Slow unloading leads to DNA-bound β₂-sliding clamp accumulation in live Escherichia coli cells

Abstract

The ubiquitous sliding clamp facilitates processivity of the replicative polymerase and acts as a platform to recruit proteins involved in replication, recombination and repair. While the dynamics of the E. coli β₂-sliding clamp have been characterized in vitro, its in vivo stoichiometry and dynamics remain unclear. To probe both β₂-clamp dynamics and stoichiometry in live E. coli cells, we use custom-built microfluidics in combination with single-molecule fluorescence microscopy and photoactivated fluorescence microscopy. We quantify the recruitment, binding and turnover of β₂-sliding clamps on DNA during replication. These quantitative in vivo results demonstrate that numerous β₂-clamps in E. coli remain on the DNA behind the replication fork for a protracted period of time, allowing them to form a docking platform for other enzymes involved in DNA metabolism.

Introduction

The multi-protein replisome complex (replisome, Fig. 1a) is responsible for the accurate and timely duplication of the genome before cell division. The sliding clamp protein complex is a key subunit of the replisome and is vital for protein–DNA interactions related to DNA metabolism in all three domains of life^1,2,3. Through their interaction with polymerases, DNA ligase, replication initiation protein DnaA, the dynamin-like protein CrfC, as well as different mismatch-repair proteins, sliding clamps have important roles in replication and repair^{4,5,6,7,8,9,10,11,12,13,14,15}. In E. coli, the β₂-sliding clamp (β₂-clamp) is a homo-dimer¹⁶ (Fig. 1a (inset)) that encircles double-stranded DNA (dsDNA) and tethers DNA Polymerase III (DNA Pol III) to the template, thereby ensuring sufficiently high processivity during synthesis^17,18.

Figure 1: The E. coli replisome and β₂-clamp assembly during replication.

(a) The position of the β₂-sliding clamp within the E. coli replisome complex. The helicase (DnaB) unwinds dsDNA ahead of the replicative polymerase (DNA Pol III), which subsequently duplicates the template strands. Different configurations of Pol IIII are potentially possible. Primase (DnaG) synthesizes short RNA primers on the lagging strand for Okazaki fragment initiation. Single-stranded binding proteins (SSB) remove the secondary structure of ssDNA and protect it from digestion. To ensure sufficient processivity during replication, Pol III is tethered to the DNA by the β₂-sliding clamp. β₂ is assembled onto primer-template junctions by the multi-protein ((τ/γ)₃δδ′χ) clamp-loader complex. (Inset) A ribbon representation of the DNA-bound β₂-sliding clamp (generated using the Protein Data Bank (PDB) file, 2POL¹⁶). The β₂-sliding clamp is a homo-dimer that consists of six globular domains¹⁶. The monomers are arranged in a ring that encircles the DNA⁷⁰ and can slide freely along it. Different proteins can bind to the two hydrophobic pockets of the β₂-clamp via a conserved sequence motif¹⁰. (b) The life cycle of the β₂-clamp during replication. (top) The β₂-clamp is actively loaded by the clamp-loader, which opens the closed clamp and places it onto dsDNA before release²². (middle) The β₂-clamp remains DNA-bound as long as an Okazaki fragment is being synthesized. (bottom) After the β₂-clamp has reached the end of an Okazaki fragment, DNA Pol III is signalled to release⁵². The β₂-clamp is believed to be disassembled by the clamp-loader.

The β₂-sliding clamp is actively assembled and disassembled onto DNA during the synthesis of the two complementary DNA strands (Fig. 1b). The loading reaction of a β₂-clamp onto each new primer-template junction¹⁹ is catalysed by an ATP-dependent heteropentameric clamp-loader complex (clamp-loader), also known as the γ-complex²⁰. The clamp-loader pries open the β₂-clamp, recognizes the primer-template junction²¹ and closes the β₂-clamp around the dsDNA before release²². The clamp-loader is also thought to chaperone DNA Pol III onto a newly loaded β₂-clamp²³ and to unload inactive DNA-bound β₂-clamps via the δ-subunit²⁴. During all of these reactions, the loader complex and the various clamp binding proteins compete for the carboxy (C)-terminal face of the clamp. In accordance with the proposed model in which the replisome includes three core DNA polymerase IIIs^25,26,27, three β₂-clamps can be active at the replication fork, one for each of the three polymerases (Fig. 1a). While leading-strand replication is thought to be continuous, utilizing only a single β₂-clamp, the lagging-stand template is copied in discrete 1–2 kb Okazaki fragments²⁸, each utilizing a separate β₂-clamp. These fragments are initiated by the continuous formation of 10–12 nt RNA primers by the primase (DnaG), which, together with the helicase (DnaB), sets the replication fork clock²⁹. Since the number of Okazaki fragments (2,000–4,000) for the 4.6 Mbp genome is roughly an order of magnitude greater than the average number of β₂-clamps per cell in a nutrient-rich culture^24,30, continuous recycling of β₂-clamps is necessary for total genome replication to occur.

Despite numerous in vitro and in vivo studies, it still remains unclear whether recycling of the E. coli β₂-clamps takes place immediately following the completion of an Okazaki fragment, or at a later time. A slow recycling could permit a β₂-clamp to fulfil additional functions, while remaining bound to the newly synthesized DNA. Quantitative in vitro unloading assays^24,31 indicate that in the absence of the clamp-loader, a loaded β₂-clamp has a long half-life of t_unload>1 h on the DNA. Although this is decreased by more than an order of magnitude to t_unload~127 s per β₂ in the presence of clamp-loader, this unloading time still remains long compared with the typical time required to complete an Okazaki fragment (on the order of seconds). Such a slow unloading time suggests that many β₂-clamps are left behind in the wake of the replication fork³². However, a recent in vitro single-molecule study indicates that lagging-strand synthesis can persist in vitro in the absence of excess β₂-clamps in solution, implying that a β₂-clamp can be directly reused at a successive primer-template junction³³. Two in vivo studies, one in Bacillus subtilis (B. subtilis)³⁴ and the other in E. coli²⁶, provided contrasting results. Hence, to understand the regulatory mechanism that underlies the recycling of β₂-clamps in E. coli, further insights into their in vivo dynamics are required.

To gain detailed insight into the in vivo recruitment and turnover of the β₂-clamp, we investigate its dynamics in individual live cells with single-molecule sensitivity. We use both conventional fluorescence microscopy and Photoactivated Localization Microscopy (PALM)^35,36, in combination with custom-built microfluidics. Single-molecule techniques have provided us with insights into the dynamics of processes—such as replication, transcription and translation—that are not readily accessible with conventional ensemble-averaging techniques^37,38. In vivo single-molecule fluorescence imaging, in particular, has provided detailed insights into the behaviour of individual molecules in live cells^39,40,41. Combining single-molecule fluorescence microscopy with microfluidics allows us to image individual molecules in live cells over multiple cell cycles, without chemical fixation that could potentially perturb the dynamic behaviour of the protein under investigation⁴².

By using this experimental approach, we have measured the number of DNA-bound β₂-clamps during chromosomal replication over the entire course of the cell cycle. In addition, we have determined the time required to unload an individual DNA-bound β₂-clamp during replication, as well as the effective time required to load a new β₂-clamp during replication. Our data reveal that the number of DNA-bound β₂-clamps accumulates on the DNA after initiation, and then levels off to a constant steady-state number of DNA-bound β₂-clamps on the order of minutes. This steady state is maintained throughout the rest of the replication process, until termination occurs and a concomitant decrease of DNA-bound β₂-clamps is observed. The number of DNA-bound β₂-clamps in steady state exceeds the estimates of a previously published in vivo study²⁶ by an order of magnitude. The measured values for the effective loading time and unloading time during replication, in the context of the live cell, are in good agreement with previous biochemical in vitro experiments²⁴. Taken together, our data indicate that a β₂-clamp remains on the DNA for a protracted period of time following the completion of an Okazaki fragment. DNA-bound β₂-clamps that are left behind during fork progression may facilitate the recruitment of additional proteins active during the cell cycle for different processes such as DNA repair.

Results

The in vivo dynamics of β₂-clamps measured in single cells

To study the dynamics of β₂-clamps by wide-field fluorescence microscopy, we perform long time-lapse imaging of labelled β₂-clamps over multiple replication cycles. During such experiments, we ensure healthy cell physiology by implementing a custom-built microfluidic device (Fig. 2a; see Methods)^43,44 in which cells growing in steady state are immobilized in micron-sized growth channels. Through a neighbouring central trench, growth medium is continuously supplied throughout an experiment (see Methods). In such a microfluidic device, cells experience minimal perturbation over the course of the time-lapse experiment, as stable growth conditions remain continuously present. This contrasts with long time-lapse experiments performed on agarose pads in which nutrients and water may become depleted, leading to non-steady state cell populations as a result. Additional benefits of such a device are that daughter cells ultimately grow out of the growth channels, preventing the accumulation of cells, and that the cells are always aligned, which facilitates data analysis.

Figure 2: Long time-lapse fluorescence microscopy of the β₂-sliding clamp at the single-cell level utilizing microfluidics.

(a) The microfluidic device used for performing long time-lapse fluorescence microscopy. E. coli cells are immobilized in growth channels perpendicular to a main trench through which growth medium is actively pumped. (inset) A brightfield image and corresponding YPet–β₂ fluorescence image (80 ms laser light exposure) are acquired every 2.5 min for the duration of the time-lapse experiment. Scale bar, 3 μm. (b) YPet–β₂ molecules that are either DNA-bound or freely diffusing are studied using wide-field fluorescence microscopy. (left) Freely diffusing YPet–β₂ molecules in the cytoplasm of a cell. This signal is representative to YPet–β₂ dynamics before and after replication. (middle) A clear focus is observed due to DNA-bound YPet–β₂ molecules. The observation of a single focus, instead of two distinct foci, shortly after initiation results from the overlap of diffraction limited spots. (right) Two distinct foci are visible, indicative of two individual replisomes. Scale bar, 800 nm. (c) Kymograph of a single growth channel during an overnight time-lapse experiment. The cells first grow the growth channel full, and maintain a steady state growth rate as can be observed from the curved shape of the fluorescence signal. The shape of the fluorescence signal is due to the individual cells growing and pushing each other in the direction of the main trench. Clear observable diffuse patterns occur at regular intervals, indicative of no DNA-bound β₂-clamps before initiation or after termination. This repeating pattern is due to the multiple cycles of replication (indicated with repeating white dashed lines). (d) A kymograph of an individual replication cycle indicated in c. The blue lines are the cell boundaries detected from the brightfield images. The illustrations on the right-hand side indicate the different stages of replication that can be observed during the cell cycle.

Labelling of the β₂-clamp was accomplished by using a functional amino (N)-terminal YPet⁴⁵ fusion²⁶ (Supplementary Fig. 1; Supplementary Table 1; see Methods) expressed from (and replacing) the endogenous E. coli dnaN gene locus. Fluorescence images are acquired under shuttered 515 nm laser excitation (see Methods; Fig. 2a (inset)). Fluorescence images of YPet–β₂ within individual cells either yielded no (Fig. 2b, left), a single (Fig. 2b, middle) or two cellular foci (Fig. 2b, right), depending on the stage of replication, in agreement with previous reports of fluorescently labelled replisome components⁴⁶. Before each fluorescence image, a brightfield image is acquired to provide details of the cell periphery (Fig. 2a (inset)). This alternating imaging sequence has a sufficiently long period to avoid giving rise to any notable deleterious growth effects, as assessed by comparing the doubling time of cell growth in a shake flask with cells grown in the microfluidic device (Supplementary Figs 1 and 2).

Using this approach, we are able to observe numerous consecutive replication (and corresponding division) cycles of cells in the different growth channels. We examine the global replication dynamics of multiple cells within a growth channel by converting the time-lapse images into a kymograph (Fig. 2c; see Methods). A distinct reoccurring pattern indicative of multiple replication cycles in the generations of cells is clearly noticeable (indicated by repeating dashed lines in Fig. 2c). Under these experimental conditions (see Methods), the analysis of individual cells (n=137) in our microfluidic system yields an average replication time of t_rep=68±10 min, and a doubling time of t_double=84±17 min (Supplementary Fig. 2). In both cases, the error is ±s.d. We further analyse these kymographs to investigate the subcellular dynamics of the YPet–β₂ molecules within the individual cells from cell birth till cell division (Fig. 2d). One can clearly observe the dynamics of the two β₂-clamp foci associated with the two independent replisomes.

The assembly and accumulation of β₂-clamps on DNA

We use the fluorescence intensity from the YPet–β₂ fusion to determine the number of β₂-clamps that are DNA-bound as well as in the total number in the cell during the life cycle of a cell. A sample montage of the YPet–β₂ fluorescence signal from cell birth till after cell division (Fig. 3a) illustrates that there is a distinct increase in the foci following the B-period⁴⁷ of the cell cycle (represented by a diffuse signal after birth) (Fig. 3a (inset)) and a similar decrease before cell division. The fraction of fluorescence that originates from DNA-bound YPet–β₂ (foci) provides clear evidence that >50% of β₂-clamps are DNA-bound shortly after the initiation of replication (Fig. 3b). The steady decline in the fraction of DNA-bound β₂-clamps that commences roughly 10 min after initiation, results from the increase of total number of β₂-clamps in the cell as the cell grows. In assessing this intensity fraction, we verified that very little out-of-focus fluorescence escapes detection (Supplementary Note 1; Supplementary Fig. 3).

Figure 3: Quantification of the in vivo β₂-sliding clamp stoichiometry during replication.

(a) A representative temporal montage of the YPet–β₂ fluorescence signal from before initiation until after cell division. A clear intensity increase is observed at the focus formation following initiation (indicated with white arrows in inset) Scale bar, 1.6 μm. (b,c) Traces of focus formation in individual cells. In b, we plot the fraction of YPet–β₂ molecules that are DNA-bound compared with the total number in the cell. More than 50% of the total YPet–β₂ molecules are DNA-bound. The gradual decline in this fraction results from the increase of β₂ during the cell cycle. The inset indicates how the DNA-bound YPet–β₂ molecules and the total YPet–β₂ in the cell are defined. In c, we plot the absolute number of DNA-bound YPet–β₂ molecules. Here, the gradual increase, steady state and gradual decrease of the DNA-bound YPet–β₂ molecules can clearly be seen. In both b and c, the traces have been aligned with respect to initiation. (d,e) The average behaviour of individual YPet–β₂ molecules measured in individual cells. (d) The fraction of DNA-bound YPet–β₂ molecules is on average >50% half way through the replication cycle. (e) The YPet–β₂ molecules in the whole cell (blue curve) approximately doubles during the cell cycle, from 60 to 120 YPet–β₂ molecules. The DNA-bound YPet–β₂ molecules (red curve) remarkably increases to a mean steady state value of 46 YPet–β₂ molecules (s.d.=12, s.e.m.=1) following initiation. This value is maintained throughout the replication process until a concomitant decrease is observed after or during termination. Individual traces have been normalized with respect to initiation and termination to make averaging possible. (inset) A histogram of the distribution of number of DNA-bound YPet–β₂ molecules during steady state. (n=137).

A constant number of DNA-bound β₂-clamps is maintained

To precisely quantify the number of DNA-bound β₂-clamps as a function of the replication cycle, we exploited a single-molecule in vitro calibration method²⁶ that allows us to reliably convert the detected YPet–β₂ signal into an absolute number of molecules (Supplementary Note 2). We immobilize single purified YPet molecules on a cover glass and determine the average intensity of a single YPet fluorescent protein under these conditions (Supplementary Fig. 4). Using this calibration, we perform control stoichiometry experiments of previously studied DNA-bound replisome components²⁶, specifically the ε-subunit of Pol III and the τ-subunit of the clamp-loader to verify that our in vitro single-molecule calibration remains reliable in vivo⁴⁸ (Supplementary Fig. 5). For both the proteins, we reproduced the stoichiometry for the pair of sister replisomes as previously published²⁶, namely 5.74±0.04 molecules in total for the ε-subunit (n=64) and 6.12±0.03 molecules in total for the τ-subunit (n=66). Here the error is ±s.e.m. We also verified that a YPet–β intensity standard provides the same mean intensity value under our experimental conditions (Supplementary Fig. 4g). Therefore, we subsequently use this average intensity value to estimate the number of β₂-clamps in our experiments. In calculating the number of β₂-molecules, we correct for photobleaching (Supplementary Note 3; Supplementary Fig. 3) and verify that the fraction of immature, dark YPet proteins is negligible (Supplementary Note 4, Supplementary Fig. 3). In our conversion from intensity to molecules, we also take into account that β₂-clamps are dimers by dividing the measured YPet signal by two. This is a realistic assumption as it is believed that β₂-clamps are in closed conformation even when they are not DNA-bound²².

Using this calibration standard, we quantify the absolute number of DNA-bound YPet–β₂ molecules for individual traces of DNA replication. Representative individual time-traces of single cells clearly demonstrate that following the initiation of replication, a gradual increase of the number of DNA-bound β₂ occurs until a steady state plateau is reached (Fig. 3c). This plateau is maintained until decrease is observed shortly after or during termination (Fig. 3c). From the individual traces, one can observe that there is significant cell-to-cell variability in the absolute number of DNA-bound β₂-clamps, but that the overall trend in which the number of DNA-bound β₂-clamps is constant for a significant fraction of the cell cycle is the same for all cells. We compared this temporal behaviour with that of a different replisome component, the τ-subunit of the clamp-loader in a strain in which both the β₂-clamp and the τ-subunit are fluorescently labelled (Supplementary Note 5). The τ-YPet fluorescence signal fluctuates strongly and does not yield a stable plateau, in contrast to the mCherry–β₂ fluorescent signal (Supplementary Fig. 6).

To obtain statistically significant values for both the total number of β₂-clamps in the cell and the number of DNA-bound β₂-clamps, we extracted the average behaviour from analysis over numerous cells (n=137; Fig. 3d,e). Figure 3d clearly depicts that for an average cell, the fraction of DNA-bound β₂-clamps is more than half of the total content in the cell, which decreases to roughly zero after termination. During the cell cycle, an average cell doubles its YPet–β₂ content from ~60 to 120 molecules. This number of β₂-clamps in the cell is in good agreement with ensemble western estimates we performed under the same growth conditions (Supplementary Note 6; Supplementary Fig. 7). Remarkably, the number of DNA-bound YPet–β₂ is held at a stable value of (s.d.=12, s.e.m.=1; Fig. 3e (inset)). We also observe that the number of DNA-bound β₂-clamps are close to zero before initiation and after termination. We ruled out the presence of an ectopic dnaN gene, by verifying that only a single copy of the dnaN gene is present in the strain that we used for these experiments (Supplementary Note 7; Supplementary Fig. 7). The experiment was successfully reproduced with a different fluorescent protein fusion (mCherry–β₂), which strengthens the argument that accumulation is unlikely the result of fluorophore aggregation^49,50, but rather due to physiological build-up of DNA-bound clamps (Supplementary Note 8; Supplementary Fig. 8). The slightly lower mean number of DNA-bound clamps (, s.d.=12, s.e.m.=1.5) as measured using the mCherry–β₂ protein fusion, in combination with the mCherry intensity calibration, possibly results from the less ideal photophysical properties of mCherry, which make it less suitable than YPet for rigorous quantitative fluorescence microscopy.

Single β₂-clamps are not rapidly unloaded in vivo

To study the in vivo unloading time of an individual β₂-clamp, we utilized single-molecule PALM (Fig. 4a). The endogenous dnaN gene was replaced with a functional N-terminal PAmCherry⁵¹ fusion (see Methods). Fluorescence images are acquired under shuttered 561 nm excitation (see Methods), while activation is performed once with low 405 nm laser illumination, such that on average less than one DNA-bound PAmCherry–β per cell is activated. Before fluorescence activation, a phase-contrast (PH) image is acquired to determine the cell’s position and its periphery. Sample pre- and post-activation images (Fig. 4b), together with the corresponding line-profile intensity plots (Fig. 4c), demonstrate successful activation of individual DNA-bound PAmCherry–β molecules in our strain. The advantage of PALM over more conventional techniques like Fluorescence Recovery After Photobleaching and Fluorescence Loss In Photobleaching for measuring protein turnover is that it allows one to directly image a single unloading event, as shown in the sample temporal montage and the corresponding integrated intensity trace (Fig. 4d). We image a different field of view of cells for each complete PALM measurement sequence (Fig. 4a) to ensure that the cell physiology and β₂-clamp behaviour are not influenced by excessive 405 nm light exposure. Using the individual analysed traces from different cells, we are able to build up a distribution for the on-time events of single PAmCherry–β molecules (Fig. 4d). To visualize a single unloading event, we only image one out of the two PAmCherry–β₂ dimer subunits. After correcting for photobleaching (Supplementary Note 9; Supplementary Fig. 9), we estimate the in vivo unloading time to be t_unload=195±58 s per β₂ (Fig. 4e). This result is in good agreement with previous in vitro experiments (127 s per β₂; ref. 24).

Figure 4: Direct measurement of the in vivo unloading time of the β₂-sliding clamp during replication.

(a) Illustration of the measurement sequence to image a single β₂-clamp unloading event. First a phase-contrast (PH) and pre-activation snapshot are taken, after which molecules are activated only once, and subsequently imaged until foci are no longer visible (b,c) Single PAmCherry–β molecules are visualized by PALM. The sample PH image together with the respective pre-activation and post-activation fluorescence images illustrate that a single PAmCherry–β molecule can successfully be photoactivated. The corresponding pre-activation (red) and post-activation (blue) line profile plots of the single DNA-bound PAmcherry–β molecule. Scale bar, 1.6 μm. (d) A representative example of a montage showing the fluorescence intensity of a PAmCherry-β molecule over time and the corresponding intensity trace of the signal. The single-step disappearance is indicative of a single molecule. (e) On-time distribution for individual PAmCherry-β molecules (n=84) fitted with an exponential (red line), and the distribution for the unloading times corrected for photobleaching (dashed green line). The inset shows the distribution of the fitted unloading time constants over the 10⁶ bootstrapped data sets from which the confidence interval for the unloading time is determined.

The effective in vivo loading rate of β₂-clamps

The in vivo loading time of a β₂-clamp during chromosomal replication provides us with insight into how frequently a new clamp, β₂, is needed for processive genome duplication. We utilize both the long time lapse and the single-molecule PALM data to compute the effective loading time in vivo of a new β₂-clamp (). The designation ‘effective’ is added as we do not directly measure the loading of an individual β₂-clamp, but rather the total loading rate of β₂-clamps onto DNA. We have shown in the preceding section that the number of DNA-bound β₂-clamps remains essentially constant () during 2/3 of the replication process. We independently determined the in vivo unloading time via PALM to be t_unload=195 s per β₂ during the replication. In the steady-state regime, the total unloading rate of β₂-clamps is balanced by the effective loading rate of β₂-clamps () onto newly formed primer for Okazaki fragment synthesis:

Using equation (1), we compute the in vivo effective loading time for a β₂-clamp during the replication to be . The reader is referred to Supplementary Note 10 and Supplementary Fig. 10 for a more detailed discussion of equation (1).

Discussion

DNA replication, orchestrated by the multi-protein replisome-complex, is a process essential to cell viability. By using in vivo single-molecule fluorescence microscopy in combination with microfluidics, we were able to investigate the detailed dynamics of an essential component of the replisome, the β₂-clamp, during DNA replication in live E. coli cells. Lagging-strand synthesis is a complex and highly dynamic process, and the sliding clamp is one of the key proteins involved. Each new primer-template junction requires a loaded β₂-clamp to ensure processive replication by DNA Pol III, which is signalled to cycle from one Okazaki fragment to the next⁵² as the replication fork progresses at ~600 bp s⁻¹. Given the average replication fork speed and the typical size of an Okazaki fragment (1–2 kb), one would expect a β₂-clamp to be necessary every ~1.5–3 s. Leading-strand synthesis might be less processive than commonly believed, which would imply that a new β₂-clamp would also need to be loaded on the leading strand. In what follows, however, we have assumed that during normal replication, the exchange of β₂-clamps on the leading strand is a much less frequent occurrence than β₂-clamp exchange for the lagging strand. Until now, it was not demonstrated in vivo whether these loaded β₂-clamps are predominantly recycled (that is, immediately unloaded and reloaded) between successive Okazaki fragments, or whether β₂-clamps remain bound to a completed fragment for a prolonged period of time.

Our results indicate that the number of DNA-bound β₂-sliding clamps increases during the course of the cell cycle, peaking at more than 20 behind an individual fork. Following the initiation of replication, we observe that the number of DNA-bound β₂-clamps gradually increases until a steady-state plateau is reached. This plateau, whose magnitude is such that ~50% of the total β₂-clamps in the cell are DNA-bound, is maintained throughout the remainder of the cell cycle. We determined the number of β₂-clamps in the cell (60–120) during replication, as well as the total number of DNA-bound β₂-clamps (). After termination, β₂-clamps are presumably no longer being loaded, and the fraction of DNA-bound β₂-clamps decays accordingly.

Notably, our measurements for the number of DNA-bound β₂-clamps differ from the value measured previously by some of us in a comprehensive in vivo study of the whole E. coli replisome complex²⁶. In this study, the number of β₂-clamps was estimated to be three for each of the two independent replisomes⁴⁶, for a total of six DNA-bound β₂-clamps present during replication. We note that stoichiometries for most of the other proteins have been duplicated independently, and, therefore, the difference of the number of DNA-bound β₂-clamps appears an isolated case²⁷. Although we cannot fully explain the difference in the number of DNA-bound β₂-clamps, we can nonetheless delineate some possible contributions. The difference may result from the cell physiology due to the immobilization method, lower statistics due to the challenging nature of the ‘slim-field’ experiments at the time, or due to inadvertent changes of the imaging system since measurements spanned across months in the earlier study. It is thus crucial to maintain healthy cell physiology and cell cycle synchronization during experiments, which highlights the utility of microfluidics in live cell single-molecule fluorescence measurements.

The substantial number of DNA-bound β₂-clamps behind each replication fork suggests that β₂-clamps are not rapidly recycled during replication. To corroborate this view, we have utilized PALM to directly measure the in vivo unloading rate of a single β₂-clamp (t_unload=195 s per β₂). Together with the number of DNA-bound β₂-clamps in steady state, this allows us to calculate the effective time of loading (equation (1)) a β₂-clamp during replication as . This result is in good agreement with our previously calculated average estimate of the primer formation time using the Okazaki fragment size range and the typical size of the E. coli genome. Also, this effective loading rate is in accordance with the model that DnaG sets the fork speed²⁹. DnaG is thought to synthesize RNA primers at a rate of approximately one primer every one to two seconds⁵³, which is in good agreement with our calculated in vivo effective loading time. We suggest that an individual β₂-clamp remains on the DNA for a protracted period of time during chromosomal replication, as has been proposed on this basis of in vitro experiments^54,24 and plasmid replication⁵⁵. Our results are in agreement with the behaviour of the sliding clamp for the Gram-positive bacterium B. subtilis³⁴. In this bacterium, the number of DNA-bound β₂-clamps was estimated at ~200 during replication, indicative of clamps being left behind during fork progression. There is a slight possibility that the loading and unloading reaction could be sterically hampered by the fusion protein. However, we have no reason to believe that this is the case since our results are in very good agreement with previous in vitro²⁴ and in vivo work³⁴. Our study shows that rapid recycling of β₂-clamps for subsequent lagging-strand synthesis³³, though observed in in vitro experiments in the absence of excess β₂-clamps in solution, is not the predominant mode in vivo. Although our data does not exclude that β₂-clamps are rapidly recycled at the replication fork, the fact that the loading rate from solution matches the estimated primer formation rate strongly suggests that direct recycling is not the dominant mode of clamp loading.

To illustrate the overall β₂-clamp dynamics during replication, we perform a Monte Carlo simulation (see Methods) that takes our experimentally determined values for , t_unload, and t_rep as input parameters (Fig. 5a). As the approximate rate of clamp removal during termination (Fig. 3e) agrees with the value measured by PALM during steady-state replication (~195 s per β₂), we simply input the latter (likely more accurate) value into the simulations. The simulation starts at t=0 with no DNA-bound β₂-clamps, after which it takes <10 min to reach a stable steady-state number of β₂ bound to DNA (Fig. 5b, left). This value is maintained for ~60 min (Fig. 5b, middle), after which termination occurs and the clamps are unloaded in <5 min (Fig. 5b, right). The number of β₂-clamps in steady state as well as the rise and fall times underline our measured results, and are depicted schematically in Fig. 5c.

Figure 5: Describing the β₂-sliding clamp recycling process during replication.

(a) A Monte Carlo simulation of the β₂-clamp assembly and disassembly reaction. For illustrative purposes, we perform a Monte Carlo simulation of the proposed model, utilizing the experimentally determined primer formation and unloading rate, as well as the replication time, and under the assumption that primer formation is rate limiting. We show the simulation results for five individual traces (coloured lines). The black curve is the analytical solution for the average number of loaded β₂-clamps. Here we divide the total trace into three time regions, namely initiation (red), steady state (green) and termination (blue). (b) A zoom of the different sections from a. (left) A build-up of loaded β₂-clamps on the DNA proceeds for ~10 min. (middle) After the gradual accumulation of loaded β₂-clamps, a steady state plateau of 46 DNA-bound β₂-clamps is maintained for ~2/3 of the replication process. (right) After termination, all DNA-bound β₂-clamps are unloaded in ~5 min. (c) A cartoon illustrating the DNA-bound β₂-clamp build-up during replication. As the rate at which β₂-clamps are loaded (one every 4 s) is much faster than the unloading rate of individual β₂-clamps (once every 195 s) during replication, there will be a dynamic reservoir of β₂-clamps that have not yet been unloaded left on the lagging strand.

The steady-state build-up of DNA-bound β₂-sliding clamps forms a β₂-landing pad³⁴ for different proteins to dock themselves to DNA during the life cycle of the cell. Numerous different proteins utilize the β₂-clamp via the same binding pocket⁵⁶ to perform their respective biological function. These other β₂-clamp-binding proteins range from DNA ligase for Okazaki fragment maturation, inactivation of DnaA through the β₂–Hda₂ interaction^{11,12,13,57,58}, potential screening of DNA damage due to sliding capability of the β₂-clamps⁵⁶, the tethering of the necessary polymerases for repair^4,5,6,7,8, overcoming replication barriers⁵⁹, as well as coupling mismatch-detection and replication by positioning MutS at newly replicated DNA⁶⁰. It is still unclear which of the above-mentioned(or other) proteins are the main users of the DNA-bound clamps that are not directly situated at the replication fork. As Okazaki fragment maturation seems to be relatively fast as assessed via Ligase and Pol I dynamics⁶¹, it is likely not these proteins that predominantly occupy the DNA-bound clamps. The extent to which DNA-bound β₂-clamps are utilized while being docked to the Okazaki fragment will most likely be dependent on the physiological state of the cell at a particular time in its cell cycle. Under stress conditions, MutS and the different repair polymerases might predominantly make use of DNA-bound β₂-clamps, whereas under minimal stress conditions, ligase CrfC and Hda₂ are the likely candidates. A thorough in vivo investigation of the stoichiometry and dynamic of different β₂-associated proteins over the course of the cell cycle would provide the quantitative underpinning required to provide further insight into these biological processes.

Methods

Strains and strain construction

All strains are derivatives of E. coli K12 AB1157. Strains were constructed either by P1 transduction⁶² or by λ-Red recombination⁶³.

The Ypet–dnaN:tetR–mCerulean was constructed using P1 transduction by transducing the YPet–dnaN fusion²⁶ together with the adjacent kanR gene into a strain that contains a tetO array (50 kb clockwise from the dif-site), as well as the chromosomal integrated chimeric gene tetR–mCerulean⁴⁶. The presence of the YPet–dnaN gene fusion was verified using the oligonucleotides: 5′- CGTTGGCACCTACCAGAAAG -3′ and 5′- ATGCCTGCCGTAAGATCGAG -3′. The sequence of the YPet–dnaN fusion was confirmed by DNA sequencing.

A chromosomal fusion of the gene encoding for the photoactivatable fluorescent protein (PAmCherry1)⁵¹ to the N terminus of the dnaN gene was created using λ-Red recombination⁶³. The gene encoding for PAmCherry1 was amplified by PCR. The forward primer used contains an XmaI restriction site (5′- GCGGGCCCCGGGATGGTGAGCAAGGGCGAGGAG -3′). The reverse primer used contains a sequence coding for an 11 amino-acid linker and a SacI site (5′- CGATCGGAGCTCCGCGCTGCCAGAACCAGCGGCGGAGCCTGCCGACTTGTACAGCTCGTCCATGCC -3′). The PCR product was cloned into the backbone of pROD44 (ref. 26) containing a kanamycin-resistance cassette, flanked by frt sites, resulting in the template plasmid PAmCherry1.

This plasmid was then used as a template plasmid for generating the insert sequence used during λ-Red recombination to create the PAmCherry–dnaN strain. The primer sequences used were: forward 5′- ACGATATCAAAGAAGATTTTTCAAATTTAATCAGAACATTGTCATCGTAACTGTAGGCTGGAGCTGCTTC -3′; reverse 5′- ACCTGTTGTAGCGGTTTTAATAAATGCTCACGTTCTACGGTAAATTTCATCGCGCTGCCAGAACCAGCGG -3′. The DNA fragment was gel purified and ~700 ng of the linear DNA was used for electroporation of AB1157 cells overexpressing λ-Red proteins from pKD46 (ref. 63). The correct insertion of the fragment into the chromosome of the resulting strain was assayed by PCR. The oligonucleotides used were 5′- CGTTGGCACCTACCAGAAAG -3′ and 5′- ATGCCTGCCGTAAGATCGAG -3′. The sequence of the fusion gene in this strain was confirmed by DNA sequencing.

Construction of the mCherry–dnaN strain. The mCherry gene was amplified by PCR. The forward primer used for this contains an XmaI restriction site (5′- TAGGCTCCCGGGATGAGCAAGGGCGAGGAGGATAAC -3′). The reverse primer used contains a SacI site and sequence coding for an 11 amino-acid linker (5′- AAGGAGCTCGCGCTGCCAGAACCAGCGGCGGAGCCTGCCGACTTGTACAGCTCGTCCATGCC -3′). The Frt-flanked kanamycin-resistant gene was amplified using the following primers: forward 5′- TTACCCGGGCATATGAATATCCTCCTTAG -3′; reverse 5′- TTAGGATCCTGTAGGCTGGAGCTGCTTCG -3′. The resulting fragment was digested with XmaI and BamHI. The mCherry fragment and the kanamycin fragment were cloned into pUC18 between SacI and BamHI sites.

The λ-Red recombination was performed as mentioned in the previous section using the primers: forward 5′- TATCAAAGAAGATTTTTCAAATTTAATCAGAACATTGTCATCGTAAACCTGTAGGCTGGAGCTGCTTCG -3′; reverse 5′- ACCTGTTGTAGCGGTTTTAATAAATGCTCACGTTCTACGGTAAATTTCATCGCGCTGCCAGAACCAGC -3′. The presence of the gene fusion was verified using oligonucleotides 5′- CGTTGGCACCTACCAGAAAG -3′ and 5′- ATGCCTGCCGTAAGATCGAG -3′. The sequence of the fusion gene in this strain was confirmed by DNA sequencing.

The dnaX(τ)–YPet:mCherry–dnaN strain was constructed using P1 transduction by transducing the dnaX–YPet fusion²⁶ together with the adjacent kanR gene into a strain that contains the mCherry–dnaN gene fusion. The presence of the dnaX–YPet gene fusion after transduction was verified using the oligonucleotides: 5′- GAGCCTGCCAATGAGTTATC -3′ and 5′- GGCTTGCTTCATCAGGTTAC -3′ and similarly the mCherry–dnaN fusion using 5′- CGTTGGCACCTACCAGAAAG -3′ and 5′- ATGCCTGCCGTAAGATCGAG -3′. The sequences of the fusions in this strain were confirmed by DNA sequencing.

Supplementary Tables 2 and 3 provide an overview of the plasmids used, as well as a summary of the different strains. The cell morphology and the doubling times of the fusion strains in LB and M9-glycerol growth medium were compared with AB1157 wild type. No significant differences were observed (Supplementary Table 1 and Supplementary Fig. 1a). The doubling times of the cells in the microfluidic device were similar (slightly faster) compared with cells grown in a shake flask (Supplementary Fig. 2). We also confirmed that in the absence of IPTG (the experimental condition used during long time-lapse microscopy), no DNA-bound foci were detected for the YPet–dnaN:tetR–mCerulean strain (Supplementary Fig. 1b).

M9 growth medium used in experiments

The M9 growth medium used in experiments is as follows. One litre of M9 growth medium contains 10.5 g l⁻¹ of autoclaved M9 broth (Sigma-Aldrich); 0.1 mM of autoclaved CaCl₂ (Sigma-Aldrich); 0.1 mM of autoclaved MgSO₄ (J.T.Baker); 0.3% of filter-sterilized glycerol (Sigma-Aldrich) as carbon source; 0.1 g l⁻¹ of filter-sterilized five amino acids, namely L-threonine, L-leucine, L-proline, L-histidine and L-arginine (all from Sigma-Aldrich) and 10 μl of 0.5% filter-sterilized thiamine (Sigma-Aldrich).

Microfluidics for extended time-lapse microscopy

We use our own design⁴³ of the previously published microfluidic device known as the mother machine⁴⁴ for cell immobilization during long time-lapse experiments. The reader is referred to Moolman et al.⁴³ for a detailed description of the complete fabrication process. Here, we only briefly outline the main steps involved. First, we use electron-beam lithography in combination with dry etching techniques to create the structure in silicon. Next, we make a negative mould of this structure in polydimethylsiloxane (PDMS). The PDMS mould is then used to fabricate the positive structure in PDMS, which is subsequently used for experiments.

Preparation of cells for microscopy

Cells were streaked on Luria-Bertani (LB) plates containing the appropriate antibiotic. Single colonies from these plates where inoculated overnight at 37 °C with shaking in M9 medium supplemented with 0.3% glycerol (Gly), essential nutrients together with the appropriate antibiotics. The subsequent day, the overnight culture was subcultured into the same medium and grown at 37 °C with shaking until an OD₆₀₀~0.2 was reached. Cells were concentrated by centrifugation for 2 min at 16,100 g. The subsequent steps are dependent on the type of microscopy experiment performed as outlined next.

For agarose pad experiments, the supernatant was decanted and the pellet was resuspended in 100 μl M9-Gly supplemented with essential nutrients. The resuspended cells were subsequently vortexed for 2 s and immobilized on an M9-Gly 1.5% agarose pad between two coverslips. (The coverslips were ultrasonically cleaned in acetone and isopropyl alcohol and burned by a flame to minimize the fluorescent background before use).

For microfluidic device experiments, the supernatant was decanted and the pellet was resuspended in 50 μl M9-Gly with essential nutrients and injected into the microfluidic device. After injection into the device, the device was centrifuged for 10 min at 2,500 g (Eppendorf 5810R) so as to load the cells into the growth channels. Following centrifugation, the device was mounted on the microscope with tubing attached and incubated for ~45 min at 37 °C. After incubation, fresh M9-Gly with essential nutrients and the appropriate antibiotics are flushed through the device. The syringe containing the medium is then attached to an automated syringe pump to continuously infuse fresh M9-Gly, essential nutrients and 0.2 mg ml⁻¹ bovine serum albumin (BSA) through the device at a rate of 0.5 ml h⁻¹.

Microscope setup

All the images were acquired on a commercial Nikon Ti microscope equipped with a Nikon CFI Apo TIRF × 100, 1.49NA oil immersion objective and an Andor iXon 897 Electron Multiplying Charge Coupled Device (EMCCD) camera operated by a personal computer (PC) running Nikon NIS-elements software. Cell outlines were imaged using the standard Nikon brightfield halogen lamp and condenser components. The fluorescence excitation was performed using custom-built laser illumination. A Cobolt Fandango 515 nm continuous wave diode-pumped solid-state laser was used to excite YPet; Cobolt Jive 561 nm continuous wave diode-pumped solid-state laser was used to excite mCherry and PAmCherry, respectively. PAmCherry was activated by a Votran Stradus 405 nm. All the three laser beams were combined using dichroic mirrors (Chroma ZT405sp-xxr, 575dcspxr) and subsequently coupled into a single-mode optical fibre (KineFLEX). The output of the fibre was expanded and focused onto the back focal plane of the objective mounted on the microscope. Notch filters (Semrock NF03-405E, NF03-514E, NF03-561E) were used to eliminate any laser light leaking onto the camera. The emission of the different fluorescent proteins was projected onto the central part of the EMCCD camera using custom filter sets: Chroma z561, ET605/52m, zt561rdc (mCherry), Chroma z514, ET540/30m, zt514rd (YPet), Chroma zet405, ET480/40m, zt405rdc (CFP). A custom design commercial temperature control housing (Okolabs) enclosing the microscope body maintained the temperature at 37 °C. Sample position was controlled with a Nikon stage (TI-S-ER Motorized Stage Encoded, MEC56100) together with the Nikon Perfect Focus System to eliminate Z-drift during image acquisition.

Cell lysate preparation for intensity calibration

The cell lysate used for single-molecule intensity calibration was prepared as follows. Cells were grown overnight at 37 °C with shaking in M9 medium supplemented with 0.3% glycerol (Gly), essential nutrients together with the appropriate antibiotics. The subsequent day, the overnight culture was subcultured into the same medium and grown at 37 °C with shaking until an OD₆₀₀~0.5 was reached. The cells were collected by centrifugation at 6,000 g (Beckman JLA 9.1000 rotor) for 15 min. Cells were subsequently resuspended in 5 ml M9-Gly and essential nutrients. The cell suspension was French pressed (Constant Systems) twice at 20,000 p.s.i. The cell lysate was then spun down at 30,000 g (Beckman JA-17 rotor ) for 35 min. The supernatant was shock-frozen using liquid nitrogen and kept at 37 °C until needed.

Data acquisition

All data acquisition was performed on the same microscope setup. Image acquisition was performed with Nikon NIS-elements software. The acquisition protocol was dependent on the type of experiment performed as outlined next.

Long time-lapse experiments were conducted as follows. The cell outlines were imaged using standard brightfield illumination. Subsequently, the sample was excited by laser excitation (515 nm) with an intensity of ~5 W cm⁻² as calculated according to Grünwald et al.⁶⁴ The exposure time was set to 80 ms. The camera gain was set to 100. Brightfield and fluorescence images were acquired every 2.5 min. Data spanning ~10 h of measurement were acquired overnight.

We conducted two types of PALM experiments. First, we determined the bleaching characteristic of PAmCherry under our experimental conditions, and second, we measured the unloading time of a single β₂-clamp. PALM images were acquired as follows. First, the cell outlines were imaged by taking a single phase-contrast (PH) image using a commercial Nikon external phase ring configuration. The sample was then excited for a single frame (400 ms exposure time) by a 561 nm laser with an intensity of ~5 W cm⁻², calculated according to Grünwald et al.⁶⁴ This image was used to determine the auto-fluorescence level due to the sample before activation. Photoactivation of PAmCherry was done with a single pulse (5 s) of 405 nm with an intensity of ~2.5 W cm⁻², calculated according to Grünwald et al.⁶⁴ Subsequently, a post-activation time-lapse of images were acquired using the 561 nm laser at the same intensity at a frame rate of either ~700 ms (bleaching experiments) or 5 s (unloading experiments) with an exposure time of 400 ms per frame. Camera gain was set to 100.

Image analysis of long time-lapse experiments

Images were analysed with custom-written MATLAB software (MathWorks). Before any analysis, we subtract the uneven background using a rolling-ball filter⁶⁵ and subsequently corrected for illumination heterogeneity by using the previously measured laser beam profile⁶⁶. We also align the brightfield and fluorescence signals with respect to each other with 1-pixel accuracy. X–Y drift is corrected in both the fluorescence and brightfield images by tracking a fiducial marker in the PDMS to within 1 pixel.

Each drift-corrected region of interest, consisting out of a single growth channel, is analysed individually. The brightfield images are used to determine the cell poles of all the cells in a given frame. For the fluorescence signal, a kymograph of the fluorescence signal is constructed by summation of the pixel intensities per image perpendicular to the channel direction for each frame. This results in summed intensity information as a function of time per growth channel (Fig. 2c). We make use of the generated kymographs to determine individual replication and division cycles per cell (Fig. 2d). A post-processing step is subsequently performed to eliminate cells that did not match the following selection criteria: correct cell length, sufficient growth characteristics, observation of a complete cell cycle, clear fluorescence signal that both starts and ends in a diffuse state (Fig. 2d).

The fluorescence images of the detected individual cells that pass the above selection criteria are analysed further. We base our fluorescence analysis on an image of an individual bacterium with its long axis aligned with the horizontal direction of the image. The width of the image is equal to the length of the bacterium. We fix the height of the image such that a sufficient area above and below the bacterium is included that is indicative of the auto-fluorescence of the sample. We analyse the fluorescent intensity counts of a single bacterium using the individual fluorescence kymographs of each cell (summed line-profiles) by calculating three types of image content for a specific bacterium, namely ‘background’, ‘foci’ and ‘cytoplasm’ (Supplementary Fig. 3a). In brief, we first estimate the background fluorescence from the sample using the signal outside the bacterium. We did not have to take into account auto-fluorescence from the bacterium itself, as we conducted our experiments using minimal medium, which results in negligible levels of cellular auto-fluorescence (Supplementary Fig. 3b). The intensity outside the bacterium is used for a threshold with the remaining pixels intensities being representative of the total bacterium fluorescent counts. We subsequently separate ‘cytoplasm’ and ‘foci’ signals by determining the median of the summed line profiles. The signal significantly above this value is attributed to foci, whereas the remainder (lower values) are treated as the fluorescence signal from the cytoplasm (Supplementary Fig. 3a). This results in an integrated intensity value for the foci and also for the cytoplasm.

Image analysis of PALM experiments

PALM data was analysed using custom-written MATLAB software (MathWorks) in combination with the freely available MicrobeTracker software⁶⁷. Before any spot analysis, the fluorescence images are subjected to illumination correction and to alignment with respect to the phase-contrast (PH) images. The resulting corrected and aligned fluorescence images are then used during further analysis.

Using the PH image, the different cells are detected in the field of view and their respective outlines are determined using MicrobeTracker. Subsequently, using the spot detection algorithm as described in Olivo-Marin⁶⁸, the spots in each individual image of the fluorescence time-lapse series are detected, and the integrated intensity is determined by summing the pixel values of each spot⁶⁹. The integrated intensities of the spots are followed as function of time. This results in individual time-lapse integrated intensity traces of single molecules (Fig. 3c). The cell outlines as determined previously are overlaid with the fluorescence images. Any foci that are not situated in a bacterial cell (false positives) are rejected from further analysis. Only cells that had a clear fluorescence intensity focus were analysed. This focus is indicative of DNA-bound clamps and thus DNA replication. Foci that exhibit multiple steps in fluorescence intensity are also rejected. For the remainder of the foci, the time it takes from the start of the data acquisition until spot disappearance is recorded (Fig. 3c). These calculated time differences are indicative of molecule unloading (or bleaching, depending on the time of acquisition) and analysed further as described in the following section.

Monte Carlo simulation of β₂-loading and unloading dynamics

For illustrative purposes, we perform Monte Carlo simulations (Fig. 5) starting with no clamps loaded and no primers formed (), and assuming the loading rate to be much faster than the rate of primer formation (N_p≈0). In each small time-step dt, we let with probability dt and with probability dt . This is repeated until the replication time is reached, upon which the primer formation rate is set to zero.