A plasmid library is provided for the surface display of proteins, to be used by experts and non experts, including a simple option for optimization. Based on a classical autotransporter protein, variants were constructed with different signal peptides, linkers and adjacent regions, set under control of a rhaBAD promoter, resulting in a library of 24 plasmids named Autodisplay ToolBox (ATB). For proof of concept, a β-glucosidase as model passenger was subjected to combinatorial permutation using the plasmids of the ATB (pATB). The best combination based on enzymatic activity was pATB_45-β-Gluc with an OprF signal peptide, a C terminal spacer (GGGDDNAAPA), and a short linker (ΔepitopeΔβ1). Activity was 4.9-fold increased compared to the initial construct. ATB can be extended to 72 variants by using two additional promoters (PRox306, ParaBAD) and to 81 variants by using an inverse autotransporter protein (YeeJ). Besides using the pATB library, individual plasmids can be picked for custom-made surface display.
Article
Autodisplay made easy: a plasmid toolbox for the surface display of recombinant proteins and its optimization.
https://doi.org/10.21203/rs.3.rs-4899034/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
The development of phage display and its application for the design of peptides and antibodies was awared the nobel prize in chemistry 2018 to George Smith and Gregory Winter1-3. Bacterial display can be used similarily for the screening of protein and peptide libraries as well as for the identification of protein ligands. Beside the option to create a new whole cell biocatalyst by displaying enzymes, bacterial display could have advantages with regard to phage display in the size of the protein displayed, in the number of displayed proteins, and in the applicable screening methods. Moreover a selected variant displayed on a bacterium can be multiplied without further ado, in contrast to phage display, for which repeated cycles of infection and selection are required to enrich a preferred variant4.
For Gram-negative bacteria, like E. coli, a simple method for the surface display of recombinant proteins is based on the secretion mechanism of type Va autotransporter proteins (AT), the so‑called classical AT. In analogy, this method was referred to as autodisplay5,6.
Classical ATs are synthesized as single polydomain/polyprotein precursors that share structural features: a signal peptide (SP) at the N-terminus, followed by a native passenger domain, a linker that usually contains at least one α-helix, and a β-barrel at the C-terminus4,7 (Fig. 1a). Despite other terms existing, the α-helix and the β-barrel are here - as mostly - referred to as “translocation unit” (TU). The SP is cleaved off during the Sec translocon dependent transport across the inner membrane8. Several chaperones (e.g. Skp, SurA) support the transit of the unfolded AT in the periplasm to the β-barrel assembly machinery (BAM)9,10. BAM is involved in folding and outer membrane integration of the TU β-barrel, most probably via the formation of a hybridbarrel pore with BamA8. Through this pore, the passenger domain is translocated to the cell surface, where it remains C-terminally connected to the TU via the linker11 (Fig. 1b). It is worth emphasizing that besides this concept, other models are discussed on b-barrel assembly in the outer membrane12. Albeit, for autodisplay, the protein of interest is replacing the native passenger in the polyprotein precursor. Crucial points for autodisplay are, that a host strain for autodisplay needs to be outer membrane protease T (OmpT) negative (e.g. E. coli BL21, E. coli UT5600), and that the passenger domain is devoid of autoproteolytic capacity to prevent release of the passenger from the surface4. In the past, SP, linker as well as the promoter have separately been investigated to enhance surface display in various bacteria8,13-17.
Here, we present a set of plasmids named the Autodisplay-ToolBox (ATB), that enables the combination of a classical TU with four different SPs and six different linkers in order to find the optimal set for a given passenger. An individual plasmid (pATB) can be picked for the surface display of distinct passengers based on structural considerations in a rational approach. Moreover, in case the rational approach is not successful, a method to screen entire pATBs as a library is provided to identify the best combination with regard to surface display of functional passengers.
To give an impression of the feasibility of this approach, the β-glucosidase CsBglA LYTH (β-Gluc) was taken as a model passenger and in a first embodiment, the pATBs were applied individually, i.e. in a step-by-step (SBS) approach for obtaining maximum activity of β-Gluc. In a second embodiment, all plasmids were applied as a library, starting either on the DNA level, named all-in-one (AIO) approach, or starting with strains containing all the different plasmids, named strain-mix (STM) approach and screened towards maximum b-Gluc activity.
Fig. 1: Schematic structure of a classical autotransporter protein & type Va secretion pathway in Gram-negative bacteria. a, A classical autotransporter (AT) of the Type Va secretion pathway comprises a N-terminal signal peptide (SP), a passenger domain, a linker and a β-barrel. b, After the AT is synthesized as precursor in the cytosol, the SP enables Sec translocon mediated transport to the periplasm. During the transport the SP is removed. While keeping the AT unfolded, chaperones target the AT to the β-barrel assembly machinery (BAM-Complex), which assists the insertion of the β-barrel into the outer membrane. The passenger domain folds into its final conformation and remains attached to the cell surface by the β‑barrel, connected via the linker.
Autotransporter fusion proteins for the Autodisplay-Toolbox
The autotransporter fusion proteins (AT-FPs) applied for the ATB (Fig. 2) contained either an SP from the cholera toxin B subunit (CtxB) from Vibrio cholerae18 or an SP from the outer membrane porin F (OprF) SP from Pseudomonas putida KT244019. The so-called pro-region – 30 aa downstream the SP cleavage site – was previously shown to affect secretion efficiency20,21. The pro-region of Esterase A from Burkholderia gladioli (EstA)22 appeared to be benificial for surface display via AT-FP with both, CtxB and OprF SP16. Hence, the first ten aa GGGDDNAAPA from the pro-region of EstA were taken as an insertion option instead of the complete EstA pro-region and termed ‘spacer’.The pro-region was complemented by an obligatory 6x His epitope (His6) and an obligatory fXa protease cleavage site (fXa). Behind this artificial pro-region, initially, the passenger domain Cel5 from Hahella chejuensis (Uniprot ID: Q2SFD8) was inserted, followed by one of six different linker between passenger and the translocation unit (TU). The first linker variant (called: epitope) contained an OmpT & a factor Xa protease cleavage sites, a linear epitope derived from HIV Nef protein (PEYFK)5 and a myc epitope. The epitope linker variant was complemented by two native domains of the EhaA AT, the β1 domain (β1) and a connecting region (CR)23. For the epitope linker, a random coil structure was predicted. In a second variant (called: flex), the aa sequence G4S-G2S-(G4S)3 was inserted instead of the epitope linker, going along with high flexibility14. In a third variant (called: rigid), the aa sequence A(EAAAK)5 replaced the epitope linker, which was described to form an α-helical structure24, going along with more rigidity. Three deletion linker variants contained either only the β1 and the CR part (called: Δepitope) or the CR part alone (called: ΔepitopeΔβ1) or finally no linker at all, i.e. Cel5 was directly attached to the TU (called: ΔepitopeΔβ1ΔCR). All the variants as described were used with the EhaA AT TU23, which comprised an additional α-helix23 required for plugging the β-barrel after transport.
Plasmids encoding the AT-FP variants (Fig. 2b) were individually constructed based on the MATE (maximized autotransporter-mediated expression) plasmid backbone25. AT-FP expression was set under control of the rhamnose inducible rhaBAD promoter (PrhaBAD), which is established for autodisplay26. This library of 24 plasmids was termed Autodisplay-ToolBox, with each plasmid (pATB) obtaining a specific two-digit identification code (xy), denoting the type of SP +/- spacer (1y – 4y) and the type of linker (x1 – x6) (Fig. 2c).
Fig. 2: Set of ATB fusion protein variants with a classical autotransporter translocation unit. a, The secondary structures of the EhaA translocation unit (TU), the connecting region (CR), the β1 domain (β1) were modeled with RaptorX27 and the variable amino acids in the linker with PEP-FOLD328. Deletion variants of the EhaA linker are denoted. b, Classical autotransporter fusion proteins (AT-FP) contained N-terminal either a cholera toxin B (CtxB) or an outer membrane porin F (OprF) signal peptide (SP). The pro-region (Pro) started after the SP and optionally (*) contained a sequence of the esterase A functioning as a spacer in addition to an obligatory 6x His-epitope (His6) and a factor Xa protease cleavage site (fXa). Linker (light blue) variants with full length contained a sequence of variable amino acids, a β1, a CR and the α-helix of the EhaA TU. The variable amino acids were either a combination of protease cleavage sites (OmpT, factor Xa) and epitopes (PEYFK, myc) (called: epitope) or the flexible, disordered structure forming G4S-G2S-(G4S)3 sequence (called: flex) or the rigid, α-helical structure forming A(EAAAK)5 sequence (called: rigid). Linker deletion variants contained either a β1 and a CR (Δepitope) or only a CR (ΔepitopeΔβ1) or Cel5 was directly fused to the α-helix of the EhaA TU (ΔepitopeΔβ1ΔCR). The number of the amino acids of the corresponding part of the AT-FP is given in brackets. c, All variants are encoded by plasmids of the Autodisplay-Toolbox (pATB).
Step-by-step preparation of a pATB-β-Gluc library & consecutive screening
For the proof of concept, pATBs were tested with the β-glucosidase variant CsBglA LYTH (β‑Gluc)29 as model passenger in E. coli UT5600 cells. For this purpose, in each single plasmid, the initial cel5 passenger was replaced by b-Gluc using restriction/ligation by the KpnI/XhoI sites, resulting in 24 pATB-β-Gluc variants, called step-by-step (SBS) variants (Fig. 3). Each pATB was used to transform E. coli, resulting in 24 strains derived from single colonies on separate LB Kan50 agar plates. Four randomly chosen single colonies per plate (n = 4) of the different variants and a reference were tested on activity (96 picks in total). As reference pATB_11-β-Gluc was selected, as it contained the frequently used combination of CtxB SP and epitope linker for bacterial surface display14,16,25,26,30-32. The cells were cultivated in a 200 µL scale in a 96-well microtiterplate (MTP) overnight and subjected to an activity assay with 4-nitrophenyl-β-D-glucopyranoside (pNPG) as substrate (Fig. 4a). Each variant that was supposed to have β-Gluc as passenger showed an activity > 0 mM pNP/mLOD578nm, whereas control cells expressing pATB_11-cel5 showed no activity against pNPG (Data not shown). These results indicated all plasmids as functional in expressing a corresponding AT-FP gene, but with different enzymatic activities. Compared to 11-β-Gluc (1.47 pNP/mLOD578nm) the highest activity was shown by the variants 43 (6.34 pNP/mLOD578nm), 42 (8.59 mM pNP/mLOD578nm), 22 (10.75 pNP/mLOD578nm) and 45 (12.06 mM pNP/mLOD578nm). Variant 45 comprised an OprF SP including the spacer in combination with the ΔepitopeΔβ1 linker (Fig. 4a). In summary, the pATB appeared to be suitable to screen for an optimal combination of SP and linker for a distinct passenger and host. Nevertheless, the SBS approach to construct variants before screening appeared laborious and time-consuming. In consequence, two easier approaches were investigated.
All-in-one & strain-mix preparation of pATB-β-Gluc libraries for screening
For the all-in-one (AIO) approach to prepare a pATB library for screening on activiy, the 24 pATB plasmids with cel5 as passenger were manually pooled in equimolar concentrations (Fig. 3). The pooled plasmids were subjected to a restriction digest with XhoI/KpnI and the DNA fragments were treated with alkaline phosphatase (FastAP) to prevent religation. Likewise, XhoI/KpnI was used to treat a plasmid carrying the β-Gluc DNA insert sequence (pβ-Gluc). After purification the molar ratio of insert to backbone was set to 3:1 for the ligation reaction. Accordingly, approx. 0.5 fmol/pATB DNA backbone equalling approx. 12 fmol DNA backbone in total (= 60 ng DNA considering an avergage backbone size of 8,152 bp) and approx. 36 fmol (= 30 ng DNA considering an insert size of 1,366 bp) of the β-Gluc DNA insert was used for ligation. The resulting pATB-β-Gluc AIO library was used to transform E. coli cells (single cell clone library size: 305). Finally, 105 clones were tested on activity corresponding a 98.5 % statistical coverage of the possible 24 variants (calculated with GLUE-IT CASTER 2.033).
All cells showed an activity > 0 mM pNP/mLOD578nm against pNPG (Fig. 4b) indicating that the digest of the pooled pATB was successful and no ancestor pATB-cel5 was included in the screening. Nevertheless, the AIO approach could have biased the screening result compared to the SBS approach. This bias could have been due to distinct pATB-cel5 variants that have been preferably addressed in the pooled restriction digest, favored ligation of particular pATB DNA backbones in the following ligation step, different transformation efficiencies of the pATB-β-Gluc or hampered growth of the freshly transformed E. coli cells on agar plates. To evaluate such bias, the result of the AIO screening was compared with the result of the SBS screening juxtaposed as dotplot (Fig. 4b,c). The quotient mean of reference activity/mean of all activities (MR/MA) was similar for the SBS (0.37) and the AIO (0.35) after screening. Therefore, the difference in the MA was attributed to deviations between individual screening runs and not considered any further. The 99 % confidence interval (CI) of MR served as a threshold to identify single cell clones with siginificant differences compared to 11 (Fig. 4b,c). 66 % (SBS) or 67 % (AIO) of the single cell activities were considered as significantly higher (red dots), 33 % (SBS) or 30 % (AIO) of the activities were considered as equal (black dots), and 1 % (SBS) or 3 % (AIO) of the activities were considered as significantly lower than the activity of 11 (grey dots). Sequence analysis of the plasmids from the five variants with the highest activity (Fig. 4b, stars) in the AIO screening unveiled variant 43, 42 twice and 45 twice (Fig. 4d). Two of them, 43 and 45, have been identified by the SBS approach as well. This means, both library approaches applied to a consecutive screening yielded a similar distribution pattern of activity values, similar rate of significantly different activities compared to reference 11 and led to an identical bestperforming variant in the activity assay (variant 45: OprF-spacer SP, ΔepitopeΔβ1 linker) (Fig. 4b,c,d). In consequence, the simplified library preparation via the AIO approach did not bias the screening result.
To go a step further, a bacterial strain-mix (STM) approach was taken into consideration, instead of manually pooling the 24 pATB variants as for the AIO approach. For this purpose, firstly, the mixture of 24 manually pooled pATB-cel5 plasmids was used as a whole to transform heat shock competent E. coli DH5α (single cell clone library size: 298). Single cell clones were taken up from the agar plates with 5 mL LB Kan50 medium, mixed with glycerol (25% v/v final concentration) and stored as a cryologically preserved cell stock in aliquots of 0.5 mL. Secondly, an aliquot of this stock was used to inoculate an overnight culture. Thirdly, plasmids were isolated from this overnight culture, subjected to restriction digestion (XhoI/KpnI) and ligated with a similarly digested β-Gluc fragment. Finally, the resulting new pATB-cel5 STM library was screened according to the protocol as described for the AIO approach. When the screening results of the STM approach were juxtaposed with the results of the SBS approach (Fig. 4b,c,d), it truned out, that their MR/MA values were similar (STM: 0.33, SBS: 0.37), they had a similar distribution pattern of activites, a similar amount of activities that were significantly different compared to 11 (STM: 68 %, SBS: 66 %) and again led to the identical bestperforming variant, concerning enzyme activity as before, variant 45. In conclusion, a bias in the STM screening result due to an overgrowth of individual strains during the pATB-cel5 STM overnight culture was not detectable. This means, that the bacterial STM appeared to be suitable for the production of a pATB-cel5 library that subsequently can be used to prepare a pATB library with a passenger of interest like β-Gluc.
Fig. 3: Concepts of different ATB-based library screenings. In all approaches to prepare the pATB libraries (step-by-step (SBS), all-in-one (AIO) & strain-mix (STM)) for the screening, the β-Gluc encoding DNA sequence was generated via restriction digest with XhoI/KpnI from pβ-Gluc, electrophoretically separated from its DNA backbone and subsequently purified. SBS: Individual digest of the 24 pATB-cel5 with XhoI/KpnI removed the initial passenger DNA cel5. pATB DNA backbones were electrophoretically separated from its DNA insert and subsequently purified. After separate ligation of pATB DNA backbone and the β-Gluc DNA insert for each variant, the 24 pATB-β-Gluc were separately produced in E. coli DH5α and the corresponding collection of plasmids was termed pATB-β-Gluc SBS library. AIO & STM: Simultaneous digest of the 24 pooled pATB-cel5 with XhoI/KpnI removed the initial passenger DNA sequence cel5. Ligation of the pATB DNA backbone and the β-Gluc DNA insert resulted in pATB-β-Gluc AIO library. All libraries were separately used to transform E. coli UT5600 cells. E. coli pATB-β-Gluc variants were cultivated and screened in a 96-well microtiter plate (MTP) with 4-nitrophenyl-β-D-glucopyranoside pNPG) as substrate. Highest performing variants were verified in 0.2 L scale. kbp: kilobase pairs, M: Marker
Fig. 4: Enzyme activities of different of pATB libraries. a, The mean activity of cells of the step-by-step (SBS) library screening assigned to the respective variants (n = 4 per varaint, biological quadruplicate). The cells were incubated with 5 mM 4-nitrophenyl-β-D-glucopyranoside (pNPG) and turnover was detected by measurement of 4-nitrophenol (pNP) absorption at 405 nm after 9 min reaction time. b,c Each dot represents the activity either of the reference (11, n = 3-4) of variants from one of 92 (SBS) or 105 (all‑in‑one (AIO) & strain-mix (STM)) 96-well MTP scale cultivations with subsequent pNPG assay as described in a). The 99 % confidence interval (CI) of the reference (11) mean activity (MR) was used to identify significantly higher (red dots), equal (black dots) and lower (grey dots) activities than 11 (black circles). The five highest activities of each screening are marked (stars). The quotient of MR and the mean of all activities (except 11)) (MA) served to compare the different screening approaches. d, The plasmids harbored by the five variants with the highest activity of each library screening were assigned to the respective code (SBS) or subjected to sequence analysis (AIO & STM) to identify the corresponding signal peptide (SP) and linker.
Analysing E. coli pATB-β-Gluc variants 22, 42, 43, 45 with highest enzymatic activity
To verify the results of the pATB library screenings in a larger scale, the four E. coli pATB-β-Gluc variants 22, 42, 43, 45, as well as reference 11 and E. coli UT5600 host without plasmid (blank) were cultivated in 200 mL LB. After harvesting, OD578nm was adjusted to 0.1 and an whole cell activity assay with pNPG was performed (Fig. 5a). Compared to 11 (193.1 mU/mLOD578nm), variant 43 (570.9 mU/mLOD578nm) showed a 3-fold increased enzyme activity, whereas variant 22 (702.5 mU/mLOD578nm) and 42 (707.7 mU/mLOD578nm) showed 3.6-fold increased enzyme activities. The highest enzymatic activity of all variants, also after large scale cultivation showed variant 45 (947.6 mU/mLOD578nm), indicating a 4.9-fold increase compared to 11. This result confirmed variant 45 as the most active in the present experimental setup.
After SDS-PAGE of outer membrane proteins and subsequent densitometric analysis, the level of surface displayed β-Gluc was estimated 0.9-fold lower for variant 22, approx. 1.3-fold higher for variants 42 and 43 and approx. 1.9-fold higher for variant 45 compared to 11 (Supplementary Fig. 1 and Supplementary Tab. 1). This indicated that the expression level at the surface of the high performing variants was different up to a factor of two and could have contributed the altered enzymatic activities.
Fig. 5:Enzyme activity of variants after large scale cultivation. The E. coli pATB-β-Gluc variants 22, 42, 43, 45, the reference (11) and E. coli UT5600 host cells were cultivated in 200 mL scale. Cells were adjusted to a final OD578nm of 0.1 and the activity of the cells was determined with 5 mM pNPG after 0, 3, 6 and 9 min reaction time by measurement of pNP absorption at 405 nm (n = 3, biological triplicates, ± SD, one-way ANOVA, post-test Dunnett's Multiple Comparison Test: ****p ≤ 0.0001).
An Autodisplay-Toolbox (ATB) consisting of plasmids (pATB) for the expression of AT-FP variants is introduced. pATBs were used to obtain the most active variant with the model passenger β‑Gluc. A 4.9-fold increased activity of the bestperforming variant 45 compared to the reference 11, can only in part be due to the approx. 1.9-fold increased level of β-Gluc surface display of variant 45. The remaining difference between increase in activity and level of surface display can be attributed to a synergistic effect of the OprF-spacer SP and the ΔepitopeΔβ1 linker. The SBS library screening result supports this assumption (Fig. 4a): The activity of bestperfoming variant 45 (12.06 mM pNP/mLOD578nm) was higher than the sum of the activites of variant 15 (reference CtxB SP & bestperformer ΔepitopeΔβ1 linker, 2.33 mM pNP/mLOD578nm) and variant 41 (bestperformer OprF-spacer SP & reference epitope linker, 3.36 mM pNP/mLOD578nm). Together these results indicate the suitability of the ATB-based screening for increasing the activity of cells displaying recombinant enzymes and for optimizing the level of surface displayed proteins.
However, it appears not reasonable that the combination of OprF-spacer SP and ΔepitopeΔβ1 linker is the optimal combination for each passenger in each expression strain. The activity of whole-cell biocatalysts with surface displayed passengers increases with the number of catalytically active proteins displayed on the cellular surface15. Thus, an optimal SP is supposed to contribute to high levels of surface displayed passengers by enabling pre-protein export across the inner membrane. Previous studies indicate that the optimal SP is related to the protein to be exported34,35 and the expression organism16,36. Since the pATB encoded EhaA TU was shown to be functional for the surface display of a variety of proteins in many expression organisms beyond E. coli like Pseudomonas strains, Ralstonia eutropha, Zymobacter palmae and Zymomonas mobilis37, an easy SP adaption becomes important. The linker of an AT-FP is supposed to present the passenger in the ‘sweet spot’ - the optimal combination of degree of flexibility, steric orientation, structural folding and translocation efficiency. As different linker were shown to be optimal for different passengers, this ‘sweet spot’ probably varies among passengers with different composition or of different origin14,15,17,38.
It turned out in previous studies, that the surface display of certain enzymes via classical AT, where the passenger is attached by its C-terminus to the TU leads to low or even no activity. In such case, usage of so-called inverse AT, where the passenger is connected by its N-terminus to the domain appeared be a suitable alternative17,38. In consequence, variants of the inverse autotransporter YeeJ39 were applied to the ATB (Supplementary Fig. 2). To optimally control the level of protein surface display, a tuneable promoter is of importance4. Therefore, in addition to PrhaBAD, the expression of AT-FP with pATB was also set under control of the arabinose-dependent ParaBAD and the cell density-dependent, auto-inducible PRox306 promoter which had already been proven suitable for autodisplay16,40.
Incorporating the variants with inverse TU, with ParaBAD and with PRox306 expands the ATB from 24 to 81 plasmids in total. Accordingly, the pATB code was expanded by a further digit (xyz), to account for the additional promoters (1yz – 3yz). All inverse AT-FP variants are summarized under its SP native to YeeJ (x1z – x5z) (Tab. 1). Each of these pATB, as well as unpooled and pooled libraries of pATB can be obtained via Addgene (Watertown, USA).
The AIO approach to prepare pooled libraries for the screening represents a simple procedure that all users can perform with a ready-to-use pATB batch. For some application the STM approach could be of advantage which allows to produce the pATB-cel5 library on demand from a single, cryologically conserved stock of strains. Although it is not excluded that general rules for optimal combinations of expression organism and SP or passenger and linker might become apparent in the future, the ATB screenings are a convenient platform for adjusting the surface display for the passenger of interest in the expression organism of choice. Finally, but not lastly, it is possible to choose one of the 81 different plasmids, as provided in the ATB, individually, in a rational approach e.g. based on critera or features of the passenger for a straightforward surface display approach.
Tab. 1: Plasmids of the Autodisplay-Toolbox.
*The DNA backbone sizes refer to the respective plasmid without initial passenger DNA cel5G which itself was not part of the classification.
**All YeeJ inverse AT-FP variants are summarized under its signal peptide native to YeeJ.
Computational methods
The secondary structures of the EhaA TU, the EhaA connecting region (CR), the EhaA β1 domain (β1), the lysin motif (LysM), the YeeJ TU and the natural linker were modelled with RaptorX27. Variable amino acids in the linker (epitope, flex, rigid) were modelled with PEP-FOLD328. Image J was used to perform the densitometric analysis of the SDS-PAGE gel41. NEBiocalculator® version 1.15.0 was used for calculations regarding ligation reactions in pATB-β-Gluc SBS and AIO library preparation. GLUE-IT CASTER 2.033 was used to calculate the statistical coverage of the 24 pATB-β-Gluc variants for the screening of the pATB-β-Gluc AIO and STM library. Data obtained from library screenings and high performing variant verification were analyzed and visualized with GraphPad Prism V6 (San Diego, USA). Visualization of Fig. 1, Fig. 2a,b and Fig. 3 was performed with Adobe Illustrator Artwork 16.0 (Dublin, Ireland).
Bacterial cultivation conditions
Escherichia coli strains DH5α and UT5600 were cultivated in 1/5 filled 100 mL or 1 L shake flasks containing lysogeny broth (LB) medium (37 °C, 200 rpm) or on LB agar plates (37°C), both containing 50 μg mL-1 kanamycin (Kan50), if selection was desired. Cultivations of E. coli pATB‑β‑Gluc strains in 1 L shake flasks were inoculated with 1 % (v/v) from the overnight culture. When the cell suspension reached an optical density at 578 nm (OD578nm) of 0.5, AT-FP gene expression was induced via addition of 2 mM L-(+)-rhamnose (final concentration) and the cultivation prolonged for additional 2 h. E. coli pATB-β-Gluc strains were also cultivated in 96-well microtiter plates (MTP). Inner wells contained the cells which were cultivated on a 96-well microplate shaker (25 °C, 600 rpm) in 200 µl LB Kan50 medium, 0.1 % (w/v) glucose and 2 mM L-(+)-rhamnose allowing auto-induction of AT-FP gene expression. Wells at the outer edge were filled with sterile ddH2O to prevent evaporation effects.
Plasmid construction
In the entire Methods section, the three digit code (x = promoter, y = SP +/- spacer, z = linker) to denote pATB is used (Tab. 1). In a first step, the plasmids of the ATB (pATB) 111, 121, 131, 141, 211, 212, 213, 214, 215, 216, 221, 231, 241, 311, 321, 331, 341 were constructed. Therefore, the DNA sequence encoding Cel5 from Hahella chejuensis (Uniprot ID: Q2SFD8)was synthesized as string DNA including 3’ kpnI and 5’ xhoI restriction sites (Thermo Fisher Scientific). Restriction/ligation via XhoI/KpnI and T4 DNA ligase was used to exchange the passenger DNA sequence in previously published pPBAD-MATE-CsbglA29 leading to pATB_111 (ParaBAD, CtxB, epitope), in poprFSP-MATE-estA16 leading to poprFSP-MATE-cel5 (PrhaBAD, CtxB, epitope) and in PRox306-MATE-CsbglA29 leading to pATB_311 (PRox306, CtxB, epitope). The following plasmids were generated using In-Fusion® Cloning as described by the manufacturer (Takara Bio©). Therefore, DNA sequences of the respective template plasmids were amplified via PCR. The designations of the primers (synthesised by Sigma-Aldrich®) that were used for PCR are given in brackets (fw = forward, rv = reverse) and respective sequences are summarized in Supplementary Tab. 3. poprFSP-MATE-cel5 (774 (fw) & 770 (rv)) was used as template plasmid to generate poprFSP‑MATE-cel5 (= pATB_121 (ParaBAD, OprF, epitope)) containing the myc epitope encoding sequence in the linker. pATB_111 (568 (fw) & 984 (rev)) and previously published pPQ2814 (931 (fw) & 534 (rv)) served as template plasmids to generate pATB_211 (PrhaBAD, CtxB, epitope). pATB_121 (1584 (fw) & 984 (rv)) and pATB_211 (931 (fw) & 1568 (rev)) were used as template plasmids to generate pATB_221 (PrhaBAD, OprF, epitope). pATB_121 (1597 (fw) & 984 (rv) and pATB_311 (932 (fw) & 915 (rv)) were the template plasmids to generate pATB_321 (PRox306, OprF, epitope). pATB_111, 121, 211, 221, 311 and 321 (3289 (fw) & 3290 (rv)) were used as template plasmids in individual reactions, to generate pATB_131, 141, 231, 241, 331 and 341 which differed from their template plasmids by the spacer DNA sequence in the pro-region coding DNA. pATB_211 was used as template in three reactions (878 (fw) & 879 (rv), 883 (fw) & 880 (rev), 814 (fw) & 882 (rev) respectively), to construct the deletion variants encoding pATB_214 (PrhaBAD, CtxB, Δepitope), pATB_215 (PrhaBAD, CtxB, ΔepitopeΔβ1), pATB_216 (PrhaBAD, CtxB, ΔepitopeΔβ1ΔCR). pATB_214 was the template in two reactions (3291 (fw) & 3293 (rev), 2858 (fw) & 2857 (rev)) to generate pATB_212 (PrhaBAD, CtxB, flex) and pATB_213 (PrhaBAD, CtxB, rigid).
In a second step, the pATBs 112-116, 122-126, 132-136, 141-146, 22-226, 232-236, 242-246, 312-316, 322-326, 332-336, 342-346 were generated based on the previously constructed pATBs using restriction/ligation via XhoI/KpnI/BglII(10 U/enzyme, 37 °C, 2 h)and T4 DNA ligase (5 U, 4 °C, overnight). pATB_111, 121, 131, 141, 211, 221, 231, 241, 311, 321, 331 and 341 were digested with BglII/XhoIto obtain DNA backbones that individually contained each possible combination of promoter DNA (ParaBAD, PrhaBAD, PRox306) and SP encoding DNA (CtxB, OprF, CtxB‑spacer, OprF-spacer). pATB_211, 212, 213, 214, 215 and 216 were digested with KpnI/BglII(10 U/enzyme, 37 °C, 2 h) to generate DNA inserts that individually encoded the linker variants (epitope, flex, rigid, Δepitope, ΔepitopeΔβ1, ΔepitopeΔβ1ΔCR). The respective DNA backbone sequences were ligated with respective DNA insert sequences. To produce the plasmids, respective ligation or In-Fusion® mixtures were used to transform heat shock competent E. coli DH5α cells. The correctness of the plasmids was validated via colony PCR (primers listed in Supplementary Tab. 3) and sequence analysis (Microsynth AG).
In a third step, the pATBs 151-153, 251-253, 351-353 were generated. The passenger DNA sequence of previously published pDG11-estA38(ParaBAD, YeeJ, epitope) was changed to CsbglA via the described restriction and ligation approach using pPBAD-MATE-CsbglA29, XhoI/KpnI and T4 DNA ligase. pDG11-CsbglA38was used as template in two PCR reactions ((I) 3676 (fw) & 3572 (rev); (II) 1815 (fw) & 2532 (rev)). pATB_212 ((I) 1230 (fw) & 2071 (rv)) and pATB_213 ((II) 2537 (fw) & 3573 (rv)) were used as template plasmids in individual PCR reactions. The respective DNA fragments were used for In-Fusion® Cloning to generate (I) pATB_152-CsbglA (ParaBAD, YeeJ, flex) and (II) pATB_153-CsbglA (ParaBAD, YeeJ, flex). Subsequently the promoter sequences of pDG11-estA38, pATB_152-CsbglA and pATB_153-CsbglA were changed from ParaBAD to PrhaBAD and PRox306 respectively via PCR and In-Fusion® Cloning. pDG11-estA38, pATB_152-CsbglA and pATB_153-CsbglA (3569 (fw) & 3570 (rv)) and pATB_211 (931 (fw) & 3544 (rev)) were used as template plasmids to generate pATB_251 (PrhaBAD, YeeJ, epitope), pATB_252 (PrhaBAD, YeeJ, flex), pATB_253 (PrhaBAD, YeeJ, rigid). pDG11-estA38, pATB_152-CsbglA and pATB_153-CsbglA (2531 (fw) & 3570 (rv)) and pATB_311 (931 (fw) & 915 (rev)) were used as template plasmids to generate pATB_351 (PRox306, YeeJ, epitope), pATB_252 (PRox306, YeeJ, flex), pATB_253 (PRox306, YeeJ, rigid). The passenger DNA sequences of pATB 151-153, 251-253, 351-353 were changed to cel5 via the described restriction and ligation approach using pATB_211, XhoI/KpnI and T4 DNA ligase. The correctness of the plasmids was validated via colony PCR (primers listed in Supplementary Tab. 3) and sequence analysis (Microsynth AG).
pATB-β-Gluc step-by-step (SBS), all-in-one (AIO) & strain-mix (STM) library preparation
Three different approaches were applied to prepare libraries with the ATB plasmids (pATB) for the subsequent screening. In an SBS approach, approx. 50 fmol (= approx. 300 ng) DNA of each of the 24 pATB containing PrhaBAD and EhaA TU coding DNA (pATB-cel5, pATB_211 – pATB_246) was treated in a separate reactions with XhoI/KpnI (10 U/enzyme) and FastAP (2 U, 37 °C, 2 h). Likewise, approx. 367.1 fmol (= approx. 2 µg) DNA of pBBR1MCS-2BAD-CsBglA-LYTH (pβ-Gluc) carrying the β-Gluc DNA sequence was digested with XhoI/KpnI but without FastAP. The DNA fragments were separated by agarose gel electrophoresis (1 % agarose (w/v), 120 V, 45-80 min) and pATB DNA backbones and β-Gluc DNA inserts were purified via silica membrane using spin columns. For the separate ligation reactions, the molar insert:backbone ratio was supposed to be 3:1. Thus, approx. 2 fmol DNA/pATB backbone (= approx. 10 ng considering an average backbone size of 8,152 bp) and approx. 6 fmol of the β-Gluc DNA insert (= 5 ng considering an insert size of 1,366 bp) was required for the ligation reaction catalysed by T4 DNA ligase (5 U, 4°C, overnight) (NEBioCalculator® version 1.15.0). Heat shock competent E. coli DH5α cells were separately transformed with 5 µL of the ligation mixture to produce the collection of 24 plasmids termed pATB PrhaBAD-β-Gluc SBS library. Correct plasmids (validation as described) were used to separately transform heat shock competent E. coli UT5600 cells.
In an AIO approach, the 24 pATB_211 – pATB_246 were manually pooled in equimolar concentrations (approx. 3.4 fmol DNA/pATB = approx. 20 ng DNA/pATB). The pooled pATB (approx. 81.6 fmol = approx. 500 ng pATB DNA) were subjected to restriction digest using XhoI/KpnI (10 U/enzyme, 37 °C, 2 h) and added alkaline phosphatase (FastAP, 2U) was supposed to prevent religation of emerging DNA fragments. Approx. 367.1 fmol (= approx. 2 µg) DNA of pβ‑Gluc carrying the β-Gluc DNA sequence was also digested with XhoI/KpnI but without FastAP. The 24 pATB DNA backbones and the β-Gluc DNA insert were separated from the remaining plasmid DNA via agarose gelelectrophoresis (inserts: 1 % agarose (w/v), 120 V, 45 min; backbones: 1 % agarose (w/v), 120 V, 80 min) with subsequent silica-membrane based purification using a spin column. For the ligation reaction, the molar insert:backbone ratio was supposed to be 3:1. Thus, approx. 0.5 fmol/pATB DNA backbone equalling 12 fmol backbone DNA in total (= approx. 60 ng DNA considering an avergage backbone size of 8,152 bp) and approx. 36 fmol (= approx. 30 ng DNA considering an insert size of 1,366 bp) of the β-Gluc DNA insert was required for the ligation reaction catalysed by T4 DNA ligase (5 U, 4 °C, overnight) (Calculations: NEBioCalculator® version 1.15.0). Ligation reaction resulted in the pATB-β-Gluc AIO library. 5 µL of the ligation mixture were used to transform heat shock competent E. coli UT5600 cells and cells were plated out on two plates to increase single cell clone library size.
In a STM approach, the mixture of the 24 manually pooled pATB-cel5 was used to transform heat shock competent E. coli DH5α. The cells were cultivated on LB Kan50 agar plates, the cells were swepped from the plates using 5 mL LB Kan50 medium, this cellsuspension was mixed with glycerol (25 % (v/v) final) and the cells were preserved cryologically in 0.5 mL stocks. An entire stock was used to inoculate an overnight culture. The plasmids harbored by cells from the overnight culture were isolated resulting in a pooled pATB-cel5 library which was further treated as described in the protocol for the AIO approach.
Screening & verification
Cells were either picked from four randomly chosen clones per variant of E. coli UT5600 cells transformed with the pATB-β-Gluc SBS library or from 105 randomly chosen clones of E. coli UT5600cells transformed with the pATB-β-Gluc AIO library. The strains were individually cultivated in a well of a MTP as described and subsequently subjected to a activity test in MTP format. The MTPs containing cultivated cells were centrifuged (3900 x g, 4°C, 15 min) and 200 µl 0.1 M Na-citrate buffer (pH 6) was used to wash twice and resuspend the cells. The cell suspensions were diluted 1:10 with 0.1 M Na-citrate buffer (pH 6) in a fresh MTP and the OD578nm was measured in a MTP reader. 30 µl of the diluted cell suspension was transferred to a fresh plate and the activity test was started by adding 30 µl of 10 mM preheated (55°C, 3 min) 4‑nitrophenyl β-D-glucopyranoside (pNPG, final concentration 5 mM). The MTP was incubated (55 °C, 9 min) and centrifuged (3900 x g, 4°C, 3 min). To quantify 4-nitrophenol (pNP), 25 μL supernatant of each sample was transferred to a fresh MTP, diluted with 25 μL 0.1 M sodium citrate buffer (pH 6), mixed with 50 μL of 2 M Na2CO3. The absorption was photometrically determined at 405 nm, adjusted for the absorption of equally treated E. coli UT5600 host cells and normalized to the previously determined OD578nm. The sequence of the plasmids harbored by high performing variants of the AIO approach was analysed (Microsynth AG). To verify high performing variants that were identified via pATB library screening in a MTP scale, respective variants were cultivated in a 1 L shake flask scale as described in previous section, harvested (3900 x g, 4°C, 15 min), washed with 0.1 M sodium citrate buffer (pH 6) and subjected to an activity test. Therefore, 15 samples were prepared. Per reaction tube, 30 μL of the cellsuspension OD578nm 0.2 (final OD578nm of 0.1) were mixed with 30 μL of 10 mM preheated (55°C, 3 min) pNPG (final concentration 5 mM) to start the reaction. The mixtures incubated (55 °C, 900 rpm) and the reaction was stopped after 0, 3, 6 and 9 min by placing the reaction tubes on ice. The reaction tubes were centrifuged (3900 x g, 4°C, 3 min) and the supernatant of each sample was transferred to a fresh MTP. Subsequently the samples were processed as described for pNP quantification. pNP concentrations determined for 0, 3, 6 and 9 min reaction time were used to determine the activity of the cells. The activity was normalized to 1 mL of a cell suspension with an OD578nm of 1 (mU/(mLOD578nm)).
Membrane protein isolation & protease accessibility
Membrane protein isolation (MPI) was performed as described previously42. Cells from 40 ml culture broth were pelleted (10 min, 3850 x g, 4°C). Prior to MPI, the surface accessibility of β‑Gluc was tested by applying proteinase K (prot. K). Therefore, the cells were resuspended in 1 ml 0.1 M PBS and incubated after addition of 250 µg proteinase K for 1 h at 37 °C. Protease digest was stopped by adding 50 µl of 0.1 M PMSF. The mixture was centrifuged (5 min, 3850 x g, 4 °C), the cells were resuspended in 1.5 ml 0.2 M Tris-HCl buffer (pH 8) and subsequently treated as described for MPI. For SDS-PAGE analysis, the samples were boiled for 20 min at 95 °C in SDS‑PAGE sample buffer containing 30 mM DTT and separated in a 12.5 % polyacrylamid gel. Coomassie Brilliant Blue G-250 was used to stain the proteins in the gel.
- Smith, G.P. & Petrenko, V.A. Phage display. Chem. Rev.97, 391-410 (1997).
- Heinis, C., Rutherford, T., Freund, S. & Winter, G. Phage-encoded combinatorial chemical libraries based on bicyclic peptides. Nat. Chem. Biol.5, 502-507 (2009).
- Smith, G.P. Phage Display: Simple Evolution in a Petri Dish (Nobel Lecture). Angew. Chem., Int. Ed.58, 14428-14437 (2019).
- Jose, J. & Meyer, T.F. The Autodisplay Story, from Discovery to Biotechnical and Biomedical Applications. Microbiol. Mol. Biol. Rev.71, 600-619 (2007).
- Maurer, J., Jose, J. & Meyer, T.F. Autodisplay: one-component system for efficient surface display and release of soluble recombinant proteins from Escherichia coli. J. Bacteriol.179, 794-804 (1997).
- Nicolay, T., Vanderleyden, J. & Spaepen, S. Autotransporter-based cell surface display in Gram-negative bacteria. Crit. Rev. Microbiol.41, 109-123 (2015).
- Drobnak, I. et al. Of linkers and autochaperones: an unambiguous nomenclature to identify common and uncommon themes for autotransporter secretion. Mol. Microbiol.95, 1-16 (2015).
- Dautin, N. Folding control in the path of type 5 secretion. Toxins13, 341 (2021).
- Sklar, J.G., Wu, T., Kahne, D. & Silhavy, T.J. Defining the roles of the periplasmic chaperones SurA, Skp, and DegP in Escherichia coli. Genes Dev.21, 2473 (2007).
- Bernstein, H.D. Looks can be deceiving: Recent insights into the mechanism of protein secretion by the autotransporter pathway. Mol. Microbiol.97, 205-215 (2015).
- Henderson, I.R., Navarro-Garcia, F., Desvaux, M., Fernandez, R.C. & Ala'Aldeen, D. Type V Protein Secretion Pathway: the Autotransporter Story. Microbiol. Mol. Biol. Rev.68, 692-744 (2004).
- George, A., Patil, A.G. & Mahalakshmi, R. ATP-independent assembly machinery of bacterial outer membranes: BAM complex structure and function set the stage for next-generation therapeutics. Protein Sci.33, e4896 (2024).
- Muñoz-Gutiérrez, I., Oropeza, R., Gosset, G. & Martinez, A. Cell surface display of a β-glucosidase employing the type v secretion system on ethanologenic Escherichia coli for the fermentation of cellobiose to ethanol. J. Ind. Microbiol. Biotechnol.39, 1141-1152 (2012).
- Quehl, P., Schüürmann, J., Hollender, J. & Jose, J. Improving the activity of surface displayed cytochrome P450 enzymes by optimizing the outer membrane linker. Biochim. Biophys. Acta1859, 104-116 (2017).
- Hörnström, D., Larsson, G., van Maris, A.J.A. & Gustavsson, M. Molecular optimization of autotransporter-based tyrosinase surface display. Biochim. Biophys. Acta1861, 486-494 (2019).
- Tozakidis, I.E.P., Lüken, L.M., Üffing, A., Meyers, A. & Jose, J. Improving the autotransporter-based surface display of enzymes in Pseudomonas putida KT2440. Microb. Biotechnol.13, 176-184 (2020).
- Tian, H. et al. Enzyme cascade converting cyclohexanol into ε‐caprolactone coupled with NADPH recycling using surface displayed alcohol dehydrogenase and cyclohexanone monooxygenase on E. coli. Microb. Biotechnol.15, 2235-2249 (2022).
- Mekalanos, J.J. et al. Cholera toxin genes: nucleotide sequence, deletion analysis and vaccine development. Nature306, 551-557 (1983).
- Kragelund, L., Leopold, K. & Nybroe, O. Outer Membrane Protein Heterogeneity within Pseudomonas fluorescens and P. putida and Use of an OprF Antibody as a Probe for rRNA Homology Group I Pseudomonads. Appl. Environ. Microbiol.62, 480-485 (1996).
- Andersson, H. & Von Heijne, G. A 30-residue-long "export initiation domain" adjacent to the signal sequence is critical for protein translocation across the inner membrane of Escherichia coli. PNAS88, 9751-9754 (1991).
- Liu, S. et al. The pro-region of Streptomyces hygroscopicus transglutaminase affects its secretion by Escherichia coli. FEMS Microbiol. Lett.324, 98-105 (2011).
- Schlacher, A. et al. Detection of a new enzyme for stereoselective hydrolysis of linalyl acetate using simple plate assays for the characterization of cloned esterases from Burkholderia gladioli. J. Biotechnol.62, 47-54 (1998).
- Wells, T.J. et al. EhaA is a novel autotransporter protein of enterohemorrhagic Escherichia coli O157:H7 that contributes to adhesion and biofilm formation. Environ. Microbiol.10, 589-604 (2008).
- Arai, R., Ueda, H., Kitayama, A., Kamiya, N. & Nagamune, T. Design of the linkers which effectively separate domains of a bifunctional fusion protein. PEDS14, 529-532 (2001).
- Sichwart, S., Tozakidis, I.E.P., Teese, M. & Jose, J. Maximized Autotransporter-Mediated Expression (MATE) for surface display and secretion of recombinant proteins in Escherichia coli. Food Technol. Biotechnol.53, 251-260 (2015).
- Quehl, P. et al. Co-expression of active human cytochrome P450 1A2 and cytochrome P450 reductase on the cell surface of Escherichia coli. Microb. Cell Fact.15 (2016).
- Källberg, M. et al. Template-based protein structure modeling using the RaptorX web server. Nat. Protoc.7, 1511-1522 (2012).
- Lamiable, A. et al. PEP-FOLD3: faster de novo structure prediction for linear peptides in solution and in complex. Nucleic acids research44, W449-W454 (2016).
- Lenz, F., Zurek, P., Umlauf, M., Tozakidis, I.E.P. & Jose, J. Tailor-made β-glucosidase with increased activity at lower temperature without loss of stability and glucose tolerance. Green Chemistry22, 2234-2243 (2020).
- Tozakidis, I.E.P., Brossette, T., Lenz, F., Maas, R.M. & Jose, J. Proof of concept for the simplified breakdown of cellulose by combining Pseudomonas putida strains with surface displayed thermophilic endocellulase, exocellulase and β-glucosidase. Microb. Cell Fact.15, 103 (2016).
- Schulte, M.F., Tozakidis, I.E.P. & Jose, J. Autotransporter-Based Surface Display of Hemicellulases on Pseudomonas putida: Whole-Cell Biocatalysts for the Degradation of Biomass. ChemCatChem9, 3955-3964 (2017).
- Obeng, E.M. et al. The workability of Escherichia coli BL21 (DE3) and Pseudomonas putida KT2440 expression platforms with autodisplayed cellulases: a comparison. Appl. Microbiol. Biotechnol.102, 4829-4841 (2018).
- Firth, A.E. & Patrick, W.M. Statistics of protein library construction. Bioinformatics21, 3314-3315 (2005).
- Degering, C. et al. Optimization of protease secretion in bacillus subtilis and bacillus licheniformis by screening of homologous and heterologous signal peptides. Appl. Environ. Microbiol.76, 6370-6376 (2010).
- Brockmeier, U. et al. Systematic Screening of All Signal Peptides from Bacillus subtilis: A Powerful Strategy in Optimizing Heterologous Protein Secretion in Gram-positive Bacteria. J. Mol. Biol.362, 393-402 (2006).
- Peng, C. et al. Factors influencing recombinant protein secretion efficiency in gram-positive bacteria: Signal peptide and beyond. Front. bioeng. biotechnol.7, 139 (2019).
- Tozakidis, I.E.P., Sichwart, S. & Jose, J. Going beyond E. coli: autotransporter based surface display on alternative host organisms. New Biotechnol.32, 644-650 (2015).
- Gercke, D., Furtmann, C., Tozakidis, I.E.P. & Jose, J. Highly Crystalline Post‐Consumer PET Waste Hydrolysis by Surface Displayed PETase Using a Bacterial Whole‐Cell Biocatalyst. ChemCatChem13, 3479-3489 (2021).
- Martinez-Gil, M. et al. YeeJ is an inverse autotransporter from Escherichia coli that binds to peptidoglycan and promotes biofilm formation. Sci. Rep.7 (2017).
- Meyers, A., Furtmann, C., Gesing, K., Tozakidis, I.E.P. & Jose, J. Cell density-dependent auto-inducible promoters for expression of recombinant proteins in Pseudomonas putida. Microb. Biotechnol.12, 1003-1013 (2019).
- Schneider, C.A., Rasband, W.S. & Eliceiri, K.W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods9, 671-675 (2012).
- Park, M. et al. Isolation and characterization of the outer membrane of Escherichia coli with autodisplayed Z-domains. Biochim. Biophys. Acta1848, 842-847 (2015).
Yes there is potential Competing Interest.
J.Jose is co-founder and shareholder of Autodisplay Biotech GmbH, involved in the application of autodisplay technology