Separation Options for Phosphorylated Osteopontin from Transgenic Microalgae Chlamydomonas reinhardtii

Correct folding and post-translational modifications are vital for therapeutic proteins to elicit their biological functions. Osteopontin (OPN), a bone regenerative protein present in a range of mammalian cells, is an acidic phosphoprotein with multiple potential phosphorylation sites. In this study, the ability of unicellular microalgae, Chlamydomonas reinhardtii, to produce phosphorylated recombinant OPN in its chloroplast is investigated. This study further explores the impact of phosphorylation and expression from a “plant-like” algae on separation of OPN. Chromatography resins ceramic hydroxyapatite (CHT) and Gallium-immobilized metal affinity chromatography (Ga-IMAC) were assessed for their binding specificity to phosphoproteins. Non-phosphorylated recombinant OPN expressed in E. coli was used to compare the specificity of interaction of the resins to phosphorylated OPN. We observed that CHT binds OPN by multimodal interactions and was better able to distinguish phosphorylated proteins in the presence of 250 mM NaCl. Ga-IMAC interaction with OPN was not selective to phosphorylation, irrespective of salt, as the resin bound OPN from both algal and bacterial sources. Anion exchange chromatography proved an efficient capture method to partially separate major phosphorylated host cell protein impurities such as Rubisco from OPN.


Introduction
The work horse of complex protein production for nearly 56% of recombinant proteins is the mammalian Chinese hamster ovary (CHO) cell line [1]. The ability of mammalian cells to properly fold and perform post-translational modifications (PTMs) of proteins, out-weighs the high cultivation costs and potential viral contamination issues. Escherichia coli is also extensively employed to produce recombinant proteins and does so at a fraction of the CHO cost but the resultant proteins often lack proper folding, PTMs and are prone to endotoxin contamination [2]. To overcome the disadvantages of mammalian and bacterial expression, the unicellular microalga Chlamydomonas reinhardtii has been explored as an alternative protein production platform. Specifically, the chloroplast of C. reinhardtii has been utilized for the production of difficult-to-express proteins that require specific folding and PTMs to exhibit biological function [3][4][5][6][7].
One such mammalian protein is osteopontin (OPN), also called bone sialoprotein-1 (BSP1), which is present in milk, bodily fluids and is responsible for bone development and regeneration. Other

Expression and Characterization of Recombinant Osteopontin in C. reinhardtii and E. coli
The transgenic cells of C. reinhardtii and E. coli expressing OPN were created by Mayfield's group at California Center of Algae Biotechnology, University of California San Diego (UCSD) [24] and biomass was grown at Texas A&M University. Quantification of OPN concentration in C. reinhardtii lysates was difficult because of the low expression levels and interference of lysate components with enzyme-linked immunosorbent assay (ELISA). In the absence of a reliable quantification method, we relied on band density analysis of western blots with ImageJ software (v.1.49, U. S. National Institutes of Health, Bethesda, MD, USA), using FLAG-affinity purified OPN as a standard. C. reinhardtii OPN expression levels varied between biomass batches and was on average 4× lower than OPN in E. coli cells, as indicated by similar band densities of 4× and 16× diluted C. reinhardtii and E. coli cell lysates, respectively ( Figure 1, lanes 2 & 3). In terms of initial OPN concentration expressed as a percent of total soluble protein (TSP), E. coli lysates contained 2.9% and algal lysates between 0.1% and 0.2% OPN. Full-length recombinant OPN migrated to an apparent molecular weight of 45 kDa. This observation was consistent with the varied molecular weight (75-44 kDa) of OPN reported in literature and was attributed to the highly acidic nature of the protein [25,26]. OPN expressed in the chloroplast of C. reinhardtii (Lane 2) consists of a major 45 kDa band and an N-terminal 35 kDa fragment. In addition to the 45 kDa band, E. coli lysate (Lane 3) contains several N-terminal fragments with apparent molecular weights corresponding to 40, 38, 35, 30 and 28 kDa. There were considerably fewer N-terminal fragments observed in algal lysates as compared to E. coli, suggesting that OPN is better protected from fragmentation in the chloroplast of C. reinhardtii than in the E. coli cytosol.
The FLAG-affinity-purified E. coli and C. reinhardtii OPNs were analyzed by 2D electrophoresis gel ( Figure 2) to estimate pI of various OPN forms. The 45 kDa OPN from C. reinhardtii resolved in a pI range from pH 3.5 to pH 4.5. A similar streak of varying pI (pH ~3.5 to 4.0) was observed with the 35 kDa fragment. The observed streaks suggest the presence of OPN isoforms that are likely due to heterogeneous post-translational phosphorylation and/or in vitro enzymatic de-phosphorylation. In contrast, E. coli 45 kDa OPN resolved into a spot localized at pH 4.5. The E. coli OPN fragments were also localized at approximately the same pH 4.5. Single spots instead of streaks of E. coli OPN is consistent with the expected lack of protein phosphorylation in E. coli [27][28][29]. Additional spots probably belonging to E. coli host protein were also detected on 2D gels.  Full-length recombinant OPN migrated to an apparent molecular weight of 45 kDa. This observation was consistent with the varied molecular weight (75-44 kDa) of OPN reported in literature and was attributed to the highly acidic nature of the protein [25,26]. OPN expressed in the chloroplast of C. reinhardtii (Lane 2) consists of a major 45 kDa band and an N-terminal 35 kDa fragment. In addition to the 45 kDa band, E. coli lysate (Lane 3) contains several N-terminal fragments with apparent molecular weights corresponding to 40, 38, 35, 30 and 28 kDa. There were considerably fewer N-terminal fragments observed in algal lysates as compared to E. coli, suggesting that OPN is better protected from fragmentation in the chloroplast of C. reinhardtii than in the E. coli cytosol.
The FLAG-affinity-purified E. coli and C. reinhardtii OPNs were analyzed by 2D electrophoresis gel ( Figure 2) to estimate pI of various OPN forms. The 45 kDa OPN from C. reinhardtii resolved in a pI range from pH 3.5 to pH 4.5. A similar streak of varying pI (pH~3.5 to 4.0) was observed with the 35 kDa fragment. The observed streaks suggest the presence of OPN isoforms that are likely due to heterogeneous post-translational phosphorylation and/or in vitro enzymatic de-phosphorylation. In contrast, E. coli 45 kDa OPN resolved into a spot localized at pH 4.5. The E. coli OPN fragments were also localized at approximately the same pH 4.5. Single spots instead of streaks of E. coli OPN is consistent with the expected lack of protein phosphorylation in E. coli [27][28][29]. Additional spots probably belonging to E. coli host protein were also detected on 2D gels. Full-length recombinant OPN migrated to an apparent molecular weight of 45 kDa. This observation was consistent with the varied molecular weight (75-44 kDa) of OPN reported in literature and was attributed to the highly acidic nature of the protein [25,26]. OPN expressed in the chloroplast of C. reinhardtii (Lane 2) consists of a major 45 kDa band and an N-terminal 35 kDa fragment. In addition to the 45 kDa band, E. coli lysate (Lane 3) contains several N-terminal fragments with apparent molecular weights corresponding to 40, 38, 35, 30 and 28 kDa. There were considerably fewer N-terminal fragments observed in algal lysates as compared to E. coli, suggesting that OPN is better protected from fragmentation in the chloroplast of C. reinhardtii than in the E. coli cytosol.
The FLAG-affinity-purified E. coli and C. reinhardtii OPNs were analyzed by 2D electrophoresis gel ( Figure 2) to estimate pI of various OPN forms. The 45 kDa OPN from C. reinhardtii resolved in a pI range from pH 3.5 to pH 4.5. A similar streak of varying pI (pH ~3.5 to 4.0) was observed with the 35 kDa fragment. The observed streaks suggest the presence of OPN isoforms that are likely due to heterogeneous post-translational phosphorylation and/or in vitro enzymatic de-phosphorylation. In contrast, E. coli 45 kDa OPN resolved into a spot localized at pH 4.5. The E. coli OPN fragments were also localized at approximately the same pH 4.5. Single spots instead of streaks of E. coli OPN is consistent with the expected lack of protein phosphorylation in E. coli [27][28][29]. Additional spots probably belonging to E. coli host protein were also detected on 2D gels.  The presence or absence of phosphorylation on the recombinant OPN was detected by phosphostaining an SDS-PAGE gel ( Figure 3). The stained band in lane 4 indicates that the FLAG-affinity purified 45 kDa OPN from C. reinhardtii is phosphorylated, while its counterpart from E. coli in lane 5 is not stained and, thus, not detectable. The phosphostained band in lane 1 belongs to 45 kDa ovalbumin, which served as a phosphorylated protein marker. The absence of phosphostaining of the 35 kDa band may indicate little or no phosphorylation. The presence or absence of phosphorylation on the recombinant OPN was detected by phosphostaining an SDS-PAGE gel ( Figure 3). The stained band in lane 4 indicates that the FLAGaffinity purified 45 kDa OPN from C. reinhardtii is phosphorylated, while its counterpart from E. coli in lane 5 is not stained and, thus, not detectable. The phosphostained band in lane 1 belongs to 45 kDa ovalbumin, which served as a phosphorylated protein marker. The absence of phosphostaining of the 35 kDa band may indicate little or no phosphorylation. The detection of phosphorylated residues in algal 45 kDa OPN provided motivation to look into the specificity of algal OPN interaction with ceramic hydroxyapatite (CHT) and immobilized transition metal (Ga, Fe, Zn) resins.

Ceramic Hydroxyapatite Chromatography
The results from equilibrium adsorption of OPN to CHT resin from clarified C. reinhardtii and E. coli lysates are summarized in Figure 4.  The detection of phosphorylated residues in algal 45 kDa OPN provided motivation to look into the specificity of algal OPN interaction with ceramic hydroxyapatite (CHT) and immobilized transition metal (Ga, Fe, Zn) resins.

Ceramic Hydroxyapatite Chromatography
The results from equilibrium adsorption of OPN to CHT resin from clarified C. reinhardtii and E. coli lysates are summarized in Figure 4. The presence or absence of phosphorylation on the recombinant OPN was detected by phosphostaining an SDS-PAGE gel ( Figure 3). The stained band in lane 4 indicates that the FLAGaffinity purified 45 kDa OPN from C. reinhardtii is phosphorylated, while its counterpart from E. coli in lane 5 is not stained and, thus, not detectable. The phosphostained band in lane 1 belongs to 45 kDa ovalbumin, which served as a phosphorylated protein marker. The absence of phosphostaining of the 35 kDa band may indicate little or no phosphorylation. The detection of phosphorylated residues in algal 45 kDa OPN provided motivation to look into the specificity of algal OPN interaction with ceramic hydroxyapatite (CHT) and immobilized transition metal (Ga, Fe, Zn) resins.

Ceramic Hydroxyapatite Chromatography
The results from equilibrium adsorption of OPN to CHT resin from clarified C. reinhardtii and E. coli lysates are summarized in Figure 4.   Step elutions were performed with sodium phosphate (NaP) buffer, pH 6.8. Data (% OPN) were generated by ImageJ analysis of anti-FLAG western blots and normalized by the amount in the lysate. Error bars indicate standard deviation of triplicate ImageJ measurements.
The adsorbed E. coli and C. reinhardtii OPN, respectively, were eluted using increasing strength (100 to 1500 mM) of NaP buffer. The bar graph in Figure 4a reveals that the majority (>70%) of OPN from both lysates was adsorbed to the CHT resin in the absence of salt. The relatively low amount of C. reinhardtii OPN detected in the supernatant (30%) and its absence in the wash fractions indicates strong interactions between OPN and CHT. The bound C. reinhardtii OPN began to elute with 100 mM NaP and continued eluting through 1000 mM phosphate concentration. A significant fraction of adsorbed OPN eluted with the 250 mM (18%) and 500 mM NaP (33%) steps, respectively. Since we could not close the mass balance on OPN in the eluted fractions, even after 5 column volumes (CV) elution with 1500 mM NaP, we resorted to harsher desorption conditions using 100 mM NaOH. The strongly adsorbed OPN, estimated at 41% of initial OPN by ImageJ analysis, was completely eluted with 100 mM NaOH. Western blots ( Figure 5b) show that the 35 kDa fraction was eluted with 100 mM NaP and that the 45 kDa form interacted even more strongly, requiring higher phosphate concertation >250 mM NaP and 100 mM NaOH for desorption.  Step elutions were performed with sodium phosphate (NaP) buffer, pH 6.8. Data (% OPN) were generated by ImageJ analysis of anti-FLAG western blots and normalized by the amount in the lysate. Error bars indicate standard deviation of triplicate ImageJ measurements.
The adsorbed E. coli and C. reinhardtii OPN, respectively, were eluted using increasing strength (100 to 1500 mM) of NaP buffer. The bar graph in Figure 4a reveals that the majority (>70%) of OPN from both lysates was adsorbed to the CHT resin in the absence of salt. The relatively low amount of C. reinhardtii OPN detected in the supernatant (30%) and its absence in the wash fractions indicates strong interactions between OPN and CHT. The bound C. reinhardtii OPN began to elute with 100 mM NaP and continued eluting through 1000 mM phosphate concentration. A significant fraction of adsorbed OPN eluted with the 250 mM (18%) and 500 mM NaP (33%) steps, respectively. Since we could not close the mass balance on OPN in the eluted fractions, even after 5 column volumes (CV) elution with 1500 mM NaP, we resorted to harsher desorption conditions using 100 mM NaOH. The strongly adsorbed OPN, estimated at 41% of initial OPN by ImageJ analysis, was completely eluted with 100 mM NaOH. Western blots ( Figure 5b) show that the 35 kDa fraction was eluted with 100 mM NaP and that the 45 kDa form interacted even more strongly, requiring higher phosphate concertation >250 mM NaP and 100 mM NaOH for desorption.
As the bar graph in Figure 4a indicates, about 23% of E. coli OPN remained in the supernatant and 3% was found in the wash fractions. In contrast to C. reinhardtii, 72% of E. coli OPN eluted from hydroxyapatite resin in the 100-500 mM phosphate buffer range. As expected from the mass balance of OPN fractions eluted with the three NaP steps changes, the application of 100 mM NaOH did not yield a significant amount of OPN (only 8% of initial amount). The significant difference in the amount of 45 kDa OPN eluted with 100 mM NaP (7% algal OPN consisting of mainly 35 kDa OPN vs. 11% of E. coli 45 kDa OPN) and 100 mM NaOH (41% algal OPN vs. 8% E. coli OPN) steps indicate that E. coli and C. reinhardtii OPN interact differently with heterogeneous binding sites of CHT resin.
could not close the mass balance on OPN in the eluted fractions, even after 5 column volumes (CV) elution with 1500 mM NaP, we resorted to harsher desorption conditions using 100 mM NaOH. The strongly adsorbed OPN, estimated at 41% of initial OPN by ImageJ analysis, was completely eluted with 100 mM NaOH. Western blots (Figure 5b) show that the 35 kDa fraction was eluted with 100 mM NaP and that the 45 kDa form interacted even more strongly, requiring higher phosphate concertation >250 mM NaP and 100 mM NaOH for desorption.  To determine if the different interaction strengths of OPN with CHT were due to phosphorylation i.e., phosphorylated (C. reinhardtii) vs. non-phosphorylated (E. coli) OPN, the adsorption and elution experiments were repeated in the presence of 250 mM NaCl (Supplementary data Figure S1). The inclusion of 250 mM NaCl in the adsorption and elution buffers was intended to reduce ionic interactions of amino acid side chains of OPN with Ca 2+ and PO 4 3− sites on the CHT without disrupting the much stronger coordinate bonding between OPN's phosphoryl groups and Ca 2+ on CHT surface. The results shown in Figure 4b clearly demonstrate a difference in the adsorption of E. coli and C. reinhardtii OPN in the presence of salt. About 83% of E. coli OPN remains in the supernatant at the end of the adsorption period compared to the barely detectable amount (~1%) OPN in C. reinhardtii supernatants. An additional 10% of E. coli OPN was removed in the wash step while none of C. reinhardtii OPN was desorbed during the washing. Comparison of the data presented in Figure 4 demonstrates that the adsorption of E. coli OPN to CHT was significantly weakened by 250 mM NaCl, whereas that of C. reinhardtii was not. The latter, somewhat atypical behavior of increased C. reinhardtii OPN binding, can be explained by NaCl suppression of electrostatic repulsions between phosphoryl groups on OPN and PO 4 3− sites on the CHT [30]. The suppression of electrostatic repulsions, results in enhanced interactions between algal OPN's phosphoryl groups with the Ca 2+ on CHT surface. Therefore, E. coli OPN probably binds to hydroxyapatite by electrostatic interactions as depicted in Figure 6 below. This observation is consistent with the reported multi-modal interaction of proteins with CHT via charge interactions between acidic and basic amino acid side chains with the Ca 2+ and PO 4 3− moieties on CHT, respectively, which can be reversed by increasing the ionic strength of the buffer [31].
phosphoryl groups with the Ca 2+ on CHT surface. Therefore, E. coli OPN probably binds to hydroxyapatite by electrostatic interactions as depicted in Figure 6 below. This observation is consistent with the reported multi-modal interaction of proteins with CHT via charge interactions between acidic and basic amino acid side chains with the Ca 2+ and PO4 3− moieties on CHT, respectively, which can be reversed by increasing the ionic strength of the buffer [31].  Based on these data, we believe that protein phosphorylation made a difference in the interaction of OPN with ceramic hydroxyapatite resin for the following reasons. First, compared to E. coli OPN, the addition of NaCl had a minimal or no measurable effect on C. reinhardtii OPN adsorption and elution from CHT (Figure 4). The presence of NaCl weakens E. coli OPN interaction with CHT as evidenced by 83% of OPN remaining unbound in supernatant and the residual 17% being desorbed during wash and first step 100 mM NaP step (Figure 4b). Second, the elution of C. reinhardtii OPN, irrespective of ionic strength, required displacement by increasing phosphate concentration and complete desorption of phosphorylated OPN by the application of 100 mM NaOH. This line of evidence suggests that C. reinhardtii OPN adsorption to CHT is mediated by the stronger Ca 2+ -PO 3 2− coordinate binding rather than simple charge interactions. And, the unusually strong adsorption of the 45 kDa C. reinhardtii OPN form, which required 100 mM NaOH for elution (Figure 5b), is probably the result of interaction of multiple, closely positioned phosphoryl (PO 3 2− ) groups on OPN molecule with calcium moieties (Ca 2+ ) of CHT resin. To assess the specificity and purification capability of CHT, we compared host protein binding and elution by SDS-PAGE (Figure 5a). Data in Figure 5a indicate that a significant number of algal proteins would co-elute with OPN. The major OPN "competitor" appears to be acidic, phosphorylated chloroplast protein, Rubisco [32], as large (~50 kDa) and small subunit (~15 kDa) can be seen in 1000 mM NaP and 100 mM NaOH fractions (Figure 5a, lane 8 and 10 respectively). The identity of Rubisco was confirmed by western blots using anti-Rubisco large and small subunit antibodies (Supplementary data Figure S3). Thus, the presence of Rubisco in C. reinhardtii lysates would eventually reduce the resin breakthrough capacity and OPN purification fold, both of which are important factors in process scale-up. A possible solution to CHT application for OPN purification is to remove Rubisco and other interfering host protein impurities from the clarified lysates before OPN binding to CHT.

Gallium-Immobilized Metal Affinity Chromatography (Ga-IMAC)
Gallium, iron and zinc immobilized on Chelating Sepharose are known to bind phospho-peptides by coordinate bonds [33][34][35]. Gallium, having higher affinity than other metals for phosphorylated peptides [36], has been selected in this study to determine if differential interactions between Ga 3+ and phosphorylated protein residues could be utilized for the separation of C. reinhardtii OPN from non-phosphorylated proteins. As before, E. coli OPN has been used to assess the specificity of OPN-Ga interactions due to phosphorylation. To eliminate any ionic interactions between the protein and Ga-IMAC resin, OPN adsorption and elution was tested in the presence of 150 mM and 500 mM NaCl.
Results from Ga-IMAC batch adsorption chromatography are shown in Figure 7. Approximately 20% of C. reinhardtii OPN remained unbound in the supernatant compared to only 7% of E. coli OPN. An additional 8% of algal OPN was removed during the resin wash step. The majority of the unbound and wash fractions contained 35 kDa form of algal OPN (Supplementary data Figure S2) which, as discussed above, did not bind strongly to CHT either. The adsorbed C. reinhardtii OPN started to elute with 100 mM NaP with 8%-12% of 45 kDa OPN continuing to elute in each step elution up to 1000 mM NaP concentration. The major fraction (21%) of bound E. coli OPN was in the second step elution with 250 mM NaP (Figure 7). The remaining bound E. coli OPN was eluted with 500 and 1000 mM NaP.  Step elutions were performed with sodium phosphate (NaP) buffer, pH 5.5. Data (% OPN) were generated by ImageJ analysis of anti-FLAG western blots and normalized by the amount in the lysate. Error bars indicate standard deviation of triplicate ImageJ measurements.

Anion Exchange Chromatography
Previous studies with E. coli OPN have used anion exchange chromatography as one of the purification steps [37,38]. The use of anion exchange adsorption is a natural choice for highly acidic proteins like OPN. We decided to compare the interaction of C. reinhardtii and E. coli OPN with a strong anion exchange resin (Capto Q, GE Healthcare Life Sciences, Marlborough, MA, USA). The analysis of binding and elution pools of OPN from both sources are summarized in the bar graph ( Figure 8) and western blots ( Figure 9). When bound at pH 7.0 and a low conductivity of 4 mS, OPN bound strongly and specifically to the resin. No C. reinhardtii OPN (<1%) was detected by western blot in the supernatants and the washes (Figure 8). The majority of C. reinhardtii 35 kDa OPN fragment, which constitutes 11% total OPN was eluted with 100 mM NaCl (Figure 9a lane 5) suggesting weaker interaction with the quaternary amine than the intact 45 kDa form. The 45 kDa OPN from C. reinhardtii eluted in 200-500 mM NaCl range reaching a peak concentration in the 300 mM NaCl elution fraction. A faint band (~1.4%) was also observed with 1000 mM NaCl elution (Figure 9a lane 9). E. coli OPN also bound on anion exchange column strongly, demonstrated by the lack of a band in the supernatant and washes (Figure 9b lane 3 and 4). The elution of intact 45 kDa E. coli OPN starts earlier than algal OPN, with 16% in 100 mM NaCl elution (Figure 9b, lane 5) and elution continues till 500 mM NaCl with the highest concentration (37%) at 300 mM NaCl elution similar to C. reinhardtii OPN. The elution region for E. coli OPN is narrower under the same conditions with very little to no OPN over 500 mM NaCl elution. This observation corresponds to the 2D gel electrophoresis results, where E. coli OPN fragments were localized to spots, reflecting homogenous pI that should result in a uniform binding strength to the Capto Q resin. C. reinhardtii OPN had horizontal streaks on the 2D gel indicating a broader pI range of OPN due to heterogeneous phosphorylation. Hence, the elution profile of the adsorbed OPN would require a broader range of salt concentration to break the "heterogeneous" strength interactions with the resin. The ionic strength difference in the onset of 45 kDa OPN elution, 200 mM (C. reinhardtii) vs. 100 mM NaCl (E. coli OPN), probably reflects higher acidity of phosphorylated C. reinhardtii OPN. Step elutions were performed with sodium phosphate (NaP) buffer, pH 5.5. Data (% OPN) were generated by ImageJ analysis of anti-FLAG western blots and normalized by the amount in the lysate. Error bars indicate standard deviation of triplicate ImageJ measurements.
The binding mechanism of C. reinhardtii OPN to Ga-IMAC apparently does not involve phosphorylated residues because non-phosphorylated E. coli OPN binds and elutes similarly to phosphorylated C. reinhardtii OPN (Figure 7). In addition, the similar binding and elution profiles observed with Fe-IMAC resin suggests that OPN interacts with Ga 3+ and Fe 3+ via the multiple carboxylic acidic containing amino acids occurring in close proximity (carboxyl clusters) present in both E. coli and C. reinhardtii OPN. Unfortunately, we could not confirm this hypothesis because adsorption at or below pH 4.5 (pKa of carboxyl groups) was not possible due to OPN precipitation at that pH range. Based on our data, it appears that protein interactions with transition metals (Ga 3+ or Fe 3+ ) are not confined specifically to phosphorylated residues and, therefore, the application of Ga-IMAC resin would not be useful for the separation of C. reinhardtii OPN from other acidic algae proteins present in the lysates.

Anion Exchange Chromatography
Previous studies with E. coli OPN have used anion exchange chromatography as one of the purification steps [37,38]. The use of anion exchange adsorption is a natural choice for highly acidic proteins like OPN. We decided to compare the interaction of C. reinhardtii and E. coli OPN with a strong anion exchange resin (Capto Q, GE Healthcare Life Sciences, Marlborough, MA, USA). The analysis of binding and elution pools of OPN from both sources are summarized in the bar graph ( Figure 8) and western blots (Figure 9). When bound at pH 7.0 and a low conductivity of 4 mS, OPN bound strongly and specifically to the resin. No C. reinhardtii OPN (<1%) was detected by western blot in the supernatants and the washes (Figure 8). The majority of C. reinhardtii 35 kDa OPN fragment, which constitutes 11% total OPN was eluted with 100 mM NaCl (Figure 9a lane 5) suggesting weaker interaction with the quaternary amine than the intact 45 kDa form. The 45 kDa OPN from C. reinhardtii eluted in 200-500 mM NaCl range reaching a peak concentration in the 300 mM NaCl elution fraction. A faint band (~1.4%) was also observed with 1000 mM NaCl elution (Figure 9a lane 9). E. coli OPN also bound on anion exchange column strongly, demonstrated by the lack of a band in the supernatant and washes (Figure 9b lane 3 and 4). The elution of intact 45 kDa E. coli OPN starts earlier than algal OPN, with 16% in 100 mM NaCl elution (Figure 9b, lane 5) and elution continues till 500 mM NaCl with the highest concentration (37%) at 300 mM NaCl elution similar to C. reinhardtii OPN. The elution region for E. coli OPN is narrower under the same conditions with very little to no OPN over 500 mM NaCl elution. This observation corresponds to the 2D gel electrophoresis results, where E. coli OPN fragments were localized to spots, reflecting homogenous pI that should result in a uniform binding strength to the Capto Q resin. C. reinhardtii OPN had horizontal streaks on the 2D gel indicating a broader pI range of OPN due to heterogeneous phosphorylation. Hence, the elution profile of the adsorbed OPN would require a broader range of salt concentration to break the "heterogeneous" strength interactions with the resin. The ionic strength difference in the onset of 45 kDa OPN elution, 200 mM (C. reinhardtii) vs. 100 mM NaCl (E. coli OPN), probably reflects higher acidity of phosphorylated C. reinhardtii OPN.
SDS-PAGE ( Figure 10) reveals that partial separation of OPN from C. reinhardtii host protein by anion exchange chromatography is possible. The majority (>99%) of recombinant algal OPN was adsorbed on the resin, about 49% of the host protein was present in the supernatant (lane 3) and wash fractions (lane 4) while the rest of the host protein was co-adsorbed with OPN. The majority of host protein was eluted in 100-300 mM NaCl fractions (lanes 5-7), whereas, OPN eluted in 200-500 mM NaCl fractions (lanes 6-8). Thus, by pooling 200 to 500 mM NaCl elution fractions at least forty-fold increase of OPN purity (from 0.2% to 8% TSP) and 80% recovery can be achieved by anion exchange chromatography. The anion exchange data also demonstrate that a substantial fraction of phosphorylated host protein (Rubisco), which was identified above as the major interfering host protein for OPN purification with CHT, can be partially separated from OPN by excluding the 200 mM NaCl elution and pooling only the 300 and 500 mM fraction for further OPN isolation. In the latter case, OPN purity would increase to 20% at the expense of reduced recovery (<60%) as calculated from Figure 10.   Step elutions were performed with sodium chloride in 50 mM Tris pH 7.0. Data (% OPN) were generated by ImageJ analysis of anti-FLAG western blots and normalized by the amount in the lysate. Error bars indicate standard deviation of triplicate ImageJ measurements. SDS-PAGE ( Figure 10) reveals that partial separation of OPN from C. reinhardtii host protein by anion exchange chromatography is possible. The majority (>99%) of recombinant algal OPN was adsorbed on the resin, about 49% of the host protein was present in the supernatant (lane 3) and wash fractions (lane 4) while the rest of the host protein was co-adsorbed with OPN. The majority of host protein was eluted in 100-300 mM NaCl fractions (lanes 5-7), whereas, OPN eluted in 200-500 mM NaCl fractions (lanes 6-8). Thus, by pooling 200 to 500 mM NaCl elution fractions at least forty-fold increase of OPN purity (from 0.2% to 8% TSP) and 80% recovery can be achieved by anion exchange chromatography. The anion exchange data also demonstrate that a substantial fraction of Sample fraction phosphorylated host protein (Rubisco), which was identified above as the major interfering host protein for OPN purification with CHT, can be partially separated from OPN by excluding the 200 mM NaCl elution and pooling only the 300 and 500 mM fraction for further OPN isolation. In the latter case, OPN purity would increase to 20% at the expense of reduced recovery (<60%) as calculated from Figure 10.

Cultivation of Recombinant Chlamydomonas reinhardtii Strain Expressing Bovine OPN
The chloroplast genome of C. reinhardtii strain cc1670 was modified to express bovine OPN by California Center of Algae Biotechnology, UCSD [24]. The psbA gene was replaced with cDNA for bovine OPN (GenBank accession number NP_776612.1) by homologous recombination. The modified strain was rescued by re-insertion of the psbA gene at a different location in the genome. A 1×-FLAG epitope tag (DYKDDDDKS) was inserted at the N-terminal of OPN sequence for identification and quantification, with an enterokinase cleavage site to facilitate removal after OPN purification.
Recombinant algal cells were grown on 15% Tris-Acetate-Phosphate (TAP) agar plates at room temperature. The colonies were transferred into TAP media containing 0.1% Hutner's trace solution (C. reinhardtii research center) and 150 µg/mL ampicillin [39,40]. The cells were grown in sterile conditions with constant 125 rpm shaking, 200-300 µmols m −2 ·s −1 of cool white light at room temperature. They were harvested in mid-log phase at OD750 = 0.5 by centrifugation at 3500 rpm and 4 °C (Allegra 25R, Beckman Coulter, Brea, CA, USA). The wet biomass was stored at −80 °C.

Recombinant E. coli Strain and Cultivation
E. coli strain BL21 was modified to express bovine OPN with an N-terminal FLAG sequence for OPN identification and quantification similar to C. reinhardtii by California Center of Algae

Cultivation of Recombinant Chlamydomonas reinhardtii Strain Expressing Bovine OPN
The chloroplast genome of C. reinhardtii strain cc1670 was modified to express bovine OPN by California Center of Algae Biotechnology, UCSD [24]. The psbA gene was replaced with cDNA for bovine OPN (GenBank accession number NP_776612.1) by homologous recombination. The modified strain was rescued by re-insertion of the psbA gene at a different location in the genome. A 1×-FLAG epitope tag (DYKDDDDKS) was inserted at the N-terminal of OPN sequence for identification and quantification, with an enterokinase cleavage site to facilitate removal after OPN purification.
Recombinant algal cells were grown on 15% Tris-Acetate-Phosphate (TAP) agar plates at room temperature. The colonies were transferred into TAP media containing 0.1% Hutner's trace solution (C. reinhardtii research center) and 150 µg/mL ampicillin [39,40]. The cells were grown in sterile conditions with constant 125 rpm shaking, 200-300 µmols m −2 ·s −1 of cool white light at room temperature. They were harvested in mid-log phase at OD 750 = 0.5 by centrifugation at 3500 rpm and 4 • C (Allegra 25R, Beckman Coulter, Brea, CA, USA). The wet biomass was stored at −80 • C.

Recombinant E. coli Strain and Cultivation
E. coli strain BL21 was modified to express bovine OPN with an N-terminal FLAG sequence for OPN identification and quantification similar to C. reinhardtii by California Center of Algae Biotechnology, UCSD. The recombinant strain was ampicillin resistant and OPN expression was controlled by the arabinose promotor.
1 mL glycerol stock of the recombinant strain was used to grow overnight inoculum (5 mL) in Luria-Bertani (LB) media (150 µg/mL ampicillin) at 37 • C and 200 rpm. The overnight inoculum was transferred to 1 L Terrific Broth (Sigma-Aldrich, St. Louis, MO, USA, Cat # T0918) and E. coli cells were grown at 37 • C, 200 rpm to OD 600 = 0.8. OPN expression was induced overnight with 0.2% (w/v) arabinose at 37 • C and 200 rpm. The wet biomass was harvested by centrifugation and stored at −80 • C.

Batch Adsorption Chromatography and Purification Studies
Frozen biomass was thawed to room temperature and suspended in 1 g biomass:10 mL lysis buffer (50 mM Tris, 150 mM NaCl, 1 mM Ethylenediaminetetraacetic acid (EDTA), pH 7.4) containing protease and phosphatase inhibitor cocktail pills (Roche, Mannheim, Germany). Cells were lysed by sonication (15 min for C. reinhardtii and 5 min for E. coli) in 30 s on/off intervals at 4 • C using a sonicator (Sonifier 250, Branson, Danbury, CT, USA) at 30% output control and 30% duty cycle with a micro probe (1/8" micro tip A3-561 Branson, Danbury, CT, USA). Cell lysates were centrifuged (10,000× g for 20 min at 4 • C, Allegra 25R centrifuge, Beckman Coulter, Brea, CA, USA) and cell-free supernatants were filtered through 0.22 µm sterile polyethersulfone membrane syringe filter (30 mm diameter, Genesee Scientific, San Diego, CA, USA) and this was considered the clarified lysate for chromatography experiments. Protease and phosphatase inhibitor cocktail pills (Roche, Mannheim, Germany) in their recommended concentrations were added to all buffers used in the chromatography experiments.
FLAG-affinity chromatography. FLAG-affinity purification was used to obtain highly purified recombinant OPN for analysis. 1 mL of anti-FLAG affinity gel (Sigma-Aldrich, St. Louis, MO, USA, Cat # A4596) was equilibrated in lysis buffer for 30 CV and mixed with 5 mL of C. reinhardtii clarified lysate or 1 mL of E. coli clarified lysate. Incubation of resin and lysate was carried out for 2 h at 4 • C with continuous end-over-end mixing in a Glas-Col rotor (Glas-Col LLC, Terre Haute, IN, USA) at~33 rpm (40% speed control). The resin was separated from the clarified lysate by centrifuging at 6000× g for 5 min and the resin was loaded on a Bio Spin disposable chromatography column (Bio-Rad, Hercules, CA, USA, Cat # 732-6008). Non-specifically bound proteins were removed by washing with 10 CV of 50 mM Tris, 150 mM NaCl, 1 mM EDTA, pH 7.4. Proteins bound to resin were eluted using 5 CV of 100 mM Glycine, 400 mM NaCl, pH 3.5. The pH of the eluted proteins was immediately raised to 8.0 by adding 50 µL of 1 M Tris-HCl, to avoid protein denaturation at low pH.
Ceramic hydroxyapatite chromatography. Clarified lysates were dialyzed against 5 mM sodium phosphate (NaP) pH 6.8 for 2 h at 4 • C using snake-skin membrane (3.5 kDa molecular weight cut-off, Thermo Fisher Scientific, Waltham, MA, USA, Cat #68035). The clarified lysate (2.5 mL of C. reinhardtii clarified lysate, 0.66 mL of E. coli clarified lysate) was incubated for 1 h at room temperature by end-over-end mixing with 0.5 mL of ceramic hydroxyapatite resin (Bio-Rad, Hercules, CA, USA, Cat # 1584000) pre-equilibrated in 5 mM NaP pH 6.8 and resin was loaded on a chromatography column. The resin was washed with 10 CV of 5 mM NaP pH 6.8 followed by protein elution using a step-wise increasing concentration (100 mM, 250 mM, 500 mM, 1000 mM and 1500 mM) of NaP buffers at pH 6.8 with 5 CV of each buffer concentration.
Immobilized Metal Affinity Chromatography (IMAC). Gallium chloride at 400 mM concentration was immobilized on Chelating Sepharose FF (GE Healthcare Life Sciences, Marlborough, MA, USA, washes with PBS-T, blocking solution of 0.3% BSA in PBS was added to the 96-will plate and incubated for 2 h at 37 • C followed by similar wash and incubation with anti-FLAG M2 antibody conjugated with horse radish peroxidase (Sigma-Aldrich, St. Louis, MO, USA, Cat # A8592) diluted 1:5000 in PBS. TMB developer (Sigma-Aldrich, St. Louis, MO, USA, Cat # T0440) was used to develop the reaction and stopped with 2 M HCl, the absorbance was read at 450 nm.
FLAG-affinity purified OPN samples at 50 µg/mL concentration were analyzed for their pI and molecular weight using 2D gel electrophoresis performed by Protein Chemistry Lab, Texas A&M University, USA. pI resolution was performed with immobilized pH gradient technology (IPG Dry Strip, GE Healthcare Life Sciences, Marlborough, MA, USA) followed by electrophoresis on GE Healthcare Tall Mighty Small system (8 × 10 cm). The resolved gel was silver stained to visualize OPN and specificity was confirmed by anti-FLAG western blot. The pI values of OPN were estimated using the standard curve established by the IPG Dry Strip manufacturer.
Phosphostaining was performed on FLAG-affinity purified OPN samples that were de-salted and de-lipidated by chloroform-methanol precipitation [41]. SDS-PAGE was performed as above and the gel was incubated in fixing solution (50% methanol, 10% acetic acid and 40% MilliQ water) overnight at room temperature on the gel rocker. The gel was washed 3× with MilliQ water and stained with Pro-Q diamond phosphostain (Thermo Fisher Scientific, Waltham, MA, USA, Cat # MPP33301) for 90 min in the dark. Destaining solution (20% acetonitrile in 50 mM sodium acetate pH 4.0) was added 3× with 30 min incubation each and the gel was finally washed 4× with MilliQ water. Gel imaging was performed on Typhoon Trio fluorescent imaging system (GE Healthcare Life Sciences, Marlborough, MA, USA) at excitation 532 nm green and 560 nm bandpass emission filter.

Conclusions
Osteopontin was successfully expressed in the chloroplast of C. reinhardtii, a novel alternative platform for recombinant protein production, capable of performing phosphorylation as a post-translational modification. The C. reinhardtii strain expressed two major OPN isoforms (45 kDa and 35 kDa) with pI ranging from 3.5 to 4.5 due to heterogeneous post-translational phosphorylation.
Several chromatography resins were evaluated for isolating OPN by utilizing interactions with the phosphorylation and the unique polyaspartate sequence. Ceramic hydroxyapatite resin demonstrated affinity for acidic and phosphorylated proteins and adsorbed the phosphorylated OPN (45 kDa) isoform strongly. The adsorption and co-elution of phosphorylated host proteins like Rubisco along with C. reinhardtii OPN, reduced the purification efficiency of CHT resin. Since CHT could not distinguish between major phosphorylated impurities (Rubisco) and OPN, it is not ideal for separation of phosphorylated recombinant proteins from C. reinhardtii.
Gallium-IMAC, did not exhibit a preferential binding of phosphorylated OPN. Both phosphorylated OPN from C. reinhardtii and non-phosphorylated OPN from E. coli appeared to have similar binding and elution behavior with Ga-IMAC. Our study suggests that Ga 3+ interacts with multiple carboxyl groups of OPN and host proteins and has no specific preference for the phosphorylated algal OPN 45 kDa form. As such, Ga-IMAC resin does not offer any advantage over CHT as a potential purification method for phosphorylated OPN forms.
Anion exchange chromatography is a viable adsorption step for capturing OPN from both E. coli and C. reinhardtii clarified lysates. The anion exchange chromatography conducted under optimal adsorption, resin-wash and elution conditions removes as much as 50% of host protein while maintaining OPN recovery yield at 80%. Other anionic protein impurities bound to the resin can be partially separated from OPN based on their strength of charge interactions with the amine ligand resulting in a 20% pure OPN fraction.
Because of low accumulation levels in our current C. reinhardtii strain leading to low starting concentration of OPN in the clarified C. reinhardtii lysate (0.1%-0.2% TSP), the development of a scalable OPN purification train will probably require three or more chromatography steps.