Our team’s solution to using a hydrophilic water-soluble chlorophyll binding protein (WSCP) to bind chlorophyll contained in a hydrophobic environment was to use emulsions. Emulsions can be designed to maximize the surface area of the interface between the oil and aqueous phases, which serves to increase interaction opportunities between chlorophyll and WSCPs at the interface. They can also be designed to maximize the extent of bulk-phase separation, which would minimize the amount of oil lost.
However, emulsions can be difficult to design because their construction relies heavily on the structure and concentration of surfactant used. Our team first turned to molecular dynamics modelling to determine the stability of our protein in different phases, then we experimentally developed and used predictive phase diagram models to determine the optimal proportions of water, oil, and surfactant to use to maximize the efficiency of our emulsified binding protein (EBP) process.
6GIX in emulsion
One of the biggest issues we faced in project design was the incompatibility between the hydrophilic WSCP and the hydrophobic oil phase. Our solution to this problem was to emulsify our WSCP. This ensures that the WSCP stays within an aqueous environment, whilst simultaneously increasing the surface area of the oil-water interface where chlorophyll binding can occur. Our team turned to protein dynamics modelling to ensure that our L. virginicum WSCP, called 6GIX, would function properly in an emulsion.
Our models confirmed our expectations that 6GIX would be stable in an aqueous solution, but would denature immediately when modeled in oleic acid, a major component of canola oil. In a biphasic environment containing both water and oleic acid, 6GIX remained within the aqueous phase as a stable structure.
Figure 1. 6GIX (purple) in water (left), in biphasic emulsion (center), and in oleic acid (right).
In addition, we ran aggregation models to see if 6GIX would aggregate after binding to chlorophyll to form a tetramer as literature suggests, and whether or not this tetramer would be more or less stable than individual monomers (Palm et al., 2018). These models showed us that 6GIX does aggregate as a tetramer after binding chlorophyll. The tetrameric structure allows 6GIX to have an increased stability in a biphasic environment.
Figure 2. 6GIX as an aggregated tetramer in emulsion.
Meeting with Richardson Oilseed
In order to design a new industrial solution for chlorophyll extraction, it was crucial to learn more about the existing acid-activated clay (AAC) method for chlorophyll removal. We set up a meeting with a prominent canola oil extraction and production company, Richardson Oilseed International. They invited us to their processing plant in Lethbridge, AB to meet with their Facility Engineering Manager Dallas Gade and Operations Supervisor Laurence Parslow. During our meetings, we learned all about canola oil processing and production, as well as some of their thoughts on how our emulsion system would integrate into their facility.
Our meeting with Richardson Oilseed yielded three main takeaways to consider in the design of our emulsion system. The first takeaway is that the AAC bleaching process is optimized to maximize chlorophyll removal and minimize oil product loss, and thus for our system to be viable, we must have superior performance in chlorophyll removal and oil retention in comparison to the AAC process. Second, we learned that our system would be more attractive if its effects on existing processing structures are minimized. Last but not least, we learned that in order for our emulsified binding process (EBP) system to be industrially considered, the economic benefits of this new system, taking into account new production costs, must exceed existing AAC system.
Taking these points into consideration, our team identified three areas of study that we must explore in order to better our design of the EBP process. First, we studied AACs and their ability to remove chlorophyll to get a better grasp on the efficiency of this system. Second, we explored the optimization of the EBP process by using phase diagrams to determine optimal concentrations of aqueous phase, oil, and surfactant in the emulsion. Optimization of this process to decrease undesirable microemulsions can decrease processing costs and allow for smoother integration into existing infrastructure. Lastly, we compared the efficiency of our EBP process against AACs to determine the viability of our system.
Quantifying chlorophyll in canola oil
Our team began our chlorophyll experiments by extracting chlorophyll pigments from spinach using 95% ethanol. Because we were extracting chlorophyll ourselves, we searched the literature for methods to quantify the chlorophyll concentration. We found papers outlining the spectrophotometric quantification of chlorophyll in 95% ethanol (Sumanta, Imranul Haque, Nishika, & Suprakash, 2014) and in mineral oil (Jaber et al., 2012).
Figure 3. Quantification of chlorophyll in ethanol.
Figure 4. Quantification of chlorophyll in mineral oil.
However, we were unable to find any literature that specifically dealt with the quantification of chlorophyll in canola oil. This was a point of concern because mineral oil is mostly composed of alkanes and cycloalkanes, while canola oil is mostly composed of fatty acids.
Our team decided to test whether the chlorophyll quantification equation for mineral oil would also apply to canola oil. In order to do this, we decided to perform a serial dilution and measure the use spectroscopy to quantify the concentration of spinach-extracted chlorophyll in 95% ethanol. Afterwards, the ethanol was mixed with commercial canola oil, then centrifuged to separate the two phases. After separation, the concentration of chlorophyll remaining in the 95% ethanol was measured, and the concentration of chlorophyll transferred to the oil was determined using the mineral oil equation. The quantity of chlorophyll lost from 95% ethanol was compared to the quantity of chlorophyll transferred to the oil as measured by the mineral oil equation to assess the accuracy of the mineral oil model for canola oil.
Figure 5. Chlorophyll lost from 95% ethanol and chlorophyll transferred to oil upon mixing and separating two equal volumes of the two phases. Values represent the mean between five replicates. Error bars represent standard deviation.
A Welch-Satterthwaite test was run on the above data, which showed that the difference between the two datasets was not statistically significant. This implies to us that the chlorophyll absorbance equation can be used for canola oil.
Through our meeting with Richardson Oilseed, we also learned about how much the chemical composition of canola oil changes throughout the refining process. In order for us to test how our EBP process works under real-life conditions, it was necessary for us to get our hands on real green seed oil pressed from green seed canola. Early in the summer, we were graciously donated a sample of green-seed oil from Milligan Biofuels for our experiments. Before we began any chlorophyll assays, we needed to quantitatively estimate the amount of chlorophyll in our oil sample. Knowing that the mineral-oil chlorophyll equation could be used for canola oil solutions, we performed a serial dilution in order to determine the linear region of the absorbance curve.
The purpose of identifying the linear region of the curve comes from the formulation of the mineral-oil absorbance equation. It is derived using the extinction coefficient of chlorophyll in a mineral oil solution, which means that it is only valid while the concentration of chlorophyll in solution is linearly correlated to its absorbance. Without knowing the original concentration of chlorophyll in the green oil, we obtain the correlation between absorbance and concentration by increasing the dilution factor.
Figure 6. Absorbance curve of green seed canola oil obtained from Milligan Biofuels. Absorbance measured at 670nm. Values represent the mean of three replicates.
The concentration range which is represented in the most linear region can be found between dilution factors 2 and 8.
Acid-Activated Clays (AACs)
Chemical separation process design is an extremely active area of research within the chemical engineering profession. In order to learn more about the design of our chlorophyll-removal assays, we met with Dr. Kazi Sumon, a separation processes professor in the chemical engineering faculty at the University of Calgary. Dr. Sumon cautioned us that the challenge with developing assays for adsorption processes is not in the experimental approach itself, but rather in identifying physically representative models for these processes.
This is because substrates bind onto adsorptive materials by both physisorption and chemisorption. Physisorption refers to the steric affinity between the substrate and active site caused by the structure of the active sites on the clay, whereas chemisorption refers to the chemical reaction that occurs to covalently bind the substrate and the active site. In order to simplify the process of identifying a physically representative model for our AAC adsorption assays, Dr. Sumon recommended that we study a clay type that is known to bind chlorophyll without causing any side-reactions that affect the properties of the oil.
Taking this advice into account, we identified an AAC called Montmorillonite K10, that had been previously characterized to non-reactively remove chlorophyll from canola oil. Montmorillonite K10 is commonly used to remove chlorophyll from vegetable and seed oils for biofuel processing (Issariyakul & Dalai, 2010).
From preliminary experiments involving mixing Montmorillonite K10 with green seed canola oil from Milligan Biofuels, we saw a clear difference in the color of the oil after separation of the oil and clay. This shows us that Montmorillonite K10 does act to remove chlorophyll from the oil.
Figure 7. From left to right: green seed canola oil, green oil-clay slurry, oil after separation of clay.
We further studied the quantitative performance of Montmorillonite K10 in terms of chlorophyll removal and oil loss by varying the clay loading, which is the percent mass of chlorophyll relative to the oil, as well as the temperature of mixing. We examined its efficiency at three different temperatures (300.15K, 365.15K, and 393.15K) in order to examine how temperature affects the adsorption curve. Different clay loadings were tested at 1%, 2.5%, 3%, 5%, and 10%. After separation of the clay and oil, the oil was diluted by a factor of four to fall within the linear region of the oil absorbance curve measured above, and the concentration of chlorophyll in the oil was quantified.
Figure 8. Average clay adsorption data. Values represent the mean of three replicates. Standard deviation error bars too small to be seen.
These graphs showed us that the concentration of chlorophyll in the oil after processing with AACs is notably smaller with increasing clay loadings. In addition, the removal of chlorophyll by the clay appears to be more efficient at higher temperatures.
In addition to chlorophyll concentration data, we measured the percentage of oil lost in each process.
Figure 9. Percent clay loss as a function of clay loading.
This figure shows us that the percent of oil lost increases with a higher clay loading.
With this data in hand, we sought out to find a physically representative model that fits this data. Chlorophyll adsorption onto clay can be modeled using adsorption isotherms (Limousin et al., 2006). Adsorption isotherms can predict the amount of chlorophyll in the clay as a function of the chlorophyll adsorbed onto the clay during processing. Since the AAC experiments were performed in a beaker, they could be considered a closed system. Applying a conservation mass balance, the amount of chlorophyll adsorbed into the clay can be calculated using the following equation:We then fit this data to various adsorption models, and found the Langmuir isotherm to be the best fitting.
Figure 10. Langmuir Isotherm.
Figure 11. Langmuir isoform fit onto experimental data points.
Once we found that our data fits the Langmuir isotherm, we sought to create a temperature-dependent performance model that would allow us to determine the chlorophyll-removal efficiency of Montmorillonite K10 from the clay loading at any temperature.
We began this process by fitting an equation to describe KL and qads as a function of temperature (Pohndorf, Cadaval, & Pinto, 2016).
Figure 12. KL and qads as a function of temperature.
Using these graphs, we were able to derive equations for KL and qads as a function of temperature. The following two equations were obtained:
Using the above equations, we derived an equation that represents the chlorophyll fraction in the oil after processing with clay.
Using this equation, we were able to create a graph outlining the relationship between the mass ratio between clay and oil and the chlorophyll fraction remaining in the oil after processing.
Figure 13. Chlorophyll fraction in oil as a function of adsorbent:oil mass ratio.
Although this graph only depicts three different temperatures, the same equation can be used to model the relationship between the adsorbent:oil mass ratio and the chlorophyll fraction in the oil at any temperature. This proved to be valuable in determining how to compare the performance between the AAC and the EBP process.
Phase Diagram Model Design
Informing Emulsion Experiments
With our AAC model in place, our team began designing our emulsion system. In order to create the most efficient emulsion system possible, our team looked into the literature to find ways in which emulsions can be manipulated. We found that varying surfactant, composition, processing temperature, method of mixing, emulsion structure can all have large impacts onto the type of emulsion being formed (Kumar Paul, Priya Moulik, & Priya Moulik, 1997).
Seeking out more expertise on the subject, we met with Dr. Nashat Nassar, a professor in the faculty of chemical engineering at the University of Calgary whose research focuses on the application of emulsions as microreactors for nanoparticles and enhanced-oil-recovery techniques.
After listening to our EBP idea, Dr. Nassar's first question was: "What type of emulsion are you trying to make?". Dr. Nassar went on to explain to us how different emulsion types can be classified in accordance to the Winsor classification system, and how emulsions can be characterized based on ternary phase diagrams. Not realizing that the emulsion-type was something that was supposed to be intentionally controlled, our team began extensive literature review on Winsor classification types (A Winsor & Hahn, 1932).
We found that the Winsor classification system classifies emulsion-solutions into four different types.
Figure 14. Various Winsor emulsion types.
Because our emulsion will be used as a step in oil processing, it was important to us that the emulsion-solution settles out to create a pure oil phase, which would ease the oil recovery process and minimize the need for extra downstream process of the oil in an industrial application. As a result, our team honed in our focus on Winsor 1 emulsions, as it settles out to produce a pure oil phase.
We sought an alternative to classical thermodynamic models through machine learning classification algorithms (Djekic, Ibric, & Primorac, 2008). The goal was to apply them to produce ternary phase diagram models which can clearly predict the Winsor type as a function of its formulation variables.
We performed an initial study to determine how well our approach of using machine-learning based models would work. We performed a series of dilution line experiments using synthetically produced green oil, polysorbate 80, and distilled water (Kahlweit et al., 1987). It is important to note that polysorbate 80 is a food-grade surfactant suitable for use in canola oil production. For each of the 80 emulsion compositions created in this experiment, a Winsor class was assigned via observation.
Figure 15. Experimental emulsion compositions determined with the dilution line method.
Using this data, we were able to generate a ternary diagram representing the different Winsor types and their compositions.
Figure 16. Ternary diagram depicting the composition of four different Winsor-type emulsions.
This experiment was repeated at 300.15 K, 310.15 K, 315.5 K, 328.15 K, 343.15 K to test the effect of temperature on emulsion composition. The results for these experiments were fed into the machine learning algorithms developed by our drylab to generate the first round of predictive phase diagram models. These models were shown to be consistent using the support vector classification (SVC) with theoretical emulsion phase diagrams described by Winsor. For more information, see our emulsion construction prediction page.
Having confirmed that the modelling approach our dry lab has taken could be used to produce phase diagrams that were consistent with physical theory, we ran a second set of dilution-line experiments using of the green oil donated by Milligan biofuels, the aqueous buffer used for protein suspension, and the co-surfactant mixture prescribed in a study emulsifying WSCPs (Bednarczyk, Takahashi, Satoh, & Noy, 2015).
The second set of experimental classification results were inputted into the SVC algorithm to produce the following phase diagram model at 300.15K. The model was determined to have a mean classification error of 20.35%.
Figure 17. SVC-generated phase diagram model from experiment 2 at 300.15 K.
Upon developing a predictive phase diagram model for our emulsion system, we used it to identify different combinations of the aqueous, organic, and surfactant phases to produce the desired Winsor 1 structure. Specifically, we sought to identify compositions from the phase diagram model with the following criteria:
- Composition is classified as a Winsor 1 type emulsion
- Model predicts the Winsor label with the highest confidence it is capable of
- Water-to-oil ratio is between 0 and 1 – this is in order to maximize the amount of green oil that can be processed and minimize the amount of aqueous phase needed while maintaining the Winsor 1 structure
- Points are chosen to vary the amount of the surfactant volume fraction for performance testing in section 3
Based off of these criteria, we identified five composition candidates to test the performance of our emulsion system.
Figure 18. C6, C7, C8, C9, and C10 emulsions. All of them show a Winsor 1 type emulsion.
We validated the emulsion structure of these candidates in the lab and found that the phase diagram model had correctly predicted the Winsor type for each.
Emulsified Binding Protein Tests
Efficacy of 6GIX
The final section in the development of our EBP process for chlorophyll removal involves examining the performance of our emulsion system using the different identified compositions from the SVC phase diagram model.
The principle theory behind emulsion equilibrium outlines that the Winsor type is defined by the bulk phase (aqueous or oil) that contains the highest concentration of surfactant. Since our compositions were designed to produce the Winsor 1 structure, the effect of surfactant concentration primarily affects the size of the oil-in-water emulsions.
We quickly realized that this is likely an important design consideration that needs to be investigated. Originally we had thought that if 6GIX needs to bind to chlorophyll on the oil/water interface, increasing the interfacial surface area between the two phases would allow chlorophyll to be removed at a higher rate. However this primarily assumes that the protein stability is unaffected by it's concentration or by the proximity to the oil/water interface. To see whether or not these were factors that would affect the performance of our emulsion system, we tested the effects of different protein and surfactant concentrations.
In order to better characterize our emulsion system and 6GIX, we wanted to identify a control protein to account for the effect of a protein on emulsion structure. We identified Bovine-Serum-Albumin (BSA) as a candidate after realizing that BSA had a well characterized electronegative core, which is attracted to the positive magnesium core of chlorophyll. The binding between BSA and chlorophyll had been previously shown to be weak (Gorza et al., 2014), making it an ideal control protein for use.
Figure 19. Various emulsions tested in lab.
The above figure shows a clear difference in color of the bottom-phase microemulsion between the BSA emulsion, 6GIX emulsion, and buffer emulsion with the 6GIX emulsion showing the strongest green color. This indicates that 6GIX is the most effective at binding and removing chlorophyll from the oil phase.
We further tested the chlorophyll removal of of our system with the five different emulsion compositions suggested by our phase diagram model, with our first results indicating a slight improvement of chlorophyll removal of the oil phase. Increasing the concentration of surfactant in solution can be shown to increase the absorbance value at 670nm in the oil, indicating that less chlorophyll is being removed.
This can be rationalized by considering the overall emulsion structure. As the amount of surfactant increases, the emulsion structure will become increasingly similar to a bicontinuous microemulsion (Kumar Paul, Priya Moulik, & Priya Moulik, 1997). This causes the volume of aqueous solution within the emulsion to be reduced, decreasing the surface area available for contact between 6GIX and chlorophyll (Lindman et al., 1989). Out of the compositions tested, the emulsion using 6GIX and a surfactant volume of 15% (C8 composition) had the highest reduction of chlorophyll content, as presented in the figure below.
Figure 20. Absorbance of oil at increasing volume fractions of surfactant at a protein:chlorophyll molar ratio of 1:10.
From the emulsions tested in Figure 19, it was possible to qualitatively tell that 6GIX was more efficient than BSA at capturing and removing chlorophyll from green oil. We quantified these values to confirm our observations.
Figure 21. Percent change in chlorophyll concentration for BSA and 6GIX emulsions of 5 different types.
The above graph shows that for all emulsion compositions, 6GIX was much more effective than BSA at removing chlorophyll form the oil phase.
Further experiments demonstrated that both BSA and 6GIX show an increase in performance as the concentration of protein in solution increases.
Figure 22. Absorbance of oil at 670nm after emulsification at increasing protein concentrations with C8 composition.
However, the slope of the graph is much steeper at lower protein concentrations. This means that increasing the concentration of protein has diminishing returns. A possible explanation for this is that the amount of protein in solution is directly correlated the amount that is denatured due to contact with the oil phase during the mixing phase.
After we were able to validate the functionality of 6GIX in emulsion, we turned to ways in which we could improve our system. As we saw above, an increase in protein concentration has diminishing returns on the performance of our system, likely due to the denaturation of the protein as it comes in contact with oil. In order to mitigate this issue, our team looked towards ways in which we could increase the stability of 6GIX in proximity to an oil-water interface.
Through modelling, we were able to identify that 6GIX has 12 amino acids that contribute to variance in its crystalline structure when functioning in an emulsion system. Particularly at the water-oil interface, the instability caused by these amino acids causes 6GIX to denature more easily.
To address this issue, we designed and used a genetic algorithm called iGAM (international genetic algorithm machine) that analyzed and modified the sequence of 6GIX to produce ModGIX, a modified 6GIX protein that has demonstrated enhanced chlorophyll binding and stability in our emulsion system in silico.
Figure 23. Molecular model of ModGIX
Currently, our team is in the process of cloning and producing the ModGIX protein. In the future, we hope to characterize the chlorophyll-binding capabilities of this protein in comparison to 6GIX.
A Winsor, B. P., & Hahn, von. (1932). HYDROTROPY, SOLUBILISATION AND RELATED EMULSIFICATION PROCESSES. PART I. Aqueous Solutions of Parajin Chain Salts (Vol. 62). Retrieved from https://pubs-rsc-org.ezproxy.lib.ucalgary.ca/en/content/articlepdf/1948/tf/tf9484400376
Bednarczyk, D., Takahashi, S., Satoh, H., & Noy, D. (2015). Assembly of water-soluble chlorophyll-binding proteins with native hydrophobic chlorophylls in water-in-oil emulsions. Biochimica et Biophysica Acta (BBA) - Bioenergetics, 1847(3), 307–313. https://doi.org/10.1016/J.BBABIO.2014.12.003
Djekic, L., Ibric, S., & Primorac, M. (2008). The application of artificial neural networks in the prediction of microemulsion phase boundaries in PEG-8 caprylic/capric glycerides based systems. International Journal of Pharmaceutics, 361(1–2), 41–46. https://doi.org/10.1016/J.IJPHARM.2008.05.002
Gorza, F. D. S., Pedro, G. C., Trescher, T. F., Da Silva, R. J., Silva, J. R., & De Souza, N. C. (2014). Morphological analysis and interaction of chlorophyll and BSA. BioMed Research International, 2014. https://doi.org/10.1155/2014/872701
Issariyakul, T., & Dalai, A. K. (2010). Biodiesel Production from Greenseed Canola Oil †. https://doi.org/10.1021/ef901202b
Jaber, H., Ayadi, M., Makni, J., Rigane, G., Sayadi, S., & Bouaziz, M. (2012). Stabilization of refined olive oil by enrichment with chlorophyll pigments extracted from Chemlali olive leaves. European Journal of Lipid Science and Technology, 114(11), 1274–1283. https://doi.org/10.1002/ejlt.201100176
Kahlweit, M., Strey, R., Haase, D., Kunieda, H., Schmeling, T., Faulhaber, B., … Jahn, W. (1987). How to study microemulsions. Journal of Colloid and Interface Science, 118(2), 436–453. https://doi.org/10.1016/0021-9797(87)90480-2
Kumar Paul, B., Priya Moulik, S., & Priya Moulik, S. (1997). MICROEMULSIONS: AN OVERVIEW. JOURNAL OF DISPERSION SCIENCE AND TECHNOLOGY, 18(4), 301–367. https://doi.org/10.1080/01932699708943740
Li, T., Xu, J., Wu, H., Wang, G., Dai, S., Fan, J., … Xiang, W. (2016). A saponification method for chlorophyll removal from microalgae biomass as oil feedstock. Marine Drugs, 14(9). https://doi.org/10.3390/md14090162
Limousin, G., Gaudet, J.-P., Charlet, L., Szenknect, S., Barthè, V., & Krimissa, M. (2006). Sorption isotherms: A review on physical bases, modeling and measurement. https://doi.org/10.1016/j.apgeochem.2006.09.010
Lindman, B., Shinoda, K., Olsson, U., Anderson, D., Karlström, G., & Wennerström, H. (1989). On the demonstration of bicontinuous structures in microemulsions. Colloids and Surfaces, 38(1), 205–224. https://doi.org/10.1016/0166-6622(89)80154-4
Pohndorf, R. S., Cadaval, T. R. S., & Pinto, L. A. A. (2016). Kinetics and thermodynamics adsorption of carotenoids and chlorophylls in rice bran oil bleaching. Journal of Food Engineering, 185, 9–16. https://doi.org/10.1016/J.JFOODENG.2016.03.028
Sumanta, N., Imranul Haque, C., Nishika, J., & Suprakash, R. (2014). Spectrophotometric Analysis of Chlorophylls and Carotenoids from Commonly Grown Fern Species by Using Various Extracting Solvents. Research Journal of Chemical Sciences (Vol. 4). Retrieved from www.isca.me