The original expression vectors for the RNA polymerase subunits A′ and B′′ from the euryarchaeal hyperthermophile Methanocaldococcus jannaschii (mjA′ and mjB′′, respectively) were generated by subcloning the full-length, PCR-generated open reading frames into pET21a (12). For this study, the pET21a-mjA′ vector was modified by replacing the N-terminal portion of the reading frame (NdeI-AleI) with a synthetic gene fragment (GenScript) encoding the identical amino acid sequence, but containing codons optimized for Escherichia coli expression and with unique restriction enzyme targets flanking the Zipper and Lid encoding sequences. The C-terminal portion (BstBI-BamHI) of pET21a-mjA′ was also replaced with a synthetic gene fragment (GenScript) containing optimized codons and unique restriction enzyme targets flanking the Bridge Helix. A similar modification (replacement of the BseYI-EcoRI fragment with a synthetic gene construct) was carried out on pET21a-mjB′′ to allow mutagenesis of the Fork Loop domains. Further details/full sequences/samples of the modified constructs are available from the corresponding author on request.

All robotic manipulations described here were carried out on either of the two Theonyx Liquid Performers (Aviso Trade GmbH), equipped with an 8-tip pipetting arm, separate gripper arm, barcode reader, vacuum manifold, microplate shaker, thermocycler, microplate reader, 96-well microdialyzer and E-Page facility (Invitrogen) and an integrated microplate centrifuge. Detailed descriptions of all robotic procedures, platform layouts and specialized hardware modifications are available from the corresponding author upon request.

Single E. coli ‘Acella’ (EdgeBio) colonies containing wildtype or mutant pET expression constructs were inoculated into 1.5 ml of autoinduction growth medium (16; Novagen ‘Overnight Express’) containing 100 μg/ml ampicillin in a 24-deepwell plate. The plates were covered with a ‘BugStopper’ venting capmat (Whatman), barcoded and incubated in a shaking incubator for 18 h at 37°C at 250 r.p.m. The plates were then transferred to the robotic platform (1–4 plates per run). After reading the barcode, each plate was transferred to the robotic shaker for 10 s (to ensure an even cell suspension) before 10 μl of each culture were transferred to a 96-well plate for automatically measuring the cell density at 600 nm (A₆₀₀) after a 1/10 dilution step. This step documents cultures that either did not grow or failed to reach the expected cell density.

All procedures described subsequently were carried out on the robotic platforms at room temperature. Nine hundred microlitres of each bacterial expression culture were transferred from the 24-well growth plates to a defined position in a 2.2 ml 96-deepwell block and then lysed directly, without prior centrifugation, by the addition of a mixture of 100 μl FastBreak reagent (Promega) and 2 μl Lysonase (Novagen). The lysates were mixed extensively by repeated up- and down-pipetting and occasional vortexing on the robotic microplate shaker at 5 min intervals for a total of 30 min at room temperature. Inclusion bodies were then separated from the lysate using the robotic centrifuge (10 min at 4700 r.p.m; ∼2300 g), washed once with 950 μl TG100 buffer (25 mM Tris base; 192 mM glycine; pH 8.9) by vigorous robotic pipetting and plate shaking. To ensure complete automation of the entire procedure we developed a custom-built microplate decanter that automatically decants the supernatants without disturbing the inclusion body pellets (R.O.J.W., unpublished data). After a further centrifugation/automatic decanting step the washed inclusion bodies were solubilized by the addition of 500 μl of TG100 buffer containing near-saturating urea (8.3 M) by robotic pipetting and plate shaking. We chose the TG buffer system to provide a vast excess of free amino groups which would compete for the carbamylation reactions that might occur in the urea-containing in vitro assembly reactions (see subsequently). The protein concentrations of the resulting solubilized inclusion body preparations were measured by mixing 10 μl with 190 μl Bicinchonic Acid (BCA) protein assay reagent (Sigma) in a clear microplate and measuring the absorption at 562 nm after incubation at 37°C for 2 h. Different concentrations of bovine serum albumin (BSA) dissolved in TG100/8.3 M urea were set up at the same time to establish a standard curve. Out of several other assays tested, we found that the BCA assay performed best in the presence of the high urea concentrations used and gave an essentially linear read-out in the 0.1–10 mg/ml range. We previously used β-mercaptoethanol (15 mM) in all of our subunit purification buffers (12), but this compound interferes with the BCA assay and was therefore omitted without any apparent loss of activity of the recombinant subunits after in vitro assembly (R.O.J.W., unpublished observations). Based on comparisons with the BSA standards and on the staining of the purified subunits on gels we typically obtained 250 μg of purified recombinant subunits from 900 μl expression culture with a standard deviation in the concentrations of individual subunit preparations of less than ±10%.

A commercially available 96-well microdialysis device (Spectrum Laboratories) was installed on top of a Hi300N magnetic stirrer (Hanna Instruments) positioned on the robotic platform. The dialyzer position is fully accessible to the pipetting and gripper arms of the robots. The perspex lid was customized for handling by the gripper (details available from the corresponding author). A peristaltic pump (Watson Marlow Sci400) was calibrated to deliver a flow-rate of 1 ml/min and switched on and off under robotic control. In vitro assembly reactions were set up in a total volume of 100 μl per dialysis well and a dialysis membrane with a 3 K cut-off point was used for all assemblies (Spectrum Laboratories; a narrow pore membrane is required to prevent the smallest subunits from leaking into the buffer chamber under denaturing conditions). The concentrations of the various RNAP subunits and subunit ratios were empirically determined to yield recombinant RNAPs displaying the same specific activities in non-specific and factor-dependent transcription assays as previously described for the wild-type enzyme (12). For the example shown here (mjA′ G825-X mutagenesis set), 90 μl of a ‘Master Mix’ containing 3.2 μg A′′, 4.2 μg B′, 6 μg B′′, 7.2 μg D, 8.7 μg L, 6.5 μg H, 5 μg N and 3 μg P were mixed with 10 μl of the robotically purified mjA′ mutant subunit (∼12 μg). The stoichiometric ratios of the subunits required for optimal in vitro assembly most likely reflect differences in the relative refolding efficiencies. Under these conditions the subunits present in the Master Mix (-A′) are rate-limiting in the assembly reactions, thus ensuring that the added mutant mjA′ subunits are present in >2 molar excess. Any minor variation in the concentrations of various mjA′ samples is thus compensated for and does not affect the amount of assembled polymerase (the presence of excess mjA′ has no detectable negative effect on the efficiency of the assembly reaction).

The following procedures were carried out automatically without human intervention. Non-specific transcription assays (measuring the incorporation of α-³²P-rUTP into acid insoluble transcripts) were set up as 50 μl reactions in a 96-well thin wall plate using reaction conditions previously described (12,14). After incubation for 45 min at 70°C (using the integrated thermocycler with heated lid) the plate was transferred to a temperature-regulated block maintained at 1°C. The precipitation of the radiolabelled transcripts was initiated by the addition of 150 μl 15% ice-cold trichloroacetic acid (TCA), followed by incubation for 30 min at 1°C. The precipitation reactions were then transferred by pipetting onto a 96- GF/F glass fiber filter plate (Whatman) positioned on the robotic vacuum platform. The soluble portion of the reactions was filtered to waste. The precipitates immobilized on the filter surface were washed eight times with 200 μl ice-cold 10% TCA to remove traces of unincorporated nucleotides, twice with 200 μl ice-cold 95% ethanol and then dried by a final application of vacuum. The incorporated α-³²P-rUTP levels were measured after addition of 50 μl of scintillant (Optifluor-O) on a microplate counter (TopCount NXT, Packard) and were found to be identical to the results obtained with the previously described manual technique (12,14).

Structural studies of RNAPs have revealed that only a relatively small proportion of amino acid residues are likely to play any direct functional role in transcription mechanisms. These residues are typically arranged within structurally distinct domains [such as Fork Loops, Bridge Helix, Lid, Zipper, Rudder (2,3)] that contact substrates (DNA, RNA, nucleotides) directly and/or interact with other protein domains within the active site (1–4,7,8). Residues located within such domains are thus prime targets for site-directed mutagenesis projects. Moreover, these functional domains tend to be either inherently unstructured [e.g. Fork and Trigger Loops (2–4)] or are part of flexible α-helical structures [e.g. Bridge Helix (1,2) and Trigger Loop (8)] so that single amino acid substitutions are unlikely to interfere with subunit assembly or overall stability of the enzyme.

The RNAP from the archaeon M. jannaschii (mjRNAP) contains 12 different subunits that are structurally and functionally highly similar to their counterparts in eukaryotic RNAPII (11,12). A high-throughput functional analysis of a particular functional domain, such as the Bridge Helix located in the archaeal A′ subunit, requires an efficient method for creating the desired mutants, followed by a robust method for purifying the mutated subunits in a highly reproducible manner. The resulting purified mutant protein preparations then need to be combined with a ‘master mix’ of all other essential subunits and assembled into separate recombinant RNAPs. Finally, the enzymes containing the mutated subunits need to be assayed and compared to the wild-type polymerase to assess the phenotypic consequence of the introduced mutations.

We started the design of the high-throughput assembly strategy by redesigning the previously used bacterial expression vectors encoding the mjA′ and mjB′′ subunits (12). These two subunits include many of the identified functional motifs surrounding the active site (mjA′: Zipper, Lid, Rudder, Switch-2, Bridge Helix; mjB′′: Fork Loop 1, Fork Loop 2; Figure 1a). The redesigned sequences were codon-optimized for optimal translation efficiency in E. coli, and also contain unique 6- and 8-bp target sites for restriction enzymes that flank the nucleotide sequences encoding the various functional motifs (Figure 1b). The introduced restriction enzyme sites are silent, i.e. encode the ‘wild-type’ primary amino acid sequence of the encoded proteins. Cutting these modified expression vectors with two restriction enzymes flanking a particular functional domain releases a short piece of double-stranded DNA and allows this sequence to be replaced by a double-stranded synthetic oligonucleotide encoding the desired codon replacement(s) (Figure 1c). Using this method, we have up to now successfully created systematic substitution mutants in several RNAP motifs, including Fork Loop 1 and 2, Lid, Zipper and the Bridge Helix (L.T., D.T., H.C. and R.O.J.W., unpublished results). Apart from the ease of creating a variety of mutations, one of the main advantages of this method of creating mutants is that it allows the saturation mutagenesis of a short region (typically 30–80 nt, depending on the length of the chosen motif) without affecting any of the other regions outside the targeted area. This mutagenesis strategy therefore offers many advantages over frequently used ‘shotgun’ mutagenesis methods, such as error-prone PCR (17), by limiting the mutagenic changes to distinct regions of an open reading frame. Figure 1.

Redesign of the bacterial expression vectors. (a) Schematic diagram of the redesigned coding regions of the mjA′ and mjB′′ bacterial expression vectors. The full-length open reading frames are shown with the codon-optimized sequences in light blue. The position of the sequences encoding key functional RNAP domains are shown in color (Zipper, Lid, Rudder, Switch-2 in red, Bridge Helix in dark green and Fork Loops in blue). The numbers refer to nucleotide positions. (b) Original and redesigned sequence encoding the Bridge Helix domain and surroundings (positions 2281-2550 in the mjA′ open reading frame). The original nucleotide sequence (as found in the M. jannaschii genome; 23) is shown on the top line with the redesigned sequence (starting at the BstBI site) beneath. The silently introduced restriction sites flanking various portions of the Bridge Helix-encoding sequence are in red, whereas other sequence alterations that improve codon usage in E. coli are shown in light blue. The amino acid sequence encoded is shown in single-letter code in black and the Bridge Helix primary sequence is highlighted in green. (c) Example of mutagenic oligonucleotides for creating targeted substitutions in Bridge Helix residues mjA′-G825. The two strands are designed to hybridize to each other to create a double-stranded ‘cassette’ bearing single-stranded extensions (shown in red) suitable for ligating into cloning vectors cleaved with suitable restriction enzymes. The cassettes shown have been used to replace the wild-type Bridge Helix sequence between the SfiI and SbfI sites (as shown in panel B) with a sequence containing a single randomized codon (shown in purple; 18).

The previous section described an approach capable of creating a broad range of mutations restricted to ‘functional’ residues. The resulting bacterial expression vectors encode recombinant RNAP subunits (mjA′ or mjB′′) that contain a variety of predetermined mutations in one or more residues. This is, however, only the first step of a more complex procedure since the mutant subunits need to be assembled in the presence of all other essential subunits into an intact recombinant RNAP before the functional consequences of a particular mutation can be assessed. These robotically implemented procedures take place in parallel (up to 96 independent purification/assemblies per run) to maximize consistency and to create optimal conditions for functional comparisons. In addition, the technical solutions outlined below take full advantage of efforts to miniaturize protein expression/purification methods and the extensive use of automation/robotic capabilities. A flowchart that summarizes the entire sequence of expression/purification procedures is shown in Figure 2. Figure 2.

Flow diagram of the automated subunit purification and assembly process. See text for further details.

Active archaeal RNAP can be reconstituted by mixing the various subunits in appropriate ratios under denaturing conditions (6 M urea), followed by the time-controlled removal of urea by dialysis against buffers containing decreasing urea concentrations (12,15). The originally described methods were time-consuming (because the buffers were changed manually according to a strictly defined timetable) and low-throughput (parallel assembly of only a few recombinant RNAP preparations per experiment). In addition, variations in the assembly conditions (e.g. due to differences in protein/buffer preparations) contributed to poor quality control in such procedures. We therefore introduced two significant technical changes aimed at increasing throughput and reproducibility. First, we changed the dialysis set-up by using a microdialyzer device on the robotic platform that allows the in vitro assembly of up to 96 recombinant RNAPs in parallel (Figure 3a). The dialysis wells on the microdialyzer are arranged in the standardized 96-well format that facilitates the robotic transfer of liquids in and out of the dialysis block. Each well typically holds 100 μl of dialyzate that is separated from a lower communal buffer chamber by a dialysis membrane. The communal buffer chamber contains inlet and outlet ports that allow dialysis buffer from an external supply to be pumped through the chamber (Figure 3a). This is an important consideration in light of the other modification we made in the procedure: instead of manual changes of dialysis buffer to lower the urea concentration during the assembly reaction we decided that the reproducibility of the procedure could be improved substantially if the buffer changes were carried out automatically (similar to an independently developed method; 18). We therefore modified the renaturation protocol by initially filling the chamber with ∼200 ml of buffer containing 6 M urea. The RNAP subunit mixtures are then robotically pipetted into the wells and dialyzed against the 6 M urea buffer. After 1 h a pump is switched on under robotic control to pump urea-free buffer into the chamber at a low rate (typically 1 ml/minute). The urea-free buffer is immediately mixed with the 6 M urea buffer in the chamber by a magnetic stirrer and excess liquid is allowed to drain through the outlet port. This simple set-up therefore leads to a slow and continuous dilution of the urea concentration in the dialysis buffer in the lower chamber, resulting in a corresponding drop of the urea concentration in the RNAP subunit mixtures located across the dialysis membrane. Over a period of around 16–17 h the urea concentration in the dialysis chamber approaches zero and the assembly reaction is complete (Figure 3b). The assembled recombinant RNAPs are robotically pipetted out of the dialysis wells and transferred to 96-well microplates for immediate use in various functional assays. The robotic systems carrying out these procedures thus constitute a genuine ‘RNA polymerase factory’ capable of purifying and assembling up to 96 recombinant archaeal multi-subunit complexes with little or no human intervention. Figure 3.

Robotic microdialysis set-up for assembly of recombinant RNAPs. (a) The photograph shows the placement of the 96-well microdialysis device on the robotic platform. A magnetic stirrer is sunk into the platform with full manual access to the stir speed control dial. The lid of the microdialysis device is held by the gripper arm of the robot during the loading and removal of the assembly reactions. On the left hand side, the peristaltic pump controlling the flow/dilution rate of the dialysis buffer can be seen. (b) Calculated urea concentration of the dilution of 200 ml 6M urea-containing buffer with urea-free buffer at a flow of 1 ml/min. After 17 h, the urea concentration in the dialysis buffer has effectively fallen to zero. The change of urea concentration (dc) over time (dt) can be expressed as a differential function (dc/dt) that is equal to the dilution (-D) of the initial concentration (c₀) of urea according to the following equation c_t = c₀ × e ^−Dxt [D, Dilution (i.e. flowrate/volume); c_t, urea concentration at time t; e, Euler's constant].

We demonstrate the effectiveness, reproducibility and high-throughput potential of the RNA polymerase factory strategy by showing examples of results obtained by saturation mutagenesis of a single residue of the Bridge Helix of archaeal RNAP. Briefly, the Bridge Helix is a highly conserved structure (Figure 4a) spanning the cleft near the active site in bacterial, archaeal and eukaryotic multi-subunit RNAPs (1–4). It is thought to play an important (though still controversial) role during the RNAP transcription cycle by defining the position of the template strand and of the DNA/RNA hybrid in the catalytic site (Figure 4b). Inspection of the yeast RNAPII X-ray structure of an elongation complex (7; PDB #1SFO) suggests that an evolutionary invariant residue (scRPB1-G835 in yeast RNAPII, orthologous to mjA′-G825 in M. jannaschi; Figure 4a) may have been specifically selected for the small size of its side chain to allow an unimpeded passage of the DNA template strand into the active site. The available structural data predicts that any other amino acid substitutions in this particular location would result in decreased catalytic activity due to steric clashes between the larger side chains and the phosphodiester backbone of the DNA template strand. To test this hypothesis we created a set of mjA′ mutants containing all 19 possible amino acid substitutions (the ‘mjA′ G825-X’ set) by substituting the SfiI-SbfI fragment of the redesigned mjA′ expression vector with a double-stranded synthetic oligonucleotide containing a randomized codon (19) in the position encoding G825 (Figure 1c). A complete set of amino acid substitutions in G825 was selected from ∼60 randomly picked clones and by cloning double-stranded oligonucleotides designed to contain any missing substitutions. Plasmids encoding the mjA′ G825-X subunit variants were grown in expression strains and processed by the RNA polymerase factory. The results obtained confirm that the size of the side chain does indeed appear to play a key role for the residue at position 825 (Figure 4c). The activities of recombinant RNAPs containing the other 19 possible residues in this position reach at the most 40% of the wild-type catalytic activity and many of the more bulky amino acid substitutions (F, Y, W, K, R, H) result in <5% activity in non-specific transcription assays. The wide dynamic range of reproducible activity measurements demonstrated by the mjA′ G825-X mutant set show that the RNA polymerase factory is a robust experimental tool capable of providing functional data at an unprecedented speed and accuracy. Figure 4.

Saturation mutagenesis of mjA′-G825. (a) The sequence alignment shows part of the central and C-terminal region of the Bridge Helix [Hs, Homo sapiens RPB1 (positions 845–863); Sc, Saccharomyces cerevisae RPB1 (positions 822–840); Mj, M. jannaschii A′ (positions 812–830)]. Bold residues are identical in all three species and the residues orthologous to mjA′-G825 are marked by an yellow box. (b) The position of the glycine residue in yeast RNAPII corresponding to mjA′-G825 is shown in yellow. The Bridge Helix (including the residues aligned in a) is in gray and the backbone of the DNA template strand is shown in light blue. Note the close spatial proximity between the G825 side chain and the phosphodiester backbone of the DNA [data from PDB #1SFO (5) visualized with Cn3D; www.ncbi.nlm.nih.gov/Structure/CN3D/cn3d.shtml]. (c) Transcription activity assays of RNAP variants containing all 19 possible substitutions of mjA′-G825 carried by the RNA polymerase factory. The two columns on the right represent the positive (WT; wild-type) and negative (RNAP containing a mjA′-T821 termination codon substitution) control reactions. Error bars shown represent standard deviations (n = 4) based on independently expressed, assembled and assayed mutants carried out in parallel.