Processing math: 100%
ARTICLE   Open Access    

Near-infrared probe as a quality control tool for milk powder blending processes

More Information
  • This study aims to evaluate the suitability and reliability of Process Analytical Tools (PAT) in monitoring milk powder blending processes. The uniformity end point was predicted using a Near-Infrared (NIR) probe, and subsequently validated using offline Fourier Transform Infrared Spectroscopy (FTIR). A standard milk formulation (SMF) made up of 50% lactose, 40% skim milk powder, and 10% whey protein concentrate was used. Additionally, the detection limit of the NIR probe was investigated using vitamin C powder. The average predicted uniformity end point using the inline NIR fixed reference (63.89 ± 2.06 min), and dynamic reference conformity test (63.00 ± 5.25 min) were comparable with the offline FTIR measurement (56.6 ± 0.71 min). A three-component Partial Least Square Regression (PLSR) model was constructed and validated for vitamin C. The detection limit is 0.11%, which is higher than the vitamin C level commonly found in most infant milk formula (0.035%).
  • The intestinal mucosal barrier is composed of epithelial cells and intercellular connections, which can effectively regulate the transportation of large molecules in the intestinal lumen, and prevent microorganisms, harmful solutes, toxins, and intraluminal antigens from entering bodies[1]. Mucosal barrier factors such as trefoil factor (TFF) family, diamine oxidase (DAO), and transform growth factor-ɑ (TGF-α) have protective and restorative effects on intestinal mucosal integrity, which are synthesized and secreted by intestinal mucosa[24]. The damage to intestinal barrier function can cause the invasion of antigens and bacteria in the lumen, and eventually lead to intestinal diseases, including diarrhea, inflammatory bowel disease (IBD), and Crohn's disease[57].

    The Notch signaling pathway plays a crucial role in a series of cellular processes, including proliferation, differentiation, development, migration, and apoptosis. Research suggests that the Notch signaling pathway is involved in intestinal development, and it can be connected to intestinal cell lineage specification[8]. Furthermore, the Notch signaling pathway can regulate intestinal stem cells, CD4+T cells, innate lymphoid cells, macrophages, and intestinal microbiota, and intervene in the intestinal mucosal barriers in cases of ulcerative colitis[9]. The activated Notch signaling pathway can suppress the goblet cells differentiation and mucus secretion, which would result in the disruption of the intestinal mucosal barrier[10]. In addition, the absence of the Notch signaling pathway would cause the dysfunction of tight junctions and adherens junctions of intestinal mucosa, which could lead to increased permeability of epithelial cells and exposure of luminal contents to the immune system and inflammation[11]. Phytochemicals, such as cucurbitacin, honokiol and quercetin have been reported to have therapeutic effects on intestinal diseases by targeting the Notch signaling pathway[1214].

    Plant-based diets rich in phytochemicals, such as phenolics, anthocyanins, and vitamins, have been related to the prevention of human diseases[15]. Anthocyanins belong to a subclass of flavonoids, and are mainly in the form of different anthocyanin combined with glucose, galactose, and arabinose[16]. The physiological action of anthocyanins, such as antioxidant activity, anti-inflammation, and anti-obesity effects have been widely reported[17,18]. M3G is found naturally in plants and is reported to be the most common anthocyanin in different blueberry varieties[19,20]. Previous researchers reported that M3G from blueberry suppressed the growth and metastasis potential of hepatocellular carcinoma cells, modulated gut microbial dysbiosis, and protected TNF-α induced inflammatory response injury in vascular endothelial cells[2123]. Anthocyanins with diverse molecular structures and from different dietary sources are bioavailable at diet-relevant dosage rates, and anthocyanins from berry fruit are absorbed and excreted by both humans and rats[24]. Studies have shown that anthocyanins have a protective effect on intestinal barrier damage by regulating gut microbiota, tight junction (TJ) protein expression, and secretion of MUC2[2527]. In addition to physiological indicators related to the intestinal barrier, the regulation of the Notch signaling pathway was involved in this study. Therefore, it is speculated that M3G could improve the colonic mucosal barrier function via the Notch signaling pathway.

    To prove this hypothesis, the effects of M3G on the colonic mucosal barrier function were investigated in DSS-induced colitis mice. The pathological morphology of the colon tissue, markers of intestinal physical barrier function and immune barrier function and the Notch signaling pathway were evaluated to reveal the mechanism of M3G in improving colonic mucosal barrier function. The results are expected to lay the foundation for the utilization of anthocyanins as a promising natural product for improving intestinal diseases.

    M3G (CAS: 30113-37-2) used in this study was purchased from Xinyi Science and Technology Instrument Business Department, Baoji, Shanxi, China (HPLC ≥ 98%).

    Five-week-old male C57BL/6J mice (Wanlei Bio Co., Ltd., Shenyang, Liaoning, China) weighing 16−18 g were housed and given AIN-93M diet (Wanlei Bio Co., Ltd., Shenyang, Liaoning, China) feeding. The animal experiment was carried out according to the guidelines of the Standards for Laboratory Animals of China (GB 14922-94, GB 14923-94, and GB/T 14925-94) and the Ethics Committee of Shenyang Agricultural University (IACUC Issue No.: 2023022401). All animal housing and experiments were conducted in strict accordance with the institutional guidelines for the care and use of laboratory animals.

    After acclimating to the breeding environment for one week, the mice were randomly divided into two groups: control group (CG group, n = 6) and model group (n = 12). On days 1−7, the mice of the model group were given 2.5% DSS (CAS: 9011-18-1) dissolved in drinking water, while the mice of the CG group were given the same volume of drinking water as the model group. On days 8−14, mice of the model group were randomly divided into two equal groups (n = 6): DSS group and M3G group. The mice of the CG and DSS groups were administered intragastrically via drinking water, and the mice of the M3G group were administered intragastrically by M3G (5 mg/kg body weight (BW)/d) dissolved in drinking water, and the liquid volume was controlled to be the same. For all groups of mice, the body weight, food consumption, stool consistency, and bloody stool were measured daily during the experiment. On day 15, mice were sacrificed after a 12 h fast, and the colon tissue samples of the mice were collected. The length of the colon tissue was measured and recorded.

    Colon tissue was embedded with paraffin and then cut into sections. After being dewaxed from paraffin, the tissue was placed in water, and stained with hematoxylin and eosin solution. The stained tissue was dehydrated and sealed for observation. A microscope (BX53, Olympus Co., Ltd., Tokyo, Japan) was used to observe the stained tissue and photographed using 100× magnification. The damage in epithelial cells of colon tissue was evaluated.

    The colon tissue was dewaxed and placed in water after being embedded with paraffin, and then stained with schiff and hematoxylin solution. The stained tissue was dehydrated and sealed for observation. A microscope was used to observe the stained tissue and photographed using 100× magnification. The mucosal thickness and the population of goblet cells of the colon tissue were measured and recorded.

    The expression of MUC2 in colon tissue was detected by RT-PCR. Total RNA was extracted from colon tissue, and the concentration was determined using an ultraviolet spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The single-stranded cDNA of the extracted RNA (0.1 μL) was synthesized using a Transcriptor First Strand cDNA Synthesis Kit (Roche Co., Ltd., Basel, Kanton Basel, Switzerland). ExicylerTM 96 (Bioneer Corporation, Daejeon, Korea) was used to analyze the fluorescence quantitative cDNA. The reaction conditions were as follows: 94 °C for 5 min, 94 °C for 10 s, 60 °C for 20 s, 72 °C for 30 s, then followed by 40 cycles of 72 °C for 2 min 30 s, 40 °C for 1 min 30 s, and then melting from 60 to 94 °C, and incubating at 25 °C for 1−2 min. The primer sequences are shown in Table 1.

    Table 1.  The primer sequences used in the RT-PCR analysis.
    Gene Prime Sequence (5'-3') Size (bp)
    MUC2 Forward TGTGCCTGGCTCTAATA 17
    Reverse AGGTGGGTTCTTCTTCA 17
    β-actin Forward CTGTGCCCATCTACGAGGGCTAT 23
    Reverse TTTGATGTCACGCACGATTTCC 22
     | Show Table
    DownLoad: CSV

    The whole proteins from the colon tissue (200 mg) were extracted using a Whole Cell Lysis Assay kit (BioTeke Co., Ltd., Beijing, China) according to the manufacturer’s protocol. Briefly, colon tissue was cut into pieces, and then mixed with phenylmethylsulfonyl fluorid (PMSF), followed by adding the protein extraction reagents A and B to prepare the tissue homogenate. Protein concentration was then determined using a Bradford Kit (BioTeke Co., Ltd., Beijing, China) according to the manufacturer’s protocol. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The membranes were blocked with 5% non-fat dry milk in TBST buffer for 1 h, followed by incubating overnight at 4 °C with the appropriate monoclonal primary antibody (detailed information in Table 2). Then membranes were washed to remove non-bound antibodies, and then incubated with the secondary antibody (goat anti-rabbit immunoglobin G-horseradish peroxidase (IgG-HRP), 1:5000; Wanlei Bio Co., Ltd., Shenyang, Liaoning, China) at 37 °C for 45 min. The enhanced chemiluminescence (ECL) western blotting detection reagent (Wanlei Bio Co., Ltd., Shenyang, Liaoning, China) was used to detect the protein bands, and then the protein bands were visualized by a Gel Imaging System (Beijing Liuyi Biotechnology Co., Ltd., Beijing, China).

    Table 2.  Details of the primary antibodies used in the experiment.
    Primary antibody Dilution ratio Manufacturer
    Claudin-1 1:500 Wanlei Bio Co., Ltd.
    Occludin 1:500 Wanlei Bio Co., Ltd.
    ZO-1 1:500 Wanlei Bio Co., Ltd.
    iFABP 1:1000 ABclonal Technology Co., Ltd.
    DLL1 1:500 Wanlei Bio Co., Ltd.
    DLL4 1:1000 ABclonal Technology Co., Ltd.
    Notch1 1:500 Wanlei Bio Co., Ltd.
    NICD 1:1000 Affinity Biosciences Co., Ltd.
    Hes1 1:1000 ABclonal Technology Co., Ltd.
    TFF3 1:1000 Affinity Biosciences Co., Ltd.
    β-actin 1:1000 Wanlei Bio Co., Ltd.
     | Show Table
    DownLoad: CSV

    The level of SIgA in colon tissue was detected by SIgA ELISA kit (Model number: EM1362, Wuhan Fine Biotech Co., Ltd., Wuhan, Hubei, China). The experiment was carried out according to the ELISA kit instructions.

    CD4+T (CD3+CD4+) cells and CD8+T (CD3+CD8+) cells in colonic lamina propria monocytes (LPMC) were detected by FCM. The colonic epithelial cells were isolated by adding digestive solution to the colon tissue. After digestion, screening, re-suspension precipitation, and centrifugation, the precipitate was collected as colonic LPMC. Labeled antibodies were added to the flow tube and incubated according to the instructions. After washing with phosphate buffered saline (PBS), the cells were re-suspended in PBS solution. The CD4+T (CD3+CD4+) cells and CD8+T (CD3+CD8+) cells were detected by flow cytometric (Agilent Technologies, Inc., Santa Clara, CA, USA).

    Data are expressed as mean ± standard deviation (SD) based on six replicates. Differences between two groups were assessed using Student's t tests. The results were considered statistically significant at p < 0.05. Data were analyzed using Graph Pad Prism 8.0 (Graph Pad Software, San Diego, CA, USA) and SPSS 17.0 (IBM Corporation, Armonk, NY, USA).

    The body weight of mice was monitored daily during the experimental period. As shown in Fig. 1a, after being fed with DSS for 7 d, the body weight of the mice in the DSS group was reduced significantly (p < 0.01) compared with that of the CG group. While the body weight of the mice in the M3G group was increased, and exhibited significantly higher (p < 0.01) body weight gain compared with the DSS group. The DAI score of the mice in the DSS group was significantly higher (p < 0.01) than that of the CG group, but M3G supplementation significantly lowered (p < 0.01) the DAI score (Fig. 1b), which demonstrates that the DSS-induced mice colitis model was established successfully and M3G inhibited colon tissue damage in DSS-fed mice. Compared with the CG group, the food intake of the mice fed with DSS was significantly reduced (p < 0.01), and M3G supplementation significantly increased (p < 0.05) the food intake (Fig. 1c).

    Figure 1.  Effect of M3G on body weight gain, DAI score, and food intake in mice. (a) Body weight gain. (b) DAI score (Fecal consistency: 0 = normal, 1 = semi-formed, 2 = soft stool, 3 = diarrhea or watery stool; bloody stool: 0 = occult blood negative, 1 = occult blood positive, 2 = visible bloody stool, 3 = massive hemorrhage; weight loss: 0 = 0%, 1 = 1%−5 %, 2 = 6%−10% ; 3 = 11%−15% reduction). (c) Food intake. Results are expressed as the mean ± SD (n = 6). * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    Representative HE and PAS-stained sections of the colon are shown in Fig. 2a & b. The morphology of colon tissue in the CG group did not exhibit any damage, while damage in epithelial cells and goblet cells, inflammatory cells infiltration, and separation of the muscle layer and mucosal muscle layer were observed in the DSS group. The above damage to the tissue was changed for the better by M3G supplementation, with only small levels of damage in epithelial cells and goblet cells, inflammatory cell infiltration, and separation of the muscle layer and mucosal muscle layer compared to that in the CG group.

    Figure 2.  Effect of M3G on the pathological morphology of colon tissue. (a) HE staining of colon tissue. Original magnifications: 100×. (b) PAS staining of colon tissue. Original magnifications: 100×. (c) The damage score of colon tissue. (1) Epithelial cell damage: 0 = normal morphology; 1 = regional destruction of the epithelial surface; 2 = diffuse epithelial destruction and/or mucosal ulcers involving submucosa; 3 = severe epithelial damage. (2) Inflammatory cell infiltration: 0 = no infiltration or less than 5 cells; 1 = mild infiltration of the inherent layer; 2 = moderate infiltration of the muscular mucosa; 3 = high infiltration of the muscular mucosa; 4 = severe infiltration involving submucosa. (3) Separation of muscle layer and mucosal muscle layer: 0 = normal; 1 = moderate; 2 = severe. (4) Goblet cell depletion: 0 = no depletion; 1 = presence of disordered goblet cells; 2 = 1 to 3 regions without goblet cells; 3 = more than 3 regions without goblet cells; 4 = complete depletion of goblet cells. (d) The colon length. (e) The mucosal thickness of colon tissue. (f) Number of goblet cells in colon tissue. Results are expressed as the mean ± SD (n = 6). * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    HE score was calculated to evaluate the degree of colon tissue damage. The HE score in the DSS group was significantly higher (p < 0.01) than that in the CG group. Supplementation of M3G significantly reduced (p < 0.01) the score, which decreased by almost 50% of that in the DSS group (Fig. 2c). The colon length in the DSS group was significantly lower (p < 0.01) than that in the CG group, which was significantly increased (p < 0.01) by M3G supplementation (Fig. 2d). The mucosal thickness and the number of goblet cells of the colon tissue were determined by PAS staining. Compared with the CG group, the mucosal thickness and the number of goblet cells of the colon tissue were significantly decreased (p < 0.01) in the DSS group, but those were significantly increased (p < 0.01) in the M3G group (Fig. 2e & f). These results suggest that M3G supplementation can decrease the pathological damage in the colon tissue induced by DSS.

    The mRNA expression level of MUC2 in the colonic mucosa tissue was determined. Compared with the CG group, DSS significantly decreased (p < 0.01) the mRNA level of MUC2, but this effect was significantly suppressed (p < 0.01) by M3G supplementation (Fig. 3a). The protein levels of claudin-1, occludin, ZO-1, iFABP, and TFF3 in the colon tissue were subsequently assessed. The protein expression levels of claudin-1, occludin, and ZO-1 in the DSS group were significantly lower than those in the CG group (p < 0.01). M3G supplementation significantly inhibited (p < 0.01) the reduction in the expression of these proteins (Fig. 3bd). Regarding the protein expression level of iFABP, it was significantly higher (p < 0.01) in the DSS group than that in the CG group. M3G supplementation significantly decreased (p < 0.01) iFABP expression level compared with the DSS group (Fig. 3e). The protein expression level of TFF3 was significantly higher (p < 0.01) in the DSS group than that in the CG group. However, M3G supplementation significantly intensified (p < 0.01) the increase of TFF3 expression compared with the DSS group (Fig. 3f). These results indicate that M3G can regulate colonic epithelial barrier function.

    Figure 3.  Effects of M3G on the colonic epithelial barrier disruption. (a) The mRNA expression level of MUC2. (b) The protein expression level of claudin-1. (c) The protein expression level of occludin. (d) The protein expression level of ZO-1. (e) The protein expression level of iFABP. (f) The protein expression level of TFF3. Results are expressed as the mean ± SD (n = 6). * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    The content of CD4+T (CD3+CD4+) and CD8+T (CD3+CD8+) cells in colonic LPMC of mice in each group were detected by FCM as shown in Fig. 4 and Supplemental Fig. S1. Compared with the CG group, the percentages of CD4+T (CD3+CD4+) and CD8+T (CD3+CD8+) cells in the DSS group were significantly increased (p < 0.01), and those were significantly decreased (p < 0.05) after M3G supplementation (Fig. 4a & b). SIgA was measured as an indicator of immune barrier function in the colon tissue. The level of SIgA in the DSS group was significantly lower (p < 0.01) than that in the CG group. Supplementation of M3G significantly raised (p < 0.01) the level of SIgA (Fig. 4c). These results indicate that M3G can improve the colonic immune dysfunction induced by DSS.

    Figure 4.  Effect of M3G on the colonic immune barrier function. (a) The percentage of CD4+T (CD3+CD4+) cells in the colon tissue. (b) The percentage of CD8+T (CD3+CD8+) cells in the colon tissue. (c) The level of SIgA in the colonic mucosal tissue. Results are expressed as the mean ± SD (n = 6). * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    Subsequently, the protein expression levels of Notch1, NICD, DLL4, DLL1, and Hes1 in the colon tissue were measured (Fig. 5). DSS significantly up-regulated (p < 0.01) these protein expression levels compared with the CG group. M3G supplementation significantly (p < 0.01) down-regulated the protein expression levels of Notch1, NICD, DLL4, DLL1, and Hes1 in comparison with the DSS group. The results indicate that M3G can inhibit the over-activation of the Notch signaling pathway.

    Figure 5.  Effects of M3G on the protein expression levels of Notch1, NICD, DLL4, DLL1, and Hes1 in the colon tissue. (a) The protein expression level of Notch1. (b) The protein expression level of NICD. (c) The protein expression level of DLL4. (d) The protein expression level of DLL1. (e) The protein expression level of Hes1. Results are expressed as the mean ± SD (n = 6). * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    To explain the relationships between them, the Spearman r correlations between biomarkers and the Notch signaling pathway were analyzed, and represented using a heatmap. As shown in Fig. 6, colon length, food intake, ZO-1, number of goblet, MUC2, SIgA, claudin-1, occludin, body weight gain, and mucosal thickness were significantly (p < 0.01 or p < 0.05) positively correlated with Notch1, NICD, DLL4, DLL1, and Hes1. Conversely, HE score, iFABP, DAI score, CD4+T, and CD8+T were significantly (p < 0.01) negatively correlated with Notch1, NICD, DLL4, DLL1 and Hes1. In particular, TFF3 was only significantly (p < 0.05) positively correlated with DLL1, and Hes1. The results indicate that M3G may ameliorate the colonic mucosal barrier dysfunction via modulation of the Notch signaling pathway.

    Figure 6.  Heatmap of the Spearman r correlations between biomarkers and the Notch signaling pathway. * p < 0.05 and ** p < 0.01 indicate significant differences between two groups.

    Within the last five years, special attention is being paid to the therapeutic effects of anthocyanins. The recent studies could give evidence to prove the therapeutic potential of anthocyanins from different sources against various diseases via in vitro, in vivo, and epidemiological experiments[28]. The anthocyanin supplementation has been demonstrated to have positive effects on intestinal health[29]. The intestinal barrier is one of the crucial factors which can affect intestinal health and normal intestinal barrier function not only maintains intestinal health but also protects overall health by protecting the body from intestine injury, pathogen infection, and disease occurrence[30]. The disruption of intestinal barrier integrity is regarded as an important factor leading to IBD, obesity, and metabolic disorders[3133]. In this study, the effects of M3G on regulating the colonic physical barrier function and colonic immune barrier function in DSS-induced colitis mice were explored.

    M3G supplementation reduced the DAI score and the HE score of colon tissue, and restored colon length, mucosal thickness, and the number of goblet cells in the colon tissue, indicating that M3G can alleviate DSS-induced colon tissue damage and colitis symptoms in mice. The DAI score was developed as a simplified clinical colitis activity index to assess the severity of colitis[34]. Zhao et al. have found that black rice anthocyanin-rich extract can significantly decrease the DAI and HE scores of colon tissue in DSS-induced colitis mice[35]. Intestinal goblet cells are mainly differentiated from multipotential stem cells. Intestinal stem cells were located at the base of the crypt and distributed in intestinal mucosal epithelial cells, composing around 50% of colon epithelial cells[36]. The mucus layer is formed by goblet cell secretion, it separates the intestinal epithelium from the intestinal lumen, thereby preventing the invasion of pathogenic microorganisms and the translocation of intestinal microbiota[37]. The results suggested that M3G might repair the colonic mucosa by increasing the number of goblet cells.

    MUC2 is the most abundant mucin secreted by goblet cells of the colon, goblet cells, and MUC2 play important roles in maintaining and protecting the intestinal mucosal barrier[38]. M3G supplementation restored the level of MUC2 by nearly double that of the DSS group. Claudin-1, occludin, and ZO-1 are TJ proteins, which is an important component of the intestinal physical barrier[39]. M3G supplementation mitigated the down-regulation of claudin-1, occludin, and ZO-1 in DSS-induced colitis mice. Wang et al. have found that Lonicera caerulea polyphenols can increase the expression levels of occludin in HFD rats[40]. Chen et al. have found that purple-red rice anthocyanins alleviated intestinal barrier dysfunction in cyclophosphamide-induced mice by up-regulating the expression of tight junction proteins[41]. iFABP can serve as a biomarker of small bowel damage in coeliac disease and Crohn's disease, and supplementation of M3G down-regulated the expression level of iFABP in colon tissue[42]. It has been reported that TFF3 alleviated the intestinal barrier function by reducing the expression of TLR4 in rats with nonalcoholic steatohepatitis, and supplementation of M3G increased the expression level of TFF3 in this research[43]. Although the effect of M3G on epithelial TJ proteins and other related proteins were limited, the beneficial effect of M3G on the colonic mucosal barrier function was supported by histological evaluation.

    The intestinal immune barrier is composed of intestinal mucosal lymphoid tissue and intestinal plasma cell secreted antibodies. Inflammatory damage in acute colitis influences the gut microbiota, epithelial barrier, and immune function in subsequent colitis[44]. Fructose can influence colon barrier function by regulating some main physical, immune, and biological factors in rats[45]. SIgA, CD4+T cells, and CD8+T cells play important roles in the functioning of the human immune system. The results indicated that M3G supplementation can maintain the colonic immune barrier function by modulating the level of SIgA, and the percentages of CD4+T cells and CD8+T cells of colon tissue.

    Notch signaling pathways are important for the maintenance of intestinal epithelial barrier integrity, and its abnormal activation is related to IBD and colon cancer[46]. Among the four Notch receptors in mammals, the most scattered in the intestine is Notch1[47]. NICD is the active form of the Notch receptor, NICD enters the nucleus, binds to the recruitment co-activator transcription complex, and then combines with the Hes gene to regulate the fate of cells[48,49]. DLL1 and DLL4 can serve as ligands for Notch signaling receptors[50]. Lin et al. found that qingbai decoction had beneficial effects on the mucus layer and mechanical barrier of DSS-induced colitis by inhibiting Notch signaling[51]. Supplementation with M3G significantly down-regulated Notch1, NICD, DLL4, DLL1, and Hes1 expression levels in the colon tissue, and there is a significant correlation between biomarkers and Notch signaling pathway-related proteins, suggesting that M3G might sustain the colonic barrier function via inhibiting the Notch signaling pathway.

    The present results showed that M3G exerted an improvement effect on the colonic mucosal barrier (physical barrier and immune barrier) function in DSS-induced colitis mice. The positive impact of M3G on the colonic mucosal barrier function may be ascribed to the modulation of permeability, stability, and integrity of the colonic mucosa by down-regulating the Notch signaling pathway. The results provide theoretical support for anthocyanins as the raw materials of functional products related to intestinal health. However, this study also has limitations, such as lacking intensive research on the effect mechanisms through cell experiments. Therefore, intensive research and clinical tests are the future direction.

    The authors confirm contribution to the paper as follows: study conception and design: Zhang C, Jiao X; data collection: Zhang L; analysis and interpretation of results: Zhang C, Zhang L, Zhang B, Zhang Y; draft manuscript preparation: Zhang C, Wang Y, Tan H, Li L, Jiao X. All authors reviewed the results and approved the final version of the manuscript.

    The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

    We would like to thank the Liaoning Provincial Department of Education Project - General Project (LJKMZ20221060).

  • The authors declare that they have no conflict of interest.

  • [1]

    Dickens JE. 2010. Overview of Process Analysis and PAT. In Process Analytical Technology: Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, ed. Bakeev KA. Second Edition. West Sussex, UK: John Wiley & Sons, Ltd. pp. 1−15. https://doi.org/10.1002/9780470689592.ch1

    [2]

    Miller CE. 2010. Chemometrics in Process Analytical Technology (PAT). In Process Analytical Technology: Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, ed. Bakeev KA. Second Edition. West Sussex, UK: John Wiley & Sons, Ltd. pp. 353−438. https://doi.org/10.1002/9780470689592.ch12

    [3]

    Cavaglia J, Schorn-García D, Giussani B, Ferré J, Busto O, et al. 2020. Monitoring wine fermentation deviations using an ATR-MIR spectrometer and MSPC charts. Chemometrics and Intelligent Laboratory Systems 201:104011

    doi: 10.1016/j.chemolab.2020.104011

    CrossRef   Google Scholar

    [4]

    Hennigan MC, Ryder AG. 2013. Quantitative polymorph contaminant analysis in tablets using Raman and near infra-red spectroscopies. Journal of Pharmaceutical and Biomedical Analysis 72:163−71

    doi: 10.1016/j.jpba.2012.10.002

    CrossRef   Google Scholar

    [5]

    Mantanus J, Rozet E, Van Butsele K, De Bleye C, Ceccato A, et al. 2011. Near infrared and Raman spectroscopy as Process Analytical Technology tools for the manufacturing of silicone-based drug reservoirs. Analytica Chimica Acta 699(1):96−106

    doi: 10.1016/j.aca.2011.05.006

    CrossRef   Google Scholar

    [6]

    Mustorgi E, Malegori C, Oliveri P, Hooshyary M, Bounneche H, et al. 2020. A chemometric strategy to evaluate the comparability of PLS models obtained from quartz cuvettes and disposable glass vials in the determination of extra virgin olive oil quality parameters by NIR spectroscopy. Chemometrics and Intelligent Laboratory Systems 199:103974

    doi: 10.1016/j.chemolab.2020.103974

    CrossRef   Google Scholar

    [7]

    Roussel S, Preys S, Chauchard F, Lallemand J. 2014. Multivariate Data Analysis (Chemometrics). In Process Analytical Technology for the Food Industry, eds. O'Donnell CP, Fagan C, Cullen PJ. pp. 7−59. https://doi.org/10.1007/978-1-4939-0311-5_2

    [8]

    Svendsen C, Cieplak T, van den Berg FWJ. 2016. Exploring process dynamics by near infrared spectroscopy in lactic fermentations. Journal of Near Infrared Spectroscopy 24(5):443−51

    doi: 10.1255/jnirs.1244

    CrossRef   Google Scholar

    [9]

    van den Berg F, Lyndgaard CB, Sørensen KM, Engelsen SB. 2013. Process Analytical Technology in the food industry. Trends in Food Science & Technology 31(1):27−35

    doi: 10.1016/j.jpgs.2012.04.007

    CrossRef   Google Scholar

    [10]

    Cobbledick J, Nguyen A, Latulippe DR. 2014. Demonstration of FBRM as process analytical technology tool for dewatering processes via CST correlation. Water Research 58:132−40

    doi: 10.1016/j.watres.2014.03.068

    CrossRef   Google Scholar

    [11]

    Rathore AS, Bhambure R, Ghare V. 2010. Process analytical technology (PAT) for biopharmaceutical products. Analytical and bioanalytical chemistry 398:137−54

    doi: 10.1007/s00216-010-3781-x

    CrossRef   Google Scholar

    [12]

    Akseli I, Mani GN, Cetinkaya C. 2008. Non-destructive acoustic defect detection in drug tablets. International Journal of Pharmaceutics 360(1):65−76

    doi: 10.1016/j.ijpharm.2008.04.019

    CrossRef   Google Scholar

    [13]

    Medendorp J, Lodder RA. 2006. Acoustic-resonance spectrometry as a process analytical technology for rapid and accurate tablet identification. AAPS PharmSciTech 7:25

    doi: 10.1208/pt070125

    CrossRef   Google Scholar

    [14]

    Mantanus J, Ziémons E, Lebrun P, Rozet E, Klinkenberg R, et al. 2010. Active content determination of non-coated pharmaceutical pellets by near infrared spectroscopy: Method development, validation and reliability evaluation. Talanta 80(5):1750−57

    doi: 10.1016/j.talanta.2009.10.019

    CrossRef   Google Scholar

    [15]

    Paris I, Janoly-Dumenil A, Paci A, Mercier L, Bourget P, et al. 2006. Near infrared spectroscopy and process analytical technology to master the process of busulfan paediatric capsules in a university hospital. Journal of Pharmaceutical and Biomedical Analysis 41(4):1171−78

    doi: 10.1016/j.jpba.2006.02.049

    CrossRef   Google Scholar

    [16]

    Wu H, White M, Khan MA. 2011. Quality-by-Design (QbD): An integrated process analytical technology (PAT) approach for a dynamic pharmaceutical co-precipitation process characterization and process design space development. International Journal of Pharmaceutics 405(1):63−78

    doi: 10.1016/j.ijpharm.2010.11.045

    CrossRef   Google Scholar

    [17]

    Moscetti R, Massantini R, Fidaleo M. 2019. Application on-line NIR spectroscopy and other process analytical technology tools to the characterization of soy sauce desalting by electrodialysis. Journal of Food Engineering 263:243−52

    doi: 10.1016/j.jfoodeng.2019.06.022

    CrossRef   Google Scholar

    [18]

    Wang X, Esquerre C, Downey G, Henihan L, O’Callaghan D, et al. 2018. Assessment of infant formula quality and composition using Vis-NIR, MIR and Raman process analytical technologies. Talanta 183:320−28

    doi: 10.1016/j.talanta.2018.02.080

    CrossRef   Google Scholar

    [19]

    Arango O, Trujillo AJ, Castillo M. 2020. Inline control of yoghurt fermentation process using a near infrared light backscatter sensor. Journal of Food Engineering 277:109885

    doi: 10.1016/j.jfoodeng.2019.109885

    CrossRef   Google Scholar

    [20]

    Nallan Chakravartula SS, Cevoli C, Balestra F, Fabbri A, Dalla Rosa M. 2019. Evaluation of drying of edible coating on bread using NIR spectroscopy. Journal of Food Engineering 240:29−37

    doi: 10.1016/j.jfoodeng.2018.07.009

    CrossRef   Google Scholar

    [21]

    Pehrsson PR, Patterson KY, Khan MA. 2014. Selected vitamins, minerals and fatty acids in infant formulas in the United States. Journal of Food Composition and Analysis 36(1):66−71

    doi: 10.1016/j.jfca.2014.06.004

    CrossRef   Google Scholar

    [22]

    Koc H, Mar MH, Ranasinghe A, Swenberg JA, Zeisel SH. 2002. Quantitation of Choline and Its Metabolites in Tissues and Foods by Liquid Chromatography/Electrospray Ionization-Isotope Dilution Mass Spectrometry. Analytical Chemistry 74(18):4734−40

    doi: 10.1021/ac025624x

    CrossRef   Google Scholar

    [23]

    Patterson KY, Phillips KM, Horst RL, Byrdwell WC, Exler J, et al. 2010. Vitamin D content and variability in fluid milks from a US Department of Agriculture nationwide sampling to update values in the National Nutrient Database for Standard Reference. Journal of Dairy Science 93(11):5082−90

    doi: 10.3168/jds.2010-3359

    CrossRef   Google Scholar

    [24]

    Saini RK, Keum YS. 2018. Carotenoid extraction methods: A review of recent developments. Food Chemistry 240:90−103

    doi: 10.1016/j.foodchem.2017.07.099

    CrossRef   Google Scholar

    [25]

    Wehling RL. 2010. Infrared Spectroscopy. In Food Analysis. 4th Edition. Boston, MA: Springer. pp. 407−20. https://doi.org/10.1007/978-1-4419-1478-1_23

    [26]

    Hof M, Macháň R. 2003. Basics of Optical Spectroscopy. In Handbook of Spectroscopy. Second Edition. Weinheim, Germany: John Wiley & Sons, Ltd. pp. 31−38. https://doi.org/10.1002/9783527654703.ch3

    [27]

    Khan A, Munir MT, Yu W, Young BR. 2021. Near-infrared spectroscopy and data analysis for predicting milk powder quality attributes. International Journal of Dairy Technology 74(1):235−45

    doi: 10.1111/1471-0307.12734

    CrossRef   Google Scholar

    [28]

    Beć KB, Grabska J, Huck CW. 2020. Near-Infrared Spectroscopy in Bio-Applications. Molecules 25(12):2948

    doi: 10.3390/molecules25122948

    CrossRef   Google Scholar

    [29]

    Fagan CC. 2014. Infrared Spectroscopy. In Process Analytical Technology for the Food Industry. First Edition. New York: Springer. pp. 73-101. https://doi.org/10.1007/978-1-4939-0311-5_4

    [30]

    Ingle PD, Christian R, Purohit P, Zarraga V, Handley E, et al. 2016. Determination of Protein Content by NIR Spectroscopy in Protein Powder Mix Products. Journal of AOAC International 99(2):360−63

    doi: 10.5740/jaoacint.15-0115

    CrossRef   Google Scholar

    [31]

    Osborne BG. 2006. Near-Infrared Spectroscopy in Food Analysis. In Encyclopedia of Analytical Chemistry: Applications, Theory and Instrumentation. New Jersey: John Wiley & Sons, Ltd. pp. 1−14. https://doi.org/10.1002/9780470027318.a1018

    [32]

    Pu YY, O’Donnell C, Tobin JT, O’Shea N. 2020. Review of near-infrared spectroscopy as a process analytical technology for real-time product monitoring in dairy processing. International Dairy Journal 103:104623

    doi: 10.1016/j.idairyj.2019.104623

    CrossRef   Google Scholar

    [33]

    Damodaran S. 2017. Amino Acids, Peptides, and Proteins. In Fennema's Food Chemistry. 5th Edition. Boca Raton: CRC Press. pp. 217−331. https://doi.org/10.1201/9781315372914

    [34]

    Damodaran S, BeMiller JN, Huber KC. 2017. Carbohydrates. In Fennema's Food Chemistry. 5th Edition. Boca Raton: CRC Press. pp. 83−155. https://doi.org/10.1201/9781315372914

    [35]

    Krimm S, Bandekar J. 1986. Vibrational Spectroscopy and Conformation of Peptides, Polypeptides, and Proteins. Advances in Protein Chemistry 38:181−364

    doi: 10.1016/S0065-3233(08)60528-8

    CrossRef   Google Scholar

    [36]

    Parker FS. 1971. Amides and Amines. In Applications of Infrared Spectroscopy in Biochemistry, Biology, and Medicine. First Edition. Boston, MA: Springer. pp. 165−72. https://doi.org/10.1007/978-1-4684-1872-9_8

    [37]

    Wiercigroch E, Szafraniec E, Czamara K, Pacia MZ, Majzner K, et al. 2017. Raman and infrared spectroscopy of carbohydrates: A review. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 185:317−35

    doi: 10.1016/j.saa.2017.05.045

    CrossRef   Google Scholar

    [38]

    Gregory III JF. 2017. Vitamins. In Fennema's Food Chemistry. 5th Edition. Boca Raton: CRC Press. pp. 439−532. https://doi.org/10.1201/9781315372914

    [39]

    Liu H, Xiang B, Qu L. 2006. Structure analysis of ascorbic acid using near-infrared spectroscopy and generalized two-dimensional correlation spectroscopy. Journal of Molecular Structure 794(1):12−17

    doi: 10.1016/j.molstruc.2006.01.028

    CrossRef   Google Scholar

    [40]

    Yang H, Irudayaraj J. 2002. Rapid determination of vitamin C by NIR, MIR and FT-Raman techniques. Journal of Pharmacy and Pharmacology 54(9):1247−55

    doi: 10.1211/002235702320402099

    CrossRef   Google Scholar

  • Cite this article

    Tristan G, Tay KH, Wong SY. 2023. Near-infrared probe as a quality control tool for milk powder blending processes. Food Materials Research 3:3 doi: 10.48130/FMR-2023-0003
    Tristan G, Tay KH, Wong SY. 2023. Near-infrared probe as a quality control tool for milk powder blending processes. Food Materials Research 3:3 doi: 10.48130/FMR-2023-0003

Figures(8)  /  Tables(1)

Article Metrics

Article views(4533) PDF downloads(638)

Other Articles By Authors

ARTICLE   Open Access    

Near-infrared probe as a quality control tool for milk powder blending processes

Food Materials Research  3 Article number: 3  (2023)  |  Cite this article

Abstract: This study aims to evaluate the suitability and reliability of Process Analytical Tools (PAT) in monitoring milk powder blending processes. The uniformity end point was predicted using a Near-Infrared (NIR) probe, and subsequently validated using offline Fourier Transform Infrared Spectroscopy (FTIR). A standard milk formulation (SMF) made up of 50% lactose, 40% skim milk powder, and 10% whey protein concentrate was used. Additionally, the detection limit of the NIR probe was investigated using vitamin C powder. The average predicted uniformity end point using the inline NIR fixed reference (63.89 ± 2.06 min), and dynamic reference conformity test (63.00 ± 5.25 min) were comparable with the offline FTIR measurement (56.6 ± 0.71 min). A three-component Partial Least Square Regression (PLSR) model was constructed and validated for vitamin C. The detection limit is 0.11%, which is higher than the vitamin C level commonly found in most infant milk formula (0.035%).

    • In industrial food processing, offline testing is normally carried out prior to batch release as part of the quality control process. Offline testing can take hours or days, leading to delays in product release. Inline PAT is an attractive alternative as data can be collected continuously, and results can be derived automatically in minutes or seconds[1]. Inline analytical sensors provide fast, automatic, and continuous chemical or physical measurements in the form of electrical signals. Using the vast data collected from the online sensors, chemometrics can be used to analyse and extract relevant patterns and information[28]. Spectroscopic sensors such as Ultraviolet-Visible spectroscopy (UV-Vis), Infrared (IR) spectroscopy, Raman spectroscopy, Fluorescence spectroscopy, etc. are often used as PAT to facilitate various research studies and experiments[9].

      PAT has been extensively implemented across many industries. In the dewatering process, precise process control using Focused Beam Reflectance Measurement (FBRM) was used to generate the relevant data to increase the yield of biosolids and energy generation[10]. In the pharmaceutical industry, PAT is widely used in both the manufacturing, and research and development process[11]. A few of the many applications in the manufacturing process include acoustic resonance spectroscopy or laser interferometric detection to rapidly and accurately identifying tablets and detect physical defects without destruction[12,13], and NIR spectroscopy to determine the active content in uncoated pharmaceutical pellets and facilitate small-scale capsule manufacturing[14,15]. In the research space, PAT has been applied across multiple platforms, especially for the characterisation of active pharmaceutical ingredients (API). Some recent examples include homogeneity testing of a silicone-based drug reservoir[5], quantification and determination of contaminants in a tablet[4], and measurement of crystal size distribution in particulate processes[16].

      Compared to the pharmaceutical industry, PAT is not as widely implemented in the food manufacturing industry. However, the implementation of PAT in the food manufacturing industry has risen in recent years. For instance, NIR spectroscopy was used to analyse the salt composition in soy sauce to facilitate the desalting process[17], Vis-NIR, Mid-infrared (MIR) and Raman spectroscopy were used to assess the quality and composition of infant milk formula[18], and NIR spectroscopy was used to investigate the composition of extra virgin olive oil[6]. In addition to composition analysis, PAT also allows precise control and monitoring of food manufacturing processes, with successes demonstrated in wine fermentation[3], yoghurt fermentation[19], and dehydration of edible coatings in bread[20].

      A large hurdle that slows down the adoption of PAT in the food manufacturing industry is the high capital investment. Therefore, the most logical step is to use PAT technology in the manufacturing of high value food products such as infant milk formula. Infant milk powder is a blend of multiple dried components and therefore, uniformity is an important process end point parameter. The existing tests for uniformity focus on offline testing of the composition of known components such as vitamins, minerals, fatty acids, etc. using High-Performance Liquid Chromatography (HPLC)[21] or other chemical tests[2224]. These offline tests are time consuming and need to be conducted by qualified personnel. In addition, it is not possible to make process end point decisions on the production floor. PAT offers continuous process monitoring with immediate test results. Online IR spectroscopy can provide results within (approx.) minutes, without the need for any sample preparation[25]. In the case of infant milk powder, the vibrational peaks due to the C-H, O-H, and N-H bonds vibrational modes can be monitored through IR spectroscopy and the concentration of the vitamins can then predicted by measuring the area under the vibrational peaks[26]. A study carried out by Khan et al.[27] demonstrated that NIR spectroscopy can effectively be used to assess the quality of milk powders. Using various data pretreatment techniques and multivariate data analysis, important milk powder qualities such as fine particle size fraction, dispersibility and bulk density could be predicted.

      In this study, the feasibility of using the NIR probe as a PAT in the milk powder blending process was investigated in two areas. Firstly, as a process monitoring tool, to predict the end point of infant milk powder blending. Secondly, as a real-time analytical tool, to measure the concentration of vitamin C in the power blend.

    • A standard milk formulation (SMF) was created to mimic the composition of infant milk powder. The SMF was made up of 50% lactose (Wee Hoe Cheng Chemical Pte Ltd, Singapore), 40% skim milk powder (NTUC FairPrice Co-operative Ltd, Singapore) and 10% whey protein concentrate (MYPROTEINTM, Manchester, UK). To test for the detection limit, vitamin C powder (MYPROTEINTM, Manchester, UK) was used. All ingredients were purchased and used within the same package.

    • A series of experiments were conducted to predict the uniformity of the milk powder blending process. To achieve this, the NIR spectra of the SMF was measured during the SMF blending process, and the end point of the was determined by the conformity test as described below. To validate the uniformity end point prediction using the NIR probe, a comparative offline measurement of a well-mixed SMF using FTIR spectroscopy was carried out as described below.

      The capability of the NIR probe in the detection of vitamin C was explored. The NIR spectra of SMF and known concentrations of vitamin C (0, 0.5%, 1.5%, 2.0%) mixtures were measured, and a PLSR model was constructed and subsequently validated with a known concentration of vitamin C (1.0%). The limit of detection of the probe was assessed by determining the minimum concentration level of vitamin C that could be predicted accurately by the PLSR model. The process is described below.

    • The schematic of the experimental setup is shown in Fig. 1. The components of the SMF were weighed separately before being transferred into the blender (Huttlin Mycromix, Bosch, Germany) and blended at 250 rpm. The NIR reflectance probe connected to the NIR spectrometer (Matrix-F NIR Spectrometer, Bruker Corporation, Massachusetts, USA) was inserted through a top opening.

      Figure 1. 

      The experimental set-up to monitor the SMF during the blending process.

    • The IR spectra collected was analysed via multivariate tools to extract relevant and useful information from the spectra. The spectra obtained were processed by 'Spectragryph 1.2.13' software (Spectroscopy Ninja, Oberstdorf, Germany). Standard Normal Variate (SNV) analysis was applied to normalise and scale the NIR spectra. For the FTIR spectra, adaptive baseline subtraction was applied to correct the baseline. The SNV works by normalizing the average of the whole spectra, while the adaptive algorithm creates a baseline that snugly fits to the bottom of the spectra, so the baseline can be effectively corrected while keeping the actual peaks.

    • For the inline monitoring of the SMF blending process, NIR spectra were collected at a resolution of 16 cm−1 and frequency of 7 s from wavenumbers 12,000–4,000 cm−​​​​​​​1. The spectra were collected for 70 min, and the room temperature was kept at an ambient temperature (25 °C).

      The end point of the SMF blending process was determined by the conformity test method[5]. The conformity test is based on mixing kinetics, which measures the degree of spectra variation from fixed references at different blending times. Protein and lactose are the key components in SMF and are used as markers of detection in the NIR spectra. The C-H, N-H and O-H vibrational modes due to lactose and protein are marked, and the absorbance values are shown in Fig. 2[2832].

      Figure 2. 

      NIR spectrum of the SMF with the wavenumber ranges that correspond to C-H, O-H and N-H bonds vibrational modes labelled.

      The conformity test method assumed that uniformity is achieved at the end of the process (70 min). As such, the last 30 spectra (66.5 to 70 min) were used as the reference points for a uniformed powder.

      The calculations to determine the uniformity end point of the blending process is summarised in Fig. 3. First, the conformity index at each specific wavenumber range (CIw) (Fig. 3a), were calculated to quantify the variation between the test spectra and the reference spectra using the following equation:

      Figure 3. 

      Summary of the conformity index analysis process. (a) NIR spectrum (at 30, 50 and 70 min) processed by SNV. (b) CIw calculated from (a) at each blending time using Eqn 1. (c) The last 30 CIw for the last 30 reference spectrum (for uniformed powder). (d) Uniformity end point occurs when 10 consecutive CIs points are zero, where CIs were calculated using Eq. 2.

      CIw=ABSref,wABStest,wSDref,w (1)

      Where CIw is the conformity index at wavelength w,ABSref,w is the average absorbance of references points at wavelength w, ABStest,w is the absorbance of test points at wavelength w, SDref,w is the standard deviation among reference's spectrums at wavelength w.

      After that, the CIw is reduced to one value (CIs) at each blending time using Eq. 2. The uniformity end point of the blending process is determined as the time when 10 consecutive CIs are consistently zero (Fig. 3b).

      CIs=CIhighn (2)

      Where CIhigh is the sum of CIw that is higher than CImax, CImax is the maximum CIw of all reference spectra, n is the number of points with CIw > CImax.

    • To validate the uniformity predicted by the NIR probe, a comparative offline measurement was conducted using a Fourier-transform Infrared (FTIR) spectrometer (Agilent Cary 630). First, a well-mixed SMF was created by blending the three components of SMF using a vortex mixer at 2,000 rpm (ZX3 Advanced, VELP Scientifica, Usmate Velatm, Italy).

      The mid-infrared spectrum of a well-mixed SMF generated from offline FTIR is shown in Fig. 4. The key characteristic peaks due to lactose and protein were identified with the wavenumbers 3,324.785, 1,643.756, and 1,535.663 cm−1, which are consistent with the literature values (3,324.785 vs 3,320 cm−​​​​​​​1, 1,643.756 vs 1,646 cm−​​​​​​​1, and 1,524.481 vs 1,515 cm−​​​​​​​1)[3337]. The wavenumber ranges of each identified peak were marked and assigned accordingly (Fig. 4). The area under the key characteristic peaks within the marked region shown in Fig. 4 were calculated using 'MATLAB 9.7 R2019b' software (The MathWorks, Inc. Massachusetts, US).

      Figure 4. 

      Mid-Infrared spectrum of a well-mixed SMF (baseline corrected) generated from offline FTIR. The wavenumber ranges of key characteristic peaks are marked.

      From the blender (Fig. 4), sampling was carried out at 10, 20, 30, 40, 50, 60, 120, 180, 210, and 240 min. At each blending time, the ratio (Rt) of the area under the O-H stretching peak (lactose) to the sum of the area under the amide I and amide II peaks (protein) was calculated. The same ratio (Rref) was calculated for a well-mixed reference, as an indication of the completion time of the blending process. When Rt is equal to Rref, the time will be considered as the uniformity end point as determined by offline FTIR. This end point was compared to the uniformity end point determined by inline NIR probe as validation.

    • To demonstrate the feasibility of vitamin C detection using the NIR probe, vitamin C was added to the SMF at varying known concentrations (0, 0.5%, 1.5%, and 2 %) in the blender (Fig. 1) and the NIR spectrum was collected. The wavenumber ranges that correspond to O-H bond vibrational modes present in vitamin C[31,3840] were labelled and identified in the NIR spectrum shown in Fig. 5. The absorbance values of the selected wavenumber ranges were then used to create a PLSR model using built-in MATLAB functions. The model was validated by fitting a known concentration of vitamin C (1%) into the model and calculating the % error.

      Figure 5. 

      NIR spectrum of the SMF with the wavenumber ranges that correspond to O-H bonds vibrational modes in ascorbic acid labelled.

      Subsequently, the detection limit of the probe with respect to vitamin C was established by adding small amounts (0.001 to 0.4%) of vitamin C sequentially into the SMF. The NIR spectrum collected were fitted into the PLSR model, and the minimum concentration level of vitamin C that could be predicted accurately by the PLSR model was determined.

    • Using the conformity test method, the average uniformity end point of the SMF blending as determined by the NIR probe was 63.89 ± 2.06 min (Table 1).

      Table 1.  The uniformity end point predicted from fixed/dynamic reference conformity test and FTIR area ratio for three experimental runs using SMF.

      Test methodsUniformity end point (min)
      Run 1Run 2Run 3Average ± SD
      Fixed reference61.7265.8064.1763.89 ± 2.06
      FTIR area ratio (interpolated)56.356.157.456.6 ± 0.71
      Dynamic reference56.9366.0366.0363.00 ± 5. 25

      The ratio, Rt, of the area under the O-H stretching peak (lactose) to the sum of the area under the amide I and amide II peak (protein) was plotted against blending time (Fig. 6). The blend was assumed to be uniform when the area ratio of the sample (Rt) falls below the uniform line, which represents the calculated area ratio, Rref. As shown in Fig. 6, the area ratio of the replicates intersects the uniform line (Rt = Rref) between 50 and 60 min and the area ratio (Rt) are constantly below the uniform line. Therefore, the average uniformity end point as determined by FTIR was 56.6 ± 0.71 min (Table 1).

      Figure 6. 

      Uniformity end point of the blending process determined from offline FTIR measurement. The uniformity end point is the time when Rt = Rref.

      The uniformity end point measured by NIR (inline, fixed frame, blender) is generally higher than the one from FTIR (offline, vortex mixing). A statistical t-test showed that the inline and offline reading are significantly different (p-value = 0.0282) at 95% confidence level. The difference, however, is of the same order of magnitude. The different mixing method and the reference spectrum at long processing time (the 'presumed' steady state) may have contributed to the higher uniformity end point from NIR probe. The fixed reference frame method looked at the difference between the current state to the final state, any variation that is introduced into the system at a later stage would change the reference spectrum, making it difficult for the current stage to be considered as the uniformity end point. The inherent system variability is also illustrated in the FTIR data (Fig. 6), after the uniformity end point (at ~57 min), the Rf fluctuates at levels below Rref, but it did not settle to a 'steady state'. In the FTIR determination, uniformity end point was determined using a fixed Rf, which is a constant.

      The results in Table 1 demonstrates the feasibility of the NIR probe as a PAT to monitor milk powder blending process. However, the uniformity end point cannot be determined inline if the reference frame's position is fixed at the end of the blending process. As such, further refinement to the definition of fixed reference frame (Fig. 3) is required to better suit inline detections and predictions.

    • To determine the uniformity end point inline, modifications to the conformity test method must be made. As opposed to having the reference points being fixed as the last 30 spectra at the end of the blending process, the reference points must be dynamic (vary with time) (Fig. 7). This also better aligns with the characteristics of an inline monitoring process using a PAT as it incorporates the variation of spectra with time.

      Figure 7. 

      Offline, fixed reference method vs. inline, dynamic reference method to determine uniformity end point.

      For inline determination of the uniformity end point, a collection of 30 spectrum and CIs values are first defined as the reference points. The variability (Vrw, Eq. 3) within the reference spectrum is examined by calculating the ratio of the sum of the average absorbance of references points (ABSref,w) and the standard deviation among reference's spectrums at each wavelength, w (SDref,w).

      Vrw=ABSref,wSDref,w (3)

      If the maximum Vrw (Eq. 3) across all wavelengths (w) is less than 3%, it can be an indication of low variability within the 30 dynamic reference points.

      Once maximum Vrw is less than 3%, uniformity is achieved if the next 10 consecutive CIs values after the 30 defined points are zero. If they are not zero, the reference points continue to move forward until the next 10 consecutive CIs values are zero. The inline determination of uniformity end point is illustrated in Fig. 7.

      The dynamic reference algorithm was deployed and applied onto the SMF the blending process. The average uniformity end point predicted using the described method was determined to be 63.00 ± 5.25 min (Table 1). A single factor ANOVA was conducted for all tests listed in Table 1. At 95% confidence level, the ANOVA tests revealed no significant difference between the three methods (p-value = 0.066). In short, the end point obtained from dynamic reference algorithm is comparable to the end point from fixed reference frame and FTIR measurements.

    • The PLSR model shown in Fig. 8 was constructed using the NIR spectrum collected at 0, 0.5%, 1.5% and 2% vitamin C levels. The PLSR model fits the data well (R2 = 0.9763). This model was also validated with the experimental data using 1% vitamin C concentration as described in the 4th sub-section of the method section. Using the NIR spectrum obtained from the experiment, the PLSR model predicted a concentration of 1.04% ± 0.18%. The predicted concentration only had a 4% error compared to the actual concentration of 1%. The validation data and the R2 (Fig. 8), demonstrated that the PLSR model performed satisfactorily, and is suitable to predict the vitamin C content in the SMF mixtures.

      Figure 8. 

      PLSR model of predicted vitamin C concentration against the actual vitamin C concentration, with measurements at lower-level concentrations of vitamin C shown.

      To establish the detection limit of the NIR probe, vitamin C was gradually dosed into the SMF, starting at a concentration of 0.001%. From Fig. 8, it is evident that the PLSR model failed to predict Vitamin C concentration level below 0.11%. At concentration level above 0.11%, the error between predicted and experimental value is generally below 2%. Below 0.11%, the error increases exponentially. This level is higher than the fortified vitamin C level in infant milk formula (approximately 0.035%)[21]. Therefore, offline trace analysis is still required for the quantification of vitamins.

    • NIR was shown to be a viable PAT that can be used to adequately predict the uniformity end point of milk powder blending process. The average predicted uniformity end point using the inline NIR fixed/dynamic reference conformity test and offline FTIR measurement were comparable. Using a series of known concentration of vitamin C, a PLSR model was constructed and validated. Overall, there is a good agreement between the predicted and actual concentration of vitamin C in the SMF. The detection limit of the NIR probe with respect to vitamin C was determined to be 0.11%, which is higher than the level in infant milk powder.

      • This work was financially supported by the Agency for Science, Technology and Research (A*STAR) under its AME Young Individual Research Grant Scheme (Project #A188c0021).

      • The authors declare that they have no conflict of interest.

      • Copyright: © 2023 by the author(s). Published by Maximum Academic Press on behalf of Nanjing Agricultural University. This article is an open access article distributed under Creative Commons Attribution License (CC BY 4.0), visit https://creativecommons.org/licenses/by/4.0/.
    Figure (8)  Table (1) References (40)
  • About this article
    Cite this article
    Tristan G, Tay KH, Wong SY. 2023. Near-infrared probe as a quality control tool for milk powder blending processes. Food Materials Research 3:3 doi: 10.48130/FMR-2023-0003
    Tristan G, Tay KH, Wong SY. 2023. Near-infrared probe as a quality control tool for milk powder blending processes. Food Materials Research 3:3 doi: 10.48130/FMR-2023-0003

Catalog

  • About this article

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return