The titrimetric method specified in AOCS method Ca 5a-40 (American Oil Chemists’ Society, 2017) was used for the FFA determination of the products. Acidity was expressed as percentage of oleic acid.

MAG, DAG and TAG contents of the structured lipids were determined according to the AOCS Cd11C-93 (American Oil Chemists’ Society, 2017) method by column chromatography. Sample was dissolved in chloroform and transferred to the column by washing with chloroform three times. Then, 250 mL of 10, 25 and 100% (v/v) diethyl ether in petroleum ether were used for the elution of TAG, DAG and MAG fractions, respectively. The fractions were collected separately in a flask and solvents were evaporated in a rotary evaporator (Laborato 4000 Heidolph, Germany) at 50 °C. The flasks were dried until constant weight. Percentages of mass fractions were calculated for each part.

The oxidation induction time was determined with Rancimat apparatus (873 Biodiesel, Metrohm, Switzerland). The sample was placed inside the glass reaction vessel for the measurement. The carrier medium was deionized water and the reaction temperature was set to 120 °C for both columns with a constant 20 L/h air flow. Stability was expressed as the oxidation induction time (h).

The fatty acid composition of the samples was determined after converting them into their corresponding fatty acid methyl esters (FAME) according to a procedure provided by IUPAC (1987). Chromatographic analyses were performed with a GC (Agilent 6890) equipped with an auto-sampler, a split/splittless (1:50) injector and an FID detector. An HP 88 capillary column (100 m x 0.25 mm ID x 0.2μm) was used in the analyses. Conditions for GC analysis are described in a previous study (Meng et al., 2010). A 37-component mixture of FAME (Sigma) was used as the standard.

Polymorphic forms of the fat crystals were determined with X-ray diffraction (Philips, Holland) using Cu as anode material (k = 1.54056 Å, voltage = 45 kV, tube current = 40 mA, fixed 1.0, 1.0, and 0.76-mm divergence, anti-scatter and receiving slits). Samples were scanned from 4 to 50° (2θ scale) at a rate of 2.0°/min. The analyses were performed at ambient temperature.

The melting points of the structured lipids were measured with a differential scanning calorimeter (DSC, Q10 TA Instruments, Crawley, UK). All samples were conditioned at 4 °C for 24 h prior to measurements. Samples were placed in hermetically sealed aluminum pans and DSC analyses were carried out from 20 to −40 °C and from −40 to 80 °C at a scan rate of 10 °C/min with respect to an empty pan (Rodriguez et al., 2001). Data analysis was performed with DSC TRIOS software (TA Instruments, Crawley, UK).

Sample consistency was determined with a penetration test using a 45^o acrylic cone fitted to a constant speed texture analyzer (TA.XT plus, UK). Samples were conditioned at 60 °C in an oven for complete melting of the crystals. Tempering was allowed to occur for 24 h in a commercial refrigerator (4 °C), then for 24 h in an oven with controlled temperature (4, 10, 15, 25 °C). Test parameters were penetration depth of 0.4 cm with 0.2 cm/s speed for 5 s testing time (Silva et al., 2009). Consistency was calculated as “yield value” (Haighton, 1959).

The solid fat content of lipids was determined by a nuclear magnetic resonance (NMR) spectrometer (Bruker, USA) according to AOCS Official Method Cd 16b-93 (American Oil Chemists’ Society, 2017). Samples were melted at 80 ^oC and recrystallized at 0 ^oC for 30 min. Then, they were stabilized for 30 min at various temperatures (10, 20, 30 and 35 ^oC) before measuring the liquid signal.

The results from the chemical analysis of the produced interesterified samples are listed in Table 1. As can be seen from the table, the FFA content in the tallow is 1.15% while the blends without interesterification have an acidity range of 0.46-1.5%. In general, the FFA percentages of the interesterified lipids (1.08-4.12%) were higher compared to the starting blends. This result is in accordance with previous studies, where an increase in FFA content was observed after chemical interesterification (Hoshino et al., 2004 ; Kowalska et al., 2005). There are some fluctuations among the samples depending on the catalyst concentration; although 1% catalyst concentration led to the formation of structured lipids with higher FFA%, especially for the corn oil-tallow samples. With the increse in catalyst concentration, there would be more FFA released from TAG, DAG and MAG structures. The correlation graphic of DAG + MAG and FFA also supports this result (Figure 1). It was also reported that the higher amount of catalysts in the reaction medium caused the formation of higher amounts of FFA and MAG + DAG, and a lower TAG content (Ledochowska and Wilczynska, 1998). ANOVA was used to investigate the effects of the factors on the chemical parameters of the structured lipids (Table 2). As the ANOVA table indicated, catalyst concentration is the most important factor affecting FFA value and the use of a chemical catalyst at higher concentrations resulted in higher FFA%.

The TAG content in tallow is approximately 98% while the blends without interesterification have lower TAG% (81.24-96.29%) values (Table 1). Both blending and chemical interesterification caused an increase in MAG and DAG contents in the samples. The TAG content of the interesterfied lipids varied between 52.62-89.76% and all interesterfied fats containing canola oil had lower TAG contents compared to their starting blends. Fluctuations in TAG contents in the samples exist depending on the catalyst concentrations; although 0.875% CH₃NaO concentration mostly led to the formation of the structured lipids with higher TAG%. As expected, the structured lipids with low TAG contents had higher DAG and MAG values. A correlation between FFA and MAG + DAG contents in the interesterified lipids was evaluated in order to better understand the changes in TAG, DAG and MAG contents after the chemical interesterification reaction (Figure 1) and the ‘r’ value was calculated as 0.69. Although this value was not especially high, there was still an increasing trend between FFA and MAG + DAG contents in the samples. Generally, samples having higher amounts of MAG + DAG content, also have higher FFA%. Therefore, the increase in both FFA% and MAG + DAG% could be associated with the activity of the chemical catalyst that snatched the fatty acids from their original place and then random distribution of the fatty acids were provided in the TAG backbone of the newly structured lipids. According to ANOVA, models developed for DAG and MAG were significant with non-significant lack of fit at 95% confidence interval (Table 2). Examination of the significance levels of the main factors and their interactions shows that the catalyst concentration, the oil type and the blend ratio did not significantly affect the TAG% of the chemically interesterified lipids (Table 2). The ANOVA table confirms the significance of the oil type (safflower) for the model of DAG%. Generally, the structured lipids interesterified with the safflower oil have lower DAG% values compared to the samples interesterified with the other oil types. Additionally, the interaction between the oil type (safflower) and the blend ratio have some significance (p ≤ 0.05) for the DAG content model. Oil type, particularly corn and safflower oils, and their interactions with the blend ratio have important effects on the MAG% of structured lipids. Increasing the blend ratio (increasing tallow amount in the blend) resulted in an increase in the MAG contents for corn oil; however, MAG contents decreased slightly with increasing blend ratio for the interesterified lipids with safflower oil according to the interaction plot (not shown).

The oxidation induction time of the tallow was 4.81 h while the non-interesterified blends showed a range of induction times from 7.98-12.25 h (Table 1). In general, the oxidation induction times of the interesterified samples decreased compared to their starting blends. Among all the samples, the tallow interesterified with corn oil had the highest oxidation induction times (3.82-12.25 h). Varying the catalyst concentration resulted in ups and downs in the oxidative stability of the samples. However, 1% CH₃NaO concentration mostly led to the formation of structured lipids with low oxidation induction times regardless of the blend ratio. There was a drastic decrease in the oxidation induction times of the samples produced with safflower oil after the CI process compared to tallow itself. This result is in accordance with the previous studies, which also observed a decrease in oxidative stability after the chemical interesterfication of beef tallow interesterified with rapeseed oil (Hoshino et al., 2004 ; Kowalska et al., 2007; Kowalski et al., 2004). The decrease in oxidative stability could be due to the high linoleic acid and low tocopherol content of safflower oil (240-670 mg/kg) (Codex, 2011). On the other hand, some of the tallow samples interesterified with canola oil had better oxidative stability than tallow itself. The presence of tocopherols in canola oil can be associated with longer induction times. According to the literature, canola oil could contain tocopherols up to in the range of 430-2680 mg/kg, which could have an effect on the improvement in the oxidative stability (Codex, 2011). The ANOVA table reveals that the blend ratio and the catalyst concentration were not significant for oxidative stability, meaning that neither factor affected the oxidative stability of the structured lipids (Table 2). The model showed that oil type was the only significant factor. Generally, corn oil samples have better oxidative stability compared to the other oil types while products which contain safflower oil are the least stable ones against oxidation.

The major fatty acid in safflower and corn oils are linoleic acid (75.12 and 54.59%, respectively) while canola oil is rich in terms of oleic acid (57.82%), a monounsaturated fatty acid. The main fatty acids in tallow, on the other hand, are palmitic (22.50%), stearic (32.10%) and oleic (36.20%) acids. The MUFA, PUFA and SFA contents in the blends of interesterified products were in the ranges as determined by their blend ratio and intial compositions of tallow and oils without any significant changes as expected and this was also confirmed by ANOVA (Table 1). As in previous studies (Meng et al., 2011), the chemical interesterification of of tallow did not result in the formation of significant amounts of trans fatty acids. Generally, the amount of fatty acids in the trans form is less than 1%, except for samples which contain canola oil (Table 1). Safflower and corn oils themselves have trans fatty acid contents of less than 1% and their interesterified forms have slightly higher percentages of trans fats. Canola oil used in blending with tallow, on the other hand, has a higher content of trans fatty acids (2.5%) compared to the other oils and the interesterified samples containing canola oil have lower trans-fat contents compared to the oil itself. Since high temperature treatments are applied during canola oil refining mostly in the deodorization step to eliminate the intense bad odor of this oil, higher amounts of trans fatty acids form during this process. The model constructed for the trans fatty acids of interesterified fats showed that catalyst concentration and oil type and interaction between oil type and catalyst concentration were the significant factors. Due to higher trans fatty acid content in canola oil itself, interesterified products containing canola oil also had higher trans fatty acid contents with respect to other samples and this is the reason for the differences between trans fatty acid contents in fats containing different oils. In addition, increasing catalyst concentration resulted in an increase in trans fatty acid contents in fats which contain canola oil according to statistical analysis. However, most of the products are within acceptable limits.

Since many parameters were determined in the study, chemical data were combined and analyzed with a multivariate statistical analysis technique, principal component analysis (PCA), to obtain a better view of the process. The model was constructed by using all measured chemical parameters with 5 principal components (PC), R²=0.84, and Q²=0.42. There was a clear discrimination among the samples with respect to oil type (Figure 2a). While the samples containing canola oil were located at the right part of the ellipse samples with corn oil were placed just right of the center and safflower-containing ones were further to the left. Therefore, a discrimination with respect to the first principal component (PC) was obtained for the oil type. This discrimination mostly resulted from the higher trans fatty acid and MUFA contents in canola oil samples as observed in the loading plot (Figure 2b). In addition, the structured lipids containing 40% safflower and corn oils were placed at the bottom part of left quartile due to their higher PUFA contents, especially linoleic acid content. Moreover, groupings among the samples with different blend ratios were observed. The samples with 80% tallow were mostly located at the top of the ellipse regardless of the oil type since they presented higher saturated fatty acid (SFA) contents (Figure 2a). The samples containing 60% tallow were in the lower part of the ellipse and 70% tallow-containing samples were placed in the middle. Therefore, a separation based on the blend ratio was also possible with respect to the second PC. As a result, multivariate analysis of the chemical data indicated that the type of oil and the blend ratio caused differences in the chemical properties of the interesterified products.

The physical properties of the samples are provided in Table 3. As far as the crystal morphology is concerned, tallow contained mixtures of β and β’ forms. α forms were not found in either structured lipids nor blends. The non-interesterified blends also contained both β and β’ forms together. After the chemical interesterification, β and β’ forms were also mostly present together especially for the structured lipids containing safflower and corn oils. Therefore, the interesterification did not cause important changes in the polymorphic structures of the lipids produced from safflower and corn oils. This result was in accordance with previous studies related to beef tallow-palm oil interesterified fats and other similar products (Lee et al., 2008 ; Meng et al., 2010). In general, a lower tallow-oil blend ratio and lower catalyst concentration combination (CA61, CA62, CA71) resulted in the formation of only β crystals while a higher blend tallow-oil ratio and catalyst concentration combination (CA72, CA81, CA82, CA83) caused the formation of β’ crystals alone for canola oil-containing samples (Table 3). SA83, along with CA72, CA81, CA82, and CA83 were the samples which contained only the β’ polymorphic form and this crystal form is important in the baking industry due to its aeration properties and smooth texture. Therefore, these lipids can be used as alternatives for bakery fats.

The melting point range of the tallow was quite high (46.5-49.5 °C) and there were small decreases in the melting points of the samples due to blending the tallow with different oils. However, after the interesterfication sharp decreases in the melting points of the structured lipids were observed regardless of the oil type. These changes in the melting points were in accordance with the results of previous studies (Meng et al., 2011; Morselli Riberio et al., 2017). All samples containing 60% tallow presented lower melting points regardless of the oil type. When the tallow ratio was raised from 60 to 80%, a clear increase was also observed in the melting points. The β’ form of crystals has a high melting point between 17-69 °C and the melting point of β form is 32-78 °C depending on the chain length of the fatty acids (Martin et al., 2010). The interesterified fats were more likely to have β and β′ crystal types together and the melting points of the samples were relevant to the melting point of the crystal types. Therefore, it was clear that polymorphic structure of the interesterified lipids was highly related to the melting points of the structured lipids. The ANOVA results for melting point (Table 4) show that the model is significant at p < 0.05 with non-significant lack of fit. Blend ratio had the most prominent effect on the melting points of the chemically interesterified lipids. Melting points reached maximum values when the blend ratio of tallow was 80%. Although not to the extent of blend ratio, catalyst concentration had some significance on the melting points of the structured lipids and a slight decrease in melting points was observed with increasing catalyst concentration.

The consistency of all the samples decreased clearly as a function of temperature (Table 3). This result can be associated with the gradual melting of the crystals that generated more fragile crystalline networks. The same behavior was observed in previous studies (Bezerra et al., 2017; Oliveria et al., 2017; Silva et al., 2009). The consistency of tallow (69.85-385.93 MPa) was quite higher than both the interesterified lipids and the non-interesterified blends at all temperatures (Table 3). The consistency of the blends in different proportions increased with increasing amounts of tallow in the blends. However, the interesterified lipids presented lower consistency values compared to their physical blends regardless of the oil type and the catalyst concentration. This decrease in the consistency of the interesterified lipids could be attributed to higher amounts of unsaturated fatty acid composition in all positions of (UUU) TAGs produced by interesterification. In addition, differences in polymorphic structure and aggregation behavior which led to the alteration in the structure of the fat crystal network of tallow could change the consistency (Silva et al., 2009). Generally, fats with consistency of 9.8-98 MPa are considered spreadable. If the consistency is between 19.6 and 78.4 MPa products are more suitable for plastic and spreadable purposes and they are classified as very hard at above 147 MPa (Haighton, 1959). Most of the samples interesterified with canola oil could be considered as spreadable and plastic. The consistency values decreased to the levels suitable for spreadability with the increasing temperature. However, the consistency of the samples with canola oil of high blend ratio and high catalyst concentration (CA82 and CA83) were very high at 4 °C, and these lipids could be classified as hard. Moreover, these structured lipids contain β’ polymorphic forms which have higher melting points. The consistency of most of the samples interesterified with safflower oil was not measurable at 25 °C; therefore, these structured lipids were viscous at ambient temperature. Moreover, the structured lipids with safflower oil had lower melting points which also explained the lower consistency values of these lipids. Interesterified samples produced with corn oil also resulted in products with mostly spreadable and plastic properties. The statistical analysis results for the consistency at all temperatures indicated that constructed models were insignificant at 4, 10 and 15 °C at p < 0.05 (Table 4). Although the models were found insignificant, the ANOVA table reveals that the blend ratio had an important impact on the consistency at these temperatures (Table 4). The model for consistency at 25 ºC was significant with non-significant lack of fit. The blend ratio and the catalyst concentration were the most prominent factors for this model and higher blend ratio meant higher consistency while the opposite effect was true for the catalyst concentration. In addition, oil type, especially safflower oil, highly affected the consistency of the structured lipids at 25 °C since safflower oil-containing interesterified lipids had lower consistency.

The solid fat contents of both the interesterified lipids and the non-interesterified blends were determined over the temperature range of 10-35 °C (Table 3). As expected, it was observed that raising the temperature caused a marked decrease in solid fat content regardless of the oil type, the blend ratio or the catalyst concentration. The solid fat content profiles of non-interesterified blends in different proportions showed increasing trends with the increasing amounts of tallow in the blends. Interesterified lipids tended to have lower solid fat content values compared to their physical blends. Same trends were also observed in previous studies (Karabulut et al., 2004; Meng et al., 2010). A decrease in the solid fat content of the interesterified lipids could be attributed to the decreased proportion of the high-melting TAGs and médium-chain TAGs in the structured lipids. This decrease in solid fat content with respect to the increase in temperature was expected and has been reported in other studies (Bezerra et al., 2017; Fauzi et al., 2013; Oliveria et al., 2017). In addition, the decrease in solid fat content in tallow and non-esterified blends could be associated with the alteration in the TAG structure caused by the interesesterfication and the melting temperature of crystals. However, a more plastic behavior was observed for both blends and structured lipids above 20 °C. The ANOVA table revealed that the blend ratio is the only effective factor for solid fat content in the temperature range of 10-35 °C. The solid fat content in the samples increased regardless of the oil type when the amount of tallow increased for all temperatures (Table 4).

The principal component analysis was also used to investigate the data pertaining to the physical properties of the interesterified lipids. The model was constructed by using all the measured physical parameters with 2 principal components, R²=0.85, and Q²=0.69. There is some discrimination among the samples with respect to the blend ratio (Figure 2c). Samples containing 60% tallow were located at the left part of the quartile while the structured lipids with 80% were placed to the right of the quartile. Samples with 70% tallow generally placed between them. Therefore, there was a rough discrimination among the samples depending on their blend ratio as far as the physical properties are concerned. Discrimination between 60% and 80% tallow-containing samples mostly resulted from the higher solid fat contents and melting points of samples as observed in the loading plot (Figure 2d). Moreover, some of the samples with 80% and 70% tallow were grouped at the bottom of the ellipse regardless of the oil type since they had higher consistency (Figure 2d). The PCA score plot of the physical properties data did not show any separation with respect to the oil type or the catalyst concentration.

In order to better characterize chemically interesterified lipids, a PCA model was also constructed using all data including both chemical and physical properties of the 5 PCs, R²=0.80, and Q²=0.45. According to the score plot, there was a clear separation of the samples with respect to oil type (Figure 3a). Samples containing canola oil were located at the right part of ellipse while the interesterified fats with safflower oil were placed at the left part of ellipse. Samples containing corn oil were placed between them. Discrimination of canola oil-containing samples mostly resulted from the higher trans fatty acids and MUFA contents and oxidative stability of the samples as observed in the loading plot (Figure 3b). Moreover, canola oil samples were grouped together in between them with respect to their blend ratios since the amount of saturated and unsaturated fatty acids changed by increasing the ratio of tallow (Figure 3b). The interesterified lipids containing safflower oil were separated from other lipids due to their higher PUFA and MAG contents. Corn oil samples were differentiated from the rest since they had higher consistency values as the loading plot confirmed. Groupings in between the interesterified samples according to blend ratio were also observed. The results of the PCA models were in accordance with ANOVA results. Generally, the catalyst concentration did not have a remarkable effect on the chemical and physical properties of the structured lipids while blend ratio and oil type were the significant factors. Among the interesterified fats, samples produced with corn oil were discriminated from the others due to their more desirable physical and chemical properties.