If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The noninvasive, longitudinal study of products and food processing is of interest for the dairy industry. Here, we demonstrated that single-sided nuclear magnetic resonance (NMR) can be used for noninvasive monitoring of the cheese ripening process. The maturation of soft-ripened Camembert-like molded cheese samples was monitored for 20 d measuring 1-dimensional and 2-dimensional NMR relaxation and diffusion data at various depths, ranging from the hard surface layer to the soft center. Gelation and gel shrinkage were observed throughout ripening, and a complete loss of free water signal was observed at the cheese rind. Transversal (T2) relaxation distributions include 3 components that evolve with ripening time and position, corresponding to water inside the casein gel network, water trapped in casein, and fat. Two-dimensional T1-T2 relaxation experiments provided enhanced resolution of the 3 components, allowing quantification of the relative proportions of each phase. Furthermore, diffusion (D)-T2 relaxation correlation experiments revealed the bimodal size distribution of fat globules. The study demonstrated that single-sided NMR can provide spatially resolved signal intensity, relaxation, and diffusion parameters that reflect structural changes during the ripening process and can be exploited to understand and monitor the ripening of cheeses.
Cheese is a dairy product derived from milk that can have a range of flavors, textures, and shapes depending on source materials and production conditions (
). Cheese production for acid-set or rennet cheeses can be divided into 3 steps: milk processing, gelation, and ripening. During milk processing, lactose is fermented into lactic acid (
). Ripening is the most time-consuming part of production and is defined as the storage of cheese over time and under specific conditions. The ripening process is a series of physical, chemical, and biological changes that take place over time (
During cheese production, a cheese gel network is formed (Figure 1). Cheese is an emulsion of milk fat in a fluid phase. The fat consists primarily of triglycerides in spherical globules. At room temperature, a portion of the fat is liquid (
). In milk, casein proteins form micelles, whereas in cheese, casein proteins form a continuous porous network. During cheese ripening, this porous network of water, liquid fat, and casein evolves based on factors such as milk quality, milk type, milk enzymes, rennet or substitute, starter bacteria, mold, and temperature (
Figure 1A schematic diagram of the cheese gel network. Cheese is an emulsion of liquid fat globules in a fluid phase composed of water and casein proteins. The casein proteins form a continuous pore network.
). Transversal (T2) relaxation is sensitive to the rotational mobility of molecules, which is affected by composition and chemical structure, as well as molecular exchange (
). T2 experiments are fast (typical experiment time is on the order of minutes) and easy to automate. Diffusion experiments provide direct information about Brownian motion of molecules, reflecting molecular and aggregate sizes. Multidimensional correlation experiments improve the resolution and information content of relaxometry and diffusometry (
). For example, T1-T2 experiments enable correlation of longitudinal (T1) and transverse relaxation times, whereas D-T2 experiments correlate the molecular diffusion coefficient (D) with T2 relaxation time, allowing, for example, the identification of water and liquid fat components (
). A single-sided NMR device is one in which the magnetic field extends from the surface of the magnet. The field decreases with increasing distance from the surface, and signal is observed from a thin layer in which the Larmor frequency of spins matches with the resonant frequency of the radiofrequency coil. With single-sided NMR devices, the sample is placed on top of the magnet instead of inside the magnet, such that there is no sample size requirement or change in condition. Using such a set-up, the depth-dependent evolution of a sample can be studied (
; Figure 2). The present work used single-sided NMR for a longitudinal, depth-dependent study of soft-ripened molded Camembert-like cheeses. Such depth-dependent information is beneficial when the maturation process differs significantly depending on the region of product. In this study, we monitored the 20-d ripening process longitudinally by measuring 1H signal intensity, T2, T1-T2, and D-T2 data daily at the surface and inside the soft center of the cheeses.
Figure 2(a) Schematic diagram of the single-sided nuclear magnetic resonance (NMR)-Mouse used for monitoring cheese ripening. (b) 1H signal intensity of sample 1 as a function of ripening time at different depths (2 to 5 mm). The intensity is the amplitude of the fifth echo of the Carr-Purcell-Meiboom-Gill (CPMG) data, corresponding to total echo time of 1 ms. (c) Normalized NMR signal intensities of samples 1 and 2 and the normalized weight of sample 2 as a function of ripening time at a depth of 5 mm. In the normalization, the maximum NMR signal and weight was set to 100. S = south pole of the magnet, N = north pole of the magnet, RF coil = radiofrequency coil.
Each cheese was made from 400 mL of pasteurized milk containing 3.2% fat (Valio Oy), 2 mg of calcium chloride (Biochem s.r.l.), 2 mg of bacteria starter culture (Biochem s.r.l.), 2 mg of mold (Biochem s.r.l.; lyophilized Penicillium candidum), and 500 µL of rennet (Åströmin Jälk. Oy). The starter culture was a lyophilized mixture of Lactococcus lactis ssp. lactis, L. lactis ssp. cremoris, L. lactis ssp. lactis biovar diacetylactis, and Streptococcus salivarius ssp. thermophilus. The milk was warmed for 15 min at 30°C, and then calcium chloride was added to the milk. The mixture was stirred for 5 min with a spatula. Then, the bacteria and mold were added, and the mixture was left to stand for 15 min at 30°C. Finally, rennet was added and the mixture was left for 30 min at room temperature. The mixture was then filtered using a custom-made colander made from a plastic box with small perforations in the bottom. The remaining solid mass in the box was pressed for 12 h to drain the excess whey. Finally, the cheese was cut into 5 × 5 cm sections and salted with non-iodized salt. The samples were cured for 20 d in a room at 21°C and 15% humidity. Ripening measurements were performed on 2 different samples. Weight measurements were taken every day on one of the samples.
NMR Experiments
The NMR experiments were performed using a Magritek Profile NMR-Mouse PM 25 with an operating 1H resonance frequency of 13.29 MHz and 7.28 Tm−1 gradient, Kea 2 spectrometer, and Prospa software (Magritek). The cheese sample was set on the top of the NMR-Mouse (Figure 2a), and T2 relaxation and T1-T2 and D-T2 correlation experiments were performed at 5 positions: 5, 4, 3, 2.5, and 2 mm from the surface of the device. The 5-mm position corresponds to the soft center of the cheese sample and the 2-mm position corresponds to the outer layer (skin) of the cheese. Experiments were performed once a day for 20 d at room temperature.
The T2 relaxation data were measured using a Carr-Purcell-Meiboom-Gill (CPMG) sequence (
The choice of optimal parameters for measurement of spin-lattice relaxation times. II. Comparison of saturation recovery, inversion recovery, and fast inversion recovery experiments.
) block followed by a CPMG sequence in which data were recorded at the center of every echo (Figure 3a). Saturation was achieved using seven 90° radiofrequency pulses with an inter-pulse delay between 100 and 750 µs. The T1 recovery time (t1) was varied from 1 to 2,500 ms in 32 log spaced steps. A 200-µs echo time (2τ), 512 echoes (n), 32 scans, and 2.5-s repetition time were used. The experiment time was 44 min.
Figure 3Pulse sequences for (a) T1-T2 and (b) D-T2 correlation experiments. Δ= diffusion encoding time, δ = time between two 90° pulses, τ = echo time.
) followed by a CPMG sequence in which data were recorded at the center of every echo (Figure 3b). The diffusion encoding was achieved by exponentially increasing the time between two 90° pulses (δ) from 0.1 to 5 ms with 32 steps. The diffusion delay (Δ) was 5 ms. The number of echoes in the CPMG was 512, the echo time was 200 µs, the number of scans was 32, and the repetition time was 2.5 s. The experiment time was 43 min. The signal (S) observed in the D-T2 is given as follows:
[2]
where γ is the gyromagnetic, g is the gradient strength, δ is the time between 90° pulses, Δ is the diffusion encoding time, and P(D, T2) is the D-T2 correlation probability distribution.
Data Analysis
1H signal intensity was defined as the signal at the fifth echo of the CPMG sequence corresponding to S(t = 1 ms). The T2 probability distribution, P(T2), was calculated from CPMG data using a 1-dimensional (1D) inverse Laplace transform (ILT) with ITAMeD implementation (
). Complete sets of D-T2 and T1-T2 maps with all depths are shown in the supplemental data (Supplemental Figures S1 and S2, respectively; https://doi.org/10.5281/zenodo.7197359;
). The center and area of the 2D peak was defined by first fitting the 2D Gaussian function to each peak. The center of the peak was extracted directly from the model, and the area was calculated by integrating the fitted Gaussian function.
The results were further compared with another 2D ILT algorithm, Upen2DTool (
) based on the uniform penalty principle, to confirm the validity of the results (shown in Supplemental Figure S3; https://doi.org/10.5281/zenodo.7197359;
Cheese ripening was studied for the first time longitudinally in 2 soft-ripened molded Camembert-like cheese samples (called samples 1 and 2), at several depths, without alteration to the samples. These measurements were made possible by using single-sided NMR because it can measure samples of any size. Based on previous research, the obtained D, T1, and T2 peaks were assigned to water, fat, and a protein-associated peak (
). There is a long history of using NMR relaxation to identify these peaks, starting in 1958 when the first NMR study of milk was published, in which the authors identified 3 distinct components present in milk, water, lactose, and fat (
In the study of cheese ripening, the effect of gel shrinkage and water evaporation is first considered using 1H signal intensity. 1H NMR signal intensity of sample 1 evolved during maturation as a function of ripening time and distance from the surface (Figure 2b). As the T2 relaxation time of nonexchangeable protons of macromolecules (20–30 µs;
) is much shorter than the 200-µs echo time, the proton signal intensity reflects the total amount of mobile water, fat, and exchangeable protons. In the soft center 5 mm from the magnet surface, the intensity decreased continuously over the 20-d period, predominantly due to water evaporation, and the total signal decrease was about 30%. In the outermost layer, at a depth of 2 mm from the magnet surface, the signal intensity at the beginning of the ripening process was about 36% less than in the center layer, and the signal vanished on d 8 due to formation of the dry skin layer. The drying rate and overall signal depression decreased with increasing depth. The loss of signal intensity is a strong indicator, perhaps the most unambiguous indication of cheese ripening in cheeses that form a rind. When the signal is lost at the surface of the cheese, it implies that the rind is forming, which is a critical step in the ripening of this type of cheese. Additionally, the depth-dependent measurements of signal showed development of a thick rind. The total decrease in NMR signal intensity (13%) in the soft center (depth 5 mm) of sample 2 during the ripening period was smaller than that of sample 1 (30%; Figure 2c). We interpret this to reflect the slower ripening of the former sample; after ripening, sample 2 was softer than sample 1, supporting this interpretation. Both the normalized NMR signal and normalized weight of sample 3 decreased during the ripening process (Figure 2c), similarly and qualitatively reflecting the drying of cheese. However, the former decrease was smaller because the NMR signal reflects spin density of mobile protons, whereas mass reflects the overall number of molecules in a sample, which depends on both density and volume of the sample.
To gain information about the evolution of the heterogeneous cheese structure (
; Figure 1), the NMR parameters T2, T1, and D are used. T2 is affected by the porous structure and exchange processes as a result of water–macromolecule interactions (
). In this study, T2 relaxation was measured as a function of cheese ripening time and showed 3 components (Figure 4). In the soft center of the cheese 5 mm from the magnet surface, the T2 distributions included a dominant peak around T2 = 30 ms and 2 smaller peaks around T2 = 10 and 100 ms. The T2 relaxation times decreased with increasing ripening time. This is interpreted to be a consequence of ripening, gel shrinkage, and water evaporation, which increase interactions between the remaining water molecules and proteins, leading to shortened T2, for example, due to enhanced chemical exchange. The examination of all 3 peaks has been reported in several studies (
). The shortest relaxation time peak, peak P (Figure 4), was ascribed to water trapped within casein, as its T2 relaxation time (about 10 ms) is close to the T2 values of the 2 casein-associated water peaks (7.2 and 16.2 ms) observed by
A low-resolution and high-resolution nuclear magnetic resonance integrated approach to investigate the physical structure and metabolic profile of Mozzarella di Bufala Campana cheese.
for mozzarella cheese. Observation of only a single casein-associated peak may be a consequence of different structure of the Camembert-like cheese, accelerated exchange between the 2 sites due to the higher measurement temperature in our studies, lower resolution in our experiment, or an effect of diffusion in the constant gradient field on the apparent observer T2 (see below). The dominant intermediate T2 peak, labeled W, was attributed to highly mobile water present in the gel structure of the cheese. The high mobility was confirmed by the D-T2 measurements described below. The T2 of the water peak (about 30 ms) was much shorter than that of corresponding peak (488 ms) observed in mozzarella cheese by
A low-resolution and high-resolution nuclear magnetic resonance integrated approach to investigate the physical structure and metabolic profile of Mozzarella di Bufala Campana cheese.
Partial substitution of 40 g/100 g fresh milk with reconstituted low heat skim milk powder in high-moisture mozzarella cheese production: Rheological and water-related properties.
). This is an effect of fast diffusion of free water in the strong constant field gradient of the single-sided instrument used in the current study. In the presence of constant gradient g, the observed apparent relaxation time T2app in the CPMG experiment can be calculated using the following equation (
A low-resolution and high-resolution nuclear magnetic resonance integrated approach to investigate the physical structure and metabolic profile of Mozzarella di Bufala Campana cheese.
, and use values relevant for our experiments (τ = 100 μs, g = 7.28 T/m, D = 2⋅10−9 m2/s), Eq. [3] gives T2app = 36 ms, which is in good agreement with the T2 of peak W observed in our experiments. The peak with the longest relaxation time was assigned to signal from protons in the lipid phase, F (
A low-resolution and high-resolution nuclear magnetic resonance integrated approach to investigate the physical structure and metabolic profile of Mozzarella di Bufala Campana cheese.
). These peaks showed the same T2 relaxation time in the center of the cheese and at the skin before formation of the skin layer. We noted that the effect of diffusion on the apparent T2 was smaller for the fat- and casein-associated water signals due to their shorter “true” T2 and slower diffusion (see Eq. [3]). During the maturation process, T2 values of all peaks of sample 1 decreased significantly more than those of sample 2 [compare Figure 4 with Supplemental Figure S4 (https://doi.org/10.5281/zenodo.7197359;
)] due to the slower ripening of sample 2. Consequently, T2 is another sensitive parameter reflecting the degree of maturation in addition to the NMR signal strength (see above).
Figure 41H T2 distributions of sample 1 as a function of ripening time. P represents water molecules trapped in casein protein, W indicates the signal of water inside the mesh of the casein gel network, and F represents the fat signal. The drying times of different layers are illustrated by black dashed lines. Drying time was defined as a signal to noise ratio <30 and reflects the rate at which different depths of the cheese rind ripened.
Determination of water self-diffusion coefficient in complex food products by low field 1H PFG-NMR: Comparison between the standard spin-echo sequence and the T1-weighted spin-echo sequence.
performed a quantitative analysis of fat and water in fatty protein concentrate and showed that T2 of the fat proton covers a wide range, which is overlaid with the water proton relaxation time. Moreover, dairy products consist not only of fat and water, but also contain protein and sugar and, depending on the preparation, the system may become more complex. To address this issue, 2D NMR correlation experiments can be used, such as T1-T2 and D-T2 methods. In correlation experiments, the data are resolved using 2 correlated variables, which can improve peak separation and identification. In T1-T2 correlation experiments, the T1:T2 ratio can be identified for each component, which reflects mobility, interactions, and the state of molecules (
). In free liquids, T1 = T2, whereas in solid-like materials, T1 is much longer than T2. On the other hand, as described above, the strong, constant gradient of the NMR-Mouse coupled with molecular diffusion accelerates the signal decay during the CPMG loop (
), resulting in measurement of a faster, diffusion-weighted T2 relaxation rate and subsequently higher T1 or T2 relaxation rate for quickly diffusing liquids but not slowly diffusing liquids such as fat.
The T1-T2 correlation map of sample 1 measured on d 1 of ripening (Figure 5a) included 3 signals. One signal (W), around T1 = 320 ms and T2 = 60 ms, was interpreted to arise from water. The second signal (P) was interpreted to be water trapped in casein, with T1 = 160 ms and T2 = 2.4 ms. The T1:T2 ratio is high, about 65, because chemical exchange is a particularly efficient relaxation mechanism for T2 relaxation (
). The fat peak was characterized by liquid-like behavior with a T1:T2 ratio ≈ 1; therefore, we interpret that fat is predominantly in liquid form in the cheese. The integrals of the peaks in the T1-T2 maps (Figure 5d) indicate that the relative amount of water decreased by about 20% during the 20-d ripening period, whereas the protein and fat peaks increased as a function of ripening time. The peak assignment used here is in agreement with previous studies (
). In addition to the 3 peaks discussed above, the T1-T2 maps include some additional peaks, which were interpreted to be artifacts arising from experimental noise. Such artifactual peaks are unavoidable due to the numerically unstable nature of ILT (
). The peaks were identified to be artifacts because their amplitude was small, they were not reproducible, and some of them were not physically acceptable due to having a longer T2 than T1.
Figure 5T1-T2 maps of sample 1 measured from the center layer at the depth of 5 mm at (a) d 1, (b) d 10, and (c) d 20. The blue dashed lines represent T1 = T2. (d) Integrals of the signals as a function of the ripening time; and (e) T1 and (f) T2 relaxation rates of the water, protein, and fat peaks. The shaded areas represent the standard deviations of the measurements of samples 1 and 2, and the values presented in panels (d)–(f) represent the mean values of the performed measurements. P represents water molecules trapped in casein protein, W indicates the signal of water inside the mesh of the casein gel network, and F represents the fat signal.
Another NMR method that can be used to understand the ripening process is D-T2; it is well suited for quantifying the fat and moisture content of food products (
). The diffusion coefficient of free water (D) = 2 × 10−9 m2·s−1 at room temperature; D decreases as molecular size increases, such as for protein or fat, and when mobility is restricted by the structure (
). During ripening, 3 peaks were observed in D-T2 maps (Figure 6). The dominant peak with the highest diffusion coefficient (around 10−9 m2·s−1, labeled “WP”) was interpreted to arise from water. We anticipate that the signal of water trapped in casein was fused with the water signal in the D-T2 maps due to their similar D and T2 values; the exchange between the casein-trapped water and gel pore water pools are expected to be relatively fast in the diffusion delay time scale, leading to similar D values for both sites. The D associated with the WP peak decreased as ripening progressed due to more restricted diffusion in the casein gel network. The 2 peaks with significantly lower diffusion coefficients (<10−10 m2·s−1, labeled F1; and <10−11 m2·s−1, labeled F2) are interpreted to originate fat globules, which are known to have D values around that range (
). The observation of 2 fat peaks indicates 2 distinct fat globule sizes The differentiation of 2 types of fat globules confirms a previous confocal laser scanning microscopy study of Camembert-type cheese (
). We expect that diffusion of fat molecules was restricted by the size of the fat globules. The mean square displacements z = (2ΔD)1/2 corresponding to D = 10−10 and 10−11 m2·s−1 (Δ = 5 ms) were 1 and 0.3 μm, respectively. Therefore, we interpreted that the lower diffusion coefficient peak F1 corresponded to smaller fat globules with the size <1 μm and the F2 peak arose from larger, >1 μm, fat globules. Similar size ranges have been observed by microscopy (
). The relative intensity of the fat signal increased with ripening time, similarly to that of the T1-T2 experiments. The D values of both fat peaks remained quite constant over the 20-d ripening period, indicating that the fat globule sizes were stable (
Figure 6D-T2 maps of sample 1 taken from the center layer at the depth of 5 mm at (a) d 1, (b) d 10, and (c) d 20; and (d) integral, (e) diffusion coefficient (D), and (f) T2 relaxation times of the water and fat peaks. The shaded areas represent the standard deviation of the measurements of samples 1 and 2, and the values presented in panels (d)–(f) represent the mean values of the performed measurements. P represents water molecules trapped in casein protein, W indicates the signal of water inside the mesh of the casein gel network, and F represents the fat signal.
In this work, we present a longitudinal study of cheese ripening using single-sided NMR. By studying ripening at the surface and inside the cheeses, the evolution of cheese gel structure was studied by following water signals associated with aqueous water, liquid fat, and casein protein. The skin formation was visible in the outermost layers (2–3 mm) and could be differentiated from the soft-ripening process in the central layers (4–5 mm) using 1H signal intensity. The overall decrease of signal in the central layer was interpreted to reflect the degree of ripening. Different components of cheeses were identified using 1D T2 relaxation and 2D T1-T2 and D-T2 correlation measurements. By monitoring the ripening process, the mobile water loss and gel shrinkage were observed, whereas the proportion of casein-associated water and fat increased. Furthermore, information was gained about the fluidity, sizes and stability of fat globules with 2 distinct sizes during ripening. Overall, this study shows the potential for single-sided NMR to be valuable tool for studying cheese ripening noninvasively.
ACKNOWLEDGMENTS
Financial support from the European Research Council (Project number 772110), Marie Sklodowska-Curie Actions (grant no. 896824), Academy of Finland (Helsinki, Finland, grant nos. 321701 and 340099), National Science Centre, Poland (Kraków, Poland, grant no. 2021/41/B/ST4/01286), the University of Oulu (Kvantum Institute, Oulu, Finland), and the KAUTE Foundation (Espoo, Finland) is gratefully acknowledged. Part of the work was carried out with the support of the Centre for Material Analysis, University of Oulu, Finland. All experimental data and Supplementary Information are deposited in Zenodo (
The choice of optimal parameters for measurement of spin-lattice relaxation times. II. Comparison of saturation recovery, inversion recovery, and fast inversion recovery experiments.
A low-resolution and high-resolution nuclear magnetic resonance integrated approach to investigate the physical structure and metabolic profile of Mozzarella di Bufala Campana cheese.
Determination of water self-diffusion coefficient in complex food products by low field 1H PFG-NMR: Comparison between the standard spin-echo sequence and the T1-weighted spin-echo sequence.
Partial substitution of 40 g/100 g fresh milk with reconstituted low heat skim milk powder in high-moisture mozzarella cheese production: Rheological and water-related properties.
00033-4&id=gr1.jpg){kind=link}
00033-4&id=gr2.jpg){kind=link}
00033-4&id=gr3.jpg){kind=link}
00033-4&id=gr4.jpg){kind=link}
00033-4&id=gr5.jpg){kind=link}
00033-4&id=gr6.jpg){kind=link}