In-cell architecture of an actively transcribing-translating expressome

Abstract

Structural biology performed inside cells can capture molecular machines in action within their native context. Here we developed an integrative in-cell structural approach using the genome-reduced human pathogen Mycoplasma pneumoniae. We combined whole-cell crosslinking mass spectrometry, cellular cryo-electron tomography, and integrative modeling to determine an in-cell architecture of a transcribing and translating expressome at sub-nanometer resolution. The expressome comprises RNA polymerase (RNAP), the ribosome, and the transcription elongation factors NusG and NusA. We pinpointed NusA at the interface between a NusG-bound elongating RNAP and the ribosome, and propose it can mediate transcription-translation coupling. Translation inhibition dissociated the expressome, whereas transcription inhibition stalled and rearranged it. Thus, the active expressome architecture requires both translation and transcription elongation within the cell.

The two fundamental processes of gene expression, transcription and translation, are functionally coupled in bacteria. While the transcribing RNA polymerase (RNAP) produces a nascent mRNA chain that can be directly translated by ribosomes (1–3), translation was shown to influence the overall transcription rate in Escherichia coli (4, 5), implying a physical link between the two processes. Accordingly, an in vitro-reconstituted E. coli RNAP-ribosome supercomplex structure was determined and termed “expressome” (6). Additional in vitro reconstitutions, some including the essential factor NusG (7) that is proposed to link RNAP and the ribosome (8, 9), reveal different structural arrangements of the supercomplex (10–12). These raise questions as to the mechanisms of coupling that could be utilized inside the cell, in the context of all regulatory factors.

To structurally analyze transcription-translation coupling inside cells, we combined in-cell crosslinking mass spectrometry (CLMS) (13) and cellular cryo-electron tomography (cryo-ET) (14). We used the small genome-reduced bacterium Mycoplasma pneumoniae, which is an ideal cell model for system-wide structural studies (15). While M. pneumoniae has undergone significant genome reduction during its evolution as a human pathogen, it has retained the core transcription and translation machineries (16–18).

To assess the topology of a putative RNAP-ribosome supercomplex and its associated regulatory factors, we performed whole-cell CLMS of intact M. pneumoniae cells (13, 19) (fig. S1 and table S1). We identified 10,552 crosslinks involving the same protein (self-links) and 1957 heteromeric crosslinks with a 5% residue-pair false discovery rate (FDR). These represented 577 distinct protein-protein-interactions (PPIs) at a 5% PPI-FDR (Fig. 1A and supplementary text). Identified crosslinks covered 83% of the detectable proteome (table S2 and fig. S1), including PPIs of membrane proteins (41% of PPIs), 76 uncharacterized proteins, the ribosome, RNAP and their associated factors (Fig. 1B and fig. S1-S3).

Fig. 1. Crosslink-based protein interaction map of M. pneumoniae proteome.

(A) 577 distinct PPIs identified at 5% PPI-level FDR (interactions to 8 abundant glycolytic enzymes and chaperones are removed for clarity). Membrane-associated proteins are shown in grey. Circle diameter indicates relative protein size. Blue: 50S ribosomal proteins; yellow: 30S ribosomal proteins; green: RNAP; orange: NusA. Each edge represents one or more crosslinks. (B) Interactors of RNAP and NusA. NusA NTD, S1, KH domains, and proline rich region (PR) are annotated. Line thickness represents the number of identified crosslinks.

The M. pneumoniae RNAP core consisting of the conserved subunits α, β and β’, was found to interact with the known auxiliary factors SigA, GreA, NusG, NusA, SpxA, and RpoE (firmicute-specific RNAP δ subunit) (20) (Fig. 1B). Additionally, two uncharacterized essential proteins, MPN555 and MPN530 (21), were found and the interaction of MPN530 with β/β’ subunits was independently validated by a bacterial two-hybrid screen (fig. S4). Despite these interactions, no direct crosslinks between the RNAP core and the ribosome were identified. Interaction between NusG and the ribosomal protein S10, previously reported in E. coli, was also not detected (8, 9). Instead, NusA, an essential transcription factor involved in elongation, termination and antitermination (21–23), was found to interact with RNAP via its N-terminal domain (NTD), and with the mRNA entry site of the ribosome via its C-terminal region (Fig. 1B). In-cell CLMS thus indicated an unexpected architecture in which RNAP and the ribosome are linked by NusA.

To investigate the structure of this potential association, cryo-ET data were acquired on unperturbed, frozen-hydrated M. pneumoniae cells (19) (Fig. 2A and fig. S5). 108,501 ribosome sub-tomograms were extracted and subjected to classification and refinement (fig. S6, S7). These exhibited large structural heterogeneity and were first sorted into classes representing the 50S subunit (30.3%), 70S ribosomes (53.3%) and 70S ribosomes in the closely-assembled polysome configuration (24) (16.4%) (Fig. 2A and fig. S7). Subjecting 73,858 70S ribosomes to a new subtomogram analysis workflow (fig. S6) (19) resulted in a 5.6 Å ribosome density (Fig. 2B, fig. S8). A ribosome homology model (based on PDB 3J9W) was fitted and the majority of M. pneumoniae ribosomal proteins could be mapped (fig. S9). Helical densities at the C-termini of L22 and L29 (Fig. 2B, insert) were unaccounted for by the homology model and assigned to two C-terminal extensions that are unique to M. pneumoniae and its close relatives (fig. S10 and S11). L23, which also contains a C-terminal extension, was predicted to be unstructured and did not produce any discernible density in the map (fig. S12). Therefore, in conjunction with CLMS, the attained high-resolution map enabled de novo assignment of secondary structures in cellulo.

Fig. 2. In-cell cryo-ET reveals the presence of an RNAP-ribosome supercomplex.

(A) Left: tomographic slice of a M. pneumoniae cell. Right: classification of 108,501 ribosome sub-tomograms from M. pneumoniae cells. (B) Left: 5.6 Å in-cell 70S ribosome density. Insert: density near the peptide exit tunnel (dashed circle) shows two helices not accounted for by the fitted homology model (L22 and L29). (C) 9.2 Å in-cell structure of RNAP-ribosome supercomplexes (2.8% in (A)), fitted with homology models. Arrowheads indicate remaining unassigned density.

Focused classification of the 70S on the mRNA entry site identified a ribosome class in complex with RNAP (70S+RNAP, Fig. 2A and fig. S7). Refinement thereof provided a 9.2 Å map (fig. S13) into which the ribosome and RNAP models fitted unambiguously (Fig. 2C). Consistent with the CLMS data, the map contained additional density at the interface between the two complexes (Fig. 2C, arrowheads), which was further resolved by multi-body refinement (fig. S14). The path of the DNA and RNA-DNA hybrid duplex showed that RNAP is in an elongating state (Fig. 3A and fig. S15C). The existence of elongating RNAP and 70S ribosome demonstrated that the supercomplex represents an actively elongating expressome with a large degree of structural flexibility (Movie S1).

Fig. 3. Integrative model of the M. pneumoniae elongating expressome.

(A) Cryo-EM density of RNAP corresponding to the DNA, RNA-DNA hybrid, upstream β’ clamp, δ subunit and NusG, accommodates one turn of the RNA-DNA hybrid consistent with an elongating RNAP. (B) Integrative model with cryo-EM density of the RNAP-NusG-NusA-ribosome elongating expressome. Structured regions are represented as colored cartoons. The electron density colors represent the subunits occupying the corresponding volumes. Coarse-grained regions are not shown. Schematic of the putative mRNA path refers to the shortest distance between mRNA exit and entry sites. (C) The mRNA exit tunnel face of RNAP is covered by NusA. Localization probability densities for NusA domains are shown in orange. Crosslinks between NusA and other proteins are shown. Satisfied crosslinks (<35 Å) are in blue, overlong crosslinks are in red.

Both CLMS and cryo-EM results showed binding of NusG to its conserved site (fig. S15, S16) (25). M. pneumoniae NusG contains large inserts of unknown structure, but retains the residues involved in the NusG-S10 interaction (8). However, the arrangement of RNAP relative to the ribosome placed NusG away from S10, indicating that this interaction does not occur in the elongating expressome (fig. S16). All other proteins found interacting with RNAP by CLMS did not fit in the elongating expressome density (fig. S17). The remaining density between RNAP and the ribosome was therefore consistent with NusA (Fig. 1B and fig. S18).

The CLMS and cryo-EM data were used to derive an integrative model of the elongating expressome (26, 19) (fig. S19, table S4 and S5). M. pneumoniae NusA contains a disordered proline-rich C-terminal region that is not found in E. coli or B. subtilis (fig. S18), which we established to be essential by mutation experiments (fig. S20). This region, which was found to be crosslinked to multiple 30S ribosomal proteins (Fig.1B), was coarse-grained and not fitted into the density. The best scoring solutions (fig. S21) showed that the NusA NTD and S1 domain bind RNAP similarly to the E. coli paused elongation complex (23), with the S1 domain near the RNAP mRNA exit tunnel (Fig. 3B,C and fig. S22). The two KH domains were positioned near S3, S4 and S5 at the ribosome mRNA entry site. The orientation of KH domains retained their RNA-binding interface in a position that can interact with the nascent mRNA (27) (fig. S23). The C-terminal domains (CTDs) of RNAP α subunits were found to be in a wide range of conformations localized between NusA NTD and the second KH domain, with one α-CTD fitting a region between the second NusA KH domain and the RNAP core (fig. S23 and supplementary text). Additionally, the firmicute-specific RNAP δ subunit was positioned below RNAP β’ CTD (fig. S24), consistent with its suggested role in regulating RNAP-DNA interactions (20).

The integrative model demonstrated that NusA bridges the elongating RNAP and ribosome in the active expressome. To determine whether this architecture requires active translation elongation, we collected cryo-ET data on cells treated with the translation inhibitor chloramphenicol (Cm). The percentage of 70S ribosomes increased dramatically compared to untreated cells (Fig. 4A and fig. S25). The resulting 6.5 Å ribosome density (Fig. 4B) had well-resolved A and P site tRNAs similar to a previous ribosome-Cm structure (28) (Fig. 4C), but do not contain any RNAP density near the ribosome mRNA entry site. Thus, stalling ribosomes led to dissociation of the expressome.

Fig. 4. Stalling translation or transcription alters the expressome architecture in cells.

(A) Classification of sub-tomograms in untreated, Cm- and PUM-treated M. pneumoniae cells revealed shifts in ribosome populations following perturbations. (B) Models of RNAP-ribosome supercomplexes and the Cm-stalled ribosome. Left: in untreated cells, the expressome compromises an actively elongating RNAP and ribosome. Center: Cm decoupled the ribosome and RNAP. Right: in PUM-treated cells, the ribosome encounters the stalled RNAP. (C) Ribosome tRNA occupancy states. In untreated cells, densities for P-site tRNA and elongation factors densities were visible, indicating a translating ribosome. Upon addition of Cm, A and P site tRNAs were observed indicating a stalled ribosome. In PUM-treated cells, presence of EF-G and hybrid A/P* and P/E site tRNAs suggested a pre-translocation stalled state.

The dependence on active transcription was probed by treating cells with the specific RNAP inhibitor pseudouridimycin (PUM) (29), which significantly increased the percentage of well-resolved expressomes (Fig. 4A). The PUM-induced expressome was refined to 7.1 Å (fig. S26, 27 and Movie S2), revealing direct interaction between the NusG-bound RNAP and the ribosome, and excluding density for NusA (Fig. 4B and fig. S28). This stalled expressome closely resembled the architecture of the E. coli expressome solved in vitro (6) (fig. S28). Interestingly, tRNAs in the ribosome were found in hybrid A/P* and P/E states, and density corresponding to EF-G was well-resolved (Fig. 4C). This suggested that the ribosome was trapped in a pre-translocation state (fig. S28) (30), unable to complete the translocation step, presumably owing to physical obstruction by the stalled RNAP.

In summary, we have determined the native architecture of the expressome in M. pneumoniae and have shown that it requires active transcription and translation elongation. At the RNAP-ribosome interface we unexpectedly found NusA, which followed the path of nascent mRNA at the nexus of transcription-translation coupling. NusA may act as a sensor of RNAP that detects an approaching ribosome and modulates transcription elongation. However, it remains to be seen whether the involvement of NusA in the M. pneumoniae active expressome represents a feature that is conserved across bacteria. Our data highlight the structural heterogeneity of the process, and the potential of integrative in-cell structural biology in elucidating dynamic machineries within their native functional context.

One Sentence Summary.

Integrative in-cell structural biology provides structural insights into bacterial transcription-translation coupling.