Ice cream, one of the world’s favourite desserts, technically is a sweetened milk-based emulsion with variable milk fat content. A pasteurized, homogenized, and aged base mix is agitated, aerated and frozen in a scraped surface heat exchanger. Freezing of ice cream occurs progressively, so the structure of the product can be analysed at different stages. The structure can be affected by the processing conditions and the types of ingredients used.
Over the last 50 yr, research has been devoted to altering fatty acid (FA) composition of milk and dairy products by manipulating the diet of lactating dairy cows. Both rumen protected and unprotected sources of lipids have proven to be useful in manipulating FA composition in milk fat (Palmquist et al., Reference Palmquist, Beaulieu and Barbano1993). In early experiments, efforts in changing milk fat composition were driven by the benefits of increasing fat content in milk and producing butters with better spreadability that could compete with margarine (Grummer, Reference Grummer1991). Later, the focus was concentrated on health concerns because of the reported effect of saturated fat on coronary heart disease, although this concern has now been discredited (Palmquist and Jenkins, Reference Palmquist and Jenkins2017). Because the composition of a typical milk fat from dairy cows consists of approximately 5% polyunsaturated fatty acids (PUFA), 70% saturated fatty acids (SFA) and 25% monounsaturated fatty acids (MFA), milk fat became a natural target of those trying to decrease consumption of saturated fats. The functionality and stability of milk fat for some dairy products might be adversely affected by compositional changes arising from the promotion of reduced SFA (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007, Reference Ortiz Gonzalez, Perkins, Schmidt and Drackley2024).
A decrease in the consumption of lauric, myristic and palmitic acids was considered to be the most effective way to reduce the serum low density lipoprotein (LDL) cholesterol concentration in humans (Hegsted, Reference Hegsted and Rios1994; Hillbrick and Augustin, Reference Hillbrick and Augustin2002). In moving toward a more nutritionally desirable milk fat composition, the MUFA content might require the largest change. Oleic acid is the most abundant MUFA in milk and in the human diet. Although olive oil probably is the most well-known source of oleic acid, the contribution of animal fats to total C18:1 intake is of much more importance. At ambient temperature olive oil has a fluid consistency, while under refrigeration it turns into a greasy spread, very similar to the butter oil obtained in our previous experiment in which we infused soy oil FA to cows (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007). In our case, we concluded that a minimum amount of palmitic acid is required in milk fat to maintain its functionality for butter oil, and that this percentage is higher than that proposed to be ideal from a human nutritional standpoint (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007). A decrease in the high melting fraction of milk fat was observed with the reduction in palmitic acid and the increase in PUFA. In some products this might be desirable, such as in whole milk powders, where low- and mid-melting fractions improved its dispersability, or in ice cream, where the low-melting fraction produced more fat destabilization than the hard fractions of milk fat (Goff et al., Reference Goff, Sherbon and Jordan1988).
An increase in the MUFA content of milk fat might lead to an improved consumer perception of milk fat-containing products. Oleic acid content of milk fat increases linearly in response to increased intestinal supply of oleic acid to cows (LaCount et al., Reference LaCount, Drackley, Laesch and Clark1994; Jenkins, Reference Jenkins1999; Drackley et al., Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007). Because of ruminal biohydrogenation of unsaturated FA, oleic acid would need to be ruminally protected to increase its intestinal supply. Feeding calcium salts of high-oleic acid sunflower oil modestly increased oleic acid in milk fat (Lin et al., Reference Lin, Staples, Sims and O’Keefe1996). The amide of oleic acid (oleamide) has been used to provide intestinally available oleic acid to cows, which resulted in increased oleic acid in milk fat (Jenkins, Reference Jenkins1999, Reference Jenkins2000). Thus, the potential exists on a practical level to produce milk and ice cream with substantially increased contents of oleic acid. It is important that the responses of functional characteristics of such milks in various dairy products be adequately described if they are to be useful in production of human food products. Ice cream made from milk from cows fed a source of oleic acid had lower viscosity but similar firmness as control ice cream (Gonzalez et al., Reference Gonzalez, Duncan, O'Keefe, Sumner and Herbein2003).
Research in the area of milk fat modification by dietary manipulation has largely focused on FA composition of milk fat (Palmquist and Jenkins, Reference Palmquist and Jenkins2017). Several studies in our laboratory (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007, Reference Ortiz Gonzalez, Perkins and Drackley2023) and others (Bornaz et al., Reference Bornaz, Novak and Parmentier1992; DePeters et al., Reference DePeters, German, Taylor, Essex and Perez-Monti2001; Gresti et al., Reference Gresti, Bugaut, Maniongui and Bezard1993; Robinson and MacGibbon, Reference Robinson and MacGibbon1998; Pacheco-Pappenheim et al., Reference Pacheco-Pappenheim, Yener, Nichols, Dijkstra, Hettinga and van Valenberg2022) have investigated the effects of milk fat composition on triglyceride (TG) composition and structure. The correlations of TG composition with functional properties of dairy products might be key for determining the type and amount of FA to be included in an altered milk fat to improve the nutritional quality of milk without risking the processing characteristics of a specific dairy product.
In light of this background, the purpose of our study was to evaluate the effect of increasing amounts of oleic acid in milk fat on functional characteristics of ice cream. Milk fat enriched with oleic acid was obtained by abomasally infusing lactating dairy cows with high oleic sunflower FA (HOSFA) to avoid ruminal biohydrogenation of unsaturated FA. The experiment was conducted as a dose-response trial, so that an optimum content of oleic acid might be identified at which functional properties are maintained for processing purposes and consumer acceptance of ice cream.
Materials and methods
In this experiment, we evaluated ice cream prepared from milk enriched with oleic acid via abomasal infusion of HOSFA. Cannulation, housing and management of cows, and preparation of treatments were as described by Drackley et al. (Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007), who reported responses of feed intake, milk yield and composition, FA composition of milk fat, nutrient digestibilities, and blood lipids. All procedures were carried out under a protocol approved by the University of Illinois Institutional Animal Care and Use Committee. Milk TG composition was reported by Ortiz Gonzalez et al. (Reference Ortiz Gonzalez, Perkins and Drackley2023). Briefly, four multiparous Holstein cows (mean number of days in milk = 116) that had been fitted previously with ruminal cannulas were housed in tie stalls fitted with rubber mats, were bedded with straw, and had continuous access to water from individual drinking cups. Cows were milked at 06:30 and 17:00 and were allowed to exercise in an outside lot from 07:00 to 09:00 daily. A typical low-fat lactation diet was mixed as a total mixed ration (TMR) and fed twice daily at the time of milking in amounts to ensure ad libitum intake.
Experimental infusions were prepared weekly at the pilot plant of the University of Illinois Food Science and Human Nutrition Department. The ingredients (described below) were mixed and heated to 72ºC in steam-jacketed stainless-steel vats and then homogenized in a 2-stage homogenizer. Homogenized mixes were cooled and stored at 4ºC until use. The treatments consisted of continuous abomasal infusions of (a) control; 240 g/d of meat solubles (Milk Specialties Co., Dundee, IL) plus 11.24 g/d of Tween 80 (Sigma Chemical Co., St. Louis, MO) in 10 L of tap water; or (b) HOSFA (91.4% C18:1cis, 2.4% C18:2cis-cis, 2.4% C16:0 and 1.8% C18:0 by weight; Henkel Corporation, Emery Division, Cincinnati, OH) at increasing amounts of 0, 250, 500, 750 and 1000 g/d, each homogenized with 240 g/d of meat solubles and 11.24 g/d of Tween 80. Each amount of HOSFA was infused for 7 d. Four cows received treatments according to a changeover statistical design; two cows were randomly selected to receive the set of increasing HOSFA infusions, while the other two served as controls (0 g/d) and were infused only with carriers. After 2 weeks of washout at the end of the sequential set of treatments, the control cows were changed to HOSFA and HOSFA cows were changed to controls.
Morning and evening milkings from day-5 and day-7 of each infusion period were pooled by day and cow and used to prepare butter oil as described in Ortiz Gonzalez et al. (Reference Ortiz Gonzalez, Perkins, Schmidt and Drackley2024). Milk fat content increased from 3.29% at the 0 g/d HOSFA infusion to 5.93% at the 1000 g/d level; milk true protein was not affected by HOSFA and averaged 3.03% (Drackley et al., Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007). Ice cream mixtures from each day and cow were prepared using butter oil as a source of fat in the mix. Ingredients (butter oil 12%, non-fat milk solids 9.75%, sugar 12%, corn syrup solids (Star-Dri 20, Staley Mfg. Co., Decatur, IL) 6%, stabilizer (150 bloom gelatin, Continental Colloids Co., Chicago, IL) 0.30%, emulsifier (CC-280 mono- di-glycerides polysorbate 80; Continental Colloids Co., Chicago, IL) 0.20%, French vanilla flavour (pure vanilla extract 2x Bourbon; Edgar Weber and Co., Wheeling, IL) 0.18%, and ice cream shade 705 (colouring agent; Universal Foods Co., Milwaukee, WI) 0.07% were mixed in stainless steel buckets and pasteurized at 72C for 15 s in a steam water bath. After pasteurization, mixes were homogenized in a 2-stage homogenizer, cooled in ice cold water, and aged inside a 4C refrigerator for 24 h. The aged ice cream mix was frozen in an ice cream machine (Taylor FreezMaster; Tekni-Craft, Rockton, IL). Three gallons of aged ice cream mix were poured into the ice cream feeding tank and the machine and a chronometer were started. Samples for overrun and fat destabilization were taken every 2.5 min for 15 min.
The overrun (O) is defined as
\begin{equation*}O = (\frac{V_2 - V_1}{V_1}) \times 100 \end{equation*}where V1 is the initial volume of the ice cream mix and V2 is the maximum volume obtained upon ice cream formation. Samples were taken from the freezing machine every 2.5 min up to 15 min. Experimentally, this variable was measured using a domestic Kitchen-Aid mixer (model K5-A; Hobart MFG Co., Troy, OH). The bowl was cooled to 4 ºC, a pre-weighed volume of ice cream (ca. 350 mL) at the same temperature was added, and the beaters (speed 8) and a stop-clock were set in motion. Once the whipping point was reached, stop-clock and mixer were stopped. The weight for the original volume was measured, and a sample (ca. 50 mL) of ice cream was transferred to a fast filtration funnel. The funnels were left for 1 h at room temperature (ca. 20 ºC), and the seepage was collected. The volume of this serum was measured with a 10 mL graduated cylinder.
The fat destabilization procedure was followed as described by Goff and Jordan (Reference Goff and Jordan1989). Three grams of melted ice cream from the samples taken from the freezing machine every 2.5 min up to 15 min were poured into a 50-mL Erlenmeyer flask. Then, 27 mL of distilled water were added and mixed. One millilitre of this 1:10 dilution was transferred to a 50-mL volumetric flask, diluted and mixed with distilled water. A sample was transferred to a glass spectrophotometer tube and centrifuged at 1000 rpm for 5 min. Measurements were made at 540 nm, 10 min after centrifugation.
Sensory evaluation was performed on ice creams prepared with milk fat obtained from two cows infused with control (0 g/d) and two cows infused with 500 g/d HOSFA. A triangle test (Roessler et al., Reference Roessler, Pangborn, Siedel and Stone1978) was conducted at South Dakota State University by Robert J. Baer and his Dairy Product Judging Team of 18 students. This group was trained and experienced in evaluating ice cream. The group also evaluated randomly coded ice cream samples for flavour and texture (Larmond, Reference Larmond1977). Panellists were asked to evaluate six attributes (vanilla flavour intensity, overall flavour acceptability, icy, coldness intensity, creaminess, and overall texture acceptability) on a scale from 1 to 9 (1 = extreme defect or poor quality; 9 = no defect or excellent quality). Panellists were encouraged to write additional comments on the evaluation worksheet.
Data were analysed using the SAS general linear models (GLM) procedure (SAS Institute, Inc., Cary, NC). Analysis of variance for a crossover design was performed, with a model where: Y = sequence + cow in sequence + period + treatment + treatment by period by cow in sequence + week + treatment by amount. We tested the interaction of infusate treatment with linear, quadratic and quartic effects of amount. Week within periods was confounded with infusate amount. Data from repeated samples were analysed as independent measures. Least squares means were calculated and are reported. Significance was declared at P < 0.05. Pearson correlation analysis was conducted to find relationships among functional properties and milk fat composition. Triangle test sensory data were analysed as described by Roessler et al. (Reference Roessler, Pangborn, Siedel and Stone1978).
During week 5 of the 12-week experiment and the 1000 g/d HOSFA infusion, the two cows receiving this treatment went off feed and developed diarrhoea, and the infusion was suspended. Therefore, results for the 1000 g/d infusion represent only two cows.
Results and discussion
Results for milk production and composition, feed intake and digestibilities, and milk FA composition were reported in Drackley et al. (Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007). Results for the functional properties of ice cream are shown in Table 1. Overrun and destabilized fat were analysed because they are the most relevant processing properties that define the characteristics of the finished ice cream, such as the dry appearance, resistance to meltdown, whipping properties, body and texture, mouth feel sensation, and size of air cells (Goff and Jordan, Reference Goff and Jordan1989; Goff et al., Reference Goff, Sherbon and Jordan1988).
Least squares means for functional properties of ice cream made from milk of cows infused abomasally with increasing amounts of high oleic sunflower fatty acids (HOSFA)a
a Values in parentheses are least squares means for control cows at each HOSFA infusion level.
The overrun results indicate that the decrease in overrun as HOSFA infusion increased was essentially linear at and after the 10 min readings. The effect on overrun for ice creams was smaller than the effect on overrun measured in whipping creams from this experiment (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Perkins, Schmidt and Drackley2024). Ingredients added to the ice cream base-mix increased the viscosity, the air retention capacity and the stability in this emulsion, but the lower melting point of milk fat as a result of the increasing HOSFA infusions increased the amount of liquid fat, thereby decreasing the ability of foam to incorporate air. Overrun results for ice creams made from 0 g/d to 1000 g/d were in the normal range for a soft serve, considering that a soft serve usually has an overrun between 30% and 60% (Marshall and Arbuckle, Reference Marshall and Arbuckle1996). However, we observed a very glossy ice cream that melted quickly for the ice creams made from milk of cows infused with HOSFA at rates higher than 500 g/d. Overrun at 10 min and after was correlated with viscosity of cream and correlated with fat melting point and firmness peak at all times (Table 2). Overrun was directly correlated with the solid fat content, indicating that as the melting point and solid fat content decreases, the overrun decreases (Table 2).
a Number at the top is Pearson correlation coefficient and at the bottom is the probability value.
b Cream and butter oil variables were reported in Ortiz Gonzalez et al. (Reference Ortiz Gonzalez, Perkins, Schmidt and Drackley2024).
Milk fat also plays an important role considering that during the freezing process it becomes gradually destabilized from its original homogenous state. This destabilization occurs from the physical beating by the dasher in the ice cream freezer. As the milk fat globules clump and partially churn, fat is removed from its finely emulsified state. The extent of this change can be measured by the decrease in turbidity in a diluted and centrifuged sample (Keeney and Josephson, Reference Keeney and Josephson1958). According to Berger et al. (Reference Berger, Bullimore, White and Wright1972) and Berger and White (Reference Berger and White1971), high melting point TG form concentric crystalline layers in the fat globules, and a highly concentrated solution of fat crystals in liquid fat exists within these crystalline layers. As a result of the freezing and beating, the high melting point TG shell is broken and liquid fat released, providing places for clusters of fat that hold the ice cream serum and give dryness to the extruded ice cream. The ideal fat destabilization should form networks of fat globules, which can strengthen the lamellae of air cells and prevent serum drainage. If fat destabilization occurs too early in the freezing process, air cells could be ruptured, thereby making it difficult to develop a good overrun (Thomas, Reference Thomas1981).
An average ice cream mix can be ready after 8 to 12 min of freezing process. Our results for fat destabilization indicate that as the amount of HOSFA infusion increased, the amount of destabilized fat increased rapidly after 5 min of freezing the ice cream mix (Table 1). Significant negative correlations were observed (Table 2) among fat destabilized at different times and the cream overrun, fat melting point, firmness peak and solid fat content at low temperatures. The decrease in high melting point TG that resulted from the increase in HOSFA infusions increased the fragility of milk fat globules, thereby facilitating the release of liquid fat and gradually reducing the overrun of the ice creams prepared from the different treatments.
As infusion of HOSFA increased, short- and medium-chain FA decreased (Drackley et al., Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007). The decreases in medium-chain FA (C12 to C16) were significantly correlated with a decrease in the overrun and an increase in the destabilized fat (Table 3). This was a result of the increasing amounts of HOSFA infused. Oleic acid had the opposite effect on overrun and fat destabilization, since its increase is associated with a decrease in short- and medium-chain saturated FA.
Correlation analysis among ice cream functional parameters and fatty acid compositiona
a Number at the top is Pearson correlation coefficient and at the bottom is the probability.
As HOSFA infusion increased, milk TG with saturated FA decreased and TG with two oleic acid moieties increased (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Perkins and Drackley2023). Considering the overrun and fat destabilized at 10 min we found that the TG fractions from Ortiz Gonzalez et al. (Reference Ortiz Gonzalez, Perkins and Drackley2023) that significantly influenced (Table 4) the overrun in a direct way were F1 to F4, F6 to F7, F9 to F11, F14, F16 to F18, F21, F24 to F27 and F29 (fractions defined in Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Perkins and Drackley2023). Negative effects were found from fractions F12, F26, F35, F36 and F39. The fraction F36 (P < 0.001) exhibited the highest correlation, probably because it was the fraction that increased the most with the increase in the amount of HOSFA infusion (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Perkins and Drackley2023). Several correlations at other time points were significant for overrun (Table 4).
Correlation analysis among triglyceride fractions and ice cream fat destabilization and overrun at different timesa
a Number at the top is Pearson correlation coefficient and at the bottom is the probability value.
b Destabilized fat.
These TG fractions represented decreases in SFA (butyrin-caprylin-palmitin [F2], butyrin-laurin-olein [F6], caproin-caproin-palmitin [F7], butyrin-laurin-olein [F8], butyrin-myristin-palmitin [F10], caproin-palmitin-linolein or caprylin-myristin-linolein [F11], butyrin-palmitin-palmitin or caproin-myristin-palmitin [F14], caproin-palmitin-palmitin [F17], butyrin-palmitin-stearin [F18], caprin-myristin-palmitin [F21], caprin-palmitin-palmitin or caprin-myristin-stearin [F25], laurin-palmitin-olein [F27], olein-myristin-myristin [F28], and myristin-myristin-palmitin [F29]). Also, the increases in dioleyl TAG (with butyric [F12], lauric [F26], palmitic [F36], or stearic [F39] acids) and triolein (F35) were some of the most highly correlated TG with ice cream functional properties as the result of increasing HOSFA infusion.
Destabilized fat at 10 min was directly correlated to fractions F12, F15, F26, and F36, all of them being dioleyl TG, but the strongest effect was from F12 (P = 0.001; Table 4). Negative correlations were obtained with fractions F7, F10 and F25. The first two fractions are associated with short- and medium-chain SFA that sensibly decreased as the HOSFA infusions increased. Fraction F38 is a distearin that was regularly found under control treatments; this TG fraction contributes to the high melting point TG that gives more resistance to fat globules under destabilization. Other TG fractions affected destabilized fat at other time points (Table 4).
Based on general observations during the conduct of this experiment, we concluded that the upper limit for the HOSFA infusion while maintaining both cow productivity and milk processing qualities for ice cream was around 500 g/d. As previously discussed, this dose is not good for a whipping cream (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2022), but since ice cream is a frozen dessert, the risks in lowering the melting point of milk fat are avoided, in agreement with Gonzalez et al. (Reference Gonzalez, Duncan, O'Keefe, Sumner and Herbein2003), and could even result in a better ice cream because of the low melting fractions from greater fat destabilization.
Results from the limited vanilla ice cream sensory evaluation showed that 8 out of 18 judges selected the odd sample out of the 3 cups of ice cream presented in each triangle test. This means that there was no significant difference (P > 0.15) between the control ice creams and the ice creams prepared from the 500 g/d HOSFA infusion. Additional comments from Dr. Robert J. Baer indicated that no judges detected any oxidized off-flavours in ice creams from either milk. Furthermore, no differences in flavour, body or texture were detected between control and HOSFA ice creams (data not shown).
Ice cream is less affected than whipping cream or butter oil by changes in the FA profile of the milk from which it is made (Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007, Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2022, Reference Ortiz Gonzalez, Perkins, Schmidt and Drackley2024). Others have reported similar results when the composition of milk fat was altered by changing the diet of cows (Gonzalez et al., Reference Gonzalez, Duncan, O'Keefe, Sumner and Herbein2003; Corradini et al., Reference Corradini, Madrona, Visentainer, Bonafe, Carvalho, Roche and Prado2014; Nadeem et al., Reference Nadeem, Abdullah, Hussain and Inayat2015; Vargas-Bello-Pérez et al., Reference Vargas-Bello-Pérez, Cancino-Padilla, Geldsetzer-Mendoza, Vyhmeister, Sol Morales, Leskinen, Romero, Garnsworthy and Ibáñez2019) or by using other fat sources (Bazmi and Relkin, Reference Bazmi and Relkin2009; Zhao et al., Reference Zhao, Chen, Liu, Zhong, Chen, Jin, Kang, Wue and Xu2023). Thus, the potential exists to markedly alter the FA composition of milk fat without compromising the functional qualities of ice cream.
We were able to experimentally produce ice creams with milk from cows where oleic acid had been increased to 58.6% of the total FA when HOSFA was infused at 1000 g/d. A more practical limit might be 500 g/d infusion; in this situation oleic acid content of milk fat was doubled and palmitic acid decreased by one-third (Drackley et al., Reference Drackley, Overton, Ortiz Gonzalez, Beaulieu, Barbano, Lynch and Perkins2007; Ortiz Gonzalez et al., Reference Ortiz Gonzalez, Jimenez Flores, Bremmer, Clark, DePeters, Schmidt and Drackley2007). Ice cream from that milk had similar or better functional and flavour characteristics as milk from control cows. Our results indicate the feasibility to produce ice cream with a more desirable FA composition from a human perception standpoint, with no compromise of eating quality.
Conclusions
Ice cream was less affected by the increase in HOSFA in the diet than other products such as cream or butter oil. Ice cream is developed and maintained under freezing conditions, which attenuates many of the effects of increased amount of lower melting point FA and TG. The poorer animal feeding behaviour and the quicker melting observed on ice creams made from milk fat obtained from the 750 and 1000 g/d infusions as compared to control indicated that the 500 g/d infusion was the highest acceptable level. Ice cream produced with milk fat from this level of infusion was not statistically different from control, according to a sensory evaluation panel. We found 23 TG that were correlated with ice cream overrun at 10 min and seven fractions correlated with destabilized fat at 10 min; many were also correlated with other time points. We were able to produce an ice cream with desirable physical and organoleptic qualities yet containing twice as much oleic acid and two-thirds as much palmitate as a control ice cream.
Funding Statement
This research was supported by funds appropriated to the Illinois Agricultural Experiment Station.
Competing interest
The authors declare that they have no competing interests.
Open access
