Steady-state polypeptide transfer from the translocon to the membrane

In concert with irreversible non-equilibrium peptide translation by the ribosome, the nascent polypeptide chain may integrate into the membrane or translocate to the other side of the membrane, facilitated by the conserved protein translocation channel SecYEG in bacteria and Sec61 in eukaryotes. Assuming equilibrium for the decision processes yielded the biological hydrophobicity scale, reflecting free-energy differences ΔG between the pore interior and membrane. Yet kinetic effects and molecular dynamic simulations suggested that a nascent polypeptide could not sample the two separate environments a sufficient number of times for partitioning in equilibrium. Here we tested the hypothesis employing purified and reconstituted SecYEG harboring a stalled ribosome nascent chain (RNC). The SecYEG-RNC complex was open in a de-energized membrane, allowing ion flow. Application of a membrane potential closed the channel if nascent chain hydrophobicity permitted membrane integration. Taking the ratio of steady-state to initial ion conductances as a measure of nascent chain hydrophobicity, we found ΔG for KvAP’s voltage sensor (4th helix harboring four arginines) and FtsQ’s transmembrane helix to be equal to 0.3 and –2.1 kcal/mol, respectively. Thus, ΔG observed in our minimalistic system agrees very well with the position-dependent amino acid contribution of the biological hydrophobicity scale. Characteristic sampling times of ~2 s appear sufficient to reach a steady state for a ~20 amino acid-long segment invalidating the hypothesis of insufficient sampling.


Introduction
The processes of (i) protein insertion into the membrane, (ii) membrane-protein folding, (iii) protein-protein interactions within membranes, and (iv) membrane-protein conformational changes are interrelated (1,2). Despite their fundamental importance, many aspects of the underlying energetic expenditure thus far remained enigmatic (3)(4)(5). As a result, many different hydrophobicity scales have been proposed (6)(7)(8). The comparison of two specific amino acid hydrophobicity scales provides an example: Measurements of the apparent free energy ΔG of transferring an amino acid from the aqueous solution into a hydrophobic solvent (octanol) gave rise to the physicochemical hydrophobicity scale (9). It indicates insertion costs of ~3.6 kcal/mol for the most hydrophilic amino acid and a gain in free energy of 2 kcal/mol for the most hydrophobic amino acid. The so-called biological hydrophobicity scale spans a smaller range of only 4 kcal/mol. It was derived in dog pancreas rough microsomes by determining the membrane insertion probability of an alanine-leucine helix doped with the amino acid of interest in its middle. The most hydrophobic amino acid yielded a gain in energy of only ~0.6 kcal/mol (10). Most strikingly, the rank order in the two scales is not the same.
The difference between the two scales is mainly attributed to their unequal reference points: bulk water versus aqueous translocon interior.
The biological scale reflects the free energy ΔGb for the amino acid residue transfer from the translocon to the membrane (10). Predicting ΔGb is difficult, partly because the physical properties of the intraluminal water molecules and bulk water are different (11). The observation is consistent with the standard rule that water molecules regain their bulk properties starting from a distance of about five water molecules from the channel wall (12). Due to differences in the hydrophobicity of the pore lining residues, variable channel diameter, and membrane inhomogeneities, ΔGb is position-dependent (13). Moreover, ΔGb for the same amino acid also depends on the translocon, e.g., there are quantitative differences between Sec61-mediated insertion of transmembrane segments into the membrane of the endoplasmic reticulum in Saccharomyces cerevisiae and mammalian cells (14).
Moreover, mutations in the hydrophobic core of the Sec61 translocon alter ΔGb (15).
The endpoints also differ for the two hydrophobicity scales. The physicochemical scale reflects the free transfer energy, ΔGpc , into octanol (oil, hexadecane) rather than into the membrane interior.
This simplification is less of a problem than the mismatched starting points because it has long been established that ΔGpc for the transfer from water into the organic solvent scales with the energetic expense for membrane partitioning. The interrelation between ΔGpc and membrane permeability is known as Overton's rule. This rule has withstood the test of time (16). It allows for predicting membrane permeability, especially when substituting the one (oil) slab membrane model for more sophisticated multi-slab models (17)(18)(19).
Calculating ΔGpc from ΔGb and vice versa appears to be very difficult as it would require many corrections. The most important ones concern the specific amino acid location, the hydrophobicity of the organic solvent, and the free energy of the transfer from bulk water to the translocon. The example of the voltage sensor of a potassium channel, more precisely of KvAP's fourth transmembrane helix (TM4) shows the complexity of the situation. The structure of the KvAP channel promised an understanding of excitability; hence membrane insertion and movement of KvAP's voltage sensor got special attention (20,21). The initially proposed movement of its four gating charges through the hydrophobic membrane core (22) seemed to be at odds with the high Born energy required for the purpose (23).
Indeed, if calculated as the sum of the amino acids' individual (position-wise uncorrected) ΔGb values, one finds the free energy ΔGs required for the membrane integration of the 19 amino acid-long segment. ΔGs for TM4 insertion amounts to ~3.9 kcal/mol (10). Such ΔGs corresponds to an insertion probability P = 0.1%. In contrast, the physicochemical scale predicts P = 99.9 %. However, the same experiments that led to the biological hydrophobicity scale demonstrated a TM4 insertion probability P = 37% (corresponding to ΔGs~0.3 kcal/mol) into the membrane of the endoplasmic reticulum (24).
These experiments gave rise to the position-dependent ΔGb values (13), yet it is unknown whether the prokaryotic translocon processing KvAP agrees in its hydrophobicity assessment with eukaryotic Sec61.
Quantitative differences in membrane integration may also arise from variations in the folding propensities of adjacent sequences and chaperone binding (25). The observation raises the question of whether the presumed equilibrium, a pre-requisition for the ΔGs calculation from ΔGb values, is met during protein integration into the membrane. Moreover, the solvation energies per unit area for partitioning of aromatic side chains are a factor 2.5 smaller than found in the physicochemical scale's settings (biphasic water-oil), suggesting that translocon-to-membrane partitioning may be biased by other effects (26). Moreover, extensive molecular dynamics simulations raised doubts that the nascent chain has sufficient time to sample the different environments of the aqueous translocon pore and the membrane interior (27). They suggested that kinetic effects may govern the membrane insertion process.
Here we used an in vitro system that provides the nascent chain with ample time to probe the hydrophobicities of the membrane interior and translocon pore (28). To allow the stalled translocation intermediate to reach a steady state location, we reconstituted purified SecYEG-RNC (ribosomenascent chain) complexes into planar lipid bilayers. Movement of the stalled nascent chain out of the aqueous channel manifested as a loss of ion channel activity due to channel closure. Electrophysiology showed renewed ion channel activity when the nascent chain returned to the lumen of the bacterial translocon SecYEG.
We monitored the partitioning of two transmembrane helices (TM) with different hydrophobicities: the strongly hydrophobic TM of FtsQ and the weakly hydrophobic TM4 of KvAP. To quantify the impact of positive charges on the partitioning, we also varied TM4's number of arginine residues. The thus obtained ΔGs for KvAP's TM4 is in perfect agreement with the ΔGs from experiments in dog pancreas rough microsomes (24), implying that sampling limitations are unlikely to bias polypeptide membrane insertion.

Materials and Methods
Electrophysiological measurements. Ag/AgCl reference electrodes were immersed into the buffer solutions on both sides of the planar bilayers (Fig. 1A). The command electrode of the patch clamp amplifier (model EPC9, HEKA electronics, Germany) localized to the cis compartment, and the ground electrode to the trans compartment. The recording filter for the transmembrane current was a 4-pole Bessel with a -3 dB corner frequency of 0.1 kHz. The raw data were analyzed using the TAC software package (Bruxton Corporation, Seattle, WA). Gaussian filters of 12 Hz were applied to reduce noise.
The data were further processed using SigmaPlot (Systat Software Inc., San Jose, CA).
Biobeads SM2 (Biorad) were added to remove the excess detergent, and the resulting turbid suspension was pelleted at 100.000 × g. The resulting pellet was resuspended and extruded through a 100 nm filter. We used a mass ratio of protein to lipid of 1:100. Reconstitution efficiency was about 3 SecYEG complexes per vesicle, which was established using fluorescence correlation spectroscopy (34).
In brief, we labeled purified SecYEG prior to reconstitution with AlexaFluor647 (ThermoFisher Scientific) and compared the number of fluorescent particles in the sample containing proteoliposomes with the number of particles (micelles) after adding 2% octyl glucoside (Sigma) and 3% deoxy big CHAP (Anatrace) to the same sample and correcting for dilution (Fig. S3).
Locking the translocation intermediates in the reconstituted translocon. Solvent-depleted planar lipid bilayers were formed by folding lipid monolayers from E.coli polar lipid extract (Avanti Polar Lipids) in the ~150 µm wide aperture of a PTFE film (Goodfellow). Water uptake from the hypotonic compartment led to vesicle swelling and fusion. The resulting stepwise rise in membrane conductivity indicated the reconstitution of RNC-SecYEG complexes into the planar bilayer (Fig. 3A). In contrast, vesicles with closed channels, i.e., vesicles in which the RNC-SecYEG complex did not form, are unable to fuse with the planar bilayer (35,36).  (Table 1), a SecM stalling sequence (orange), and two affinity tags: TEV cleavage site (blue) for Western blot analysis and CBP.

SecYEG reconstitution into planar bilayers proceeded via osmotically induced lipid vesicle fusion
The linking amino acids are depicted in green (see the SI for the exact amino acid sequence).
In most of the experiments, we added 7 µM calmodulin (CM) to the hypotonic compartment to prevent nascent chain backsliding. With the calmodulin bound to the calmodulin-binding sequence at one side and SecM arrested in the ribosome on the other side, the nascent chain was locked. That is, neither retrograde nor forward movements of the nascent chains were possible. They had only one remaining degree of freedom: they could move in and out of the lateral gate. Control measurements confirmed that calmodulin did not cause SecYEG openings in the absence of RNCs.

Results
Voltage-driven SecYEG closure triggers nascent chain partitioning into the lipid bilayer from the stalled translocation intermediate.
We  After the ψ was switched to 0 mV, two of the intermediates reopened, as seen when ψ = -120 mV was reapplied (right part of the graph). B: Cumulative closing kinetics of SecYEG-RNC complexes with FtsQ-based RNC was built from individual traces comprising 35 translocation intermediates. We fitted the trace with a single-exponential function (dashed dotted line, see Table 2).
Fitting the conductance with a single exponential function yielded a time constant of 1.6 ± 0.1 s. That is, those of the 19 amino acid-comprising nascent chains that partitioned into the lipid bilayer either already had left the aqueous environment or made that decision at a rate of roughly 10 amino acids per second. Since the translocon closure occurs much faster than empty ribosomes are bound, the latter scenario is the most probable. The rate of 10 amino acids per second is remarkably close to the rate previously observed for nascent chain secretion (37). It is also close to the ribosome's translation rate of 5 -10 a.a./s at a high growth rate for yeasts (38) and 15 a.a./s in E. coli (39). Our observation indicates that the rate at which the nascent chain may sample the different environments of the translocation pore and the lipid bilayer suffices to integrate a transmembrane helix in thermodynamic equilibrium. Unfortunately, the possibility of backsliding does not allow us to judge whether the steady state reached in the experimental settings of Fig. 2 mirrors that in a living cell.

Reconstitution of locked SecYEG-RNC complexes
We performed a new series of experiments where we prevented backsliding by adding CM to the trans-compartment. As in the case without CM, channel activation enabled the osmotically induced fusion of SecYEG-RNC(FtsQ) containing vesicles with the planar bilayer (Fig. 3A). As the FCS experiments indicated the presence of three SecYEG channels per vesicle (Fig. S3), and vesicle fusion inserts all of them at once, we expected to see a correspondingly large increment in planar bilayer conductivity. In line with the prediction, we mostly found fusion events where the incremental current comprised from 2 (blue arrows) to 4 times (red arrow) the current through a unitary SecYEG pore (Fig.   3A). the nascent chain to partition into the lipid bilayer (Fig. 3B). Fitting the cumulative conductance curve with the single exponential yielded the characteristic time for channel closure (Fig. 3C). With 11.6 s it was nearly tenfold slower than in the absence of calmodulin. The observed calmodulin effect is in line with the observation that chaperon binding affects membrane helix integration (25). While chaperons with ATPase activity lower the hydrophobicity threshold for membrane integration (25), the binding of a passive ligand merely prolonged the time required to establish a steady state.
We defined the quasi-equilibrium probability Pqe as the ratio between the number Np of RNCs partitioning into the bilayer and the total number Ntotal of reconstituted SecYEG-RNC complexes: where I0 and I∞ are the steady-state ionic currents through the bilayer immediately after applying ψ (at time t = 0) and after the steady-state has been reached (at time t = ∞). Pqe for FtsQ was equal to about 93%. The related quantity PSS is equal to the ratio between the numbers of RNC partitioning into the bilayer and RNCs that did not. PSS allows calculating ΔGS: The resulting ∆ = -1.5 kcal/mol suggests that the segment is less hydrophobic as predicted by the position-dependent bio-scale (13) and by the physicochemical scale (9) ( and that is about 0.04, we arrive at the corrected ΔGs values given in Table 2 in parentheses. For moderately hydrophobic sequences, this correction is insignificant. Our electrophysiological approach allows monitoring how Pqe approaches its steady-state value. To quantify the process, we define the time-dependent partitioning coefficient Pdyn(t): where Pdyn(∞) = Pqe. I(t) is the current at a given time t. Pdyn significantly deviates from Pqe for short observation times.

The steady-state probability of SecYEG-RNC complex closure is voltage-dependent
We reconstituted the translocon-RNC complexes KvAP 3R, a mutant KvAP-voltage sensor helix in which one of the four arginines was replaced by an alanine. Large membrane potentials, ψ = -110 mV, led to rapid channel closure, indicating membrane integration of the nascent chain (Fig. 4). In   Table 1) from translocation intermediates at physiologically high (-110 mV), middle (-65 mV), and low (-15 mV) transmembrane potential ψ. The apparent equilibrium partitioning for each trace was calculated using Eq. 1. It was 0.8 for ψ = -110 mV, 0.6 for -65 mV, and 0.1 for -15 mV

Determing ΔGS in quasi-equilibrium
The third RNC construct contained the KvAP's wild-type sensor helix TM4. To distinguish the four arginines bearing wild-type segment from its variants, we labeled the nascent chain as KvAP 4R (see Table 1). To enable quantitative analysis, we added up all available individual traces (such as the traces in Fig. 5A) to yield a cumulative trace (Fig. 5B). Extracting from the transmembrane current (Eq. 3) allowed calculation of the characteristic partitioning time τ, according to Eq. 4: for the 19 amino acid-long segment was equal to 12.4 s. At high ψ values, the resulting Peq was equal to 37 % (Fig. 5A, B). Accordingly, ∆ amounted to 0.3 kcal/mol. This perfectly agrees with the value predicted by the position-dependent biological scale ( Table 2).
The charge effect on the partitioning probability.
From membrane integration experiments with microsomes (24), it is known that the removal of the central arginine has a profound effect on the partitioning of the TM4 segment of KvAP (see Table 2).
We compared constructs analogous to KvAP 4R-RNC, where one of the central arginines (KvAP 3R-RNC) or all arginines (KvAP 0R-RNC) were substituted with alanines (Fig. 5). The respective ∆ values amounted to -0.8 kcal/mol for KvAP 3R-RNC and -1.3 kcal/mol for KvAP 0R-RNC. The ∆ for the KvAP 3R sequence is in good agreement with the value predicted by the position-dependent bio-scale.
However, as in the case of FtsQ-RNC, our ∆ underestimates the energy gain in the case of strongly hydrophobic sequences (see Table 2).

Discussion
We determined the steady-state probability of polypeptide membrane integration in Our conclusion is also in line with the observed insertion kinetics. For FtsQ's transmembrane segment, we found an insertion rate of 0.2 amino acids per second. This rate is comparable with previously observed secretion rates and measured amino acid translation rates at the ribosome.
Longer integration times recorded upon restricting polypeptide movement by allowing calmodulin to bind to a tag on the nascent chain in the receiving compartment only reflect an artificially increased activation barrier for membrane integration. It is very likely that the large proteinaceous anchor decreased polypeptide mobility. In consequence, every move of the segment between the translocon and the membrane may take more time. Thus, the nascent chain bound to calmodulin could take more time to perform the minimum 100 sampling motions that the molecular dynamics simulations suggest it would need to approach the correct equilibrium probability of membrane integration (27).
Our calmodulin binding experiment displays interesting parallels with chaperone binding experiments carried out in yeast (25). In contrast to our reconstitution experiments, the yeast assay did not allow determining integration rates, but it revealed changes in integration probability depending on (i) chaperone binding in the receiving compartment or (ii) misfolding of upstream segments. Both modifications altered the environmental hydrophobicity probed by the nascent chain.
In case (i), the chaperone removed a potential interaction partner for the emerging transmembrane segment, and in case (ii), the misfolded structure presented a hydrophobic surface capable of attracting the hydrophobic segment otherwise destined for membrane integration. Thus, this study does not contradict our conclusion that thermodynamics instead of kinetic factors govern membrane integration.
Our electrophysiological assay may underestimate ΔGs for hydrophobic sequences (FtsQ-RNC and KvAP 0R-RNC), i.e., the position-dependent biological scale predicts higher ΔGs_b values (see Table   2). The deviation would be caused by the leakage current passing through an uncertain number of closed intermediates. However, the smaller hydrophobicity range may also be due to a compression of the hydrophobicity scale caused by the prokaryotic translocon relative to the mammalian translocon. Such alterations in translocation thermodynamics are known from yeast (14). Moreover, they would reflect corresponding changes in protein folding thermodynamics (41).
Our results show quasi-equilibrium partitioning of TM segments in a minimal system where SecYEG lacks accessory proteins. They are reasonably close to the prediction by the positiondependent biological hydrophobicity scale obtained in dog pancreas rough microsomes (13). We thus demonstrate that TM insertion in vivo may well satisfy thermodynamic equilibrium.
Funding: This work was supported by grants of the Austrian Science Fund (FWF): P 34584 to PP, P29841 to DGK.
Competing Interests: The authors declare no competing interests.