FtsZ does not initiate membrane constriction at the onset of division

The source of constriction required for division of a bacterial cell remains enigmatic. FtsZ is widely believed to be a key player, because in vitro experiments indicate that it can deform liposomes when membrane tethered. However in vivo evidence for such a role has remained elusive as it has been challenging to distinguish the contribution of FtsZ from that of peptidoglycan-ingrowth. To differentiate between these two possibilities we studied the early stages of division in Escherichia coli, when FtsZ is present at the division site but peptidoglycan synthesizing enzymes such as FtsI and FtsN are not. Our approach was to use correlative cryo-fluorescence and cryo-electron microscopy (cryo-CLEM) to monitor the localization of fluorescently labeled FtsZ, FtsI or FtsN correlated with the septal ultra-structural geometry in the same cell. We noted that the presence of FtsZ at the division septum is not sufficient to deform membranes. This observation suggests that, although FtsZ can provide a constrictive force, the force is not substantial at the onset of division. Conversely, the presence of FtsN always correlated with membrane invagination, indicating that allosteric activation of peptidoglycan ingrowth is the trigger for constriction of the cell envelope during cell division in E. coli.

In most bacteria a large, highly dynamic protein complex called the 'divisome' assembles at the midcell, constricts the cell envelope and finally separates the mother cell into two daughter cells [1][2][3][4] . Despite considerable effort, it is not completely clear which proteins in the divisome that trigger membrane invagination, or contribute to constriction of the cell envelope.
At the center of the divisome complex is FtsZ, a protein that forms a structural framework for the rest of the divisome 5 . Numerous in vitro studies posit that FtsZ is a major force generator during constriction [6][7][8][9] . Perhaps the most convincing observation backing this postulate is that, when artificially fused to the membrane, FtsZ exerts a contractile force on liposomes 7,8,10 . Whether this contractile force is significant in vivo remains an important but unanswered aspect of cell division. A secondary source for envelope constriction most likely comes from peptidoglycan (PG) ingrowth. PG forms a rigid layer that needs to be remodeled at the division site; this process is carried out by amidases (AmiA, AmiB, AmiC) and their respective activators (EnvZ, NlpD), transglycosylases (PBP1A, PBP1B, MgtA) which together with FtsW forms the septal PG synthase complex, the sensory subcomplex FtsQ/L/B, a transpeptidase (FtsI) and its allosteric activator (FtsN) [11][12][13][14][15] . Perturbations to many of these PG synthesising/regulating enzymes inhibit division and result in filamentous cells. These filaments often contain multiple FtsZ-rings, that appear not to constrict, as judged by the relatively crude resolution obtainable by fluorescence microscopy [16][17][18][19][20] . However, it is not possible to determine if geometrical changes take place at a membrane level using fluorescence microscopy alone.
In order to determine whether a divisome protein is contributing to constriction of the envelope in vivo, it is essential to be able to monitor its localization at the divisome, and at the same time the ultra-structure of the membranes (to detect small invaginations). To achieve this we have, in this study, used correlative cryo-fluorescence and cryo-electron microscopy (cryo-CLEM) as it is currently the only method that enables tracking of fluorescently labeled proteins in vivo and high-resolution cryo-electron microscopy (cryo-EM) imaging of membrane ultra-structure in the same cell 21,22 (Supplementary Fig. 1 for an overview of the workflow). Using this approach we have monitored the contributions of FtsZ, FtsI and FtsN during the initial stage(s) of envelope constriction.
tethers FtsA 23 and ZipA 24 (aka the proto-ring) are present at midcell (but PG synthesizing enzymes are not). We cryogenically preserved Escherichia coli cells expressing a chromosomal copy of FtsZ-GFP by vitrification and imaged them by cryo-CLEM ( Supplementary Fig. 2). When examining cells in a pre-divisional stage, indicated by the typical helical arrangement of FtsZ-GFP 25,26 , we never observed membrane invagination (Fig. 1a). In cells where FtsZ-GFP had condensed to a single band (as judged from the cryo-fluorescence images) we noted that 27% lacked a visible invagination (as judged from the cryo-EM images) (Fig. 1b). The cells that had visible membrane invaginations, could be further classified as having either minor (~17%) (Fig. 1c) or major invaginations (~56%) (Fig. 1d). In all cells where membrane invagination was observed we noted that both the inner and outer membranes were equally deformed (see also Supplementary Fig. 3). These observations suggest that the presence of FtsZ-GFP at the division septum is not sufficient to deform the inner membrane in vivo.
The proto-ring assembles at one point in time. The inability of FtsZ to deform the inner membrane in vivo could be explained by the possibility that it was arriving at the midcell before ZipA and FtsA, and was therefore not yet tethered to the inner membrane. Although previous studies have shown FtsZ, ZipA and FtsA arrive at the division site at essentially the same point in time 23,27 , we validated this point using our experimental setup. We co-expressed FtsZ-GFP and ZipA-mCherry and monitored their localization at the mid-cell by live cell dual color fluorescence microscopy imaging (Fig. 1e). In all cells analyzed (n > 100), both fusions co-localized at the mid-cell. Or stated in an alternative way, there was no instance where FtsZ-GFP was present at the midcell and ZipA-mCherry was not. The same observation was made when we co-expressed a plasmid-encoded version of FtsZ-mCherry and a chromosomally encoded version of FtsA-GFP (Fig. 1f). As a result we can conclude that the whole proto-ring is assembled in our cryo-CLEM experiments on FtsZ-GFP. The most plausible explanation for our earlier observations using cryo-CLEM is that FtsZ was tethered by ZipA and FtsA, but that it was simply not generating a sufficient force to deform the inner membrane.

Cephalexin treated cells does not show membrane invagination at potential division sites.
Additional evidence that the proto-ring containing FtsZ, FtsA and ZipA is not generating a sufficient force to deform the inner membrane was obtained by adding the antibiotic Cephalexin (15 μ g/ml) to cultures of cells expressing FtsZ-GFP and analyzing potential division sites by cryo-CLEM ( Supplementary Fig. 7). Cephalexin allows the early division proteins to assemble, but inhibits the catalytic activity of a down stream protein (FtsI) which in turn blocks cell division and renders cells filamentous 28,29 . We reasoned that if FtsZ would exert a large enough force to constrict membranes in vivo we would detect membrane invaginations in the filaments at potential division sites using cryo-CLEM. Cephalexin treated cells did not contain membrane invaginations at spots where FtsZ-GFP was accumulated (n = 27) ( Supplementary Fig. 7). These data taken together with the previous cryo-CLEM data strongly support a scenario in which FtsZ-GFP does not generate a large enough force to invaginate membranes in vivo.
Significantly, immunofluorescence microscopy has shown that FtsZ-GFP mirrors the localization patterns of its native counterpart, thus we can be confident that also the native FtsZ was present in cells lacking visible invaginations [30][31][32] .

FtsN accumulation at midcell triggers membrane invagination.
If FtsZ is not sufficient to initiate the membrane invagination, what is? The other plausible candidate is peptidoglycan (PG) ingrowth 1 , a process largely governed by two divisome proteins that arrive at a later time point than FtsZ, FtsI and FtsN 33 . FtsI is a PG synthase that is thought to be allosterically activated by FtsN 12 . FtsN is the last essential protein to arrive at midcell and is generally believed to be responsible for the initiation of membrane invagination 12 , however it has not been possible to unequivocally resolve this point because available methods such as super resolution fluorescence microscopy were unable to correlate the presence of FtsN with membrane invagination. Moreover recent data has suggested that a portion of the septal localized FtsN is recruited early by FtsA 34 , and that this portion of FtsN is independent of FtsI localization to midcell 35 . These observations raised the possibility that membrane invagination could be initiated earlier than previously thought.
To determine if the accumulation of FtsI or FtsN to midcell was sufficient to initiate constriction of the inner membrane in vivo we expressed either a chromosomal encoded GFP-FtsI or GFP-FtsN and analyzed cells by cryo-CLEM. Here we present pseudo time-lapse images of respective protein during different stages of division ( Fig. 2a-h). In cells where GFP-FtsI was accumulated at the septum, we observed that roughly 35% were without a visible membrane invagination at the midcell (Fig. 2a). Thus accumulation of FtsI at the septum is clearly not enough to initiate membrane invagination. The remaining 65% of cells showed both GFP-FtsI accumulation and membrane invagination (Fig. 2b). These numbers were comparable to what we observed for FtsZ-GFP, suggesting that FtsI is not sufficient to initiate constriction of the membranes.
Conversely, in cells where GFP-FtsN was accumulated at midcell we observed that 100% of cells (n = 65) contained visible membrane invaginations (Fig. 2f,g). These invaginations again indicated a uniform constriction of both the inner and outer membranes and could either be minor (Fig. 2f) or major (Fig. 2g). Even though we cannot formally rule out the possibility that membrane invagination is initiated prior to FtsN arrival at the septum, the data presented here strongly suggests that FtsN accumulation at midcell is a trigger for membrane invagination. Finally, we also noted that both GFP-FtsI and GFP-FtsN remained at the septum until complete closure of the cell envelope (Fig. 2d,h).

Discussion
In summary, we have utilized cryo-CLEM to study the early stages of bacterial cell division. This novel approach allowed us to correlate the localization of specific proteins at the division septum with the underlying membrane geometry. The latter allowed us to go far beyond the current resolution limit of light microscopy and observe minor membrane invaginations that mark the onset of envelope constriction. Our data unequivocally demonstrate that the presence of FtsZ (together with ZipA and FtsA) at midcell, is not sufficient to initiate membrane constriction in vitro. This conclusion does not contradict previous studies that show that membrane anchored FtsZ polymers can generate a force on membranes in vitro 7,8,36 . But it does suggest that the force generated is not strong enough to initiate membrane invagination in vivo. Taking into account other recent findings, which show that (1) FtsZ activity is not a rate-limiting factor during membrane constriction 30 , (2) that it is not present at a late stage of division 31 , and (3) the fact that cell wall free E. coli can undergo division in absence of FtsZ 37 , brings its role as a main force generator into question. Nonetheless it is still possible (even probable) that it contributes to constriction.
In the course of our experimentation we also observed that the assembly of FtsI was not sufficient to initiate constriction. However, once its allosteric activator FtsN had assembled we observed that all cells had visible invaginations. The simplest interpretation of this latter observation is that FtsN acts as the trigger for envelope constriction, by activating FtsI, other sensory proteins (FtsQ/B/L) and PG in-growth. Although this claim has been made previously 14,15,28 , it was based on low-resolution fluorescence imaging, which cannot resolve the ultra-structure of the membrane. Our cryo-CLEM data are therefore the first to emphatically prove this point. Methods Cryo-CLEM workflow. The workflow of Koning et al. 38 was followed loosely. Each individual step is detailed below and schematically depicted in Supplementary Fig. 1. Briefly, cells were grown and the respective fluorescent protein fusion was produced by the addition of Isopropyl β -D-1-thiogalactopyranoside (IPTG), thereafter an aliquot was transferred to a Zeiss Elyra SP.1. for a pre-plunging fluorescence check using Structured Illumination Microscopy (SIM). Following SIM imaging, another aliquot from the same culture was mixed (10:1) with 20 nm gold beads, and 1-5 μ l sample was directly applied on glow discharged EM girds and plunge frozen in liquid Ethane using a Vitrobot (FEI company, The Netherlands). Cryo-fluorescence images were acquired using a Linkam cryo-stage (Linkam Scientific, UK) mounted on an upright Zeiss Axio Imager A2 microscope, and cells with different fluorescence localization were identified. The grids were subsequently transferred to a Talos ARCTICA cryo-TEM (FEI Company, The Netherlands), where the same cells again were identified in low magnification mode before high magnification imaging was preformed. Data correlation was finally performed using ImageJ and Adobe Photoshop. Note that EM grids are round and may rotate during sample transfer, thus cryo-fluorescence and cryo-TEM images may be at slightly different angles with respect to each other. Strains. Construction of strains BS001, BS003 and BS007 carrying chromosomal encoded copies of FtsZ-GFP, FtsA-GFP and GFP-FtsI respectively, is described elsewhere 31 . Strain EC1213 carrying a chromosomal encoded copy of GFP-FtsN (MG1655 attHK22::pDSW560 (lacIq, P207::gfp-ftsN)) 32 was a gift from David Weiss (University of Iowa). All strains and plasmids used in this study can be found in Supplementary Table 1. Growth conditions. Cells were grown (37 °C) in LB overnight and back-diluted 1:200 in fresh media and grown for 1 hour before inducing production of fluorescent proteins. Antibiotics were added at the appropriate concentration (Ampicillin 25 μ g ml −1 , Kanamycin 50 μ g ml −1 , Cephalexin 15 μ m ml −1 ). Cells in early exponential growth phase were harvested and imaged by SIM and cryo-CLEM. Fluorescent protein production. As described before 31,32 , production of FtsZ-GFP, FtsA-GFP, GFP-FtsI and GFP-FtsN was induced with Isopropyl β -D-1-thiogalactopyranoside (IPTG) at the following concentrations; FtsZ-GFP, 2.5 μ M; FtsA-GFP, 100 μ M; GFP-FtsI and GFP-FtsN, 5 μ M. FtsZ-mCherry was produced from pEG4 39 and ZipA-mCherry 32 , as described.
Structured Illumination Microscopy (SIM) Imaging. ~6 μ l of cell culture was placed on a microscope glass coated with thin pre-made agarose pads (1% (w/v) agarose). A cover slip was added and the cells were left for ~5 minutes so that they had sufficient time to immobilise. SIM images were acquired on a Zeiss ELYRA PS.1 system using a 100X NA 1.4 oil immersion plan-Apochromat objective. GFP was excited at a wavelength of 488 nm and mCherry at 561 nm. To minimize bleaching and photo damage to the cells during imaging, the laser power output was kept below 2,5% for the 488 nm laser, and 5% for the 561 nm laser (total laser power was 5 and 10 mW respectively). The calibrated SIM system gave a lateral (xy-) resolution of ~100 nm. For each fluorophore the illumination grids were phase shifted 3 times over the field of view (1024 × 1024) and then rotated 180/3 degrees (i.e. three rotations per image) to generate a data set containing 9 raw images per fluorophore. Appropriate laser and filter settings for the respective fluorophore were chosen in the software ZEN2011 Black. The SIM images were reconstructed and subsequently analysed in ZEN2011 Black or ImageJ (N.I.H.). Imaging was preformed at controlled room temperature of 23.5 °C.
Vitrification. Quantifoil grids (R3,5/1 or R2/2 Cu 200 mesh) were glow discharged and 3-5 μ l of sample solution was directly applied. Grids were blotted using Whatman filter papers and plunge frozen using a Vitrobot (FEI Company, The Netherlands) in pre-cooled liquid Ethane. The samples were subsequently stored under liquid nitrogen conditions until further examination.

Cryo-Fluorescence Microscopy imaging. A Linkam CMS196 cryo-stage was mounted on an upright
Zeiss Axio Imager A2 microscope body equipped with an Olympus DP73 Peltier cooled CCD camera, a (cryo compatible) Zeiss air LD EC Epiplan-NEUFLUAR 100 × 0,75 NA objective with a working distance of 4.0 mm and an 'GFP' filter cube (Ex. 470/40. FT 495. Em. 525/50). Images were recorded using Olympus cellSens software (Olympus, Japan). Once a grid was examined it was directly removed form the cryo-stage, mounted into a FEI autogrid and again stored in liquid nitrogen until further examination.

Cryo-Transmission Electron Microscopy imaging.
Grids that already had been examined by cryo-fluorescence imaging and found to contain interesting regions of interests were transferred to a Talos ARCTICA 200 keV cryo-transmission electron microscope equipped with a Falcon II direct electron detector. The regions of interests identified in the cryo-fluorescence step were manually relocated using low magnification mode (84× -210× ). Once the same area as in the fluorescence images (i.e. the cells) had been localized, high magnification (4300× -57000× ) imaging was preformed using low dose mode.