De novo synthetic biliprotein design, assembly and excitation energy transfer

Abstract

Bilins are linear tetrapyrrole chromophores with a wide range of visible and near-visible light absorption and emission properties. These properties are tuned upon binding to natural proteins and exploited in photosynthetic light-harvesting and non-photosynthetic light-sensitive signalling. These pigmented proteins are now being manipulated to develop fluorescent experimental tools. To engineer the optical properties of bound bilins for specific applications more flexibly, we have used first principles of protein folding to design novel, stable and highly adaptable bilin-binding four-α-helix bundle protein frames, called maquettes, and explored the minimal requirements underlying covalent bilin ligation and conformational restriction responsible for the strong and variable absorption, fluorescence and excitation energy transfer of these proteins. Biliverdin, phycocyanobilin and phycoerythrobilin bind covalently to maquette Cys in vitro. A blue-shifted tripyrrole formed from maquette-bound phycocyanobilin displays a quantum yield of 26%. Although unrelated in fold and sequence to natural phycobiliproteins, bilin lyases nevertheless interact with maquettes during co-expression in Escherichia coli to improve the efficiency of bilin binding and influence bilin structure. Bilins bind in vitro and in vivo to Cys residues placed in loops, towards the amino end or in the middle of helices but bind poorly at the carboxyl end of helices. Bilin-binding efficiency and fluorescence yield are improved by Arg and Asp residues adjacent to the ligating Cys on the same helix and by His residues on adjacent helices.

1. Introduction

Phycobiliproteins (PBPs) are a superfamily of proteins that bind linear tetrapyrrole chromophores, i.e. bilins, and that assemble into supramolecular complexes known as phycobilisomes in cyanobacteria and red algae [1,2]. PBPs have intense visible absorption and exhibit high fluorescence quantum yields, consistent with their roles as light-harvesting proteins. PBPs have been extensively studied for decades, and details concerning their biogenesis, structure, assembly and function are well established (e.g. [1–8]). A recent cryoelectron microscopic study of the approximately 17 MDa phycobilisomes of the red alga Griffithsia pacifica establishes the organization of approximately 860 protein subunits and approximately 2048 individual bilin chromophores in each light-harvesting complex at a resolution of approximately 3.5 Å [9]. In spite of the very large numbers of chromophores (approx. 250 to greater than 2000) found in phycobilisomes, energy transfer efficiencies of approximately 95% have typically been observed. PBPs generally absorb visible light, but recently PBPs absorbing light in the far-red wavelength range between 700 and 750 nm have been identified [10–12]. Each subunit of these brilliantly coloured proteins carries one or more linear tetrapyrrole (i.e. bilin) chromophores that are usually bound to the protein through a single cysteinyl thioether linkage to a vinyl side chain on the bilin chromophore [1]. Over the past 20 years, the bilin lyases responsible for the specific attachment of bilin chromophores to PBP apoproteins have been identified [8,13]. Functionally, it is self-evident that PBPs are remarkably versatile proteins that are exquisitely evolved to perform a specific task, harvesting light in the wavelength range 400–750 nm for oxygenic photosynthesis [2].

A second family of bilin-containing proteins are the phytochromes and cyanobacteriochromes that are associated with light perception in cyanobacteria, algae and plants [14–17]. In these proteins, the bilin chromophore of a GAF domain may be singly or doubly bound to cysteine residues to modify the number of conjugated double bonds and thus the light-wavelength response of the chromophore. Light-induced isomerization of the chromophore can trigger conformational changes in the GAF domain that may be transmitted to an output domain, often a histidine kinase domain that can lead to downstream signalling to a response regulator that controls the response to light [15]. They have been recently engineered as opto-genetic infrared reporters for deep-tissue and low-background fluorescence imaging [18–20].

In both types of biliproteins, protein environment profoundly influences bilin optical properties. By constraining the flexibility and orientation of the pyrrole rings, the pi-conjugation of the highest occupied and lowest unoccupied molecular orbitals can be extended or shortened, shifting absorption and emission to longer or shorter wavelengths [3,21,22]. Restricted bilin mobility is associated with greater fluorescence yield [3]. Moreover, the binding pocket for the phycocyanobilin (PCB) chromophore of ApcD rigidly holds the PCB chromophore in a planar configuration, which increases the effective conjugation and leads to the red-shifted absorption spectrum of AP-B compared to AP [22]. The protein may also provide hydrogen bonds, provide aromatic residues for pi–pi stacking with the bilin, alter the local dielectric environment and change pyrrole nitrogen pKa values with corresponding effects on the absorption properties [3]. As a result of knowledge gained over many years of study, structure-focused, targeted mutational studies have recently succeeded in increasing fluorescence yield substantially [23–27].

In this report, we have tested the feasibility of drawing on the advances made with natural biliproteins to reconstruct and engineer biliproteins in minimal, compact protein frames supporting selected functions that are expressible in living cells with minimal metabolic load on cellular metabolism. The challenge of dramatically redesigning natural biliproteins along these lines is daunting. Most PBPs require specific interactions with lyase proteins that must be preserved to attach the bilin(s) covalently to the polypeptide [8]. Furthermore, natural PBPs only bind bilin chromophores, which constrains the design of multifunctional proteins. They exhibit complex quaternary interactions that are difficult to overcome without affecting the properties of the globin domain that binds the chromophore(s) [9]. Changes to the bilin-binding environment to adjust absorption and fluorescence properties often have unintended consequences on the folding of the protein. De novo protein design offers an alternative means to enhance biliprotein designability to achieve compactness and adaptable function.

De novo protein design uses first principles of protein folding to intentionally build minimal, structurally transparent protein frameworks that have been proved robust to extensive amino acid sequence variation. α-Helical bundles with a binary patterning of hydrophobic and hydrophilic residues are favoured by de novo protein engineers because of their designability, i.e. many different sequences fold into the same structure [28,29]. Structural stability is built in, resisting temperatures up to boiling [30]. These frames are readily adapted to accommodate many natural and non-natural cofactors to achieve diverse functionality through informed, iterative design [31–37]. They readily lend themselves to progressive engineering directed towards further synthetic applications [30,36,37]. This approach allows the exploitation of design and sequence space undiscovered by natural selection.

The work reported here seeks to define some initial principles to determine whether bilins can be effectively and efficiently bound to completely artificial proteins with no sequence similarity to natural proteins, and if so, how the optical properties of the resulting proteins can be manipulated and optimized. Here we report initial designs to artificial biliproteins to optimize three factors that control the key optical properties of absorption, emission and fluorescence yield: (i) the type of bilin; (ii) Cys localization for bilin binding; and (iii) the bilin binding environment. This work initiates the development of basic design rules for these choices and tests the performance of initial de novo biliprotein light-harvesting maquettes in excitation energy transfer (EET) from different bilins to cyclic tetrapyrrole acceptors bound within the same maquette.

2. Results and discussion

2.1. De novo biliprotein design choices

There are three principal choices in designing de novo biliproteins to control the key optical properties of absorption, emission and fluorescence yield. The most important choice is bilin type. The different conjugation lengths of the various bilins set the expected absorption and emission range, with the longest absorption and emission wavelengths of red to near IR for biliverdin (BV), shorter wavelengths for PCB and phycoerythrobilin (PEB), and the shortest wavelengths for tripyrroles (figure 1). Although not discussed here, other bilin choices are available [39,40] for in vivo ligation to the de novo bundle scaffold.

Figure 1.

Spectral diversity of bilins and increase in the visible absorption band upon binding to protein. (a) Structure of free bilins. (b) Absorbance of free bilins in phosphate-buffered saline, pH 7. (c) Absorbance of bilins bound to protein. TPB and BV are bound to a maquette, while PCB and PEB are bound to the CpcA subunit fused to a maquette [38]. The shorter wavelength major absorption bands are referred to as Soret or UV and the longer wavelength major bands as Q or Vis.

The second choice is the location of the ligating Cys residue within the 4-helix bundle framework and the identity of adjacent amino acids that control the yield of in vivo bilin binding during expression. These selections modulate the accessibility of the Cys residue for interacting with free bilin or bilin lyase and the ability of the region of the bundle to unfold and permit access to the bundle hydrophobic core during the ligation between the Cys sulfhydryl group and bilin vinyl moiety.

The third choice is to select residues that control the environment of the bound bilin to modulate the bilin conformation—specifically, the extension of the tetrapyrrole chain, the angle of rotation between the pyrrole rings [22] and the bilin rigidity. These factors critically influence the effective degree of pi-electron conjugation in the molecular orbitals of the bilin and thereby its optical properties [3]. This is a special design challenge presented by bilin cofactors that is not a concern for previously studied rigid maquette cofactors, including porphyrins, chlorins, iron–sulfur clusters and flavins [31].

2.2. Binary patterning in helical bundle design

Maquettes exploit first principles of protein folding by using repeating amino acid heptads (figure 2) [42]. In an α-helical fold, seven amino acids form complementary hydrogen bonds to complete two helical turns. Binary patterning of these heptads places non-polar residues along one face of the helix with polar residues on the other. Ends of helices are connected with simple glycine (Gly)-rich loops. In aqueous solution, this single polypeptide chain spontaneously sequesters the non-polar residues to form the core of a 4-helix bundle. Maquette frames are highly tolerant to changes in sequences that maintain the general binary patterning. We have the design freedom to move the bilin ligating Cys around the helical and loop regions of the maquette, make additional sequence changes around the Cys, change the net charge of individual helices, alter lengths of loops and make changes in the hydrophobic core, all without compromising the highly tolerant folding of the protein.

Figure 2.

α-Helical-bundle designs of maquettes based on binary patterning of heptad-based sequences (upper left) connected by loops. Non-polar residues (purple) line the interior face of the helix, while polar (black), positive (blue) or negative (red) residues cover the exterior face. An end-on view of the bundle based on a maquette NMR structure [41] is shown at the upper middle, showing non-polar residues (purple) buried in the bundle core. A side view (upper right) colours helices 1, 2, 3 and 4 in red, orange, yellow and green, respectively. A flattened amino acid sequence map (middle) shows non-polar heptad positions a, d and e as purple stripes. The designs of this work use either four net negative helices or alternate negative and positive helices connected by either long or short loops (bottom).

2.3. Bilin lyase co-expression

We exploited the expression system developed previously [40,43,44], which imports bilin synthesis and attachment machinery for ligating diverse phycobilins to PBP apoproteins expressed into E. coli. The plasmids code for the following: haem oxygenase-1 to catalyse BV synthesis from haem; a reductase to reduce BV to either PCB or PEB; and a bilin lyase. We added plasmid PJ414 encoding a maquette to this system. Despite the completely unnatural sequence of the designed bundle, successful coupling between the bilin lyase and the maquette is clear. The higher copy number of the maquette PJ414 expression vector tends to outstrip the bilin cofactor biosynthetic machinery on the low copy number pACYCDuet vector. Nevertheless, expression of the bilin S-type lyase CpcS from Thermosynechococcus elongatus [44] greatly increases the amount of bilin bound. Without PCB lyase, purified maquettes carried minimal amounts of bound PCB along with small amounts of bound BV and haem (see the electronic supplementary material, figure S1, for spectra).

In this study, we show the bilins BV, PCB and PEB all bind covalently to Cys residues in maquettes in vivo (figures 1 and 3; and electronic supplementary material, figure S20) and in vitro (figure 6). We selected the PCB plasmid for the most extensive analysis.

Table 1.

Relative binding yields and optical qualities of maquette designs shown in figure 3. Designs vary residues immediately downstream from the bilin-ligating Cys and Cys positions in the middle loop or second helix.

ID	Cys insertion	Cys location			His position on helix				relative binding yield in vivo	Vis/UV
		helix	loop	position	1	2	3	4	PCB
1	CGEI		2	3	6	6	6	6	36%	0.38
2	CGRI		2	3	6	6	6	6	72%	0.54
3	CGRD		2	3	6	6	6	6	54%	0.47
4	CLRD	2		1	6	6	6	6	100% (0.9% absolute)	0.62
5	CLRD	2		2	6	6	6	6	72%	0.57
6	CLRD	2		3	6	—	6	6	82%	0.70
7	CLRD	2		23	6	6	6	6	21%	0.43
8	CLRD	2		24	6	6	6	6	16%	0.21
9	CLRD	2		25	6	6	6	6	15%	0.31

Figure 3.

(a) Effect of introducing downstream Arg near the loop Cys anchor. PCB binding yield increases and the 650 nm band is enhanced. (b) Effect of placing a CLRD motif near amino (yellow) or acid (brown) terminals of helix 2. Protein concentrations, assayed by tryptophan absorbance at 280 nm, are held constant. Variants are grown in parallel to the same cell densities and purified in parallel with comparable protein yields. Protein is in excess of bound bilin, thus larger absorbances correspond to larger ligation efficiency. Binding yields are given in table 1.

Figure 6.

Absorption (a) and emission (b) properties of representative bili-maquettes ligated in vitro with BV, PEB or PCB with Cys in the helix core (blue) or loop (green) and HPLC-purified. Absorption spectra are normalized for protein concentration. Emission spectra are normalized for absorbance at excitation wavelength: 600, 560 and 580 nm for BV, PEB and PCB, respectively. Excitation spectra are normalized for 600, 575 and 575 nm for BV, PEB and PCB, respectively. Binding yields are given in table 4.

2.4. Modulating bilin binding through Cys exposure

To test the hypothesis that surface exposure of the ligating Cys residue may modulate bilin binding, Cys residues were inserted at different positions in the helical heptad. Even though maquettes are expressed in surplus over bilin, we can compare the relative yields of bound bilin to determine which binding sites are most favourable. When individual Cys residues are inserted in the middle of a helix, both hydrophobic core sites (heptad positions a or d) and surface exposed sites (heptad positions b or c) lead to similar bilin-binding yields (see the electronic supplementary material, figure S2 and table S1). On the other hand, Cys residues inserted in a conformationally flexible loop position displayed greater absorbance and enhanced binding yield. Spontaneous PCB binding without lyase is negligible. These observations support the hypothesis that local unfolding of the protein sequence is more important than Cys exposure in interfacing with bilin lyases in vivo. Local unfolding has been proposed in a structural model that docks the lyase CpcT with the natural biliprotein CpcB; in this model the Cys-containing sequence is inserted into the bilin-carrying cleft of the lyase, which brings the Cys-sulfhydryl and the bilin vinyl into apposition for catalytic ligation [45].

2.5. Modulating bilin binding through nearby Arg and Asp

Cys-containing sequences in natural biliproteins are highly diverse and different sequence classes are associated with specific bilin lyase families [8,13]. Restricting a biliprotein BLAST search to those associated with the S-type lyase used here indicates that Arg and Asp residues are usually found downstream from the Cys in a Cys-Xxx-Arg-Asp (CXRD) pattern, where Xxx is typically a non-polar residue [46]. Arginines can form charge-pairs with bilin propionates, while Asp can interact with pyrrole nitrogens [44]. Figure 3a shows that introducing an Arg two positions after a loop Cys boosts the bilin-binding yield by a factor of 2 and adds structure to the absorption spectrum. Longer wavelength absorption features in natural biliproteins are associated with more extensive pi-orbital conjugations of the more extended bilin conformations and those with smaller dihedral angles between the pyrrole rings, especially between the rings C and D. However, adding an Asp in a loop CXRD motif did not improve binding yield.

2.6. Modulating bilin-binding yield with CXRD position

While cysteines in CXRD motifs are comparably effective in binding PCB at both surface-exposed or core buried helix heptad positions (figure 3b), there is a conspicuous asymmetry in bilin-binding efficiency of the amino end of the helix compared to the carboxyl end of the helix. Bilin-binding yield increases by nearly fivefold when CXRD is placed near the amino rather than the carboxyl terminal of helices. Helical unwinding for insertion of Cys into the lyase for ligation is expected to be easier near either end of the helix. However, when the Cys was placed within two turns or 10.5 Å from the carboxyl terminus of a helix, bilin-binding yield decreased and the absorption broadens (brown spectra in figure 3b). Cys anchoring close to the carboxyl terminus (less than the 13 Å of an extended bilin) may force a twisted geometry.

2.7. Modulating bilin binding with CXRD order

When the Cys is positioned in the middle of a helix, introduction of a negatively charged Asp two residues downstream greatly improves bilin binding in maquettes relative to positively charged lysine (Lys; figure 4a). This is evidence that in contrast to Asp in loops, the carboxylate of Asp in a helix is positioned for favourable interactions with the nitrogens of pyrrole ring B/C. When the Asp carboxylate is replaced with a Glu carboxylate, analogous to the CXRE motif found in some BV-binding bacteriophytochromes, PCB binds less favourably, while comparable amounts of BV are bound (electronic supplementary material, figure S21).

Figure 4.

Comparable binding of CLRD and DRLC motifs in loop and helix 4. (a) Sequence position (black, red, blue). (b) Spectra: replacing D with K reduces yield dramatically (magenta). (c) Overlay of X-ray structure of natural allophycocyanin (green backbone ribbon, purple bilin) with molecular dynamics structure of the maquette (grey) by superimposing allophycocyanin CLRD onto maquette DRLC. Cys (yellow), Arg (blue), Asp (red). (d) Corresponding spectra.

Unexpectedly, the reverse motif, DRXC, appears equally effective in driving bilin ligation in maquettes (figure 4a). While the amount of bilin bound to the maquette varies depending upon details of growth conditions and the relative rate of cellular production of both the bilin chromophore and the maquette protein frame, under controlled conditions of parallel growth and purification, there is comparable PCB binding yield for the DRLC and CLRD motifs (table 2). These results indicate that within a helix, proximity of the Arg and Asp to the ligating Cys is more important than the sequence order.

Table 2.

Relative binding yields and optical qualities of maquette designs shown in figure 4. CXRD and the reverse DRXC sequences are compared. Absolute yield of bilin binding was determined by protein unfolding in acidic urea (see the electronic supplementary material, figure S3) and analytical HPLC.

ID	Cys insertion	Cys location			His position on helix				binding yield in vivo	abs. quantum yield (%)	Vis/UV
		helix	loop	position	1	2	3	4
4	CLRD	2		1	6	6	6	6	0.9	2.9	0.76
10	CLRD		2	3	6	6	6	6	4.2	0.7	1.65
11	CLRD	4		9	6	6	6	6	1.1	1.0	2.03
12	CLRK	4		9	6	6	6	6	0.3	2.5	0.61
13	DRLC	4		9	6	6	6	—	3.1	1.3	1.27
14	DRAC	4		9	6	6	6	—	3.2	0.9	1.02
15	DRLC	4		9	—	6	6, 26	—	3.4	1.6	1.68

Structural insight on this equality can be found by superimposing the bilin-binding Cys-containing helices of the crystal structure of the alpha subunit of the natural biliprotein allophycocyanin from the cyanobacterium Phormidium sp. A09DM (4RMP) [47] with a molecular dynamics structure of the maquette helical bundles (figure 4c) [48]. Despite the reverse order of the sequences, the arginine of both the allophycocyanin and the maquette are in positions close enough to interact with the B ring propionate, while the aspartate of both motifs are close enough to interact with the ring B and C pyrrole nitrogens. Furthermore, the nearly parallel orientations of adjacent helices near the bound bilin in both allophycocyanin and the maquette accommodate intercalation of roughly planar PCB pyrroles between adjacent helices.

2.8. Modulating bili-maquette fluorescence

Figure 5a compares the absorption and fluorescence emission (excitation at 580 nm) spectra of three maquette sequences described in figure 4d. A higher Vis/UV absorbance ratio is correlated with a more rigid and extended bilin conformation [3,49,50]. The DRLC motif binds more PCB than the DRAC motif and has a higher fluorescence quantum yield, consistent with constraint by a better packed, less mobile hydrophobic core.

Figure 5.

(a) The DRLC motif (13,15) binds more PCB and has higher fluorescence quantum yield than a DRAC motif (14). (b) Changing external charge patterning and loop length (4, 16, 18) had minimal effects on bilin attachment and quantum yield. Breaking the 4-helix bundle into a dimer of two helices (17) increases bilin ligation efficiency but lowers fluorescence quantum yield. Spectra were normalized to 50 µM protein.

If maquette-bound bilin orients similarly to the PCB of the alpha subunit of allophycocyanin, the D ring pyrrole should approach position 26 in adjacent helix 3. Indeed, replacing Leu with His in this position noticeably sharpens the Vis absorbance band and increases the Vis/UV ratio (table 2 and figure 4c). This is consistent with a pi-stacking or hydrogen bonding interaction between the helix 3 His and ring D of PCB. This third-helix His improves both binding and fluorescence quantum yield.

Figure 5b shows that changing the external charge patterning from near-neutral to negative modestly improves the fluorescence quantum yield from 2 to 3%. Quantum yields are comparable to that observed for the PCB-binding domain of a natural cyanobacteriochrome [51]. Shortening the loops has little effect, but breaking the central loop (17), so that the 4-helix bundle assembles from two separate subunits, doubles the yield of bilin binding per Cys, presumably through greater lyase accessibility. However, fluorescence quantum yield was relatively poor, probably reflecting higher mobility of the PCB when a single maquette chain is divided into two.

2.9. Lyase strengthens CLRD motif effectiveness

The chaperoning influence of bilin lyase during in vivo expression is absent with in vitro ligation. In such cases, loop Cys position provides the best yield for attachment for all bilins (figure 6, sequence (2)). Without a lyase, cysteine exposure facilitates binding. Although the binding yield was lower, burying the Cys in the core (20) generally enhanced fluorescence quantum yield, indicating that the core restricts quenching associated with bilin mobility.

To test the hypothesis that CLRD motifs improve in vitro bilin binding, we inserted the motif downstream from the buried helical core position at the beginning and end (sequences 5 and 9) of the second helix. Neither CLRD position improved bilin binding or fluorescence quantum yields compared to Cys alone. Indeed, BV binding decreased. This contrast with the results for in vivo binding supports the view that a CLRD motif gains significance through modulating interactions with bilin lyases.

2.10. Lyase–maquette interactions

The tendency of the hydrophobic core of elementary 4-helix bundle synthetic protein designs to associate with bilins even without Cys present probably facilitates the autocatalytic thioether linkage formation to bilin vinyl groups when Cys is introduced. This reaction is analogous to the previously demonstrated ligation of haem vinyls to Cys during the formation of cytochrome c maquettes [52]. However, the autocatalytic bilin reaction in vitro is generally slow compared to in vivo lyase-catalysed ligation. By co-expressing lyases with bilin synthases, the total amount of bilin available for ligation improves.

In natural systems, lyases boost bilin ligation yield and modulate the stereochemistry of bilin ligation and corresponding bilin absorption and emission properties [8,13]. The helical conformations seen in free bilins have relatively weak long wavelength absorptions; more extended conformations imposed by bilin-binding sites provide longer pi-conjugation for more intense absorptions, particularly when torsion angles allow pi-orbitals to overlap optimally [3,49,50,53]. Suppression of bilin flexibility by binding-site confinement and specific interactions with nearby residues inhibits excited state quenching and supports higher fluorescence emission quantum yields [19].

Because maquettes have minimal sequence similarity to natural biliproteins, they are not expected to form specific complexes with natural lyases. Nevertheless, lyases do confer different maquette bilin geometries than seen in lyase-free autocatalysis. Autocatalytic PCB binding to maquettes produces spectral forms with relatively large absorption bands around 650 nm, similar to the absorption seen in autocatalytic binding of PCB to the apo-biliprotein PecA [54]. Four examples are shown as solid lines in figure 7. Maquette designs with a Cys in the central loop show a similar spectrum (dashed black line figure 7) upon in vivo expression with the lyase. However, for maquette Cys in helix positions, lyase-assisted bilin binding generates shorter wavelength, relatively broad and less-structured absorption bands (orange, purple and blue dashed lines figure 7). Apparently for helical ligation sites in maquettes, lyase interaction favours less extended and potentially more varied bilin geometries. These absorption maxima are well within the 535–700 nm range observed for natural PCB biliproteins [26] and match the typical 610–710 nm range of the majority of natural PCB biliproteins (see electronic supplementary material, figure S12) [1,12].

Figure 7.

Maquette-bound PCB has longer wavelength absorption when ligated in vitro (solid lines) compared to lyase-assisted in vivo (dashed). Colours correspond to maquette sequences in figure 6.

Lyases also appear to influence E–Z isomerization at the C15–C16 or possibly the C4–C5 double bond of PCB in maquettes. Under denaturing conditions that largely remove the influence of the protein environment on the bilin absorption (see electronic supplementary material, figures S13 and S14), PCB largely assumes a Z isomer (absorbance maximum 661 nm [3]) when ligated to maquettes in the presence of lyase in vivo. By contrast, in vitro ligation of the Z isomer of PCB purified by high-performance liquid chromatography (HPLC) yields predominantly E isomers (593 nm [55]). Although photoisomerization between Z and E configurations of PCB is a critical part of the signalling process in phytochromes [15,56], there is so far no evidence of analogous photoisomerization in bili-maquettes.

2.11. Tripyrrole and phycocyanobilin tetrapyrrole to chlorin excitation energy transfer

The Q band absorbance maxima and the UV/Vis absorbance ratios for high-yield bili-maquettes lie within the range of natural CpcB and PecB previously produced in E. coli [57], an indication that PCB is bound rigidly. However, fluorescence quantum yields in maquettes of 1 to 3% (compared to 12 and 32% of a range of PCB-binding PBPs [57,58]) show room for improving chromophore rigidity. Nevertheless, these fluorescence levels are adequate for adapting bili-maquette designs towards voltage sensors and energy-transfer agents.

We demonstrate EET from a maquette-bound bilin to another maquette-bound chromophore, analogous to EET in maquettes fused to a PCB-containing phycocyanin α subunit [38]. We created a Zn-tetrapyrrole-macrocycle binding site by inserting a potential His ligand at position 6 on helix 3, approximately 32 Å from the bilin-ligating Cys. Two remaining interior His residues were replaced with alanine to preclude competing ligation sites (sequence (19); table 3). A range of light-active Zn-tetrapyrrole macrocycles ligate to His in this site [59].

Table 4.

Binding yields and optical qualities of maquette designs shown in figure 6.

ID	Cys insertion	Cys location (%)			binding yield in vitro quantum yield (%)
		helix	loop	position	BV	PEB	PCB
2	CGRI		2	3	100 (0.3)	100 (1.8)	100 (1.3)
5	CLRD	2		2	9.7 (0.2)	65 (1.7)	82 (1.1)
9	CLRD	2		25	7.1 (0.7)	77 (1.5)	97 (0.9)
20	CEDA	2		6	70 (1.1)	69 (1.8)	88 (2.0)

Table 3.

Binding yields and optical qualities of maquette designs shown in figure 5b.

ID	Cys insertion	Cys location		His position on helix				binding yield in vivo (%)	abs. quantum yield (%)	Vis/UV
		helix	pos.	1	2	3	4
16	CLRD	2	1	6	6	6	6	1.3	3.0	0.96
17	CLRD	2	1	6	6			5.6	0.9	4.49
18	CLRD	2	1	6	6	6	6	2.1	2.3	1.45
19	CLRD	2	1	—	6	6	—		4.4	2.0

The EET distance between maquette-bound bilin and Zn tetrapyrrole is much shorter than the calculated Förster energy-transfer distance of approximately 50 Å [38]. If chromophores were isotropically oriented, EET efficiency would be above 90%. The transition dipoles of bilins, commonly assumed to be along the direction of the B to D pyrroles [60], would be constrained to be mostly parallel to the long helical axis if buried in the bundle core in an extended conformation. A test of this orientation uses a chromophore with a transition dipole unfavourably oriented nearly perpendicular to the helical-bundle axis, Zn-chlorophyllide a (ZnChl).

For EET donors, we compared two bilin chromophores: PCB and a blue-shifted tripyrrole bilin (TPB) derived from PCB. By raising the pH of a denaturing 6 M guanadinium solution of PCB-bound maquette-19–8.5, the absorption redshifts to 739 nm (figure 8), presumably because of chromophore deprotonation [3]. Over 30 min, the absorption blue-shifts to 586 nm. Mass spectrometry revealed a mass loss of 109 Da, consistent with scission of one pyrrole ring to form a TPB [3,61] (see the electronic supplementary material, figure S10). The detailed chemical structure of this readily generated tripyrrole derivative has not yet been determined, but the high quantum yield of fluorescence (26%) facilitates energy-transfer measurements.

Figure 8.

Formation of TPB by alkaline scission of PCB bound to maquette 19. Bound PCB in 6 M guanidinium (pH 8) has a broad absorption around 650 nm (cyan). Adjusting the pH to 8.5 rapidly shifts to 739 nm (blue); this converts on a minutes timescale to the 586 nm absorbing species (purple) associated with a tripyrrole.

For the energy-transfer measurements, the bilin-bound maquette was HPLC-purified to enrich the fraction of maquette containing bound TPB to about 50%. ZnChl binding (Kd 4 ± 3 nM) to either the PCB or TPB maquette does not distort the bilin absorption spectrum (see the electronic supplementary material, figure S11). Figure 9 shows absorption spectra for individual energy-transfer pigments. As ZnChl was titrated into the TPB maquette, singular value decomposition (SVD) of the absorption spectra confirmed the presence of two spectral forms corresponding to the bound and unbound ZnChl. Figure 10a,b shows binding and saturation of the site at a 1 : 1 stoichiometry. Fitting to a single binding-site model confirms nM binding of ZnChl with and without bound TPB.

Figure 9.

Spectra of individual energy transfer pigments in maquette 19: tripyrrole (pink), PCB (blue) and ZnChl (green).

Figure 10.

Binding titration of ZnChl into TPB maquette demonstrates EET by fluorescence quenching of the tripyrrole derivative of PCB. (a) Absorbance of bili-maquette (purple) and successive additions of ZnChl. PBS buffer, 1 M NaCl (pH 7.4). (b) SVD analysis fits to a single His-ligating site with a nanomolar range Kd. Spectra of His-ligated (red) and non-His ligated ZnChl are superimposed on graph A. (c) Emission spectra of each addition upon excitation at 550 nm. (d) SVD analysis of emission spectra reveals three components fitting to a single His-binding model. The spectra corresponding to bilin-only maquette (red), His-ligated ZnChl (green) and non-His-ligated ZnChl (blue) are superimposed on graph C.

To characterize EET from the bilin to the Zn tetrapyrrole, each titration step was excited at 550 nm. Figure 10c,d shows that SVD analysis of the emission spectra reveals three components fitting a simple binding-site model. The first spectrum corresponds to TPB emission in the maquette without ZnChl with a maximum near 590 nm (red, figure 10c). As ZnChl is added, this emission spectrum is replaced with a second emission spectrum with two peaks of approximately equal intensity at 590 and 675 nm corresponding to emission from TPB and histidine-ligated ZnChl, respectively (green, figure 10c). As excess ZnChl is added, a third spectral component grows that quenches the remaining TPB emission and shows a blue-shifted emission maximum at 665 nm, corresponding to ZnChl not ligated to the maquette histidine (blue, figure 10c). At a 1 : 1 stoichiometry of His-ligated ZnChl per maquette, approximately 30% of the TPB fluorescence is quenched by energy transfer, indicating that the transition dipoles are relatively constrained and unfavourably oriented. At high concentration, ZnChl as a relatively hydrophobic pigment will partition into the hydrophobic core. Once the His site is filled with ZnChl, the additional loosely bound ZnChl will not be held as rigidly or as far from the bound bilin, permitting more efficient fluorescence quenching. Figure 11 shows that ZnChl is more effective at quenching the fluorescence of PCB. At a 1 : 1 stoichiometry of ZnChl per maquette, PCB is 75% quenched, thus the PCB transition dipole is oriented less orthogonal to the dipole of ZnChl.

Figure 11.

EET is demonstrated by fluorescence quenching of PCB. (a) Absorbance of ZnChl titrated into 2.7 µM maquette, 10 mM NaPO₄ (pH 7.4). Insert: PCB and ZnChl peak emission intensities at 626 (red) and 676 nm (blue). (b) Corresponding emission spectra during titration.

2.12. Compact biliprotein design motivations

When natural phytochrome biliproteins are exploited for use as red to near-infrared fluorescent proteins, their utility and performance can be limited due to intrinsic biophysical reasons. As photosensory signalling proteins, they have not evolved to be naturally fluorescent, but rather, their structures were selected to convert their photonic energy into large-scale structural rearrangements. While they have been engineered by directed evolution for enhanced fluorescence by minimizing this Nature-prescribed motion and restricted water access in the bilin-binding GAF domain [18], there is a limit to size reduction [62] due to the structural requirements for bilin binding and stabilization. By using the compact artificial protein platform presented here (approx. half the molecular weight of green fluorescent protein), energy-transfer distances can be shortened and the cellular metabolic burden can be limited when applied as cell-expressible transcriptional reporters and fusion tags.

This work establishes key determinants of biliprotein holo-protein formation in E. coli using genetically encoded components. Companion eukaryotic work has established bili-maquette expression in mammalian neuronal cells. That work, aimed at developing maquette biliproteins for opto-genetic studies, shows that maquette technology can be compacted to a 9 kDa maquette while preserving useful BV fluorescence.

3. Summary

This work extends the fundamental rules for designing protein maquettes to include a new and versatile class of light-active cofactors, bilins, thus enabling the engineering of a broad spectral selection of light-activated artificial proteins (e.g. [40]). Unlike the His ligation used to bind metalated cyclic tetrapyrrole cofactors to selected positions in maquette helical bundles, Cys ligation is used to anchor bilins, analogous to the Cys binding to haem vinyls by maquettes with CXXCH motifs to produce c-type cytochromes. Ligation of various bilins in vitro is favoured by placing Cys in relatively exposed and flexible positions, especially bundle loops, although helical positions are also competent. Arg placed two residues away presumably interacts with the propionate side chains of rings B/C to constrain bilin orientation partially and improve fluorescence yield.

The relatively simple in vitro rules for creating bili-maquettes are altered in the more complex environment of the cell where interactions with natural bilin lyases boost the speed, and alter the yield and stereochemistry of bilin ligation. The S-type lyases operate by partially unfolding both natural and artificial proteins to expose the ligating Cys to the A-ring vinyl of the bilin carried by the lyase. Such unfolding and insertion occurs for Cys in most regions of the bundle with the exception of the carboxy terminal ends of the helices. In the other helical locations along the helices, Arg and Asp two and three residues away are well positioned to interact with the propionate carboxylates and pyrrole nitrogens, respectively, of the bilin to constrain motion and favour more extended chromophore geometries with desirable optical changes in absorption and fluorescence emission quantum efficiency. Arg and Asp are equally effective in a natural CXRD and an unnatural DRXC motif.

Engineering bilin binding in maquettes is a modular, local design issue that leaves the rest of the protein framework free for other functions, such as introduction of a second chromophore binding site for EET, or for fusion to natural proteins for cellular localization. Improvement in optical performance can be expected by adding to these rules, as required for any specific design application.

4. Methods

4.1. Gene constructions

Codon optimized synthetic genes were obtained from DNA2.0 in a PJ414 expression vector. Mutations to parent vector were made with primers synthesized by IDT or Invitrogen (see electronic supplementary material, tables S2–S4). Sequences were verified using UPENN DNA sequencing core.

4.2. Maquette expression

Bilin biosynthetic plasmids were CpcS lyase [44] and PcyA and HO1 [43]. His-tagged maquettes and bilin biosynthetic and attachment enzymes were expressed in E. coli BL21 (DE3). Strains containing bilin biosynthetic plasmids were made competent to take up the maquette plasmid using the Hanahan method [63]. Cells were grown in Terrific Broth [64] to OD at 600 nm between 0.8 and 1 at 37°C at 205 r.p.m. and then induced with 1 mM isopropyl-thiogalactopyranoside for 21 h at 20°C at 260 r.p.m. in bevelled flasks with culture volumes of 100 ml.

4.3. High-performance liquid chromatography methods

Solvents used for pigment and protein purification were a mixture of acetonitrile (ACN) and 0.1% trifluoroacetic acid (TFA) in water. A Waters HPLC was used with a Vydac C18 analytical and preparatory reverse-phase column.

4.4. Protein purification

Proteins expressed with His tags were purified by resuspending pelleted cells in 300 mM NaCl, 20 mM imidazole, 50 mM NaH₂PO₄ (pH 8) buffer, homogenized and microtip sonicated. Lysate was centrifuged at 25 000g for 25 min, with supernatant applied to a Qiagen Ni-NTA superflow resin and purified via the gravity column method. One litre of cell culture yields approximately 20 mg of maquette. PCB bili-maquette 19 without His tag was purified via solvent extraction, by washing cell pellets with 0.1% TFA in water and then sonicating with 50% ACN, 0.1% TFA in water. Initial extractions lacked blue colour and were discarded; after approximately 60 ml, subsequent blue extractions were pooled and rotovapped until dry.

Unless otherwise noted, both tagged and untagged proteins were then HPLC-purified using a 30 ml gradient from 30 to 42% ACN; spectra collected during the gradient quantified the efficiency of PCB attached to mutants in figure 4a (see electronic supplementary material, figure S4). Purified bili-maquette was collected, lyophilized and stored at −20°C in the dark until use. Molecular weight was assayed by both SDS PAGE (Invitrogen Novex NuPAGE electrophoresis system, 4–12% Bis-Tris gels, MES running buffer) and MALDI [30] (see the electronic supplementary material, figure S5).

4.5. Pigment purification

PCB was extracted using a modified procedure of [65] via overnight methanol refluxing of 50 mg of a gene fusion between CpcA and maquette [38]. The rotovapped reflux mixture was extracted with 1 ml of ethanol, and purified by HPLC using a 500 ml 20–70% ACN gradient on a C-18 column (see the electronic supplementary material, figure S16). MALDI-MS verified the mass of PCB (see the electronic supplementary material, figure S17).

PEB was extracted from a His-tagged CpcA modified to delete the covalently ligating Cys (electronic supplementary material, table S5). After His tag purification, 25 mg of protein was precipitated by adding methanol at 50% followed by a half volume of CHCl₃ and then drying by rotary evaporation at 40°C. MALDI mass spectrometry used 2,5-dihydroxybenzoic acid matrix (see electronic supplementary material, figure S18). HPLC verified purity (see electronic supplementary material, figure S19). Pigment concentration assay used extinction coefficients (ɛ) for free PEB of 25.2 mM⁻¹ cm⁻¹ at 591 nm [66] and for free PCB ɛ = 37.9 mM⁻¹ cm⁻¹ at 690 nm in 5% HCl/methanol.

ZnChl was prepared by addition of Zn to pheophorbide a (Frontier Scientific) as described [67]. DMSO-solubilized ZnChl stock concentrations were determined by dilution in methanol using an extinction coefficient of 64 000 M⁻¹ cm⁻¹ at 656 nm [68].

4.6. In vitro bilin attachment

Maquettes were reduced with 5 mM DTT, followed by PD-10 size-exclusion chromatography before addition of bilin. BV reaction was performed in PBS buffer at pH 7.4 for 4 h with 100 µM protein and 500 µM BV. Free BV was separated from BV-maquette by a size-exclusion column equilibrated with PBS (pH 7.4). PEB in vitro treatment was performed overnight in 60 µM protein 50 mM Tris (pH 8), while PCB attachment was performed in 30 µM protein 50 mM NaPO4 (pH 7) as described in [69]. In vitro PEB/maquette attachment reactions were purified using a 700 ml gradient from 30 to 60% ACN for sequences 5, 9 and 20 and a 500 ml gradient from 20 to 70% ACN for sequence 2. PCB in vitro maquette attachment reactions were purified using a 500 ml gradient from 20 to 70% ACN. To form the tripyrrole, sequence 19 with bound PCB was solubilized in 6 M guanadinium-HCl (pH 8) and raised to pH 8.5 with NaOH. After 30 min, the solution was HPLC-purified using a 500 ml gradient from 20 to 70% ACN.

4.7. Protein, phycocyanobilin concentration and quantum yield determination

UV/Vis absorbance was measured using a Varian Cary-50 spectrophotometer. Fluorescence spectra were obtained using a Horiba Fluorolog 2 fluorimeter at 20°C. The PCB concentration was calculated by denaturing in 8 M urea (pH 2), using an ε 660 of 35.4 mM⁻¹ cm⁻¹ [70]. HPLC spectra show that PCB absorbance at 280 nm is minor compared to Trp absorbance. Relative bilin yield was assayed by normalizing 660 nm absorbance in acidic urea to the 280 nm Trp absorbance in PBS after purification.

The quantum yield of PEB and PCB bound to maquettes used the method in [43], after PD-10 size-exclusion chromatography using cresyl violet perchlorate (quantum yield 0.54 in methanol) [71] as a standard. For BV, four dilutions of each BV-maquette and Cy5 standard in PBS were made. Absorbance at 600 nm was plotted against integrated emission from 635 nm to 830 nm. Quantum yield was determined using the slope of the dilution lines relative to that of Cy5 (quantum yield of Cy5 of 0.27) [72].

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

J.A.M. engineered and performed bili-maquette molecular biology, purification and physical and spectroscopic characterization, and contributed to manuscript drafting. M.S. assayed BV binding. G.K. assisted in experimental design and engineering of bili-maquettes and synthesized ZnChl pigment. B.Y.C. supervised in vitro BV binding measurements. D.A.B. provided CpcS, haem oxygenase and PcyA overexpression bacterial strains. P.L.D. supervised maquette design. C.C.M. assisted in bili-maquette design and analysis, and directed manuscript drafting. All the authors gave their final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

This research was carried out as part of the Photosynthetic Antenna Research Center (PARC), an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award DESC0001035 supporting J.A.M., M.S., G.K., D.A.B., P.L.D. and C.C.M. in maquette construction and characterization as well as chlorin synthesis. National Science Foundation (CBET 126497) supported M.S. and B.Y.C. in biliverdin maquette design development.