Assembly properties of bacterial tubulin homolog FtsZ regulated by the positive regulator protein ZipA and ZapA from Pseudomonas aeruginosa

Bacterial tubulin homolog FtsZ self-assembles into dynamic protofilaments, which forms the scaffold for the contractile ring (Z-ring) to achieve bacterial cell division. Here, we study the biochemical properties of FtsZ from Pseudomonas aeruginosa (PaFtsZ) and the effects of its two positive regulator proteins, ZipA and ZapA. Similar to Escherichia coli FtsZ, PaFtsZ had a strong GTPase activity, ~ 7.8 GTP min-1 FtsZ-1 at pH 7.5, and assembled into mainly short single filaments in vitro. However, PaFtsZ protofilaments were mixtures of straight and “intermediate-curved” (100–300 nm diameter) in pH 7.5 solution and formed some bundles in pH 6.5 solution. The effects of ZipA on PaFtsZ assembly varied with pH. In pH 6.5 buffer ZipA induced PaFtsZ to form large bundles. In pH 7.5 buffer PaFtsZ-ZipA protofilaments were not bundled, but ZipA enhanced PaFtsZ assembly and promoted more curved filaments. Comparable to ZapA from other bacterial species, ZapA from P. aeruginosa induced PaFtsZ protofilaments to associate into long straight loose bundles and/or sheets at both pH 6.5 and pH 7.5, which had little effect on the GTPase activity of PaFtsZ. These results provide us further information that ZipA functions as an enhancer of FtsZ curved filaments, while ZapA works as a stabilizer of FtsZ straight filaments.


Results
PaFtsZ assembles into both straight and curved filaments. Although PaFtsZ has been used in some studies, its detailed biochemical characteristics are still not fully reported. Here, we first examined its assembly using both negative-stain electron microscopy (EM) and light-scattering assays. In these studies, we used several different buffers. HMK buffer contains 50 mM HEPES, pH 7.5, 100 mM KAc and 5 mM MgAc. The pH value of HMK buffer is close to the physiological condition. MMK buffer contains 50 mM MES, pH 6.5, 100 mM KAc and 5 mM MgAc. Usually, EcFtsZ assembly is better in MMK buffer and is referred to as the assembly buffer. HEK buffer and MEK buffer use 1 mM EDTA to replace MgAc of HMK and MMK buffers; FtsZ assembly still occurs in the absence of Mg, but GTP hydrolysis is completely blocked 9,45 . Similar to EcFtsZ, PaFtsZ assembles into single pfs in HMK buffer (Fig. 1A,B). Most PaFtsZ pfs are straight but some are arc-shaped, with a 100-200 nm diameter characterized as "intermediate curvature" 5 . The intermediate curved conformation was enhanced when PaFtsZ assembled in MEK buffer, which contains no Mg 2+ at pH 6.5 (Fig. 1E). Occasionally, PaFtsZ formed closed circles in MEK buffer (Fig. 1E, arrowed). Assembly in HEK buffer (pH 7.5) was much weaker than in MEK (pH 6.5), producing few pfs even at 10 µM PaFtsZ (Fig. 1F).
In HMK buffer the assembly gave mostly single pfs at 3 µM PaFtsZ, while at 5 µM these tended to form thin bundles, as indicated by the higher contrast (Fig. 1A,B). In MMK buffer assembly produced mostly single pfs after 1 min, but these aggregated into thick bundles by 5 min (Fig. 1C,D). Light scattering confirmed the bundling and showed that assembly in MMK followed two stages. Assembly of single pfs gave an initial rise in light scattering at ~ 10 s both in HMK and MMK buffer (Fig. 1G), which is similar to the assembly of EcFtsZ detected by fluorescence assays 9,10 . This was followed after ~ 60 s by a large increase in light scattering attributed Scientific Reports | (2020) 10:21369 | https://doi.org/10.1038/s41598-020-78431-x www.nature.com/scientificreports/ to bundling in MMK buffer (Fig. 1G,H). To observe this second stage more fully we reduced the slit width of the spectrofluorometer. In Fig. 1H the initial formation of single pfs is hidden in the lag phase, but after ~ 100 s light scattering increases strongly due to bundling.  PaZipA enhances PaFtsZ assembly and stabilizes PaFtsZ curved conformation in GTP. ZipA only exists in γ-proteobacteria. The identity of ZipA protein between E. coli and P. aeruginosa is around 30% ( Figure S1A). In general, ZipA is considered to be a Z-ring stabilizer based on early studies showing that ZipA from E. coli induced the FtsZ filament bundling 28,39 . However, in our recent study, we demonstrated that ZipA effects on EcFtsZ assembly varied with pH and may function as curved filaments enhancer, not Z-ring stabilizer 41 . To this end, we investigated the effects of PaZipA. PaZipA used in this study is an N-terminally truncated to remove the transmembrane domain for better solubility. Here we report similar results for PaZipA. PaZipA induces PaFtsZ bundling formation in MMK buffer at pH 6.5 (Fig. 2B,C). However, in HMK buffer (pH 7.5), which is close to the physiological condition, PaFtsZ pfs are still single, but more curved ( Fig. 2A) than that of PaFtsZ alone (Fig. 1A,B). The diameter of PaFtsZ curved filaments assembled with PaZipA is around 50-100 nm. It is worth noting that PaFtsZ-PaZipA mixture could also assemble into curved filaments in MEK (Fig. 2D) and HEK buffer (Fig. 2E); PaFtsZ alone could not assemble in HEK buffer (Fig. 1E). This indicates that PaZipA can both enhance PaFtsZ assembly and promote PaFtsZ curved conformation.
Similar to previous studies on ZipA from E. coli, PaFtsZ-PaZipA assembled into huge, tight bundles in pH 6.5 solution (Fig. 2B,C). We observed the single or double PaFtsZ filaments form tight bundles through the lateral contacts, cross-linked by most likely PaZipA proteins in MMK buffer (Fig. 2B,C, arrowed). In MEK buffer, the addition of PaZipA to PaFtsZ greatly enhanced assembly, producing thick filaments and bundles (Fig. 2D).
PaZapA induced PaFtsZ to assemble into large straight sheets and bundles. Unlike PaZipA, PaZapA induced PaFtsZ to form large straight sheets and bundles in most buffers we used, including HMK, MMK and MEK buffer ( Fig. 3A-D). No assembly was observed in HEK buffer (Fig. 3E), similar to PaFtsZ alone, but different from PaFtsZ with PaZipA. Apparently, PaZapA induced PaFtsZ to form loose bundling structures, and the distance between each PaFtsZ filament is relatively large, from 5 to 11 nm. We also observed some striations across the bundles, which might be ZapA cross-links (Fig. 3B,D, arrowed). The distances between ZapA molecules are different in Fig. 3B,D. It is around 3.5 nm in Fig. 3D, which suggests that ZipA binds to each FtsZ molecule. Meanwhile, it is about 7.9 nm in Fig. 3B, which indicates that every two FtsZ molecules have a ZapA binding.
Assembly kinetics of PaFtsZ are modulated by PaZipA or PaZapA. We compared the assembly kinetics of PaFtsZ, PaFtsZ plus PaZapA, and PaFtsZ plus PaZipA using the light-scattering assay. In HMK buffer (pH 7.5), PaZipA has mild effects on the kinetics of PaFtsZ assembly (Fig. 4A); it is consistent with our EM observation that PaZipA has small effects on the sizes of PaFtsZ filaments (Figs. 1A, 2A). It is a rapid assembly that took around 10 s to the plateau. The weak light-scattering signal is reflected in the single pfs formation. Meanwhile, PaZapA promoted PaFtsZ bundling formation, corresponding to a strong light-scattering signal. For this, we used the reduced slit width in the spectrofluorometer to follow the formation of large bundles. After Table 1. GTPase activity of PaFtsZ in the absence of or in the presence of ZipA or ZapA in MMK and HMK buffer. Cc critical concentration. a HMK buffer: 50 mM HEPES, 100 mM KAc, 5 mM MgAc, pH 7.5. b MMK buffer: 50 mM MES, 100 mM KAc, 5 mM MgAc, pH 6.5.

Samples
HMK buffer (pH 7.5) a MMK buffer (pH 6.5) b GTPase activity (GTP min -1 FtsZ -1 ) Cc (µM) GTPase activity (GTP min -1 FtsZ -1 ) Cc (µM)  There is an obvious contradiction between ZapA inducing FtsZ to form a huge bundle structure and only causing a small change in the GTPase activity of FtsZ. Therefore, we are curious about the effect of ZapA on the polymerization and subunit exchange of FtsZ pfs. Light scattering is strongly affected by the size of the scattering polymers, and therefore measures primarily bundling. We have previously devised fluorescence assays that report  The identity between FtsZ from E. coli and P. aeruginosa is around 62%. Even though both ZipA and ZapA from E. coli and P. aeruginosa are only less than 30% identical (around 50% positives) ( Fig. S1A and B), we found that their effects on the FtsZ pfs, including the sizes, shapes and properties, are very similar. EcZapA also had only a mild effect on the GTPase activity of EcFtsZ, a less than 20% reduction 27,41 . Therefore, we think that the effects and kinetic characteristics of ZapA from E. coli and P. aeruginosa are similar. Both EcZapA and PaZapA can induce FtsZ assembles into large straight bundles, but both of them have only mild effects on GTP hydrolysis activity of FtsZ. So we turned to EcFtsZ to apply the fluorescence assays to study the effects of ZapA. There are two fluorescence assays we used to measure the kinetics of assembly and subunits turnover of EcFtsZ. One assay uses a bodipy fluorophore was labeled at Cys151 of the double mutant EcFtsZ-T151C/Y222W in the N-terminal subdomain which is quenched by a nearby engineered Trp222 within van der Waals distance in the C-terminal subdomain 24,46 . Upon assembly a conformational change moves the bodipy away from the Trp, increasing its fluorescence. Another FRET (fluorescence resonance energy transfer) assay, which labeled EcFtsZ-L268C with fluorescein as donor and with tetramethylrhodamine as acceptor, can measure initial kinetics of assembly, and also the kinetics of subunit exchange at steady state 10 .
We then used the bodipy assay to see how EcZapA and EcZipA affected the kinetics of EcFtsZ assembly in HMK buffer (Fig. 5A). To minimize the impact of protein labeling, the samples we used contain only less than 10% Bodipy labeled FtsZ, mixed with 90% wild type FtsZ. Assembly of EcFtsZ rose rapidly to a plateau within 10 s, consistent with the rapid assembly of EcFtsZ single pfs. Assembly was almost identical with EcZipA, while EcZapA slowed EcFtsZ assembly somewhat, reaching a plateau in ~ 15 s (Fig. 5A). Figure 5B compares the assembly of EcFtsZ pfs measured by bodipy fluorescence with bundling measured by light scattering for the mixture of EcFtsZ and EcZapA. The bodipy reached a plateau by ~ 15 s, however, the light scattering showed a lag of ~ 10 s, and then rose more slowly to reach a plateau after 30 s. These results suggest that bundling formation might follow the single filaments assembly through the lateral contact of single filaments.
The exchange of subunits at steady state was measured by our FRET assay 10 . For this experiment EcFtsZ-L268C pfs labeled with fluorescein and tetramethylrhodamine were pre-assembled separately and then mixed. The FRET signal, tracked as the decrease in donor fluorescence, measures the exchange of subunits to form mixed pfs. Figure 5C shows that EcFtsZ pfs preassembled with EcZipA exchanged subunits at the same rate as EcFtsZ alone, with halftime around 3.0 ± 1.0 s. EcFtsZ pfs assembled with EcZapA showed four times slower exchange, with halftime about 12.3 ± 1.5 s. Apparently, ZapA induced FtsZ to form huge bundles is consistent with the four times slower of FtsZ pfs subunits exchanges, but there is a contradiction that there is only a small effect on the GTPase activity of FtsZ.

Discussion
ZapA induces FtsZ to form straight bundles and PaZipA enhances FtsZ curved conformation. Both ZapA and ZipA are positive regulators of the Z-ring. ZipA is an essential protein for cell division and we failed to knock it out from P. aeruginosa. It is reported that some FtsA mutants from E. coli may bypass the need of ZipA 42,43 . In contrast, ZapA is not necessary in bacterial division. Similar to the E.coli studies, knockout of zapA from P. aeruginosa has no obvious effect on bacterial growth and division (Fig. S2).
Previous studies showed both ZipA and ZapA to induce FtsZ to form bundles and thus considered both of them to be Z-ring stabilizers 28,29,37,39 . However, recent studies showed that ZipA from E. coli only induced FtsZ bundling when pH below 7 40,41 . Our results here using proteins from P. aeruginosa provide further information to confirm these results. PaZapA induced straight PaFtsZ bundling formation in vitro; meanwhile, PaZipA only caused bundling of PaFtsZ at pH 6.5, but not at pH 7.5. This ZipA effect is similar to our previous results with FtsZ and ZipA from E. coli 41 . In that study, we further demonstrated that EcZipA promoted and stabilized the FtsZ-GDP highly curved miniring conformation 41 . In the present study of P. aeruginosa proteins, we did not find In the previous study of the effects of ZipA from E. coli, we mostly focused on the highly curved filaments of diameter 20-30 nm. These highly curved structures appeared mostly at the end of the filament, or sometimes minirings , which is considered as the FtsZ-GDP bending conformation after GTP hydrolysis 41 . We checked it again, and we could also observe many pfs moderately curved, referred to as intermediated curved filaments in the EM images of EcFtsZ-ZipA in our previous published paper 41 . It is consistent with our reports here that ZipA promotes and stabilizes the FtsZ-GTP intermediated curved filaments. The intermediate curved conformation may also be the physiologically important one for forming the ring structures and generating the constriction force 2 . PaFtsZ assembly in HEK buffer gives us a good example of PaZipA also enhancing PaFtsZ assembly: 10 µM PaFtsZ alone is not enough to polymerize; however, 5 µM PaFtsZ plus 10 µM ZipA can assemble into long curved filaments in HEK buffer.
An apparent contradiction: ZapA bundling hardly affects GTPase, but significantly slows subunit exchange measured by FRET. One might expect that when FtsZ pfs are associated into bundles, the lateral contacts would slow treadmilling and subunit exchange. Indeed this appears to be the case for bundles induced by divalent cations (Ca 2+ and high concentration Mg 2+ ), which reduces significantly the GTPase activity and the recycling of subunits 11,47,48 . However, this seems not to be the case for bundles induced by ZapA. Previous studies showed that ZapA-induced bundles from B. subtilis and from E. coli only slightly inhibited GTPase activity of FtsZ 29,37 . Our results confirm this for PaFtsZ. Recently, two studies that measured the effect of ZapA on treadmilling of FtsZ, and found that ZapA did not slow the FtsZ treadmilling rate in vitro 49 and in vivo 50 .
Overall these results suggest that the bundling of FtsZ pfs by ZapA is loose enough that it does not significantly alter treadmilling. The interval distance between each filament of PaFtsZ-ZapA bundles is around 5-11 nm from our measurements, close to the previous results of 11-18 nm of EcFtsZ-ZapA 37 . Nucleotide hydrolysis is generally considered to be a measure of subunit exchange, because once an FtsZ subunit has hydrolyzed its GTP it needs to dissociate from the polymers in order to exchange GDP for GTP and undergo another round of assembly and hydrolysis. One possible explanation is the GTP/GDP exchanges may occur in the middle of filaments. Earlier suggestions that nucleotide might exchange within pfs are discussed in the previous publications 5, 51,52 and are finally contradicted by the structure of pfs of FtsZ from Staphylococcus aureus 53,54 . The structure shows that the GDP is trapped in a pocket with no channel that could permit its escape. Now that treadmilling is established as the mechanism for assembly dynamics [14][15][16]49,50 , we can assume that the rate of GTP hydrolysis at steady state measures the rate at which subunits associate to the bottom and dissociate from the top of pfs.
In contrast to the minimal effect of ZapA bundling on GTP hydrolysis, our FRET assay suggested a fourfold slowing of exchange of fluorescently labeled FtsZ subunits. The subunits exchanges include the cycles of pfs disassembly and re-assembly. The much slower subunits turnover rate is parallel to the longer and larger bundling formation, which is considered to decrease the recycling of subunits. However, it is not proportional to the changes in GTP hydrolysis activity and the rates of treadmilling. During treadmilling process, subunits are bound to one end and release from the other end. This apparent discrepancy between GTPase and FRET is still unclear. It suggests that the rate of GTP hydrolysis determines the rate of treadmilling, but it is not affected significantly by the length of FtsZ pfs, meanwhile, the rate of subunits turnover is related largely to the length of FtsZ filaments.

Three kinds of FtsZ filaments.
This study provides us more evidence about the bending conformation of FtsZ pfs. More and more data supported that there are 3 different kinds of FtsZ pfs: straight, intermediate curved and highly curved 2,41,55-57 . Early reports emphasized that EcFtsZ assembled into mostly straight pfs; however, Erickson and Osawa suggested recently that EcFtsZ pfs actually are a mixture of straight and intermediate curved 2 . These intermediate curved filaments of diameters around 100-300 nm may play important roles in Z-ring formation and force generation 2,5 .
Here we report that PaFtsZ filaments assembled with GTP are mixtures of straight and intermediate curved filaments, and that ZipA induces FtsZ to form more curved pfs. The intermediate curved filaments widely exist across the species. FtsZ from Caulobacter crescentus assembles into similar curved filaments 58,59 ; Cyanobacterial FtsZ assembles into toroid-like circle bundles of similar curvature 60,61 . Toroids and spiral bundles also were observed when EcFtsZ was assembled in crowding agents 62 , and FtsZ from B. subtilis assembled with GMPCPP or with PC190723, a cell division inhibitor 63 . How straight filaments change to the intermediate curved conformation is still unclear, since it is not coupled with GTP hydrolysis. These two types of filaments often exist together, and a single filament can have a straight and a curved segment.
Another type of curved FtsZ pfs is the highly curved, miniring conformation, with a diameter of 20-30 nm. This highly curved conformation seems to be favored following GTP hydrolysis. In our previous study of EcFtsZ, we observed that the miniring structures could be enhanced and stabilized by ZipA 41 . However, in the present study, PaZipA did not seem to enhance minirings of PaFtsZ with GTP, suggesting that this effect is highly dependent on species. Minirings-like structures were also observed in the cyanobacterial FtsZ pfs 61 .
How the Z-ring generates a constrictive force is still controversial. Clearly, the conformation changes among these three types of filaments may provide the force to bend the membrane. How these bending conformations work together with FtsZ's treadmilling dynamics in vivo is still unclear. We suggest that the GDP subunits at the minus end of FtsZ filaments switch to a highly curved structure before release, which may produce a continuous contraction force.

Conclusions
In conclusion, we investigated the unique assembly properties of FtsZ from P. aeruginosa and the effects of its two positive regulator proteins, ZipA and ZapA. We found that PaFtsZ pfs are mixtures of straight and intermediate curved filaments. The effects of ZipA on PaFtsZ assembly varied with pH. In physiological conditions, PaFtsZ-ZipA protofilaments were not bundled but promoted more curved filaments. It is consistent with our recent findings in E. coli 41 . It provides us further information that ZipA functions as an enhancer of FtsZ curved filaments during cytokinesis, which may contribute to the force generation to bend the membrane. Comparable to ZapA from other bacterial species, ZapA from P. aeruginosa induced PaFtsZ pfs to associate into long straight bundles and/or sheets, which had little effect on the GTPase activity of PaFtsZ. This suggests that the bundles are loose structures and it is consistent with ZapA has little effect on the treadmilling activity of FtsZ filaments 49,50 .

Methods
Protein purification. Expression vectors for PaFtsZ, PaZapA, N-terminal truncated PaZipA , and EcZapA and N-terminal truncated EcZipA  were constructed in the plasmid pET15b at the NdeI/ BamHI sites. The expression vectors were transformed into an E. coli strain BL21. Protein expression was induced at 16 °C overnight by the addition of 0.5 mM isopropyl β-d-1-thiogalactopyranoside. After sonication and centrifugation, the soluble His6 proteins were purified by affinity chromatography on a Talon column (Clontech Lab, Inc.). After washing with 0-30 mM imidazole, proteins were eluted with the elution buffer containing 50 mM Tris pH 7.7, 300 mM KCl, 300 mM imidazole. The purified His6-PaFtsZ was incubated with 2 units/ml of thrombin for 2 h at room temperature to remove the His-tag. A further purification followed by chromatography on a source Q 10/10 column (GE healthcare) with a linear gradient of 50-500 mM KCl in 50 mM Tris, pH 7.9, 1 mM EDTA, 10% glycerol.
The purified proteins were dialyzed into HMK buffer (50 mM HEPES, pH 7.5, 5 mM MgAc, 100 mM KAc), and stored at − 80 °C. GTPase activity measurement. GTPase activity was determined by a continuous assay coupled with GTP regeneration system, as described previously 41 . Our assay mixture included 1 mM Phosphoenolpyruvic acid monopotassium (PEP), 0.9 mM NADH, 10 units/ml pyruvate kinase and lactate dehydrogenase (Sigma-Aldrich), and 0.5 mM GTP. In this assay, when GTP is hydrolyzed to GDP, a NADH is consumed in the subsequent reaction, and the GDP in the solution is rapidly regenerated to GTP. The GTP hydrolysis rate is determined through the decrease in absorption of NADH using the extinction coefficient 0.00622 μM -1 cm -1 at 340 nm. A 3 mm path cuvette was used for measurement. Hydrolysis was plotted as a function of FtsZ concentration, and the slope of the line above the critical concentration (Cc) was taken as the hydrolysis rate. Measurements were made at room temperature with a Shimadzu UV-2401PC spectrophotometer. Each measurement repeats 2 or 3 times.
Negative stain electron microscopy. Negative stain electron microscopy was used to visualize FtsZ filaments as described previously 41 . Samples of PaFtsZs with or without ZipA or ZapA were incubated with GTP to polymerize for 1-5 min at room temperature. Then, 10 µl samples were applied to a carbon-coated copper grid for about 5 s and then quickly dried with filter papers. Grids were immediately stained with several drops of 2% uranyl acetate. Images were obtained on a Philips EM420 equipped with a CCD camera.
Light-scattering assay. The assembly kinetics of FtsZ filaments and bundles were measured by light scattering with a Shimadzu RF-5301 PC spectrofluorometer at room temperature. Both excitation and emission were set to 340 nm. Each measurement repeats 2 or 3 times.
Bodipy fluorescence quenching assay. EcFtsZ assembly kinetics were measured using the Bodipy fluorescence quenching assay as described previously 46 . Bodipy fluorescence can be efficiently quenched by a tryptophan that is close enough to form van der Waals contacts. A double mutant EcFtsZ-T151C/Y222W was labeled with Bodipy-FL N-(2-aminoethyl)maleimide (Thermo Fisher Scientific) at Cys 151, and nearby Trp 222 could quench Bodipy fluorescence efficiently. After FtsZ assembly, the bodipy fluorescence increased about 60% due to the conformational changes. Tracking the Bodipy fluorescence changes, we could obtain FtsZ assembly kinetics. For these measurements, the labeled EcFtsZ protein was mixed with a ninefold excess of unlabeled wild-type protein to avoid the changing properties of labeled protein. Bodipy fluorescence was measured at 515 nm, with excitation at 490 nm. Fluorescence measurements were taken with a Shimadzu RF-5301 PC spectrofluorometer at room temperature. Each measurement repeats 2 or 3 times. Scientific Reports | (2020) 10:21369 | https://doi.org/10.1038/s41598-020-78431-x www.nature.com/scientificreports/ FRET (fluorescence resonance energy transfer) assay. EcFtsZ filaments turn-over rate was measured using a FRET assay, as described previously 10,11 . A single cysteine mutant of EcFtsZ-F268C was labeled separately with Fluorescein 5-maleimide (Thermo Fisher Scientific) as donor and Tetramethylrhodamine 5-maleimide (Thermo Fisher Scientific) as acceptor. FtsZ subunits labeled with each fluorescent dye were mixed with EcZapA (or EcZipA) and polymerized separately for 5 min. The preassembled protofilaments and bundles were then mixed and subunit exchange of FtsZ was tracked using the decrease in donor fluorescence at 515 nm, with excitation at 470 nm. The data were fitted by a single exponential decay, F(t) = Fo + a × e −t/τ . Each measurement repeats 2 or 3 times.