A Human Milk–Based Protein Concentrate Developed for Preterm Infants Retains Bioactive Proteins and Supports Growth of Weanling Rats

ABSTRACT

Background

Bovine milk–based protein modulars are currently available to nutrient-enrich enteral feedings; however, they have limitations for use in very-low-birth-weight infants.

Objectives

Our objectives were to develop a human milk–based protein (HMP) concentrate and to conduct a preclinical assessment of the HMP concentrate in weanling rats.

Methods

An HMP concentrate was produced from donor milk using pressure-driven membrane filtration processes and high hydrostatic pressure processing. Protein and lactoferrin concentrations and lysozyme activity were determined by Kjeldahl, HPLC, and turbidimetric assay, respectively. Male Sprague Dawley rats 24 d old (n = 30) were randomly assigned to 1 of 3 isocaloric AIN-93G diets for 4 wk containing 100% casein (control) or with 50% of the casein replaced with the HMP concentrate (treatment) or a bovine whey protein isolate (treatment). Body weight, food intake, fat mass, plasma amino acid profiles, and organ weights were measured. Data were analyzed using linear regression models.

Results

Raw donor milk contained (mean ± SD) 101 ± 6 g protein/kg and 5 ± 1 g lactoferrin/kg of milk solids. Postprocessing, protein and lactoferrin concentrations were 589 ± 3 g/kg and 29 ± 10 g/kg, respectively. Lysozyme activity was initially 209 ± 4 U/kg and increased to 959 ± 39 U/kg in the HMP concentrate. There were no statistically significant differences in body weight, food intake, fat mass, or plasma amino acid profiles between rats fed diets containing the HMP concentrate and the control diet. Full cecum weights were higher in rats fed the HMP concentrate than in those fed control diets (mean difference: 5.59 g; 95% CI: 4.50, 6.68 g; P < 0.0001), likely reflecting the concentration of human milk oligosaccharides. No differences were found for other organ weights.

Conclusions

The HMP concentrate retained important bioactive proteins and supported normal rat growth in the preclinical assessment.

Introduction

Although mother's milk with supplemental human donor milk is the recommended way to feed very-low-birth-weight (VLBW; <1500 g) infants, human milk alone is insufficient to meet their elevated nutritional requirements (1, 2). As such, human milk feeds are enriched with bovine milk– or human milk–based multinutrient fortifiers during initial hospitalization (3). Commercially available multinutrient fortifiers differ in their nutrient composition; however, for smaller VLBW infants and infants with fluid restriction, an additional protein modular is frequently added to human milk to improve growth and lean body mass accretion (4, 5). Currently, the protein modulars available to fortify human milk are limited. In North America, bovine milk–based protein modulars are available and routinely used in neonatal intensive care units. One commonly used bovine milk–based protein modular is a whey-based powder; although it has the advantage of only minimally diluting mother's own milk, it is not sterile and possesses the risk of transmitting potential pathogens to a vulnerable infant. It is recommended by the WHO, federal agencies, and professional associations that powders should be avoided when a nutritionally appropriate sterile alternative is commercially available (6–9). It is also used off-label for VLBW infants because it is only recommended for individuals >3 y of age. A second commonly used commercially available fortifier is a liquid bovine casein hydrolysate; although it is sterile, it disproportionately displaces the volume of mother's milk (10). Further, there is significant interest to replace bovine milk–based protein modulars in the neonatal intensive care unit with human milk–based products. The proteins in human milk not only provide a source of amino acids, but they also possess various bioactive components that serve a variety of important functions including facilitating the absorption of other nutrients (e.g., haptocorrin, α-lactalbumin, and lactoferrin) and providing a complex set of protective mechanisms against pathogenic bacteria and viruses (e.g., lactoferrin, lysozyme, and secretory IgA) (11, 12).

Human milk–based protein (HMP) concentrates were first explored >30 y ago during the early development of multinutrient fortifiers. Polberger et al. (13) randomly assigned VLBW infants to receive unfortified human milk or human milk fortified with an HMP concentrate. The concentrate was produced by passing de-fatted milk through an ultrafiltration membrane, followed by freeze-drying it into a powder, which contained ∼70% protein. Infants that received the HMP concentrate (mean protein intake: 3.6 g · kg⁻¹ · d⁻¹) had a mean weight gain of 20.0 g · kg⁻¹ · d⁻¹, which was significantly higher than the growth rates of infants allocated to receive unfortified human milk (weight gain: 15.3 g · kg⁻¹ · d⁻¹; protein intake: 2.1 g · kg⁻¹ · d⁻¹). Further development of human milk–based fortifiers was abandoned owing to the concern over HIV transmission in human milk during the early 1980s, and the impracticality of securing large volumes of donor milk. Given the re-emergence of human milk banking in recent years, donor milk is now available in large volumes and is pasteurized by the Holder method (62.5°C for 30 min) to eliminate known pathogens including HIV (14).

Although this varies, it is estimated that even after careful screening of donors and issuing guidance on hygienic practices to follow when pumping breast milk, ∼15% of the milk collected by milk banks is discarded because it contains a higher bacterial load than pre- or postpasteurization cutoffs allow (15, 16). Given the significant investment made by mothers, costs incurred by milk banks to screen and process the milk, and limitations with currently available protein modulars, investigating whether this milk could be diverted toward the production of an HMP concentrate is warranted. The aims of this study, therefore, were 1) to develop a prototype for an HMP concentrate and 2) to conduct a preclinical assessment of the product to determine whether its use can support normal growth in weanling rats.

Methods

Development of the HMP concentrate

Before collecting human donor milk from the Rogers Hixon Ontario Human Milk Bank, research ethics board approval was obtained from Sinai Health. Mothers consented at the time of donation to use their milk for research. Further, all donors had screening blood tests completed to rule out HIV, Human T-lymphotropic Virus, hepatitis C, hepatitis B, and syphilis before their first milk donation and at regular intervals thereafter (16). Raw human milk (∼180 L) that had a bacterial load >5 × 10⁷ CFU/L was used for the present study. According to local standard operating procedures at the time, milk with a prepasteurization bacterial load >5 × 10⁷ CFU/L was neither processed nor distributed to feed infants.

Frozen donated milk was shipped by courier to the Dairy Science and Technology Research Centre (STELA) in Quebec, Canada, and subsequently thawed at 4°C overnight and pooled for processing. Supplemental Figure 1 provides a graphical depiction of the processing steps required to produce the HMP concentrate. First, fat was removed from the milk by centrifugation (1700 × g at 4°C for 15 min) followed by microfiltration (Tetra Alcross MFS-7 pilot plant, Tetra Pak Filtration solutions) using a membrane pore diameter of 1.4 μm to remove bacteria and the remaining fat globules. The microfiltration permeate was subsequently processed by ultrafiltration (Koch Membrane Systems Inc.) with a membrane molecular weight cutoff of 10 kDa to reach a volume concentration factor of 4X. During ultrafiltration, suspended solids and solutes with molecular weights >10 kDa are retained, whereas water and solutes with molecular weights <10 kDa pass through the membrane and are filtered out. The retentate was then diafiltered using 6 diavolumes, which further concentrated the proteins by removing lactose, oligosaccharides, and salts (17). The HMP concentrate was then pasteurized by high hydrostatic pressure (HHP) processing (Hiperbaric 135, Hiperbaric). Based on published work conducted by Pitino et al. (18), a pressure of 500 MPa at 4°C for 8 min was selected for pasteurization because this was shown to reduce bacteria to below detectable concentrations, while better preserving the bioactivity of proteins (e.g., lactoferrin, lysozyme, and bile salt–stimulated lipase). For the purpose of incorporating the concentrate into rat diets, the product was then freeze-dried (Sp Scientific/REEP Virtis) into a powder. Samples were collected after each processing step to determine macronutrient concentrations and to confirm the retention of bioactive proteins.

Nutrient, bioactive component, and microbiological analyses of milk, HMP, and rat diets

The protein contents of raw donor milk, the final HMP concentrate, and samples collected at intermediary steps (e.g., after centrifugation and microfiltration) were determined by the Kjeldahl method (19). As is recommended for human milk, total protein was calculated from total nitrogen content multiplied by the protein conversion factor of 6.25 (20). The macronutrient content of samples was determined using a mid-infrared human milk analyzer (Miris), calibrated against wet-chemistry methods. The precision of the analyzer was confirmed by analyzing an internal quality control donor milk sample (n = 55) and the interassay CV for each macronutrient was <6%.

Lactoferrin concentration was determined by HPLC (Agilent 1260 Infinity II system) using human milk lactoferrin (Sigma-Aldrich) to generate a standard curve as previously described (18, 21). Measurement of lactoferrin in a homogeneous pool of donor milk (n = 4) yielded a mean ± SD of 0.89 ± 0.06 g/L with an interassay CV of 6.22%. Lysozyme activity was determined by turbidimetric assay in 96-well plates using Micrococcus lysodeitikus as the test organism (18, 22). Lysozyme isolated from chicken egg white (99% purity) (Sigma-Aldrich L6876) was used as a reference material in this analysis, producing a mean ± SD recovery of 99% ± 10% and an interassay CV of <5% (n = 3).

Samples of human milk and concentrate collected at each processing step were cultured to determine the total bacterial count at Sinai Health. Blood and MacConkey agar incubated for 48 h at 37°C with 5% CO₂ were used to determine total bacterial counts. The minimum detection concentration for this microbiological analysis was 1 × 10³ CFU/L. Bacterial species were then identified using Matrix Assisted Laser Desorption Ionization Time-of-Flight technology (Vitek-MS).

Rat bioassay

The animal experiments were approved by the local Animal Care Committee at The Hospital for Sick Children. All animal procedures followed the established guidelines for the care and handling of laboratory rats. Weanling male Sprague Dawley rats (n = 30) weighing 50–60 g at 22 d of age were obtained from Charles River Canada and acclimated for a 48-h period to a standard AIN-93G purified diet (Research Diets) (23, 24). Rats were housed individually in experimental rooms maintained at 22 ± 2°C, at 45%–60% relative humidity with a 12-h light/dark daily cycle. Rats were then randomly assigned to 1 of 3 groups using a computer-generated randomization schedule and fed for 4 wk AIN-93G rodent diets differing in their protein source: 1) 100% casein (AIN-93G standard, control), 2) 50% HMP concentrate and 50% casein, or 3) 50% commercially available bovine whey protein isolate (BWPI) and 50% casein (Table 1). The third experimental group was added to compare the effect of feeding the HMP concentrate with the BWPI commonly used to fortify feeds in neonatal intensive care units. Because a protein modular in clinical practice would not be used to replace endogenous proteins found in mother's own or donor milk, the HMP concentrate and the BWPI were incorporated at 50% of the total protein in the rat diets with the remaining 50% provided as casein, methionine, and cystine. Beneprotein (Nestlé Health Science), reported to have a purity of 85.7% protein, was used as the commercially available BWPI. Sulfur amino acids were added to the diets, as recommended by the AIN, to address the elevated requirements of rats for these amino acids (Table 2) (23, 24). Diets were then analyzed for total protein content by the Kjeldahl method to ensure they were isonitrogenous (25). Amino acid concentrations of the HMP concentrate and the BWPI were determined by an ultra-performance liquid chromatography system as described below. The total amino acid content of diets was then determined using the analyzed values for HMP concentrate and BWPI by ultra-performance liquid chromatography and reference amino acid concentration values for casein provided by Research Diets Inc. Diets were γ-irradiated to ensure the elimination of any remaining viruses and bacteria.

TABLE 1.

Composition of the control, HMP concentrate, and BWPI diets fed to male Sprague Dawley rats¹

Components, g/kg	Control diet	HMP diet	BWPI diet
Casein	200	99	99
dl-Methionine	—	1	1
l-Cystine	3	2	2
HMP	—	142	—
BWPI	—	—	103
Corn starch	398	371	398
Maltodextrin	132	132	132
Sucrose	100	100	100
Cellulose	50	50	50
Soybean oil	70	52	70
tBHQ	0.01	0.01	0.01
Mineral mix2	35	35	35
Vitamin mix2	10	10	10
Choline bitartrate	3	3	3
Analyzed total protein,3 % of dry matter	17	18	17

¹BWPI, bovine whey protein isolate; HMP, human milk–based protein; tBHQ, tert-butylhydroquinone.

²Prepared according to AIN-93G formulation (23).

³Analyzed total protein content determined by the Kjeldahl method (total nitrogen × 6.25).

TABLE 2.

Concentrations of amino acids in the control, HMP concentrate, and BWPI diets fed to male Sprague Dawley rats¹

Amino acids, g/kg	Control diet	HMP diet	BWPI diet
Alanine	5.0	5.7	7.0
Arginine	5.9	6.6	4.8
Aspartic acid	12.0	15.0	15.2
Cysteine	4.2	4.2	4.2
Glutamic acid	37.7	35.1	33.9
Glycine	3.0	3.5	2.9
Histidine	4.5	4.9	3.7
Isoleucine	7.5	8.9	9.2
Leucine	15.7	17.2	17.3
Lysine	13.0	11.1	16.3
Methionine	5.0	5.0	5.0
Phenylalanine	8.3	7.5	6.6
Proline	17.6	17.5	13.6
Serine	9.9	9.7	8.8
Threonine	7.1	7.8	9.3
Tryptophan	2.1	3.0	2.5
Tyrosine	9.0	9.1	7.0
Valine	9.2	10.1	9.5

¹Amino acid concentrations of diets were determined using the analyzed values for HMP and BWPI (HPLC, n = 3 each) and reference amino acid concentration values for casein provided by Research Diets Inc. BWPI, bovine whey protein isolate; HMP, human milk–based protein.

Throughout the 4-wk study period, rats were provided food and water ad libitum. Throughout the intervention the body weight of, and amount of feed consumed by, each rat were assessed daily. After 4 wk, rats were deprived of food for a 12-h period before blood collection. Rats were then anesthetized using isoflurane delivered at 4% and killed by exsanguination (26). Blood was collected into EDTA-treated tubes. The plasma was immediately separated after blood collection by centrifugation (2000 × g for 15 min at 4°C) and the supernatant was stored at −80°C until analysis.

A subset of rats (n = 5/group) was randomly selected using a computer-generated random number table to undergo MRI to determine fat mass. Selected rats were immediately transported after being killed to the MRI facility at The Hospital for Sick Children. Imaging was conducted by a research MRI technologist on a 3 Tesla MRI Scanner (Siemens PrismaFit) using a Tx/Rx 15-channel knee coil. Coronal images were acquired and provided coverage from neck to tail. The pulse sequence used was a T1-weighted turbo spin echo with repetition time = 947 ms; echo time = 15 ms; number of slices = 25; slice thickness = 2 mm; field of view (FOV) = 175 mm; FOV phase = 50%; acquisition matrix = 352 × 704; resolution = 0.2 × 0.2 × 2.0 mm³; flip angle = 150°; number of singles averaged = 6; and acquisition time = 16 min 50 sec. OSIRIX Lite version 9.0 (Pixmeo) medical imaging software was then used to quantify the total body fat from the Digital Imaging and Communications in Medicine (DICOM) images obtained. Semiautomated fat segmentation was done using a threshold-based region growing algorithm (27). The computed fat volume was then multiplied by the assumed density of adipose tissue (0.92 g/cm³) to obtain total fat mass (28).

Necropsies were conducted on all rats following the procedures outlined by Fiette and Slaoui (29). The brain, heart, liver, kidneys, spleen, stomach, and small and large intestine (with contents) were removed, cleaned, and weighed. All gastrointestinal organs were then rinsed with 0.9% sterile saline and reweighed.

Plasma amino acid profiles were determined at Sparc BioCentre Facility at The Hospital for Sick Children using procedures previously described by Bidlingmeyer et al. (30). Plasma samples were first deproteinized using acetonitrile; norleucine (internal standard) was then added and the solution was centrifuged (10,000 × g for 15 min at 4°C) to precipitate proteins. Thereafter samples were derivatized with a methanol: water: triethylamine: phenylisothiocyanate (7:1:1:1) solution for 15 min at room temperature. Derivatized amino acid profiles were determined using a Waters Acuity UPLC system consisting of a binary solvent manager, a sample manager, a tunable UV detector, and a BEH C18 column (2.1 × 100 mm). A modified Pico-Tag gradient, with proprietary buffers (Pico-Tag Eluent 1 and 2) from Waters, was used (31). The system was operated at a flow rate of 0.5 mL/min at 48°C and 6 μL of each sample was injected into the column. Derivatized amino acids were detected at 254 nm and Empower 3 (Waters) software was used for system control and data acquisition.

Statistical analyses

Statistical analyses were conducted using SAS version 9.4 (SAS Institute). Differences in weight gain and the amount of feed consumed between treatment groups were evaluated using repeated-measures linear regression models (PROC MIXED). Autoregressive covariance structures were used to assess longitudinal changes in body weight adjusting for treatment, time, and the interaction term between these 2 variables. If the interaction term was not statistically significant, it was removed from the model and the analyses were rerun. Data collected at 1 time point (i.e., organ weights, plasma amino acid profiles, fat mass) were analyzed using a 1-factor ANOVA followed by post hoc pairwise comparisons using least-squares means (LS-MEANS). For all statistical analyses, a P value < 0.05 was considered statistically significant.

Results

Nutrient, bioactive component, and microbiological analyses of milk and HMP concentrate

Table 3 summarizes the macronutrient and lactoferrin contents and lysozyme activity of milk collected at each processing step. Compared with an initial mean ± SD protein content of 101 ± 6 g/kg of milk solids in raw donor milk, the final concentrate contained 589 ± 3 g/kg. Centrifugation removed a large proportion of the fat, which was then further reduced through microfiltration: a processing step incorporated for the removal of bacteria from the de-fatted milk. For carbohydrate content, a significant amount was removed through ultrafiltration-diafiltration, although the final concentrate still contained 188 ± 53 g/kg of milk solids. Lactoferrin concentration was initially 5 ± 1 g/kg of milk solids in raw milk and was concentrated to 29 ± 10 g/kg in the final product. Lysozyme activity in the raw donor milk was measured at 209 ± 4 U/mg of milk solids, whereas the final protein concentrate contained 959 ± 39 U/mg.

TABLE 3.

Macronutrient and lactoferrin concentrations and lysozyme activity after each processing step in the production of a human milk–based protein concentrate¹

	Analyte
Processing step	Total protein, g/kg	Fat, g/kg	Carbohydrate, g/kg	Lactoferrin, g/kg	Lysozyme, U/mg
Raw donor milk	101 ± 6	272 ± 6	621 ± 19	5 ± 1	209 ± 4
Postcentrifugation	129 ± 4	62 ± 8	774 ± 8	9 ± 2	299 ± 24
Postmicrofiltration	132 ± 7	17 ± 8	780 ± 25	10 ± 2	306 ± 12
Post–ultrafiltration- diafiltration	592 ± 16	108 ± 17	229 ± 17	32 ± 12	999 ± 26
Postpasteurization	589 ± 3	125 ± 0	188 ± 53	29 ± 10	959 ± 39

¹Values are mean ± SD, n = 3 for each processing step. Total protein was determined by the Kjeldahl method (total nitrogen × 6.25). Fat and carbohydrate concentrations were determined using a mid-infrared human milk analyzer calibrated against wet-chemistry methods. Lactoferrin concentrations and lysozyme activities were determined by HPLC and turbidimetric assay, respectively. Data are displayed on a dry basis, determined by dividing the amount of macronutrient, lactoferrin, or lysozyme by the total milk solid content.

Table 4 summarizes results for the microbiological analysis of milk samples (n = 3) at each processing step. In raw donor milk, a median of 7 × 10⁶ [min: 6 × 10⁶ and max: 7 × 10⁶] CFU/L were measured; however, the spore-forming bacteria Bacillus cereus and Bacillus sphaericus were not detected. Postultrafiltration, the total colony count increased to 2 × 10⁸ [min: 2 × 10⁸ and max: 3 × 10⁸] CFU/L and no spore-forming bacteria were detected. After pasteurization, no bacteria were isolated apart from B. cereus and B. sphaericus, which were then further concentrated after freeze-drying to 8 × 10⁵ [min: 6 × 10⁵ and max: 1 × 10⁶] CFU/L.

TABLE 4.

Results for the microbiological analysis of milk samples at each processing step¹

	Processing step
Species, CFU/L	Raw donor milk	Post-ultrafiltration-diafiltration	Post-pasteurization	Post-freeze-drying
Enterobacter cloacae complex	2 × 106 [2 × 106 – 2 × 106]	1 × 108 [1 × 108 – 1 × 108]	—	—
Serratia liquefaciens	—	0 [0 – 1 × 108]	—	—
Serratia marcescens	5 × 106 [5 × 105 – 5 × 106]	3 × 107 [3 × 107 – 1 × 108]	—	—
Acinetobacter baumannii complex	0 [0 – 5 × 106]	—	—	—
Pantoea agglomerans	—	1 × 108 [0 – 1 × 108]	—	—
Bacillus cereus	—	—	1 × 104 [8 × 103 – 1 × 104]	3 × 105 [2 × 105 – 5 × 105]
Bacillus sphaericus	—	—	3 × 104 [2 × 104 – 6 × 104]	5 × 105 [3 × 105 – 8 × 105]
Total bacterial count	7 × 106 [6 × 106 – 7 × 106]	2 × 108 [2 × 108 – 3 × 108]	4 × 104 [3 × 104 – 8 × 104]	8 × 105 [6 × 105 – 1 × 106]

¹Values are median [min - max], n = 3 for each processing step.

Rat bioassay

The initial and final body weights of rats did not differ between treatment groups (Supplemental Table 1). Likewise, there were no significant differences in daily body weight gain (P = 0.81) or daily feed intake (P = 0.13) over the 4-wk study period (Figure 1). The protein efficiency ratio, calculated as weight gain of the test group in grams divided by the total amount of protein consumed, also did not differ between groups (P = 0.13). The mean ± SD percentages of total body fat at the end of the intervention for rats fed the control, HMP concentrate, and BWPI diets were 10.2% ± 1.9%, 9.1% ± 1.6%, and 11.0% ± 2.4%, respectively. Again, no statistically significant differences were observed between groups (P = 0.35).

FIGURE 1.

Mean body weight (A) and food intake (B) of male Sprague Dawley rats fed the control, HMP concentrate, and BWPI diets. Data are means ± SDs, n = 10/group. Although daily weights and food intake were used in the statistical analysis, for the purpose of clarity, they are presented here as averaged values for every 2 d. Differences in weight gain and the amount of food consumed between treatment groups were evaluated using repeated-measures linear regression models (PROC MIXED). NS interaction terms were removed from final models. BWPI, bovine whey protein isolate; HMP, human milk–based protein; NS, nonsignificant.

In contrast, mean cecum weight (cecal sac plus contents) differed between groups (Table 5). Rats fed the HMP concentrate had a significantly heavier mean cecum weight than those fed the control and BWPI diets (P < 0.001). After the cecum was emptied, rinsed with sterile saline, and weighed, these differences were no longer statistically significant (P = 0.26). No other statistically significant differences in organ weights were observed between groups.

TABLE 5.

Organ weights of male Sprague Dawley rats fed the control, HMP concentrate, and BWPI diets¹

	Treatment
Organ, g	Control	HMP	BWPI
Brain	1.9 ± 0.1	1.8 ± 0.2	1.9 ± 0.2
Heart	1.3 ± 0.2	1.4 ± 0.1	1.5 ± 0.3
Lungs	1.9 ± 0.5	1.7 ± 0.4	1.7 ± 0.4
Liver	12.9 ± 1.2	13.7 ± 1.1	14.2 ± 1.4
Kidneys	2.7 ± 0.2	2.7 ± 0.1	2.7 ± 0.2
Spleen	0.8 ± 0.1	0.7 ± 0.1	0.9 ± 0.2
Stomach2	1.3 ± 0.3	1.2 ± 0.3	1.2 ± 0.2
Small intestine2	4.0 ± 0.8	4.0 ± 1.1	3.5 ± 0.5
Large intestine2	1.7 ± 0.4	1.8 ± 0.6	1.4 ± 0.3
Cecum (sac + contents)	3.8 ± 0.4b	9.4 ± 1.9a	4.4 ± 0.7b

¹Values are mean ± SD, n = 10/group. Data were analyzed using linear regression models (PROC MIXED) followed by post hoc pairwise comparisons using least-squares means (LS-MEANS). Labeled means in a row without a common superscript letter differ, P < 0.05. BWPI, bovine whey protein isolate; HMP, human milk–based protein.

² n = 8/group.

Table 6 presents the mean plasma amino acid concentrations. Plasma threonine concentrations were significantly higher in rats fed the BWPI diet than in those fed the HMP concentrate and control diets (P = 0.005). No significant difference was found in plasma threonine concentrations between rats fed the control and those fed the HMP diet. All other plasma amino acid profiles did not differ between the 3 groups.

TABLE 6.

Plasma amino acid profiles of male Sprague Dawley rats fed the control, HMP concentrate, and BWPI diets¹

	Treatment
Amino acid, μmol/L	Control	HMP	BWPI
Alanine	662 ± 158	683 ± 113	720 ± 235
Arginine	21 ± 17	17 ± 5	26 ± 16
Asparagine	92 ± 17	98 ± 14	95 ± 19
Aspartic acid	11 ± 3	12 ± 4	12 ± 3
Cysteine	10 ± 2	12 ± 3	12 ± 3
Glutamic acid	174 ± 56	178 ± 44	192 ± 47
Glutamine	789 ± 111	762 ± 73	861 ± 186
Glycine	266 ± 6	286 ± 56	284 ± 69
Histidine	23 ± 8	23 ± 6	26 ± 9
Isoleucine	123 ± 26	127 ± 17	142 ± 24
Leucine	196 ± 30	192 ± 29	211 ± 41
Lysine	248 ± 54	272 ± 39	253 ± 46
Methionine	84 ± 20	87 ± 18	100 ± 30
Phenylalanine	94 ± 16	90 ± 10	96 ± 15
Proline	91 ± 18	85 ± 20	85 ± 13
Serine	286 ± 56	307 ± 37	318 ± 71
Threonine	419 ± 110b	458 ± 101b	615 ± 164a
Tryptophan	3 ± 1	3 ± 1	2 ± 1
Tyrosine	182 ± 32	169 ± 23	165 ± 25
Valine	251 ± 43	245 ± 36	261 ± 51

¹Values are means ± SDs, n = 10/group. Data were analyzed using linear regression models (PROC MIXED) followed by post hoc pairwise comparisons using least-squares means (LS-MEANS). Labeled means in a row without a common superscript letter differ, P < 0.05. BWPI, bovine whey protein isolate; HMP, human milk–based protein.

Discussion

The aims of this study were, first, to develop a prototype for an HMP concentrate made from raw donor human milk that would not meet local criteria for dispensing to VLBW infants and, second, to conduct a preclinical assessment of the product by determining whether it can support normal growth in weanling rats. Using a unique combination of centrifugation, microfiltration, ultrafiltration (×4), diafiltration (6 diavolumes), and HHP processing steps, we were able to produce an HMP concentrate containing ∼59% protein on a dry weight basis.

In comparison, a recent study by Oliveira et al. (32) also aimed to develop a lyophilized human milk concentrate. Their process involved freeze-drying 50 mL donor milk to produce a freeze-dried human milk concentrate, which was then immediately reincorporated into 75 mL donor milk. The mean protein and total solid contents of the donor milk before and after incorporating the human milk concentrate were reported. Using these values, we calculated their human milk concentrate to be ∼11.2% protein on a dry weight basis. Similarly, human milk–based multinutrient fortifiers produced as a sterile liquid are now commercially available. These fortifiers are produced by separating the milk into cream and skim milk, removing some water from the skim milk, reincorporating a portion of the cream, followed by thermal double pasteurization (33). Both the commercially available human milk–based multinutrient fortifier and the concentrate developed by Oliveira et al. are intended to be used as multinutrient fortifiers, which raise the concentration of various nutrients. In contrast, our HMP concentrate was developed by combining several membrane filtration techniques to concentrate the protein in human milk. Our protein concentrate more closely resembles the HMP concentrate developed in the 1980s by Polberger et al. (13), which was produced by passing de-fatted milk through an ultrafiltration membrane and was reported to be ∼70% protein. Future modifications to our HMP concentrate's processing procedures include evaluating the use of different ultrafiltration molecular weight cutoffs to increase protein purification.

Rather than using heat treatment, HHP processing was used to pasteurize our HMP concentrate. Based on the study published by Pitino et al. (18), Holder pasteurization appears to significantly reduce the bioactivity of proteins including bile salt–stimulated lipase, lysozyme, and lactoferrin compared with HHP processing. Preserving the bioactivity of human milk proteins is crucial because they provide a multitude of functions for the infant including facilitating the absorption of other nutrients, stimulating growth, modulating the immune system, and providing defense mechanisms against pathogens (11). HHP processing was therefore deemed as the superior alternative and was used in the development of our protein concentrate. Previous studies that have produced human milk–based concentrates did not examine the concentration of bioactive components retained. As noted in Table 3, we found that lactoferrin and lysozyme were concentrated by 5–6-fold. Modeling the actual enteral feedings of a VLBW infant receiving fortified donor milk, the HMP concentrate would incrementally provide ∼1.5 g protein, ∼76 mg lactoferrin, and ∼2,494,180 units of lysozyme (Supplemental Figure 2). This dosage of lactoferrin (76 mg · kg⁻¹ · d⁻¹, assuming a 1-kg infant) would be comparable with that provided in recent human and bovine lactoferrin trials, which showed favorable results. Infants allocated to receive 100 mg/d of bovine lactoferrin had a decreased occurrence of late-onset sepsis and necrotizing enterocolitis, compared with those given a placebo (34). Similarly, in a randomized controlled trial, VLBW infants allocated to receive a human recombinant lactoferrin at a dose of 150 mg/kg every 12 h had a reduced risk of hospital-acquired infections when compared with infants given a placebo (35). Notably, our own HMP concentrate would provide not only a source of lactoferrin, but other important bioactive proteins as well.

Although not the primary aim of our study, we conducted an exploratory analysis of the microbiology results after each processing step to gauge the future work necessary to produce a safe product for infants. Previous studies looking to produce a human milk concentrate did not report their microbiological results (13, 32). With respect to commercially available products, the BWPI is a powder, making it unsterile, whereas the bovine casein hydrolysate and the human milk–based multinutrient fortifier are heat treated to ensure sterility. Of note, all donor milk used in this study was known to have bacteria above the prepasteurization threshold for clinical use at the milk bank. Compared with a median total bacterial colony count of 7 × 10⁶ [min: 6 × 10⁶ and max: 7 × 10⁶] CFU/L measured in raw donor milk, the postultrafiltration samples contained 2 × 10⁸ [min: 2 × 10⁸ and max: 3 × 10⁸] CFU/L, thereby suggesting that both the protein and bacteria were concentrated. Theoretically, the microfiltration step should reduce bacterial contamination in the permeate (Supplemental Figure 1). We hypothesize that the membrane pore size used during the microfiltration step (1.4 μm) was too large to successfully remove bacterial spores as well as the bacterial species identified after ultrafiltration listed in Table 4. In light of these findings, future modifications to optimize the production of the protein concentrate would include reducing the microfiltration membrane pore size to 0.8 μm (36). Notably, B. cereus colonies were not detected in the prepasteurized samples likely owing to their tendency to coalesce; they are also often masked by the presence of other bacteria, thereby hindering the accuracy of their enumeration (37).

To address the second aim of the study, which was to ensure that the HMP concentrate would support growth, we conducted a 4-wk preclinical feeding study where rats were fed AIN-93G diets with half the casein replaced with HMP or BWPI in the experimental groups. Results from this study show that weanling rats fed 50% of their protein as a concentrated HMP did not grow differently from those fed exclusively the AIN-93G casein-based diet nor those with 50% of their protein intake as a BWPI. Likewise, there was no evidence that their organ weights or total fat mass differed by the end of the 4-wk study period. These findings confirm that the processing procedures used to produce the HMP concentrate did not inadvertently affect its protein quality. This preclinical assessment also provides confidence that the use of the HMP concentrate would support the growth of infants in a future clinical trial.

The carbohydrate content of the HMP concentrate contributed ∼18% of total solids after ultrafiltration. The oligosaccharides in human milk, which have molecular weights ≤8 kDa, may not have been entirely filtered out (38). We speculate that the presence of oligosaccharides in the HMP concentrate resulted in the observed increase in full cecum weight of rats fed the HMP diet. This finding is consistent with the literature, where rats and pigs fed diets containing various sources of fiber or resistant starch, including human milk solids rich in oligosaccharides, also had larger full cecum weights (39, 40). This may in part be due to increased microbial fermentation of dietary fibers in the large intestine and the high water-holding capacity of oligosaccharides. Human milk oligosaccharides have become a recent area of interest given the fact that they have been associated with a lower incidence of necrotizing enterocolitis in human milk–fed infants (41). These complex sugars provide a wide range of benefits to the infant such as preventing the adhesion of pathogens and serving as substrates for desired bacteria in the infant's intestinal lumen. Future studies are needed to assess whether the dose of oligosaccharides provided in the HMP concentrate would provide these added benefits without having unintended adverse effects.

Plasma threonine concentrations in rats fed the BWPI diet were significantly higher than in those fed the HMP concentrate and control diets. The elevated plasma threonine concentrations in rats fed the BWPI may, in part, be explained by the higher amount of threonine in the BWPI diet (Table 2). In a study by Boehm et al. (42), increased threonine intake was significantly associated with elevated plasma threonine concentrations in growing rats. Rigo and Senterre (43) have also shown that plasma threonine concentrations of preterm infants increased with higher threonine intake, particularly in infants with the lowest gestational age, suggesting that both the dose of threonine and metabolic maturity play a role in blood concentrations. A study by Darling et al. (44) also showed that despite similar threonine intakes, preterm infants fed a bovine milk–based formula did not oxidize threonine as well as infants fed human milk. This resulted in higher plasma threonine concentrations among infants fed the bovine milk–based formula. High plasma threonine has been shown to increase brain glycine concentrations in rat models and affected neurotransmitter balance in the brain (42). Given the rapid brain growth that occurs during early infancy, excess threonine intake should be avoided because it may influence brain development. Future clinical studies are warranted to assess whether the addition of our HMP concentrate compared with a commercially available BWPI, to preterm milk, improves the modulation of infants’ plasma threonine concentrations.

To our knowledge, this is the first study to combine several processing techniques including the application of HHP to produce an HMP concentrate. Strengths of this preclinical assessment also include precise longitudinal monitoring of body weights and food intake of rats and the determination of body composition using MRI scans and fat segmentation. We do acknowledge a few limitations of our study. First, the nonprotein nitrogen content in the HMP concentrate was not determined and subtracted from the total nitrogen content to obtain the true protein concentration of the concentrate. However, we speculate that the nonprotein nitrogen content in the HMP concentrate would be negligible because most, if not all, of it would have been filtered out during ultrafiltration and diafiltration owing to its small molecular weight.

There are, of course, many differences between the weanling rat and the human neonate. Among these differences are the higher sulfur-containing amino acid requirements of a rat, which exceed the amounts needed by a human (i.e., due to requirements for fur). It is important to note that protein efficiency ratio values in this study do not accurately reflect the protein digestion and utilization of humans, and that the extrapolation of these findings to neonates warrants necessary caution. Nonetheless, the rat bioassay is still recommended by several regulatory authorities to assess the quality of infant products and formulas (45, 46). This study has verified that the various processing steps used to develop the concentrate have not inadvertently changed its effectiveness as a protein supplement.

In conclusion, an HMP concentrate was feasibly developed from raw donor milk by combining several dairy processing techniques and supported normal growth in weanling rats. The opportunity to use donor milk, which would have otherwise been discarded, would recapture a significant investment made by mothers and milk banks. Although careful systematic research is required, an HMP concentrate may help overcome the limitations of bovine milk–based protein modulars for VLBW infants and provide an important source of bioactive components including lysozyme, lactoferrin, and oligosaccharides.

Notes

Supported by Canadian Institutes of Health Research Foundation Grant CIHR FDN #143233 (to DLO, SU, and YP). The graduate student stipend of SS was supported by a Canadian Graduate Scholarship (CGS-Master's Program) and an Ontario Graduate Scholarship. The sources of support had no role in the design or conduct of the research study, statistical analysis, data interpretation, or writing of the manuscript.

Author disclosures: DLO is a member of The Journal of Nutrition’s Editorial Board and Chair of the Advisory Board of the Rogers Hixon Ontario Human Milk Bank (unpaid positions). SU is the Medical Director of the Rogers Hixon Ontario Human Milk Bank. All other authors report no conflicts of interest.

Supplemental Figures 1 and 2 and Supplemental Table 1 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: BWPI, bovine whey protein isolate; FOV, field of view; HHP, high hydrostatic pressure; HMP, human milk–based protein; VLBW, very low birth weight.

Data Availability

The data that support the findings of this study are available from the corresponding author, [DLO], upon request.