Sortase-mediated ligation of cytochrome c peroxidase and cytochrome c highlights the roles of dynamics and conformational specificity for interprotein electron transfer

Abstract

The complex between cytochrome c peroxidase (CcP) and cytochrome c (Cc) is an important model system for studying interprotein electron transfer (ET). Low ionic strength conditions stabilize the CcP:Cc complex, but promote unfavorable second-site binding of Cc. Conversely, high ionic strengths favor the 1:1 complex but promote its dissociation. We sought to stabilize the complex and minimize second-site binding by linking the two proteins together via sortase-mediated transpeptidation. Ligation efficiency of the two proteins depends on the length of the flexible linker and the conditions of the ligation reaction. Structural comparisons and AI-based predictions indicate that the conformations assumed by the fusion proteins depend substantially on the linker length. A short linker allows the association mode found in the 1:1 non-covalent complex but favors more extended states. Longer linkers are more conducive to productive complex formation but still sample other conformational states unconducive to interprotein ET. The degree of predicted conformational variation in the fusion proteins agrees well with their ET reactivity and structural analyses by crystallography and small-angle x-ray scattering. Our findings underscore that specific interaction modes between redox partners influence their electronic communication and reveal that interdomain linkers have the potential to control intermolecular reactions and alter the sampling of productive interfaces.

1. Introduction

Researchers have long been fascinated by the long distances over which electron transfer (ET) in proteins can manifest; these reactions occur with high rates, remarkable specificity and relatively small losses in free energy [1]. Over the years, Gray, Winkler and co-workers have made tremendous contributions to our understanding of protein ET, both with respect to chemical mechanism and biological functionality [2–5]. Recent work includes the study of ET across protein interfaces, where dynamics can play an important role in reactivity [6].

Cytochrome c peroxidase (CcP) and cytochrome c (Cc) from yeast have long stood as a model system for understanding interprotein electron transfer [7–11]. CcP reduces H₂O₂ to H₂O with two equivalents of Fe(II) Cc (Scheme 1). Peroxide initially reacts with CcP to form Compound I (Cpd I), which contains a ferryl heme (Fe(IV)=O) and a Trp191 radical cation (Trp191^•+). Trp191^•+ oxidizes the Fe(II) Cc heme, which is ~20 Å away. The second Fe(II) Cc is oxidized at a slower rate due to the intramolecular electron exchange between Trp191^•+ and Fe(IV)=O in Compound II (Cpd II), which lies toward the ferryl species at physiological pH [12,13]. Photoinitiated electron transfer can be induced by substituting the CcP heme with Zn(II) protoporphyrin IX (ZnP), a closed shell-metal porphyrin (Scheme 2) [7,14–18]. The photoexcited ³ZnP in the substituted protein (ZnCcP) ejects an electron across the ZnCcP:Cc interface to produce a charge-separated species composed of ZnP^•+ and Fe(II) Cc. The reduced Cc will then transfer an electron back to CcP through a hole-hopping mechanism involving the formation of Trp191^•+, analogous to peroxide reduction in the native heme system. The growth and decay of the participating species can be spectroscopically monitored through time-resolved spectroscopy, although in ZnCcP, the back ET rate exceeds the forward rates and hence little charge-separated intermediate builds up. One potential factor influencing the slower forward ET rate compared to the recombination back rate is the dynamic sampling of related conformations within the associated proteins, of which only a small subset are reactive [19–21].

Scheme 1.

Scheme 2.

The photoexcited reaction depends on one molar equivalent of Cc. However, in low ionic strength buffers, which stabilize the CcP:Cc complex against dissociation, secondary Cc binding affects the ET mechanism [22]. Conversely, high ionic strength buffers do not form the active complex as readily, but favor the 1:1 complex and thereby eliminate secondary Cc binding effects [20,23]. We sought to covalently ligate CcP and Cc together to stabilize the CcP:Cc complex, limit secondary Cc binding, and favor the productive configurations within the associated proteins. The CcP:Cc complex has been linked previously through disulfide crosslinking at the interface; the bonded complex was highly active but the interprotein linkage was redox sensitive and limited dynamics of the fused proteins [24]. Covalent ligation through joining the flexible termini of the proteins with variable linkers promotes association of the CcP and Cc domains due to their high effective concentrations, discourages secondary binding interactions of Cc, and allows for dynamics about the interface.

Sortase-mediated transpeptidation is a useful tool for protein modifications. S. aureus sortase A has been used for the attachment of fluorescent and spin active probes, the immobilization of proteins to surfaces, protein oligomerization, and the production of unnatural chimeric proteins [25–30]. Furthermore, sortase A ligation has been employed to create specific protein fusions and for attaching small NMR-silent protein probes to proteins of interest [31,32]. Herein, we take advantage of sortase-A ligation to couple two proteins together that require individual expression and purification for proper folding and cofactor incorporation (Fe-heme or Zn-porphyrin for CcP, c-type heme for Cc).

For these studies, we optimize the conditions of sortase A protein ligation by varying temperature, time, and linker length. Our results show that linker length plays a role in the association of the redox partners. Despite longer linkers allowing the native association mode of the CcP:Cc complex, ET rates only approach that of the free complex in the fusion proteins. Through structural analysis and computational predictions, we find that the shortest linker disfavors the native electrostatically driven complex, whereas longer linkers allow the complex, but also sampling of inactive conformations. These studies further highlight the sensitivity of ET reactions between CcP and Cc to the native protein association mode and provide insight into the influence of linker regions on redox reactivity of fused protein domains.

2. Results

2.1. Rosetta assessment of ligating the CcP:Cc complex with sortase A

Sortase A-mediated transpeptidation reactions have been used previously for creating protein fusions, ligating unnatural chimeric proteins, and tagging proteins of interest with small protein tags [30–32]. To predict if sortase A would be a viable candidate for ligating CcP and Cc, the crystal structure of the unligated complex was evaluated with the loop modelling protocol of Rosetta [33,34] with the residue sequence LPATGG as the ligating loop between the C-terminus of CcP and the N-terminus of Cc. Loop modeling in the context of the 1:1 CcP:Cc complex was carried out for the LPATGG loop and 5–14 residues of the CcP C-terminus and the first 6 residues of the Cc N-terminus (Figure 1a). The predictions gave promising results for a stable ligation in the CcP:Cc native association mode and accommodated the LPATGG motif, which is required for recognition by sortase A (Figure 1b). Small perturbations in salt-bridge and H-bonded residues predicted in the linked structure could potentially affect complex association and ET kinetics (Figure 1c). Based on these predictions, CcP and Cc variants were designed and produced such that CcP was C-terminally appended with the sortase A recognition motif (LPATG) either alone or extended by GGSG motifs and Cc was N-terminally modified with a poly-glycine motif, also needed for ligation.

Figure 1.

Rosetta modeling of CcP:Cc fusion proteins. a) Rosetta predicted loop models using the last 5 (purple) or 14 (blue) residues of CcP, the first 6 residues of Cc, and the LPATGG sequence for loop building compared to the crystallized complex (PDB: 1u74, white). b) Structural comparison of the loops in lowest energy generated models using different portions of the CcP C-terminus. Rosetta loop modelling predicts that the LPATGG linker can assume a low-energy conformation to ligate the CcP:Cc active complex. c) Interface region between CcP and Cc in the crystallized complex vs in the Rosetta predicted models highlighting salt-bridge and weak H-bonded residues. Slight perturbations from the native complex observed in the predicted structure may have an impact on ET and complex association in the linked protein.

2.2. Characterization and optimization of sortase-mediated ligation of CcP-X_n-Cc

Sortase A catalyzes the ligation of the LPATG motif to a poly-glycine containing peptide (Scheme 3) [25]. The ligation of CcP-X_n (X_n = LPATG, GGSGLPATG, or GGSG(2x)LPATG) to GGG-Cc was optimized by varying temperature, time, and CcP C-terminal length. Initially, the reaction of CcP-X_n, GGG-Cc, and sortase was allowed to proceed overnight at 4 °C. A small amount of the ligated product was confirmed via SDS-PAGE with low yields. Increasing the reaction temperature to 25 °C was expected to increase reaction yields; however, the higher temperature caused sortase A to precipitate. Instead, the reaction was allowed to proceed at 17 °C where sortase A remained soluble. The reaction was monitored over the course of a week and SDS-PAGE samples were taken each day (Figure 2a). The reaction appeared to reach equilibrium at ~5 days. The ligated reaction product was purified and its sequence confirmed via mass spectrometry (MS), which detected the tryptic peptide TLEEQGLLPATGGGSAK for the linker region with y-ions 4–10, and b-ions 8–14 observed.

Scheme 3.

Figure 2.

Time course for sortase A ligation. SDS PAGE of a) CcP-LPATG and b) CcP-GGSGLPATG ligating to GGG-Cc by sortase A over time at 17 °C. CcP-LPATGGG-Cc yield (46 kDa) grew in slowly over 7 days, whereas the CcP-GGSGLPATGGG-Cc ligation completed after only 1–2 days.

To improve the ligation reaction the CcP-LPATG variants were extended with the addition of one or two GGSG motifs between CcP and the C-terminal LPATG (CcP-GGSGLPATG and CcP-GGSG(2x)LPATG, respectively). CcP-GGSGLPATG and CcP-GGSG(2x)LPATG were allowed to react with GGG-Cc and sortase A at 17 °C. SDS-PAGE samples of the CcP-GGSGLPATGGGG-Cc ligation product revealed a higher yield of the ligated protein than with the shorter C-terminal CcP sequence (Figure 2b). Additionally, the reaction to generate the longer linker variant completed after only one day, presumably owing to the increased flexibility of the longer CcP C-terminal tail and its better accessibility to sortase.

2.3. Reactivity of the ligated redox partners

In yeast, heme bound CcP (FeCcP) reduces H₂O₂ to H₂O via formation of Cpd I on CcP. Fe(II) Cc reduces Cpd I to Cpd II, which can then be reduced by a second molecule of Fe(II) Cc (Scheme 1). Single-turnover reactions were monitored spectroscopically for the shortest variant, FeCcP-LPATGGG-Fe(II) Cc and the free complex (Figure 3a,b). At t = 0 s, the sample displayed a characteristic Soret peak in the UV-vis spectra at 410 nm (ε_{410 nm} = 200 mM⁻¹ cm⁻¹) arising from presence of the two hemes in CcP and Cc [35,36]. The Q-bands of Fe(II) Cc were also observed in the spectra at 520 and 550 nm (Figure 3a). Addition of peroxide to the ligated complex formed CcP Cpd I, which was stable for minutes in the presence of Fe(III) Cc (Figure 3b) [37]. With Fe(II) Cc, CcP Cpd I formed to a much lesser extent, being actively reduced by the tethered Fe(II) Cc, which converted to Fe(III) Cc on the same time scale. Thus, the ligated FeCcP-LPATG-Cc complex was enzymatically active, with the rate of Fe(II) Cc oxidation limited by the mixing of peroxide and formation of CcP Cpd I. At low ionic strength and equal concentrations, CcP and Cc will form a 1:1 complex. Indeed, the rates of Fe(II) Cc oxidation for the free complex are similar as those of FeCcP-LPATG-Cc (initial rate ~0.45 μM/s); however, the behavior of CpdI is quite different when the proteins are unlinked, with some 434 nm species persisting as it does with Fe(III) Cc (Figure 3a). This unreacted Cpd I indicates that the free complex allows for a 2:1 reaction between Fe(II) Cc and a portion of CcP (Scheme 1), leaving some unreacted CpdI. Thus, the reaction of FeCcP-LPATG-Cc proceeds unimolecularly because CcP in the fusion protein only reacts with the tethered Fe(II) Cc and, in doing so, reduces all of the CpdI. The fate of CpdII (420 nm) after initial reduction is difficult to discern owing to Soret band overlap, but likely involves slow side oxidations, resulting in, for example, intermolecular CcP disulfide formation.

Figure 3.

Single-turnover reaction of the oxidation of Fe(II) Cc in CcP-LPATGGG-Cc. a) Comparison of the initial spectra of CcP-LPATGGG-Fe(II) Cc (0.0 s) to CcP-LPATGGG-Fe(III) Cc (70.0 s). Q-bands from the Fe(II) Cc at 520 and 550 nm, diminish as Cc is oxidized by CcP. b) Oxidation of Fe(II) Cc monitored at 550–540 nm and the formation and decay of CpdI CcP at 434 nm for both the CcP-LPATGGG-Cc fusion and the unligated complex. Fe(II) Cc is oxidized in the fusion protein on the same timescale as the growth and consumption of CpdI CcP. In the 1:1 stoichiometric bimolecular reaction with Fe(II) Cc, some of the 434 nm remains unreacted over the course of the measurement, similar to what is observed with Fe(III) Cc, due to a portion of Fe(II) Cc reacting with CcP in a 2:1 Cc:CcP ratio. Initial rates of Fe(II) Cc oxidation were estimated from the slope of the first 3–5 time points in the plot.

Substitution of heme with ZnP in CcP allows for photoexcitation and spectral monitoring of ET between ZnCcP and Fe(III) Cc (Scheme 2). The rate constant for the ZnCcP-X_n-Fe(II) Cc triplet state decay (k_D) in the longer ligated variants (CcP-GGSGLPATGGG-Cc and CcP-GGSG(2x)LPATGGG-Cc) was faster than for CcP-X_n alone, thereby indicating that the linker influences the electronic environment of ZnP to some extent. Note that because Fe(II) Cc cannot be further reduced by ³ZnCcP, it makes no contribution to the decay rate. However, Fe(III) Cc is able to accept an electron from ³ZnCcP, and therefore the rate constant of triplet state quenching in the presence of Fe(III) Cc (k_p) reflects both the decay to the ground state (k_D) and ET to Fe(III) Cc (k_e). In the ZnCcP-LPATGGG-Cc variant, the equivalence of k_p and k_D indicated no ET to Cc from ³ZnP (Table 1), despite the ability of FeCcP-LPATGGG-Cc to oxidize Fe(II) Cc with peroxide. The oxidation of Fe(II) Cc by Cpd I has a similar driving force as the reduction of Cc Fe(III) by ³ZnP (E^0’ ~ 0.7 V); however, the former reaction involves Fe(II) Cc reducing Trp191^•+; whereas the latter involves ³ZnP reducing Fe(III) Cc, which are further separated and more weakly coupled [38,39]. Nonetheless, when the linker in the ligated complex is extended by one or two GGSG sequences, k_p increased due to ET from ³ZnP to Fe(III) Cc. Indeed, the quenching rates for Fe(III) Cc in the variant with the longest linker approached that of the non-covalent native complex. Thus, despite all of the variants being predicted to assume the native electrostatically driven mode of association, longer linker lengths substantially increase interprotein ET within Zn CcP-X_n-Cc.

Table 1.

Rate constants for Zn CcP-X_n-Cc decay kinetics. k_p is the rate constant for ³ZnP decay in the presence of Fe(III)Cc and reflects the sum of ³ZnP decay in the absence of Fe(III) Cc (k_D) and the rate constant for ET from ³ZnP to Fe(III) Cc (ke), i.e. k_p = k_D + k_e.

Variant	kD (s−1)	kp (s−1)	ke (S−1)
wt CcP-LPATGGG	114 ± 5	-	-
wt CcP-LPATGGG-Cc	120 ± 4	121 ± 2	-
wt CcP-GGSGLPATG	111 ± 16	-	-
wt CcP-GGSGLPATG-Cc	173 ± 20	213 ± 6	40 ± 21
wt CcP-GGSG(2x)LPATG-Cc	192 ± 26	248 ± 7	56 ± 27
wt CcP:Cc [18]	114 ± 6	230 ± 10	116 ± 13

2.4. AlphaFold predicts differences in the CcP-X_n-Cc structures

Although Rosetta loop modeling suggested that the LPATGGG covalent tether between CcP to Cc was stable within the native non-covalent complex, AlphaFold predicted CcP-LPATGGG-Cc to preferentially disfavor the native complex (Figure 4a). In the AlphaFold structure, Cc is dissociated from the interface but constrained in space by the linker. The region of the flexible LPATGGG linker was predicted with the least confidence (by pLDDT scores) and given the little contact between the two domains, in solution, Cc is likely conformationally heterogeneous with respect to CcP (Figure 4b). Of the five highest confidence models, the one closest in domain arrangement to the native complex still gives a metal-to-metal separation that is 18 Å longer than the 26.5 Å separation of the native complex (Figure 4c). Such a large increase in distance between the heme in Cc and CcP will decrease the ET rate by many orders of magnitude, thus rendering CcP-LPATGGG-Cc inactive, as observed in the triplet decay rates.

Figure 4.

a) AlphaFold models of the CcP-X_n-Cc sortase A ligated complex (CcP-LPATGGG-Cc: purple, CcP-GGSGLPATGGG-Cc: yellow, CcP-GGSG(2x)LPATGGG-Cc: brown) versus the free CcP:Cc complex (PDB: 1u74, white). Only CcP-LPATGGG-Cc is not predicted to form the active complex. b) AlphaFold model of CcP-LPATGGG-Cc colored by pLDDT value. Lowest confidence values are observed in the flexible LPATGGG linker. c) Comparison of the five predicted AlphaFold models for CcP-LPATGGG-Cc (purple) vs the free complex (white). All models contain Cc displaced from the interface of the CcP:Cc active complex. Inset: Distance difference between the complex and CcP-LPATGGG-Cc.

2.5. Small Angle X-ray Scattering (SAXS) and X-ray crystallography of CcP-X_n-Cc variants

SAXS was employed to investigate the solution conformational properties of the variants and differentiate distended from globular states. SAXS profiles of the ligated proteins CcP-X_n-Cc and the free CcP:Cc complex along with FoXS fits of the AlphaFold predicted models revealed substantial changes in the scattering profile between CcP-LPATGGG-Cc and the CcP:Cc complex, whereas CcP-GGSGLPATGGG-Cc matched closely to the native CcP:Cc profile (Figure 5a). In particular, Kratky plots, which are especially sensitive to polymer flexibility, revealed shape changes in the low q region of CcP-LPATGGG-Cc that were indicative of an extended state compared to the more globular native complex (Figure 5c).

Figure 5.

SAXS data and model-fit profiles calculated in FOXS using the AlphaFold predictions. a) SAXS scattering profiles fit well to model profiles calculated in FOXS as indicated by low residuals. b) Kratky plot profiles of the experimental data and respective fits from panel (a). CcP-LPATGGG-Cc exhibits features of more extended conformations in the low q region compared to the free CcP:Cc complex.

Of the linker variants, only CcP-GGSG(2x)LPATGGG-Cc was able to be crystallized and diffracted via X-ray crystallography. The crystals were small and diffracted poorly and only produced a 3.2 Å resolution structure of low to medium quality (Table 2). Nonetheless, the structure confirms that in the crystal lattice the longest linker variant assumes the same association mode between CcP and Cc as the native complex (Figure 6). Moreover, there is clear electron density for only the beginning GGSG portion of the linker, but much weaker density for the remaining loop. Thus, the linker is flexible and likely disordered in solution. Side-chain densities at the CcP:Cc interface are difficult to discern in the electron density, but likely similar to the native complex, given that the backbone positioning is nearly identical.

Table 2.

Data collection and refinement statistics. Values in parentheses are for the outer shell.

PDBid	9Q39
Resolution range (Å)	44.42 –3.24 (3.49–3.24
Space group	C121
a, b, c (Å)	160.66 60.60 107.40
α, β, γ (°)	46.41 90 90
Total reflections	17576 (3661)
Unique reflections	6454 (1337)
Completeness (%)	94.4 (95.7)
Redundancy	2.7
Wilson B-factor (Å2)	61.64
Rmerge (%)	25.7 (62.6)
Rmeas (%)	31.5 (76.6)
Rpim (%)	17.9 (43.4)
CC1/2 (%)	96.7 (67.9)
CC* (%)	99.2 (89.9)
Mean I/σ (I)	4.4 (1.8)
Reflections used in refinement	6422 (1283)
Reflections used for R-free	650 (126)
R-work (%)	27.6 (32.5)
R-free (%)	33.1 (39.2)
Number of non-hydrogen atoms	3354
macromolecules	3263
ligands	91
solvent	0
Protein residues	413
RMS(bonds)	0.004
RMS(angles)	0.74
Ramachandran favored (%)	89.5
Ramachandran allowed (%)	9.0
Ramachandran outliers (%)	1.5
Rotamer outliers (%)	0.58
Clashscore	12.96
Average B-factor	61.7
macromolecules	61.8
ligands	57.4
solvent	0

Figure 6.

Superposition of CcP-GGSGGGSGLPATG-Cc (PDB: 9Q39) and CcP:Cc complex (PDB: 1u74) crystal structures. Inset: 2F_O-F_C map contoured to 0.7 σ around the linker region. The ligated protein structure shows that the fusion protein is able to adopt the electrostatic complex conformation with the longer linker.

2.6. AFsample2 predicts alternate conformations for the ligated variants

AFsample2 is a recently developed tool for broadening the scope of predicted protein structures by AlphaFold and for sampling different conformational states of these structures [40]. We utilized AFsample2 to predict potential conformational states of the ligated proteins and compared these results to the unligated native complex (Figure 7). Free CcP and Cc was predicted to favor one primary conformation, the active complex, whereas the CcP-X_n-Cc variants assume multiple conformations with Cc positioned in different orientations relative to CcP. Interestingly, the longer linker protein favored fewer of these extended states and samples the active complex more than the shortest linker. Furthermore, when the linker is extended to the extreme such that X_n = (GGSG)₆LPATGGG, the structure converges toward the native association mode, and the distribution of heme-to-heme distances narrow accordingly. These results correlate with the triplet quenching data observed in ZnCcP-X_n-Cc, in that CcP-LPATGGG-Cc is unable to eject an electron across the interface from ³ZnP to Fe(III) Cc, whereas ET can be partially rescued with longer linkers to that of the free complex, likely owing to the sampling of other states with large heme-to-heme distances.

Figure 7.

AFsample2 predictions of conformational heterogeneity. a) AFsample2 models of the free CcP:Cc complex (white) and the sortase A ligated complexes: CcP-LPATGGG-Cc (purple), CcP-GGSGLPATGGG-Cc (yellow), and CcP-GGSG(6x)LPATG-Cc (brown). b) Distances between heme coordinating His residues in CcP and Cc with respect to model confidence. AFSample2 predicts that the CcP:Cc complex prefers one dominant conformation, with even small deviations degrading confidence of the predictions, whereas the fused proteins sample multiple conformations of similar confidence, that include the native complex to varying degrees. When the linker is extended to be very long, conformational sampling converges on the CcP:Cc native complex with preference for one dominant conformation.

3. Discussion

The complex between cytochrome c peroxidase and cytochrome c in yeast has been extensively studied for understanding the underlying mechanism of interprotein ET [7–11]. The CcP:Cc redox active complex is stabilized against dissociation in low ionic strength buffers, but the secondary binding of Cc complicates reactivity [20,23]. Furthermore, the native reaction requires two equivalents of Cc to reduce peroxide and thus multiple dissociation/association steps of the redox partners are required. In order to promote and constrain the association of the complex, but avoid second-site and sequential binding, sortase-mediated transpeptidation was utilized to link the C-terminus of CcP to the N-terminus of Cc. Our ligation reaction results support the utility of sortase-ligation for tethering small proteins like CcP (34 kDa) and Cc (12 kDa) together. The reaction kinetics of the ligation is highly dependent on the temperature and length of linker sequence used, with CcP-LPATGGG-Cc taking 5 days to complete at 17 °C and the longer variants (CcP-GGSGLPATG-Cc and CcP-GGSG(2x)LPATG-Cc) only taking a mere 1–2 days (Figure 2). This may be due to the association of the electrostatic complex prior to ligation making it difficult for the sortase A enzyme to catalyze the ligation when the linker is small. However, when the linker is extended by as little as 4 residues, the CcP C-terminus has sufficient flexibility to reach the associated Cc and for sortase A to catalyze transpeptidation.

Importantly, we find that the linker length in CcP-X_n-Cc influences ET rates between CcP and Cc. In FeCcP-LPATGGG-Cc, CcP Cpd I oxidizes the ligated Fe(II) Cc, although the reaction is likely slow compared to the ET reaction between optimally associated CcP Cpd I and Fe(II) Cc (k_e = 2 × 10⁶ sec⁻¹ [39]). We suspect based on the computational modeling and SAXS data that the shortest linker variant does not readily sample the native complex owing to constraints of the linker. Thus, the Fe(II) Cc oxidation rate primarily reflects domain association dynamics. This rate of sampling is so slow that in the Zn-substituted protein, ³ZnP is not sufficiently long-lived to allow oxidative quenching by the tethered Fe(II) Cc (Table 1). Again, the reactions in the two systems differ; Cpd I is longer-lived than ³ZnP and better coupled to the Cc heme via Trp191^•+ (Schemes I and II). Moreover, in the ZnP case, because the rate constant for forward ET (k_e) is much smaller than the recombination rate constant (k_eb), little charge-separated species accumulates and thus, we are unable to monitor the back ET reaction. Electron transfer between CcP and Cc is sensitive to the protein interface, and slight changes to Cc positioning upon oxidation and reduction can impact the ET rates of both the forward and back ET steps differentially [7,14–16,41,42]. Therefore, it is reasonable that no forward ET is observed for the Zn substituted system in the case of the shorter linker variant, but Fe(II) Cc oxidation can still proceed provided that the native complex can be sampled, if infrequently.

The rate of forward ET from ³ZnP to Fe(III) Cc can be partially rescued by introducing a longer linker in the ligated proteins. AlphaFold3 predictions suggest that by extending the linker by a GGSG motif, the electrostatic complex becomes the preferred association mode, unlike for the CcP-LPATGGG-Cc protein (Figure 4). Kratky profiles of SAXS data confirm that CcP-LPATGGG-Cc forms a distended state relative to the other variants (Figure 5). This change in conformational preference could be quite substantial considering that even a small deviation from the native complex can affect the ET kinetics, as observed when single residue substitutions were made at the protein interface [42]. Despite the SAXS profiles and predicted structures being nearly identical, the longer linker variants are unable to fully recover k_e to the value exhibited by the native non-covalent complex (Table 1). However, AFsample2, which broadens the scope of conformer and state prediction by introducing more noise into the prediction neural network [40] reveals a distribution of conformational states for the longer-linker variants that mostly favor the native association mode (Figure 7a). However, a few of these states have heme-to-heme distances too large for efficient interprotein ET. This result potentially explains why the ligated protein cannot fully rescue the forward ET rate for the Zn substituted variants. If the linked proteins are sampling other states even a fraction of the time relative to the associated state, then ET from CcP to Cc will be diminished, which is what was observed in the ZnCcP-X_n-Cc redox kinetics. These results reiterate the sensitivity of interprotein ET to the specific interfaces that redox partners form between them and how the partners sample productive conformations. When considering that homologous redox partners can be found interacting either covalently or non-covalently in related systems, our findings highlight the roles that linkers may play in modulating dynamics of association and hence overall reactivity. In general, disordered linkers of optimal length act to increase the effective concentration of one protein module with respect to another and thereby enhance interactions between weakly interacting partners [43]. However, shorter linkers can restrict conformations and excluded volume effects entropically favor the more extended states of fused domains [43,44]. Such thermodynamic consequences may have to be overcome by tuning association affinities, linker flexibility and/or dynamic properties when evolving covalently coupled systems.

4. Conclusions

In this work, we successfully optimized the ligation of the redox partners CcP and Cc through a sortase-mediated transpeptidation reaction by varying linker length and reaction conditions. Although variants with a range of linker lengths between the CcP:Cc pairs all allow redox communication between the partners, there is a large variation in the sampling of productive association modes among them. These domain dynamics rate-limit interprotein ET. Variants with the longest linkers, which are presumably unconstrained to form the active complex, approach, but do not achieve, the rates of the native non-covalent complex. The AI-based prediction method of AFsample2 generates distributions of conformational states that agree remarkably well with the reactivity patterns and conformational properties of the variants. Our results provide insight into the utility of tethering redox pairs to prevent intermolecular reactions and the effects disordered linkers may have on forming productive interfaces within protein complexes.

5. Materials and Methods

5.1. Mutagenesis of -LPATG and GGG- tagged mutants

The sortase-recognition motif, LPATG was cloned into the ppSUMO vector containing the CcP gene at the C-terminus of CcP (CcP-LPATG). To extend the flexible linker a GGSG motif was introduced between the C-terminus of CcP and LPATG in CcP-LPATG (CcP-GGSGLPATG) and was further extended with an additional GGSG sequence (CcP-GGSG(2x)LPATG). A TEV cleavable sequence followed by an N-terminal poly-glycine (GGG) motif was also introduced into a PBTR1 vector containing Cc. Overlap extension PCR cloning was performed and PCR products were digested with DpnI to remove template DNA. DNA sequences were confirmed via Sanger sequencing.

5.2. Protein expression and purification

Escherichia coli BL21(DE3) cells were transformed with the CcP-X_n gene in a ppSUMO vector and grown at 37 °C in LB with 50 μg/mL kanamycin. When the O.D._600nm reached ~0.8, the cells were induced with 25 μg/mL isopropyl β-D-1-thiogalactopyranoside (IPTG) and expressed overnight at room temperature. Cells were pelleted via centrifugation and stored at −20 °C. Resuspended cell pellets were lysed via sonication and soluble protein was isolated through centrifugation. The lysate was then passed through a Ni-NTA column and His-SUMO tags were cleaved by ULP-1 (R3) protease. The cleaved protein was then purified on a Ni-NTA column and CcP was collected in the flow through. For FeCcP-X_n, purified apo-protein was then added to 1 molar equivalent of free heme dissolved in 100 mM NaOH and allowed to incorporate overnight at 4 °C. The reaction was neutralized with acetic acid and the precipitated protein and insoluble heme were removed via centrifugation. For Zn CcP-X_n, the apo-protein was added to 5 molar equivalents of free ZnP and 1,1’-carbonyldiimidazole (CDI) dissolved in DMF and stirred for 5 days at 4 °C. The soluble holo-proteins were further purified via size exclusion column and anion exchange chromatography. Purified protein with Abs_408nm/Abs_280nm ≥ 1 and A_432nm/A_280nm ≥ 2 were flash frozen and stored at −80 °C for FeCcP and ZnCcP variants respectively.

GGG-Cc was expressed from a PBPTR1 vector containing CcHL for covalent c-type heme incorporation in 6 L of BL21(DE3) and grown simultaneously with 1 L of BL21(DE3) cells transformed with the TEV plasmid. When O.D._600nm reached ~0.8 the cells expressing TEV were induced with 25 μg/mL IPTG and all proteins were expressed at 30 °C overnight. Cells were pelleted via centrifugation and stored at −20 °C. Resuspended cell pellets were lysed via sonication and cell detritus was spun down to isolate soluble protein. Lysate was then incubated at RT for 1 hr to allow the TEV protease to cleave Cc. The cleaved protein was then filtered and purified through cation exchange and size exclusion chromatography. The protein was flash frozen in liquid N₂ and store at −80 °C.

Sortase A was expressed in BL21(DE3) as described above for CcP-X_n. Cells were lysed via sonication and soluble protein was isolated by centrifugation. The His-tagged protein was purified from the lysate using a Ni-NTA column and flash frozen and stored at −80 °C.

5.3. Ligation of CcP and Cc via sortase A

Sortase A catalyzes the ligation of the N-terminal poly-glycine motif on Cc to the C-terminal LPATG motif on CcP (Scheme 3). Purified proteins were mixed in a 1:5:2–5 ratio of CcP-X_n: GGG-Cc: sortase A in 20 mM Tris (pH 8), 150 mM NaCl, and 5 mM CaCl₂. The mixture was then allowed to react for 1–7 days at 4, 17, or 20 °C. Precipitated protein was then removed via centrifugation and the soluble ligated product was purified using size exclusion and anion exchange chromatography. CcP-X_n-Cc ligation was confirmed by SDS-PAGE and mass spectrometry.

5.4. Cc oxidation kinetics in Fe CcP-LPATGGG-Cc

Cc in Fe CcP-LPATGGG-Cc was reduced with excess dithionite and the protein was buffer exchanged into 25 mM KP_i buffer, pH 7.5. To ensure that no remaining dithionite was affecting the observed reduction of CcP, equal-molar dithionite was added to CcP CpdI under conditions of the Cc reaction: no noticeable reduction was observed. Single turnover measurements were conducted with 2 μM FeCcP-LPATGGG-Fe(II) Cc and 2 μM H₂O₂ in 1800 μL of 25 mM KP_i buffer, pH 7.5 at 10 °C. Reactions were initiated by the addition of H₂O₂ to the reduced protein while stirring at 600 RPM. The rate of Fe(II) Cc oxidation was monitored at the heme Q bands Abs_{550 nm} - Abs_{540 nm} (ε = 19.2 mM⁻¹ cm⁻¹) and, Fe(IV)=O CcP at Abs_{434 nm} (a Fe(II/III) Cc isosbestic point).

5.5. Transient absorption spectroscopy of Zn CcP-X_n-Cc kinetics

To avoid triplet state quenching from ground state oxygen, ZnCcP-X_n-Cc samples were prepared anaerobically. The rate-constant for decay of photoexcited ³ZnP in the absence of oxidative quenching by Cc (k_D) was determined in two ways: 1) by monitoring the triplet state decay at 475 nm with CcP-X_n alone and 2) by monitoring the kinetics of the ligated products in the presence Cc reduced by dithionite (i.e. CcP-X_n-Fe(II) Cc). The decay rate constant for ³ZnP in the presence of oxidized Fe(III) Cc (k_p) reflects both k_D and the additional decay process resulting from ET to Cc (k_e). To determine k_e, Cc in Zn CcP-X_n-Cc was oxidized by excess K₃[Fe(CN)₆]. All samples were buffer exchanged into a degassed 100 mM KP_i buffer, pH 6. Samples were prepared in a 1 mm cuvette with a 4 μL drop containing 75 μM Zn CcP-X_n-Fe(III) Cc. Transient absorption (TA) measurements were performed as described previously [18]. Samples were excited at 560 nm by an Opotek Opolette Nd:YAG laser with 2 mJ of power per 8 ns pulse firing at 20 Hz. Absorption from a 75 W Xe-arc lamp probe light was measured using a Hamamatsu Photonic Multichannel Analyzer (PMA) and averaged 50 times. The time delays between the excitation pump source and probe light were controlled by a Digital Delay Generator DG355 (Stanford Research Systems) which was triggered by the laser Q-switch. Time delays were varied between 0 to 40 ms randomly to reduce effects of photobleaching. Difference spectra were calculated as ΔA = −log(excited/reference). Data were subjected to global analysis and singular value decomposition (SVD) was used to deconvolute the evolution-associated difference spectra in Glotoran [45].

5.6. Structural predictions for CcP-X_n-Cc proteins

Rosetta was used to model in the LPATGG loop sequence into the CcP:Cc complex with PDB: 6P41 as the starting model [33,34]. The loop modelling protocol was performed using the last 5–14 residues of CcP, the LPATGG sequence, and the first six residues of Cc [46,47]. Of the 135 – 500 generated models, the lowest energy models were chosen to run the relax protocol [48–51]. After running relax, 100 models were generated and the lowest energy models were selected.

The amino acid sequence for CcP-X_n-Cc ligated variants were input into the AlphaFold3 AlphaFold server to predict the effects flexible linker length have on the conformational state of the ligated complex [52]. The predicted local distance difference test (pLDDT) confidence scores for the predicted structure models were stored in the B-factor field of the output pdb files.

AFsample2 was carried out on the same sequences to predict other possible conformers in the ligated and un-ligated proteins. AFsample2 outputs both the predicted models as well as the confidence score for each individual model. AlphaFold2 is unable to incorporate ligands into the prediction so the distances between the Fe coordinating His in both CcP and Cc (His175 and His23 respectively) were used to measure “heme-to-heme” distances.

5.7. Small Angle X-ray Scattering (SAXS)

The CcP:Cc complex and ligated CcP-X_n-Cc protein variants were prepared in 10 mM Tris, pH 7 and 5% glycerol at 1–9 mg/mL concentrations for SAXS data collection. Batch and SEC SAXS data were collected at the Cornell High Energy Synchrotron Source (CHESS) ID7A1 beamline. Photon flux was 2.3 × 10¹² ph/s and scattering data was collected 0.005 Å⁻¹ ≤ q ≤ 0.7 Å⁻¹ on an EIGER 4M detector (Dectris). For SEC SAXS, samples were separated via size exclusion chromatography 4°C using a Superdex 200 5/150 column. Image integration, normalization, and subtraction was performed using the BioXTAS RAW 2 program [53]. The processed data were then input into the FoXS server along with the AlphaFold3 models to compute a theoretical scattering profile and compare it to the experimental profiles [54,55].

5.8. X-ray crystallography data collection and model building

To promote complex formation, purified CcP-GGSG(2x)LPATGGG-Cc was buffer exchanged into filtered nanopure water. The concentrated protein was then diluted to 0.5 mM and initial crystal hits were obtained using a Gryphon robot (Arts Robbins Instrument). Crystals were optimized via vapor diffusion in 4 μL sitting drops against a well solution containing 14–20 % PEG 8000, 40 mM KP_i (pH 3–5.5), and 20 % glycerol. Small, red crystals grew within a week. Prior to data collection, crystals were soaked briefly in a 20 % ethylene glycol cryoprotectant and flash frozen. Diffraction data were collected with The Northeastern Collaborative Access Team (NE-CAT) facility at the Advanced Photon Source (APS) on beamline 24-ID-E on an EIGER 16M pixel array detector. Diffraction data were indexed, integrated, scaled, and merged in RAPD2, which uses the CCP4 suite of programs [56]. The scaled sets were phased via molecular replacement with an AlphaFold3 generated model of the CcP-GGSG(2x)LPATGGG-Cc structure refined using Phaser and phenix.refine in Phenix, respectively [57]. Model building was performed in Coot [58].