Hostname: page-component-7b9c58cd5d-sk4tg Total loading time: 0 Render date: 2025-03-15T12:21:32.577Z Has data issue: false hasContentIssue false

Prediction efficiency by near-infrared spectroscopy of immunoglobulin G in liquid and dried bovine colostrum samples

Published online by Cambridge University Press:  07 September 2016

M. Jordana Rivero*
Affiliation:
Escuela de Agronomía, Facultad de Recursos Naturales, Universidad Católica de Temuco, Temuco 4780000, Chile Núcleo de Investigación en Producción Alimentaria, Facultad de Recursos Naturales, Universidad Católica de Temuco, Temuco 4780000, Chile Escuela de Graduados, Facultad de Ciencias Agrarias, Universidad Austral de Chile, 5090000 Valdivia, Chile
Daniel Alomar
Affiliation:
Instituto Producción Animal, Facultad de Ciencias Agrarias, Universidad Austral de Chile, 5090000 Valdivia, Chile
Ximena Valderrama
Affiliation:
Instituto de Ciencia Animal, Facultad de Ciencias Veterinarias, Universidad Austral de Chile, 5090000 Valdivia, Chile
Yannick Le Cozler
Affiliation:
AGROCAMPUS-Ouest, UMR1348 Physiology, Environment and Genetics for Animal and Livestock Systems, 35000 Rennes, France INRA, UMR1348 Physiology, Environment and Genetics for Animal and Livestock Systems, 35590 St-Gilles, France
Alejandro Velásquez
Affiliation:
Escuela de Agronomía, Facultad de Recursos Naturales, Universidad Católica de Temuco, Temuco 4780000, Chile Núcleo de Investigación en Producción Alimentaria, Facultad de Recursos Naturales, Universidad Católica de Temuco, Temuco 4780000, Chile
Deborah Haines
Affiliation:
Department of Veterinary Microbiology, Western College of Veterinary Medicine, University of Saskatchewan, Saskatoon S7N 0M3, Canada
*
*For correspondence; e-mail: mjrivero@uct.cl
Rights & Permissions [Opens in a new window]

Abstract

The objective of this study was to compare the prediction efficiency of IgG concentration in bovine colostrum by NIRS, using liquid and dried (Dry-Extract Spectroscopy for Infrared Reflectance, DESIR) samples by transflectance and reflectance modes, respectively. Colostrum samples (157), obtained from 2 commercial Holstein dairy farms, were collected within the first hour after calving and kept at −20 °C until analysis. After thawing and homogenisation, a subsample of 500 mg of liquid colostrum was placed in an aluminium mirror transflectance cell (0·1 mm path length), in duplicate, to collect the spectrum. A glass fiber filter disc was infused with another subsample of 500 mg of colostrum, in duplicate, and dried in a forced-air oven at 60 °C for 20 min. The samples were placed in cells for dry samples to collect the spectra. The spectra in the VIS-NIR region (400–2500 nm) were obtained with a NIRSystems 6500 monochromator. Mathematical treatments, scatter correction treatments and number of cross-validation groups were tested to obtain prediction equations for both techniques. Reference analysis for IgG content was performed by radial immunodiffusion. The DESIR technique showed a higher variation in the spectral regions associated with water absorption bands, compared with liquid samples. The best equation for transflectance method (liquid samples) obtained a higher coefficient of determination for calibration (0·95 vs. 0·94, respectively) and cross validation (0·94 vs. 0·91, respectively), and a lower error of cross validation (9·03 vs. 11·5, respectively) than the best equation for reflectance method (DESIR samples). In final, both methods showed excellent capacity for quantitative analysis, with residual predictive deviations above 3. It is concluded that, regarding accuracy of prediction and time for obtaining results of IgG from bovine colostrum, NIRS analysis of liquid samples (transflectance) is recommended over dried samples (DESIR technique by reflectance).

Type
Research Article
Copyright
Copyright © Proprietors of Journal of Dairy Research 2016 

Colostrum is the first secretion produced after calving and contains many substances important to the health and survival of the newborn including immunoglobulins (Ig, Conneely et al. Reference Conneely, Berry, Sayers, Murphy, Lorenz, Doherty and Kennedy2013). Calves are born agammaglobulinemic, without any measurable circulating IgG or IgM in plasma (Morrill et al. Reference Morrill, Conrad, Polo, Lago, Campbell, Quigley and Tyler2012) as these macromolecules are not able to cross the placental barrier. By absorbing Ig from colostrum, calves acquire passive immunity (Morrill et al. Reference Morrill, Conrad, Polo, Lago, Campbell, Quigley and Tyler2012) and this process is strongly related to the immunoglobulin G (IgG) concentration of the maternal colostrum (Nocek et al. Reference Nocek, Braund and Warner1984). Therefore, the concentration of IgG in bovine colostrum is considered the hallmark for evaluating colostrum quality (Godden, Reference Godden2008).

Concentration of IgG is extremely variable in dairy cattle but good-quality colostrum is defined as colostrum that has an IgG concentration above 50 g/L (McGuirk & Collins, Reference McGuirk and Collins2004). Hence, an accurate method to measure IgG content is essential to assess colostrum quality. Techniques developed to measure or estimate IgG concentration include radial immunodiffusion, developed by Mancini et al. (Reference Mancini, Carbonara and Heremans1965) and considered to be the gold standard (Bielmann et al. Reference Bielmann, Gillan, Perkins, Skidmore, Godden and Leslie2010), weight of the first milking (Chigerwe et al. Reference Chigerwe, Tyler, Schultz, Middleton, Spain, Dill and Steevens2008), colostrum colour (Argüello et al. Reference Argüello, Castro and Capote2005), refractometry (Bielmann et al. Reference Bielmann, Gillan, Perkins, Skidmore, Godden and Leslie2010), mid infrared spectrometry (Elsohaby et al. Reference Elsohaby, McClure, Hou, Riley, Shaw and Keefe2016), hydrometry (Rudovsky et al. Reference Rudovsky, Locher, Zeyner, Sobiraj and Wittek2008), size-exclusion HPLC (Dolman & Thorpe, Reference Dolman and Thorpe2002), electrophoresis (Page & Thorpe, Reference Page, Thorpe and Walker2002), enzyme-linked immunosorbent assay (ELISA) (Gelsinger et al. Reference Gelsinger, Jones and Heinrichs2015) and protein affinity liquid chromatography (Abernethy et al. Reference Abernethy, Otter, Arnold, Austad, Christiansen, Ferreira, Irvine, Marsh, Massom, Pearce, Stevens, Szpylka, Vyas, Woollard and Wu2010). Near-infrared spectroscopy (NIRS) has also been presented as a suitable and valuable method to predict IgG concentration in fresh bovine colostrum (Rivero et al. Reference Rivero, Valderrama, Haines and Alomar2012). This method of prediction can give a precise measurement of IgG concentration a few hours after the first milking.

The strong water absorptions observed and the wide variation found in water absorption signals reported (Rivero et al. Reference Rivero, Valderrama, Haines and Alomar2012), could imply that water content in liquid samples can disturb the calibration for other constituents (Núñez-Sánchez et al. Reference Núñez-Sánchez, Garrido, Serradilla, Ares, Davies and Cho2002), e.g., IgG. Therefore, dried samples of colostrum could improve efficiency of IgG prediction equations.

The Dry-Extract Spectroscopy for Infrared Reflectance (DESIR) is a direct technique of liquid analysis based on measuring diffuse infrared reflectance by solutes isolated on solid supports (Alfaro et al. Reference Alfaro, Meurens and Birth1990). This method consists of drying a glass fibre filter previously impregnated with the liquid under test. Regarding milk quality, DESIR was successfully used with goat's milk to calibrate NIRS equations for milk components, i.e., total protein, total casein, αs-, β- and κ-caseins, fat and lactose (Díaz-Carrillo et al. Reference Díaz-Carrillo, Muñoz-Serrano, Alonso-Moraga and Serradilla-Manrique1993). Spectra from dried ewe's milk samples were also collected to determine total protein, total casein, fat, total solids, and somatic cell count (Núñez-Sánchez et al. Reference Núñez-Sánchez, Garrido, Serradilla, Ares, Davies and Cho2002). Since NIRS is a reliable method to predict IgG content in bovine colostrum, and taking into account that dried samples could improve prediction equations, DESIR method was tested to improve the accuracy of prediction of colostrum IgG by NIRS over results obtained by transflectance on liquid samples.

The objective of this study was to compare the prediction efficiency of IgG in colostrum bovine by NIRS, using liquid and dried samples by transflectance and reflectance (DESIR) modes, respectively.

Materials and methods

Colostrum samples collection and storage

The study was carried out using a total of 157 samples of colostrum collected from both primiparous and multiparous Holstein-Friesian cows, from two commercial dairy farms with spring and autumn calving. These farms were located in southern Chile (Valdivia, Región de Los Ríos). Samples were collected the first hour after calving, mixed and divided into two subsamples: one for spectra collection and one for reference analysis (from now on both will be referred as ‘samples’). All samples were then stored at −20 °C.

Samples preparation for NIRS analyses

Prior to NIRS analysis, each of the 157 samples stored for spectra collection were thawed inside a refrigerator at 2 to 5 °C for 24 h and then, kept for 10 min at room temperature (18–20 °C). Thereafter, they were agitated for 1 min with a vortex mixer at maximum speed to ensure an uniform dispersion of colostrum components. From each sample, four aliquots of 500 mg each were taken for spectra collection. In the case of the liquid analysis, two aliquots were directly situated in a pair of cam-lock cells for liquid products. For the two aliquots destined to DESIR method, two glass microfibre filters (37 mm diameter, Whatman) were infused with colostrum (500 mg each) and dried at 60 °C for 20 min in a forced-air oven. Thereafter, dried samples were kept in a desiccator to reduce the remaining humidity.

Spectra collection

All samples were scanned in the visible and near infrared (VIS-NIR) region (400–2500 nm, at 2-nm intervals) in a NIRSystems 6500 monochromator (Silver Springs, MD). The software WINISI II (Infrasoft International, Port Matilda, PA) was used for scanning and calibration. All samples were scanned in duplicate; liquid samples were scanned by transflectance (folded transmission) and dried samples were scanned by reflectance. Each duplicate of liquid samples was situated in the cam-lock cell with an aluminium reflector and 0·1-mm pathlength (part number IH-0345-1). Each duplicate of dried samples were situated in a ring cell with quartz window (part number NR-7072). The cells were inserted in a spinning module (part number NR-6506) for the readings. Each scan was composed of 32 readings of the sample plus 32 of a white ceramic tile provided in the instrument as a reference. The resulting spectra were recorded and stored in suitable files as log (1/R), where R is the reflectance energy recovered by detectors positioned at 45 ° of the sample, in the VIS (silicon detectors) and NIR (lead sulphide detectors) range.

Reference analysis

Colostrum samples reserved for reference analysis were defrosted at 2–5 °C during 24 h and analysed by radial immunodiffusion (RID) at the Laboratory of Immunology, Western College of Veterinary Medicine, University of Saskatchewan (Canada). The RID assay was performed following the protocol of the laboratory (Chelack et al. Reference Chelack, Morley and Haines1993), on agar gel plates with rabbit anti-bovine IgG (Jackson Laboratories Inc., West Grove, PA). Dilution series of test colostrum and reference colostrum were dispensed (5 replicates of 4 µl) in wells (2 mm diameter) cut on each agar plate. Plates were incubated at room temperature (20–25 °C) for 18–24 h in a humid chamber and the ring diameters of precipitin surrounding the wells were recorded. Immunoglobulin G content was determined against a regression line generated with known standards incubated on the same plates. The final IgG concentration for each sample was determined by calculating the average of the 5 replicates. Results were expressed as g/L.

Calibrations and cross validations

Calibrations were performed with the full VIS and NIR segments of the electromagnetic range considering absorption values recorded every 1 data point (2 nm), which resulted in computation of a total of 1050 variables (wavelengths), although the final number of variables included in a model resulted from the mathematical treatment applied to the spectral data. Prediction equations were obtained for each treatment (transflectance and DESIR) by modified partial least squares regression (MPLS) and testing different mathematical treatments on the spectral data to resolve overlapping peaks and remove linear baselines (Hruschka, Reference Hruschka, Williams and Norris2001). These treatments consisted in applying (or not) a first or second derivative (subtraction) order across different intervals and with different smoothing segments of the spectral data. Standard normal variate (SNV) and detrending (DT) treatment were also applied for scatter correction (Barnes et al. Reference Barnes, Dhanoa and Lister1989). The option of no scatter correction was tested. Equations were evaluated by one-out full cross-validation, testing 4, 5 or 6 cross-validation groups. If 4 groups are used, the first group is created from samples in position 1, 5, 9, and so on, the second group with samples in position 2, 6, 10, and so on. Consecutively, each group is predicted by an equation developed with the remaining three groups, until all groups are ‘externally’ predicted. At each stage, a coefficient of determination and a standard error are computed, which are averaged at the final pass, resulting in a standard error of cross-validation, also referred as root mean square error of cross-validation (RMSECV) and a coefficient of determination of cross-validation (R2CV) (Shenk & Westerhaus, Reference Shenk, Westerhaus, Davies and Cho1996). Two passes of elimination of outliers were applied to remove aberrant predictions (‘T’ outliers). The T term represents the residual value to standard error of calibration (SEC) ratio for the group of samples (InfraSoft International, 1992). A critical value of 2·5 was set for T outliers, which is frequently used by default (InfraSoft International, 1992). Equations were selected according to the lowest RMSECV, the residual predictive deviation (RPD, relation between standard deviation of reference data and the RMSECV), and the range error ratio (RER, relation between the range of reference data and the RMSECV) (Williams & Sobering, Reference Williams, Sobering, Davies and Williams1996).

Results

Reference data

IgG concentration of colostrum samples by RID methods indicated a mean value (±sd) of 93·3 (±37·9) g/L, ranging from 1·7 to 185·4 g/L. Only 12·7% of the samples had a concentration below 50 g/L (Fig. 1).

Fig. 1. Immunoglobulin content of 157 Holstein Frisian bovine colostrum samples.

Spectra analysis

Raw (log1/R) spectra and average spectrum of liquid and dried colostrum samples are shown on Fig. 2. In liquid samples, two absorption peaks at around 1445 and 1940 nm were observed (Fig. 2, a1), in close connection with absorbance of O-H groups in water. In dried samples, some spectra showed prominent absorption peaks in the same wavelength as liquid samples (Fig. 2, b1). There was also a great variation of absorption in those ranges and some samples did not show a high absorption. A wider variation in absorbance in the visible range (standard deviation of the spectra) was noticed in the spectra from dried samples than that from liquid samples (Fig. 2, a2 vs. b2), especially in the wavelengths related to water absorption bands (1445 and 1940 nm).

Fig. 2. Visible and near infrared transflectance spectra of colostrum samples. (a1) Complete calibration set of liquid samples; (b1) Complete calibration set of dried samples; (a2) average spectrum and standard deviation of liquid samples; (b2) average spectrum and standard deviation of dried samples. R, reflectance.

Calibrations and cross-validations

Calibrations and cross-validations statistics for the best equations developed are presented in Table 1. For liquid samples, eight samples with unusual prediction residuals and 1 with an abnormal spectrum were excluded from the calibration in 2 passes of elimination of outliers, resulting in a final inclusion of 148 samples. For DESIR samples, nine samples with unusual prediction residuals were excluded from the calibration in 2 passes of elimination of outliers, resulting in a final inclusion of 148 samples. For both techniques, the best equations were obtained with five cross-validation groups.

Table 1. Calibration and prediction statistics of reference data and best equation obtained

PLS, partial least squares; Range, difference (highest minus lowest); sd, standard deviation of reference data (g/L); RMSECV, root mean square error of cross-validation (g/L); R2CV, coefficient of determination of cross validation; residual predictive deviation (RPD), SD divided by RMSECV; RER, range error ratio: range divided by RMSECV

MT, mathematical treatment = derivative order (first digit)-subtraction gap (second digit)-smoothing segment (third digit), standard normal variate (SNV) and detrending (DT)

For liquid samples, a second derivative order was performed on the spectral values across a subtraction gap of 16 nm (8 data points) and a smoothing segment of 8 nm (4 data points) (Table 1). The SNV and DT algorithms were applied to correct scatter. For dried samples (DESIR technique), a second derivative order was also performed on the spectral values but across a subtraction gap and a smoothing segment of 10 nm each (5 data points). No scatter treatment was applied. The prediction equations included 7 and 6 PLS terms, for liquid and dried samples, respectively.

The best calibration for transflectance (liquid samples) obtained a higher R2, R2CV, and RER and a lower RMSECV than the best calibration for dried (DESIR) samples (Table 1). Both RER terms were greater than 15 and both obtained RPD values greater than 3, with a better value for liquid (transflectance) samples. Regarding RMSECV, less is better, liquid samples obtained a lesser value (9·8% of the mean) than dried samples (12·6% of the mean). Reference values were plotted against NIRS predictions of IgG content obtained with the selected equations for transflectance (liquid samples) and reflectance (DESIR) techniques (Fig. 3).

Fig. 3. Near infrared reflectance spectroscopy (NIRS) predicted versus reference (radial immunodiffusion) values for IgG content (g/L) in bovine colostrum for liquid (a) and dried (b) techniques.

Discussion

The hypothesis of this work that the NIR spectra of bovine colostrum samples scanned by the DESIR method could result in better IgG predictions, compared to the spectra collected by transflectance of liquid samples is not confirmed.

In agreement with Murray (Reference Murray1988), a suitable group of samples was available for this study, with a wide range of IgG content and a high variability (coefficient of variation between 41 and 42%, Table 1). Another advantage of this set of samples is that they were obtained from herds under commercial conditions, the same conditions where predictions could eventually be performed as routine work. Regarding colostrum quality, most of the samples were considered to have a good level of IgG (McGuirk & Collins, Reference McGuirk and Collins2004). Related to the calibration value of the samples, even though there were relatively few samples under the critical value (50 g/L), those samples had an even distribution along IgG content (Fig. 1). With respect to fibreglass filters, any significant interference with the colostrum spectrum is discarded as this material has the advantage of practically no NIR absorption. Besides, light refraction and diffraction of fibreglass is very low (Díaz-Carrillo et al. Reference Díaz-Carrillo, Muñoz-Serrano, Alonso-Moraga and Serradilla-Manrique1993).

Even though the DESIR method obtained poorer calibration statistics than the transflectance method, the accuracy of the former was fairly good; R2CV over 0.90 means an excellent capacity for quantitative analysis (Núñez-Sánchez et al. Reference Núñez-Sánchez, Garrido, Serradilla, Ares, Davies and Cho2002). Different authors had reported good R2CV for calibration equations for diverse constituents obtained for milk from different species: total casein, alphas, beta and kappa caseins from goat's milk (0·87, 0·86, 0·92 and 0·86, respectively; Díaz-Carrillo et al. Reference Díaz-Carrillo, Muñoz-Serrano, Alonso-Moraga and Serradilla-Manrique1993); somatic cell count (SCC) from cow's milk (0·85, Tsenkova et al. Reference Tsenkova, Atanassova, Kawano and Toyoda2001); casein from cow's milk (0·96 to 0·98, Laporte & Paquin, Reference Laporte and Paquin1999; 0·93, Šustová et al. Reference Šustová, Kuchtík and Kráčmar2006). Particularly, Núñez-Sánchez et al. (Reference Núñez-Sánchez, Garrido, Serradilla, Ares, Davies and Cho2002) compared the ability to predict milk constituents (casein, protein, fat, total solids and SCC) by NIRS with DESIR (reflectance) and with liquid samples (folded transmission) of ewe milk. They found that for protein content and SCC, DESIR showed slightly better R2CV than liquid samples (0·94 vs. 0·92 for protein and 0·89 vs. 0·88 for SCC, respectively). However, for fat and total solids content, DESIR technique had lower R2CV than liquid samples (0·94 vs. 0·99 for fat and 0·98 vs. 0·99 for total solids content, respectively), whereas there was no difference for casein R2CV (0·88 for both methods). They concluded that, regarding RMSECV, the accuracy of folded transmission (liquid samples) was significantly higher than the corresponding reflectance equations (DESIR) only for fat and total solids content. In the present study, a similar trend was found, i.e. regarding R2CV and RMSECV, folded transmission was slightly better than reflectance, though both methods offer excellent results. According to Rivero et al. (Reference Rivero, Valderrama, Haines and Alomar2012), although the models could probably be improved if more samples were included in the range of 40 to 60 g/L and, also, some samples in the lowest range of IgG content were slightly overestimated by both models, from a practical point of view, those colostrum samples would still be predicted as poor quality by the model.

The small superiority in the transflectance technique compared with the DESIR method might be due to the greater variation in the water absorption bands of the dried samples. The high water content in some products can reduce the quality of the predictions of parameters, especially in samples with relatively low concentrations. This might be due to the interference of the wide bands of vibration of water (Rodriguez-Saona et al. Reference Rodriguez-Saona, Fry, McLaughlin and Calvey2001). This high variation in the water content of the dried samples was probably generated by the heterogeneity of the initial total solids/water content of the colostrum samples. Even though the colostrum moisture was not assessed previous to the spectra collection, the aspect of the colostrum samples, regarding apparent water content, was very heterogeneous, varying from thin and milky to thick and creamy, varying from light yellow to intense yellow, probably related with lower and higher fat content, respectively (Gross et al. Reference Gross, Kessler and Bruckmaier2014). As temperature and time of drying were equal for all samples, the final water content was probably different for each sample and related to the initial water content. This great variation in the moisture level of dried samples could be inferred from the spectra, given the higher coefficient of variation in the wavelengths related to water absorption bands (1445 and 1940 nm) of the dried samples compared with liquid samples. Complementarily, the appearance of dried samples on the fibreglass filter was also very heterogeneous, including samples from mildly humid filters and slightly yellow coloured to very dry filters with intense yellow material and wrapped edges.

The drying method of fruit juice samples proposed by Alfaro & Meurens (Reference Alfaro and Meurens1989) consisted in keeping the filter for 15 min in a fan forced oven at 70 °C. In the case of colostrum, IgG shows two peaks of denaturation at different temperatures: 61 and 71 °C (Vermeer & Norde, Reference Vermeer and Norde2000). That finding determined the modification of the drying method in this study, reducing the air-forced oven temperature to 60 °C and increasing the time to 20 min. For goat's milk analysis it has been proposed to modify the typical drying conditions of this method (70 °C, 15 min) to 40 °C during 4 h (Díaz-Carrillo et al. Reference Díaz-Carrillo, Muñoz-Serrano, Alonso-Moraga and Serradilla-Manrique1993), hampering the main advantage of the NIRS technique: quickness in results. For ewe's milk the proposed method was drying the infused filters for 24 h at 40 °C (Núñez-Sánchez et al. Reference Núñez-Sánchez, Garrido, Serradilla, Ares, Davies and Cho2002), which is a much longer drying time than the proposed in this study. A drying curve could be attempted to overcome the heterogeneous final moisture content, i.e. adjusting drying time according to the water content of the samples. This alternative would imply a previous analysis of each sample to assess the water content and, thereby, applying the required time to completely dry the sample. However, this additional step would increase the time required for obtaining results as well as costs. Alfaro et al. (Reference Alfaro, Meurens and Birth1990) proposed the DESIR dryer as an alternative to reduce coefficient of variation (CV) of the spectra. They found that the DESIR dryer, using sugar solutions to simulate fruit juices, had the lowest CV, 0·76 to 1·49%, the air flow oven has an also low CV, 2·28 to 3·14%, and the microwave drying technique has the highest variation, a CV greater than 10%. This indicates that the drying technique can affect the spectra variation. Therefore, a short drying time could not be possible to implement in very heterogeneous samples, as is the case of bovine colostrum, given that instead of reducing CV, this pre-treatment would increase it. Longer times of drying could overcome this problem. However, that practice would increase time for obtaining results, which in a field level could be impracticable or useless to make quick and objective decisions.

As proposed by Rodriguez-Saona et al. (Reference Rodriguez-Saona, Fry, McLaughlin and Calvey2001), the high water content of some products can affect the accuracy of prediction of constituents presented in relatively low concentrations. Moreover, Thyholt & Isaksson (Reference Thyholt and Isaksson1997) concluded that NIRS technique give good composition quantification for the analysis of major constituents with moderate moisture content products; i.e. lower than 80% water content (weight basis). According to Davis & Drackley (Reference Davis and Drackley1998) bovine colostrum obtained in the first milking contains 239 g/kg of total solids, with 140 g/kg of proteins (70 to 80% are immunoglobulins), containing 32 g/L of IgG (75% of total immunoglobulins). Therefore, bovine colostrum water content could not be interfering significantly with NIRS determination of IgG obtained in the first milking after calving.

Removing water by DESIR means removing the H bonding interference and giving small molecules more characteristic spectra (Thyholt & Isaksson, Reference Thyholt and Isaksson1997). On the other hand, to overcome the problem of water, other commonly employed strategy, which prevents desiccation of the product, is based on the elimination of the areas near the four intense absorption bands of water molecules (970, 1190, 1450 and 1950 nm), reducing the reading range (Rambla et al. Reference Rambla, Garrigues and de la Guardia1997). However, it is expected that the IgG concentration of colostrum obtained during the first milking after calving is directly related with total solids content. Hence, IgG concentration should be inversely related to moisture content, given that IgG is the most abundant immunoglobulin in colostrum, and proteins the most concentrated solid in that stage of lactation (Davis & Drackley, Reference Davis and Drackley1998). Therefore, water absorption could provide valuable indirect predictive information when considering the whole NIR spectrum. If so, partial moisture elimination from colostrum by DESIR could reduce the possible contribution of water to IgG prediction capacity of NIRS. Moreover, working with liquid samples has the additional advantage of making possible the application of NIRS in-field by using more robust and portable instruments, i.e. taking the instrument to the sample (Cattaneo & Holroyd, Reference Cattaneo and Holroyd2013).

In conclusion, regarding accuracy of prediction and time for obtaining results of IgG from bovine colostrum, and considering that both methods show excellent analytical capacity, NIRS analysis of liquid samples (transflectance) is recommended over dried samples (DESIR technique by reflectance). Besides, usefulness of NIRS analysis for liquid samples to select or discriminate dams according to their colostrum quality is highlighted. Moreover, a model like this could be strengthened by adding new samples from different years and diverse farms.

This work was part of the PhD studies of M. Jordana Rivero, funded by a Doctoral Fellowship provided by CONICYT (Chilean National Commission for Scientific and Technological Research). The support of the personnel of the Laboratory of Animal Nutrition (Institute Animal Production, Faculty of Agricultural Sciences, Universidad Austral de Chile, Valdivia) and of the Laboratory of Immunology (Department of Veterinary Microbiology, Western College of Veterinary Medicine, University of Saskatchewan, Saskatoon, Canada) for the NIRS and RID analysis, respectively, are gratefully acknowledged.

References

Abernethy, G, Otter, D, Arnold, K, Austad, J, Christiansen, S, Ferreira, I, Irvine, F, Marsh, C, Massom, LR, Pearce, K, Stevens, J, Szpylka, J, Vyas, P, Woollard, D & Wu, C 2010 Determination of immunoglobulin G in bovine colostrum and milk powders, and in dietary supplements of bovine origin by protein G affinity liquid chromatography: collaborative study. Journal of AOAC International 93 622627 CrossRefGoogle ScholarPubMed
Alfaro, G & Meurens, M 1989 Determination of individual sugar and aminoacid by dry extract NIR reflectance spectroscopy on glass fiber disk. Proceedings of the Second International Conference on NIR Spectroscopy, pp. 204207 Google Scholar
Alfaro, G, Meurens, M & Birth, GS 1990 Liquid analysis by dry-extract near-infrared reflectance on fiberglass. Applied Spectroscopy 44 979986 Google Scholar
Argüello, A, Castro, N & Capote, J 2005 Short communication: evaluation of a color method for testing immunoglobulin G concentration in goat colostrum. Journal of Dairy Science 88 17521754 Google Scholar
Barnes, RJ, Dhanoa, MS & Lister, SJ 1989 Standard normal variate transformation and de-trending of near infrared diffuse reflectance spectra. Applied Spectroscopy 43 772777 Google Scholar
Bielmann, V, Gillan, J, Perkins, NR, Skidmore, AL, Godden, S & Leslie, KE 2010 An evaluation of Brix refractometry instruments for measurement of colostrum quality in dairy cattle. Journal of Dairy Science 93 37133721 Google Scholar
Cattaneo, TMP & Holroyd, E 2013 New applications of near infrared spectroscopy on dairy products. Journal of Near Infrared Spectroscopy 21 307310 Google Scholar
Chelack, BJ, Morley, PS & Haines, DM 1993 Evaluation of methods for dehydration of bovine colostrum for total replacement of normal colostrum in calves. The Canadian Veterinary Journal 34 407412 Google Scholar
Chigerwe, M, Tyler, JW, Schultz, LG, Middleton, JR, Spain, JN, Dill, JS & Steevens, BJ 2008 Comparison of four methods to assess colostral IgG concentration in dairy cows. Journal of the American Veterinary Medical Association 233 761766 Google Scholar
Conneely, M, Berry, DP, Sayers, R, Murphy, JP, Lorenz, I, Doherty, ML & Kennedy, E 2013 Factors associated with the concentration of immunoglobulin G in the colostrum of dairy cows. Animal 7 18241832 Google Scholar
Davis, CL & Drackley, JK 1998 The Development, Nutrition, and Management of the Young Calf. Ames, Iowa: Iowa State University Press.Google Scholar
Díaz-Carrillo, E, Muñoz-Serrano, A, Alonso-Moraga, A & Serradilla-Manrique, J 1993 Near infrared calibrations for goat's milk components: protein, total casein, αs-, β- and κ-caseins, fat and lactose. Journal of Near Infrared Spectroscopy 1 141146 Google Scholar
Dolman, C & Thorpe, R 2002 Analysis and purification of IgG using size-exclusion high performance liquid chromatography (SE-HPLC). In The protein protocols handbook, pp. 995–997 (Ed. M Walker). Totowa, NJ, USA: Springer Humana Press Google Scholar
Elsohaby, I, McClure, JT, Hou, S, Riley, CB, Shaw, RA & Keefe, GP 2016 A novel method for the quantification of bovine colostral immunoglobulin G using infrared spectroscopy. International Dairy Journal 52 3541 Google Scholar
Gelsinger, SL, Jones, CM & Heinrichs, AJ 2015 Effect of colostrum heat treatment and bacterial population on IgG absorption and health of neonatal calves. Journal of Dairy Science 98 46404645 Google Scholar
Godden, S 2008 Colostrum management for dairy calves. Veterinary Clinics: Food animal practice 24 1939 Google Scholar
Gross, JJ, Kessler, EC & Bruckmaier, RM 2014 Colour measurement of colostrum for estimation of colostral IgG and colostrum composition in dairy cows. Journal of Dairy Researc 81 440444 Google Scholar
Hruschka, WR 2001 Data analysis: wavelength selection methods. In Near-Infrared Technology in the Agricultural and Food Industries, pp. 39–58 (Eds Williams, P & Norris, K). St Paul, Minneapolis: American Association of Cereal Chemists Google Scholar
InfraSoft International 1992 NIRS 3. Routine Operation, Calibration and Network System Management Software for Near Infrared Instruments. Maryland, USA, Perstorp Analytical inc., Silver Spring Google Scholar
Laporte, MF & Paquin, P 1999 Near-infrared analysis of fat, protein, and casein in cow's milk. Journal of Agricultural and Food Chemistry 47 26002605 Google Scholar
Mancini, G, Carbonara, AO & Heremans, JF 1965 Immunochemical quantitation of antigens by single radial immunodiffusion. Immunochemistry 2 235254 Google Scholar
Morrill, KM, Conrad, E, Polo, J, Lago, A, Campbell, J, Quigley, J & Tyler, H 2012 Estimate of colostral immunoglobulin G concentration using refractometry without or with caprylic acid fractionation. Journal of Dairy Science 95 39873996 Google Scholar
McGuirk, SM & Collins, M 2004 Managing the production, storage, and delivery of colostrum. Veterinary Clinics: Food Animal Practice 20 593603 Google Scholar
Murray, I 1988 Aspects of the interpretation of Near Infrared Spectra. Food Science and Technology Today 2 135140 Google Scholar
Nocek, JE, Braund, DG & Warner, RG 1984 Influence of neonatal colostrum administration, immunoglobulin, and continued feeding of colostrum on calf gain, health, and serum protein. Journal of Dairy Science 67 319333 Google Scholar
Núñez-Sánchez, N, Garrido, A, Serradilla, JM & Ares, JL 2002 Near Infrared analysis of liquid and dried ewe milk. In Proceedings of the 10th International Conference of Near Infrared Spectroscopy, pp. 225228 (Eds Davies, AMC & Cho, RK). Chichester, UK: NIR Publications Google Scholar
Page, M & Thorpe, R 2002 Analysis of IgG fractions by electrophoresis. In The protein protocols handbook, pp. 10051007 (Ed. Walker, M). Totowa, NJ, USA: Springer Humana Press Google Scholar
Rambla, F, Garrigues, JS & de la Guardia, M 1997 PLS-NIR determination of total sugar, glucose, fructose and sucrose in aqueous solutions of fruit juices. Analytica Chimica Acta 344 41 Google Scholar
Rivero, MJ, Valderrama, X, Haines, D & Alomar, D 2012 Prediction of immunoglobulin G content in bovine colostrum by near-infrared spectroscopy. Journal of Dairy Science 95 14101418 Google Scholar
Rodriguez-Saona, LE, Fry, FS, McLaughlin, MA & Calvey, EM 2001 Rapid analysis of sugars in fruit juices by FT-NIR spectroscopy. Carbohydrate Research 336 6374 Google Scholar
Rudovsky, A, Locher, L, Zeyner, A, Sobiraj, A & Wittek, T 2008 Measurement of immunoglobulin concentration in goat colostrum. Small Ruminant Research 74 265269 Google Scholar
Shenk, JS & Westerhaus, M 1996 Calibration the ISI way. In Near-Infrared Spectroscopy: The Future Waves, pp. 198202 (Eds Davies, AMC & Cho, RK). Chichester, UK: NIR Publications Google Scholar
Šustová, K, Kuchtík, J & Kráčmar, S 2006 Analysis of ewe's milk by FT near infrared spectroscopy: measurement of samples on Petri dishes in reflectance mode. Acta Universitatis Agriculturae et Silviculturae Mendelianae Brunensis 53 131138 Google Scholar
Thyholt, K & Isaksson, T 1997 Near infrared spectroscopy of dry extracts from high moisture food products on solid support – a review. Journal of Near Infrared Spectroscopy 5 179193 Google Scholar
Tsenkova, R, Atanassova, S, Kawano, S & Toyoda, K 2001 Somatic cell count determination in cow's milk by near-infrared spectroscopy: a new diagnostic tool. Journal of Animal Science 79 25502557 Google Scholar
Vermeer, AW & Norde, W 2000 The thermal stability of immunoglobulin: unfolding and aggregation of a multi-domain protein. Biophysical Journal 79 21502154 Google Scholar
Williams, PC & Sobering, DC 1996 How do we do it: a brief summary of the methods we use in developing near infrared calibrations. In Near Infrared Spectroscopy: the Future Waves, pp. 185188 (Eds Davies, AMC & Williams, P). Chichester, UK: NIR Publications Google Scholar
Figure 0

Fig. 1. Immunoglobulin content of 157 Holstein Frisian bovine colostrum samples.

Figure 1

Fig. 2. Visible and near infrared transflectance spectra of colostrum samples. (a1) Complete calibration set of liquid samples; (b1) Complete calibration set of dried samples; (a2) average spectrum and standard deviation of liquid samples; (b2) average spectrum and standard deviation of dried samples. R, reflectance.

Figure 2

Table 1. Calibration and prediction statistics of reference data and best equation obtained

Figure 3

Fig. 3. Near infrared reflectance spectroscopy (NIRS) predicted versus reference (radial immunodiffusion) values for IgG content (g/L) in bovine colostrum for liquid (a) and dried (b) techniques.