Linking variation in the casein fraction and salt composition to casein micelle size in milk of Dutch dairy goats

The casein composition, salt composition, and micelle size varies substantially between milk samples of individual animals. In goats, the links between those casein characteristics are unknown and could provide useful insights into goat casein micelle structure. In this study, the casein and salt composition of 42 individual Dutch goats from 17 farms was studied and linked to casein micelle size. Micelle size, the proportions of individual caseins, and protein content were associated with one another. Milk with smaller casein micelles was higher in protein content, salt content, and proportion of α s1 -CN, but lower in α s2 -CN and β-CN. The higher salt content in milk with small casein micelles was mainly attributed to a higher protein content, but changes in casein composition might additionally contribute to differences in mineralization. The nonsedimentable casein content in goat milk correlated with nonsedimentable fractions of β-CN and κ-CN and was independent of micelle size. Between large and small casein micelles, goat casein micelles showed more differences in casein and salt composition than bovine micelles, indicating differences in internal structure. Nevertheless, the casein mineralization in goat milk was similar to casein mineralization in bovine milk, indicating that mineralization of casein micelles follows a general principle. These results can help to better understand how composition and micelle structure in goat milk are related to each other, which may be useful to improve processing and product properties of goat milk in the future.


INTRODUCTION
Variation in casein micelle composition and micelle size can influence various technological properties of milk, such as heat aggregation or coagulation behavior (Pesic et al., 2016;Hewa Nadugala et al., 2022).As reviewed by Huppertz et al. (2018) and Horne (2020), a large body of scientific literature therefore focuses on casein composition and casein micelle structure, size, and stability (Huppertz et al., 2018;Horne, 2020).Because cattle are responsible for ~83% of the world's dairy production, casein micelles haven been extensively studied in bovine milk, but little attention has been given to the properties of caseins and casein micelles and their variation in other dairy animals, such as goats (Faye and Konuspayeva, 2012).In recent years, interest in goat milk has increased (Miller and Lu, 2019).Therefore, studying goat casein micelles has become more relevant, as this can lead to a better understanding of differences in processing behavior and product properties, such as heat stability or coagulation properties, between bovine and goat milk.
In addition to differences between species, the relative casein composition can also differ greatly between individual goats (Pierre et al., 1999;Alichanidis et al., 2016).
In goats especially, the α s1 -CN fraction varies extensively due to a high degree of genetic polymorphism, which includes deleterious variants (internal deletion of part of the amino acids sequence) and null variants (Mora-Gutierrez et al., 1991;Martin et al., 2002;Selvaggi et al., 2014).These polymorphisms lead to differences in the amount of α s1 -CN (Grosclaude and Martin, 1997).Differences in the α s1 -CN fraction have furthermore been reported to correlate with protein content (Remeuf, 1993;Barbieri et al., 1995) and casein micelle size in goat milk (Remeuf, 1993;Pierre et al., 1998;Tziboula and Horne, 1999).It is therefore not surprising that earlier studies on goat casein composition and micelle size focused on α s1 -CN, especially on milk samples lacking α s1 -CN (Pierre et al., 1998;Devold et al., 2011;Inglingstad et al., 2014).However, it is not clear how other factors affecting casein composition relate to casein micelle structure in goat milk.Because κ-CN, the only glycosylated casein, is located at the outside of the casein micelle, κ-CN is believed to affect casein micelle size (Dalgleish et al., 1989;Day et al., 2015).However, there is no general agreement in studies on bovine milk and goat milk about whether a higher fraction of κ-CN increases, reduces, or has no effect on the casein micelle size (Sullivan et al., 1959;Pierre et al., 1998;O'Connell and Fox, 2000;de Kruif and Huppertz, 2012;Inglingstad et al., 2014).In addition, the level of glycosylation of κ-CN may also affect casein micelle size (Bijl et al., 2014).The level of glycosylation of goat milk κ-CN (36%; Javier Moreno et al., 2001) is lower compared with bovine milk (60%; Vreeman et al., 1986), but it is unknown whether variation within the glycosylation of goat κ-CN influences micelle size.More research is therefore necessary to better understand how the different casein fractions relate to micellar properties in goat milk.
Even though most casein is incorporated into the casein micelles, a small fraction is present in the serum phase, which can be measured as nonsedimentable casein by centrifugation techniques.In bovine milk, variation in casein composition has been associated with differences in nonsedimentable casein (Huppertz et al., 2021).Previous research on goat milk indicates that ~10% of β-CN is in the serum phase (O'Connor and Fox, 1973).However, little is known about the extent to which caseins other than β-CN contribute to the nonsedimentable casein fraction in goat milk, and to what extent the nonsedimentable casein fraction and composition varies among individual goats.As the nonsedimentable fraction of casein influences the stabilization of interfaces in emulsion and foam (Zhou et al., 2022), this knowledge can be helpful to better understand interfacial properties of goat milk casein.
In addition to casein composition, other properties play a crucial role in the structure and stability of casein micelles.One factor of particular interest is the salt composition of milk and casein micelles.The major salt fractions in bovine milk and goat milk are represented by the cations calcium, magnesium, potassium, and sodium, as well as the anions phosphate, citrate, and chloride (Jenness, 1980;Holt, 1985;de la Fuente et al., 1997a).These salts are present in milk in various forms, either dissolved as free ions or ion pairs in milk serum or dispersed in colloidal calcium phosphate nanoclusters (CPN; de Kruif et al., 2012).These nanoclusters, which also contain fractions of magnesium and citrate, are key elements for stabilizing the substructures of casein micelles by binding to the casein phosphate groups (de Kruif and Holt, 2003;Holt et al., 2013;Huppertz et al., 2017).Differences in salt concentrations were therefore suggested to correlate with nonsedimentable casein in bovine milk (Huppertz and Lambers, 2020).
The salt concentration in milk can vary greatly between different species (Holt and Jenness, 1984;Holt, 2011).Previous research on the salt composition in goat milk found that the micellar fraction is similar to bovine milk, with ~67% Ca, 34% Mg, and 62% P located in the micelle (O'Connor and Fox, 1977;de la Fuente et al., 1997b;Little and Holt, 2004).However, the salt composition differs between individual animals of the same species, for instance depending on the casein content (Holt and Jenness, 1984;Bijl et al., 2013).Previous research on goat milk assumed that salt balances, as well as casein mineralization for casein micelles as observed in individual bovine milk, also apply to goat milk.However, evidence is still lacking, as studies on goat milk are often based on bulk milk, limited to a very low sample size, or limited to Ca and P (O'Connor and Fox, 1977;Ormrod et al., 1982;de la Fuente et al., 1997b;Pierre et al., 1998).
Several researchers investigated how differences in salt composition relate to bovine casein micelle size, but they found somewhat contradictory results, including positive, negative, or no correlation between micelle size and salt composition (Devold et al., 2000;Glantz et al., 2010;Bijl et al., 2014;Huppertz et al., 2021).In goats, a detailed investigation is still missing on both the variation in salt composition among individual animals and on how salt composition is related to casein fractions and casein micelle size.
This research therefore aims to first examine the variation in casein and salt composition in goat milk, and second to investigate how casein and salt composition relates and differs in milks with different casein micelle size.Additionally, findings are compared with literature on bovine casein micelles to gain a better understanding of which aspects of bovine and goat milk casein micelles show similarities or differences.The casein micelle size was determined on 234 individual samples out of 1,000 goat milk samples from 19 goat farms across the Netherlands.Based on the average micelle size, 42 milk samples were subsequently selected to study variation in casein and salt composition of goat milk with small, medium, or large casein micelles.This study gives new insights into how casein micelle composition and structure in goat milk are related.This knowledge may be useful to improve processing and functional properties related to the casein fraction of goat milk in the future.

Sample Collection and Sample Selection
Goat milk was collected from ~1,000 animals from 19 farms across the Netherlands from May to September 2021 during the morning milking.From each farm, 11 to 14 goats were randomly selected for casein micelle size measurement, and samples were preserved with 0.02% sodium azide.Based on the frequency distribution of average casein micelle size of the individual animals (234 goats in total), samples were grouped into 3 classes: small (<180 nm), medium (180-190 nm), and large casein micelles (>190 nm).Samples from 2 farms were excluded from further analysis because of inconsistent results during casein micelle size measurement.Next, we selected a subset of 14 samples per casein micelle class (42 samples in total) from 17 farms (Figure 4), which represent a similar micelle size distribution and mean micelle size (Supplemental Figure S1, see Notes), for further analysis of casein and salt composition.All 42 goats were in their first parity.To minimize possible seasonal variation or farm effects on micelle size and casein and salt composition, each size group consisted of samples of various farms sampled at different time points.

Macronutrient Analysis and Sample Preparation
The fat, protein, and lactose content of fresh whole milk was determined at an external Dutch laboratory for milk control (Qlip, Zutphen, the Netherlands) by Fourier-transform infrared spectroscopy.Skim milk was prepared by centrifugation (3,000 × g, 20°C, 30 min).The pH was determined at 20°C on skim milk.To separate nonsedimentable casein and soluble salts, skim milk ultracentrifugate was obtained by ultracentrifugation of skim milk (100,000 × g, 20°C, 90 min) on a Beckman Coulter L-60 ultracentrifuge with a 70 Ti rotor (Beckman Coulter GmbH) and stored at −20°C.Skim milk samples that were used for particle size measurements on the next day were preserved with 0.02% sodium azide (Merck Life Sciences N.V.) and stored at room temperature.The remaining skim milk was stored at −80°C until further analysis.

Casein Micelle Size Determination
To determine the average casein micelle size, skim milk samples were diluted 100 times in Milli-Q water and subsequently filtered by using a 1.2-µm glass fiber membrane syringe filter (Phenomenex).Measurements were performed immediately after filtration of skim milk by using a Zetasizer Ultra (Malvern Panalytical Ltd.).Samples were measured at 22°C by using a scattering angle of 173° and a refractive index of 1.35 for the dispersant and 1.57 for the casein micelle (Bijl et al., 2014).For each milk sample, 2 samples were prepared, which were measured in triplicate with 5 sub-runs per measurement.
For the analysis of anions, the phosphate, citrate, and Cl content in skim milk and skim milk ultracentrifugate was determined by diluting 500 times and subsequently filtering using a 0.45-µm cellulose acetate syringe filter (Phenomenex).Samples were analyzed in duplicate by ion exchange chromatography (Gaucheron et al., 1996).
The micellar salt content was determined by subtracting the salt content of skim milk ultracentrifugate from salt content in skim milk.The salt content for the skim milk ultracentrifugate was corrected by a factor (f = 0.95), which was previously determined based on the hydration level of the casein pellet in goat bulk milk.

Protein Composition
Reversed-phase HPLC was used to determine casein composition in goat skim milk and skim milk ultracentrifugate.For protein determination, the method described by Miranda et al. (2020) on bovine milk was transferred and slightly modified to improve the separation of goat milk proteins.First, 100 µL of samples were prepared in 1 mL 0.1 M Bis-Tris buffer, 8 M urea, 5.37 mM sodium citrate, and 19.5 mM dithiothreitol and incubated for 1 h at room temperature.After incubation, the samples were centrifuged at 16,000 × g for 5 min and subsequently filtered with a 0.2-µm syringe filter (Phenomenex).Proteins were separated on a Discovery BIO Wide Pore column C5 (150 × 2.1 mm, 300 Å; Supelco, Sigma-Aldrich, St. Louis, MO).Elution was achieved using a multistep gradient of solvent A (0.05% trifluoroacetic acid [TFA] in water) and solvent B (0.05% TFA in acetonitrile): 30% to 30.5% B, 4 min, 30.5% to 34% B in 9.2 min, 34% to 37% B in 1 min, 37% to 42% B in 19 min, 42% to 95% B in 0.1 min, isocratic elution at 95% B for 5 min, return to start condition in 0.1 min, and recalibration at 30% B for 10 min.The first elution step was performed at a flow rate of 0.16 mL/min, followed by 0.2 mL/min for subsequent steps.The column oven was set to 42°C.
The relative percentage of individual caseins was estimated with the Chromeleon 7 software (v.7.2.10; Thermo Scientific, Waltham, MA) by dividing the integrated peak area by the total integrated peak area of intact whey and casein.Peaks were previously assigned by liquid chromatography coupled with electrospray ionization quadrupole time-of-flight mass spectrometry (Shimadzu, Japan) by comparing the most the abundant masses to an in-house database with theoretical masses of casein, α-lactalbumin, and β-lactoglobulin.An example of peak assignment of individual goat milk is given in Figure 1.Nonsedimentable casein content was calculated by subtracting the corresponding peak areas of skim milk ultracentrifugate from the peak areas in corresponding skim milk samples.The casein content of individual samples was estimated using the relative casein content based on the integrated peak area and the protein content of corresponding samples.

Statistical Analysis
The software R (v.4.2.1; https: / / www .r-project .org/ ) was used for statistical analysis.Values are reported as mean ± SD.Parameters for which Pearson correlation coefficients (Supplemental Figure S2) were less than −0.50 and greater than 0.50 were considered to indicate significant correlations.Only these data were selected to perform a simple linear regression analysis subsequently.Significance in protein content, casein composition, micelle size, and salt composition of different micelle size groups was determined by one-way ANOVA followed by Tukey's honestly significant difference test, or by Kruskal-Wallis test followed by Dunn's multiple comparison test, with differences considered significant when P < 0.05.

Casein Composition
The relative casein and nonsedimentable casein composition for 42 goats was determined by HPLC, as presented in Table 1.
All 42 milk samples did contain the 4 caseins α s1 -CN, α s2 -CN, β-CN, and κ-CN.As expected, α s1 -CN showed the largest variation among all caseins, ranging from 5.64% to 26.68% of total casein, whereas κ-CN showed the smallest variation, ranging from 9.65% to 19.25% of total casein.Linear regression analysis between individual casein fractions (Figure 2A) showed that the proportion of α s1 -CN is negatively related to the proportions of β-CN and α s2 -CN in goat milk, and to a smaller extent with the proportion of κ-CN (Figure 2A), which is in agreement with Pierre et al. (1998).This indicates that a decrease in the proportion of α s1 -CN is mainly compensated for by an increase in the β-CN and α s2 -CN fraction.Substantial variation was also observed for the glycosylated fraction of κ-CN when expressed as a percentage of the total κ-CN, ranging from 21.21% to 49.45% of total κ-CN.On average, the glycosylated fraction accounted for ~35% of total κ-CN, which is in line with previous findings for bulk goat milk (Javier Moreno et al., 2001).
Regarding the nonsedimentable casein content in goat milk, 3.78% to 15.77% of the total casein in goat milk was found to be present in the serum phase.It is interesting to note that the average content, as well as the variation of nonsedimentable casein observed in this study, was found to be considerably lower than in bovine milk, where 10% to 60% of casein was found to remain in the serum phase (Davies and Law, 1983;Huppertz et al., 2021).Gener- ally, β-CN is considered to be the most soluble casein, as compared with the other caseins, due to its relatively weak association via hydrophobic interactions (Bingham, 1971;Dalgleish and Law, 1988;O'Connell et al., 2003).It is therefore not surprising that the total amount of nonsedimentable casein in goat milk was positively correlated with the nonsedimentable fraction of β-CN (Figure 2B).In addition to β-CN, total nonsedimentable casein was also positively correlated with the nonsedimentable fraction of κ-CN.Because κ-CN is located at the outside of casein micelles, it may be released from the casein micelles relatively easily in comparison to α s -CN.The fact that α s -CN are more phosphorylated and thus able to bind more calcium phosphate than κ-CN may additionally contribute to the α S -CN being more strongly attached to the casein micelle than κ-CN.In contrast to  bovine milk (Huppertz et al., 2021), the nonsedimentable fractions of α s2 -CN and α s1 -CN in goat milk were not correlated with the nonsedimentable fractions of other caseins.It thus appears that goat and bovine milk do not only differ in the overall casein composition, but also in the percentage of individual nonsedimentable caseins.

Salt Composition
Next to the casein composition, the salt composition between 42 individual goat milk samples was determined by using ICP-OES and ion exchange chromatography and compared with what is known on bovine milk casein mineralization.The contents of total salts in goat skim milk as well as the contents of the micellar salts are presented in Table 2.We found that ~70% Ca, 55% P, and 35% Mg were present as micellar fractions.This is in line with expectations based on goat bulk milk from other studies with a small sample size (Holt and Jenness, 1984;de la Fuente et al., 1997b;Park et al., 2007).Additionally, the SD values in Table 2 indicate that the salt composition varies substantially between goats.
Relations between salts and other milk components of individual goat skim milk were further analyzed by linear regression analysis.These trends, presented in Figure 3, were clustered based on the correlations discussed by Holt (1985) for bovine milk.The first cluster consisted of the relation of protein content and multivalent salts (Figure 3A).Total Ca, P, and Mg of individual samples showed a positive relation with protein content.Micellar contents of Ca, P, and Mg showed a similar relation with protein content (Supplemental Figure S2).This relation between protein content and salt content can be explained by the fact that milk with a higher protein content also contains more caseins, which can therefore bind a higher amount of salts in milk via casein phosphate sites.Furthermore, total Ca, P, and Mg showed positive correlations with their micellar contents (Supplemental Figure S2).Strong positive correlations were also found among the micellar contents of Ca, P, and Mg.Despite considerable variation in mineralization, the Ca/P ratio (1.18 ± 0.06) and the (Ca + Mg)/P ratio (1.29 ± 0.05) of the micellar fraction were quite constant.This indicates that these salts are present at a fixed ratio in CPN in goat milk, as presented on bovine milk, even though values obtained in this study on goat milk appear to be slightly lower (Holt, 1982;Dalgleish and Law, 1989).
Although multivalent ions are correlated with the protein content in milk, the second correlation cluster consists of the relation between monovalent ions and lactose in milk (Figure 3B), as they are together involved in maintaining the osmotic pressure of milk.In the present study, the lactose content was negatively correlated with Cl and K, but not with Na.Such negative correlations between lactose and monovalent ions have previously been described in bovine milk (White and Davies, 1958;Bijl et al., 2013).
Previous research has assumed that the salt balance and casein mineralization, as observed in bovine milk, are applicable to the milk of other species, including goat milk.However, clear evidence on this point is still missing.Earlier studies have indicated an interspecies correlation between, for instance, the Ca and P contents in the milk of different dairy species (Holt, 2011;Holt and Jenness, 1984;O'Connor and Fox, 1977).However, studies on the intraspecies variation in milk salts are rare.In this study on the variation in salt composition in goat milk, similar correlations have been obtained as compared with bovine milk (Bijl et al., 2013).This therefore indicates that the inter-and intraspecies variation in salt composition are connected, following similar trends.Hence, these results indicate that casein mineralization might follow a general principle applicable to casein micelles of different dairy species.

Casein Composition in Casein Micelles of Different Size
As goat milk showed considerable variation in salt and casein composition, it was investigated whether this composition differed between milks varying in micelle size.The frequency distribution of average casein mi-  celle of the 42 milk samples is shown in Figure 4.The mean micelle size (187.33 ± 13.17 nm) was generally found to be lower than that reported previously for goat milk, where the mean size ranged typically from 200 to 280 nm (Pierre et al., 1995;Tziboula and Horne, 1999;Inglingstad et al., 2014).Our data therefore also suggest that goat casein micelle size might not necessarily vary in comparison to the mean size of bovine casein micelles (~190 nm; Remeuf and Lenoir, 1986;Pierre et al., 1995;Bijl et al., 2014).
These contrasting findings may be due to a high frequency of deleterious or null variants of α s1 -CN in previous studies.Both have been associated with an increased casein micelle size (Pierre et al., 1995;Tziboula and Horne, 1999;Inglingstad et al., 2014).In this study, the specific casein variants have not been investigated, but from HPLC analysis, it is apparent that all goat milk samples contained a considerable amount of α s1 -CN, though with substantial variation (Table 1).Furthermore, differences in sample treatment before measurement could explain differences observed for goat casein micelle size.It was noted in this study that larger particles, such as possible remaining fat, could cause significant interference during micelle size measurements, resulting in nonrepeatable results.Without using a syringe filter before micelle size determination, a multimodal size distribution was observed with a second peak around the size typical for goat milk fat globules (8.5 µm; Attaie and Richter, 2000), which caused a shift of the mean micelle size toward higher values (Supplemental Figure S3).By using a syringe filter, the number of these larger particles could be reduced.
To compare whether the casein and salt composition of goat milk with different micelle sizes varied, milk samples were divided into 3 groups consisting of 14 samples per group, containing either small micelles (<180 nm), medium micelles (180-190 nm), or large micelles (>190nm).The comparison between small and large micelles regarding their macronutrients, casein composition, and as salt composition is presented in Table 3.Before discussing the casein and salt composition, it should be noted that milk with smaller casein micelles showed significant differences in its macronutrient composition, especially in protein content.Milk with smaller casein micelles showed a higher protein content in comparison to milk with larger casein micelles.Not found not only a difference in protein content between groups, but also a negative correlation between protein content and micelle size among all samples (Figure 5A).In bovine milk with differing micelle sizes, differences in protein content were not observed (Bijl et al., 2014;Day et al., 2015).
When comparing the relative casein composition between small and large casein micelle groups, a difference in the proportion of α s1 -CN, α s2 -CN, and β-CN was found (Table 3).Linear regression analysis revealed that the fraction of α s1 -CN was negatively correlated with casein micelle size (Figure 5B).The α s2 -CN content was positively associated with micelle size.The β-CN fraction was not strongly associated with micelle size, regardless of the differences between the micelle size groups.In contrast to the other caseins, differences in κ-CN were not found between milks varying in micelle size, regardless of the negative correlation observed between α s1 -CN and κ-CN (Figure 2).These results therefore indicate that changes in micelle size in goat milk mainly related to changes in the α s -CN and β-CN fractions, especially α s1 -CN.Moreover, these results support previous findings on studies on larger sets of individual cows and goats, that the percentage of κ-CN might not be the main factor contributing to differences in casein micelle size between individual animals (Pierre et al., 1998;de Kruif and Huppertz, 2012;Bijl et al., 2014).Furthermore, even though the glycosylation level of κ-CN varied extensively between individual animals, no differences in glycosylated κ-CN between the groups were found.In bovine milk, a higher glycosylation level was correlated with a smaller micelle size (Bijl et al., 2014;Day et al., 2015), which was not the case for goat milk in this study.Even though κ-CN might play a role in determining micelle growth, it seems thus that in goat milk the other caseins, and especially α s1 -CN, determine the mean micelle size (Pierre et al., 1998;Tziboula and Horne, 1999;Inglingstad et al., 2014).
Moreover, the α s1 -CN fraction was not only associated with micelle size, but also with the protein content in milk, which itself showed a linear relation with micelle size (Figure 5).In contrast, α s2 -CN and β-CN were negatively associated with protein content, though to a lower extent than α s1 -CN (Supplemental Figure S2).The content of α s1 -CN has also previously been associated with protein content (Grosclaude and Martin, 1997;Tziboula and Horne, 1999).We therefore hypothesize that changes in protein content and casein micelle size can be attributed to changes in the α s1 -CN fraction, whereas the protein content itself does not directly influence the micelle size.
Both proportion of α s1 -CN and protein content seem to depend on genetic variants of α s1 -CN (Grosclaude and Martin, 1997;Selvaggi et al., 2014).The null variants and defective variants of α s1 -CN in particular have previously been described to be associated with a larger casein micelle size and lower protein content (Remeuf, 1993;Pierre et al., 1995;Grosclaude and Martin, 1997).Samples from the 42 goats in this study, however, all contained considerable amount of α s1 -CN, indicating that significant difference in casein micelle size and protein content in goat milk exist also at the presence of all 4 caseins.Nevertheless, it can be hypothesized that the differences in protein content and casein fractions between micelle size groups observed in the Dutch goat population are a result of differences in genetic variants of the individual caseins, but this needs further investigation.
Overall, the results suggest that differences in mean casein micelle size between individual goats are mainly a results of changes in relative compositions of the α s -CN and β-CN fractions.The differences in casein composition between casein micelles of different sizes seem to be a characteristic of goat milk caseins.In bovine milk, differences in the casein fraction between casein micelles differing in size, if observed, were rather attributed to the κ-CN fraction (Bijl et al., 2014;Day et al., 2015).Only 1 study on individual cows also observed differences in α s1 -CN, but in contrast to goat milk micelles, larger casein micelles in bovine milk contained a higher amount of α s1 -CN (Day et al., 2015).How the changes in casein composition may cause changes in goat milk micelle size, however, still remains unclear.An explanation may lie within the structure of individual caseins that differ for instance in amino acid composition, number of structure-breaking proline residues, and number of phosphate groups.In addition, caseins are highly polymorphic; in goats especially, α s1 -CN is known to exhibit extensive variation.Changes in the proportions of caseins and differences in genetic variants might therefore affect how these caseins interact with each other to form casein micelles.

Salt Composition in Casein Micelles of Different Size
In addition to macronutrient and casein composition, the salt composition between the micelle size groups was compared.Overall, goat milk with smaller casein micelles had a higher total and micellar content of Ca, P, inorganic phosphate, and Mg (Supplemental Table S1).This can be explained by the higher protein content in samples with smaller casein micelles and the positive correlation between the content of protein and several salts (Figure 3A).Because the protein content correlated with salt composition, the contents of total, micellar, and soluble salts presented in Table 3 were corrected for casein content (mmol•g −1 casein) to determine whether differences between the micelle size groups at similar casein contents existed.After this correction for casein content, goat milk samples with small casein micelles had a higher total P, inorganic phosphate, and citrate content, as well as a higher micellar P, inorganic phosphate, and Mg content than large casein micelles.Total and micellar Ca content between the different size groups did not differ significantly, probably due to large standard deviations, regardless of the strong relation between P and Ca observed (Figure 3A).In bovine milk, differences in salt composition between milk with different micelle sizes have not been reported; however, the bovine samples did also not vary significantly in their protein content (Bijl et al., 2014;Day et al., 2015).In this study, more compositional differences were thus observed in goat milk compared with bovine milk for different micelle size groups.This indicates differences in the contribution of individual caseins to casein micelle formation and differences in internal micelle structure.However, this topic needs further investigation.
The differences in salt composition between goat casein micelles differing in size suggest that these might be a result of differences in casein composition at a similar casein content.However, total and micellar salts were not correlated with the proportion of the casein fractions or nonsedimentable contents of individual caseins.The only exception is that a negative correlation was observed for nonsedimentable β-CN with micellar Ca, P, and inorganic phosphate at similar casein contents.Even though the proportions of casein were not correlated with changes in mineralization at similar protein content, one has to keep in mind that proportions of β-CN, α s1 -CN, and α s2 -CN differed between micelle size groups.When considering the ratio of the most abundant casein (β-CN) and the most variable casein fraction in goat milk (α s1 -CN), the α s1 -CN/β-CN ratio differed greatly between small casein micelles (0.40 ± 0.08) and large casein micelles (0.23 ± 0.08).Changes in relative casein composition could also affect the capacity of casein micelles to bind calcium phosphate, which might be related to the different phosphorylation levels of the casein fractions.In goat milk, α s1 -CN has 5 clustered phosphorylation sites, which are closely located to one another, and an overall higher phosphorylation level than β-CN, which has 4 clustered phosphorylation sites (Martin et al., 2013).Milks high in α s1 -CN might therefore bind more inorganic salts at a similar casein content.This might explain the higher levels of P and inorganic phosphate in milk with small casein micelles when values were corrected for casein content.

CONCLUSIONS
The casein and salt composition of 42 goats was analyzed to explore their variation and to investigate their link to casein micelles size.Goat milk showed an extensive variation in casein composition and mineralization.The results indicate that changes in micelle size are mainly a result of changes in the casein composition.Compositional differences between micelle size groups seem to relate to an interplay of casein composition, protein content, and salt composition.Milk with smaller micelles had a higher α s1 -CN / β-CN ratio and protein content; the higher salt content did not seem to be directly linked to micelle size, but was rather attributed to a higher protein content.Compared to bovine milk, goats showed more compositional differences between micelle size groups, indicating differences in contribution of individual caseins to micelle formation and structure.Future studies on casein variants and their post-translational modifications could give further insights into the variations in micelle size, casein, and salt composition among goats.

NOTES
This work was financially supported by the Ausnutria B.V. (Zwolle, the Netherlands).The authors thank the goat farmers for their support and enthusiasm in collecting and providing the milk samples.Supplemental material for this article is available at https: / / doi .org/ 10 .18174/644583.This research adhered to the animal experimentation guidelines of Wageningen University & Research.Aspects concerning the welfare and treatment of animals in this study did not require approval from an Ethics Committee.The authors have not stated any conflicts of interest.

Figure 1 .
Figure 1.Peak assignment of major proteins (CN, α-LA, β-LG) based on the HPLC chromatogram of an individual goat milk sample.

Figure 3 .
Figure 3. Linear regression analysis of total and micellar salt, protein (A), and lactose (B) content for 42 individual goat skim milk samples.
Figure 4. Distribution of average casein micelle size (nm) of 42 Dutch goat milk samples.

Figure 5 .
Figure 5. Linear regression analysis between protein content, casein micelle size, and α S1 -casein (percentage of total casein) of 42 goat milk samples.

Table 1 .
Breunig et al.: VARIATION IN GOAT MILK CASEIN FRACTION Average values of casein (CN) composition and nonsedimentable casein of corresponding caseins (%) in the milk of 42 Dutch goats

Table 2 .
Breunig et al.: VARIATION IN GOAT MILK CASEIN FRACTION Average values of total and micellar salt content (mmol/L) of 42 Dutch goat skim milk samples Breunig et al.: VARIATION IN GOAT MILK CASEIN FRACTION

Table 3 .
Breunig et al.: VARIATION IN GOAT MILK CASEIN FRACTION Comparison of total, colloidal and nonsedimentable salt correct for casein content, and relative casein composition in goat skim milk with small, medium, and large casein micelles (n = 42) Mean values in the same row with different superscripts differ by P < 0.05 in multiple comparison test.
1 P-values indicate statistical significance of ANOVA or Kruskal-Wallis test.