Modulation of DNA polymerase IV activity by UmuD and RecA* observed by single-molecule time-lapse microscopy

DNA polymerase IV (pol IV) is expressed at increased levels in Escherichia coli cells suffering high levels of DNA damage. In a recent single-molecule imaging study, we demonstrated that elevating the pol IV concentration is not sufficient to provide access to binding sites on the nucleoid, suggesting that other factors may recruit pol IV to its substrates once the DNA becomes damaged. Here we extend this work, investigating the proteins UmuD and RecA as potential modulators of pol IV activity. UmuD promotes long-lived association of pol IV with the nucleoid, whereas its cleaved form, UmuD’, which accumulates in DNA-damaged cells, inhibits binding. In agreement with proposed roles for pol IV in homologous recombination, up to 40% of pol IV foci colocalise with a probe for RecA* nucleoprotein filaments in ciprofloxacin-treated cells. A hyperactive RecA mutant, recA(E38K), allows pol IV to bind the nucleoid even in the absence of exogenous DNA damage. In vitro, RecA(E38K) forms RecA*-like structures that can recruit pol IV, even on double-stranded DNA, consistent with a physical interaction between RecA and pol IV. Together, the results indicate that UmuD and RecA modulate the binding of pol IV to its DNA substrates, which frequently coincide with RecA* structures.


37
DNA polymerase IV (pol IV), encoded by dinB, is one of three specialised DNA 38 polymerases that are produced at increased levels in Escherichia coli cells suffering DNA 39 damage (Napolitano et al., 2000). In vitro, DNA polymerase IV is capable of translesion known as recA730), which constitutively produces RecA*-like activity (Cazaux et al., 1993; 113 Wang et al., 1993;Ennis et al., 1995), promotes the binding activity of pol IV to the nucleoid, 114 even in the absence of DNA damage. We further showed that pol IV physically interacts with 115 RecA(E38K), which forms RecA*-like structures on single-stranded as well as double-stranded 116 DNA, suggesting that pol IV might also associate with these RecA*-like structures in cells.  123 In a previous study, we carried out time-lapse measurements on E. coli cells treated with 124 DNA damaging agents (Henrikus et al., 2018b). We found that the colocalisation of pol IV foci 125 with replisome markers started at ~10% prior to treatment, and dropped to < 5% (i.e. baseline 126 levels) at a time-point 90-100 after the onset of treatment. In a separate study, we observed that 127 pol V (UmuC-mKate2) enters the cytosol and forms foci on the nucleoid at this same 90 min 128 time-point (Robinson et al., 2015). This spatial re-distribution of the UmuC-mKate2 marker  were determined at 0, 30, 60, 90 and 120 min time points (Fig 1). Colocalisation between DinB-151 YPet foci and -mKate2 foci was also monitored. In order to enhance diffusional contrast in our 152 images we used longer exposure times when capturing DinB-YPet signal (300 ms) than in our 153 previous study (50 ms; (Henrikus et al., 2018b). We nonetheless recorded a complementary set 154 of colocalisation measurements with the shorter exposure time of 50 ms in order to better capture 155 transient foci and allow for more direct comparison with our previous results (Fig S1). 156 We first monitored pol IV behaviour in cells expressing wild-type levels of UmuD and 157 UmuC (EAW643 , Table 1). Cells exhibited very few pol IV foci prior to ciprofloxacin treatment markedly between the 90 min and 120 min time-points (Fig 1A, middle panel). From 0-90 min 163 after ciprofloxacin addition, 5% of pol IV foci colocalised with the replisome marker . From 164 120-150 min this decreased to < 2%. These values are somewhat lower than those we reported 165 previously (10% dropping to < 5%) and is attributable to the longer image exposure times used 166 in the current study (Fig S1). The percentage of  foci that contained a pol IV focus followed a 167 similar trend (Fig 1A, lower panel); from 0-90 min after ciprofloxacin addition, 0.5% of  foci 168 contained a pol IV focus, dropping to ~0.1% (indistinguishable from chance colocalisation) from 169 120-150 min. 170 We next examined the effect of deleting the umuDC operon (and thus eliminating UmuD 171 and UmuC) on the number of pol IV foci and the extent of colocalisation with  foci (SSH007, 172 Table 1). From 30 min, 10-15% of pol IV foci colocalised with replisomes. Compared to 173 umuDC + cells, ΔumuDC cells exhibited a three-fold increase in the number pol IV foci per cell 174 with ~0.3 foci per cell from 60 min after ciprofloxacin addition (Fig 1B, upper panel). Moreover, 175 deletion of umuDC led to a three-fold increase in the percentage of pol IV foci that colocalise 176 with a  focus (Fig 1B, middle panel). Interestingly, pol IV- colocalisation now persisted above 177 10% for the 90, 120 and 150 min time points. The percentage of  foci that contained a pol IV 178 focus was also elevated in the ΔumuDC background (Fig 1B,  The increased numbers of pol IV foci and increased pol IV- colocalisation in ΔumuDC 188 than in umuDC + cells could manifest through two scenarios: 1. the deletion of the umuDC 189 operon, which encodes for pol V, eliminates competition between pol IV and pol V for binding 190 sites on the nucleoid. 2. a subunit of pol V has a regulatory effect on pol IV focus formation and 191 pol IV- colocalisation. It has been shown previously that UmuD 2 and UmuD′ 2 physically 192 interact with pol IV and modulate its mutagenic activity (Godoy et al., 2007). To that end, we 193 tested if UmuD or UmuD′ affect the extent of pol IV focus formation and the colocalisation 194 between pol IV with , in the absence of UmuC (and thus pol V). 195 We constructed two strains, both of which include the dinB-YPet and dnaX-mKate2 196 alleles: i) umuDC (SSH007) expressing the non-cleavable UmuD(K97A) protein from a low-197 copy plasmid (SSH007 + pJM1243), and ii) SSH007 expressing the 'cleaved' UmuD′ protein 198 from a low-copy plasmid (SSH007 + pRW66). The amount of UmuD(K97A) and UmuD′ 199 produced from each plasmid is 4-5-fold higher than UmuD expressed from its native 200 chromosomal locus (Churchward et al., 1984). Time-lapse analysis was repeated as described 201 above. 202 We first explored the effects of expressing the non-cleavable UmuD(K97A) mutant in 203 dinB-YPet dnaX-mKate2 umuDC cells (SSH007 + pJM1243, Table 1). At the 90 min time 204 point, cells contained on average 0.6 pol IV foci per cell -a six-fold increase over umuDC + 205 cells (Fig 1C, upper panel). This sixfold increase in pol IV foci per cell was accompanied by a 206 three-fold increase in colocalisation with the replisome marker -mKate2 (Fig 1C, middle panel).

207
From 30 min after damage induction, 13% of pol IV foci overlapped with a  focus. This 208 colocalisation remained relatively constant during the later stages of the SOS response; 209 colocalisation did not drop below 9% from 90-120 min as observed in umuDC + cells. These 210 observations reveal that UmuD(K97A), and by inference uncleaved UmuD, promote the binding 211 of pol IV to DNA and do not limit pol IV- colocalisation beyond 90 min.

212
During the later stages of the SOS response (90 min after SOS induction), UmuD is 213 cleaved to UmuD′ (Woodgate et al., 1989). To explore the effects of UmuD′ on pol IV 214 behaviour, we imaged umuDC cells expressing UmuD′ directly from a plasmid (SSH007 + 215 pRW66, Table 1). These cells produced ~0.1 DinB-YPet foci per cell at 60 min (Fig 1D, upper   216 panel), similar to umuDC + cells. In the cells expressing UmuD′, colocalisation of pol IV with  217 was generally low, but highly variable (Fig 1D, middle panel). Two large spikes in colocalisation 218 were apparent at the 30 and 90 min time points. However, due to the low number of foci 219 available for analysis at these time-points, there was very large error associated with these 220 values. No spikes in colocalisation were observed when measuring the proportion of  foci that 221 contained a pol IV focus (Fig 1D, lower panel). Importantly, the colocalisation of pol IV with  222 decreased to < 1% after 90 min (Fig 1D, middle panel). Similarly, the percentage of  foci that 223 contained a pol IV focus drops between the 90 and 120 min time points (Fig 1D, lower panel).

224
From 30-90 min, ~1% of  foci contained a pol IV focus. By 120 min < 0.1% of  foci contained 225 a pol IV focus. Overall, the introduction of UmuD′ into umuDC cells restores rates of focus 226 formation and colocalisation with the replisome marker  to near wild-type (umuDC + ) levels.   (Fig 2A), 240 ΔumuDC ( Fig 2B) and UmuD′-expressing cells (Fig 2D), most foci were relatively faint and 241 diffuse. In contrast, cells expressing UmuD(K97A) produced brighter, and much more distinct, 242 pol IV foci. Reasoning that these differences might reflect differences in the nature of pol IV 243 interactions with the substrates, we next measured the binding lifetime of pol IV at these sites.  For the umuDC + , ΔumuDC and UmuD′-expressing cells, intensity trajectories collected at 251 the positions of  foci predominantly exhibited short-lived binding events (Fig 2, second row).

252
Cells expressing UmuD(K97A), on the other hand, often produced long-lived pol IV binding 253 events. To comprehensively assess the binding lifetimes of pol IV with respect to the UmuD 254 status at sites of the replisome marker, mean autocorrelation functions were calculated for foci 255 within each strain (Fig 2, third row; Fig S2). This approach allows us to extract characteristic  For both umuDC + and umuDC cells, most pol IV binding at  positions appeared to be 262 short-lived (Figs 2A, B). In the early stages of ciprofloxacin exposure (25-45 min) the 263 components of the autocorrelation function were 80% short-lived (< 0.03 s, shorter than a frame 264 of 34 ms), 10% medium (0.3 s) and 10% long-lived (3.3 s). In the later stages, (120-150 min), 265 the proportion of medium-long lived events increased to 40%. In cells expressing UmuD(K97A) 266 long-lived events were much more common: by the 120-150 min period medium and long-lived 267 events comprised 80% of the autocorrelation function (Fig 2C). In stark contrast, cells 268 expressing UmuD′ produced almost exclusively short-lived events (Fig 2D). UmuD′ appeared to 269 supress the medium and long-lived pol IV binding events that occur in wild-type umuDC + 270 background following ciprofloxacin treatment. field-of-view was recorded.

296
Following ciprofloxacin treatment, cells typically contained multiple mCI foci (Fig 3A). 297 At later time points, some cells contained more elongated "bundle" structures as described 298 previously (Ghodke et al., 2019). We next determined the percentage of DinB-YPet foci that 299 colocalised with mCI foci and bundle-like structures (Fig 3B). Prior to the introduction of   366 We therefore directly tested whether RecA(E38K) interacts with pol IV on filaments 367 assembled dsDNA in vitro. Using an identical SPR experimental setup as described above, we 368 assembled RecA(E38K) on a 60-mer dsDNA substrate (Figs S3C, D). We found that pol IV 369 associates with RecA(E38K)-ATPγS filaments formed on dsDNA (Fig S3E), producing a much 370 stronger response than measurements in which pol IV was exposed to dsDNA in the absence of 371 RecA(E38K) (Fig S3F). Unfortunately, despite our attempts to further optimise the assay, non-372 specific binding of pol IV to the chip surface hampered our attempts to extract binding 373 parameters from the sensorgrams. Nevertheless, these results clearly demonstrate that the 374 association of pol IV with the nucleoid is promoted by the presence of RecA*-like structures.

375
Returning to the live-cell single-molecule data, we next examined fluctuations in the 376 DinB-YPet signals that occur as pol IV binds to, or dissociates from, binding sites on the 377 nucleoid. We monitored pol IV binding events within cells, both close to and away from  foci.  and  is highest in cells that lack UmuD and UmuC altogether (ΔumuDC). One possibility is that 448 in wild-type umuDC + cells pol V competes with pol IV for binding to replisome-proximal 449 binding sites, however this explanation seems unlikely for two reasons: 1. pol IV- 450 colocalization is low in cells that express UmuD′, but lack UmuC and therefore cannot produce 451 pol V (Fig 1D); 2. fluorescently labelled pol V colocalises with replisomes even less frequently 452 than pol IV does (Robinson et al., 2015). Another explanation, which is more consistent with the 453 data, is that the accumulation of UmuD′ in response to treatment with DNA damaging agents 454 inhibits the binding of pol IV at replisome-proximal sites in wild-type (umuDC + ) cells.   To investigate the influence of UmuD mutants on pol IV activity, SSH007 was 492 complemented with plasmids that express UmuD(K97A) (pJM1243) or UmuDʹ (pRW66).      The wild-type E. coli RecA protein was purified as described (Craig and Roberts, 1981).

588
The E. coli RecA(E38K) protein was purified as previously described (Cox et al., 2003) 589 with the following modifications. After washing the protein pellet with R buffer plus 2.1 M 590 ammonium sulfate, the pellet was resuspended in R buffer plus 1 M ammonium sulfate. The monitoring pol IV association (Fig S3E). From 220 s, buffer containing 0.5 mM ATPγS was 648 31 flowed in at 5 μL min -1 and fast dissociation of pol IV was observed. Similarly, pol IV 649 association with dsDNA was monitored, giving a lower response curve (Fig S3F). We also 650 observed non-specific binding of pol IV to the chip surface, making it impossible to measure 651 binding kinetics of pol IV.