Engineering Success

Engineering success for us was not a single clean build. Each of our three cycles began with a method that should have worked, failed in a way we did not predict, and ended with a change we would keep.

One iteration · eight days from plating to fix (June 22–30)

Design

• Recover transformants by plating on antibiotic selection, then grow single colonies in liquid culture for miniprep.
• Chose ampicillin at 100 µg/mL, the standard concentration and what the lab had prepared.

Build

• Heat-shock transformed the products of reactions 1+2, 3+4 and 5+6 on June 22 and plated on LB-ampicillin.
• June 27: picked colonies into twelve tubes of LB with ampicillin, incubated with shaking.

Test

• Colonies appeared on all three plates and looked entirely normal.
• June 28: no growth in any of the twelve liquid cultures. We returned them to the incubator assuming slow growth.
• June 29: still nothing. We examined the plates again and tried to establish the colonies a second time.

Learn

• They were satellite colonies: cells growing in the zone where a real transformant has already degraded the surrounding ampicillin through secreted beta-lactamase. They carry no resistance of their own and are indistinguishable from real transformants on a plate.
• Re-plated on chloramphenicol at 25 µg/mL on June 30. Growth was normal and the problem never recurred.
• Rule we adopted: a colony on a plate is a hypothesis, not a transformant. Nothing counts until it grows in liquid selection.

Three iterations · twelve days from first PCR to gel purification (June 19–30)

Design

• Amplify the backbone fragment for reaction #7 with the standard IVA protocol: 18 cycles, 60 °C annealing, 4 min extension.

Build

• June 19 and June 22: ran #7 alongside the other eleven reactions. Six constructs were in the panel; because backbone and insert amplify separately, that is twelve reactions.

Test

• No usable product. June 25: ran a 55–67 °C gradient at 17 cycles, against a standing fix of 55–65 °C and a standard 18 cycles. Product appeared at every temperature, but every lane carried extra bands above the expected size.

Learn

• The reaction was not failing to amplify. It was failing to be specific, and annealing temperature alone would not fix it.

Iteration 2 — more cycles, better reference

Build

• June 29: same gradient at 20 cycles, with the pEM106 parent plasmid run as a size reference lane.

Test

• Stronger product sitting near the expected size when read against the 10 kb and 3 kb plasmid bands. The extra bands remained.

Learn

• Reading band size off the ladder alone was not reliable for a fragment this size. Running the parent plasmid in the same gel became routine.

Iteration 3 — stop optimising, start purifying

Design

• Accept the extra bands and separate the correct product physically instead of chemically.

Build

• June 30: ran #7 at 65 °C on 0.5% agarose to maximise separation between the two upper bands, then excised each band and gel-extracted it.

Test

• Recovered 8 ng/µL from the upper band and 13 ng/µL from the lower band, roughly 30 µL each. Low but enough to transform.
• Sequencing later showed the upper band carried the correct insert and the lower band was parent plasmid.

Learn

• There is a point at which optimising a reaction costs more than purifying its product. For us that point arrived after two gradients and a cycle-number change.
• Cutting both bands rather than guessing which was correct is what made the sequencing result interpretable.

Three iterations · eight weeks · resolved

Sequencing returned on July 20 showed our ΔTm-FAAH clones carried the insert in frame, but with one extra adenine at the junction. The reading frame downstream was wrong. Removing a single base became the longest engineering problem of our season, running eight weeks across July, August and September.

Iteration 1 — let the cell close the plasmid

Design

• Amplify the entire plasmid with primers that anneal either side of the unwanted base so it is excluded from the product.
• Destroy the parent with DpnI, transform the linear blunt product, and let E. coli circularise it.

Build

• July 30: ran the deletion PCR alongside the 6xHis tagging PCR, under identical conditions.
• August 19: repeated at 60 °C and 62 °C. August 26: repeated across a 61–68 °C gradient.
• DpnI treatment at 37 °C for 30 minutes each time, then transformation.

Test

• July 30: no product at all, while the 6xHis reaction in the same run gave clean product.
• August 26 and September 4: colonies grew, but sequencing returned either unchanged parent plasmid or an ambiguous ‘N’ at the target position. Three clones, none correct.

Learn

• Amplification was never the failing step. Something after the PCR was wrong.
• Our working explanation: we were relying on blunt-end ligation inside the cell, and it was not happening efficiently enough to compete with surviving parent plasmid.

Iteration 2 — design sticky ends instead

Design

• September 8: redesigned the deletion primers so the product would carry complementary single-stranded overhangs, checking the Tm of the homologous region.

Build

• September 9: ran the redesigned reaction across eight conditions.

Test

• Bands came out between 1.0 and 2.0 kb against an expected product of about 3 kb.

Learn

• The redesigned primers were amplifying something other than the full plasmid. We abandoned this route rather than optimise it further.
• Checking product size on a gel before committing to a ligation saved us an entire round.

Iteration 3 — take the ligation out of the cell

Design

• Keep the original blunt PCR, but circularise the plasmid in vitro with purified enzymes before transforming anything.
• Address the surviving parent plasmid directly with a much longer DpnI digest.

Build

• September 10: treated the PCR product with fresh DpnI overnight at 37 °C instead of 30 minutes.
• September 21: ran the product preparatively and excised the ~3 kb band.
• Gel-extracted the fragment, phosphorylated the 5′ ends with T4 PNK, circularised with T4 DNA ligase, transformed into E. coli Stable cells, and picked five colonies for miniprep and sequencing.

Test

• Five colonies (#191 to #195) were submitted for Sanger sequencing. Sequencing confirmed removal of the unwanted adenine and an intact reading frame in clone #193, while #194 returned unchanged parental plasmid.
• A final preparative gel excision and in vitro ligation run on September 21–23 confirmed the reproducibility of the workflow ahead of moving the verified construct into the target chassis.

Learn

• PCR products carry no 5′ phosphate. Without the kinase step there is nothing for a ligase to join, which is why the in-cell attempts returned parent plasmid. This was the missing step, not the annealing temperature.
• A 30-minute DpnI digest leaves enough methylated parent to dominate a transformation. Overnight digestion became standard for every whole-plasmid reaction afterwards.
• Colony count reports nothing about whether an edit took. Only sequencing does.

• Build the catalytic mutant first, not last. It is the control that makes both activity assays interpretable, and we left it until the end. Every activity experiment is now waiting on it.
• Phosphorylate by default. Any whole-plasmid edit we attempt from now on goes through T4 PNK and T4 ligase rather than relying on the cell.
• Name clones once and keep the name. The two July sequencing rounds used the same clone numbers for the same clones, re-sequenced, and we spent weeks believing they were two different numbering schemes. A clone tracker started in August fixed this, and starting it in June would have saved the confusion.
• Move off the plasmid. A chloramphenicol resistance marker on a plasmid in a gut-colonising organism is acceptable for lab work and not acceptable for anything further. Chromosomal integration is the next design change.

The figures behind each cycle are on our Results page, and the dated record on the Notebook page.