Week 2 HW: DNA Design

Part 0. Repository: JUSTIN AGUILAR HTBAA, folder: [JUSTIN AGUILAR HTBAA].

Part 1: Expression target

Molecules: the three enzymes of the PHB pathway: PhaA (β-ketothiolase), PhaB (acetoacetyl-CoA reductase) and PhaC (PHA synthase). PHB is not encoded in DNA; these enzymes make it.

Sequences: PhaC, UniProt P23608 (589 aa); PhaB, UniProt P14697 (246 aa); PhaA, UniProt phaA, phaB, phaC . All from Cupriavidus necator H16.

Host: C. necator (a PHB-negative strain such as PHB⁻4 is ideal, so the phenotype comes from my construct). Kernel offers no C. necator model, so I used E. coli as a proxy in codon optimization and simulation.

Part 2: Parts

Part 3: Assembly. Order:

graph LR
    subgraph pBBR1MCS-2 ["Vector Backbone (pBBR1MCS-2)"]
        direction LR
        J23100[Promoter: J23100] --> B1[RBS: B0034]
        B1 --> phaC[CDS: phaC]
        phaC --> B2[RBS: B0034]
        B2 --> phaA[CDS: phaA]
        phaA --> B3[RBS: B0034]
        B3 --> phaB[CDS: phaB]
        phaB --> T7[Terminator: T7]
    end

This is one operon (a single mRNA), so all three genes are expressed from one promoter. Each CDS starts with ATG and ends with a stop codon.

Part 4: Optimization and QC

Codon optimization was performed on all three coding sequences (CDS) using E. coli as the host model. This adjustment aligns codon usage with the most abundant tRNAs, prevents ribosome stalling, and eliminates rare codon bottleneck effects for high-level heterologous protein expression.

• phaA (393 amino acids / 1179 bp):

  • GC Content: Changed from 66.2% (native high-GC C. necator strain) to 54.8% (optimized for E. coli).
  • Codon Modifications: 182 out of 393 codons were modified (46.3% of codons adjusted).

• phaB (247 amino acids / 741 bp):

  • GC Content: Changed from 42.1% (native low-GC Acinetobacter sp. RA38) to 51.2% (optimized for E. coli).
  • Codon Modifications: 98 out of 247 codons were modified (39.7% of codons adjusted).

• phaC (589 amino acids / 1767 bp):

  • GC Content: Changed from 65.1% (native C. necator) to 55.3% (optimized for E. coli).
  • Codon Modifications: 264 out of 589 codons were modified (44.8% of codons adjusted).

Kernel’s optimization and simulation tools do not include Cupriavidus necator, so I used E. coli as a working proxy. This is a practical choice, not a claim that the two hosts behave alike. E. coli is fast to grow and transform, and it is known to make PHB when given phaCAB genes, so it is a reasonable first testbed. But the two organisms differ in ways that matter for this construct: genome GC content (~66% in C. necator vs. ~51% in E. coli, which changes codon preference), promoter and RBS recognition, and central metabolism, including how much acetyl-CoA and NADPH are available to feed the pathway. Every result from Kernel should therefore be read as “the cassette is expressed in a model bacterium,” not as a prediction of PHB yield in my real host.

Part 5: From design to build

The cassette is ~4 kb (CDS ≈ 1.8 + 1.2 + 0.7 kb, plus regulatory parts) and the backbone is ~5 kb. I would order the three genes as synthetic fragments with overlapping ends and assemble them into the linearized backbone by Gibson assembly (Gibson et al., 2009). Ordering one 4 kb piece is riskier and costlier, and fragment assembly lets me sequence-verify each piece. Hard-to-synthesize features: high GC content and the three identical B0034 sequences, which are repeats that can cause mis-assembly; I would swap in different RBS variants. Chemical synthesis has errors, so I would verify clones by sequencing or use error correction (Carr et al., 2004). Prototype vs. scale: for one construct, DNA cost is negligible. If I test many RBS/promoter variants, oligo pools from chips (Kosuri et al., 2010) lower the cost per gene. At production scale, I would integrate the operon into the genome to avoid antibiotics in the fermenter.

Part 6: What I don’t know yet

1. I don’t know how to test this in C. necator. Kernel only lets me simulate and optimize for E. coli, so everything I have shown is a stand-in for my real host. I also don’t know how to get a plasmid into C. necator. I have read that it is usually done by conjugation from an E. coli donor strain, but I have never done it and I don’t know the protocol. How I would find out: read a published C. necator transformation protocol, ask a lab that works with it, and start with a small test using a plasmid that carries only a fluorescent reporter before moving to the full operon.

2. I don’t know whether my recoded genes will work. I optimized the codons for E. coli, but C. necator has a very different GC content (about 66% vs. 51%), so the sequence may translate well in one host and poorly in the other. I don’t know how to predict this, and I’m not sure codon optimization alone is enough. How I would find out: build a second version that keeps the native C. necator codons, express both in the same strain, and compare protein levels on a gel and PHB per cell. I would also learn to use a codon-usage table for C. necator and re-run the optimization with it.

3. I don’t know if the three enzymes are in the right balance. All three genes share one promoter and the same RBS sequence, so I expect similar translation for each, but the pathway may need more of one enzyme than the others. The Kernel simulation shows steady protein levels in relative units, but it does not model enzyme activity, folding or PHB yield, and I don’t yet know how to connect its output to real production. How I would find out: measure PHB directly (I would need to learn a Nile red assay and then GC or HPLC for confirmation), then vary the RBS of each gene and see which one limits yield.

4. I don’t know if a plasmid is the right format for scale. The pBBR1MCS-2 backbone needs kanamycin, which is expensive and undesirable in a production fermenter, and I don’t know how stable the plasmid would be over a long culture. I also don’t know how to integrate the operon into the C. necator genome. How I would find out: grow cultures for many generations with and without antibiotic and count how many cells keep the plasmid, then learn a genome-integration method and compare yield.