Structural Origins of FRET-Observed Nascent Chain Compaction on

Subscriber access provided by Iowa State University | Library

Article

Structural Origins of FRET-Observed Nascent Chain Compaction on the Ribosome Daniel A. Nissley, and Edward P. O'Brien J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.8b07726 • Publication Date (Web): 28 Sep 2018 Downloaded from http://pubs.acs.org on September 30, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 23 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Structural Origins of FRET-Observed Nascent Chain Compaction on the Ribosome Daniel A. Nissley1 and Edward P. O’Brien1,2,3* of Chemistry, Penn State University, University Park, PA 16802, USA 2Bioinformatics and Genomics Graduate Program, The Huck Institutes of the Life Sciences, Penn State University, University Park, PA 16802, USA 3Institute for CyberScience, Penn State University, University Park, PA 16802, USA 1Department

*To whom correspondence should be addressed: [email protected] Abstract A fluorescence signal arising from a FRET process was used to monitor conformational changes of a domain within the E. coli protein HemK during its synthesis by the ribosome. An increase in fluorescence was observed to begin 10 s after translation was initiated, indicating the domain became more compact in size. Since fluorescence only reports a single value at each time point it contains very little information about the structural ensemble that gives rise to it. Here, we supplement this experimental information with coarse-grained simulations that describe protein conformations and transitions at a spatial resolution of 3.8 Å. We use these simulations to test three hypotheses that might explain the cause of domain compaction: (1) that poor solvent quality conditions drive the unfolded state to compact, (2) that a change in the dimension of the space the domain occupies upon moving outside the exit tunnel causes compaction, or (3) that domain folding causes compaction. We find that partial folding and dimensional collapse are both consistent with the experimental data while poor-solvent collapse is inconsistent. We identify alternative dye labeling positions on HemK that upon fluorescence can differentiate between the domain folding and dimensional collapse mechanisms. Partial folding of domains has been observed in C-terminally truncated forms of proteins. Therefore, it is likely that the experimentally observed compact state is a partially folded intermediate consisting, according to our simulations, of the first three helices of the HemK N-terminal domain adopting a native, tertiary configuration. With these simulations we also identify the possible co-translational folding pathways of HemK.

1

ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 23

Introduction Förster resonance energy transfer (FRET) has recently been used to probe the conformations of nascent proteins on both translationally arrested1 and continuously translating2 ribosomes. The efficiency of energy transfer from the FRET donor dye to the FRET acceptor dye is related to the distance between them, providing a means of monitoring protein conformational states3. In one such study Rodnina and co-workers2 used a FRET-based assay, which measures acceptor dye fluorescence but not FRET efficiency, to monitor the synthesis of the N-terminal domain (NTD) of the E. coli N5-glutamine methyltransferase protein HemK. The HemK NTD is 73 residues in length and contains five helices that form a bundle in the native state (Fig. 1A). In these experiments the HemK NTD was fused to a 39-residue C-terminal segment composed of the inter-domain linker and a small portion of the HemK C-terminal domain (CTD). A FRET donor/acceptor pair was incorporated into the HemK NTD via chemically modified amino acids at positions 1 and 34 (Figs. 1B and C). Time series of acceptor fluorescence due to FRET during continuous synthesis were obtained for six HemK constructs by monitoring the fluorescence in the FRET acceptor channel in a stopped-flow apparatus. The acceptor fluorescence began to increase after 10 s of synthesis, indicating that the FRET dyes moved closer together due to compaction of the domain. These data, in combination with limited proteolysis assays and photoinduced electron transfer experiments, were used to support the hypothesis that HemK NTD co-translationally folds through a compact state. Subsequent fitting to a chemical kinetic model suggested that partially folded structures can transiently form during synthesis, although alternative mechanism such as the formation of a molten globule or a non-specific collapsed state was not ruled out4. The observed compact state could arise due to partial domain folding5 or collapse of the unfolded state in either a good or poor solvent. The ribosome exit tunnel is about 100 Å in length and has an average diameter of 15 Å, which is similar to the 10 Å persistence length of unstructured proteins6–8. For these reasons, nascent protein segments are confined in an effectively 1-dimensional tube inside the tunnel and translocate to a 3-dimensional space outside of it. Polymers tend to collapse when there is an increase in the space dimension in which they exist9–11. The radius of gyration, , of the unfolded state of a -monomer chain in a good solvent scales with the space dimension, , as the exit tunnel

’s scaling switches from

∝

. Thus, as a domain moves from inside to outside (

1) to

(

3), such that the unfolded state of

the 73-residue HemK NTD would compact in size by or 82%. Indeed, such unstructured collapse has been seen in simulations of other ribosome-nascent chain complexes12. For a domain in a poor solvent,

∝

, and so

will go from

1 to

3 when the

or 94%. Such collapse would domain enters the cytosol, such that HemK NTD will collapse by bring the fluorescent dyes in HemK closer together and therefore could explain the increased FRET fluorescence. Here, we present results from coarse-grained, low-friction Langevin dynamics simulations of HemK synthesis that were performed with three different Hamiltonians to assess which driving force yields a compaction process that is most consistent with the experimental data. The first Hamiltonian, which models domain folding, is a Gō-based model previously used to model cotranslational folding in silico13–15. The second Hamiltonian approximates the influence of good solvent on the nascent protein to model dimensional collapse. The third Hamiltonian models the 2


Page 3 of 23 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


influence of a poor solvent to permit non-specific collapse. With this approach we are able to provide a molecular interpretation of FRET measurements on ribosome-nascent chain complexes, suggest optimal dye positions for future experiments, and determine the cotranslational folding pathway of HemK NTD. Methods Construction of HemK N-terminal domain folding model. A structure-based Gō model of the HemK N-terminal domain (residues 2-73) was constructed from PDB ID: 1T43 in a manner previously described16. Each amino acid was represented as a single spherical interaction site with its center of mass at the location of the Cα atom in the crystal structure. Transferable bond, angle, and dihedral terms were used. Native contacts were defined based on the PDB ID: 1T43 crystal structure. The ETEN potential was used for all simulations. The stability of this HemK NTD Gō model was tuned with replica exchange simulations in the molecular dynamics package CHARMM17 to reproduce the experimental value of the folded state stability at 298 K within error (Experiment: 4.66 0.10 kcal/mol, Simulation: 4.58 0.11 kcal/mol) when Lennard-Jones well depths were globally increased by a factor of 2.2. A total of 50,000 exchanges were attempted between replicates, the first 5,000 discarded, and the weighted histogram analysis method18 used for analysis. The root-mean-square deviation (RMSD) from the native state was calculated for the five helices of 1T43 as identified by STRIDE. The HemK NTD Gō model was defined as folded when the RMSD was 7 Å in comparison to the native state. Construction of dye-modified HemK model. Explicit coarse-grained representations of the FRET dyes were constructed2. Structures of BodipyFL (BOF, Life Technologies D6140) and Bodipy 576/589 (BOP, Life Technologies D2225) and their linkers were built in Gaussian/g09d01 and their geometry optimized with B3LYP/6-31G[19]. These structures include all atoms of the dyes and linkers with the Cα position represented by a –CH3 group. The coordinates of the optimized structures were used to generate coarse-grained representations at a similar level as has been done previously for FRET dyes20. The approximate locations of the coarse-grain centers are displayed in Fig. 1C. Each of the rings was reduced to a single interaction site at the centroid of the ring’s heavy atoms. Heavy atoms in two rings were included in the calculation of both ring’s centroids. Linker interaction sites were placed at bond midpoints. The total mass of the dye, linker, and Cα beads was redistributed evenly between all of the beads constituting each dye-modified amino acid representation. BOF-Met and BOP-Lys beads were thus assigned masses of 67.666667 AMU and 48.777778 AMU, respectively, such that the total mass of each dye-modified amino acid was conserved. A full accounting of the parameters chosen for the BOF-Met and BOP-Lys is provided in Tables S1-S5. Bond lengths were extracted by measuring the distances between the coarsegrain beads in the geometry-optimized structures. Force constants of 50.0 kcal/mol*Å2 were used for all dye and linker bonds. Bond angles were extracted from the B3LYP/6-31G-optimized structures, and all force constants were taken as 30 kcal/mol*degree2. A double-well angle potential21 was used for all angles in the linker. Bond angles formed by adjacent nascent chain beads, the dye-modified Cα beads, and the first bead of the linker (e.g., the L1-A1-A2 bond angle for BOF-Met) were assigned the average Cα-Cα-Cβ bond angle from an ALA-ALA-ALA tripeptide 3



Page 4 of 23

constructed in CHARMM in a right-handed helical conformation. Gly-Gly dihedral terms were used for the linkers. Single-well dihedral potentials were used for the dye dihedrals, with the minimum energy position located at the value of the dihedral calculated from the B3LYP/6-31G-optimized /2 values for the linker beads were taken from Merchant et al. 2007[20]. The structures. The /2 of the ribose interaction site was used for each of the five-membered rings, while the /2 used for six-membered rings by Merchant et al. 2007 was used for the six-membered rings here. The dye and linker representations interact in a purely repulsive manner with one another, the rest of the protein representation, and the ribosome representation. As the PDB ID: 1T43 structure does not contain an N-terminal Met residue the BOF-Met coarse-grain representation was appended to the N-terminus of the HemK NTD model. A structured linker, consisting of the next 39 residues in HemK’s sequence, was appended to the C-terminus of the N-terminal domain model as was done in the original experiments. This coarse-grain structure was used for all continuous-synthesis trajectories and is referred to in following Methods sections as the HemK NTD Gō-dye model. Construction of good- and poor-solvent collapse Hamiltonians. The folding Hamiltonian was modified to approximate the influence of good solvent by removing all intra-NTD contacts while preserving inter-NTD/linker and inter-NTD/CTD contacts. Similarly, the poor-solvent Hamiltonian was constructed by first removing all intra-NTD contacts while preserving inter-NTD/linker and inter-NTD/CTD contacts and then adding a non-specific pairwise attractive term. The functional form of the potential for these attractive interactions is taken to be the same as for contacts within the folding Hamiltonian. The well-depth for these added interactions was tuned with simulations of the HemK Gō–dye model NTD in bulk solution with well-depths of ε = {0.0, -0.2, -0.4, …, -19.6, -19.8, -20.0}, in units of kcal/mol. Ten trajectories were run for 750 ns at each well-depth and system coordinates recorded every 1.5 ns. The first 150 ns were discarded and the mean radius of gyration calculated over the remaining 600 ns of each simulation. The resulting plot of mean radius of gyration as a function of the well-depth is shown in Fig. S1. The well-depth corresponding to the midpoint of the collapse transition, -3.4 kcal/mol, was used in our continuous-synthesis of 6.17 Å, which is the median of all the intra-NTD contacts for the folding simulations. An Hamiltonian, was used for these non-specific interactions. Selection of mean in silico translation elongation time. The use of low-friction Langevin Dynamics with a coarse-grain model greatly accelerates dynamics. Because the relative timescales of folding and amino acid addition are critical to accurately capturing co-translational 〉, as folding behavior, we calculated the in silico timescale of amino acid addition, 〈 〈 where 〈

〉 and 〈

〉

〈

〉∗

〈

〉

〈

〉

,

[1]

〉 are the experimental amino acid addition and folding times, respectively,

〉 is the simulation folding time. This equation ensures that the experimental ratio of and 〈 〉 was determined with temperature quenching timescales is maintained. The value of 〈 simulations. The HemK NTD Gō-dye model was first equilibrated at 800 K for 40 ns to ensure complete unfolding and then instantaneously cooled to 298 K for 200 ns. Three hundred and fifty trajectories were run with this protocol and the time-dependent survival probability of the unfolded state calculated. The HemK NTD Gō-dye model was considered to have folded during the quench 4


Page 5 of 23 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


at 298 K if the RMSD of the residues within its five helical elements was 5 Å. The survival probability curve was fit with a single-exponential function of the form exp , where is the survival probability of the unfolded state, is the time since the start of the quenching period, and is a fit parameter representing the rate of folding. The value of 〈 〉 was determined with curve-fitting in python to be 0.281 ns-1 (Pearson of fit: 0.987), corresponding to a mean 〉 3.56 ns. Holtkamp et al. 2015 determined that 〈 〉 0.278 s and folding time of 〈 〈 〉 1.95 10 s; with 〈 〉 3.56 ns, Eq. [1] gives 〈 〉 5080 ns. Simulating the synthesis of the HemK constructs with an explicit representation of the E. coli 50S ribosome is computationally intractable for a dwell time of this duration. To reduce the required computational time, we applied an acceleration factor, , such that 〈

〉

〈

〉

∗

〈

〉

〈

〉

∗〈

〉.

[2]

As increases the simulation mean dwell time decreases linearly. A factor of speed up is 〉≪〈 〉 , such that the simulation is in a quasi-equilibrium regime. We used reasonable if 〈 Gillespie Algorithm simulations of two-state co-translational folding to determine the highest value of that does not significantly alter the behavior of the system. In these simulations, the HemK NTD was only permitted to fold once the entire domain and an additional 30 residues, to simulate the effect of the exit tunnel, were synthesized. An initial set of 10,000 trajectories with the experimental rates of folding, unfolding, and amino acid addition were run for comparison. Additional sets of simulations were then run with the in silico rates calculated from temperature quenching with 1, 10, 100, 200, 500, 1000, 10000 (see Fig. S2). These simulations suggest that an of up to ~1,000 is reasonable, consistent with the notion that significant departure from 〉≅〈 〉, and there is ~1400-fold quasi-equilibrium behavior is only expected when 〈 〉 and 〈 〉. We chose 500 as a compromise between computational difference between 〈 expense and preservation of the experimental ratio of timescales as closely as possible. Thus, 〈

〉

10.2 ns was used for all continuous synthesis simulations.

Continuous synthesis simulations. A coarse-grained representation of the E. coli ribosome’s large subunit was constructed from PDB ID: 3UOS as described previously15. A partial sphere of 26 ribosomal interaction sites within 12 Å of the point (6, 0, 0) and with x-coordinate greater than 3 Å were harmonically restrained using CONS HARM with a force constant of 0.5 kcal/mol*Å2. A single ribosome interaction site corresponding to U2585’s uracil ring (by PDB ID: 3UOS numbering) was deleted from the PDB structure in order to avoid steric clashes with the nascent chain. All interactions between ribosome sites were deactivated with the BLOCK module in CHARMM. The C-terminal bead of the nascent chain was held at the point (6, 0, 0) by a spherical harmonic restraint with a force constant of 50 kcal/mol*Å2. Five planar restraints were positioned on five sides around the nascent chain using the GEO PLANE functionality in the MMFP module of CHARMM for nascent chain lengths up to and including 15 residues to direct the nascent chain into the exit tunnel. These planar restraints use the potential form EXPONENTIAL and parameters FORCE 50, DROFF 2.0, and P1 0.050. One yz-plane, passing through the point (1, 0, 0), was used to approximate the steric bulk of the P-site tRNA. Four additional planes were used to define a box around the nascent chain, open only on the face pointing out of the exit tunnel, to help guide it into the exit tunnel correctly. Two xz-planes, with one passing through the point (0, 20, 0) and 5



Page 6 of 23

the other through the point (0, -10, 0), were used. Two xy-planes, one passing through the point (0, 0, 20) and the other through the point (0, 0, -20), were also used. For nascent chain lengths greater than 15 residues only the yz-plane through (1, 0, 0) was retained to approximate the steric bulk of the tRNA. This combination of planar and harmonic restraints prevents early termination due to steric clashes that lead to SHAKE algorithm errors. A modified version of a previously published continuous synthesis protocol for CHARMM was used22. Simulations were begun with the Cα bead of the N-terminal residue (BOF-Met) restrained at the point (6, 0, 0). The dwell time at each nascent chain length was randomly 〉 10.2 ns. A temperature of 310 sampled from a single-exponential distribution with mean 〈 -1 K, frictional coefficient of FBETA = 0.050 ps , an integration time step of 0.015 ps, and the SHAKE algorithm were used for all continuous synthesis simulations. System coordinates were saved every 5000 integration time steps (75 ps). Two-hundred, one-hundred, and eighty synthesis trajectories were completed using the folding, good-solvent, and poor-solvent Hamiltonians, respectively. Each poor-solvent Hamiltonian trajectory was initiated from a different configuration from the folding Hamiltonian simulations at nascent chain length 33. Starting at a nascent chain length of 33 ensures that the nascent chain will be equilibrated by the time the BOP-Lys residue is added at position 34 and FRET becomes possible. The time series for continuous synthesis trajectories synthesized with the poor-solvent Hamiltonian were adjusted by appending thirty-two dwell times selected randomly from an exponential distribution with mean of 10.2 ns. This adjustment accounts for the time that would be required to synthesize the first 32 residues of the nascent chain that were not explicitly simulated with Langevin dynamics. The size of the harmonically restrained region of ribosomal interaction sites was also increased to a total of 95 ribosome interaction sites within the box bounded by [3, 30], [-15,10], and [-15,10]. The poorsolvent simulations were otherwise identical to the folding and good-solvent simulations. Trajectories for HemK98, HemK84, HemK70, HemK56, and HemK42 with each of the three Hamiltonians were extended by an appropriate amount to bring their total duration to ~35 s when mapped to experimental time. Mapping of simulation timescales. In order to compare our simulation results to experimental time series we mapped our simulation time course to experimental time. We assume that there exists a uniform scaling factor, , that maps the timescale of our simulations to experimental time: ∗ . [3] If we assume that

〈

〉 and

〉, we have

〈 〈

〉

〈

〉

.

[4]

As we employed an acceleration factor, , this equation must be adjusted to the form 〈

〉

〈

〉

〈

〉

〈

〉

.

[5]

Thus, we mapped our simulated time series in units of nanoseconds onto the experimental time regime in units of seconds by the procedure in which

has units of ns and both 〈

∗ 〉 and 〈

〈

∗〈

〉 〉

,

〉′ have units of s.

6


[6]

Page 7 of 23 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Calculation of

and

ensemble average time series. The Förster equation was used to

calculate the FRET efficiency as a function of the inter-dye distance, , according to Equation 1. A Förster radius, , of 54 Å was used19. The distance between the donor and acceptor was calculated as the distance between the interaction sites representing the six-membered rings in BOF and BOP at each frame of a trajectory. The FRET efficiency time series for each trajectory was then time averaged into 15-ns bins. Finally, the binned trajectories were averaged together to produce the ensemble average FRET time series. Statistics were generated by determining the standard error of the average for each bin across all trajectories and then calculating the corresponding 95% confidence interval. Time series of

were produced by dividing the

ensemble average time series for each construct by the value in the final bin for the HemK112 ensemble average time series for the corresponding Hamiltonian. These ensemble average trajectories were then projected onto experimental time by multiplication by the time adjustment factor, . Calculation of

and

time series for alternative dye positions was carried out in the

same fashion except the inter-dye distance was taken to be the distance between the Cα interaction sites for the dye locations under consideration. Fraction of native contacts analysis. The fraction of native contacts formed by the five helices within the HemK Gō-dye model was calculated from the folding Hamiltonian continuous synthesis simulations. First, the set of contacts within the coarse-grain, native-state reference structure was determined. Two interaction sites are considered to share a contact if two criteria are satisfied: (1) both residues are within structured regions and separated by at least three residues in the primary sequence (i.e., → 4 or greater separation) and (2) the coarse-grain residues in the reference structure are no more than 8 Å apart. A contact was considered to be formed during a continuous synthesis trajectory if it satisfied criteria (1) above and the distance , where is the distance between the between the residues was less than or equal to 1.2 ∗ residues in the reference structure and the factor of 1.2 is to adjust for thermal fluctuations. The total number of contacts formed at frame of the simulation, , was divided by the total number of contacts formed in the native state, , to produce the value of . Time series of were then binned, averaged, and the corresponding statistics determined in the same manner as described for the

and

time series.

Comparing simulated and experimental time series. Pearson between simulated

or

time series and experimental

values were calculated

time series. The experimental

time point temporally closest to each of the simulation time point (after appropriate mapping of the simulated time series to the experimental time frame of reference) was determined and the calculated between this set of experimental values and the simulation time series. All Pearson -values are

Structural Origins of FRET-Observed Nascent Chain Compaction on

Recommend Documents