Single-molecule live-cell imaging visualizes parallel pathways of prokaryotic nucleotide excision repair

In the model organism Escherichia coli, helix distorting lesions are recognized by the UvrAB damage surveillance complex in the global genomic nucleotide excision repair pathway (GGR). Alternately, during transcription-coupled repair (TCR), UvrA is recruited to Mfd at sites of RNA polymerases stalled or paused by lesions. Ultimately, damage recognition is mediated by UvrA, culminating in the loading of the damage verification enzyme UvrB. We set out to characterize the differences in the kinetics of interactions of UvrA with Mfd and UvrB. We followed functional, fluorescently tagged UvrA molecules in live cells and measured their residence times in TCR-deficient or wild-type cells. We demonstrate that the lifetimes of UvrA in Mfd-dependent or Mfd-independent interactions in the absence of exogenous DNA damage are comparable in live cells, and are governed by UvrB. Upon UV irradiation, we found that the lifetimes of UvrA strongly depended on, and matched those of Mfd. Here, we illustrate a non-perturbative, imaging-based approach to quantify the kinetic signatures of damage recognition enzymes participating in multiple pathways in cells.


Introduction 24
Across the various domains of life, the recognition and repair of bulky, helix-distorting lesions in 25 chromosomal DNA is coordinated by nucleotide excision repair (NER) factors. Damage detection 26 occurs in two stages: a dedicated set of damage surveillance enzymes (reviewed in ref. 1,2,3 ) (namely 27 the prokaryotic UvrA, and the eukaryotic UV-DDB, XPC, XPA and homologs) constantly survey genomic 28 DNA for lesions. At sites of putative DNA damage, these enzymes load specific factors (UvrB in 29 prokaryotes, TFIIH and homologs in eukaryotes) that unwind the DNA and verify the location of the 30 damage with nucleotide resolution (Fig. 1a) (reviewed in ref. 2,3 ). Subsequently, specialized 31 endonucleases (prokaryotic UvrC and homologs, and the eukaryotic XPF-ERCC1/XPG and homologs) 32 are recruited to the site of the DNA, resulting in cleavage of the single-stranded DNA (ssDNA) patch 33 containing the lesion (reviewed in ref. 2,3 ). 34 In several studied organisms (barring certain archaea 4, 5 ), removal of DNA damage also occurs via the 35 triggering of strand-specific DNA repair following the stalling of RNA polymerase (RNAP) at sites of 36 lesions (reviewed in ref. 6 ). In this case, a ternary elongation complex (TEC) of RNAP that is unable to 37 catalyse nascent RNA synthesis manifests as an ultra-stable protein-DNA roadblock.  repair coupling factors such as the prokaryotic Mfd, and the eukaryotic homologs Rad26/CSB are 39 dedicated factors that recognize these TECs and remodel them 7,8,9,10,11 . In prokaryotes, Mfd is 40 recruited to the site of a failed TEC, and in turn it recruits the UvrA2B complex via interactions between 41 its UvrB-homology module (BHM) and UvrA. (Fig. 1) 9,11,12,13 . Similarly, in eukaryotes, CSB is recruited 42 to the site of a stalled RNAPII complex, and recruits the TFIIH complex 14 . 43 DNA repair triggered by stalling of TECs is termed transcription-coupled repair (TCR), in contrast to the 44 direct detection of lesions by the UvrAB damage sensor (global genomic repair, or GGR). Studies 45 investigating the rate of repair during TCR vs. GGR, have reported an enhancement in the rate of 46 removal of UV-induced lesions from the template strand in transcribed DNA compared to non-47 transcribed DNA 15,16,17,18,19,20 . This observation has sparked several studies targeted at understanding 48 the mechanistic basis of rate enhancement 13,21,22 . A recent single-molecule in vitro study reported 49 that the time to incision in TCR is approximately three-fold faster than in GGR under certain conditions 50 13 . 51 A diverse set of intermediates is readily formed in vitroranging from a translocating RNAP-Mfd 52 complex, arrested RNAP-Mfd-UvrA2 and the complete Mfd-UvrA2-UvrB handoff complex in the 53 presence of both UvrA and UvrB 12,13 . To understand which of these intermediates are formed inside 54 cells, we have recently visualized Mfd in cells and quantified its lifetime in the TCR reaction 12 . A recent 55 study failed to detect an influence of Mfd on the behaviour of fluorescently tagged UvrA in living 56 cells 23 . Therefore, in vitro studies notwithstanding, how TCR is orchestrated by UvrA in cells remains 57 unclear. 58 In this work, we revisited this question in the context of live cells and applied high-resolution single-59 molecule imaging methods that permit accurate measurements of DNA binding lifetimes over a broad 60 timescale ranging from a few hundred milliseconds to several minutes 12,24  parallel pathways in vivo using non-perturbative, single-molecule imaging approaches. 74

75
The nucleotide excision repair reaction can be considered to occur in five discrete stages: damage 76 recognition, damage verification, incision, repair synthesis and ligation (reviewed in ref 2 To visualize the binding of UvrA to DNA in cells, we replaced the uvrA gene with a C-terminal fusion of 85 uvrA to the gene for the yellow fluorescent protein (YPet 27 ) in MG1655 cells using l Red recombination 86 ( Fig. 2a) 12,28 . This strategy enabled observation of fluorescent UvrA-YPet expressed from the native, 87 SOS-inducible uvrA promoter (Supplementary Movie 1). We first performed UV-survival assays to 88 assess the ability of UvrA-YPet to execute nucleotide excision repair (NER). Compared to wild-type 89 cells, uvrA-YPet cells exhibited somewhat poorer survival upon exposure to UV (Supplementary Fig. 90 1a). Considering that C-terminal fusions of UvrA are fully functional in NER 23, 29 , this modestly lower 91 survival of uvrA-YPet cells may be attributable to a lower efficiency of protein translation. 92 Therefore, we set out to measure the copy numbers of UvrA-YPet in uvrA-YPet cells grown in EZ-rich 93 defined media supplemented with glucose at 30 °C. Exponentially growing cells were deposited on a 94 modified glass coverslip at the bottom of a flow cell and visualized by illumination with 514-nm laser 95 light under continuous flow of growth medium (Fig. 2b). Images of uvrA-YPet cells revealed DNA-96 bound UvrA-YPet molecules that manifested as static foci and diffusive molecules contributing to 97 cellular background fluorescence (Fig. 2c). These observations are consistent with its role as a damage 98 surveillance protein. 99 Exposure to laser light led to rapid loss of YPet signal due to photodarkening and photobleaching of 100 the chromophore (Supplementary Movie 1). We used this loss of signal to measure copy numbers of 101 UvrA-YPet in cells. Dividing the background-corrected cellular fluorescence intensity by the intensity 102 of a single YPet molecule revealed a copy number of 16 ± 4 copies of UvrA-YPet per cell 103 . Copy numbers of UvrA are strongly influenced by the carbon source 104 present in the growth medium, ranging from 9 -43 copies (minimal media) to 129 copies (rich media) 105 per cell 30 . These estimates exceed the copy numbers of UvrA-YPet detected in the uvrA-YPet strain 106 grown in rich medium. The lower copy numbers of UvrA-YPet are consistent with the minor 107 deficiencies in survival observed after UV exposure ( Supplementary Fig. 1a). 108 To confirm the hypothesis that the deficiencies in UV survival observed for the uvrA-YPet strain are 109 attributable to lower copy numbers as opposed to a catalytically deficient protein, we created a low 110 copy plasmid (pSC101 origin of replication, 3-4 copies/cell 31 ) expressing the C-terminal YPet fusion of 111 UvrA under its native promoter (pUvrA-YPet). We then transformed DuvrA cells with pUvrA-YPet. 112 These cells express UvrA-YPet to the extent of 120 ± 28 copies per cell ( Supplementary Fig. 1b-d). UV 113 survival assays ( Supplementary Fig. 1e) revealed that UvrA-YPet expressed from the plasmid is able to 114 fully complement the DuvrA phenotype. Therefore, we conclude that the copy numbers of UvrA-YPet 115 expressed from the endogenous uvrA promoter represent lower copy numbers compared to untagged 116 UvrA expressed in wild-type cells, likely reflecting a poorer efficiency of translation of the uvrA-YPet 117 gene. 118 Interval imaging strategy to measure DNA binding kinetics 119 Continuous imaging of UvrA-YPet could not be used to measure DNA binding lifetimes, since the 120 apparent lifetime of a focus represents UvrA-YPet molecules dissociating from the site or bound  YPet molecules that are photobleached during the imaging. Consequently, measurement of 122 interactions that last longer than the photobleaching lifetime is impossible. Instead we imaged UvrA-123 YPet using a strategy of performing time-lapse imaging with dark periods of varying intervals (for 124 brevity, we term this mode of imaging 'interval imaging') 12, 24, 32, 33 (Fig. 2d) that elegantly deconvolutes 125 the lifetime of the interaction of UvrA-YPet with DNA and the lifetime of the fluorescent probe. Briefly, 126 the introduction of a dark interval (td) between consecutive frames of duration (tint) extends the 127 observation time window. Here the time between consecutive frames is denoted as time-lapse time 128 (ttl). By acquiring the same number of frames in each video collected with a different dark interval, 129 the photobleaching rate (kb) is maintained constant (see Table 1 and Methods) while the observation 130 window is extended arbitrarily. From these videos, cumulative residence time distributions of DNA-131 bound UvrA-YPet are constructed. Since these distributions reflect a mixture of two populations (UvrA 132 molecules that dissociate and YPet molecules that photobleached), fitting them to an exponential 133 function yields an effective off rate (keff) 32 . The product (keffttl) is a linear function of true off rate (koffttl) 134 and the normalized photobleaching rate (kbtint) 32 . For purposes of illustration, the data can be 135 represented as keffttl vs. ttl plots where the slope reveals the off rate koff and the Y-intercept reveals kb. 136 This interval imaging strategy enables accurate quantification of binding lifetimes over three orders of 137 magnitude from 0.1 s to several minutes 12, 32, 34 . 138 UvrA is long-lived on DNA in the absence of UvrB and Mfd 139 First, we interrogated UvrA binding kinetics in the absence of its two major interacting partners UvrB 140 and Mfd in growing cells. To that end, we transformed cells lacking UvrA, UvrB and Mfd (DuvrA DuvrB 141 Dmfd cells) with pUvrA-YPet. In these cells, we expected that interactions of UvrA-YPet with 142 chromosomal DNA would reflect two of its key activities: binding to non-damaged DNA and binding 143 to endogenous DNA damage produced as a by-product of cellular metabolism (Fig. 3a). Indeed, 144 measurements of UvrA-YPet kinetics of dissociation in these cells revealed two lifetimes that are an 145 order of magnitude apart -a fast lifetime (tUvrA|DuvrA DuvrB Dmfd, fast) of 1.6 ± 0.1 s (72 ± 2 %) and a slow 146 lifetime (tUvrA|DuvrA DuvrB Dmfd, slow) of 24 ± 1 s (28 ± 2 %) (summarized in Fig. 3c; Supplementary Figure 2a Table 1; error bars represent standard deviation of the bootstrap distribution of values obtained 148 by performing global fits to CRTDs ten times). To eliminate the possibility that this measured lifetime 149 is affected by cellular copy numbers of UvrA, we additionally created a strain that expresses uvrA-YPet 150 from its endogenous promoter, and lacks the genes for uvrB and mfd (uvrA-YPet DuvrB Dmfd). The 151 measured lifetimes of UvrA-YPet in this strain were found to be identical within error (tUvrA|uvrA-YPet 152 DuvrB Dmfd, slow = 22 ± 8 s (27 ± 4 %) and tUvrA|uvrA-YPet DuvrB Dmfd, fast = 1.4 ± 0.4 s (73 ± 4%)) to those in the 153 DuvrA DuvrB Dmfd/pUvrA-YPet strain ( Fig. 3a and 3c, Supplementary Figure 2c-d and Table 1). Further, 154 we also measured the binding lifetime of a mutant UvrA that is deficient in its interactions with UvrB 155 and Mfd (Fig. 3b). Since UvrA interacts with both UvrB and Mfd via the interface formed by residues 156 131-250 35, 36, 37 , we expected that the labelled mutant UvrA lacking residues 131-250, UvrA(D131-157 250)-YPet, would be a faithful reporter of binding of kinetics of UvrA alone in uvrB + mfd + cells (Fig. 3b).  Table 1). It is 162 noteworthy that deletion of these residues leads to a complete abolishment of the short-lived species. 163 The reasons for this loss may lie in structural differences between the wild-type and mutant proteins.

UvrA is longer lived on DNA in mfd + cells compared to mfdcells in the absence of exogenous damage 211
We next set out to measure the residence time of DNA-bound UvrA in cells carrying both Mfd and 212 UvrB in the absence of exogenous DNA damage. UvrA is recruited to DNA via Mfd to form the 213 asymmetric handoff complex Mfd-UvrA2-UvrB that couples failed TECs to the repair machinery, unlike 214 the symmetric UvrB-UvrA2-UvrB complex formed during damage surveillance in the absence of Mfd 215 ( Fig. 1) To distinguish between these three scenarios, we imaged UvrA-YPet in wild-type cells. Interval imaging 224 of UvrA-YPet revealed a short-lived species with a lifetime (tUvrA, fast) of 1.9 ± 0.2 s (79 ± 0.2%) and a 225 long-lived species of UvrA with a lifetime (tUvrA, slow) of 12.0 ± 0.8 s (21 ± 2%) ( Fig. 4a Table 1). 249 In cells, UvrA is involved in target search (1.6 ± 0.1 s and 24 ± 1 s lifetimes, Fig. 3c) and damage 250 surveillance as part of UvrA2B2 (8.7 s lifetime, Fig. 3f) in addition to Mfd-dependent UvrA(B) complexes 251 (with lifetime of at least 12 s). To identify whether this long-lived UvrA species (19 ± 1 s lifetime) 252 interacts with Mfd, we treated DuvrA/pUvrA-YPet cells with rifampicin. Under this condition, we 253 expected to recover the lifetime of UvrA as part of UvrA2 or UvrA2B2 complexes. Indeed, 254 measurements of lifetimes of UvrA-YPet in rif-treated DuvrA/pUvrA-YPet cells revealed a lifetime 255 (tUvrA|↑rif, slow) of 11.5 ± 0.6 s (25 ± 2%) and a short lifetime (tUvrA|↑rif, fast) of 1.7 ± 0.1 s (75 ± 2%) (Fig. 4a, 256 e, Supplementary Figure 4g- Notably, rif treatment of DuvrA/pUvrA-YPet cells yielded a lifetime (11.5 s) that is longer than that 260 measured for rif-treated uvrA-YPet cells (9.6 s), and cells lacking mfd (8.7 s). The simplest explanation 261 consistent with these observations is that under conditions of high relative UvrA/UvrB abundance the 262 population is composed of UvrA2B (2)  window and delivered a dose of 20 Jm -2 of damaging 254-nm UV light (Fig. 5a). This was followed by 281 acquiring a single snapshot upon laser illumination with 514-nm light, every five minutes for three 282 hours. Quantification of cellular fluorescence intensities revealed that the integrated fluorescence 283 intensities of single uvrA-YPet cells increase 30 minutes after UV exposure by three-fold, consistent 284 with the rapid deregulation of the SOS inducible uvrA promoter (Fig. 5b) 47,48 . UvrA copy numbers have 285 been suggested to increase from 25 to 250 copies per cell after SOS induction 49 . We note that since 286 the experimental conditions associated with these measurements are not available in the published 287 literature, we are unable to effectively compare our measurements with these numbers. YPet corresponding to a short-lived species with a lifetime (tUvrA| Dmfd, UV fast) of 1.6 ± 0.1 s (77 ± 3%) and 297 a long-lived species of UvrA corresponding to a lifetime of (tUvrA| Dmfd, UV slow) 13.1 ± 0.6 s (23 ± 2%) (Fig.  298 5c, and Supplementary Figure 5, Table 1). Strikingly the lifetime of the slowly dissociating species was 299 larger than that detected in the absence of exogeneous DNA damage (8.7 s). 300 Seeking an explanation for the increase in binding lifetime of UvrA-YPet following UV exposure, we 301 wondered if the longer lifetime of UvrA detected in these experiments represented temporally 302 averaged measurements. Since each set of interval measurements lasted 25 min, we proceeded to 303 disaggregate each data set into the four constitutive 25-min intervals after UV exposure. Analysis of 304 the resulting data from each time window revealed that the measured lifetime of UvrA changes as a 305 function of the experimental timeline after the UV pulse (Fig. 5c, and Supplementary Fig. 5, Table 1). 306 Indeed, in the first 25 minutes, the slow lifetime of UvrA (9.6 ± 1 s) matched that measured in the 307 absence of DNA damage (8.7 ± 0.4 s). This lifetime increased to a maximum of 15 ± 2 s in the 50-75 308 minute time window, finally plateauing to 15 ± 4 s in the 75-100 minute time window after UV 309 exposure. 310 There are two main takeaways from these experiments. First, the lifetime of short-lived UvrA does not 311 change upon UV exposure, and is identical to that measured in the absence of any exogenous DNA 312 damage. We therefore conclude that this species is involved in binding undamaged DNA. Second, since 313 the lifetime of long-lived UvrA changes upon UV exposure, we conclude that this species is engaged 314 in DNA damage recognition. 315 Next, we repeated the interval imaging experiments on wild-type cells (uvrA-YPet) following exposure 316 to a 20 Jm -2 pulse of 254-nm UV light provided in situ. In this case, UvrA-YPet exhibited two kinetic 317 populations after UV-exposure, a short-lived population with a lifetime of 1.5 ± 0.1 s (74 ± 2 %) and a 318 second, longer lived population with a lifetime of 9.9 ± 0.4 s (26 ± 2%) (Fig. 5d, and Supplementary 319 Figure 6, Table 1). As before, we disaggregated each data set into the four constitutive 25-min intervals 320 after UV exposure. In contrast to TCR-deficient cells, the measured lifetime of UvrA in wild-type cells 321 remained low (8.4 ± 0.6 s) in the 75-100 minute time window after UV exposure. These data indicate 322 that UvrA is turned over faster in an Mfd-dependent manner during the SOS response. In this work, we set out to visualize the binding behaviour of fluorescently tagged UvrA in living cells 344 over the experimentally accessible, and biologically relevant timescale of 0.1 -1000 s. We found that 345 in live cells, UvrA exhibits foci with residence times ranging from a few hundred milliseconds, to tens 346 of seconds. Here, using single-molecule imaging in combination with chemical and genetic tools, we 347 characterized the DNA-binding lifetimes of UvrA in its interactions with its key binding partners in live 348

cells. 349
We identified that UvrA is long-lived on DNA in the absence of Mfd and UvrB with a lifetime of 24 s. UvrA complexes in cells (Fig. 5f). A prediction of this hypothesis is that elevated concentrations of UvrB 367 would lead to an enhanced turnover of Mfd in the Mfd-UvrA2-UvrB complex. Third, experiments on 368 cells exposed to UV light revealed that the binding kinetics of UvrA mirrored those of Mfd suggesting 369 that these two proteins participate in a TCR handoff complex during the SOS response. Further, UvrA 370 is turned over faster in an Mfd-dependent manner during the SOS response compared to untreated 371 cells (Fig. 5). The elevated copy numbers of UvrA and UvrB following UV treatment ( Fig. 5b and refs 47, 372 48 ) may explain the faster turnover of UvrA during the SOS response. 373 The lifetime of UvrA in TCR-deficient UV-treated cells increased to 15 s, but remained low in UV-374 treated wild-type cells. An explanation for this striking difference may be found in work by Crowley 375 and Hanawalt 46 . The differences observed in our experiments may fundamentally be attributable to 376 the efficiencies with which UvrA recognizes CPDs and 6-4 photoproducts. CPD lesions 51, 52 being less 377 helix distorting than 6-4 photoproducts 53, 54 , are generally inefficiently recognized by damage 378 recognition enzymes in NER across organisms 55,56 . CPD recognition by RNA polymerase and 379 subsequent TCR at such sites is a major mechanism for CPD removal after UV treatment 46 . Compared 380 to untreated cells that remove greater than 80% of CPD lesions within 40 minutes following UV 381 treatment, rif-treated cells removed less than 50% of CPD lesions at the same time point 46

Construction of strains and plasmids 405
Escherichia coli MG1655 uvrA-YPet was constructed using λ Red recombination as previously 406 described for Mfd using the linker sequence described previously 12 . Deletion constructs were created 407 by replacing the indicated gene with a kanamycin cassette flanked by FRT sites as described 408 previously 12 . Sequence specified wild-type uvrA and uvrA(D131-250) geneblocks (including the native 409 uvrA promoter) were ordered from IDT (Coralville, USA) and subcloned into pHH001 12 using standard 410 molecular biology techniques. Plasmids were sequenced on both strands prior to use. Strain 411 expressing mutant UvrB was created using CRISPR-Cas9 assisted λ Red recombination as previously 412 spectinomycin (50 μg per mL) was added to the growth media. Cells in early exponential phase were 418 loaded in flow cells at 30 °C, followed by a constant supply of aerated EZ-rich defined media at a rate 419 of 30 µL per min, using a syringe pump (Adelab Scientific, Australia). 420

Single-molecule live-cell imaging 421
Single-molecule fluorescence imaging was carried out with a custom-built microscope as previously 422 described 12 . Briefly, the microscope comprised a Nikon Eclipse Ti body, a 1.49 NA 100x objective, a 423 514-nm Sapphire LP laser (Coherent) operating at a power density of 71 W cm -2 , an ET535/30m 424 emission filter (Chroma) and a 512 x 512 pixel 2 EM-CCD camera (either Photometrics Evolve or Andor 425 iXon 897). The microscope was operated in near-TIRF illumination 59 and was controlled using NIS-426 Elements (Nikon). PAmCherry-tagged proteins were imaged as described previously 12 . 427 Fluorescence images were acquired in time-series format with 100 ms frames. Each video acquisition 428 contained two phases. The first phase (50 frames) aimed to lower background signal by continuous 429 illuminating, causing most of the fluorophores to photo-bleach or to assume a dark state. The second 430 phase (single-molecule phase lasting for 100 frames) is when single molecules can be reliably tracked 431 on a low background signal. In the second phase, consecutive frames were acquired continuously or 432 with a delay time (td). 433

Image analysis 434
Image analysis was performed in Fiji 60 , using the Single Molecule Biophysics plugins (available at 435 https://github.com/SingleMolecule/smb-plugins), and MATLAB. First, raw data were converted to TIF 436 format, following by background correction and image flattening as previously described 12 . Next, foci 437 were detected in the reactivation phase by applying a discoidal average filter (inner radius of one pixel, 438 outer radius of three pixels), then selecting pixels above the intensity threshold. Foci detected within 439 3-pixel radius (318 nm) in consecutive frames were considered to belong to the same binding event. 440

Interval imaging for dissociation kinetics measurements 441
Interval imaging was performed as described previously 12 . Briefly, the photobleaching phase 442 contained 50 continuous 0.1-s frames. In phase II, 100 discontinuous 0.1-s frames (tint = 0.1 s) were 443 collected with time-lapse time (ttl) ranging from ttl = (0.1, 0.2, 0.5, 1, 2, 4, 8, 10). In each experiment, 444 videos with varying td were acquired. Foci were detected using a relative intensity threshold of 7 or 8 445 above the background as appropriate. Depending on the construct being imaged, between 3-15 446 repeats of each experiment were collected for each strain. Cumulative residence time distribution of 447 binding events detected in all data sets were generated for each ttl.