Kinetics of CagA type IV secretion by Helicobacter pylori and the requirement for substrate unfolding

Type IV secretion of effector proteins is an important principle for interaction of human pathogens with their target cells. The corresponding secretion systems may transport a multitude of effector proteins that have to be deployed in the respective spatiotemporal context, or only a single translocated protein, as in the case of the CagA effector protein produced by the human gastric pathogen Helicobacter pylori. For a more detailed analysis of the kinetics and mode of action of CagA type IV secretion by H. pylori, we describe here, a novel, highly sensitive split luciferase‐based translocation reporter which can be easily adapted to different end‐point or real‐time measurements. Using this reporter, we showed that H. pylori cells are able to rapidly inject a limited amount of their CagA supply into cultured gastric epithelial cells. We have further employed the reporter system to address the question whether CagA has to be unfolded prior to translocation by the type IV secretion system. We showed that protein domains co‐translocated with CagA as protein fusions are more readily tolerated as substrates than in other secretion systems, but also provide evidence that unfolding of effector proteins is a prerequisite for their transport.


| INTRODUC TI ON
Interaction of pathogenic bacteria with host cells is often facilitated by dedicated protein secretion systems that have evolved to deploy their respective substrate proteins as pathogenicity determinants.
A common theme in this respect in Gram-negative bacteria is the direct injection of effector proteins into target cells, which is mediated by type III, type IV, or type VI secretion systems (Galán & Waksman, 2018). In addition, some type IV and type VI secretion systems are used to transport effector proteins into other bacterial cells. Type IV secretion systems (T4SS) are known as macromolecule transporters with a wide range of transported substrates, from small monomeric effector proteins to nucleoprotein complexes of considerable size. As an adaptation to different substrates, functions and host cell types, the corresponding transporters also vary considerably in their composition and the mechanistic details of substrate transport (Li et al., 2019). Transport capabilities range from only one or a few to several hundred different effector proteins. The Cag T4SS of the human gastric pathogen Helicobacter pylori is at an extreme position within this range, as it represents an expanded T4SS containing several non-canonical components in addition to the core subunits (Costa et al., 2020), but seems to transport only one effector protein, the cytotoxin-associated antigen CagA, into gastric target cells .
Translocation of the CagA protein to gastric epithelial cells is considered a hallmark of H. pylori pathogenesis, and due to the | 795 LETTL ET aL. multitude of its cellular interaction partners and its association with development of gastric cancer, CagA has been described as a protein with oncogenic potential (Hatakeyama, 2014;Takahashi-Kanemitsu et al., 2020). The CagA protein itself, as well as interactions of its domains with eukaryotic target proteins have been examined in structural detail (Hayashi et al., 2012;Kaplan-Türköz et al., 2012;Nesic et al., 2010Nesic et al., , 2014, and numerous functional consequences of CagA interactions with target proteins have been identified (Hatakeyama, 2014;Tegtmeyer et al., 2017), but the mechanisms of its transport across the bacterial envelope or the host cell membrane are still only poorly understood. Although CagA is not secreted into the culture supernatant in the absence of eukaryotic target cells, it is able to interact with integrin molecules on the surface of host cells (Jiménez-Soto et al., 2009;Kaplan-Türköz et al., 2012;Koelblen et al., 2017). Moreover, it requires a positively charged region for binding to phospholipids and uptake into cells, which has been interpreted as an interaction of bacterial surface-exposed CagA with membrane lipids that constitutes an intermediate step in translocation (Murata-Kamiya et al., 2010). Similar to effector proteins of other T4SS (Christie et al., 2014), the secretion signal is partly located at the CagA C-terminus, although the C-terminal region is not sufficient to direct the protein to the secretion machinery (Hohlfeld et al., 2006;Schindele et al., 2016).
We have recently reported that a β-lactamase (TEM-1) fused to CagA can be used as a quantitative translocation reporter system for CagA type IV secretion into gastric epithelial cells (Schindele et al., 2016). Meanwhile, this reporter system has been used with different cell types and under different conditions, and turned out to be a robust and reliable tool (Behrens et al., 2020;Bonsor et al., 2018;Zhao et al., 2018). Successful application of TEM-1 fusions implies that the Cag T4SS is able to transport the β-lactamase domain in an efficient way; however, TEM-1-CagA fusions are produced much less than wild-type CagA, and translocated TEM-1-CagA could only weakly be detected by tyrosine phosphorylation, and did not exhibit typical CagA-dependent phenomena, such as the hummingbird phenotype (Schindele et al., 2016). Another limitation of the TEM-1-CagA reporter system is that it represents an end-point assay and cannot easily be adapted for monitoring translocation in real time. Here, we have used N-terminal fusions of CagA with an 11-residue split-luciferase (HiBiT) tag, and we show that luciferase reconstitution in gastric epithelial cells is a highly sensitive method to monitor CagA translocation in end-point and real-time assays. HiBiT-tagged CagA is well-tolerated and transported by H. pylori, and does not interfere with downstream phenotypes such as CagA tyrosine phosphorylation.

| Evaluation of a NanoLuc HiBiT fusion as CagA type IV secretion reporter
Several epitope tags or protein domains with different amino acid sequences have been demonstrated to be co-translocated to target cells, when fused to the N-terminus of CagA (Hohlfeld et al., 2006;Murata-Kamiya et al., 2010;Schindele et al., 2016). To examine the possibility of utilizing a split-luciferase system for monitoring translocation, we generated an H. pylori strain producing a CagA variant with an N-terminal NanoLuc HiBiT tag by integrating the HiBiT sequence without a resistance marker in the cagA locus ( Figure 1a).

Immunoblots of bacterial lysates that were probed with recombinant
LgBiT protein confirmed that the HiBiT-CagA fusion can be easily detected via its HiBiT tag (Figure 1b). To evaluate translocation of HiBiT-CagA into target cells, we first determined HiBiT-CagA tyrosine phosphorylation upon co-incubation with AGS cells. Immunoblot analysis showed that HiBiT-CagA was indeed phosphorylated, albeit at a lower efficiency in comparison to wild-type CagA, whereas no phosphorylation was detected with a HiBiT-CagA-producing, but secretion-deficient strain (ΔcagT; Figure 1c). In contrast, tyrosine phosphorylation of a TEM-1-CagA fusion protein was not visible under these conditions, probably due to lower protein levels, as shown previously (Schindele et al., 2016). Microscopic analysis of infected AGS cells confirmed that translocated HiBiT-CagA was also capable of eliciting the hummingbird phenotype, which is dependent on phosphorylated CagA, to a similar extent as P12 wild-type ( Figure 1d). In contrast, the reporter strain producing a TEM-1-CagA fusion did not show clear signs of hummingbird cells under these conditions. To be able to measure HiBiT-CagA translocation independently of its tyrosine phosphorylation, we stably transfected AGS cells with an expression plasmid encoding the LgBiT sequence coupled to a self-labeling HaloTag (Los et al., 2008). Single trans-  (Figure 1g). These results show that the HiBiT tag is a useful reporter for detecting CagA translocation at high sensitivity, and with minimal disturbance of downstream cellular processes.

| Quantification of CagA translocation rates
In marked contrast to effector proteins of type III or other type IV secretion systems (Schlumberger et al., 2005;Schroeder, 2017), CagA is produced at rather high levels; in fact, CagA represents one of the most abundant H. pylori proteins under standard culture conditions (Jungblut et al., 2000). However, translocation competition experiments using strains that co-produce tagged and untagged versions of CagA, and estimation of translocation by quantitative immunoblot analysis, suggested that only a minor fraction of the bacterial CagA pool gets translocated into cultured target cells (Jiménez-Soto & Haas, 2016;Schindele et al., 2016). To obtain a more quantitative view on CagA translocation rates, we first quantified translocated HiBiT-CagA after co-incubation with AGS [LgBiT] cells, as described  -CagA] was generated by inserting a translational HiBiT-cagA gene fusion in a two-step procedure, using the rpsL- between translocated CagA and total CagA, we obtained values between 3.5% and 12.5%, depending on the bacterial to target cell ratio ( Figure 2c). Thus, H. pylori is able to translocate up to 12.5 ± 2% of its CagA content when present at one bacterial cell per target cell or less, but only lower fractions at higher concentrations, although the total amount of CagA per bacterial cell does not seem to be reduced at higher MOIs.

| Kinetics of CagA type IV secretion
Since the HiBiT translocation assay requires only the addition of substrate to the bacteria-target cell co-incubation mixture, we modi- In order to translocate CagA, H. pylori has to establish target cell contacts that are productive for type IV secretion. For example, binding to CEACAM receptors at the target cell surface via the outer membrane protein HopQ represents such a productive interaction (Königer et al., 2016). To assess the influence of this interaction during early stages of translocation, we compared a P12

| Requirement of protein unfolding for CagA translocation
Taking advantage of the high sensitivity of the HiBiT-CagA translocation assay, we next used it to address the impact of protein folding on CagA translocation. It is generally assumed that CagA has to be unfolded before passage through the secretion channel (Pattis et al., 2007), but experimental evidence for this has been lacking so far. To address this question, we generated N-terminal CagA fusions with the HiBiT tag and further passenger domains with different folding characteristics. First, we examined fusions with ubiquitin (Ub), which is known to fold rapidly and to impede protein translocation across membranes in its folded state, or with a ubiquitin variant (Ub I3G, I13G ) with decreased conformational stability that can be unfolded more easily (Johnsson & Varshavsky, 1994) (Figure 4a).
As a control, we generated HiBiT-GFP-CagA, reasoning that the showed that HiBiT-Ub-CagA was partly processed, and that only the processed protein was tyrosine-phosphorylated. Since only translocated CagA can get phosphorylated, and HiBiT-Ub-CagA produced a clear luminescence signal (Figure 4c), this indicates that proteolysis occurred after translocation. Furthermore, since no processing, but tyrosine phosphorylation was observed for HiBiT-Ub I3G, I13G -CagA ( Figure S4a), this proteolysis step obviously depends on Ub folding, either suggesting that Ub is translocated in a folded state, or that it folds rapidly upon arrival in the AGS cell, and then gets recognized by a cellular protease, presumably a deubiquitinase.
To corroborate and extend these findings with an alternative approach, we generated a strain producing a fusion of the HiBiT tag together with murine dihydrofolate reductase (DHFR) to the N-terminus of CagA ( Figure 5a). DHFR is a 22 kDa single-domain protein which folds rapidly and can be stabilized in its folded conformation by folate analogs such as methotrexate (Eilers & Schatz, 1988). The HiBiT-DHFR-CagA fusion protein was produced well by strain P12, but detection by HiBiT blotting showed partial processing and the release of low molecular-weight HiBiT-containing fragments ( Figure S4b).

Nevertheless, translocation of HiBiT-DHFR-CagA to AGS [LgBiT]
cells could be detected with the HiBiT translocation assay, albeit at much lower rates than HiBiT-CagA ( Figure S4c). Importantly, HiBiT-DHFR-CagA translocation was efficiently blocked by adding methotrexate in a dose-dependent manner, whereas methotrexate did not inhibit translocation of HiBiT-CagA lacking DHFR (Figure 5b). Using a phosphotyrosine immunoblot after immunoprecipitation from infected cells, as described above, we were able to detect a very weak phosphorylation of DHFR-CagA lacking the HiBiT tag ( Figure S4a).
This phosphorylation was not seen after infection in the presence of methotrexate, as expected, but we noted that the sensitivity of the phosphotyrosine assay was much lower than that of the HiBiT translocation assay (Figure 5b). Since methotrexate does not bind to

| D ISCUSS I ON
Translocation reporters that allow quantitative monitoring of substrate transport have facilitated the identification of effector proteins, and the analysis of molecular details of transport in different T4SS. For example, Cre recombinase (Lang et al., 2010;Vergunst et al., 2000), adenylate cyclase (Chen et al., 2004;Nagai et al., 2005), or β-lactamase fusions (de Felipe et al., 2008;Schindele et al., 2016) have been used to demonstrate type IV translocation of effector proteins, and to elucidate the nature of type IV secretion signals.
NanoLuc luciferase or split-luciferase fusions have been employed as reporters to study Sec-dependent transport (Pereira et al., 2019), or type III secretion in Yersinia (Lindner et al., 2020) and Salmonella (Westerhausen et al., 2020). The advantage of the split-luciferase system used here is that the reporter tag is small and thus does not interfere with the secretion process, as confirmed by assessing CagA-dependent phenotypes, and that detection is very sensitive and does not require any lytic or other cumbersome post-infection treatment. The results of this study show that an efficient monitoring of type IV secretion by the H. pylori Cag system is achieved with both end-point and real-time protocols. Our data indicate that the luminescence signal is directly proportional to the amount of translocated HiBiT-CagA, as may be expected since there is no accumulation of reaction products that influence the readout signal.
Furthermore, we were able to establish a protocol including a lysis step to obtain additional quantitative information about the fraction of bacterial effector protein that gets translocated. The only limitation of this reporter system is that it depends on LgBiT production Deletion of the hopQ gene resulted in only slightly delayed translocation kinetics, but in lower overall translocation rates. This is in line with the previous conclusion that other ligand-receptor interactions which are conducive to type IV secretion may form between H. pylori and certain target cells (Königer et al., 2016), but it also shows that these other putative routes to type IV secretion do not reach full efficiency, corroborating the finding that the HopQ-CEACAM interaction is necessary for full translocation rates .
Generally, the kinetics of CagA translocation featured an initial burst of effector protein delivery, followed by a plateau phase after CagF (Bonsor et al., 2013). In contrast to this, a rapid exhaustion of intrabacterial effector protein pools was often observed in type III secretion systems of Salmonella enterica, Shigella flexneri, or enteropathogenic Escherichia coli, where a similar immediate onset of effector protein delivery upon target cell contact has been described as well (Enninga et al., 2005;Mills et al., 2008;Schlumberger et al., 2005). For example, the pool of the S. enterica effector protein SipA was found to be quickly translocated within 2 to 10 min, clearly indicating limited amounts of this effector (Schlumberger et al., 2005). Furthermore, only a minor fraction (26%)  an optimum at about one bacterium per epithelial cell, which might be in a similar range as bacterial densities observed in gastric glands of colonized mice (Fung et al., 2019). At higher bacterial densities, the total amount of translocated CagA would reach a maximum and then decrease again (Figures 2a, S3a), possibly to prevent inflicting too much damage. As CagA has been described to be involved in acquisition of nutrients for microcolony formation (Tan et al., 2011), it is also conceivable that CagA translocation becomes dispensable at higher colonization densities.
It is well-established that T4SS comprise a wide variety of transported substrates or secretion apparatus architectures (Christie et al., 2014), and substantial variations exist for their molecular secretion mechanisms as well. Nevertheless, most T4SS probably secrete their substrates directly from the cytoplasm, without periplasmic intermediates, and are believed to depend on substrate unfolding prior to transport (Christie et al., 2014). Indeed, substrates of different T4SS show strongly reduced transport rates, or no transport at all, when they are fused to rapidly folding protein domains such as Ub or DHFR (Amyot et al., 2013;Trokter & Waksman, 2018).
We show here that the Cag system exhibits strongly reduced translocation rates for a DHFR fusion protein, and no translocation of a GFP fusion, but is still able to transport Ub-CagA fusions at almost wild-type rates (Figure 4e). Thus, at least transport of the small Ub domain by the Cag system is markedly different from the situation in the R388 conjugation system, where addition of folded Ub to a fusion of Cre recombinase with the relaxase TrwC resulted in a decrease of transport efficiency by three orders of magnitude (Trokter & Waksman, 2018), or from type III secretion systems of Yersinia enterocolitica or S. enterica, where corresponding fusions were not detectably secreted (Lee & Schneewind, 2002;Radics et al., 2014). Our data therefore suggest that Ub is either translocated in a folded state within the HiBiT-Ub-CagA fusion, or that the secretion machinery is able to provide enough energy to unfold Ub (but not GFP or DHFR bound to methotrexate) prior to translocation. Although we cannot completely rule out the first possibility, we favor the alternative explanation that folded Ub is not stable enough to resist the substrate unfolding activity of the Cag secretion system, for several reasons: First, although detailed structural analysis of whole Cag secretion apparatus complexes by cryo-electron tomography (Chang et al., 2018;Hu et al., 2019), or of isolated outer membraneassociated core complex particles by cryo-electron microscopy (Chung et al., 2019;Sheedlo et al., 2020) has been performed, there is currently no structural evidence that the translocation channel is able to accommodate folded substrates. Second, the difference between the strongly reduced, but still substantial translocation rates of HiBiT-DHFR-CagA, and the complete incapacity of HiBiT-GFP-CagA to be translocated, is most likely not due to the (minor) size differences of the corresponding folded domains, but probably rather due to a general resistance of GFP to become unfolded, as shown for the Salmonella SPI-1 type III secretion system (Akeda & Galan, 2005). Indeed, GFP fusions have recently been used to trap a type III-secreted substrate inside the translocation channel, where it could be visualized in an unfolded state (Miletic et al., 2021). Our interpretation of the differences in translocation rates is, thus, that the DHFR and Ub domains require less energy for the unfolding process. Third, the inhibitory activity of methotrexate observed here indicates on the one hand that the DHFR domain is actually folded in H. pylori, since methotrexate would not bind otherwise. On the other hand, it shows that stabilization of DHFR folding prevents translocation of HiBiT-DHFR-CagA, an effect that can hardly be explained by assuming transport of folded domains. These observations are also in contrast to the Yersinia Ysc type III secretion system, where a YopE-DHFR fusion can only be secreted when it is kept in an unfolded state by the SycE chaperone (Feldman et al., 2002). Nevertheless, we assume that unfolding of CagA itself is facilitated by its interaction with the secretion chaperone CagF (Pattis et al., 2007), which is able to bind to several CagA domains (Bonsor et al., 2013), but probably not to DHFR. Finally, our observation that co-production of HiBiT-CagA and DHFR-CagA led to more severe inhibition of HiBiT-CagA translocation in the presence of methotrexate, also supports the assumption that substrate unfolding is required. In this notion, DHFR-CagA would be introduced into the secretion apparatus, which would be unable to process the fusion protein in the presence of methotrexate, due to stabilized folding of DHFR, resulting in physical obstruction of the secretion channel. This would also imply that CagA is introduced with its C-terminus first into the secretion channel, as speculated previously (Woon et al., 2013), and would be consistent with the notion that the CagA C-terminus contains the major, although not the only, type IV secretion signal (Hohlfeld et al., 2006;Schindele et al., 2016).
In conclusion, the HiBiT-CagA translocation assay developed here is a powerful tool to explore mechanistic details of type IV secretion by quantitative measurements with a high sensitivity, and optionally in real-time experimental setups. As a complement to the previously described TEM-1-CagA translocation assay (Schindele et al., 2016), it has the potential to enable advanced functional studies of the widespread family of bacterial type IV protein transporters.

| Plasmid constructs
All PCR amplification, cloning and DNA analysis procedures were performed according to standard protocols (Sambrook & Russell, 2001). Gene fusions encoding N-terminally tagged CagA variants were introduced in the cagA locus via a two-step protocol using the rpsL-erm counterselection system (Rohrer et al., 2012).

| Cell lines
AGS cells were cultivated under standard conditions in 75 cm 2 tissue culture flasks (BD Falcon) and subcultivated every 2-3 days in 6-well, 24-well (tissue culture treated, Costar, Corning Inc.), or 96-well microtiter plates (black, transparent bottom, tissue cultured treated, 4ti-tude®), as described previously (Fischer et al., 2001). For microscopy, cells were seeded in IBIDI 8-well μ-slides. To generate the HaloTag- LgBiT-producing reporter cell line, AGS cells were stably transfected with plasmid pCS1956B02 (Promega), as described previously . To obtain a cell line derived from a single cell clone, trans-

| Antibodies, SDS-PAGE and immunoblotting
Rabbit polyclonal antisera AK257, AK299, and AK270 directed against CagA, the CagA EPIYA region, and CagT, respectively, have been described previously (Fischer et al., 2001;Schindele et al., 2016). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting was performed as described previously (Fischer et al., 2001). For the development of immunoblots, polyvinylidene difluoride (PVDF) filters were blocked with 5% non-fat milk powder in TBS (50 mM Tris-HCl, pH 7.5, 150 mM NaCl), 0.1% (v/v) Tween 20 (TBS-T), and incubated with the respective antisera at a dilution of 1:5,000 in TBS-T with 5% non-fat milk powder. Alkaline phosphatase-conjugated protein A, or horseradish peroxidase-conjugated anti-mouse IgG antiserum, was used to visualize bound antibody. Where appropriate, membranes were stripped by incubation in 25 mM glycine, pH 2.0, 1% SDS, for 10 min, and reprobed with a different antiserum. HiBiT-tagged CagA was further detected using the Nano-Glo® HiBiT blotting system (N2410, Promega), according to the manufacturers' protocol. Briefly, proteins were blotted on nitrocellulose membranes, which were incubated for at least 10 min in TBS-T to solubilize the HiBiT tag. Subsequently, membranes were incubated overnight at 4℃ under gentle agitation with 1x Nano-Glo® blotting buffer containing recombinant LgBiT protein at a dilution of 1:200. After adding Nano-Glo® luciferase assay substrate, and incubation at room temperature for 5 min, bands were visualized using a chemiluminescence imager.

| AGS infection and immunoprecipitation of phosphorylated CagA
AGS cells were co-incubated with the respective H. pylori strains and subsequently samples for phosphotyrosine immunoblotting were prepared as described (Fischer et al., 2001).    (4titude) were infected with 40 µl preculture, and 10 µl 5× luciferase substrate (9.5 µl assay buffer +0.5 µl furimazine substrate) were added. Luminescence measurements at 470 nm (20 nm bandwidth, 10 s integration time, gain 4,095, bottom optics) were started immediately and detected every 3-5 min for up to 2.5 hr in a prewarmed Clariostar plate reader with CO 2 levels adjusted to 5% using an atmospheric control unit. Optionally, bacteria were brought into contact with target cells by centrifugation Luminescence was recorded, as described above, using top optics.

| Translocation and HiBiT quantification assays
For the blank value, luminescence of uninfected cells was measured.
"Total HiBiT-CagA" values were calculated considering a dilution factor of five from the original sample.

| Killing assay
AGS cells seeded in 12 well culture plates were infected with 800 µl H. pylori P12 in PBS/FCS at an OD 550 of 0.1. After 30 min of incubation at 37℃, 5% CO 2 , inhibitors or DMSO were added at the indicated concentrations, and co-incubation was continued. Cells with adherent bacteria were scraped off at 0, 30, 60, or 120 min after compound administration, and aliquots of the infection mixture were spread on nonselective serum agar plates, and incubated to determine total numbers of viable bacteria.

| Statistical analysis
Unless indicated otherwise, quantitative data shown are means with standard deviations of at least three independent experiments. In case of the real-time HiBiT-CagA translocation assay, representative results with means and standard deviations derived from two technical replicates are depicted. Statistical analysis was performed using the GraphPad Prism 5 software. The significance of differences was determined using One-way ANOVA with the indicated post-hoc tests. Linear regression analysis was calculated considering individual replicate values of each Y point. The goodness of fit is indicated as R 2 .

ACK N OWLED G EM ENTS
This work was supported by grants from the German Center for Infection Research (DZIF Project 8025806810 to RH) and the Deutsche Forschungsgemeinschaft (HA2697/16-2 to RH). The authors wish to thank Nikola D. Tomić for his contribution during HiBiT-CagA translocation assay development. Open access funding enabled and organized by ProjektDEAL.

CO N FLI C T O F I NTE R E S T
The 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.

AUTH O R CO NTR I B UTI O N S
Conception and design of the study: RH and WF. Acquisition of data: CL. Analysis and interpretation of data: CL and WF. Writing of the manuscript: CL and WF.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study are available in the supplementary material of this article.