Specific N-terminal amino acids potentiate the periplasmic expression of single-chain variable fragments in Escherichia coli

Antibodies are indispensable for therapy, diagnosis and research. Recombinant antibody fragments can be easily engineered and small, single-chain variable fragments (scFvs) have proven particularly advantageous [1,2]. Escherichia coli is the standard scFv expression host; however, the yield is low due to formation of inclusion bodies and cell lysis [3,4]. Various approaches have been attempted to improve yields, including fed-batch culture [5], a cold-shock promoter [6], a solubility-enhancing tag [7–9], host selection [10] and specific N-terminal sequences [11], yet none of these methods is sufficiently versatile in E. coli.

Here we describe using a single-step colony assay [12,13] to efficiently screen scFvs whose N-terminal sequence enhanced periplasmic expression in E. coli. Three specific amino acid sequences capable of boosting periplasmic expression of scFvs were identified.

A random library (Figure 1A) was constructed by inserting the degenerate codons (NNK)₃, corresponding to three N-terminal amino acids, between the pelB signal peptide and anti-rabbit IgG scFv (A10B) [14]. The pET-22b expression vector containing the random library was then transformed into E. coli BL21 (DE3) cells.

The procedure used in the single-step colony assay to screen for clones with higher expression is depicted schematically in Figure 1B. A hydrophilic polyvinylidene difluoride filter and an antigen-coated nitrocellulose membrane were placed on agar plates. E. coli transformed with the library were grown to exponential phase and suspended in autoinduction solution containing glucose and lactose. The cells were then spread on the filter and cultivated until colonies appeared on the filter surface. ScFvs were automatically expressed by allolactose once glucose was depleted. The expressed scFvs diffused into the antigen-coated membrane, where those with affinity bound to the antigen beneath the colonies. The filter containing the colonies was removed, placed on a fresh plate and stored for later recovery of the clones. Antigen-bound scFvs on the membrane were detected by chemiluminescence using a horseradish peroxidase-conjugated anti-His antibody. The filter harboring the colonies was superimposed on the image with chemiluminescence results. Clones with stronger signals were identified and their reactivity against the antigen was measured by ELISA. Plasmids encoding scFvs were purified and their scFv and N-terminal sequences were read.

Figure 2 shows a nitrocellulose membrane resulting from the single-step colony screening. E. coli were transformed with the scFv library, and 2 × 10⁴ cells were spread on ten 15-cm plates. After incubation for 16 h at 30°C, 1.16 × 10⁴ colonies were detected, corresponding to approximately 1000 colonies per plate. A10B scFvs with the three N-terminal amino acids obtained from each colony bound to the antigen on the nitrocellulose membrane and were detected by chemiluminescence. The library generated from (NNK)₃ degenerate codons was estimated to contain 20 × 20 × 20 = 8 × 10³ clones. We screened 1.16 × 10⁴ clones, which was sufficient to cover the library size. Signal intensity, corresponding to scFv expression, varied between colonies. The 25 colonies with the highest intensities (indicated by arrow in Figure 2) were selected for further experiments.

The 25 clones were picked from the upper filter and cultured overnight at 37°C in Luria–Bertani medium (Sigma-Aldrich, MO, USA) containing 50 μg/ml ampicillin. The resulting culture was inoculated in a 200-ml flask containing 20 ml Terrific Broth medium (Sigma-Aldrich) supplemented with 50 μg/ml ampicillin, and incubated at 37°C and 200 r.p.m. Once the optical density at 600 nm reached 0.5, expression was initiated by addition of isopropyl β-d-1-thiogalactopyranoside to a final concentration of 0.5 mM. The cells were incubated overnight at 26°C, harvested by centrifugation at 6000 × g and 4°C, washed with phosphate-buffered saline, and the periplasmic fraction from the pellet was extracted by cold osmotic shock.

ScFv activity was measured by ELISA. Rabbit IgG was used to coat a 96-well ELISA plate, and the periplasmic fraction was applied. Antigen binding was detected using a horseradish peroxidase-conjugated anti-His-tag antibody. As shown in Figure 3, the signal from clones 7, 8 and 12 was more than 1.7-times higher than that of the control scFv without the N-terminal sequences.

The DNA of each scFv from the 25 clones was sequenced. Every clone contained a complete scFv structure consisting of heavy and light chain variable domains plus a linker. The N-terminal amino acid sequences of the clones are listed in Table 1.

Amino acids at each position were classified on the basis of their characteristics (Table 2). In the first position, most amino acids (68%) were either alanine (n = 10) or glycine (n = 7), indicating a bias toward smaller residues. In the second position, most amino acids (56%) were either glutamic acid or aspartic acid, suggesting a preference for residues with a negative charge. In the third position, all amino acids were either neutral or polar, with tyrosine being the predominant species (32%). No preference for charge or polarity was found at this position. The frequencies of amino acids at each position are listed in Table 3. Alanine in the first, glutamic acid in the second and tyrosine in the third position were the most frequent amino acids, with probabilities of 40, 40 and 32%, respectively.

For a quantitative comparison of scFv expression, the top three sequences (AND, AEY and GDN) and a synthetic sequence (ADY), comprising the most frequent amino acid at each position, were selected. The cells were cultured as described previously, and the scFvs in the periplasmic extract were purified using a Ni column. Cell growth, volumetric yield and relative yield were also analyzed (Table 4). Although the numbers of cells were almost identical in all cases, the volumetric yield was two-times larger and the relative yield (i.e., scFvs expressed per cell) was 2.2-times larger with ADY at the N-terminus. The other three sequences also presented expression levels two-times higher than that of the control. These results showed that specific amino acid sequences enhanced periplasmic scFv expression. Given that ADY achieved the highest yield, our screening was successful.

A single-step colony assay for antibody fragments was developed previously by our group to quickly and easily identify scFv clones with high affinity against antigens [12]. Here, the assay was used successfully to identify clones in a library of 1 × 10⁴ clones. We used the degenerate codon NNK to generate a library with random sequences. This method can be easily applied in various contexts because it necessitates only primers with degenerate codons to introduce diversity [15]. In this study the probability of a stop codon being introduced by NNK was 9.0%, which was within the acceptable range. Screening with a longer random sequence could help identify a more effective sequence for periplasmic scFv expression. However, NNK is not suitable for longer sequences because the probability of a stop codon increases substantially with length, thereby interfering with the screening. To avoid this scenario, trimer oligos can be used to construct a highly diverse library [16].

Here we identified specific N-terminal sequences that increased periplasmic expression. In the future, it will be important to confirm this effect on other scFvs and proteins, as well as to determine whether the identified sequence is valid for other signal sequences. This would help elucidate the mechanism underlying the observed potentiation.

The second amino acid, after methionine, is important for the cytosolic expression of recombinant proteins. Specifically, alanine, cysteine, proline, serine, threonine or lysine at the +2 position may result in significantly higher (up to tenfold) expression of recombinant Ig-α, while methionine, histidine or glutamic acid may have the opposite effect [17]. Ojima-Kato et al. reported that N-terminal SKIK augmented the cytosolic expression of scFvs [11]. Kim et al. reported that N-terminal DDDDD improved the secretion of recombinant proteins [18,19]. The fusion of green fluorescent protein, which contains 15 negatively charged amino acids, to scFvs also improves secretion [20]. The effect of N-terminal sequences on secretion should be addressed in the future.

In summary, we constructed a library with three random amino acids at the N-terminus of scFv, and screened this library using a simple and rapid single-step colony assay. Three specific amino acids capable of potentiating the periplasmic expression of scFvs were identified.