Cheese yield is largely determined by the concentrations of protein, particularly casein, and fat in milk (Lawrence, Reference Lawrence1993). An important indicator of milk suitable for cheese production would therefore be casein number (weight percentage of casein to total protein). It is, however, possible that rather than regarding caseins as a homogenous group, some of the caseins may play a more significant role for the cheese yield, i.e. there may be room for improvement of the casein composition. Rather than analysing concentration of the various milk proteins, many studies have been looking for associations between polymorphisms in the milk protein coding genes and milk coagulation, thereby addressing the impact of structural variation of allelic milk protein variants. Studies have shown that selection for genetic variants of milk proteins could be an option to change the protein composition of milk (Ikonen et al. Reference Ikonen, Ojala and Syväoja1997; Lodes et al. Reference Lodes, Buchberger, Krause, Aumann and Klostermeyer1997; Bobe et al. Reference Bobe, Beitz, Freeman and Lindberg1999; Hallén et al. Reference Hallén, Wedholm, Andrén and Lundén2008), thereby possibly obtaining improved processing properties resulting in a higher dairy product yield and quality (Rahali & Ménard, Reference Rahali and Ménard1991; Boland & Hill, Reference Boland and Hill2001; Ikonen et al. Reference Ikonen and Ruottinen1999).
Several of the studies on the detailed protein composition and cheese making potential of milk have focused on rheological properties such as curd firmness (Storry et al. Reference Storry, Grandison, Millard, Owen and Ford1983; van den Berg et al. Reference van den Berg, Escher, de Koning and Bovenhuis1992; Ikonen et al. Reference Ikonen, Ojala and Syväoja1997; Jõudu et al. Reference Jõudu, Henno, Kaart, Püssa and Kärt2008), whereas fewer have also related the protein composition to actual cheese yield (Ikonen et al. Reference Ikonen and Ruottinen1999; Auldist et al. Reference Auldist, Johnston, White, Fitzsimons and Boland2004; Wedholm et al. Reference Wedholm, Larsen, Lindmark-Månsson, Karlsson and Andrén2006). Curd firmness at cutting has been positively associated with cheese yield (Bynum & Olson, Reference Bynum and Olson1982; Riddell-Lawrence & Hicks, Reference Riddell-Lawrence and Hicks1989), but in practice gel firmness may have minor consequences for cheese yield as long as the coagulation process is relatively consistent (Lucey & Kelly, Reference Lucey and Kelly1994). Hurtaud et al. (Reference Hurtaud, Rulquin, Delaite and Vérité1995) showed that actual cheese yield (Camembert) was more accurately predicted by laboratory scale cheese yield than through coagulation measures obtained by Formagraph. Being the main constituents of cheese, measures of caseins lost to the whey may be more relevant for actual cheese yield than rheological properties.
The present work studied how the milk protein composition and the genetic polymorphism of milk proteins were associated with the retention of casein in curd at chymosin-induced coagulation after syneresis and simulated pressing of the curd.
Materials and Methods
Milk samples
Individual morning milk samples were collected during five non-consecutive months between November 2003 and March 2005 from 110 cows in the experimental dairy herd of the Swedish University of Agricultural Sciences (Jälla, Uppsala, Sweden). Seventy cows were of the Swedish Red breed (SRB) and 40 of the Swedish Holstein breed (SLB). Of the SRB cows, 37 belonged to a selection line for high milk fat percentage (SRB H) and 33 to a selection line for low milk fat percentage (SRB L), but with equivalent total milk energy production in both lines. One milk sample from each cow was analysed. To reduce effects of lactation stage or mastitis, the sample was collected when the cow was in lactation week 10–35 and an upper limit was set for somatic cell count (SCC ⩽250,000). Samples were cooled directly after milking and kept at 4°C. Fat, protein and lactose concentration was analysed by mid-infrared spectroscopy (MilkoScan FT120, A/S Foss Electric, Hillerød, Denmark) and SCC by flow cytometry (Fossomatic 5200, A/S Foss Electric). Protein content and composition was analysed by reversed phase HPLC (RP-HPLC). Information regarding morning milk yield, parity number, lactation week and time of sampling was collected for each sample.
Typing for protein variants
Typing for variants of the β-CN (A1, A2, A3, B) and κ-CN (A, B, E) genes was carried out by Pyrosequencing™ (Biotage AB, Uppsala, Sweden) as previously described by Hallén et al. (Reference Hallén, Allmere, Näslund, Andrén and Lundén2007).
Milk sample and chymosin preparation
On the day of sample collection, fresh milk samples were pre-warmed (30°C, 30 min) and defatted by centrifugation (2465 g, 3°C, 25 min) (Centrifuge 5810R, Eppendorf AG, Hamburg, Germany). Samples were then kept refrigerated. Chromatographically pure chymosin (Andrén et al. Reference Andrén, Björck and Claesson1980), 174,000 International Milk Clotting Units/g, was used to prepare a working solution of 1·5 mg chymosin/ml in a 0·1 m-phosphate buffer (pH 5·7).
Curd and whey preparation
Milk coagulation was performed the day after sample collection, using a procedure similar to Hurtaud et al. (Reference Hurtaud, Rulquin, Delaite and Vérité1995) and Othmane et al. (Reference Othmane, Carriedo, de la Fuente Crespo and San Primitivo2002). Defatted milk samples (10 ml) were incubated in test tubes in a shaking water bath (30°C, 30 min). Chymosin solution (25 μl) was added to each sample, after which they were vortexed and incubated for another 30 min. The coagulum was vertically cut in four equally sized sections, using a four-edged knife specifically made to fit the tubes. After another 30 min incubation the tubes were removed from the water bath and a 300 μl sample of the expelled whey was withdrawn by pipette (Whey1). Pressing of the curd was simulated by centrifugation at room temperature (1258 g, 15 min) (Centrifuge 5810R, Eppendorf AG). Expelled whey was decanted by a standardized protocol and measured by weighing (Whey2). Fresh curd yield (Yf) was calculated as the weight difference between the initial milk sample and the expelled whey, and expressed as grams of curd per 100 g milk. The repeatability was 0·8 for the Whey2 measurement, calculated according to International Standardization Organization guidelines (ISO, 2005) at an initial trial conducted on 30 samples analysed in triplicate. Samples of defatted milk, Whey1 and Whey2 were stored at −80°C pending analysis of protein composition by RP-HPLC.
HPLC analysis
Skim milk and whey samples were analysed for milk protein composition by RP-HPLC. The method, including equipment, reagents and buffers, was as described by Hallén et al. (Reference Hallén, Wedholm, Andrén and Lundén2008). Concentration of ‘major proteins’ was calculated as the sum of concentrations of the individual proteins (αs1-CN, αs2-CN, β-CN, κ-CN, β-lg and α-lactalbumin; α-la). Casein (CN) ratio was calculated as the sum of individual concentrations of the analysed caseins (αs1-CN, αs2-CN, β-CN and κ-CN) divided by concentration of ‘major proteins’. Casein concentrations in Whey1 and Whey2 (CNwhey1 and CNwhey2) were calculated as the sum of individual concentrations of the analysed caseins in the respective whey fraction. Casein retention in curd (retCN) was calculated as the weight difference between total amount of casein in the initial milk sample and total amount of casein in Whey2 expelled from this milk sample.
Statistical analysis
Effects of milk protein composition of the original milk on CNwhey1, CNwhey2 and Yf were analysed using the Mixed procedure of the statistical software SAS (SAS Institute Inc, Cary, USA). The time of sampling parameter was entered as a random effect. Fixed effects of parity, lactation week, breed, β-lg genotype, β-/κ-CN genotype, milk yield, SCC, initial concentration of fat and lactose were not significant and dropped from the subsequent analyses. Models were also run where the individual protein concentrations were exchanged with concentration of major proteins, total casein, and CN ratio, respectively.
The following statistical model was used:
where:
yijklmnop=CNwhey1 or CNwhey2 or Yf for cow ijklmnop
μ is the general mean
b1, b2, …b6=regression coefficients of yijklmnop on the respective protein concentration in milk
αs1CNi, αs2CNj, βCNk, κCNl, βlgm, αlan=individual protein concentration i/j/k/l/m/n in milk of cow ijklmnop
samplo=random effect of time of sampling o (o=1, 2,…5)
εijklmnop=random residual effect
The above model was also used in the analysis of CNwhey1 and CNwhey2 as categorical traits (no measurable casein in whey=0, casein in whey=1) using the GLIMMIX procedure of SAS (SAS Institute Inc, Cary).
Effects of genetic polymorphism of milk proteins on protein composition of milk were analysed using the general linear model (GLM) procedure (SAS Institute Inc, Cary). Due to the close genetic linkage between the casein loci (Ferretti et al. Reference Ferretti, Leone and Sgaramella1990; Threadgill & Womack, Reference Threadgill and Womack1990), aggregate β-/κ-CN genotype was entered as a fixed effect in the model. Genotypes comprising less than three cows were dropped from the subsequent analysis (see Table 3). The group variable specified cow breed and selection line and consisted of three classes: SRB H, SRB L and SLB. Parity was grouped into four classes; first, second, third, and fourth or higher parity. Fixed effects of milk yield, SCC, fat and lactose concentration were not significant and dropped from the subsequent analyses.
where:
yijklmno=milk protein variable for cow ijklmno
β/κCNi=fixed effect of β-/κ-CN genotype i (i=1, 2, …10; see Table 3)
βlgj=fixed effect of β-lg genotype j (j=AA, AB, BB)
groupk=fixed effect of group k (k=SRB H, SRB L, SLB)
parityl=fixed effect of parity l (l=1, 2, …4)
samplm=fixed effect of time of sampling m (m=1, 2, …5)
b1=regression coefficient of yijklmno on lactation week
lactwkn=lactation week n of cow ijklmno
Results and Discussion
Means and measures of variation for gross composition of milk and for detailed protein composition of milk, Whey1, and Whey2 are given in Table 1. About one third of the Whey1 and Whey2 samples, respectively, contained no measurable amounts of casein and the majority contained <2 g/l, whereas 6% of the Whey1 samples and 3% of the Whey2 samples contained >10 g/l (Fig. 1). Although mean levels of casein were similar in Whey1 and Whey2 (Table 1), a higher proportion of the Whey1 samples contained less than 2 g casein/l compared with Whey2 (90% and 60%, respectively). The casein lost in whey at cheese making has been reported to be around 1 g/kg milk (Lucey & Kelly, Reference Lucey and Kelly1994). In this trial mean casein loss in Whey2 was 1·8 g/kg milk (SD 2·5, range 0–13·9 g/kg). Of the 110 milk samples, four (3·6%) did not aggregate to form a curd within the set time of the coagulation experiment (>1 h). These non-coagulating samples were not included in further calculations.
Distribution of values for casein content of whey after cutting (CNwhey1) and after simulated pressing (CNwhey2), respectively, at chymosin-induced coagulation of individual defatted milk samples (n=106).
Mean composition (with standard deviations) of morning milk samples and of the whey after cutting (Whey1) and after simulated pressing (Whey2), respectively, along with proportion of protein retained in the curd, at coagulating individual defatted milk samples with chymosin
† αs1-CN+αs2-CN+β-CN+κ-CN+β-LG+α-LA
‡ αs1-CN+αs2-CN+β-CN+κ-CN
§ β-LG+α-LA
¶ Non-coagulating samples excluded
†† para-κ-casein
Mean fresh curd yield was 23·9 g/100 g milk (sd 4·3, range 14·5–34·1 g/100 g). Similar results were reported by Othmane et al. (Reference Othmane, Carriedo, de la Fuente Crespo and San Primitivo2002), and Hurtaud et al. (Reference Hurtaud, Rulquin, Delaite and Vérité1995), who also analysed small volumes of milk (26·5% in 10 ml, and 30·6% in 30 ml, respectively). Average protein content of whey was 0·85% (8·48 g/l), equivalent to 27·3% of the protein originally present in the milk (Table 1). These percentages correspond well to the 1% and 30·7%, respectively, reported by Ng-Kwai-Hang et al. (Reference Ng-Kwai-Hang, Politis, Cue and Marziali1989). About 27% of the protein in whey consisted of casein, which can be considered to constitute yield losses. Of the casein and major proteins from the original milk, 93·3% and 72·8%, respectively, was retained in the curd. In the production of cheddar cheese, recoveries of casein ranging from 93 to 99% and of total protein of approximately 74% have been reported (Lucey & Kelly, Reference Lucey and Kelly1994).
Influence of protein composition of milk on CNwhey1 and CNwhey2 is given in Table 2. A higher concentration of κ-CN in milk was associated with lower levels of casein in whey after cutting, CNwhey1, whereas concentrations of the other individual proteins, total casein, or major proteins showed no significant effect. Milk with high κ-CN content has been shown to contain smaller casein micelles compared with milk with low κ-CN content (Niki et al. Reference Niki, Kohyama, Sano and Nishinari1994; Walsh et al. Reference Walsh, Guinee, Reville, Harrington, Murphy, O'Kennedy and Fitzgerald1998). This allows for a more compact and uniform arrangement of the gel network, which may reduce losses in whey by an improved entrapping ability (Nuyts-Petit et al. Reference Nuyts-Petit, Delacroix-Buchet and Vassal1997; Walsh et al. Reference Walsh, Guinee, Reville, Harrington, Murphy, O'Kennedy and Fitzgerald1998). A faster coagulation reaction of samples with a high κ-CN concentration (van den Berg et al. Reference van den Berg, Escher, de Koning and Bovenhuis1992; Nuyts-Petit et al. Reference Nuyts-Petit, Delacroix-Buchet and Vassal1997) and thus firmer curd at cutting might have reduced the losses of casein in Whey1, as previously suggested by Ng-Kwai-Hang et al. (Reference Ng-Kwai-Hang, Politis, Cue and Marziali1989). CNwhey2, i.e. losses after simulated pressing, was negatively associated with CN ratio and positively associated with levels of major proteins and α-la concentration in milk, whereas concentrations of individual caseins or total casein, respectively, were not significant (Table 2). Somewhat contrasting results were reported by Verdier-Metz et al. (Reference Verdier-Metz, Coulon and Pradel2001), who found protein losses in whey to be lower in milk with high protein (and fat) content, although casein content was not analysed. Consequently, according to our results increasing the overall protein content of milk may not have the desired effect, whereas a high proportion of casein to total protein would decrease casein losses in whey. Why κ-CN concentration was found significant for casein loss in Whey1 as described above, but not in Whey2, might have been due to additional factors affecting casein loss as the pressing force was introduced. Samples with CNwhey1 above 2 g/l were found to have CNwhey2 values lower than this (data not shown), suggesting that casein aggregates in the whey were caught in the coagulum during simulated pressing. Further, 50% of the samples with CNwhey1 below 2 g/l exhibited increased CNwhey2 values compared with CNwhey1 (data not shown), indicating weak gels losing casein during simulated pressing. Further analyses would have been necessary to explain these occurrences.
Regression coefficients (±standard error) of casein concentration in whey after cutting (CNwhey1) and after simulated pressing (CNwhey2), and of fresh curd yield (Yf) at chymosin-induced coagulation of defatted milk on milk protein composition
† αs1-CN+αs2-CN+β-CN+κ-CN+β-LG+α-LA
‡ αs1-CN+αs2-CN+β-CN+κ-CN
§ (αs1-CN+αs2-CN+β-CN+κ-CN)/(αs1-CN+αs2-CN+β-CN+κ-CN+β-LG+α-LA)
¶ not significant
* P<0·05; **P<0·01; **P<0·001
Of the protein components in milk, only concentrations of αs1-CN and β-CN were positively associated with Yf (Table 2). This may have been due to the lower accuracy of the HPLC method when analysing small protein fractions, resulting in comparatively large standard errors of the estimates for αs2-CN and κ-CN. Marziali & Ng-Kwai-Hang (Reference Marziali and Ng-Kwai-Hang1986) found αs-CN (αs1-CN+αs2-CN) and β-CN concentrations to have a positive effect on actual cheese yield, whereas Wedholm et al. (Reference Wedholm, Larsen, Lindmark-Månsson, Karlsson and Andrén2006) in addition found a positive effect of κ-CN concentration. It has also been reported that whereas total concentration of casein in milk is positively related to cheese yield, variation in casein composition is not (Christian et al. Reference Christian, Grainger, Sutherland, Mayes, Hannah and Kefford1999). This is supported by the present results, where total casein and major proteins concentrations were positively associated with Yf (Table 2), whereas CN ratio and relative proportions of the different caseins showed no association (data not shown).
Aggregate β-/κ-CN genotype was not found to have effect on Yf, CNwhey1 or CNwhey2. However, β-/κ-CN genotype was associated with concentration of κ-CN in milk (Table 3) (see Hallén et al. Reference Hallén, Wedholm, Andrén and Lundén2008). The BB genotype of β-lg was associated with increased retCN compared with AB in the present study (P<0·01, data not shown). This effect was probably due to the association of β-lg BB with CN ratio (Table 3), since if CN ratio was added to the model, β-lg genotype did not remain significant. Higher cheese yield has been associated with β-lg B in several other studies (Marziali & Ng-Kwai-Hang, Reference Marziali and Ng-Kwai-Hang1986; Wedholm et al. Reference Wedholm, Larsen, Lindmark-Månsson, Karlsson and Andrén2006) possibly through its impact on CN ratio (Rahali & Ménard, Reference Rahali and Ménard1991; van den Berg et al. Reference van den Berg, Escher, de Koning and Bovenhuis1992; Boland & Hill, Reference Boland and Hill2001).
Least squares means (±standard error) for effect of β-lactoglobulin (β-LG) genotype on casein ratio and β-LG concentration, and for effect of aggregate β-/κ-casein (β-/κ-CN) genotype on κ-CN concentration in individual milk samples
† Number of cows
‡ (αs1-CN+αs2-CN+β-CN+κ-CN)/(αs1-CN+αs2-CN+β-CN+κ-CN+β-LG+α-LA)
a, b, c values within column with differing letters in superscript are statistically significant (P<0·05)
Fresh curd yield (Yf) was dependent on amount of casein available for curd formation, reflecting the milk casein content (Table 2), whereas there was no association between casein content of milk and casein content of whey. The retCN parameter, representing amount of curd forming casein, was therefore a better predictor of Yf (R2=0·60). An increased casein retention of 0·1 g/l milk resulted in 1 g/l higher fresh curd yield in this study. There was also a high correlation between retCN and casein content of milk (R2=0·51). The measure of CNwhey2 (casein loss in whey after simulated pressing) showed a weak association with Yf, as milk with a high casein concentration and a large loss of casein in whey could still result in a large yield, whereas milk with a low casein concentration exhibiting only a small loss would also give a small yield.
Aiming to keep the losses of casein in whey to a minimum, milk samples with no (measurable) losses of casein in whey could be considered desirable. In an attempt to find characteristics in the protein composition which might distinguish ‘casein loss in whey’ from ‘no/negligible casein loss in whey’ CNwhey1 and CNwhey2 were analysed as binary traits. The results showed that a higher concentration of κ-CN in milk reduced the risk of casein loss in whey, both after cutting (Whey1) and after simulated pressing (Whey2) (P<0·05, data not shown), whereas the other protein fractions analysed showed no association with casein loss. The result for κ-CN is consistent with previously reported positive effects of κ-CN concentration on milk coagulation and cheese yield (e.g. Storry et al. Reference Storry, Grandison, Millard, Owen and Ford1983; Rahali & Ménard, Reference Rahali and Ménard1991; van den Berg et al. Reference van den Berg, Escher, de Koning and Bovenhuis1992; Ikonen et al. Reference Ikonen, Ojala and Syväoja1997; Walsh et al. Reference Walsh, Guinee, Reville, Harrington, Murphy, O'Kennedy and Fitzgerald1998; Wedholm et al. Reference Wedholm, Larsen, Lindmark-Månsson, Karlsson and Andrén2006) and supports the possibility to select on protein genotype for improving the coagulation and thereby cheese yield potential of milk.
This work was funded by the Swedish Farmers' Foundation for Agricultural Research. Jessica Näslund at the Department of Animal Breeding & Genetics, Swedish University of Agricultural Sciences, is thanked for performing the genotype analyses. Lena Hagenvall and Gudrun Franzén at the same department are thanked for handling the milk sample collection.
