Combining Protein Ligation Systems to Expand the Functionality of Semi-Synthetic Outer Membrane Vesicle Nanoparticles

Bacterial outer membrane vesicles (OMVs) attract increasing interest as immunostimulatory nanoparticles for the development of vaccines and therapeutic agents. We previously engineered the autotransporter protein Hemoglobin protease (Hbp) into a surface display carrier that can be expressed to high density on the surface of Salmonella OMVs. Moreover, we implemented Tag-Catcher protein ligation technology, to obtain dense display of single heterologous antigens and nanobodies on the OMVs through coupling to the distal end of the Hbp passenger domain. Here, we aimed to further expand the versatility of the Hbp platform by enabling the coupling of heterologous proteins to internal sites of the Hbp passenger. Inserted SpyTags were shown to be accessible at the Salmonella OMV surface and to efficiently couple SpyCatcher-equipped fusion proteins. Next, we combined distally placed SnoopCatcher or SnoopTag sequences with internal SpyTags in a single Hbp molecule. This allowed the coupling of two heterologous proteins to a single Hbp carrier molecule without obvious steric hindrance effects. Since coupling occurs to Hbp that is already exposed on the OMVs, there are no limitations to the size and complexity of the partner proteins. In conclusion, we constructed a versatile modular platform for the development of bivalent recombinant OMV-based vaccines and therapeutics.


INTRODUCTION
Outer membrane vesicles (OMVs) are nanoparticles (20-100 nm) that shed from the outer membrane of Gram-negative bacteria in a natural continuous process. They are non-replicating and non-invasive, hence safe, but still contain the immunostimulatory properties that are a facsimile of the parental cells (Alaniz et al., 2007). This makes OMVs an attractive platform for the development of recombinant vaccines (Bitto and Kaparakis-Liaskos, 2017;Gnopo et al., 2017). One effective strategy involves the display of heterologous antigens on the surface of OMVs for which various concepts have been described (Gerritzen et al., 2017).
We have previously engineered a protein display system based on the bacterial autotransporter Hemoglobin protease (Hbp; Daleke-Schermerhorn et al., 2014) to allow surface decoration of bacterial cells and derived OMVs with recombinant proteins. To cross the complex Gram-negative bacterial cell envelope, autotransporters are organized in three domains (Pohlner et al., 1987; Figure 1A): (i) an N-terminal signal peptide that targets the protein to the Sec translocon for translocation across the inner membrane, (ii) a secreted passenger domain that carries the effector function, and (iii) a C-terminal β-domain that integrates into the outer membrane (OM) and facilitates translocation of the passenger from the periplasm into the extracellular space. The latter process also involves the host-derived β-barrel assembly machinery (Bam) complex (Ieva and Bernstein, 2009;Sauri et al., 2009). Following translocation, autocatalytic cleavage in the β-domain interior results in cleavage of the passenger from its β-domain and release into the extracellular environment (Barnard et al., 2007). The cleaved and released passenger domain is a long β-helical stem structure from which small loops and larger functional domains extend (Otto et al., 2005). We previously identified five subdomains (1-5; Figure 1A) that can be replaced by heterologous polypeptides without compromising Hbp expression and secretion. Furthermore, mutagenesis of residues critical for autocatalytic cleavage created an Hbp passenger that remains anchored to the OM to allow permanent exposure of fused antigens (Jong et al., 2012). This platform was successfully used to display multiple inserted antigens from Mycobacterium tuberculosis, Streptococcus pneumonia, and pathogenic Escherichia coli (ETEC) on the surface of OMVs derived from a hypervesiculating and genetically LPS-detoxified Salmonella Typhimurium strain (Jong et al., 2012;Daleke-Schermerhorn et al., 2014;Jong et al., 2014;Kuipers et al., 2015;Hays et al., 2018). These OMVs induced strong protective responses in animal models (Kuipers et al., 2017;Hays et al., 2018), underscoring the potential of recombinant antigendecorated OMVs for vaccination.
Although Hbp tolerates integration of heterologous protein sequences, it has a limited capacity to translocate very bulky and/or complex fusion partners (Jong et al., 2007;Daleke-Schermerhorn et al., 2014;Jong et al., 2014;Kuipers et al., 2015). With the development of protein ligation systems, such as the Tag-Catcher technology that creates covalent links between proteins (Zakeri et al., 2012;Veggiani et al., 2016), the display of complex proteins on the surface of OMVs has become much easier. This so called-bacterial superglue is based on domains of streptococcal adhesins that spontaneously form intramolecular isopeptide bonds. These domains can be split into a small Tag and a larger Catcher that reconstitute and covalently bind upon mixing, even when fused to other proteins. We have recently shown that the SpyTag and the SnoopTag, as well as the SpyCatcher and the SnoopCatcher, can be incorporated into the Hbp display platform when fused to the distal end of the Hbp passenger, thereby replacing the N-terminal protease domain. Full-length complex antigens as well as functional nanobodies could be linked to the OMV platform (Van Den Berg et al., 2018). Moreover, the coupling procedure appeared versatile and robust, allowing fast production of experimental vaccines, and therapeutic agents through a modular plug-anddisplay procedure.
Here, we aimed to further expand the versatility of the Hbp display platform by incorporating a second protein ligation site. We show that heterologous proteins can be efficiently coupled to Hbp on the surface of Salmonella OMVs via SpyTag at an internal position of the passenger domain. Importantly, the coupling did not affect SnoopCatcher-SnoopTag ligation at the distal end of the same passenger. The ability to couple multiple heterologous proteins simultaneously, for instance complex antigens and/or targeting moieties like antibodies and nanobodies, is of considerable interest for the development of more effective OMV-based vaccines and biomedicines.

Construction of Plasmids
Plasmid constructs and primers used in this study are listed in Table 1 and Supplementary Table S1, respectively.

OMV Isolation, Analysis of OMV Integrity, and HbpD Surface Display
The OMV production strain S. Typhimurium SL3261 tolRA msbB carrying one of the HbpD expression plasmids ( Table 1) was grown at 30 • C in TYMC supplemented with glucose (0.2%), chloramphenicol (30 µg/ml), and kanamycin (25 µg/ml). Overnight precultures were used to inoculate fresh medium to an to an optical density at 660 nm (OD 660 ) of 0.07. After 7 h of incubation and reaching an OD 660 of approximately 1.0 this culture was used to inoculate fresh medium containing 50 µM of Isopropyl β-D-1-thiogalactopyranoside (IPTG) to an OD 660 of 0.02. Growth under these inducing conditions was continued overnight. To isolate OMVs, cells were removed by two successive centrifugation steps at 5,000 × g. The supernatant was passed through 0.45 µm-pore-size filters (Millipore) and centrifuged at 235,000 × g for 1 h to sediment the OMVs. The OMVs were finally resuspended in PBS containing 15% glycerol (1 OD unit of OMVs per µl). An amount of 1 OD unit of OMVs is derived from 1 OD 660 unit of cells. The integrity of the OMVs and the surface display of HbpD variants on OMVs was analyzed using a Proteinase K accessibility assay as described previously .

Protein Ligation to HbpD on OMVs
To OMVs displaying a variant of HbpD a >4-fold molar excess of purified SpC-MBP, SpC-TrxA, SnC-MBP, SnC-TrxA, SnT-MBP, and/or SnT-TrxA was added. After 24 h of incubation at 4 • C, the reaction mixtures were analyzed by SDS-PAGE and Coomassie staining.

Analysis of Protein Content of OMVs
Protein profiles of OMV samples were analyzed using SDS-PAGE and Coomassie G-250 (BioRad) staining. Densitometric analysis on Coomassie-stained gels was carried out using a Molecular Imager GS-800 Calibrated Densitometer and ImageJ software 1 . To quantify protein ligation efficiencies, the respective densities of protein bands corresponding to ligated and nonligated Hbp fractions were calculated after correcting for the difference in molecular mass. Quantifications were performed on samples of a representative experiment and correspond to the Coomassie-stained gels shown, where appropriate. For immunodetection of protein samples after Western blotting, monoclonal anti-FLAG M2 antibody (F3165; Sigma), and HA tag monoclonal antibody (2-2.2.14; ThermoFisher Scientific) were used.

Purification of Recombinant Proteins for Coupling to OMVs
Escherichia coli BL21(DE3) cells harboring a pET28a plasmid for expression of recombinant proteins carrying Spy or Snoop elements were grown in LB containing glucose (0.2%) and kanamycin (50 µg/ml) to early log phase. Protein expression was induced by the addition of IPTG to a final concentration of 0.5 mM, and the cells were incubated for a further 2 h. After centrifugation, the cells containing pellet was resuspended in buffer A (pH7.4) containing 50 mM sodium phosphate (Na 2 HPO 4 + NaH 2 PO 4 ), 300 mM NaCl, and 125 µM phenylmethylsulfonyl fluoride (PMSF). The cells were disrupted by two passages through a One Shot cell disruptor (Constant Systems, Ltd.) at 1.2 × 10 8 Pa. Cell debris and membranes were removed by centrifugation at 10,000 × g and 293,000 × g, respectively, at 4 • C. His 6 -tagged proteins were isolated from the cleared lysate using Talon Superflow medium (GE Healthcare Life Sciences) according to the manufacturer's instructions. Eluates were dialyzed overnight at 4 • C against 500 volumes of PBS (pH 7.4). After dialysis, glycerol was added to 10%, and aliquots were stored at −80 • C.

Protein Ligation to Internal Sites of the Hbp Passenger
In previous work we have shown highly efficient Spy and Snoop-based ligation of proteins to the non-cleaved Hbp display construct (HbpD) at the surface of OMVs. Tag and Catcher sequences were positioned at the "tip" of the Hbp passenger through genetic replacement of the native subdomain 1 at the N-terminus of the passenger (Van Den Berg et al., 2018; Figures 1A,C). Four additional subdomains (2-5) are present in the Hbp passenger that protrude from the β-helical stem (see Figure 1C) and are also tolerant toward substitution by heterologous sequences (Jong et al., 2014). To investigate whether protein ligation can be achieved at these internal positions in the Hbp passenger, we genetically integrated SpyTag (SpT) at domain 2 (d2) and 4 (d4) (Jong et al., 2007;Sauri et al., 2012;Jong et al., 2014), both located almost halfway down the β-helical stem on opposite sides (Figure 1C). To optimize its ability to contact the SpyCatcher in this context, the SpyTag was flanked by flexible 5-amino acid Gly-Ser linkers. Given the small spatial distance of the fusion sites in the Hbp passenger structure (6 Å apart; see PDB 1WXR), we considered the possibility that these short linkers (5/5) would force the 13 amino acid SpyTag sequence into a bent conformation, inhibiting functional interaction of the SpyCatcher-SpyTag pair. Hence, longer 10-residue linkers (10/10) were also tested that allow more conformational freedom, but may be more prone to proteolysis.
To test ligation to the HbpD-SpT fusions on the surface of the OMVs, we used two soluble model proteins of different size, the relatively small thioredoxin1 (TrxA, 12 kDa, roughly 40 × 40 × 40 Å; PDB 5HR3), and the larger maltose-binding protein (MBP, 40 kDa, roughly 45 × 65 × 70 Å; PDB 1LLS). Both were equipped with a SpyCatcher (SpC) for coupling, an HAtag for immunodetection and a polyhistidine tag for purification ( Figure 1B and Supplementary Figure S1). The OMVs were incubated overnight with purified SpC-MBP or SpC-TrxA and analyzed by SDS-PAGE (Figure 2 and Table 2). Adducts with the expected molecular mass (see Table 2) appeared upon addition of SpC-MBP and SpC-TrxA (Figure 2, lanes 2-3) at the expense of the HbpD-(d1)SpT carrier (cf. lane 1). These are specific for the presence of a SpyTag as no adducts are formed upon addition of Catchered model proteins to Salmonella OMVs carrying HbpD without a SpyTag (Van Den Berg et al., 2018). Densitometric scanning analysis showed 95% and 87% ligation efficiency, respectively, similar to previous ligation experiments HbpD-(d4)SpT 10/10 (145) (199) 84% (170) 86% * Protein molecular mass in kDa is given between brackets. # Percentage of Hbp ligated with SpC-fusion protein. Frontiers in Microbiology | www.frontiersin.org with the same carrier (Van Den Berg et al., 2018). Efficient ligation (≥74%) to the model proteins was also seen for the HbpD constructs carrying internal SpyTags, irrespective of the length of the flanking linkers (Figure 2 and Table 2). Yet, the corresponding adducts ran with lower mobility in SDS-PAGE than expected for their molecular mass (see Table 2) and different running forms of the same adduct were observed for TrxA. This aberrant and unpredictable migration may be due to the branched nature of polypeptides upon linkage to SpyTag at the internal domain 2 and 4 positions of the Hbp passenger. Of note, immunodetection showed that these ligation products contained an HA-tag, confirming the identity of the HbpD-MBP and HbpD-TrxA adducts (Supplementary Figure S2). In conclusion, the internal domain 2 and domain 4 positions of HbpD are suitable sites for Spy-based protein ligation at the surface of OMVs.

Simultaneous Ligation of Multiple Proteins to the Hbp Passenger
Being able to link two independent proteins at the same time would be a valuable addition to the functionality of the HbpD display platform. To achieve this, we investigated whether elements of the Spy and Snoop ligation system could be functionally combined in a single HbpD molecule. Importantly, cross-talk between these systems is absent (Veggiani et al., 2016), which should allow site-specific coupling of recombinant proteins to the Hbp platform.
Previously, we have shown efficient protein ligation using the SnoopCatcher (SnC) or the SnoopTag (SnT) at the N-terminal domain 1 (d1) position of Hbp (Van Den Berg et al., 2018). Likewise, to construct bivalent Spy-Snoop HbpD variants, either SnoopCatcher, or SnoopTag was placed at the domain 1 position of HbpD constructs already containing an internal SpyTag at either the position of domain 2 or 4 ( Figure 1A). We chose to build on the ten-amino acid-long linker (10/10) variants only, which seemed to yield marginally more efficient protein ligation compared to the variants carrying shorter linkers (5/5; see Table 2). Moreover, future cargo proteins other than MBP or TrxA may be sterically more restricted in the coupling reaction, requiring longer linkers.
Model fusion proteins were produced to enable (immuno)detection of coupling to the various positions in the dual Spy-Snoop HbpD constructs. In addition to the above-described HA-tagged SpC-MBP and SpC-TrxA proteins  (Figure 1B and Supplementary Figure S1). Of note, during purification of the model fusion proteins we noticed that SnC-MBP was rather prone to aggregation, the reason for which is unclear. This protein was not further analyzed in coupling experiments. To investigate Spy and Snoop-based protein ligation, OMVs displaying one of the four Hbp variants (see Figure 3) were incubated overnight with the cognate MBP and TrxA versions ( Table 3). In the cases where simultaneous coupling to both attachment sites in the HbpD passengers was studied, the Snoop-and Spyequipped model proteins were added together in a single reaction mix. The reaction mixes were analyzed by SDS-PAGE (Figures 4A,C) and the coupling efficiencies were determined by densitometry of the Coomassie stained carriers and adducts ( Table 3). The identity of the adducts, both single and double, was confirmed by immunodetection of the HA-tag and the FLAG-tag (Figures 4B,D).
Initially we focused on coupling to the individual sites. Successful Snoop-based ligation of SnC-TrxA and SnT-TrxA to the N-terminus of Hbp in the absence of a SpyCatcher protein (Figure 4A, lane 2 and 8; Figure 4C, lane 2, 6, 10, and 14) showed that the SnoopTag as well as the SnoopCatcher can be coupled efficiently at this position (up to 92 and 69%, respectively). However, SnT-MBP ligated to the SnoopCatcher at the domain 1 position with a much lower efficiency (∼25%; Figure 4C, lane 6, and 14), which is puzzling since similar fusions to SpyTag couple more efficiently (Van Den Berg et al., 2018).   On the other hand, Spy ligation to the internal SpyTag was highly efficient (between 93 and 100%) both with TrxA and MBP fusion proteins. In fact, coupling appeared to proceed more efficiently than to the comparable HbpD-(d2)SpT and HbpD-(d4)SpT constructs (see Table 2). These fusions still carried the native 27 kDa protease domain 1 at the N-terminus (Figure 1A), which possibly interferes with coupling to SpyTags integrated further downstream in the Hbp passenger. Remarkably, SpC-MBP adducts migrated quite differently, depending on whether coupling took place to the position of domain 2 or 4 (Figure 4A, cf. lanes 3 and 9; Figure 4C, cf. lanes 3 and 9). Coupling to either of these internal positions of the Hbp passenger did result in branched structures with different arm lengths that apparently have distinct mobilities in SDS-PAGE. Yet, these products were all detected by α-HA, confirming their identity as HbpD-MBP adducts (Figures 4B,D).
Next, we analyzed whether the TrxA and MBP model proteins could be simultaneously linked to the dual Spy and Snoop sites in the same HbpD derivative. Clearly, in all tested combinations, adducts were detected by SDS-PAGE at a higher apparent molecular mass than those formed upon incubation with the corresponding single TrxA and MBP model proteins (Figures 4A,C). Moreover, immunodetection (Figures 4B,D) demonstrated the presence of both an HA-tag and FLAGtag in the highest molecular mass adducts in all but one sample ( Figure 4C, lane 16), confirming their identity as coupling products comprising HbpD and both TrxA and MBP. Interestingly, the efficiency of formation of these adducts was roughly equivalent to the product of the two individual ligation efficiencies (Figure 4 and Table 3), suggesting that ligation to the N-terminal Snoop elements and the internal SpyTag in Hbp did not affect each other.
In conclusion, Spy and Snoop elements can be functionally combined in single HbpD molecules and exploited for the simultaneous coupling of different heterologous proteins at the surface of OMVs.

DISCUSSION
We combine OMV technology with autotransporter-based display of recombinant proteins in order to create semisynthetic, safe and immunogenic nanoparticle vectors for vaccine formulation and therapeutic purposes. To this end, we previously genetically removed TolRA in an existing attenuated ( aroA) S. Typhimurium strain to overproduce OMVs for easy and cost-effective harvesting . Secondly, we deleted the msbB gene, resulting in pentaacylated lipidA to reduce the reactogenicity of OMVs while retaining the intrinsic adjuvant effect (Kuipers et al., 2017). We further engineered the autotransporter Hbp into a scaffold that allows display of multiple integrated heterologous protein sequences (Jong et al., 2012;Jong et al., 2014). Although the Hbp display platform is relatively tolerant toward genetically inserted polypeptides, increasing the number, size and structural complexity of insertions generally lowers the expression of the fused constructs at the OMV surface due to the limited capacity of translocation across the Salmonella cell envelope.
To overcome this limitation, the Spy and Snoop protein ligation technologies (Veggiani et al., 2016) were successfully used to covalently link purified heterologous proteins to Hbp passengers that are already present on the surface of OMVs (Van Den Berg et al., 2018).
Whereas Catchers and Tags of the ligation systems were previously located at the distal domain 1 position of the Hbp passenger, we show here that Tags also function at the domain 2 and domain 4 position, roughly halfway down the β-helical stem structure. Consistent with other studies using internally placed Tags (Zhang et al., 2013;Sun et al., 2014;Gao et al., 2016), integration of the SpyTag in Hbp did not limit its potential to couple SpyCatchered model proteins. Also, the position closer to the OM surface did not appear to hinder efficient protein ligation, as almost all Hbp-(d2)SpT and HbpD-(d4)SpT copies could be coupled ( Table 3). Although not tested here, we expect that these results can be extrapolated to the integration of Tag sequences of other ligation systems such as Snoop, which was shown to function under a wide range of conditions, similar to Spy (Zakeri et al., 2012;Veggiani et al., 2016). Catchers, on the other hand, are expected to be less compatible because their integration in the Hbp passenger will likely interfere with translocation across the OM (Jong et al., 2007;Daleke-Schermerhorn et al., 2014;Jong et al., 2014;Kuipers et al., 2015). Also, fusion of the Catchers to internal positions may well affect their folding into a functional conformation, more than fusion to the passenger N-terminus where they have full conformational freedom. Of note, this limitation can be overcome by making use of the robust tripartite variants of both Spy and Snoop in which two short ligation Tags are covalently bonded by a separate active ligase (Fierer et al., 2014;Andersson et al., 2019).
Next, we examined the simultaneous coupling of two proteins to a single display platform molecule. To have independent control over the positioning and degree of occupation of the two separate ligation sites we chose to use the SnoopTag-SnoopCatcher pair as the secondary system. In this proof-ofprinciple study a non-optimized protocol was used in which OMVs carrying the HbpD fusions were simply mixed with both complementing model proteins at the same time and left to incubate. The efficiencies of dual adduct formation (Table 3) were approximately equivalent to the product of the ligation efficiencies of the individual reactions. The observation that ligation to one site is not intrinsically restrictive to ligation to the second site is important in cases where both components are needed at the highest possible density on the surface of the OMVs. Moreover, it allows high densities of dual adducts in which both attachments are fixed in close proximity to each other. A potential application for this is the simultaneous engagement of different receptors, for instance to stimulate T cells (Wu et al., 2020). In any case, the coupling of multiple proteins to one Hbp allows for a simple and cost-efficient production process in which only one batch of OMVs needs to be produced, without the need to express different display systems in parallel, which may be tedious. It also circumvents approaches involving the simultaneous expression of multiple Hbp constructs with a single attachment site in the same OMV production host, which may lead to instability at the genetic level due to homologous recombination events.
One obvious application of our technology is the attachment of multiple different complex antigens from any source to a single OMV for multivalent vaccine production. However, antigen display can also be combined with the coupling of for instance antibodies, nanobodies or adhesins for targeting to certain tissues and immune cells to tailor immune responses. The compatibility of our OMV platform with the attachment of SpyTagged nanobodies (Van Den Berg et al., 2018) and antibodies (data not shown) indeed suggests such a dual functionality is possible, not only for prophylactic but also for therapeutic use (Bitto and Kaparakis-Liaskos, 2017;Zhang et al., 2019). In conclusion, by functionally incorporating two protein ligation sites in HbpD we created a versatile modular platform for the development of recombinant OMV-based vaccines and therapeutics.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
HB, DH, JL, and WJ made major contributions to the conception and design of the study. HB, DH, CK, JL, and WJ contributed to the acquisition, analysis, or interpretation of the data. HB, DH, JL and WJ contributed to the writing of the manuscript.

FUNDING
Part of this research was carried out in the framework of the Bac-Vactory program that was funded by the domain of Applied and Engineering Sciences (TTW) of the Netherlands Organization for Scientific Research (NWO). Part of this work was funded by an EU/Eureka Eurostars grant to project OptiFemVac (ref: 11019).