Escherichia coli K-12 strain TOP10F′ (Invitrogen, UK) and MC4100 (Casadaban, 1976) were used for plasmid-based protein expression. MC4100ΔtatA/E (Sargent et al., 1998) was used when indicated. When grown in culture, cells were grown in LB medium in a shake incubator in the presence of ampicillin (100 μg/ml). Cultures of cells containing pOFX-tac (derivative) plasmids also contained kanamycin (50 μg/ml) and glucose (0.2%). Cultures were grown at 37°C in tubes or flasks in a shake incubator (200 rpm) using a 5:1 tube/flask:culture volume ratio. Growth on solid medium was performed at 37°C making use of LB-agar (1.5%) plates supplemented with ampicillin (100 μg/ml). Plates also contained chloramphenicol (30 μg/ml) if indicated.

The Rapid Dephos & Ligation Kit was obtained from Roche Applied Science. Phusion High Fidelity DNA polymerase was purchased from New England Biolabs (New England Biolabs; NEB). Low fidelity DNA polymerase Pfu-Pol(exo⁻) D473G (Biles and Connolly, 2004) was obtained as a kind gift from B. Connolly (University of Newcastle, Newcastle upon Tyne, UK). DNA restriction enzymes were purchased from Roche or NEB. Anhydrotetracyclin was from IBA GmbH. Isopropyl β-D-1-thiogalactopyranoside (IPTG) as purchased from Bioline Reagents Ltd. Lysozyme and all other chemicals were purchased from Sigma Aldrich.

The plasmids and primers used in this study can be found in Supplementary Tables 2, 3, respectively. Details on plasmid construction can be found as Supplementary Materials and Methods.

Plasmid-based protein expression was induced using anhydrotetracycline (0.2 μg/ml) when cell cultures reached an OD₆₆₀ of ~0.3. For analysis, cells and cell-fractions were resuspended in SDS-sample buffer (125 mM Tris–HCl pH 6.8, 4% SDS, 20% glycerol, 0.02% bromophenolblue, 83 mM DTT) and incubated at 96°C for 10 min. Proteins were analyzed by SDS-PAGE and Coomassie Brilliant Blue G (Jansen Chimica) staining. Commercial Bis–Tris NuPAGE (Invitrogen) or TGX gels (Biorad) were used where appropriate. Imaging of Coomassie-stained gels was carried out using a GS-800 densitometer (Biorad) in combination with Quantity-One software (Biorad).

To separate IBs from the soluble cell content, a culture volume containing the number of cells that gives an OD₆₆₀ of 1.5 in a 1 ml suspension was subjected to centrifugation. The pelleted intact cells were resuspended in 750 μl ice-cold lysis buffer (5 mM Tris–HCl, pH 7.6, 1 mM EDTA, 100 mM NaCl). Lysozyme was added to a final concentration of 17 ng/ml and cells were incubated on ice for 15 min. Subsequently, the cells were disrupted by freeze thawing and tip sonication (Branson Sonifier 250). The resulting lysate was centrifuged (4,500 × g, at 4°C for 10 min) to sediment IBs and other dense, insoluble material. The resulting pellet was subjected to SDS-PAGE analysis directly, whereas the supernatant was trichloroacetic acid precipitated first. Intact cells directly subjected to SDS-PAGE analysis served as a control for total cell content.

For microscopy analysis, cells were fixed in-culture with formaldehyde (3%) at room temperature for 15 min, collected by low-speed centrifugation and re-suspended in PBS. Cells were then photographed with an Olympus F-view II camera mounted on an Olympus 5 BH-2 microscope through a DApo100UV PL 1.30 oil 160/0.17 objective.

Cells carrying either pIBA-ssTorA/TrxA, pIBA-ssTorA(Δ5–9)/TrxA, or pIBA-ssTorA(Δ5–34)/TrxA were induced with anhydrotetracycline for 1 h. Total RNA was isolated from 1.5 OD₆₆₀ units of cells using the RNeasy Mini kit (Qiagen), RNAprotect Bacteria reagent (Qiagen) and DNaseI treatment (Qiagen) according to manufacturer's protocols. For quantitative PCR (qPCR), we used an iCycler and SYBR green Supermix (Bio-Rad, Hercules, CA, US). Expression of trxA in each sample was analyzed in duplicates, and expression levels were normalized to the expression of bla gene present as an antibiotic resistance marker on the pIBA vector backbone. Primer pairs used were qPCR TrxA fw/qPCR TrxA rv and qPCR bla fw/qPCR bla rv both yielding ~160 bp products.

To create a library of randomly mutagenized DNA encoding ssTorA, error-prone PCR was carried out using Pfu(exo⁻) D473G polymerase, Pfu ultra HF reaction buffer (Agilent), 600 μM of dNTPs and 20 mM of added MgSO₄. Plasmid pIBA-ssTorA/TrxA (Jong et al., 2017) was used as the template and the primers were ssTorA mut fw and ssTorA mut rv. The PCR product was XbaI/NheI digested and the resulting ssTorA^mut fragment–encoding residues 1–36 of ssTorA and the upstream Shine-Dalgarno sequence of the template—was ligated to the corresponding sites of pIBA-ssTorA/CAT. Following electroporation of the ligation mixture into E. coli TOP10F′ cells, bacterial colonies were scrape-harvested from LB-agar plates containing ampicillin and pooled. The resulting cell material was suspended in LB with glycerol (12%) to establish a TOP10F′(pIBA-ssTorA^mut/CAT) library. Prior to harvesting, plasmid DNA from individual colonies was subjected to DNA sequencing and a mutation rate of 1–5 mutations per ssTorA encoding fragment was established for 90% of the pIBA-ssTorA^mut/CAT ligation products.

Liquid medium was inoculated with part of the TOP10F′(pIBA-ssTorA^mut/CAT) library cell suspension to an OD₆₆₀ of 0.05 and incubated at 37°C with shaking. When the culture reached an OD₆₆₀ value of 0.3 cells were induced for expression of ssTorA^mut/CAT by addition of anhydrotetracycline (0.2 μg/ml) and growth was continued. After 2 h samples were withdrawn from the culture and spread on solid medium with or without chloramphenicol (30 μg/ml), respectively, and the resulting plates were incubated at 37°C. To allow comparison of mutational loads in cells grown in the presence or absence of selective pressure, similar numbers of colonies (10,000 and 16,000 CFUs) were harvested from plates with and without chloramphenicol. For either condition, colonies that formed overnight were scrape-harvested and pooled. Plasmid DNA was isolated (GeneJet, ThermoFisher Scientific) from 5 OD₆₆₀ units of cells of either suspension as well as from 5 OD₆₆₀ units of original TOP10F′ (pIBA-ssTorA^mut/CAT) library cells not subjected to the chloramphenicol survival assay. Using the respective plasmid isolates as a template, a 222 bp fragment including the ssTorA region of interest was PCR amplified using primers ssTorA MiSeq fw and ssTorA MiSeq rv. Following QIAquick gelextraction (Qiagen, Hilden, Germany) the amplicons were purified using the Agencourt AMPure XP PCR Purification system (Beckman Coulter, Pasadena, CA, US) and the resulting starting DNA library was quantified using the Qubit dsDNA BR Assay system (Invitrogen, Carlsbad, CA, US). The sequencing library construction was performed according to the TruSeq Nano DNA sample prep kit (Illumina, San Diego, CA, US). The barcoded libraries were subsequently mixed in equimolar amounts. The multiplexed library pool was spiked with 40% PhiX control to improve base calling during sequencing and was loaded at 7.5 pM. Sequencing was conducted using a paired-end, 2 × 250-bp cycle run on an Illumina MiSeq sequencing system and MiSeq Reagent Kit V2 (500 Cycle) chemistry. After sequencing was complete, image analysis, base calling, and error estimation were performed using Illumina Real-Time Analysis software (version 2.2.0.2).

We previously found the Tat signal sequence of TorA (ssTorA) to function as an inclusion body (IB) tag (Figure 1A) when fused to a cargo protein and expressed at high levels (Jong et al., 2017). To determine whether this capacity is related to its function as a Tat signal sequence, we substituted the twin arginine motif (RR) of ssTorA that is essential for export via the Tat translocon by a non-conservative alanine pair (RR/AA) (Supplementary Table 1). The resulting ssTorA mutant was fused to TrxA that is soluble in E. coli even at high concentrations (LaVallie et al., 1993; Jong et al., 2017). IB formation of the fusion proteins was analyzed using an IB sedimentation assay. Briefly, E. coli cells expressing the fusion constructs were broken and the lysate was subjected to low-speed centrifugation to separate the dense cell material including IBs from the soluble proteins. The resulting fractions and corresponding whole-cell lysate samples were analyzed by SDS-PAGE and Coomassie staining (Figure 1B). As we demonstrated previously (Jong et al., 2017), the native ssTorA sequence drives efficient IB formation, as evident from the presence of the majority of the ssTorA/TrxA fusion in the low-speed pellet (Figure 1B, lane 2). The RR/AA substitution did not affect the amount of fusion protein in the low speed pellet (Figure 1B, c.f. lane 5 and 2), suggesting that Tat-mediated targeting is not a prerequisite for the functioning of ssTorA as an IB formation tag. Moreover, we expressed ssTorA/TrxA in a strain that lacks the key Tat translocon components TatA and TatE (Sargent et al., 1998). Also in this genetic background, most ssTorA/TrxA fusion protein ended up in the low speed pellet (Figure 1C, lane 2). Together, the data show that ssTorA-mediated IB formation does not require a functional Tat pathway.

To analyze whether the capacity to induce IB formation is a universal feature of Tat signal sequences, a panel of these was selected for fusion to TrxA. The panel represents a cross-section of E. coli Tat signal sequences, ranging from relatively short sequences (e.g., ssCueO, 28 aa) to long ones (ssYagT, 53 aa), and including sequences known to bind a cognate chaperone (ssTorA, ssDmsA, FdnG, ssHyaA, ssHybO, NapA) (Chan et al., 2009) and ones that do not (Supplementary Table 1). Despite being cloned in the same expression context, the various signal sequences affected expression differently (Figure 2A). Three chimeras did not show detectable levels of expression (ssFdnG, ssHybO, ssNapG). Another six showed reduced levels compared to the ssTorA/TrxA benchmark (AmiC, ssHyaA, ssNapA, ssWcaM, ssYbaK, ssYagT) whereas two yielded high levels of expression, comparable to the ssTorA/TrxA (ssCueO, ssDmsA). The origin of these differences is unclear, but seems unrelated to length or chaperone binding. Next, we selected TrxA fusions of two chaperone binding (ssDmsA and ssHyaA) and two non-chaperone binding signal sequences (ssCueO and ssWcaM) for analysis of IB formation (Figure 2B). TrxA fused to ssDmsA, ssHyaA, and ssWcaM was predominantly found in the soluble fraction in contrast to ssCueO/TrxA that appeared in the insoluble IB fraction similar to ssTorA/TrxA (Figure 2B, lane 11). The IB induction by ssCueO is surprising as it shares no obvious features with ssTorA. In comparison, ssDmsA seems more closely related to ssTorA with respect to length, hydrophobicity and chaperone binding (Tullman-Ercek et al., 2007; Shanmugham et al., 2012). In conclusion, the features that determine the IB formation capacity of ssTorA are not inherent to all Tat signal sequences.

In native E. coli cells ssTorA functions as a signal sequence to trigger targeting and translocation of TMAO reductase (TorA) via the Tat export pathway (Sargent, 2007). A dedicated cytoplasmic chaperone TorD is co-expressed with TorA when TMAO is present in the medium (Mejean et al., 1994). TorD interacts with ssTorA and co-ordinates the export of the reductase by preventing its premature targeting to the bacterial inner membrane (Hatzixanthis et al., 2005). In our assays, constructs are expressed under conditions of low availability of TorD as no TMAO is added to the growth medium. We therefore considered the possibility that the lack of TorD plays a role in the aggregation of fusions carrying ssTorA. To investigate this, ssTorA/TrxA was produced in cells already expressing high-levels of TorD from a compatible expression plasmid (Figure 3A). Under these co-expression conditions the fusion protein was still efficiently produced but recovered almost exclusively from the soluble cell fraction (Figure 3A, lane 6). In contrast, control cells not induced for TorD expression produced similar amounts of ssTorA/TrxA but located in the insoluble fraction as before (Figure 3A, lane 2). Apparently, overexpressed TorD binds to ssTorA and precludes ssTorA/TrxA IB formation.

Previously, it was shown that ssTorA also interacts with the generic cytosolic chaperone DnaK (Oresnik et al., 2001). Accordingly, using the LIMBO algorithm (Van Durme et al., 2009) DnaK binding sites were predicted for ssTorA at residues 12–18 (RRFLAQL) and 30–36 (SLLTPRR). To investigate whether a disbalance between DnaK and ssTorA under the overexpression conditions used may influence the aggregation process, we monitored IB-formation upon co-overexpression of the chaperone. To this end, ssTorA/TrxA was produced in cells that also express high levels of plasmid-encoded DnaK and its co-chaperone DnaJ (Figure 3B). Insoluble production of ssTorA/TrxA could be confirmed in control cells carrying either an empty co-expression plasmid (Figure 3B, lane 2) or in cells co-overexpressing the unrelated maltose binding protein (MBP) (Figure 3B, lane 8). In contrast, a strongly reduced amount of insoluble ssTorA/TrxA was recovered from cells expressing excess amounts of DnaK/J (Figure 3B, lane 5), indicating impaired IB formation under these conditions. Conceivably, like TorD, DnaK associates with ssTorA and interferes with the aggregation process.

We next focused on sequence motifs and regions not related to the Tat system that influence IB formation by ssTorA. The triple asparagine sequence (NNN) directly downstream of the initiator methionine (Supplementary Table 1) was considered potentially important because prions typically possess a region rich in asparagine and/or glutamine residues that enhances their aggregation (Du, 2011). Initially we substituted the asparagine residues by alanine residues. However, this modification was found to impair expression (data not shown), possibly due to the codon usage in the encoding mRNA. Whereas, in prokaryotes A/T-rich codons are preferred around the start of the coding sequence, alanine codons intrinsically have a high G/C-content, which may interfere with efficient translation initiation and expression (Kozak, 2005). To prevent impaired expression, the NNN motif was substituted with a KTK sequence as present in ssDmsA (Figure 4A; Supplementary Table 1) that sustained high expression levels upon fusion to TrxA but without inducing IB formation (Figure 2). With the NNN/KTK substitution incorporated, expression and IB formation of ssTorA/TrxA was on par with the wild-type ssTorA/TrxA (Figure 4A). Vice versa, replacement of the KTK motif of ssDmsA by NNN did not induce insoluble expression of ssDmsA/TrxA fusion (data not shown). Together, the data indicate that the NNN sequence in ssTorA is neither required nor sufficient for IB formation. The presence of A/T rich codons at the start of ssTorA seems important to achieve high-level expression, which likely is a prerequisite for efficient IB formation.

Rather than mutating individual motifs in ssTorA, we next applied an unbiased deletion strategy to identify sequences that are critical for IB formation. We set out making deletions based on the natural N-H-C domain structure of the signal sequence (see Figure 1A). The basic N-domain of ssTorA (AA 2–13) was removed, leaving the C-terminal domain with hydrophobic core and polar C-domain intact [ssTorA(C′)] (Supplementary Table 1). Alternatively, we kept the N-domain intact and removed the C-terminal part of ssTorA (except C-terminal ASA; AA 14–36)[ssTorA(N′)]. A TrxA fusion carrying ssTorA(N′) was exclusively expressed in the soluble fraction (Figure 4B, lane 9). In contrast, ssTorA(C′)/TrxA accumulated in IBs (Figure 4B, lane 5), although the yield appeared relatively low for this construct due to reduced expression (Figure 4B, c.f. lanes 4 and 1). In an attempt to enhance the expression of ssTorA(C′)/TrxA we reinserted the N-terminal NNN motif (Supplementary Table 1), the codon usage of which appeared favorable for translation initiation (see above). Indeed, improved expression of the resulting ssTorA(C′+NNN)/TrxA chimera was observed restoring the IB yield to the level of ssTorA/TrxA (Figure 4C, c.f. lanes 4–5 and 1–2). In conclusion, the C-terminal part of ssTorA (AA 14–39) appears sufficient for induction of IB formation.

To further narrow down the regions important for IB formation, we followed a strategy in which incremental parts of ssTorA in the ssTorA/TrxA construct were deleted, starting from the N- and C-terminus, respectively. The set of N-terminal deletion mutants started at position 5 to keep the NNN motif at position 2–4 intact for optimal expression. The number of deleted residues increased by 5 for each consecutive mutant (Supplementary Table 1). Remarkably, whereas fusions carrying deletions up to residue 9 or extending beyond residue 29 were efficiently expressed (Figure 5A, lanes 1–2, 6–9), versions with deletions Δ5–14, Δ5–19, and Δ5–24 were hardly detectable in whole cell lysates by SDS-PAGE (Figure 5A, lanes 3–5). To investigate whether the poor expression was due to problems at the mRNA level, we performed quantitative PCR using trxA-targeting primers to determine the levels of mRNA encoding the non-detectable ssTorA(Δ5–19)/TrxA vs. the strongly expressed ssTorA/TrxA and ssTorA(Δ5–34)/TrxA (Supplementary Figure 1). A considerable reduction in encoding mRNA was observed specifically for ssTorA(Δ5–19)/TrxA, which might suggest that deletions Δ5–14, Δ5–19 and Δ5–24 introduce mRNA instability, explaining their poor expression. The highly expressed constructs were analyzed for IB formation as before. In line with the above observation that the N-terminal part of ssTorA is dispensable for the production of IBs (see Figure 5A), insoluble expression was observed for ssTorA(Δ5–9)/TrxA (Figure 5B, lane 5). In contrast, fusions carrying deletions Δ5–29, Δ5–34 or Δ5–39 were soluble (Figure 5B, lanes 9,12,15). It follows that a region critical for the induction of IBs must be localized downstream of residue 9 and upstream of residue 29 of ssTorA.

A similar deletion strategy was applied on ssTorA but now starting from the C-terminus. Leaving the C-terminal ASA motif intact, we removed residues 36 to 17 by deleting three or four residues per consecutive mutant (Supplementary Table 1; Figure 6A). Efficient expression was now found for all ssTorA/TrxA mutants allowing analysis of IB formation. Deletions up to proline 29 of ssTorA (Δ29–36) did not interfere with IB formation (Figure 6A, lane 8) whereas derivatives with larger deletions were completely soluble (Figure 6A, lanes 11, 14, 17). Next, the largest N- and C-terminal deletions that still formed IBs were combined in one ssTorA/TrxA construct (Supplementary Table 1) and tested (Figure 6B). Indeed, ssTorA(Δ5–9/Δ29–36) still induced IB formation to a similar extent as ssTorA(Δ5–9) (Figure 6B, c.f. lane 11 and 5) but less efficiently than the intact ssTorA sequence (Figure 6B, lane 2). This suggests that residues 5–9 are not required for IB formation per se, but do enhance aggregation. Together, the data limit the region of ssTorA required for IB formation roughly to residues 10–28.

Given the importance of residues 10–28 that roughly correspond to the hydrophobic H-domain of ssTorA (AA 14–32) (Cristobal et al., 1999), a potential scenario for the initiation of IB formation would involve hydrophobic interactions between neighboring ssTorA sequences or between ssTorA and fused cargo proteins. However, the hydrophobicity of ssTorA is relatively modest and similar or lower than that of other Tat signal sequences that did not form IBs (Tullman-Ercek et al., 2007) (see Figure 2). Alternatively, the tendency of signal sequence H-domains to attain an α-helical conformation was considered. Indeed, substitution of Ala16 in the core of ssTorA by a proline (A16P) prevented insoluble expression of ssTorA/TrxA (Supplementary Figure 3B, c.f. lanes 2 and 5), suggesting that α-helix formation is a prerequisite for IB formation. Interestingly, helical wheel representation of the ssTorA revealed a concentration of hydrophobic residues at one side of the helix (F7-F14-L18-L21) (Supplementary Figure 3A, boxed). Non-conservative substitution of either one (F14 > S14) or two (F7 > A7 + F14 > S14) of these residues reduced the amount of ssTorA/TrxA recovered in the insoluble cell fraction (Supplementary Figure 3C, lanes 5 and 8), suggesting the importance of this feature in IB formation. Possibly, the imposed clustering of hydrophobic residues nucleates the assembly of fusion proteins into IBs.

To identify residues important for IB induction in a more unbiased way, we developed a library screen with randomly mutagenized ssTorA (Supplementary Figure 4). Chloramphenicol acetyl transferase (CAT) was used as a fusion partner for the ssTorA library reasoning that ssTorA/CAT can only confer resistance toward the antibiotic chloramphenicol when expressed in soluble form (Maxwell et al., 1999). Hence, only cells with mutant ssTorA that has lost its capacity to deposit CAT in inclusion bodies are expected to grow on chloramphenicol containing plates.

First, we confirmed that CAT is expressed in IBs when fused to wild-type ssTorA (Supplementary Figure 5, lane 6), whereas the majority of fusion protein remains soluble when a non-functional variant [ssTorA(N′)] is used (Supplementary Figure 5, lane 2). Next, we determined the conditions that allow on-plate discrimination between cells expressing the soluble and insoluble CAT fusions. To ensure optimal protein expression and IB formation cells were first grown and induced in liquid culture before being spread on LB-plates containing chloramphenicol for overnight selection of resistant cells (Supplementary Figure 4). When testing cells expressing either insoluble ssTorA/CAT or soluble ssTorA(N′)/CAT on plates without the antibiotic, growth was observed in both cases (Supplementary Figure 5B). In contrast, growth on LB-agar supplemented with 30 μg/ml chloramphenicol yielded colonies only for the cells expressing the soluble ssTorA(N′)/CAT fusion (Supplementary Figure 5B), providing proof of concept for the screening assay.

Next, we fused a library of randomly mutagenized ssTorA fragments to CAT and used the on-plate chloramphenicol survival assay to select for cells that express soluble ssTorA/CAT mutants. Next-generation sequencing was used to define the mutations in individual ssTorA encoding DNA fragments (Supplementary Figure 4). In our analysis we focused on a 138 bp sequence including 21 upstream bases that contain the ribosome binding site. Interestingly, mutation hotspots were identified in DNA obtained from chloramphenicol resistant cells for the R11, R12, L15, L18, and G19 codons (Supplementary Figure 6). In all these cases the mutations were in the second nucleotide of the codons, consequently resulting in amino acid substitutions. Importantly, the detected bias against mutagenesis in the Shine-Dalgarno sequence, which is needed for efficient expression of ssTorA/CAT, is consistent with the validity of the screening procedure. Taken together, we identified R11, R12, L15, L18, and G19 of ssTorA as potentially critical residues for IB formation (Figure 7).

A bias for specific changes in the mutational hotspots suggested amino acid substitutions potentially responsible for enhanced ssTorA/CAT solubility (Supplementary Figure 7). To further validate the screening procedure and analyze the importance of R11, R12, L15, L18, and G19 these residues were individually replaced with the selected residues (Figure 7). Using the same conditions as above, all mutants indeed induced a chloramphenicol-resistant phenotype in E. coli on plate (Supplementary Figure 8), implying expression of ssTorA/CAT in a (partially) soluble fashion.

Next, we analyzed the solubility of the mutant fusion proteins using the sedimentation assay (Figure 8). Impaired IB formation compared to non-mutated ssTorA/CAT (Figure 8, lanes 19–33) could be confirmed for L15, L18, and G19 mutants, which were recovered largely in the soluble fraction (Figure 8, lanes 19–33). In contrast, changes at R11 and R12 did not clearly affect IB formation despite the use of rather non-conserved substitutions (Figure 8, c.f. lanes 1–15 and 16–18). As a control, we also substituted L15, L18, and G19 with residues that were expected not to interfere with the functionality of the tag based on the Next-generation sequencing analysis (Figure 7; Supplementary Figure 7). Indeed, the resulting ssTorA/CAT fusions (M15, Q17, I18, V18, and G19) accumulated almost exclusively in IBs (Supplementary Figure 9A). Together, these data suggest that our method effectively discriminates between solubilizing and non-solubilizing mutations in ssTorA/CAT.

To exclude that our findings were specific for fusions to CAT, the same substitutions were tested in the context of a second cargo protein, TrxA, with very similar results (Supplementary Figures 9B, 10). Whereas mutation of positions R11 and R12 did not affect IB formation (Supplementary Figure 10, c.f. lanes 1–15 and 16–18), substitutions at positions 15, 18, or 19 resulted in elevated levels of soluble ssTorA/TrxA (Supplementary Figure 10, c.f. lanes 19–33 and 37–39) except for M15, Q17, I18, V18, and G19 (Supplementary Figure 9B). In conclusion, our combined mutagenesis and screening approach identified five potentially important residues for ssTorA-mediated IB formation. Three of these (L15, L17, and G19) were confirmed to be critical using IB sedimentation analysis.

Previously, we have shown that suboptimal deposition of cargo proteins in IBs could be enhanced by fusing three ssTorA in tandem [ssTorA(3X)] (Jong et al., 2017). Since for biotechnological applications a short IB tag is preferred, a triple tag of reduced length was made by linking three ssTorA sequences lacking residues 29–36, which appeared dispensable for IB formation (see under Regions Critical for IB Formation) (Supplementary Table 1). To test ssTorA(Δ29/36|3x) it was not only fused to TrxA but also to another highly soluble E. coli protein, MBP, and to a small recombinant protein of human origin, hEGF. All fusion proteins were almost exclusively detected in the insoluble pellet upon sedimentation of IBs (Figure 9A, lanes 8, 11, 14), demonstrating that ssTorA(Δ29/36|3x) indeed functions as an IB formation tag with a higher efficiency than single ssTorA(Δ29/36). Surprisingly, when compared side-by-side to TrxA and MBP carrying full-length ssTorA(3X), fusions to the new truncated tag were expressed at a higher level (Figure 9A, c.f. lanes 7 and 1; c.f. lanes 10 and 4) and yielded higher amounts of fusion protein in IBs (Figure 9A, c.f. lanes 8 and 2; c.f. lanes 11 and 5). Consistently, analysis of cells by phase contrast microscopy showed more pronounced IB structures for the fusions carrying the truncated triple tag (Figure 9B). To exclude that the altered (E. coli optimized) codon usage of the synthetic ssTorA(Δ29/36|3x) coding sequence is responsible for the effect, ssTorA(Δ29/36|3x; NO) with a codon usage identical to that of the full length ssTorA(3X) (Supplementary Figure 2A) was shown to induce similar elevated expression and IB yields (Supplementary Figure 2B). In conclusion, the shortened triple fusion tag has improved IB-formation properties compared to the original ssTorA(3X) sequence.

WJ, J-WG, and JL are involved in Abera Bioscience AB that aims to exploit the ssTorA inclusion body formation tag for biotechnology purposes. DV is employed at Xbrane Biopharma AB, a forerunner of Abera Bioscience AB. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.