The Crystal Structure of the Escherichia coli RNase E Apoprotein and a Mechanism for RNA Degradation

Summary RNase E is an essential bacterial endoribonuclease involved in the turnover of messenger RNA and the maturation of structured RNA precursors in Escherichia coli. Here, we present the crystal structure of the E. coli RNase E catalytic domain in the apo-state at 3.3 Å. This structure indicates that, upon catalytic activation, RNase E undergoes a marked conformational change characterized by the coupled movement of two RNA-binding domains to organize the active site. The structural data suggest a mechanism of RNA recognition and cleavage that explains the enzyme's preference for substrates possessing a 5′-monophosphate and accounts for the protective effect of a triphosphate cap for most transcripts. Internal flexibility within the quaternary structure is also observed, a finding that has implications for recognition of structured RNA substrates and for the mechanism of internal entry for a subset of substrates that are cleaved without 5′-end requirements.


RNase E Complexed with Small Fragments of RNase P at 3.5 Å
Crystallographic data and model statistics for RNase E bound to nucleotides of M1 RNA are presented in Table S1. Although crystallizations of the complex were carried out with the addition of either a 377 or 327 nucleotide derivative, the crystals possess only a few nucleotides. An elemental analysis of the crystals by microPIXE (Proton Induced X-ray emission spectroscopy) (Garman and Grime, 2005) places an upper limit of ~40 phosphates per tetramer (see below for methods).
We could identify ten nucleotides in the structure. Additional bases of RNA may be present, but these could not be accounted for given the resolution and the presence of disordered regions. Two of the four S1 sub-domains could not be built as well as one of the zinc-coordinating peptides.
The structure of the M1-fragment/RNase E complex is in many respects intermediate between that of the holo-form (Callaghan et al., 2005) and the apo-form described in the main text. Although the quaternary organization is deformed relative to the D 2 symmetry of the original holo-form, it is still nearly planar and only possesses a slight ~10° bend as compared to the much greater ~40° bend of the apo-form (Figures S1A and S1B). Aside from the differences in the heterologous domain-domain interactions that produce this bending of the tetramer, the tertiary organization of the individual protomers is nearly identical to that of the apoprotein structure.
A short strand of two to three nucleotides is seen in each protomer within the structure ( Figure S2). Just as the putative sulfates in the apoprotein are hydrogen-bonded to residues R169 and T170, so is the 5'monophosphate group for each strand. Relative to the consolidated 5/S1 sub-domain, the first two nucleotides are oriented identically to those found at the 5' end of the previously solved holoprotein structures.
Where present, the third nucleotide base of M1 RNA binds to a hydrophobic surface patch on the S1 sub-domain (residues F57 and F67) of an adjacent tetramer ( Figure S2B). There, the base is rotated by about 90° relative to the holoprotein substrate, but in the same plane as the nucleotide rings. Thus, the purine rings remain parallel to the side chain of F67 and the π-π stacking interaction is maintained between the two structures.

Processing
For the co-crystallisation of RNase E catalytic domain to fragments of M1 RNA, a truncated form of M1 RNA lacking bases 156-205 and with total length 377 or 327 nucleotides was mixed with RNase E in a roughly 1:4 RNA to protein ratio and incubated for 30 minutes. The complex was concentrated and crystals grown using the sitting drop vapor diffusion method in Hampton Membfac screen condition number 34 (10% PEG 4000, 100 mM (NH 4 ) 2 SO 4 , and 100 mM HEPES pH 7.5). A single crystal was serially transferred into 10, 15, 20, 25, and 30% v/v ethylene glycol cryoprotectant before rapid cooling in liquid nitrogen.

Solution of RNase E -M1 RNA Complex Structure
PHASER and MOLREP were used to construct a model of RNase E within the crystal by molecular replacement. As in the apoprotein, individual protein domains from the existing holoprotein tetramer (PDB entry 2BX2) were used as search models to iteratively build the structure.
The DNase I domains were placed first within the tetramer using PHASER, followed by two RNase H domains, two 5/S1 domains, a small domain, the remaining RNase H domains, and the remaining 5' sensors.
The second small domain was placed using a MOLREP search within the difference map of the structure and the third small domain found using PHASER. The fourth small domain was added directly by hand, based on the relative positions of the small domains already present. The model was refined using MOLREP rigid-body refinement, and both the Zn-link domains and initial RNA fragments were added in Coot based on homology to the holoprotein structure. Missing residues were then added to the structure using Coot to the extent possible from the electron density map. The structure was refined using TLS combined with restrained refinement and structure idealization was carried out in Refmac. The structure was examined and validated using PROCHECK (Laskowski et al., 1993) and SFCHECK (Vagin et al., 1999), and shown to have appropriate chemical bond lengths, phi-psi backbone angles, and chi angles. No residues have disallowed Ramachandran geometry.

MicroPIXE
Data were collected at the Surrey Ion Beam Centre (University of Surrey). A 2.5 MeV proton beam of 2 µm diameter was used to induce characteristic X-ray emission from protein samples under vacuum. The sample to detector distance was 25 mm. The X-rays were detected in a solid-state lithium-drifted silicon detector with high energy resolution, placed at an incident angle of 45° with respect to the incident beam. The crystals examined were RNase E catalytic domain with RNase P M1.
RNase P M1 contains 377 nucleotides, i.e. 377 phosphorus atoms. From the crystal structure of RNase E catalytic domain, it is known that there is 1 zinc atom per RNase E dimer. The protein buffer was 20 mM Tris, 500 mM NaCl, 0.5 mM EDTA, 10 mM MgCl 2 , 10 mM DTT and glycerol.
RNase P was stored in water. The crystals were grown in 10% PEG 4K, 100 mM (NH 4 ) 2 SO 4 , and 100 mM HEPES pH 7.5. The crystal of the RNase E catalytic domain/15-mer single-stranded RNA complex for which the structure has been determined, were used as a control. The 15mer RNA contains 15 phosphorus atoms, and there is one molecule of RNA per RNase E protomer. One zinc is expected per dimer. The protein buffer was 20 mM Tris, 500 mM NaCl, 0.5 mM EDTA, 10 mM MgCl 2 and 10mM DTT. These crystals were grown in 100 mM Tris pH 7.8, 20 mM Mg formate 2% PEG 8000 and 0.1 mM Na azide. Normally the ratio of sulphur to phosphorus signal would be used to determine the RNA species present per protein molecule. However, in this case there is extra sulphur in each sample from corresponding buffers that would interfere with the sulphur signal from the methionine and cysteine residues.
Therefore the Zn/P ratio was measured. This assumes that in the RNase E/RNase P complex crystals, RNase E binds the same number of Zn atoms per protein molecule as the control RNase E/15-mer complex. One RNase E/15-mer RNA complex crystal and two RNase E/RNase P complex crystals were prepared for the experiment. They were transferred from the crystallization drop to deionized water and washed three times in order to remove the buffer. The crystals were then transferred onto 2μm thick mylar film sample holders (Garman and Grime, 2005). The samples were aligned in the beam and a broad two dimensional scan (1 mm x 1 mm) was performed to localize the crystal to be examined. After location of the crystal, a 200 x 200 µm scan was performed. In order to measure element concentrations, point spectra were collected at three selected points in the crystal and also at a point on the backing foil at different positions in each of the crystals for 10 min. The spectra were analysed by the DAN32 and GUPIX software (Grime, 2004, Campbell et al., 2000.   Shown here are (a) two large domains in adjacent unit cells within the RNase E -M1 RNA crystal structure, only one of which (top) possesses an ordered S1 domain. A three-nucleotide fragment of RNA stretches between these two monomers and interacts with the hydrophobic surface patch on the S1 sub-domain at one end (b) and the 5' sensing pocket at the opposite terminus (c). The electron density shown represents a simulated annealing omit map for the RNA, generated using PHENIX starting from the refined model in which the RNA was removed.  [Vagin et al., 1999]. Refinement statistics were calculated by Refmac [Winn et al., 2001]. No Ramachandran outliers are present in the model.