Design and characterization of the separation process
The present work reports a novel approach for the separation of proteins. This is divided into two sequential liquid-liquid extraction steps, a first step based in ATPS and a second step based in AMTPS.
Measurement of the ATPS phase diagrams and tie-lines
The ATPS were characterized through the measurement of the phase diagrams and tie-lines (TLs), aiming at understanding the effect of different inorganic salts and the copolymer nature along with the influence of surfactants used as adjuvants on their formation. The phase diagrams were determined for all the ATPS studied, as depicted in Figs. 1, 2, and 3. All curves were determined using the cloud-point titration method at 25 ± 1 °C and atmospheric pressure. The experimental points were correlated using the Merchuk equation [17]. Its parameters (A, B and C) used on the description of the experimental binodal data as well as the experimental data for the phase diagrams are reported in Additional file 1: Tables S1-S4. The experimental TLs, along with their respective length (tie-line length, TLLs), are reported in Additional file 1: Table S5. The TLL is a numerical indicator of the difference between the compositions of the two phases and it is generally used to correlate trends in the partition of solutes between both phases. The mixtures with total compositions along a specific TL have different mass or volume ratios from those of the two coexisting phases, though the composition of each phase is maintained [18].
Regarding the effect of inorganic salts in the ATPS formation, their aptitude to promote the phase separation was studied for potassium phosphate salts, namely K2HPO4, KH2PO4, K3PO4 and K2HPO4/KH2PO4. The study of inorganic salt nature has been performed on ATPS composed of Pluronic L-35 as the phase former in presence of small amounts of Triton X-100 (circa of 1 wt%) - Fig. 1. Herein, the ability to promote the two-phase formation follows the order: K3PO4 > K2HPO4/KH2PO4 ≈ K2HPO4 > KH2PO4. In general, the potassium phosphate salts with higher salting-out strength exhibit a wider biphasic region. This observation corroborates the qualitative trend on the salt cations ability to induce the salting-out nature of the copolymer, which follows closely the Hofmeister series [19] with KH2PO4 and K3PO4 being the weakest and strongest salting-out agents, respectively. Considering the buffer capacity of the potassium phosphate buffer (K2HPO4/KH2PO4), a very attractive aspect for the proteins separation, along with its larger biphasic region, this system was adopted in the following studies.
The presence of a surfactant as adjuvant was evaluated in terms of its ability to promote the two-phase formation by using small amounts (circa of 1 wt%) of two non-ionic surfactants, namely Triton X-114 and Triton X-100, whose characteristics and chemical structure are present in Additional file 1: Table S6. These surfactants possess a similar chemical structure, varying only in the number of ethoxylate groups forming the surfactant’s crown and thus, its hydrophilicity (cf. the hydrophilic-lipophilic balance (HLB) of the surfactants is presented in Additional file 1: Table S6). The surfactants’ influence was analysed in a Pluronic L-35 + potassium phosphate buffer-based ATPS and compared with the conventional system (without any adjuvant present) - Fig. 2. The results show that the use of these co-surfactants does not significantly affects the binodal curves, and thus the phases separation in this system
The copolymer nature (normal versus reverse) and composition (weight percentage of PEG units, cf. Additional file 1: Table S6) were two other aspects explored on the phase diagrams. Three different copolymers were selected, namely Pluronics 17R4, 10R5 and L-35 and studied using a pseudo-ternary system composed of potassium phosphate buffer (pH = 6.6). The respective phase diagrams are present in Fig. 3, where a tendency can clearly be established, considering their capacity to form two phases, as Pluronic 17R4 > Pluronic 10R5 > Pluronic L-35.
Herein, Pluronic 17R4 holds the wider biphasic region, due to its more hydrophobic nature, considering the 60 wt% of PPG in its composition compared with the 50 wt% in the remaining copolymers. In contrast, Pluronic L-35 displays the narrowest biphasic region, though with only a small difference for Pluronic 10R5. This difference is a result of the copolymer structural rearrangement, i.e. Pluronic L-35 is composed of repetitive units of PEG-PPG-PEG, while Pluronic 10R5 presents sequences of PPG-PEG-PPG. Therefore, the normal copolymer evidences a higher hydrophilicity owing to the two PEG units, resulting in a lower ability to form the two-phases.
Measurement of the AMTPS coexisting curves
As thermo-responsive copolymers, these systems can be induced to form two-phases using temperature as the driving force. The cloud points were determined, and the phase diagrams are presented in Fig. 4. These results show that the copolymer nature and composition display a major effect on the AMTPS formation, namely upon the cloud points. Once again, the ability of the copolymers to form the biphasic region follows the tendency of Pluronic 14R4 > Pluronic 10R5 > Pluronic L-35. The main difference is that now, there is the formation of an AMTPS instead of a simpler ATPS, which means that the phase separation occurs due to the micelles coalescence in one phase and is not a result of the copolymer being salted-out by the salt. Herein, there is a complex balance of distinct interactions (electrostatic interactions, hydrophobic associations, hydrogen bonds and van der Waals forces), which in turn affects both solute-solute and solute-solvent interactions [20]. Moreover, it is well known that the addition of a co-surfactant can, not only reduce the system cloud points, but also improve the system extractive performance [20, 21]. This is the best option in terms of the cloud point extraction of labile proteins, and thus, the quaternary system composed of Pluronic L-35 + potassium phosphate buffer (+ water) + Triton X-114 was also characterized and depicted in Fig. 4 (dashed line and square symbol).
Through these results, it is visible a slight reduction of the cloud point temperatures of this system in comparison with the pseudo-ternary system composed of Pluronic L-35, but in absence of Triton X-114 as co-surfactant. Since both Pluronic L-35 and Triton X-114 are non-ionic surfactants above their CMC, non-ionic mixed micelles are formed. Nevertheless, it seems that there is a dominance of the copolymer in the aggregate’s formation, since it is present in higher concentration.
Optimization of the proteins partition applying ATPS and AMTPS
Once the phase diagrams had been characterized, a mixture point was selected, considering two criteria, the water content, and an appropriate temperature, above the system cloud point, but not too high to maintain the proteins thermal stability. As previously mentioned, cytochrome c, azocasein and ovalbumin were the model proteins selected (cf. properties in Additional file 1: Table S7). The ternary system composed of Pluronic 17R4 was not used due to experimental restrictions imposed by its very low cloud point (25 °C).
Thus, the systems studied in the partition of proteins were the ones constituted by Pluronic L-35 and Pluronic 10R5 and the quaternary system composed of Pluronic L-35 + potassium phosphate buffer + water + Triton X-114. The ATPS and AMTPS prepared to perform the partition tests are exemplified by the case of Pluronic L-35 as presented in Additional file 1: Figure S1.
The recovery and partition coefficient data obtained for each model protein in both (top and bottom) phases of the ATPS and AMTPS were determined, and the results presented in Figs. 5, 6, Additional file 1: Figure S2 and S3. From the Recovery results displayed in Figs. 5 and 6 and corroborated by the partition coefficient data (Additional file 1: Figures S2 and S3), it is clear the cytochrome c (red bars) preferential partition to the bottom/salt-rich phase whereas azocasein (blue bars) was completely recovered in the top/copolymer-rich phase. Contrarily, the ovalbumin (green bars) partition was found to be dependent on the system, since for Pluronic L-35-based AMTPS, ovalbumin is mainly recovered in the top phase, while for Pluronic 10R5, this protein partitions preferably for the bottom-phase of the ATPS.
The cytochrome c preferential partition to the salt-rich phase can be improved by the proper choice of the copolymer, being this partition more pronounced for Pluronic 10R5 (%Rec Bottom = 95 ± 5%). It is also clear that electrostatic interactions between proteins and the buffer are not the only parameter influencing the proteins’ partition behaviour since both cytochrome c and ovalbumin partition varies with the copolymer applied. For instance, when the normal is replaced by the reverse Pluronic, the ovalbumin partition tendency completely changed with around 60% of this protein being concentrated not in the polymeric phase but in the salt-rich phase. This leads to the conclusion that some more specific interactions between the copolymers and ovalbumin should be occurring and dictating its partition. Likewise, cytochrome c recovery is also improved with this copolymer replacement, suggesting that the more hydrophobic character of Pluronic 10R5 might be forcing more cytochrome c to migrate towards the more hydrophilic phase.
Regarding the presence of Triton X-114 as co-surfactant, it was found that the ovalbumin recovery is enhanced by 20% to the copolymer-rich phase. This reinforces the notion that some specific interactions between the system phase formers and the proteins contribute to their partition.
To further elucidate the ability of these systems to separate the proteins, the ATPS selectivity was also determined. As expected, higher selectivity values were obtained for the Pluronic L-35 in the partition of ovalbumin and cytochrome c. Even though the presence of Triton X-114 affects the partition of proteins, a negligible effect is observed when the proteins selectivity (especially SOva/Cyt c) is investigated. Nevertheless, outstanding selectivity values were obtained for the partition of azocasein and cytochrome c in all the studied systems (S > 1250).
Sequentially, the ATPS top phase was submitted to a temperature above the cloud point of each system and allowed it to separate into two macroscopic phases, aiming at separating ovalbumin and azocasein in the end (Fig. 6). Once again, azocasein migrated completely towards the top/surfactant-rich phase while ovalbumin partitioned mostly to the bottom/surfactant-poor phase. The ability to fractionate both model proteins in the AMTPS is described by the trend: Pluronic 10R5 < Pluronic L-35 + 1 wt% Triton X-114 < Pluronic L-35. The differential partition between the two proteins can be explained by their molecular weights and hydrophobic/hydrophilic character [21, 22]. The smallest and more hydrophobic protein, in this case azocasein, is recovered inside the micelles, while ovalbumin, due to its higher molecular weight and more hydrophilic character, is excluded to the most hydrophilic phase, the surfactant-poor phase. As far as the pseudo-ternary and quaternary systems with Pluronic L-35 are concerned, it can be assumed that the micelle complexity of the quaternary AMTPS hinders the partition of ovalbumin towards the surfactant-rich phase. Therefore, the addition of a co-surfactant is not so selective as it was in the first separation step, probably by the nature of the mixed micelles created [21]. Taking these results into account, the system with Pluronic L-35 was identified as the most selective system for the two fractionation steps.
Sequential fractionation of the protein mixture
The separation of the three proteins present in a single mixture was performed for the most selective system composed of Pluronic L-35 + potassium phosphate buffer + water. The isolation of each protein from the phase formers was a step also investigated in this work and corroborated by distinct techniques. Herein, two different parameters were considered to analyse the proteins purification owing to the use of a complex protein mixture, namely the proteins recovery (Rx) in each phase and their purity (Px). This data is presented in Fig. 7 and Additional file 1: Figure S4. As expected, the protein partition of ovalbumin, azocasein and cytochrome c maintained almost the same partition profile as previously observed for each protein when individually tested (section 2.2 of this work). Herein, cytochrome c was completely recovered in the salt-rich phase of the ATPS (RCyt c = 100% and PCyt c = 14%), which was an improvement compared to the results previously obtained for each protein tested individually. However, ovalbumin partitioned almost completely to the salt-rich phase contrarily to the expected (ROva = 96% and POva = 86%), being the remaining concentration separated from azocasein in the second fractionation step, applying the AMTPS. In this case, azocasein was completely recovered in the surfactant-rich phase (RAzo = 100% and PAzo = 100%), whereas the ovalbumin still present in the system was totally concentrated in the surfactant-poor phase (ROva = 4% and POva = 100%). In the end, an ultrafiltration step was applied to isolate cytochrome c and ovalbumin, obtaining a cytochrome c recovery of 89% with a 74% purity and recovering 97% of an almost pure (99%) ovalbumin. Aiming at the industrial potential of the integrated process developed, an isolation step was considered in this integrated process enabling the reuse of copolymer for additional cycles of purification. For this step, an acid precipitation of Azo was carried using TCA (0.1 M at pH 3–4) to promote the azocasein isolation from Pluronic L-35, as represented in Additional file 1: Figure S5. After Azo precipitation, the pellet was microfuged at maximum speed and the supernatant was discharged, being applied a solution of NaOH (0.1 M) to further dissolve the precipitated protein. Both the azocasein isolation and copolymer recovery were confirmed through 1H NMR and FTIR, as shown in Additional file 1: Figures S6 and S7, respectively. The NMR data seems identical before and after the polishing step, Additional file 1: Figure S6, however, the FT-IR spectra were different. In Additional file 1: Figure S7, the main differences between the aqueous copolymer solution (black line) and the supernatant (red line) presenting essentially Pluronic L-35 correspond to the amount of water present in the solutions (cf. 1625 cm− 1: water H-O-H bend and 3400–3200 cm− 1: water O-H stretch). The potassium phosphate buffer was not recovered since it is a common media used to stabilize proteins.
Overall, high purities (> 74%) were obtained for the four distinct polished streams: iv), v), vii) and ix), as presented in Fig. 7. It should be stressed that circa of 5 wt% of Pluronic L-35 is still present in stream ix); yet, this copolymer concentration is at an acceptable concentration approved by FDA [23].
Summing up, a high-performance separation process was here developed by the sequential application of ATPS and AMTPS to separate ovalbumin (maximum yield and purity of 97 and 99%, respectively), azocasein (maximum yield and purity of 100%) and cytochrome c (maximum yield and purity of 89 and 74%, respectively).
Environmental assessment
Figure 8 shows the results of the carbon footprint of the novel purification platform of proteins complex matrices per 1 kg of aqueous system. The total carbon footprint is equal to 117 kg CO2 eq. The contribution of the fractionation process (79 kg CO2 eq.), which includes the ATPS, AMTPS and ultrafiltration steps, represents ~ 67% of the total carbon footprint, is dominated by the ultrafiltration step (~ 49%). The proteins isolation process contributes with 39 kg CO2 eq. This process encompasses the acid precipitation step, representing ~ 33% of the total carbon footprint. The main contribution to the carbon footprint comes from the electricity consumption, more precisely, the electricity consumption in the centrifugation processes of the ultrafiltration and acid precipitation steps (representing a contribution of 99.6 and 99.9% of the carbon footprint for each step, respectively). The carbon footprint of the ATPS step is also dominated by electricity consumption, mainly by the centrifuge, contributing to 95% of the carbon footprint. However, it should be noted that the energy consumption of some equipment should be reviewed in view of the system industrial implementation.