Targeted mutagenesis of the last four codons of the celF gene was initially accomplished by creating a PCR primer set with the reverse primer (anti-sense strand) containing any nucleotide in the first and second positions and a pyrimidine in the last position (NNY). Consequently the sense strand contains the complementary NNR motif where R is a purine. This strategy generates a library of open reading frames that code for 14 of the possible 20 amino acids at each of the last four positions in the CelF mutants, or 144 = 38,416 possible combinations. Phe, Tyr, His, Asn, Asp, and Cys are not possible since they have only NNY codons. Also, because of the degeneracy of the code, almost every amino acid is represented by several codons (synonyms). In the NNR strategy the amino acids with the most synonyms are leucine and arginine. Although the amino acids with more synonyms have a theoretically greater likelihood of appearing, it is possible that they do not improve the activity of the mutant and may even eliminate activity and thus are not seen in the screening process.
A subsequent large-scale screening run was conducted in which the site-directed PCR mutagenesis strategy was modified to obtain NNY as well as NNR codons in the sense strand. This was accomplished by producing PCR primers with an NNR motif in the anti-sense strand in addition to the primers with the NNY motif. A library of open reading frames is obtained from the NNR primers coding for the six amino acids having only NNY codons and also for the nine other amino acids having both NNY and NNR codons. So the NNY codon strategy gives a total of 154 = 50,625 possible CelF variants. Together with the 38,416 possible CelF variants from the NNR codon strategy this gives a total of 89,041 theoretically possible CelF variants.
The forward primer contained the CACC sequence in front of the ATG start codon to allow the library of mutagenized clones to be moved into the pENTR D TOPO vector system directionally. The resulting library was transformed into E. coli cells, spread onto selective plates and incubated overnight. In the initial NNR screen, approximately 10 to 80 colonies were obtained on each of the 100 plates in this set when each plate was spread with one 25-μL competent cell transformation. In the subsequent large-scale NNR plus NNY screening run, 178 spread plates were prepared from the transformation reactions containing clones with NNY codons and 262 spread plates were prepared from the transformation reactions containing clones with NNR codons. One 25-μL competent cell transformation was used for each spread plate.
Multiplexed colony picking
Colony picking was performed on the automated workcell using the workcell picker (Figure 1 #4) to give multiplexed wells on a 96-well deep well culture plate. The first run entailed picking eight colonies from the spread plates to inoculate each well of an ABgene deep well block containing 1.6 mL of TB KAN 25 medium, giving a total of 768 clones for one plate (8 per well × 96 wells). From the 768 clones screened in this run, five (5) multiplexed cultures with improved activity were identified. The subsequent large-scale run entailed picking 80 colonies to inoculate each well of three 96-well deep well blocks and 4 colonies to inoculate each well of another 96-well block, giving a total of 23,424 clones screened (3 × 96 × 80) plus (96 × 4). Of this total, 9,486 were taken from colonies containing NNY clones and 13,938 were taken from colonies containing NNR clones. The number of clones screened was increased to verify the high-throughput capability of the automated platform.
Integration of the picking operation onto the robotic workcell permits automated multiplexing, which conserves reagents and time. After colony picking was performed, the resulting multiplexed cultures were incubated overnight and processed using the plasmid preparation protocol on the liquid handler of the workcell (Figure 1 #9). Inserts were directionally cloned into pENTR D TOPO using topoisomerase ligation, then into pEXP-1 DEST vector via LR clonase reaction and transformed into TOP 10 E. coli cells . The resulting multiplexed plasmid library with mutagenized inserts was used for in vitro transcription/translation.
Screening of multiplexed mutants
The expressed CelF proteins from the multiplexed transcription/translation reactions were assessed for cellulase activity at several pH levels using an automated azo-CMC plate assay. The results at pH 5.8 showed several cleared zones after incubation for 10 hours at 37°C, indicating cellulase activity (Figure 2). It can be determined from the diameter and intensity of the cleared zones on the plate that protein from wells B11, C12, G2, G11, and H12 had elevated activity at this pH. The negative control placed on the plate (pUC19) in well E1 showed no cellulase activity. Activities of positive controls, wild-type (CelF #5) and an improved mutant (CelF #62) obtained previously , are shown in wells B3 and C5, respectively.
Control experiments were performed to determine the effect of increasing amounts of CelF mutants in a multiplexed well or increasing numbers of multiplexed colonies on assay sensitivity. Figure 3 illustrates that as the amount of an individual active CelF mutant spotted on the azo-CMC plate is increased from 27.7 to 277 ng, the activity measured by the diameter of the cleared zone produced on the plate increases in a linear fashion. The cleared zones on the azo-CMC plate at pH 5.8 are shown in the inset at the top of the graph (incubation at 10 hours, 37°C). In addition, if the number of colonies multiplexed in a well is increased relative to an individual CelF mutant, the activity of that mutant is still detectable. Figure 4 depicts the activities obtained when an individual active CelF mutant is added to a mixture containing an increasing number of background colonies multiplexed as described in this paper, from no background colonies (active mutant alone) to 96 background colonies per well. The cleared zones on the azo-CMC plate at pH 5.8 are shown in the inset at the bottom of the graph (incubation at 10 hours, 37°C).
Identification of individual clones
The multiplexed wells containing mutants with improved activity that were identified on the multiplexed azo-CMC screening plate (Figure 2) were linked back to the corresponding multiplexed cultures stored in glycerol. The identified glycerol stocks were spread on LB AMP 50 plates and single colonies from these plates were picked by the workcell colony picker into a 96-well plate to isolate individual clones from the selected high-activity wells, B11, C12, G2, and G11. The single colonies were picked into TB AMP 50 medium and incubated for 20 hours. In the automated picking process, the colony picker was programmed to select colonies between 0.2 to 10 millimeters in size. This restriction was fixed for a given run but could be reprogrammed if necessary. To determine how this automated picking process compared with manual picking by an individual who has greater leeway in selecting colonies and who may attempt to optimize the variety of colonies being picked, a set of colonies was also picked manually.
The manual picking was performed using 4 groups of 24 wells on the 96-well plate and taking 24 colonies corresponding to each of the identified wells (process outlined in Figure 5). After expansion of the 96-well plate into four 24-well plates, large-scale plasmid preparation was performed on the liquid handler and the resulting plasmids were eluted into one 96-well Matrix collection plate. The plate was moved to the Bio-Tek UV/VIS microplate reader and the absorbance at 260 nm of each well was adjusted to 1.0 (equivalent to a concentration of 0.05 μg/μL) by diluting with Qiagen EB buffer on the liquid handler deck using the liquid handler pipets and buffer from the cold block on the deck. In vitro transcription/translation reactions were performed using the entire plate of plasmids. Protein was spotted on four azo-CMC plates at pH 4.0, 4.5, 5.0, and 5.8. Protein heated to 50°C for 1 hour was spotted on a fifth plate (at pH 5.8). An improved CelF enzyme was isolated from each of the identified multiplexed wells (Figure 6). Individual mutants are identified by row (letter) and column (number) on the plate and by "MP" indicating they were manually picked. Columns 1–3 on each plate were picked from colonies corresponding to multiplexed well B11 (see Figure 2) (high activity variant F2MP boxed in purple is in row F and column 2), columns 4–6 from multiplexed well C12 (high activity variants are E5MP boxed in yellow and F5MP boxed in blue), columns 7–9 from multiplexed well G2 (high activity were E7MP in green and F9MP in orange), columns 10–12 from multiplexed well G11 (F12MP in pink). Clones F2MP, F5MP, F9MP, and F12MP had activity at lower pH than wild-type CelF. Clones F2MP, F9MP, and F12MP were also stable after heating to 50°C for 1 hour. F9MP (boxed), which exhibits the most exceptional activity, was shown to have the same sequence as B9MP, C9MP, and E8MP (all not boxed), which all exhibit activity similar to F9MP. The boxed mutants were selected for Western blot analysis.
The same multiplexed glycerol stock spread plates from which colonies had been picked manually were also used for automated picking into a 96-well plate (see process in Figure 5). The colonies were picked and identified as controls or samples. Expansion of the 96-well plate into the first of the four 24-well plates involved inoculation on the workcell with 84 colonies from five of the identified high-activity multiplexed stocks (B11, C12, G2, G11, H12). Control cultures were placed in 12 of the 24 wells (boxed at top left of Figure 7), including pUC19 (negative control for expression), GUS gene (positive for expression; negative for activity), wild-type CelF #5 and improved mutant CelF #62  (positive controls for activity). The 24-well process was repeated for three more plates. After expansion of the 96-well plate into the four 24-well plates, large-scale plasmid preparation was performed on the liquid handler and the resulting plasmids were combined into a 96-well Matrix collection plate. Plasmid DNA concentrations were adjusted to 0.05 μg/μL using the A260 reading from the UV/VIS microplate reader so all clones were present in sufficient quantity for consistent expression, and then in vitro transcription/translation reactions were performed for the entire plate of plasmids. The resulting set of recombinant proteins was spotted on azo-CMC plates at pH 4.0, 4.5, 5.0, and 5.8. A fifth plate (at pH 5.8) was spotted with protein heated at 50°C for 1 hour. Results were similar to those for the manual picking procedure and indicated that an improved CelF protein was present in each of the identified multiplexed wells. Two mutants having high activity at lower pH and showing stability at elevated temperature (G4AP boxed in brown, G12AP boxed in red) were identified (Figure 7) in the automated picking process ("AP" indicates clones from this process). Several other CelF variants were seen that had high activity at pH 5.8 and 5.0 but not at pH 4.5 or 4.0 (for example, C1AP and E2AP boxed in aqua). One mutant (H1AP in aqua) had activity only at pH 5.0 and two variants were not stable after heating to 50°C (E2AP and D6AP in aqua). Several clones were aberrant because no insert was detected (B8AP), no plasmid DNA was recovered (H10AP) or an unsuccessful recombinational event occurred (F9AP) as seen when the plasmids were subjected to Bsr GI restriction analysis (data not shown).
Sequences of mutants
Sequences for the mutagenized clones that gave CelF variants with improved activity at lower pH are listed in Table 1. Sequencing results for the manually picked clones identified five unique sequences from the 768 clones originally picked into the multiplexed wells. Clones G12AP and G4AP selected in automated picking had the same sequence as F2MP and F9MP, respectively, from manual picking. The CelF variant (F9MP/G4AP) that had the greatest activity at lower pH and was stable at elevated temperature also contained a point mutation at amino acid 213 in addition to the four amino acid changes at the C-terminus generated by mutagenizing the last four codons of the celF gene. A CelF mutant, C7HTS ("HTS" denotes high-throughput screen), that was active at restrictive pH 4.0 and very stable to heating at elevated temperature, with a sequence differing in the last four amino acids from those in the mutants obtained previously, was identified in the large-scale run from the NNY plus NNR mutagenesis strategy. C7HTS also had a premature stop codon, shortening the sequence by one amino acid.
Western blot analysis of the expressed proteins is provided in Figure 8. Wild-type CelF (control #5) and variants of the same length as wild-type give a band at 47.7 kD. E5MP in the first lane at 52.6 kD had a mutated stop codon resulting in a CelF variant with 54 more amino acids than wild-type (486 versus 432) (sequence not shown). It had decreased stability after heating and some cellulase activity down to pH 4.5. The G4AP CelF variant, which was shown to have the same sequence as F9MP, demonstrated activity similar to F9MP in that it had greater stability after heating than wild-type CelF and retained cellulase activity down to pH 4.0. Variants F5MP, F12MP, F2MP and E7MP all had activity at lower pH (4.0 or 4.5) but none was as stable after heating as G4AP/F9MP. Variants H1AP, C1AP, E2AP and D6AP demonstrated less cellulase activity than wild-type CelF. Control #62 is an improved CelF variant obtained previously . No cellulase activity and, in fact, no CelF variants were obtained from locations B8AP, F9AP and H10AP. Restriction digest analysis (data not shown) showed primer-dimer insertion, a recombinational event or no plasmid was present in those cases, respectively.
Figure 9 provides a comparison of the activity of wild-type CelF and CelF mutants, #62, F9MP, and C7HTS, at pH 4.0. C7HTS from the large-scale screen demonstrates the greatest activity compared with all other clones obtained to date. The Western blot analysis in the inset at the top shows that the improved CelF variants C7HTS, F9MP and #62 were present at the same concentration for all comparisons.