of retinol before and after the dose, increasing the acceptability of the test, because blood draws can be a significant deterrent to participation in studies.³⁹ In this study, serum and milk achieved this equilibrium between d 10 and d 21, yielding 21-d milk- and serum-derived TLR estimates that did not differ, and was likely maintained because a VA-deficient feed was consumed.¹¹

Second, we discovered during analysis that the studied sows lost significant amounts of the labeled retinol dose to milk, which impacts the mass balance calculations. Previous studies have shown that chylomicron delivery of retinol to milk could account for loss of 25% of a newly ingested dose within 3 d²³ in addition to the redistribution of body stores to milk as needed from retinol bound to retinol-binding protein. This effect is compounded by the high concentration of retinol in milk (>1 μmol/L) early in the lactation cycle, which did not decrease until 5–6 d after farrowing (2 d after dosing), an effect consistent with previous observations.⁴⁰ Using α-retinol, a positional isomer of retinol that cannot bind retinol-binding protein and therefore represents only lipoprotein delivery, it was demonstrated that dietary retinol delivery to milk peaks around 8 h,²³ which was before the first post-dose milk retinol ¹³C-enrichment measured in this study (0.5 d). This could result in significant quantities of dose lost before correction was possible. While we estimated the fraction of dose lost at any given point that data were collected, and corrected the TBS calculations, these values varied substantially between individual sows. Monitoring breast-milk tracer concentrations more frequently in the initial period after isotope administration could result in more accurate loss calculations. In a population survey setting, it may be inaccurate to correct by a single factor for loss of dose to breast milk. On the other hand, milk production of sows [7.5–13.6 L/d²⁰] is much higher than that for humans [0.5–0.9 L/d⁴¹] due to the size of the litter. The loss of dose, therefore, should be quantified in human studies before conclusions can be drawn about the RID in lactation, given that losing 7–37% of the dose results in changes in predicted TBS by the same percent. In this regard, non-human primates that usually have singletons would be a more applicable model to use for lactation studies than multiparous mammals, such as swine. Furthermore, waiting a short period of time for milk retinol to decrease following initiation of lactation, could reduce the quantity of dose lost in the initial period following dosing.

Despite correcting for this loss, a large discrepancy remained between calculated and measured stores. In all cases, the differences are attributable either to uncaptured dose loss (spit out/not swallowed, not absorbed, not seen due to time between milk samples) or an inaccurate assumption of the proportion of VA stores in the liver to be 50%. For individual sows with especially large discrepancies (4 and 6), the first cause seems more likely, while the difference between predicted and measured liver stores across all five animals with data points to VA stores that were traced but incorrectly predicted to be in the liver. As noted, the sows weighed 271 ± 21 kg and might have ∼25% body fat,⁴² indicating significant total adipose mass. Future studies in swine, should consider collecting adipose tissue to assist in data interpretation. In rats for example, despite an approximately 50-times lower concentration of retinol in adipose than liver (20–24 nmol/g tissue), adipose has been predicted to contain a proportion as high as one-fifth the total VA found in the liver due to relative mass of the tissues.⁴³ In obese humans, 50% of TBS could be in adipose.¹¹ In these sows, if we assume similar adipose retinol concentrations as determined in the rats (22 nmol VA/g tissue) given that liver concentrations are comparable, and assume Sow 1 to have 25% body fat of 303 kg total weight, we would expect approximately 1667 μmol of VA, with liver accounting for another 440 μmol if retinol concentrations are uniform throughout the liver, for a total of 2107 μmol of that sow’s RID-predicted 3700 μmol (i.e., 57% explained by liver and adipose). It is not, therefore, implausible that a significant store of retinol could be found in the adipose tissue; however, this still leaves 43% of the predicted retinol unaccounted for, indicating further research is needed in lactating women.

Furthermore, during VA deficiency, kidney may contain a similar retinol concentration to liver.⁴⁴ We did not anticipate that the sows would have this degree of VA deficiency and did not collect the kidneys. Future studies could investigate carcass and other organ concentrations of retinol, as well as unaccounted-for dose loss (e.g. more frequent milk collections for milk loss, feces analysis for unabsorbed dose or intramuscular injection as a ∼100% absorption control) to determine whether these stores exist or are artifacts of the RID method in lactating sows.

Newborn infants are typically born with low VA stores,⁴⁵ and VA comes from colostrum and breast milk in the early stages of life. Due to its function in growth and the immune system, VA deficiency can be devastating in countries with poor maternal VA status. Thus, the data presented above supports the utility of the RID in large-scale infant and child surveys, adjusting for their shorter VA half-life^15,19 and low initial stores. The RID has been applied in Ghanaian infants, as a method of comparing liver stores between a group receiving high-dose VA supplements and controls.¹⁸ Our study provides support for the accuracy of the data showing that the infants in both groups were VA sufficient, which was corroborated by the baseline measure of VA sufficiency in both groups by the modified relative-dose-response test, another biomarker of VA liver status.³ A similar equation was used in that study to these piglets with an absorption factor of 80% for a 1.0 μmol dose, 0.8 serum to liver enrichment because diet was not controlled, and 80% storage in the liver because of adequate status.¹⁸

The RID in lactating women, however, needs to be further investigated in humans to correct for dose losses to milk as well as potential extrahepatic storage. It is important to note that extrahepatic storage that is turning over with the plasma pool (i.e., being traced), likely represents usable VA, and so TBS may prove to be a more relevant metric than TLR. If the estimations with RID in lactation prove to be more qualitative than quantitative, dose response tests may be more appropriate than RID during lactation. Studies are currently underway in Zambian and Thai lactating women to evaluate the utility of the RID test during this life stage.

Authors’ contributions

JS interpreted data, reanalyzed a subset of samples, and wrote the first draft of the manuscript. RLS conducted research with the swine. SAT designed the studies, analyzed the piglet samples, revised the paper, and had primary responsibility for final content. All authors read and approved the final manuscript.

DECLARATION OF CONFLICTING INTERESTS

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

FUNDING

This research was supported by NIHNIDDK 61973 and the National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, grant number USDA NRI 2007–35200-17729.