Colorful Packages: Encapsulation of Fluorescent Proteins in Complex Coacervate Core Micelles

Encapsulation of proteins can be beneficial for food and biomedical applications. To study their biophysical properties in complex coacervate core micelles (C3Ms), we previously encapsulated enhanced green fluorescent protein (EGFP) and its monomeric variant, mEGFP, with the cationic-neutral diblock copolymer poly(2-methyl-vinyl-pyridinium)n-b-poly(ethylene-oxide)m (P2MVPn-b-PEOm) as enveloping material. C3Ms with high packaging densities of fluorescent proteins (FPs) were obtained, resulting in a restricted orientational freedom of the protein molecules, influencing their structural and spectral properties. To address the generality of this behavior, we encapsulated seven FPs with P2MVP41-b-PEO205 and P2MVP128-b-PEO477. Dynamic light scattering and fluorescence correlation spectroscopy showed lower encapsulation efficiencies for members of the Anthozoa class (anFPs) than for Hydrozoa FPs derived from Aequorea victoria (avFPs). Far-UV CD spectra of the free FPs showed remarkable differences between avFPs and anFPs, caused by rounder barrel structures for avFPs and more elliptic ones for anFPs. These structural differences, along with the differences in charge distribution, might explain the variations in encapsulation efficiency between avFPs and anFPs. Furthermore, the avFPs remain monomeric in C3Ms with minor spectral and structural changes. In contrast, the encapsulation of anFPs gives rise to decreased quantum yields (monomeric Kusabira Orange 2 (mKO2) and Tag red fluorescent protein (TagRFP)) or to a pKa shift of the chromophore (FP variant mCherry).


Introduction
Fluorescent proteins (FPs) are nowadays indispensable in life sciences [1][2][3][4]. The discovery of FPs started in the early 1960s with studies on the identification of the glow of jellyfish from Aequorea victoria by Osamu Shimomura [5]. The protein emitting the green light was called green fluorescent protein (GFP) [6] and its sequence was obtained in 1992 by Prasher [7]. In the following years, a wide variety of GFP variants with different colors and improved brightness and stability were developed. However, there were no GFP variants with emission maxima above 527 nm [4]. This limitation was overcome by cloning of GFP homologs from non-bioluminescent reef corals of the Anthozoa class [8][9][10][11]. From this class, a palette of FPs became available emitting at longer wavelengths. Consequently, the number of FP has a high extinction coefficient compared to other FPs (ε SYFP2 = 101 000 M −1 ·cm −1 , see Table S2) making it a very suitable acceptor in FRET-pairs [24].
The chromophore structures of FPs from Anthozoa species generally have more extended π-systems, enabling higher excitation and emission wavelengths. Such a type of fluorophore is found in mKO2, which evolved from a fluorescent protein of the mushroom coral Fungia concinna, with a cysteine located at position 65 (mEGFP numbering, Figure 1F) [9,25]. mKO2 is very useful in multicolor imaging applications as it can be combined with cyan, green, yellow, and red FPs. A protein with an almost similar excitation maximum as mKO2 is TagRFP ( Figure 1G), but this protein has been derived from the sea anemone Entacmaea quadricolor [10]. TagRFP has an even more extended π-system than mKO2, because it has a methionine located at position 65 (mEGFP numbering, Figure 1G). Next to that, TagRFP is one of the few FPs bearing a trans-isomerized chromophore. A protein that also contains a methionine at position 65 (mEGFP numbering) is mCherry, one of the "mFruit" FPs derived from Discosoma species [26,27]. mCherry shows a high photostability and its chromophore is rapidly formed ( Figure 1H), which makes it very suitable as a FRET acceptor in combination with EGFP in fluorescence-lifetime imaging microscopy (FLIM) studies [28].
In this research, we used the diblock copolymers P2MVP 41 -b-PEO 205 and P2MVP 128 -b-PEO 477 to form C3Ms in combination with the above-mentioned FPs. We characterized the C3Ms with dynamic light scattering (DLS) and fluorescence correlation spectroscopy (FCS), and explored the effects of packing on the FPs with circular dichroism (CD) and fluorescence spectral analysis. The experimental data, and in particular the observed encapsulation efficiencies, are discussed in relation to what is known about the structural features of the FPs. The chromophore structures of FPs from Anthozoa species generally have more extended πsystems, enabling higher excitation and emission wavelengths. Such a type of fluorophore is found in mKO2, which evolved from a fluorescent protein of the mushroom coral Fungia concinna, with a cysteine located at position 65 (mEGFP numbering, Figure 1F) [9,25]. mKO2 is very useful in multicolor imaging applications as it can be combined with cyan, green, yellow, and red FPs. A protein with an almost similar excitation maximum as mKO2 is TagRFP ( Figure 1G), but this protein has been derived from the sea anemone Entacmaea quadricolor [10]. TagRFP has an even more extended π-system than mKO2, because it has a methionine located at position 65 (mEGFP numbering, Figure 1G). Next to that, TagRFP is one of the few FPs bearing a trans-isomerized chromophore. A protein that also contains a methionine at position 65 (mEGFP numbering) is mCherry, one of the "mFruit" FPs derived from Discosoma species [26,27]. mCherry shows a high photostability and its chromophore is rapidly formed ( Figure 1H), which makes it very suitable as a FRET acceptor in combination with EGFP in fluorescence-lifetime imaging microscopy (FLIM) studies [28].
In this research, we used the diblock copolymers P2MVP41-b-PEO205 and P2MVP128-b-PEO477 to form C3Ms in combination with the above-mentioned FPs. We characterized the C3Ms with dynamic light scattering (DLS) and fluorescence correlation spectroscopy (FCS), and explored the effects of packing on the FPs with circular dichroism (CD) and fluorescence spectral analysis. The experimental data, and in particular the observed encapsulation efficiencies, are discussed in relation to what is known about the structural features of the FPs.  [29]); (B) Chromophore of SBFP2 made from Ser65, His66, and Gly67; (C) chromophore of mTurquoise2 made from Ser65, Trp66, and Gly67; (D) Chromophore of mEGFP made from Thr65, Tyr66, and Gly67 in the anionic form; (E) Chromophore of SYFP2 made from Gly65, Tyr66, and Gly67 in the anionic form with Tyr203 to extend the π-system; (F) Chromophore of mKO2 made from Cys65, Tyr66, and Gly67 in the anionic form; (G) Chromophore of TagRFP made from Met65, Tyr66, and Gly67 in the anionic form and in the trans-conformation and (H) Chromophore of mCherry made from Met65, Tyr66 and Gly67 in the anionic form and in the cis-conformation. Absorption maxima are indicated on the spectral bar and fluorescence colors are indicated as box colors.

Results
In this work, we purified seven FPs using either the intein/chitin-binding-domain system (for mTurquoise2, SYFP2-His, and mCherry) or metal affinity chromatography (for SBFP2, mEGFP, mKO2, and TagRFP, see Section 4.2). The influence of the His-tag was tested by encapsulation of mTurquoise2 and of mTurquoise2-His [30]. We did not observe any differences in encapsulation properties between the two proteins; therefore, it is presumed that His-tags have no effects on our experiments using other FPs. The purified FPs have distinctive spectral properties. Figure 2 shows  [29]); (B) Chromophore of SBFP2 made from Ser65, His66, and Gly67; (C) chromophore of mTurquoise2 made from Ser65, Trp66, and Gly67; (D) Chromophore of mEGFP made from Thr65, Tyr66, and Gly67 in the anionic form; (E) Chromophore of SYFP2 made from Gly65, Tyr66, and Gly67 in the anionic form with Tyr203 to extend the π-system; (F) Chromophore of mKO2 made from Cys65, Tyr66, and Gly67 in the anionic form; (G) Chromophore of TagRFP made from Met65, Tyr66, and Gly67 in the anionic form and in the trans-conformation and (H) Chromophore of mCherry made from Met65, Tyr66 and Gly67 in the anionic form and in the cis-conformation. Absorption maxima are indicated on the spectral bar and fluorescence colors are indicated as box colors.

Results
In this work, we purified seven FPs using either the intein/chitin-binding-domain system (for mTurquoise2, SYFP2-His, and mCherry) or metal affinity chromatography (for SBFP2, mEGFP, mKO2, and TagRFP, see Section 4.2). The influence of the His-tag was tested by encapsulation of mTurquoise2 and of mTurquoise2-His [30]. We did not observe any differences in encapsulation properties between the two proteins; therefore, it is presumed that His-tags have no effects on our experiments using other FPs. The purified FPs have distinctive spectral properties. Figure 2 shows our recorded normalized fluorescence excitation and emission spectra of the FPs, which display maxima in agreement with those listed in literature (Table S2) our recorded normalized fluorescence excitation and emission spectra of the FPs, which display maxima in agreement with those listed in literature (Table S2).

Fluorescent Protein Charge Determination
Measurements on encapsulated FPs in C3Ms are commonly performed at the preferred micellar composition (PMC), which is the ratio between protein and polymer at which the highest concentration of micelles is obtained [18,19,31]. The PMC is defined in terms of the total concentration of positively charged groups on the polymers and the net concentration of negative charges on the protein (see Equation (1), Section 4.5). The positive charge on the polymers is fixed due to quaternization, but the charge of the proteins varies with pH. The amino acid residues on the protein surface determine to a great extent the net charge of the protein, which can be deduced from the protein's three-dimensional structure using the PROPKA software package [32,33]. For four of the studied FPs (mTurquoise2, mEGFP, TagRFP, and mCherry), a crystal structure is available in the RCSB Protein Data Bank [34] (Table 3). For the three other FPs (SBFP2, SYFP2, and mKO2), a Blast search was performed to obtain the most suitable template, which was then used for building homology models (Table 4).
The pI value for the four avFPs and for two anFPs (mKO2 and mCherry) is about 5.5, while TagRFP has a significantly higher pI value, ~7.6 (Table S2). To achieve similar electrostatic interactions between the polymers and the different FPs, we encapsulated all FPs at the pH at which they have a net negative charge of about 10 unit charges. Thus, TagRFP was encapsulated at pH 10 and the other FPs at pH 9 (Table 1). At these conditions, all FPs are stable [35]. Table 1. Results of PROPKA 3.1 analysis and preferred micellar composition (PMC) determination. The charge of the proteins was determined at pH 9, except for TagRFP (pH 10), with PROPKA 3.1 [32,33]. PMC (F + ) and hydrodynamic radii (with standard deviations) were determined with dynamic light scattering for all used fluorescent protein variants encapsulated using the two diblock copolymers.

Fluorescent Protein Charge Determination
Measurements on encapsulated FPs in C3Ms are commonly performed at the preferred micellar composition (PMC), which is the ratio between protein and polymer at which the highest concentration of micelles is obtained [18,19,31]. The PMC is defined in terms of the total concentration of positively charged groups on the polymers and the net concentration of negative charges on the protein (see Equation (1), Section 4.5). The positive charge on the polymers is fixed due to quaternization, but the charge of the proteins varies with pH. The amino acid residues on the protein surface determine to a great extent the net charge of the protein, which can be deduced from the protein's three-dimensional structure using the PROPKA software package [32,33]. For four of the studied FPs (mTurquoise2, mEGFP, TagRFP, and mCherry), a crystal structure is available in the RCSB Protein Data Bank [34] ( Table 3). For the three other FPs (SBFP2, SYFP2, and mKO2), a Blast search was performed to obtain the most suitable template, which was then used for building homology models ( Table 4).
The pI value for the four avFPs and for two anFPs (mKO2 and mCherry) is about 5.5, while TagRFP has a significantly higher pI value,~7.6 (Table S2). To achieve similar electrostatic interactions between the polymers and the different FPs, we encapsulated all FPs at the pH at which they have a net negative charge of about 10 unit charges. Thus, TagRFP was encapsulated at pH 10 and the other FPs at pH 9 (Table 1). At these conditions, all FPs are stable [35]. Table 1. Results of PROPKA 3.1 analysis and preferred micellar composition (PMC) determination. The charge of the proteins was determined at pH 9, except for TagRFP (pH 10), with PROPKA 3.1 [32,33]. PMC (F + ) and hydrodynamic radii (with standard deviations) were determined with dynamic light scattering for all used fluorescent protein variants encapsulated using the two diblock copolymers.

Preferred Micellar Composition (PMC)
The seven FPs were encapsulated using two diblock copolymers with different lengths (P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 ). As a start, dynamic light scattering (DLS) experiments were performed to determine the PMCs. The results of SBFP2 with P2MVP 41 -b-PEO 205 and P2MVP 128 -b-PEO 477 are shown in Figure 3. The highest concentration of micelles is found at the maximum of the scattered light intensity. For SBFP2, PMCs are found at F + values of 0.75 and 0.70 for P2MVP 41 -b-PEO 205 and P2MVP 128 -b-PEO 477 , respectively ( Figure 3 and Table 1). Similar DLS experiments were performed on the other six FPs with both diblock copolymers ( Figure S2) and their respective PMCs are listed in Table 1. For all FPs and with both diblock copolymers, optimal F + values ranging between 0.60 and 0.75 were found. Samples with this optimal composition were used in all other spectroscopic analyses: fluorescence correlation spectroscopy (FCS), circular dichroism (CD), and steady-state fluorescence spectroscopy.

Preferred Micellar Composition (PMC)
The seven FPs were encapsulated using two diblock copolymers with different lengths (P2MVP41-b-PEO205 or P2MVP128-b-PEO477). As a start, dynamic light scattering (DLS) experiments were performed to determine the PMCs. The results of SBFP2 with P2MVP41-b-PEO205 and P2MVP128b-PEO477 are shown in Figure 3. The highest concentration of micelles is found at the maximum of the scattered light intensity. For SBFP2, PMCs are found at F + values of 0.75 and 0.70 for P2MVP41-b-PEO205 and P2MVP128-b-PEO477, respectively ( Figure 3 and Table 1). Similar DLS experiments were performed on the other six FPs with both diblock copolymers ( Figure S2) and their respective PMCs are listed in Table 1. For all FPs and with both diblock copolymers, optimal F + values ranging between 0.60 and 0.75 were found. Samples with this optimal composition were used in all other spectroscopic analyses: fluorescence correlation spectroscopy (FCS), circular dichroism (CD), and steady-state fluorescence spectroscopy. The fluctuations of the scattered light intensities were used to calculate the hydrodynamic radii of the C3Ms. For all seven FPs, the hydrodynamic radii of the C3Ms are quite constant over a relatively wide range of F + compositions (0.40 < F + < 0.80, Figures 3 and S2). In general, radii of the formed C3Ms vary between 30 and 38 nm, except for C3Ms formed with mKO2, which are somewhat smaller with radii of about 27 nm (Table 1).

Encapsulation Efficiency
Next to DLS, FCS can be used for the determination of PMC values [18]. An advantage of FCS is that it gives, amongst other parameters, the average number of fluorescent particles in the confocal volume (N, Equation (2)). In this study, the fluorescent particles observed are free FPs and C3Ms with multiple FPs encapsulated. We quantified the number of free FPs before addition of polymers (Nbefore) and of fluorescent particles after addition of polymers (Nafter), and expressed the encapsulation efficiency per FP according to the following relation: Eencap = 1 − (Nafter/Nbefore). The encapsulation efficiencies per FP are shown in Figure 4 and the corresponding graph with the number of fluorescent particles is shown in Figure S3. FCS was not performed on samples containing SBFP2 because no suitable excitation source for this FP was available on the used confocal microscope.
For all avFPs, the encapsulation efficiencies are almost 100% with both diblock copolymers, meaning that virtually all protein molecules are packed in C3Ms. However, we observed lower The fluctuations of the scattered light intensities were used to calculate the hydrodynamic radii of the C3Ms. For all seven FPs, the hydrodynamic radii of the C3Ms are quite constant over a relatively wide range of F + compositions (0.40 < F + < 0.80, Figure 3 and Figure S2). In general, radii of the formed C3Ms vary between 30 and 38 nm, except for C3Ms formed with mKO2, which are somewhat smaller with radii of about 27 nm (Table 1).

Encapsulation Efficiency
Next to DLS, FCS can be used for the determination of PMC values [18]. An advantage of FCS is that it gives, amongst other parameters, the average number of fluorescent particles in the confocal volume (N, Equation (2)). In this study, the fluorescent particles observed are free FPs and C3Ms with multiple FPs encapsulated. We quantified the number of free FPs before addition of polymers (N before ) and of fluorescent particles after addition of polymers (N after ), and expressed the encapsulation efficiency per FP according to the following relation: E encap = 1 − (N after /N before ). The encapsulation efficiencies per FP are shown in Figure 4 and the corresponding graph with the number of fluorescent particles is shown in Figure S3. FCS was not performed on samples containing SBFP2 because no suitable excitation source for this FP was available on the used confocal microscope.
For all avFPs, the encapsulation efficiencies are almost 100% with both diblock copolymers, meaning that virtually all protein molecules are packed in C3Ms. However, we observed lower encapsulation efficiencies for anFPs (50% to 75%, see Figure 4), which implicates that, for these FPs, more protein molecules remain free in solution ( Figure S3). encapsulation efficiencies for anFPs (50% to 75%, see Figure 4), which implicates that, for these FPs, more protein molecules remain free in solution ( Figure S3).  (2) and Figure S3).

Fluorescence Properties
Previously, we have shown that encapsulation of EGFP and mEGFP resulted in different spectral properties compared to that of the proteins free in solution [19]. The spectral properties of EGFP upon encapsulation do not changes more than that of mEGFP, which is due to the pKa shift of the chromophore of EGFP. To investigate if encapsulation changes the spectral properties of the FPs, absorption and fluorescence excitation and emission spectra for all FPs free in solution, as well as encapsulated in C3Ms were recorded ( Figures 5 and S4).
We observed that encapsulation of the FPs leads to minor differences in their absorption and fluorescence properties and these are dependent on the kind of FP and the type of polymer used. Figure 5H,I shows that, for SBFP2, both the absorption and the fluorescence intensity increases upon encapsulation. Encapsulation of mTurquoise2, mEGFP, and SYFP2 resulted in a decrease of the fluorescence intensity, whereas the absorption remained the same. Both the absorption and fluorescence intensity decreases upon encapsulation of TagRFP. For mCherry, the fluorescence intensity increases and the absorption and excitation maxima become blue-shifted upon encapsulation (for absorption spectra see Figure S4). The absorption and fluorescence results were combined in the determination of relative quantum yields of FPs encapsulated in C3Ms (Equation (4) and Table 2). Table 2 shows that the quantum yield of SBFP2 does not change; that of mCherry increases; and that of the other FPs decreases upon encapsulation.
To address if the observed spectral changes are due to a pH-related phenomenon, fluorescence excitation and emission spectra at different pH values were acquired of all FPs free in solution (see Figure S5). SBFP2, mEGFP, SYFP2, and mKO2 have a pKa of 5.5-6.0 and show a large decrease in their fluorescence intensity at pH 5. For the latter three proteins, this effect is caused by protonation of the phenolic oxygen of the chromophore ( Figures 1D-F and S5C-E). mCherry shows a stronger susceptibility to changes in pH ( Figure S5G): at increasing pH values (from pH 5 to 10), the spectra are blue-shifted and the fluorescence intensity increases. These changes resemble the changes observed upon encapsulation of mCherry.
The only two FPs showing no significant effect upon changes of pH are mTurquoise2 and TagRFP, which can be explained by their rather low pKa values (pKa ~3.5, see Figure S5F). It is therefore remarkable that the fluorescence intensity of TagRFP decreases about 40% upon encapsulation compared to the free protein ( Figure 5H), even though the encapsulation efficiency is about 60% (Figure 4). This suggests that the fluorescence of TagRFP is highly affected upon encapsulation. In solution, TagRFP tends to dimerize with a KD of 38.4 μM [36]. Assuming a protein  (2) and Figure S3).

Fluorescence Properties
Previously, we have shown that encapsulation of EGFP and mEGFP resulted in different spectral properties compared to that of the proteins free in solution [19]. The spectral properties of EGFP upon encapsulation do not changes more than that of mEGFP, which is due to the pK a shift of the chromophore of EGFP. To investigate if encapsulation changes the spectral properties of the FPs, absorption and fluorescence excitation and emission spectra for all FPs free in solution, as well as encapsulated in C3Ms were recorded ( Figure 5 and Figure S4).
We observed that encapsulation of the FPs leads to minor differences in their absorption and fluorescence properties and these are dependent on the kind of FP and the type of polymer used. Figure 5H,I shows that, for SBFP2, both the absorption and the fluorescence intensity increases upon encapsulation. Encapsulation of mTurquoise2, mEGFP, and SYFP2 resulted in a decrease of the fluorescence intensity, whereas the absorption remained the same. Both the absorption and fluorescence intensity decreases upon encapsulation of TagRFP. For mCherry, the fluorescence intensity increases and the absorption and excitation maxima become blue-shifted upon encapsulation (for absorption spectra see Figure S4). The absorption and fluorescence results were combined in the determination of relative quantum yields of FPs encapsulated in C3Ms (Equation (4) and Table 2). Table 2 shows that the quantum yield of SBFP2 does not change; that of mCherry increases; and that of the other FPs decreases upon encapsulation.
To address if the observed spectral changes are due to a pH-related phenomenon, fluorescence excitation and emission spectra at different pH values were acquired of all FPs free in solution (see Figure S5). SBFP2, mEGFP, SYFP2, and mKO2 have a pK a of 5.5-6.0 and show a large decrease in their fluorescence intensity at pH 5. For the latter three proteins, this effect is caused by protonation of the phenolic oxygen of the chromophore ( Figure 1D-F and Figure S5C-E). mCherry shows a stronger susceptibility to changes in pH ( Figure S5G): at increasing pH values (from pH 5 to 10), the spectra are blue-shifted and the fluorescence intensity increases. These changes resemble the changes observed upon encapsulation of mCherry.
The only two FPs showing no significant effect upon changes of pH are mTurquoise2 and TagRFP, which can be explained by their rather low pK a values (pK a~3 .5, see Figure S5F). It is therefore remarkable that the fluorescence intensity of TagRFP decreases about 40% upon encapsulation compared to the free protein ( Figure 5H), even though the encapsulation efficiency is about 60% (Figure 4). This suggests that the fluorescence of TagRFP is highly affected upon encapsulation.
In solution, TagRFP tends to dimerize with a K D of 38.4 µM [36]. Assuming a protein concentration of about 10 mM in the C3Ms, this implies that TagRFP associates into dimers or tetramers inside C3Ms, which might cause the drastic decrease of quantum yield of the chromophore upon encapsulation.
Next to these differences between the FPs, we also observed an effect depending on the length of diblock copolymer used: if the fluorescence increases upon encapsulation, the increase is larger with the longer polymer (P2MVP 128 -b-PEO 477 ) than with the shorter one (P2MVP 41 -b-PEO 205 ). Conversely, if the fluorescence decreases upon encapsulation, the decrease is larger with the shorter polymer than with the longer one, except for TagRFP ( Figure 5H). This dependency, however, is not observed in the absorption spectra ( Figure 5I). concentration of about 10 mM in the C3Ms, this implies that TagRFP associates into dimers or tetramers inside C3Ms, which might cause the drastic decrease of quantum yield of the chromophore upon encapsulation. Next to these differences between the FPs, we also observed an effect depending on the length of diblock copolymer used: if the fluorescence increases upon encapsulation, the increase is larger with the longer polymer (P2MVP128-b-PEO477) than with the shorter one (P2MVP41-b-PEO205). Conversely, if the fluorescence decreases upon encapsulation, the decrease is larger with the shorter polymer than with the longer one, except for TagRFP ( Figure 5H). This dependency, however, is not observed in the absorption spectra ( Figure 5I).  Figure S4.   Figure S4.

Secondary Structure
To investigate whether the differences in encapsulation efficiencies are due to structural perturbations of the FPs, far-UV circular dichroism (CD) experiments were performed. Figure 6 shows CD spectra of all seven FPs free in solution and encapsulated with P2MVP 41 -b-PEO 205 or with P2MVP 128 -b-PEO 477 . The CD spectra of the FPs are not affected by the increase in pH, as the CD spectra at pH 9.0 or 10.0 do not show any differences compared to those at pH 7.0 [37].
For all FPs, a negative mean residue ellipticity near 220 nm was observed, which is in good agreement with the prominent β-barrel architecture of these proteins ( Figure 1A) and in line with previous observations [19,38]. The spectrum of mKO2, however, resembles more a α-helical architecture with two negative peaks near 210 and 220 nm [39,40].
The CD spectra of the four avFPs free in solution are quite similar in shape ( Figure 6A). The CD spectra of the anFPs are remarkably different compared to those of the avFPs (SBFP2 was taken as a representative reference, see Figure 6B). To our knowledge, these differences have not been reported before, and are further addressed in Section 3.2.
Upon encapsulation of the FPs, the CD spectra alter to a greater or lesser extent compared to that of the proteins free in solution, especially in the range where the spectra switch ellipticity ("zero crossing", between 205 and 215 nm). For the encapsulated avFPs and for encapsulated mKO2, the zero crossing shifts to higher wavelength compared to that of the respective free FPs ( Figure 6C-G). On the other hand, the zero crossings of encapsulated TagRFP and mCherry change to lower wavelengths compared to that of the free proteins ( Figure 6H,I). In general, the zero crossings of all encapsulated FPs shift to ±210 nm. Apart from the zero crossings, the changes upon encapsulation of mTurquoise2, SYFP2, and mKO2 are moderate. More pronounced deviations in CD spectra after encapsulation are observed for SBFP2, mEGFP, and TagRFP. The largest change, however, can be observed for mCherry, with a significant positive decrease and a negative increase in ellipticity around 200 and 220 nm, respectively ( Figure 6I).

Discussion
Previously, we found that the encapsulation of EGFP in C3Ms stimulates protein dimerization and changes the spectral properties of the EGFP chromophore [19]. Because mEGFP mainly remains monomeric in the densely packed C3Ms, encapsulation of this protein hardly affects its spectral properties. In this work, we studied the encapsulation of four avFPs and three anFPs in C3Ms. All investigated FPs were successfully encapsulated using two diblock copolymers (P2MVP 41 -b-PEO 205 and P2MVP 128 -b-PEO 477 ) with F + values ranging between 0.60 and 0.80. For strong polyelectrolytes, stoichiometric C3M systems are formed at a F + value of 0.50 [41]. Proteins, however, are weak polyelectrolytes and therefore their charge may change upon interaction with the diblock copolymer. Moreover, coacervation between polymer and protein does not necessarily arise from the overall charge of the protein, but rather from specific charge patches on the protein surface [42]. Both effects can even lead to coacervation between similarly charged proteins and polyelectrolytes [43][44][45][46].

Encapsulation Efficiency
The encapsulation efficiencies of avFPs (mEGFP, SBFP2, SYFP2 and mTurquoise2) were almost 100%, whereas those of anFPs (mKO2, TagRFP and mCherry) varied between 50% and 75%. This implicates that the interactions between the anFPs and the diblock copolymers to form C3Ms are less favorable. The formation of C3Ms requires an interaction between the FPs and the polymers, which can be dependent on the surface charge distribution and/or the shape of the protein. For the investigation of the presence of specific charge patches on the protein surface, we determined the surface potential distribution of the FPs on the acquired protein structures. For this, homology modeling was used to obtain the protein structures of SBFP2, SYFP2, and mKO2, next to the crystal structures of mTurquoise2, mEGFP, TagRFP, and mCherry. In Figure 7, the surface potentials of the FPs are visualized at the pH value at which they were encapsulated. All avFPs share a negative surface patch, as displayed on the side view at 90 • , with an expansion to half of the molecule displayed in the side view at 180 • . The amino acid residues with negative charge belonging to this patch are located on β-strands 1 and 2. The three anFPs do not contain a similar negative patch displayed on the side view at 90 • , as observed for avFPs. Negative patches for mKO2 and TagRFP are mainly present in the side view at 0 • . For TagRFP, the amino acid residues with negative charge are more distributed over the entire protein surface than for the other proteins. For mCherry, there is not a side entirely filled with negatively charged amino acid residues. It is key for the positively charged polyelectrolyte to bind to a local negative charge patch on the protein while minimizing the repulsive effect arising from the positively charged amino acid residues. Therefore, the interactions between the diblock copolymers and mKO2, TagRFP, and mCherry might not be optimal, thus affecting their encapsulation efficiencies. displayed in the side view at 180°. The amino acid residues with negative charge belonging to this patch are located on β-strands 1 and 2. The three anFPs do not contain a similar negative patch displayed on the side view at 90°, as observed for avFPs. Negative patches for mKO2 and TagRFP are mainly present in the side view at 0°. For TagRFP, the amino acid residues with negative charge are more distributed over the entire protein surface than for the other proteins. For mCherry, there is not a side entirely filled with negatively charged amino acid residues. It is key for the positively charged polyelectrolyte to bind to a local negative charge patch on the protein while minimizing the repulsive effect arising from the positively charged amino acid residues. Therefore, the interactions between the diblock copolymers and mKO2, TagRFP, and mCherry might not be optimal, thus affecting their encapsulation efficiencies.

Elliptical Symmetry of FP Barrels
During this study, we uncovered clear differences in the far-UV CD spectra between avFPs and anFPs free in solution ( Figure 6B). It is well known that all FPs share a similar 11-stranded β-barrel fold ( Figure 1A). However, it is hardly reported in the literature that the elliptical symmetry between avFPs and anFPs is diverse [47]. Figure 8 shows the ribbon structures of the studied FPs in three different orientations: the broad side, the narrow side, and the top. From the top views, it is clear that the barrels of the FPs are not completely round, but form elliptic cylinders. The avFPs are rounder than the anFPs, which is depicted by the difference in aspect ratio:~0.85 for the avFPs and~0.74 for the anFPs. Especially mKO2 is the most "squeezed" of the anFPs. We hypothesize that these differences are the cause for the observed differences in the far-UV CD spectra. Micsonai et al. reported that CD spectra are influenced, among others, by degree of twist and distortion of the β-sheets [48]. The variance in the elliptical symmetry is another apparent difference between avFPs and anFPs, and could also be influencing their encapsulation efficiencies.

Biophysical Properties of Encapsulated Proteins
Encapsulation of the avFPs hardly influenced their secondary structural properties and only minor changes in absorption and emission characteristics were observed. All avFPs bear the A206K mutation, which favors their monomeric state. This supports that the minor spectral changes observed are caused by the electrostatic interactions between the polymers and the protein surfaces of these FPs.
All anFPs are found as tetramers in their hosts [9,10,26]. The anFPs used here are all modified to enhance their tendency to remain monomeric. In literature, this tendency is expressed in terms of dissociation constants and "monomeric quality" (see Table S2). Previously, we calculated the number of EGFP molecules present in a C3M to be around 400, yielding a local protein concentration of about 10 mM [18,19]. Since the FPs used in this study form C3Ms with PMCs (~0.65) and radii (~34 nm) similar to EGFP-C3Ms, it is a reasonable assumption that the protein concentration in the various FP-C3Ms is about the same. Hence, we expect that mCherry with a monomeric quality of 95% remains mostly monomeric upon encapsulation. However, mKO2 and TagRFP with monomeric qualities of 68% and 58%, respectively, and a dissociation constant of 0.038 mM for TagRFP, will likely form oligomers in the C3Ms (Table S2). This oligomerization causes the large decrease in quantum yield of the encapsulated forms of mKO2 and TagRFP (Table 2).
For encapsulated mCherry, the absorption spectrum changes according to a pK a shift of its chromophore ( Figure S4). For EGFP it was proposed that the pK a shift of its chromophore is caused by a reorientation of Glu222 due to the dimerization of EGPF in C3Ms [19]. For free mCherry, the equivalent Glu215 is also linked to the pH-dependent spectral shifts ( Figure S5G) [27]. If mCherry, however, remains monomeric in the C3Ms, the reorientation of Glu215 can only occur due to the interaction between protein and polymer.

Future Research
We show that encapsulation of structurally similar FPs in C3Ms is dependent on the origin of the FPs and can give rise to different encapsulation efficiencies. Moreover, the spectral and structural perturbations observed are dependent on the kind of FP and the type of polymer used. In future research, we plan to investigate the stability and dynamics of encapsulated FPs. This can be accomplished by mixing two appropriate FPs using FRET as a readout. Some requirements should be considered choosing an optimal FP FRET-pair: First, use fluorescent proteins with similar encapsulation efficiencies. Second, use FPs that show minor changes in their absorption and fluorescence properties upon encapsulation into the C3Ms. Third, use the diblock copolymer which has the least effect on the fluorescence properties of the FPs. According to the present results, the ideal partners of an FRET-pair in C3Ms would be mTurquoise2 and SYFP2 in combination with P2MVP 128 -b-PEO 477 .  [49,50] in MacPyMOL 1.4 [51].   [49,50] in MacPyMOL 1.4 [51].  [49,50] in MacPyMOL 1.4 [51].
Protein concentrations were determined with a Pierce BCA protein assay (Pierce Biotechnology, Rockford, IL, USA), using a bovine serum albumin standard as a reference. The purity of the FPs was checked by SDS-PAGE.

Modeling
Homology models were built from existing crystal structures using SWISS-MODEL [55][56][57][58]. Table 3 shows the proteins used in this paper and their corresponding PDBs. Table 4 shows the proteins used in this paper and their respective templates used for the homology modeling. The chromophores were placed in the model structure at the same position and orientation as the chromophore in the template structure. Pairwise sequence alignments of the FPs are listed in Figures S7-S13. Because some N-and C-termini were missing in the created homology models (for SBFP2, mEGFP, and SYFP2), these termini were modeled manually using the PDB entry 3ZTF as a template. The A206K mutants were created by mutagenesis of Ala206 into Lys206 in PDB entries 4EUL and 3ZTF to construct mEGFP and mTurquoise2, respectively.

C3M Preparation
Encapsulation of FPs with polymers was achieved by first diluting the FP stock solution in 10 mM borate buffer at pH 9.0 for SBFP2, mTurquoise2, mEGFP, SYFP2, mKO2, and mCherry; and at pH 10.0 for TagRFP to the desired concentration, followed by addition of the polymer. After mixing, samples were stored at room temperature for 24 h before measuring. All experiments were performed in 10 mM borate buffer at the encapsulation pH.

Dynamic Light Scattering (DLS)
DLS measurements were performed on an ALV instrument equipped with a 300 mW Cobolt Samba-300 DPSS laser operating at 660 nm and 100 mW, and static and dynamic enhancer fiber optics for an ALV/High QE APD (high quantum efficiency avalanche photo diode) single photon detector connected to an ALV5000/60X0 External Correlator (ALV-Laser Vertriebsgesellschaft m-b.H., Langen, Germany). The detection angle θ was set at 90 • and all measurement were performed at room temperature.
DLS measures fluctuations in scattered light intensities caused by the diffusion of particles. The diffusion time of particles is dependent on their size: proteins diffuse faster than the encapsulated proteins in C3Ms. Furthermore, larger particles scatter more light, because the scattered light intensity is proportional to R 6 , where R is the particle radius. The formation of more C3Ms leads to higher light intensities, which results in a maximum in the scattered light intensity versus composition plot (I vs. F + ). The composition at the maximum in scattered light intensity is denoted as the preferred micellar composition (PMC). For determination of the PMC, 500 µL solutions with different polymer/protein compositions were prepared. The protein concentration was kept constant at 1 µM for each composition. The amount of P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 was varied to obtain the desired values of F + : where [n + ] = c + N + refers to the total concentration of positively charged groups on the polymer and [n − ] = c − N − is the total concentration of negatively charged groups on the protein molecules. The number of charged groups on the diblock copolymer (N + ) taking the degree of quaternization into account, is +33.1 for P2MVP 41 -b-PEO 205 and +112.0 for P2MVP 128 -b-PEO 477 , which is used to calculate [n + ]. The net charge of the proteins as a function of pH was calculated using the software package PROPKA 3.1 [32,33]. The charges of the native proteins at pH 9 or 10 (N − ) are listed in Table 1, which are used to calculate [n − ].

DLS Data Analysis
DLS autocorrelation curves were generated from 10 intensity traces and averaged. The CUMULANT method [63,64] was used to analyze the mean apparent hydrodynamic radius (R h ) as: where q is the scattering vector, k is the Boltzmann constant, T is the absolute temperature, n is the viscosity of the solvent, and Γ is the measured average decay rate of the correlation function. The CONTIN method [65,66] is used to analyze the distribution of the radii of the C3Ms. The data were analyzed with the AfterALV program (AfterALV 1.0d, Dullware, The Netherlands).

Fluorescence Correlation Spectroscopy (FCS)
FCS was performed on a Leica TCS SP8 X SMD system equipped with a 63× 1.20 NA (numeric aperture) water immersion objective with coverslip thickness correction collar. Samples with FPs were excited using a diode laser (emits at 440 nm) or a super continuum laser (emits a continuous spectrum from 470 to 670 nm). The lasers were set at a pulsed frequency of 40 MHz. The size-adjustable pinhole was set at 70 µm for all measurements. Fluorescence emission was detected using bandpass-adjustable spectral filters. In Table 5 the used laser lines and range of the spectral filters are given per fluorescent protein. Fluorescence was recorded via the internal hybrid detector, which was coupled to a PicoHarp 300 TCSPC module (PicoQuant, Berlin, Germany). With this system, it was not possible to measure SBFP2, because its excitation maximum is below 440 nm. Rhodamine 110 (D = 4.3 × 10 −10 m 2 s −1 ) was used to calibrate the confocal volume of the setup. A diffusion time of 18 µs and a structural parameter (a, expressed as (ω xy /ω z )) between 5 and 10 were obtained, resulting in a confocal volume of approximately 0.2 fL. Measurements were performed in a µ-Slide 8-wells chambered coverslip (Ibidi ® ).
Samples with concentrations of 1 µM FP were measured free in buffered solution as well as encapsulated with P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 at their respective PMCs. For each sample, 5 fluorescence intensity fluctuation traces of 30 s each were collected. All measurements were performed at room temperature.

FCS Data Analysis
For the FCS data analysis, the FFS-data processor version 2.3 (Scientific Software Technologies Software Centre, Minsk, Belarus) was used [67]. The equation used to fit translational data, which includes triplet state, is as follows [68]: where N represents the average number of fluorescent particles in the confocal volume. The exponential term describes the triplet state behavior of the molecule, in which F trip is the fraction of molecules in the triplet state and T trip is the average time a molecule resides in the triplet state. The last part of the equation describes the diffusion behavior of the molecules, where F i is the fraction of species i, τ diff,i is the diffusion time of species i, ω xy and ω z are the equatorial and axial radii of the detection volume, respectively. Equation (3) was used to obtain N for the different samples.

Steady-State Fluorescence Spectroscopy
Fluorescence excitation and emission spectra were measured using a Cary Eclipse spectrofluorimeter (Varian). Excitation and emission slits were set to yield bandwidths of 5 nm.
All measurements were performed at 20 • C. Samples with concentrations of 1 µM FP were measured free in buffered solution as well as encapsulated with P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 at their respective PMCs.
The relative quantum yields are calculated using the following equation [69]: where QY represents the quantum yield, FA the integrated area under the corrected emission spectrum, and A the absorbance at the excitation wavelength. The subscripts C3M and P refer to the proteins in the C3Ms and the proteins free in solution, respectively.

Circular Dichroism (CD)
CD experiments were performed on a JASCO J-715 spectropolarimeter with a Jasco PTC 348 WI temperature controller set at 20 • C. The far-UV CD spectra (195-260 nm) were obtained from samples in a 0.3 mL quartz cuvette with an optical path length of 1 mm. Thirty spectra, each recorded with a resolution of 1 nm, a scan speed of 50 nm/min and a response time of 1 s, were accumulated and averaged. Samples with concentrations of 2.5 µM FP were measured free in buffered solution as well as encapsulated with P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 at their respective PMC. The polymers did not show any CD signal over the measured range, therefore, buffer blank spectra, obtained at identical conditions, were subtracted.

Conclusions
We have studied the encapsulation efficiency of SBFP2, mTurquoise2, mEGFP, SYFP2, mKO2, TagRFP, and mCherry and determined their spectral and structural properties as protein free in solution and upon encapsulation with P2MVP 41 -b-PEO 205 or P2MVP 128 -b-PEO 477 . This revealed that avFPs are almost 100% encapsulated, while anFPs show encapsulation efficiencies ranging between 50% and 75%. Upon encapsulation, all FPs show differences in spectral properties compared to their respective protein free in solution: the chromophores of SBFP2, mKO2, and mCherry are affected in their molar extinction coefficient and the chromophores of mTurquoise2, mEGFP, SYFP2, TagRFP, and mKO2 are affected in their fluorescence quantum yield. Only for mCherry, the changes in spectral properties upon encapsulation are similar to changes observed as a result of pH variation and are, therefore, related to a shift in the pK a . Even though all FPs have an 11-stranded β-barrel fold, the CD spectra differ between avFPs and anFPs. This is most likely due to a different shape of the cylinders between the two groups of FPs, where the β-barrel structures of avFPs are almost round cylinders and that of anFPs elliptic ones. This variation in structure, together with the difference in charge distribution on FP surfaces, potentially causes the differences in encapsulation efficiency.