Calcium is the most abundant mineral in the human body, with 99% of adult body calcium located in bone and the remaining 1% located in soft tissues and extracellular fluid. The full-term newborn has about 30 g of body calcium, whereas a 24-week preterm infant has only 10% to 15% of this value at 3.0 to 4.5 g body calcium (1). From 28 to 40 weeks of gestation, fetal calcium content quadruples due to increased bone mineralization (2). Approximately, 85% of body phosphorus is located in bone and 15% in soft tissues and extracellular fluid (3). The full-term newborn body contains about 16 g of phosphorus (3).
Approximately, 80% of calcium and phosphorus accretion occurs during the third trimester, between 24 and 40 weeks of gestation (3), and infants born preterm miss this period of calcium and phosphorus accretion. Failure of the preterm infant to meet the mineral requirements may lead to metabolic bone disease, also called rickets or osteopenia of prematurity (3). To optimize bone mineralization and prevent metabolic bone disease, the American Academy of Pediatrics recommendations for preterm infant formulas are 140 to 160 mg of calcium per 100 kcal and 95 to 108 mg of phosphate per 100 kcal (4). The European Society for Pediatric Gastroenterology, Hepatology, and Nutrition recommendations for preterm infant formulas are 70 to 140 mg calcium per 100 kcal and 50 to 87 mg phosphate per 100 kcal (4). These recommendations are meant to serve as guidelines to ensure adequate calcium and phosphorus intake for preterm infants; however, they do not take into account the variable absorption of calcium from different sources, which depend upon many factors other than the amount of calcium provided. For example, these recommendations do not take into account the type of calcium salt, the amount or type of fat in the diet, the type of milk, and the processes used for infant formula manufacturing (5).
It is important to understand the bioavailability of calcium and phosphate in donor human milk (DHM) with nutritional additives because premature infants are often given this type of feeding in the neonatal intensive care unit (NICU) and are at risk for metabolic bone disease. Ideally, bioavailability should be studied in humans; however, these studies are time-consuming, expensive, and may not be practical to perform in premature infants (6). As alternative to measuring bioavailability, in vitro methods have been used consisting of a gastrointestinal digestion followed by measurement of mineral dialyzability through a membrane (6–12). Measuring mineral dialyzability estimates the fraction of the mineral soluble in the gastrointestinal tract and available for in vivo absorption (12). Although dialyzability estimates calcium and phosphate that is soluble in gastrointestinal fluids, sometimes referred to as bioaccessibility, it may provide a good estimate of the potential bioavailability of the minerals to preterm infants who depend primarily on passive calcium absorption through the leaky tight junctions of the intestinal epithelium.
We developed a simulated premature infant digestion system to measure the dialyzability of calcium and phosphate in DHM with and without common nutritional additives. We hypothesized that adding these minerals, HMF, and the 2 powdered infant formulas to DHM would decrease the dialyzability of calcium and phosphate when prepared according to present hospital protocols.
Preparation and In Vitro Digestion of DHM Samples
Preterm DHM was shipped from the San José Mother's Milk Bank (San José, CA) to our laboratory. Mothers who donated the milk gave signed consent that their milk may be used for research purposes, as is policy for all donations to milk banks of the Human Milk Banking Association of North America. The institutional review board at North Carolina State University approved the study. Five batches of preterm DHM were used in the study and treated according to standard fortification protocols used in NICUs (13). It is standard protocol for Human Milk Banking Association of North America milk banks to pool milk from 4 to 6 mothers to reduce variability of the nutritional composition. Therefore, milk from 20 to 30 mothers was included in the analyses. Each sample was studied 7 ways: no additives (control); addition of 0.15 mL calcium glubionate (Rugby, Duluth, GA) per 1 mL milk; 0.23 mL sodium potassium phosphate (sodium phosphate and potassium phosphate) per 1 mL milk; 0.15 mL calcium glubionate and 0.23 mL sodium potassium phosphate per 1 mL milk; 0.064 g Enfamil Enfacare (Mead Johnson, Glenview, IL) per 1 mL milk; 0.1563 g Similac Human Milk Fortifier (Abbott Nutrition, Columbus, OH) per 1 mL milk; and 0.072 g Similac NeoSure (Abbott Nutrition) per 1 mL milk. The amounts of calcium and phosphate content added to each treatment are listed in Table 1. The quantities of the infant formula added were based upon recommendations in NEOFAX(13), a nutritional guide used commonly in NICUs. The guidelines for different nutrition fortifiers supply different amounts of calcium and phosphorus to premature infants. Quantities of calcium glubionate and sodium potassium phosphate were based on protocols used at Lucile Packard Children's Hospital at Stanford University.
In Vitro Digestion Protocol
An in vitro digestion model was developed to simulate the gastrointestinal tract of the premature infant. The model was modified from those described previously (14,15). In the gastric phase, 0.2 g pepsin (Sigma, St Louis, MO) was dissolved in 5 mL of 0.1 N HCl and 0.25 mL was added to each 4-mL sample of DHM. Additionally, 1.7 g lipase with similar specificity as human milk lipase (16) (Sigma) was dissolved into 15 mL 0.1 N HCl and 1.5 mL was added to each DHM sample. Lipase was added in the gastric phase because it has been shown that there is a high degree of gastric lipolysis in premature infants (17). The low pH optimum (2.5–6.5), the absence of requirements for cofactors or bile salts, and resistance to pepsin digestion enable lipase to remain active in the infant's stomach and contribute significantly to fat digestion (16,17). The pH of each nutritional additive was adjusted to 5 before addition to the milk. All of the DHM samples were adjusted to pH 5.0 by addition of HCl or NaOH and then placed in a shaking water bath at 37°C for 2 hours. They then were placed on ice for 10 minutes to stop digestion.
In the intestinal phase, 0.05 g pancreatin (Sigma) and 0.3 g bile extract (Sigma) were dissolved in 25 mL of 0.1 mol/L NaHCO3 and 1.25 mL of this solution was added to each DHM sample. To add 17.2 mU of lactase (Sigma) to each DHM sample, 0.25 g of lactase was dissolved in 200 mL H2O, and 2 μL was added to each DHM sample. DHM samples were adjusted to a pH of 7.0 by 1 mol/L NaHCO3 and to a final volume of 10 mL by addition of cell culture grade water (Sigma). The DHM samples were placed in a shaking water bath at 37°C for 2 hours. The DHM samples were placed on ice for 10 minutes to stop digestion and they were adjusted to pH 7.0.
After the gastric and intestinal phases, the samples were centrifuged at 3500g for 1 hour at 4°C. Aliquots of the supernatant were transferred to tubes and stored at −20°C until further analysis, unless samples were analyzed within 24 hours, in which case they were stored at 4°C.
Dialysis was completed using Spectra/Por Float-A-Lyzer G2 (Model G235067, Spectrum Labs, Rancho Dominguez, CA) dialysis tubing with a molecular weight cutoff of 8000 to 10,000 Da. The Spectra/Por Float-A-Lyzer was submerged and allowed to soak in deionized water for 15 to 30 minutes. The hydrated membrane was not allowed to dry out.
Using a pipette, 10 mL of the previously digested sample was added to the inside of the membrane. The cap was replaced and the membrane was placed inside a glass tube that contained 25 mL of either a solution of 0.9% NaCl with 1% albumin or 0.9% NaCl, pH 7. In this experiment, all of the samples were dialyzed with and without albumin in the buffer to determine the effects of having a protein with moderate calcium binding in the dialyzate on the final distribution of calcium and phosphate.
The solutions dialyzed for 24 hours, except in the case of the time course experiments, after which time the volumes in the inside and outside of the membranes were measured. The contents on the inside and outside of the dialysis membrane were removed and analyzed for total calcium concentration.
In a separate experiment to determine the rate and kinetics of calcium and phosphate equilibration across the dialysis membrane, total calcium and phosphate concentrations were measured at 1, 4, 8, 12, and 24 hours during dialysis with and without albumin in the buffer. The initial conditions placed the digested mixtures inside the dialysis tubing, and either 0.9% NaCl with 1% albumin or 0.9% NaCl as dialyzate on the outside. Additionally, total calcium and phosphate concentrations were measured after days 1, 2, 3, 4, and 5 of dialysis with albumin in the buffer to look for long-term changes in the redistribution of calcium and phosphate in the dialysis system.
After dialysis, total dialyzable calcium was measured by analyzing the calcium content of the dialyzate using an atomic absorption spectrophotometer (Perkin Elmer Model 3100, Norwalk, CT). Calcium standards were made at calcium concentrations of 0.5 to 10 ppm in buffer containing 0.01 N HCl with 0.5% lanthanum oxide. The digested samples were diluted in buffer containing 0.01 N HCl with 0.5% lanthanum oxide so they could be measured within the range of the calcium standards.
Calcium dialyzability, which is the amount of dialyzable calcium, was estimated by the equilibrium dialysis of calcium in this system and it is the amount of soluble calcium capable of moving through the dialysis tubing during the 24-hour time period of dialysis, when equilibrium was achieved.
Percent Calcium Dialyzability
Percent calcium dialyzability was calculated according to the following formula:
Total Phosphate Assay
Phosphate concentration was determined using a phosphate colorimetric assay kit (BioVision K410–500, Mountain View, CA). The assay uses a preparation of malachite green and ammonium molybdate, which forms a chromogenic complex with phosphate ions, resulting in an absorption band around 650 nm. The kit can directly determine phosphate concentrations between 1 μmol/L and 1 mmol/L, with a lower limit of detection of approximately 0.1 nmol.
The assay was performed after in vitro digestion on all of the DHM samples. A microplate reader (Multiskan EX, Thermo Electron Corp, Vantaa, Finland) was used to measure absorbance at 650 nm. A standard curve was created by plotting absorbance at 650 nm as a function of phosphate concentration. The standard curve was used to determine the phosphate concentration of each unknown sample.
First, 0 to 200 μL of each standard and sample was added to the wells of the 96-well microplate and the volume of each well was adjusted to 200 μL with distilled water. Next, 30 μL of the phosphate reagent was added to all of the standard and sample wells. The samples were mixed for 30 seconds on a plate shaker and incubated for 30 minutes at room temperature. After the 30-minute incubation period, the absorbance of each sample was measured at 650 nm in triplicate.
The equation resulting from the standard curve was used to determine the phosphate concentration of each sample. The equation is as follows:
where absorbance = average of the absorbencies of wells for each sample; intercept = y-intercept from the standard curve graph; slope = slope from the standard curve graph; DF = dilution factor used to dilute DHM samples (dilution factor used varied among samples).
Experiments were performed in triplicate for analysis of total calcium and phosphate. Statistical analysis was performed using JMP (SAS Inc, Cary, NC). One-way analysis of variance with the Tukey post hoc test to describe the relation between means was used, and P < 0.05 was regarded as statistically significant.
Percent Dialyzable Calcium
When albumin was included in the dialysis buffer, the percentage of dialyzable calcium in DHM without additives (33.14% ± 8.3) and with added calcium (47.10% ± 8.7), phosphate (32.88% ± 4.9), and calcium and phosphate together (35.25% ± 7.3) was significantly greater than the percentage of dialyzable calcium in DHM with added Similac human milk fortifier (10.15% ± 1.4%) (P < 0.0001). Percentage of dialyzable calcium in DHM with added calcium and with calcium and phosphate together was significantly greater than in DHM with added Similac Neosure (14.32% ± 2.3%), Enfamil Enfacare (14.28% ± 0.68%), or Similac human milk fortifier (P < 0.0001). Percentage of dialyzable calcium was significantly greater in DHM without additives than in DHM with added Similac human milk fortifier (Table 2).
When albumin was not used in the dialysis buffer, the percentage of dialyzable calcium in DHM with added calcium (27.56% ± 4.0%), phosphate (32.74% ± 3.7%), and calcium and phosphate together (33.55% ± 4.8%) was significantly greater than the percentage of dialyzable calcium in DHM with added Enfamil Enfacare (13.43% ± 1.6)%, with added Similac NeoSure (13.24% ± 0.75%), and with added Similac human milk fortifier (12.58% ± 0.5%) (P < 0.0001).
When the 2 methods of dialysis are compared, with and without use of albumin in the dialysis buffer, percentage of dialyzable calcium is significantly greater in DHM with added calcium when albumin is included in the dialysis buffer (47.10% ± 8.7%) than when albumin is not included in the buffer (27.56% ± 4.0%) (P = 0.0018) (Table 2).
Total Dialyzable Calcium
When albumin was used in the dialysis buffer, the total dialyzable calcium was significantly greater in DHM with added calcium (1694 ± 313.4 μg/mL) and with added calcium and phosphate together (1268.01 ± 261.8 μg/mL) than in all other treatment groups (P < 0.0001). Dialyzable calcium in DHM without additives (67.68 ± 16.9 μg/mL) was not significantly different from dialyzable calcium in DHM with added phosphate (67.13 ± 9.9 μg/mL), Similac human milk fortifier (135.84 ± 18.3 μg/mL), Similac NeoSure (74.29 ± 12.36 μg/mL), and Enfamil Enfacare (75.15 ± 3.6 μg/mL).
When albumin was not used in the dialysis buffer (Table 2), dialyzable calcium was significantly greater in DHM with added calcium (991.22 ± 145.2 μg/mL) and with added calcium and phosphate together (1206.82 ± 171.6 μg/mL) than in all other treatment groups (P < 0.0001). Dialyzable calcium in DHM without additives (49.94 ± 9.9 μg/mL) was not significantly different from dialyzable calcium in DHM with added phosphate (66.85 ± 7.5 μg/mL), Similac human milk fortifier (168.32 ± 7.1 μg/mL), Similac NeoSure (68.70 ± 3.9 μg/mL), and Enfamil Enfacare (70.66 ± 8.4 μg/mL).
When the 2 methods of dialysis are compared, with and without use of albumin in the dialysis buffer, total dialyzable calcium was significantly greater in the DHM with added calcium when albumin was included in the dialysis buffer (1694 ± 313.4 μg/mL) compared with when albumin was not included in the buffer (991.22 ± 145.17 μg/mL) (P < 0.0019).
Total Dialyzable Phosphate
There were no statistically significant differences in total dialyzable phosphate between the treatment groups, both with (P = 0.1389) and without albumin in the dialysis buffer (P = 0.1661) (Table 3). When these 2 methods of dialysis were compared, there were no statistically significant differences in total dialyzable phosphate between the 2 groups. Although it is not statistically significant, there was a trend for an increase in total dialyzable phosphate for DHM with added calcium or powdered infant formulas.
Time Course Experiments
Change in Dialyzable Calcium Concentrations During 24 Hours of Dialysis
Regardless of whether albumin was used in the dialysis buffer, dialyzable calcium was greatest in DHM with calcium and phosphate added (Table 4). When digests of DHM and DHM with Enfamil Enfacare were dialyzed for increasing lengths of time up to 24 hours, with and without albumin in the buffer, dialyzable calcium concentrations came to equilibrium at or before 24 hours. Dialyzable calcium in DHM with calcium and phosphate together came to equilibrium after 24 hours both when albumin was included in the dialysis buffer and when it was not included in the buffer (Table 4).
Change in Dialyzable Phosphate Concentration During 24 Hours of Dialysis
Regardless of whether albumin was used in the dialysis buffer, dialyzable phosphate was greatest in DHM with calcium and phosphate added (Table 5). Dialyzable phosphate plateaued between 8 and 12 hours when digests DHM, DHM with calcium and phosphate, and DHM with Enfamil Enfacare were dialyzed for 24 hours. Dialyzable phosphate concentrations reached equilibrium between 8 and 12 hours.
Quantity of Calcium Dialyzable During Successive 24-hour Intervals
The results of the time course data shown in Table 6 indicate that when albumin was used in the dialysis buffer and was replaced every 24 hours, most of the calcium was dialyzed within the 5 days. In DHM with calcium or calcium plus phosphate added, there was an exponential washout of the calcium during a period of 4 days, but most was removed within the first 24 hours (Table 6).
Change in Phosphate Dialyzability During 5 Days of Dialysis
The results of the time course data shown in Table 7 indicate that when albumin was used in the dialysis buffer and was replaced every 24 hours, most of the phosphate was dialyzed within 24 hours (Table 7).
To understand the effect of preparing DHM according to present hospital feeding protocols on the dialyzability of calcium and phosphate in DHM, we developed a simulated premature infant digestion system to measure the dialyzability of calcium and phosphate in DHM before and after adding common nutritional additives used in NICUs, including calcium glubionate, sodium potassium phosphate, calcium glubionate and sodium potassium phosphate together, Enfamil Enfacare, Similac human milk fortifier, and Similac NeoSure. Following in vitro digestion, samples were analyzed for dialyzable calcium and phosphate, which estimate calcium and phosphate that is soluble in the gastrointestinal tract and available for in vivo absorption.
A comprehensive literature review was used to design optimal in vitro digestion procedures, including the appropriate digestive pH, enzyme types, enzyme levels, and transit time for the gastrointestinal tract of a premature infant. In vitro digestion has been used to measure dialyzability of minerals in human milk (11,14,18) and in bovine milk and infant formulas (6,11,15); however, the method has not been sufficiently modified to model a premature infant's gastrointestinal tract and applied to donor milk containing nutritional additives.
Modifications to previously described methods of in vitro digestion (6,14,15) were made. Lipase was added in the gastric phase because it has been shown that there is a high degree of gastric lipolysis, even in premature infants (17), and that gastric lipase is resistant to pepsin (16). Lipase activity in preterm infants is equal to that of healthy adults consuming a high-fat diet (23 ± 5 vs 23 ± 3 U/kg body weight, respectively) and is higher than in adults consuming a low-fat diet (5.2 ± 1.3 U/kg) (19). The lipase used was Rhizopus niveus because it has a similar specificity for fatty acids as the gastric lipase, preferentially hydrolyzing the fatty acids at the Sn-1 and Sn-3 positions of glycerol, whereas human gastric lipase preferentially hydrolyzes at the Sn3-position (16). Additionally, the amount of lipase used was based on previously reported lipase activity and output in premature infants (19). A gastric pH of 5 was used because gastric contents of gavage-fed premature infants maintain pH >5 for the entire postprandial period (16).
In our initial design for this in vitro experiment, albumin was used in the buffer for dialysis to create a physiological simulation of intestinal calcium transport into plasma. As calcium ions leave the gut through the paracellular pathway, they can bind to plasma albumin and reduce the effect of a calcium ion gradient. In the present experiment, all of the samples were dialyzed with and without albumin in the buffer to compare the albumin effect on the redistribution of calcium and phosphate in the dialysis system. The only statistically significant difference caused by using albumin in the dialysis buffer was in the percentage of dialyzable calcium in DHM with added calcium, which was significantly greater when albumin is included in the dialysis buffer (47.10% ± 8.7%) than when albumin is not included in the buffer (27.56% ± 4.0%) (P = 0.0451) (Table 2). When albumin was not present in the dialysis buffer, phosphate increased the dialyzable calcium obtained from added calcium glubionate but not the native calcium from the DHM perhaps by binding to compounds that would otherwise bind to calcium, rendering it less dialyzable.
Percentage of dialyzable calcium was significantly greater in DHM with added calcium when albumin was used in the dialysis buffer (Table 2); however, the same effect was not observed when calcium and phosphate were added to DHM together, suggesting that calcium preferentially binds to phosphate over albumin, and diffusion out of the dialysis tubing could be in the form of a calcium phosphate complex. Conversely, when albumin was not present in the dialysis buffer, phosphate increased the dialyzability of the calcium added as calcium glubionate, suggesting that perhaps the phosphate prevents calcium binding to compounds that would impair its dialyzability.
Addition of the powdered formulas and HMF decreased the percentage of dialyzable calcium when compared with DHM with or without added calcium, phosphate, and calcium and phosphate together. The values measured for percentage of dialyzable calcium in DHM without additives (33.14% ± 8.3%) were higher than those reported (17,18), that is, 13.6% ± 0.8% and 19.6% ± 2.1%, respectively (8,11). This most likely reflects differences in in vitro digestion methods. The percentage of dialyzable calcium in DHM without additives and without albumin (24.46% ± 4.8%) is closer to these previous measurements. Comparisons between results with different methods are hard to interpret; however, the changes we observed with different additives used the same methods, and thus comparisons should be valid.
Addition of HMF and the powdered infant formulas to DHM did not affect the quantity of total dialyzable calcium but did decrease the percentage dialyzable. Consistent with the findings in the present study, other authors have found significantly lower values for calcium absorption from infant formulas than from human milk (11,20–22). Infant formulas made from bovine milk have more micellar casein. When casein clots, it binds calcium and is less well digested than the nonmicellar proteins of human milk. Additionally, the formulas and HMF used in the present study have differing fatty acid profiles, and certainly have different fatty acid compositions than are found in human milk. Fortifiers or formulas with lipid profiles more similar to human milk may result in better calcium dialyzability. Future studies on human-derived fortifiers may help elucidate this.
Calcium absorption percentages for human milk measured in vivo by various authors are consistently higher than the percentage of dialyzable calcium measured in the present study (33.14% ± 8.3%). Previous in vivo studies reported values of 65% (23), 67.2% ± 3.6% (20), 76% (22), and 61.3% ± 22.7% (24). Possible reasons for this discrepancy are better mixing in vivo, active transport of Ca2+ from the duodenum into a larger blood volume relative to the design of the Float-a-Lyzer, which would allow for further dissociation of Ca2+ from large molecular weight complexes, or direct absorption of larger molecular weight complexes in the lower small intestine in vivo than in the dialysis system.
Adding calcium, phosphate, HMF, and powdered infant formula to DHM did not affect phosphate dialyzability. As the calcium content of the additives increased, there was a slight trend for an increase in phosphate dialyzability, but the trend is not statistically significant.
It appears that calcium interacts with phosphate and albumin, and that phosphate alone extracts calcium from nondialyzable binding sites in milk, but it does not allow it to associate with albumin. The increased calcium would have little effect on dialyzable phosphate if the calcium-phosphate complex passes through the pores in the dialysis membranes and phosphate does not bind to albumin.
When total dialyzable calcium and phosphate concentrations were measured in digests of DHM, DHM with calcium and phosphate, and DHM with Enfamil Enfacare at 1, 4, 8, 12, and 24 hours during dialysis with and without albumin in the buffer, dialyzable calcium and phosphate come to equilibrium after 12 hours in most cases. The exception was the milk sample with both calcium glubionate and sodium potassium phosphate added and with albumin in the buffer, which was approaching equilibrium at 24 hours, suggesting slower dissociation of calcium phosphate complexes and binding to albumin than in the case of the other samples. When total dialyzable calcium and phosphate concentrations were measured on days 1, 2, 3, 4, and 5 of dialysis with albumin in the buffer and daily dialyzate replacement, >75% of the calcium and >50% of the phosphate were dialyzed within 24 hours. Measurements of gut transit times in neonates indicate that this is <24 hours (25). Therefore, 24 hours should be a sufficient time for equilibrium dialysis to occur, and a logical termination for the experiment because anything left in the infant's gut after 24 hours would most likely be excreted in the feces.
Results obtained from the applied in vitro digestion methods show an increase in calcium dialyzability in DHM with added calcium glubionate and calcium glubionate with sodium potassium phosphate together. Addition of powdered infant formulas did not increase dialyzable calcium. When premature infants are at risk for bone disease, adding calcium glubionate, or calcium glubionate with sodium potassium phosphate together, is an option that may provide the most dialyzable calcium.
Our method models calcium digestion in premature infants, for which most calcium absorption is a passive, paracellular transport process (26). For adults, vitamin D is important for calcium homeostasis and bone mineralization; however, fetal mineral ion homeostasis appears to be independent of vitamin D (26).
The development of an in vitro model to simulate digestion in the preterm infant is a reliable way to compare the relative dialyzability of minerals in human milk with and without common additives used in NICUs. The method is simple and inexpensive compared to methods using animals and cell culture. Future research should focus on improving premature infant feeding protocols to provide optimal growth, bone mineralization, and short-term and long-term health outcomes while still providing and preserving the desirable nonnutritive effects of human milk. Future research should use the in vitro system to examine the effects of liquid infant formulas as well as formulas made from human milk on dialyzability of calcium and phosphate as well as their interaction with other nutrients, such as protein, fat, and other minerals.
The authors acknowledge Jerry Spears, Laurie Dunn, Miriam Labbok, and the late Mary Rose Tully for their significant contributions to this work in providing their knowledge and expertise.