The Effect of Protein Electrostatic Interactions on ... - ACS Publications

Aug 2, 2016 - ABSTRACT: Mutation of a superfolder green fluorescent protein (GFP) was used to design GFP variants with formal net charges of 0, −8, ...
1 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/Biomac

The Effect of Protein Electrostatic Interactions on Globular Protein− Polymer Block Copolymer Self-Assembly Christopher N. Lam, Helen Yao, and Bradley D. Olsen* Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States S Supporting Information *

ABSTRACT: Mutation of a superfolder green fluorescent protein (GFP) was used to design GFP variants with formal net charges of 0, −8, and −21, providing a set of three proteins in which the total charge is varied to tune protein−protein interactions while controlling for the protein size and tertiary structure. After conjugating poly(N-isopropylacrylamide) (PNIPAM) to each of these three GFP variants, the concentrated solution phase behavior of these three block copolymers is studied using a combination of small-angle X-ray scattering (SAXS), depolarized light scattering (DPLS), and turbidimetry to characterize their morphologies. The electrostatic repulsion between supercharged GFP suppresses ordering, increasing the order−disorder transition concentration (CODT) and decreasing the quality of the ordered nanostructures as measured by the full width at half-maximum of the primary scattering peak. By contrast, the charge distribution of the neutrally charged GFP results in its largest dipole moment, calculated about the protein’s center of mass, among the three GFP variants and a self-complementary Janus-like electrostatic surface potential that enhances nanostructure formation. The different electrostatic properties result in different protein−protein interactions that affect the high temperature morphologies, including the formation of macrophase separated or homogeneous micellar phases and the smaller hexagonal ordering window of the supercharged GFP. Small improvements in the quality of the ordered nanostructures of GFP(−21)-PNIPAM can be achieved through protein−divalent cation interactions. Therefore, varying protein charge and electrostatics is demonstrated as a method of tuning the magnitude and directionality of protein−protein interactions to control self-assembly.

■

INTRODUCTION The biodegradability and renewability of enzymes, their ability to function under mild reaction conditions, and their high degree of specificity toward substrates make them very desirable materials to address a diversity of challenges in energy,1,2 medicine,3−6 environmental sustainability,7 and national security.8−10 Enzymes have enabled the advancement of sensor technologies, with glucose oxidase used to pioneer the development of biosensors for biomedical applications3 and other sensing technologies developed for environmental monitoring11 and neurotoxin detection and decontamination.10 Protein−polymer conjugates, in particular, PEGylated (modified by poly(ethylene glycol) (PEG)) proteins, have garnered much attention as therapeutic agents and drug delivery platforms.4−6,12−14 Many design criteria need to be considered when designing enzyme-containing materials, including engineering the material environment to protect the enzyme structure and function15−17 and accessibility of the active site through control over enzyme position and orientation to allow facile transport of substrate, product, and cofactors.18,19 Direct self-assembly of proteins as blocks in block copolymers provides an elegant solution to achieving the necessary nanostructural control within a multicomponent material.20,21 Inspiration for direct self-assembly as an approach © XXXX American Chemical Society

for patterning enzymes is drawn from the rich phase behavior of coil−coil block copolymers.22,23 These self-assembly principles have been applied to form nanostructured materials from globular protein−polymer block copolymers, demonstrating the ability to make highly active biocatalytic materials.24 In addition to a number of studies of self-assembly in dilute solution25−30 and in thin films,24 self-assembly principles have been applied to form solid-state nanostructured plastics and gels from globular protein−polymer block copolymers, forming many nanostructured phases observed in coil−coil block copolymerslamellae, perforated lamellae, hexagonal cylinders, and disordered micelles.31−35 Dissimilarities in the phase behavior of coil−coil block copolymers and globular protein−polymer block copolymers reflect differences in the underlying physics governing their selfassembly. Self-consistent mean field theory predicts that the Flory−Huggins interaction parameter χ, the degree of polymerization N, and the volume fraction fA, representing the fraction of block A, govern the phase behavior. In concentrated block copolymer solutions, according to the “dilution approximation” Received: April 26, 2016 Revised: June 23, 2016

A

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules Scheme 1. Electrostatic Properties of GFP Variants with Different Net Chargea

a

(a) Cartoon of the superfolder GFP crystal structure (PDB 2B3P);49 (b) Dipole moments calculated using the Protein Dipole Moments Server50 and illustrated for GFP(0) (left), GFP(−8) (middle), and GFP(−21) (right), which have magnitudes of 1.6 × 10−27 (C·m), 5.3 × 10−28 (C·m), and D = ∑⇀ ri qi about the proteins’ centers of mass, and the lengths of the 5.5 × 10−28 (C·m), respectively. The dipole moments are calculated as ⇀ arrows are drawn to scale in atomic units. (c) Different views of the electrostatic surface potential (±10 kT/ec) at the solvent-accessible surface of GFP(0), GFP(−8) (PDB 2B3P), and GFP(−21) rendered from solutions of the linearized Poisson-Boltzmann equation using the adaptive PoissonBoltzmann solver (APBS).51−53 Red and blue represent negative and positive values, respectively.

global net charge and protein surface charge distribution, electrostatic complementarity can assist in the formation of protein assemblies.47,48 Herein, the effect of protein electrostatic interactions is investigated by comparing the phase behavior of globular protein−polymer conjugates containing structurally homologous proteins using a series of green fluorescent protein (GFP) variants with different values of formal net charge. Using the superfolder GFP49 as the reference globular protein (Scheme 1), a series of mutants is prepared with variable charge, electrostatic surface potential, and electric dipole moment. While conjugates of all three GFP variants to PNIPAM are able to form long-range ordered nanostructures, the electrostatic repulsion of the supercharged protein is shown to affect the phase behavior significantly, suppressing nanostructure formation and reducing the quality of ordering in the selfassembled nanostructures. Enhanced nanostructure formation is observed due to the net neutral charge distribution and selfcomplementary Janus-like electrostatic surface potential of the net neutral GFP variant.

a neutral solvent will uniformly screen unfavorable A−B interactions, leading to the term ϕχN governing the phase behavior, where ϕ represents the block copolymer volume fraction in solution.36,37 In contrast to a random coil polymer, the diverse shape, size, and composition of enzymes result in a complex set of anisotropic hydrophobic, ionic, and hydrogen bonding interactions. Globular protein−polymer block copolymers have been shown to undergo lyotropic and thermotropic order−disorder transitions (ODTs) and order−order transitions (OOTs) in a rich and highly asymmetric phase space that is significantly different from that of coil−coil block copolymers33 due to the interaction potentials of proteins38,39 and the difference in shape between proteins and coil polymers. Detailed studies have elucidated many of the relevant effects governing the self-assembly process in the complex protein− polymer hybrid systems. Changes in polymer chemistry40 have led to significant changes in phase transition lines and ordered morphologies, yielding for the first time a cubic phase consistent with a bicontinuous gyroid structure; change in polymer topology has also been observed to have a significant effect on phase behavior and micellar stability.41 In contrast, comparative studies between mCherry and EGFP, two proteins with nearly identical tertiary structure but very different surface potentials, demonstrated very little change in their phase behavior. This suggests that coarse-grained properties such as the second-virial coefficient, protein size, and colloidal shape of the protein are critical to understanding protein−polymer block copolymer phase behavior. However, among these coarsegrained parameters, the role of protein electrostatic interactions has not been explored. Electrostatic interactions are important to self-assembly in many protein-based systems.42 Interactions between a charged surface and a protein with the same net charge can result from charge heterogeneity and oppositely charged protein surface patches,43 and overall protein charge is important for the formation of protein dimers with complementary charge.44 Supramolecular nanostructures can also form through electrostatic interactions.45,46 In particular, polystyrene-b-poly(acrylic acid) (PS-b-PAA) was observed to form polymerosomes with positively charged proteins, whereas negatively charged proteins led to electrostatic repulsions that only formed micelles. A neutrally charged myoglobin led to approximately a 1:1 ratio of polymerosomes to micelles.45 In addition to

■

MATERIALS AND METHODS Synthesis. Poly(N-isopropylacrylamide) (PNIPAM) was synthesized by reversible addition−fragmentation chain transfer (RAFT) polymerization as previously described.35 The molar mass and dispersity were analyzed by gel permeation chromatography using an Agilent Technologies 1260 Infinity system equipped with two ResiPore, 7.5 × 300 mm columns (Agilent Technologies), each with a molecular weight range up to 500 000 g/mol, in N,N-dimethylformamide (DMF) with 0.02 M lithium bromide (LiBr) as the mobile phase (Supporting Information, Figure S1). 1H NMR shows that the synthesis procedure yields a maleimide-functionalized PNIPAM (Figure S2). The superfolder green fluorescent protein (GFP) (overall formal net charge of −8) and two variants with overall formal net charges of −21 and 0 (DNA and amino acid sequences shown in Supporting Information, Figures S3−S5), each containing an N-terminal 6xHis tag, were expressed in the Escherichia coli strain Tuner(DE3). Henceforth, the GFP variants will be referred to by their charge as GFP(0), GFP(−8), and GFP(−21). GFP(0) and GFP(−8) were expressed by inoculating 1 L of LB medium with 5 mL of B

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules

Scheme 2. General Schematic Showing Conjugation of Maleimide-Functionalized PNIPAM to the N-Terminal Cysteine in the GFP Variants

overnight culture and incubating at 37 °C until OD600 ≈ 0.8− 1.0, after which the cultures were induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and harvested after growing for 9−10 h at 25 °C. The supercharged variant, GFP(−21), was expressed for 11−12 h at 37 °C after inoculating and then harvested. Cell pellets were resuspended in lysis buffer, incubated with 1 mg/mL lysozyme at 4 °C for 30 min, and sonicated. After clarifying the lysate by centrifugation, the proteins were purified by Ni-NTA metal affinity chromatography (Supporting Information, Figure S6). GFP(−8) and GFP(−21) were dialyzed to 20 mM Tris-Cl, pH = 8.0, and GFP(0) was dialyzed to 20 mM Tris-Cl, 250 mM NaCl, pH = 8.0 due to its low solubility in low ionic strength buffers. Bioconjugation. The coupling reactions between the GFP variants and maleimide end-functionalized PNIPAM (Scheme 2) were performed in 20 mM Tris-Cl, pH = 8.0 for GFP(−8) and GFP(−21) and in 20 mM Tris-Cl, 250 mM NaCl, pH = 8.0 for GFP(0). In addition to the N-terminal cysteine residue, there are two additional cysteine residues in the GFP variants: one near the edge of the β-barrel, and another buried inside near the chromophore. These three cysteines are also located in similar positions (in terms of amino acid sequence and protein crystal structure) in enhanced green fluorescent protein (EGFP), and previous studies established that conjugation reactions of PNIPAM to EGFP showed selective bioconjugation to the N-terminal cysteine.34 Protein solutions at approximately 1−2 mg/mL were first incubated with a 10x molar excess of tris(2-carboxyethyl)phosphine hydrochloride (TCEP·HCl) at 4 °C for approximately 45−60 min. A 6x molar excess of maleimide-functionalized polymer was then added to the solution and allowed to dissolve and react at 4 °C overnight for up to 20−24 h. After completion, the conjugation reaction was precipitated three times in 1 M ammonium sulfate solution at pH = 8.0 to remove unreacted protein. Unreacted polymer was subsequently removed by Ni-NTA metal affinity chromatography. Purified bioconjugate was then dialyzed against Milli-Q water. Purity was confirmed by both SDSPAGE and Native PAGE (Supporting Information, Figure S7). Two sets of bioconjugates with different PNIPAM coil fractions (Mn = 21.6 kDa, ϕPNIPAM = 0.52, and Mn = 30.7 kDa, ϕPNIPAM = 0.60) (Supporting Information) were synthesized using GFP(0), GFP(−8), and GFP(−21), as detailed in Table 1. These two coil fractions were chosen to be similar to a nearsymmetric block copolymer to target coil fractions that have exhibited lamellar-hexagonal OOTs and lamellar morphologies in previously studied mCherry-PNIPAM and EGFP-PNIPAM block copolymers to see if supercharging has a significant effect on phase behavior.33,34 The bioconjugates were also characterized by MALDI (Figure S8) at the Biopolymers and Proteomics Laboratory at MIT using a Bruker model MicroFlex MALDI-TOF instrument. Samples were dissolved in 20 μL of hexafluoroisopropanol, mixed in a 1:1 (v/v) ratio with the sinapinic acid matrix solution, and spotted prior to analysis.

Table 1. Composition of GFP(0)-, GFP(−8)-, and GFP(−21)-PNIPAM Block Copolymers molecule

PNIPAM Mn (kg/mol)

ĐPNIPAM

bioconjugate Mn (kg/mol)

ϕPNIPAM

GFP(0)PN21 GFP(−8)PN21 GFP(−21)PN21 GFP(0)PN30 GFP(−8)PN30 GFP(−21)PN30

21.6 21.6 21.6 30.7 30.7 30.7

1.06 1.06 1.06 1.09 1.09 1.09

50.1 50.0 50.1 59.2 59.1 59.2

0.52 0.52 0.52 0.60 0.60 0.60

UV−vis spectrophotometry and circular dichroism (CD) spectroscopy show retention of protein optical activity and secondary structure after conjugation and after self-assembly and rehydration (Supporting Information, Figures S9−S12). Sample Preparation and Characterization. Concentrated solution samples were prepared by rehydrating the solid globular protein−polymer block copolymer to the desired concentration. The concentrated solution phase behavior of GFP-PNIPAM conjugates was characterized using a combination of small-angle X-ray scattering (SAXS) (Supporting Information, Figures S13 and S14), depolarized light scattering (DPLS), and turbidimetry (Supporting Information, Figures S15−S20), and differential scanning calorimetry (DSC) (Supporting Information, Figure S21). Further details of these characterization techniques are provided in the Supporting Information.

■

RESULTS AND DISCUSSION Electrostatic Effects on Self-Assembly of Structurally Homologous Proteins. Concentrated Solution Phase Behavior. A series of three proteins was synthesized with a range of formal charge: 0, −8, and −21, referred to as GFP(0), GFP(−8), and GFP(−21) (Scheme 1). The different total net charge and charge distribution of the three GFPs result in their different electric dipole moments. Because the dipole moment of a protein with a nonzero monopole will depend upon the placement of the protein within the coordinate system, the D = ∑⇀ ri qi with the dipole moments are calculated as ⇀ proteins centered at their center of mass, enabling a meaningful comparison of the dipole moments of the three GFP variants. The electric dipole moments are calculated using the Protein Dipole Moments Server.50 The magnitude of the electric dipole moment of GFP(0), 1.6 × 10−27 C·m, is approximately three times larger than the magnitudes of the electric dipole moments of GFP(−8) and GFP(−21), which are 5.3 × 10−28 C·m and 5.5 × 10−28 C·m, respectively. The dipole reflects the greater polarization of GFP(0), which results in a Janus-like electrostatic surface potential, promoting complementary electrostatic patchy directional interactions that can potentially improve selfassembly. Comparison of the self-assembly in concentrated solution of the different GFP variants shows common morphologies C

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules

Figure 1. Phase diagrams of (a) GFP(0)PN21, (b) GFP(−8)PN21, (c) GFP(−21)PN21, (d) GFP(0)PN30, (e) GFP(−8)PN30, (f) GFP(−21)PN30 as a function of temperature and concentration. Phases are identified as disordered (Dis), disordered micellar (DM), lamellar (Lam), and hexagonal (Hex). Open symbols represent regions where macrophase separation between a conjugate-rich ordered phase and a waterrich phase is observed. Semitransparent symbols demarcated by the dashed lines denote ordered morphologies that are nonbirefringent.

The slopes of the CODT transition lines are different between the sets of conjugates at the near-symmetric coil fraction, reflecting the differences in monopole-monopole interactions of the GFP variants. Within the resolution of the measured temperatures and concentrations, the slopes of the CODT lines for ϕPNIPAM = 0.52 (Figure 1a−c) are approximately 7.5 °C/wt % for GFP(0) (over the concentration range of 35−37 wt %), 3.0 °C/wt % for GFP(−8) (over the concentration range of 35−40 wt %), and 2.1 °C/wt % for GFP(−21) (over the concentration range of 33−40 wt %). The gradual tilting of the CODT transition line with increasing protein net charge shows that there is a greater propensity toward disordering due to the increasing monopole−monopole interactions with increasing protein charge. However, in these lower coil fraction conjugates, this is manifest most strongly at slightly elevated temperatures as the PNIPAM hydration starts to decrease, decreasing the unperturbed size of the coil block. At the larger coil fraction of ϕPNIPAM = 0.60, the CODT occurs at a higher packing density for GFP(−8) and GFP(−21) than for GFP(0), reflecting the stronger propensity toward self-assembly in the neutral net charge GFP. However, the transition lines are nearly vertical for all three of these higher coil fraction conjugates. Increasing temperature is observed to lead to a disordered micellar phase for all three GFP variants above the PNIPAM desolvation transition, and the micellar stability shows trends due to differences in net charge and polymer coil fraction. At the smaller coil fraction of ϕPNIPAM = 0.52, GFP(0)-PNIPAM forms homogeneous micellar phases. In contrast, GFP(−8) and GFP(−21) show a decrease in transmission below 90% of the initial transmission, the criterion chosen to determine the macrophase separation temperature, and only remain macrophase homogeneous at sufficiently high concentrations. For GFP(−8), at concentrations of 35−40 wt %, the transmission recovers to greater than 90% of the initial transmission (Table 2 shows the temperature range in which the two-phase region is

formed by all three variants but also reveals the impact that total net charge and electrostatic surface potential have on selfassembly. These similarities and differences are illustrated in the phase diagrams for ϕPNIPAM = 0.52 and ϕPNIPAM = 0.60 (Figure 1) in which the same molecular weight PNIPAM is conjugated to the proteins GFP(0), GFP(−8), and GFP(−21) to enable a direct comparison. Phase identifications are performed using a combination of SAXS to identify nanodomain symmetry, DPLS to measure the optical anisotropy of nanostructures, turbidimetry to identify macrophase separation, and DSC to measure the PNIPAM thermal transition (Table 2). At low temperature where water is a good solvent for both protein and polymer, the different charge of the GFP variants leads to a significant shift in the disorder-to-order concentration (CODT) at the larger coil fraction of ϕPNIPAM = 0.60. All three GFPs show a lyotropic CODT from an isotropic disordered phase to a lamellar morphology, and the CODTs are similar at ϕPNIPAM = 0.52: 33 wt % for GFP(0) and GFP(−8) and 33 wt % for GFP(−21). The quality of the lamellar ordering is observed to improve with increasing concentration, becoming birefringent at 45 wt % for all three GFP variants at a PNIPAM coil fraction of 0.52. However, the birefringence of the supercharged GFP(−21) is more than an order of magnitude less than that of GFP(0) and GFP(−8) (Supporting Information, Figures S15−S17). At ϕPNIPAM = 0.60, a significant difference in the CODTs is observed; the CODT of GFP(0) is 33 wt %, while the CODT of GFP(−8) and GFP(−21) is significantly greater at 47 wt %. The lamellar nanostructures of all three GFP variants at ϕPNIPAM = 0.60 are birefringent at their respective CODTs. For GFP(−21), SAXS shows a single peak at 43 wt %, but at 25 and 30 °C, a reflection at 2q* is observed; it is likely that at low temperatures at 43 wt % that the 2q* peak may be suppressed due to symmetry between the protein and polymer nanodomains.54 D

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules

formation of a narrow two-phase region during the transition to the high temperature, macrophase homogeneous, disordered micellar nanostructure was previously observed in PNIPAM conjugates of the similar protein enhanced green fluorescent protein (EGFP).34 Similarly to GFP(−8), GFP(−21) shows a drop in transmission that recovers with further heating up to 40 °C, but the transmission does not recover to at least 90% of the initial transmission. Hence, the phase diagram for GFP(−21) in Figure 1c denotes the formation of a two-phase region for all concentrations less than 50 wt %. At the larger coil fraction of ϕPNIPAM = 0.60, the propensity toward macrophase separation is opposite to that at the smaller coil fraction. For GFP(0), in which solutions remained homogeneous at all concentrations studied at the smaller coil fraction, macrophase separation is now observed at all concentrations studied at the larger coil fraction. GFP(−8) and GFP(−21) display greater regions of micellar phases that remain homogeneous at the larger coil fraction. GFP(−8) remains homogeneous at all concentrations, and GFP(−21) forms a narrow two-phase region at intermediate concentrations of 30−40 wt %, the transmission recovering above 90% above approximately 36 °C, and shows no decrease in transmission at concentrations greater than 40 wt %. In contrast to the other two GFPs, GFP(−21) forms a homogeneous disordered phase at high concentration that persists even up to 40 °C at 43 and 47 wt %. The net protein charge is observed to affect micellar phase stability at high temperature, with the intermediate charge of GFP(−8) showing the greatest propensity toward formation of homogeneous micellar phases overall. When the GFP is overall net neutral, at the larger coil fraction, the PNIPAM desolvation driving force at high temperature is sufficient to result in macrophase separation at all concentrations. While GFP(−8) and GFP(−21) both form micellar phases at both coil fractions, the greater charge and monopole-monopole interactions in GFP(−21) lead to a slightly greater tendency toward micelle aggregation, resulting in GFP(−21) having a larger two-phase region than GFP(−8). The PNIPAM desolvation temperatures of the GFP conjugates are also measured by DSC, which shows that the desolvation temperature decreases with increasing concentration for conjugates of all three GFP variants (Supporting Information, Figure S21). This is consistent with the behavior of PNIPAM homopolymer55,56 and of previously studied mCherry-PNIPAM and EGFP-PNIPAM conjugates.34,35 The macrophase separation transition temperatures, by contrast, increase with increasing solution concentration. More water is accommodated through hydration of the protein domains with increasing conjugate concentration, which results in an increase in the macrophase separation transition temperature above the PNIPAM desolvation temperature needed to achieve a higher degree of PNIPAM collapse for macrophase separation to occur.35 Differences in the phase behavior among the GFP variants at high concentration and high temperature are also observed, where the deswelling of the PNIPAM domain results in an effective reduction in polymer coil fraction. At ϕPNIPAM = 0.52, an OOT from lamellar to hexagonal is observed at 45 and 50 wt % with increasing temperature for both GFP(0) and GFP(−8). GFP(−21) has a smaller region of hexagonal morphologies, with the transition occurring at 50 wt %. At the larger coil fraction of ϕPNIPAM = 0.60, GFP(0) has an OOT from lamellar to hexagonal at 40 and 50 wt % from 30 to 40 °C. GFP(−8)

Table 2. Thermal Transitions for GFP(0)-, GFP(−8)-, and GFP(−21)-PNIPAM Block Copolymers molecule GFP(0)PN21

GFP(−8)PN21

GFP(−21)PN21

GFP(0)PN30

GFP(−8)PN30

GFP(−21)PN30

conc. (wt %)

TDPLSa (°C)

Ttb (°C)

TDSC (°C)

25 27 30 33 35 37 40 45 50 25 27 30 33 35 37 40 45 50 25 27 30 33 35 37 40 45 50 20 25 27 30 33 37 40 50 27 30 33 37 40 43 47 50 25 30 33 40 43 47 50 55

○ ○ ○ ○ ○ ○ ○ 32.9 ● ○ ○ ○ ○ ○ ○ ○ 34.1 34.9 ○ ○ ○ ○ ○ ○ ○ ● ● ○ ○ ○ ○ ○ 32.8 33.0 35.8 ○ ○ ○ ○ ○ 30.4 ● ● ○ ○ ○ ○ ○ ● ● ●

32.2 32.2 32.2 31.5 31.6−39.3 32.7−39.6 32.6−38.5 33.1 33.2 32.9 32.9 33.0 33.0 33.3 33.4 32.8 32.9 33.2 32.7 32.8 32.8 33.0 35.8 32.6 33.2−36.6 33.2−36.3 -

31.3 30.9 30.4 29.4 29.6 28.8 28.7 26.9 26.2 31.8 31.7 31.4 29.9 28.4 28.0 27.3 27.6 26.5 29.8 29.8 29.3 29.1 28.7 28.8 28.3 27.2 25.4 31.7 31.3 31.3 30.8 30.2 29.8 29.5 28.9 30.4 30.1 29.1 27.7 25.7 27.0 26.5 26.3 31.7 31.1 30.3 29.7 28.7 27.3 26.4 25.8

a

A temperature listed for a transition in DPLS denotes that the sample is initially birefringent at 10 °C and loses birefringence at the listed temperature. ○ denotes samples that never display birefringent behavior within the studied temperature range. ● denotes samples that remain birefringent throughout the entire studied temperature range. bThe symbol “-” signifies that no thermal transition is observed.

observed), and no decrease in transmission is observed at 45 and 50 wt % (Supporting Information, Figure S16). The E

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules

Figure 2. (a) SAXS data for birefringent lamellar morphologies of GFP(0), GFP(−8), and GFP(−21) conjugates at ϕPNIPAM = 0.52. Fitting a threeparameter Lorentzian function to the primary peak shows that GFP(0) and GFP(−8) form more well-ordered nanostructures than the supercharged GFP(−21), denoted by smaller fwhm values of their primary peaks. (b) Evolution of the fwhm of the primary peak of GFP(0), GFP(−8), and GFP(−21) as a function of concentration at 10 °C normalized to the CODT. Results for ϕPNIPAM = 0.60 are shown in panels c and d.

similarly transitions from lamellar to hexagonal at 30 °C between 47 and 50 wt %. The GFP(−21) conjugate transitions from lamellar to a disordered phase above 35 °C at 43 and 47 wt %, but at higher concentrations of 50 and 55 wt %, only lamellar morphologies are observed throughout the entire temperature range. The superfolder GFP and EGFP share a high identity (96%) in their sequence alignment, and previously studied EGFPPNIPAM at a coil fraction of ϕPNIPAM = 0.55 shows good agreement with the phase behavior of GFP(−8)-PNIPAM in this work with small shifts in transition lines due to differences in coil fraction. Similar to GFP(−8)-PNIPAM, EGFP-PNIPAM undergoes a lyotropic transition from a disordered to a lamellar morphology, forms a narrow two-phase region, and forms homogeneous disordered micelles at high temperatures, and undergoes an OOT from lamellar to hexagonal at high concentration and temperature. However, the hexagonal window for EGFP-PNIPAM at ϕPNIPAM = 0.55 is smaller than those for GFP(−8) at the studied coil fractions of 0.52 and 0.60. Previous comparative studies of EGFP, mCherry, and mCherry variants with single amino acid substitutions demonstrated very similar phase behavior,34 suggesting that large changes in total charge are required to see a noticeable effect. Quality of Ordering. Despite undergoing a similar lyotropic transition from a disordered to a lamellar phase, the different supercharged GFP variants show significant differences in the quality of the lamellar morphology at both coil fractions studied. The quality of the ordering can be assessed by the full width at half-maximum (fwhm) of the primary scattering peak. Figures 2a and 2c show a comparison of the fwhm of the primary peaks for birefringent lamellar phases formed by

GFP(0), GFP(−8), and GFP(−21) at ϕPNIPAM = 0.52 and 0.60, respectively. The background of the SAXS data is fit using the functional form Aq−4 + Bq−2 + C and then subtracted from the SAXS data, after which a three-parameter Lorentzian function is ⎡ ⎤ γ2 used to fit the primary peak: f (q , q0 , γ , I0) = I0⎢ (q − q )2 + γ 2 ⎥. ⎣ ⎦ 0 Here, q0 refers to the position of the primary peak, I0 denotes the height of the peak, and 2γ gives the value of the fwhm. As Figure 2a shows, for ϕPNIPAM = 0.52, while the fwhm values of lamellar morphologies of GFP(0) and GFP(−8) at 45 wt % are similar (0.028 and 0.022 nm−1, respectively), the fwhm of GFP(−21) at 50 wt % is greater by approximately a factor of 2 (0.048 nm−1). Figure 2c shows similar results for ϕPNIPAM = 0.60, where the fwhm values of lamellar morphologies of GFP(0) at 37 wt % and GFP(−8) at 47 wt % are 0.018 and 0.024 nm−1, respectively, and the fwhm of GFP(−21) at 50 wt % is again greater by approximately a factor of 2, 0.043 nm−1. These results clearly indicate a preference for better ordered phases in the bioconjugates with the net neutral GFP with its charge distribution and complementary electrostatic surface potential. The weak tendency toward microphase separation and the decreased quality of ordering in the supercharged GFP(−21) conjugates is observed across the entire concentration range explored. Figure 2b,d shows the evolution of the fwhm of the primary peak as a function of concentration at 10 °C for both coil fractions. The concentrations plotted along the x-axis have been normalized to the CODT. Even at low concentrations within the disordered regime of the bioconjugates, the sharpness of the single peak of the solutions follows the trend GFP(0) > GFP(−8) > GFP(−21), indicating stronger concentration inhomogeneity in the samples with near neutral F

DOI: 10.1021/acs.biomac.6b00522 Biomacromolecules XXXX, XXX, XXX−XXX

Article

Biomacromolecules

Figure 3. Cartoon schematic representations of results from ClusPro 2.059,60 of the five largest clusters representing the most probable configurations for interactions between two GFP molecules as a function of total net charge. The orientation angle between proteins is displayed underneath each configuration and is calculated as the angle between the vectors in the directions of the principal axes aligned with the β-barrel of the two proteins, represented as the black arrows.

packing of the GFPs as a function of charge and how this packing might influence the propensity to self-assemble into nanostructured phases. The protein docking server ClusPro 2.0 by Vajda and co-workers59,60 uses PIPER, an FFT-based docking program that calculates the pairwise interaction potential E = Eattr + w1Erep + w2Eelec + w3Epair, where Eattr and Erep represent attractive and repulsive contributions to the van der Waals interaction energy, respectively (shape complementarity), Eelec represents the electrostatic energy, and Epair represents the desolvation energy. The coefficients w1, w2, and w3 represent the weights of the corresponding terms that are optimally set for different types of docking problems.59 Without a priori knowledge of which interactions are most important, the balanced model is used to enable calculations in which all of the interaction terms in the pairwise interaction potential are considered. By rotating and translating one protein relative to another fixed protein, approximately 109 configurations are sampled, from which the 1000 most energetically favorable configurations are selected and grouped using a root-mean-square deviation (RMSD)-based clustering algorithm.61 Among the 1000 most favorable configurations from the 109 configurations sampled returned by ClusPro, the scoring function using the pairwise potential cannot meaningfully distinguish between the 1000 configurations, and instead the cluster size is a more reliable measure of model prediction.59 Figure 3 shows cartoon representations of the top 5 clusters (groups of configurations that are similar to within a root-mean-squared deviation radius of 0.9 nm) of the most favorable configurations, ranked by size, for GFP(0), GFP(−8), and GFP(−21). The most favorable configurations show qualitatively that GFP(0) and GFP(−8) are better able to pack laterally in a lamellar morphology than the supercharged GFP(−21) (Figure 3). The principal axis that is aligned with the β-barrel of the GFP variants (defined to point toward the N-terminus and represented as black arrows in Figure 3) is used to quantify the relative orientation between the two proteins in each interacting configuration. While most of the configurations have orientation angles that do not correspond to nearly parallel alignment of the beta barrels, most of the configurations for GFP(0) and GFP(−8) result from interactions along the side of the β-barrel that could facilitate lateral packing. Recent studies in which the protein colloidal shape in protein−polymer

charge. While Figure 2a,c indeed show better ordering of lamellar nanodomains formed by GFP(0) and GFP(−8) than that of GFP(−21) at high concentrations within the wellordered regime, the eventual convergence of the fwhm of the primary peaks of GFP(−21) conjugates to those of GFP(0) and GFP(−8) suggests that at sufficiently high concentration, molecular crowding leads to contributions from short-range excluded volume and van der Waals interactions and also electrostatic interactions that govern self-assembly.57 A similar decrease in protein−protein electrostatic repulsion owing to increasing molecular crowding effects has been observed with only a moderate increase in concentration beyond the dilute regime (