Shuttling along DNA and directed processing of D-loops by RecQ helicase support quality control of homologous recombination

Significance

RecQ helicase and its eukaryotic homologs are thought to play crucial roles in the quality control of homologous recombination (HR)-based DNA repair. These enzymes have multiple functions in processes that can either promote or suppress HR. A major role suggested for RecQ is the selective inhibition of illegitimate recombination events that could lead to loss of genome integrity. How can RecQ enzymes perform an exceptionally wide range of activities and selectively inhibit potentially harmful recombination events? Here, we propose a model in which the conserved domain architecture of RecQ senses and responds to the geometry of DNA substrates to achieve HR quality control.

Abstract

Cells must continuously repair inevitable DNA damage while avoiding the deleterious consequences of imprecise repair. Distinction between legitimate and illegitimate repair processes is thought to be achieved in part through differential recognition and processing of specific noncanonical DNA structures, although the mechanistic basis of discrimination remains poorly defined. Here, we show that Escherichia coli RecQ, a central DNA recombination and repair enzyme, exhibits differential processing of DNA substrates based on their geometry and structure. Through single-molecule and ensemble biophysical experiments, we elucidate how the conserved domain architecture of RecQ supports geometry-dependent shuttling and directed processing of recombination-intermediate [displacement loop (D-loop)] substrates. Our study shows that these activities together suppress illegitimate recombination in vivo, whereas unregulated duplex unwinding is detrimental for recombination precision. Based on these results, we propose a mechanism through which RecQ helicases achieve recombination precision and efficiency.

The ability to repair DNA damage is essential for all living organisms. Homologous recombination (HR) is the most accurate DNA repair pathway, which is used to correct DNA lesions including double-strand breaks, single-strand gaps, and interstrand cross-links (1, 2). Ideally, HR occurs between allelic loci (legitimate HR), thereby avoiding genome rearrangements. However, in some cases, HR can take place between nonallelic segments containing short homologous regions, resulting in illegitimate recombination (IR). IR can lead to genome rearrangements, loss of genetic material, and altered gene functions, which ultimately lead to cell death or malignant transformation. HR must therefore be tightly controlled to preserve genome integrity. Despite their paramount importance, the mechanisms of HR quality control are largely unknown.

RecQ-family helicases are essential HR regulators in organisms ranging from bacteria to humans (1). Mutations in RecQ helicases are associated with chromosome aberrations, elevated recombination frequency, more frequent sister chromatid exchange, and cancer predisposition in higher organisms (3, 4). The capability of these helicases to unwind noncanonical DNA structures of HR intermediates is proposed to be crucial for HR progression (5, 6). In addition, RecQ helicases are thought to efficiently inhibit HR via disruption of displacement loop (D-loop) structures, early HR intermediates in which a short single-stranded DNA (ssDNA) segment invades a homologous double-stranded DNA (dsDNA) segment (7, 8). Recent studies showed that ds DNA regions are repetitively unwound and reannealed by RecQ helicases (shuttling activity), highlighting that these enzymes unwind DNA via mechanisms that are more complex than simple unidirectional strand separation (9–11). These unique shuttling and D-loop disrupting activities may serve HR quality control.

At least one RecQ homolog of each investigated organism [e.g., Escherichia coli RecQ, Saccharomyces cerevisiae Sgs1, and human Bloom’s syndrome helicase (BLM)] shares a conserved modular domain architecture. It comprises two RecA-like motor domains required for ATP hydrolysis-driven ssDNA translocation with 3′–5′ directionality, a protein structure-stabilizing zinc-binding domain (ZBD) and two auxiliary DNA-binding elements: the winged-helix domain (WHD) and the helicase-and-RNaseD–C-terminal (HRDC) domain (6) (Fig. 1A). The WHD generally facilitates DNA unwinding and substrate recognition, and mediates protein–protein interactions (6). The HRDC domain also contributes to substrate specificity (12–14). Despite their importance in DNA processing by RecQ, the detailed mechanisms of how these accessary domains coordinate with, and modulate, RecQ activity have not been elaborated.

Fig. 1.

The HRDC domain induces geometry-dependent pausing and shuttling during DNA hairpin unwinding. (A, Left) Structure of E. coli RecQ fragment comprising the core (RecA-like and ZBD domains) and the WHD [Protein Data Bank (PDB) ID code 1OYW], shown alongside that of the isolated HRDC domain (PDB ID code 1WUD). (Right) Domain structure of RecQ constructs used in this study. (B) Schematic of hairpin (main panel) and gapped (Inset) DNA substrates used in magnetic-tweezers experiments. The 3′-biotinylated end of the ssDNA segment of the hairpin DNA substrate was attached to a streptavidin-coated 2.8-µm magnetic bead, whereas the 5′ end of the DNA substrate was attached to a digoxigenin-labeled dsDNA handle bound to the anti-digoxigenin–coated flow cell surface. A pair of magnets pulled the bead upward, imposing a constant tension on the DNA. DNA extension (Ex) was measured by tracking the 3D position of the bead. The translocation direction is indicated by the green drop symbol. (C) Example traces of 174-bp hairpin DNA unwinding by RecQ constructs. Each trace includes multiple unwinding events typified by a gradual increase in the number of unwound base pairs followed by a rapid reannealing when the enzyme dissociated from the hairpin (Left). Individual unwinding events (blue shaded regions) are shown in the panels at the Right. The duration of each event is indicated at the Top of the individual plots. Fig. S1A includes additional hairpin DNA unwinding traces. (D) Mean and step rates of hairpin DNA unwinding by RecQ constructs. (E) Example traces of gapped DNA unwinding by RecQ, RecQ*, and RecQ-dH. Fig. S3 includes additional gapped DNA unwinding traces. (F) Mean and step rates of gapped DNA unwinding by RecQ constructs. Error bars in D and F correspond to the SD of Gaussian fitting parameters (Figs. S2 and S4).

To elucidate the mechanistic basis of DNA processing and HR quality control by E. coli RecQ helicase, we used a magnetic-tweezers–based single-molecule assay to determine the contributions of the accessory domains to the unwinding activity of RecQ by measuring DNA unwinding by wild-type (WT) RecQ and RecQ variants harboring mutations in the HRDC and WHD (Fig. 1A). In this assay, we used two different DNA constructs to understand the effect of DNA geometry on RecQ activity: a DNA hairpin in which both ends of the DNA are under tension and a gapped DNA in which only the translocating strand is under tension (Fig. 1B). Both DNA constructs contain a ssDNA–dsDNA junction providing a specific high-affinity binding site for a single RecQ molecule (15), which allowed us to perform the unwinding assays at low RecQ concentrations (typically, 50 pM). In addition, we developed a transient kinetic assay to monitor D-loop disruption directly. Using the engineered RecQ variants (Fig. 1A) permitted us to further define the mechanistic roles of the HRDC and WHD domains in disrupting D-loop HR intermediates. We show that the WHD acts as a dsDNA-binding processivity factor, whereas the HRDC domain interacts with the displaced DNA strand, inducing pausing and shuttling during unwinding of dsDNA in a hairpin, but not a gapped, geometry. Moreover, the HRDC domain orients the helicase on D-loops to promote specific disruption of the invading strand. Together, the WHD and HRDC domains enhance shuttling and D-loop processing in specific DNA geometries.

The ability to engineer the complex activities of RecQ by altering specific domains, while preserving dsDNA unwinding, enabled us to test the role of these activities in vivo. We show that shuttling and directed D-loop disruption activities mediated by the HRDC enable disruption of IR invasions, thereby supporting HR quality control. Indeed, in the absence of these complex activities, dsDNA unwinding alone is detrimental to HR precision in vivo. These results reveal that sensing of the DNA substrate geometry mediated by the HRDC and WHD domains, is a key feature enabling precise and efficient recombination by RecQ helicase.

Results

The WHD Promotes Processive Unwinding, Whereas the HRDC Domain Mediates Pausing and Shuttling.

To understand how the accessory domains affect dsDNA unwinding, we generated a series of E. coli RecQ protein constructs (Fig. 1A). In addition to WT RecQ (denoted as RecQ), we generated a point mutant harboring the Y555A substitution within the HRDC domain (denoted as RecQ*) that abolishes the HRDC–ssDNA interaction (13). We also made a HRDC deletion construct (RecQ-dH, lacking amino acid 524 onward), and a WHD-HRDC deletion construct (RecQ-dWH, lacking amino acid 415 onward) (16, 17).

We measured the unwinding activity of the constructs in a magnetic-tweezers assay using a 174-bp DNA hairpin substrate (Fig. 1B) at saturating ATP concentration (1 mM; Table S1). Unwinding by RecQ was typified by numerous pauses often accompanied by periods of repetitive unwinding and reannealing, collectively referred to as shuttling. Reannealing events occurred on two timescales: (i) slow reannealing, suggesting closure of the hairpin behind the helicase as it translocated on ssDNA after completely unwinding the hairpin; and (ii) rapid incomplete reannealing (abrupt reduction of DNA extension followed by a gradual increase in DNA extension), suggesting transient release or backsliding of the helicase core along ssDNA (Fig. 1C and Fig. S1). Rapid incomplete reannealing events spanning more than 50 bp were occasionally observed but were not considered as shuttling as they are indistinguishable from events in which RecQ dissociates from the hairpin junction and the hairpin is prevented from completely reannealing by a second RecQ bound to distal ssDNA (SI Materials and Methods). In contrast to the frequent pausing and shuttling by RecQ throughout the hairpin, we found that the HRDC point mutant RecQ* unwound the hairpin efficiently with substantially shorter pauses and dissociated rapidly with less frequent shuttling events once the hairpin was unwound (Fig. 1C, Fig. S1, and Table 1). Deletion of the HRDC domain (RecQ-dH) further reduced the pausing and shuttling frequency, suggesting that the effects of the HRDC domain are largely, but not entirely, due to ssDNA binding mediated by Tyr⁵⁵⁵ (Fig. 1C, Fig. S1, and Table 1). In particular, RecQ* exhibited shuttling behavior at the end of DNA hairpin similar to that of RecQ, consisting of pausing and short-distance unwinding followed by fast reannealing (Fig. S1B). On the other hand, RecQ-dH exhibited either rapid disassociation or slow reannealing and unwinding over larger extents (>50 bp) at the end of the hairpin (Fig. S1B). Further deletion of the WHD (RecQ-dWH) significantly decreased the processivity (number of DNA base pairs unwound during a single binding event), indicating that the WHD stabilizes RecQ on DNA during unwinding (Fig. S2).

Table S1.

Additional mechanistic parameters of RecQ constructs

Parameter	RecQ	RecQ*	RecQ-dH	RecQ-dWH
ATPase activity
DNA-free ATPase, s−1†	0.21 ± 0.08	0.22 ± 0.04	0.26 ± 0.08	0.14 ± 0.06
KATP, µM‡	16 ± 2	46 ± 11	22 ± 6	23 ± 3
kend, s−1§	16 ± 1	16 ± 1	51 ± 0.1	42 ± 0.6
koff,end, s−1§	3.2 ± 0.2	1.9 ± 0.1	3.8 ± 0.1	8.4 ± 2.2
ATPase-translocation coupling, nt/ATP§	1.0 ± 0.1	1.0 ± 0.1	1.0 ± 0.1	1.0 ± 0.3
DNA binding Kd, nM
ss54¶	16 ± 4	40 ± 6	70 ± 20	>3,000
3T#	180 ± 30 (20 ± 6)||	200 ± 20 (40 ± 4)||	1,000 ± 100 (42 ± 2)||	N.D.
DL3#	8 ± 1	8 ± 4	20 ± 5	N.D.
DL4#	8 ± 2	6 ± 1	26 ± 4	N.D.
Apparent DNA unwinding and rebinding rate constants, s−1††
kU‡‡	0.13 ± 0.01 (0.12 ± 0.01)§§	0.14 ± 0.04	0.15 ± 0.03	0.12 ± 0.04
kR	0.02 ± 0.001 (0.018 ± 0.01)§§	0.017 ± 0.001	0.017 ± 0.001	0.009 ± 0.001
kR′	0.02 ± 0.01 (0.018 ± 0.01)§§	0.018 ± 0.01	0.06 ± 0.005	0.010 ± 0.001
Fraction of DNA binding configurations††
fDLI	0.94 ± 0.12 (0.58 ± 0.11)§§	0.73 ± 0.16	0.35 ± 0.04	0
fDLE	0 (0)§§	0	0	0
fDLI′	0 (0)§§	0.04	0.22	0.06 ± 0.01
fDLN	0.05 ± 0.01 (0.41 ± 0.03)§§	0.22 ± 0.01	0.42 ± 0.08	0.94 ± 0.02
f3TE	0.60 ± 0.06 (0.14 ± 0.05)§§	0.58 ± 0.05	0.20 ± 0.05	0.26 ± 0.03
Shuttling at the end of DNA hairpin
Relative frequency of each shuttling pathway (A, B, C)¶¶	0.09, 0.70, 0.21	0, 0.59, 0.41	0.56, 0.24, 0.20	N.D.

N.D., not determined.

^†ATPase activity of RecQ constructs (500 nM) in the absence of DNA at saturating ATP concentration (1 mM). Determined from PK/LDH-coupled ATPase experiments.

^‡Michaelis constant of RecQ constructs (10 nM) for ATP in the presence of saturating dT₅₄ concentration (1 µM; except 3 µM for RecQ-dWH). Determined from PK/LDH-coupled ATPase experiments.

^§Determined from the oligo-dT length dependence of ATPase k_cat values (Fig. S5D) by fits using Eq. S3.

^¶Determined from direct fluorescence anisotropy titrations using ss54-FLU (Fig. S7A).

^#Determined from competitive titration experiments (Fig. S7 B–D).

^||Values in parentheses were determined from direct titrations with 3T monitoring the intrinsic RecQ fluorescence signal (Fig. S7E; using 100 nM RecQ constructs).

^††Values reported are determined from global fitting of the model shown in Fig. 2A to DL4, DL3, and 3T unwinding data of each helicase construct (Fig. 2 B–F).

^‡‡Values of k_U also represent lower bounds for the rate constant of nonproductive unwinding runs.

^§§Values in parentheses were determined from control experiments performed at elevated ssDNA trap strand concentration (15 μM) (Fig. S6E).

^¶¶A, HRDC independent resetting; B, tethered shuttling; C, untethered sliding (Fig. S9). Combined shuttling events (Fig. S1B), that is, exhibiting both A and C behavior, were scored as both pathways. Other parameters determined for shuttling are shown in Table 1.

Fig. S1.

Additional single-molecule records of hairpin DNA unwinding by RecQ constructs. (A) Traces of hairpin DNA unwinding by RecQ, RecQ*, RecQ-dH, and RecQ-dWH. Experiments were performed as in Fig. 1 B and C (Materials and Methods). Individual traces are displaced along the y axis for clarity. (Scale bar: 100 bp.) (B) Examples of shuttling behavior occurring at the end of the hairpin DNA for RecQ, RecQ*, and RecQ-dH. Note the differences in the timescales represented by the black bars. The presence of the HRDC domain in WT RecQ increases the frequency and duration of shuttling, while decreasing its extent: long-duration shuttling events occur over a short extent (∼30 bp or less) with a high probability of rapid reannealing (HRDC-tethered shuttling, pathway B in Fig. S9). For RecQ* and RecQ-dH, the frequency and duration of shuttling decrease significantly, whereas the extent increases (∼50 bp or more). For RecQ*, rapid reannealing occurs frequently similar to RecQ (untethered sliding, pathway C in Fig. S9), whereas for RecQ-dH slow reannealing events corresponding to single-strand translocation are dominant (HRDC-independent resetting, pathway A in Fig. S9). Different shuttling behaviors are color-coded for differentiation [HRDC-independent resetting (pathway A), red; HRDC-tethered shuttling (pathway B), blue; untethered sliding (pathway C), green; mixed shuttling behaviors A and C, brown; B and C, dark green].

Table 1.

Mechanistic parameters of RecQ constructs

Construct	RecQ	RecQ*	RecQ-dH	RecQ-dWH
DNA unwinding†
Mean rate, bp/s	5.0 ± 2.6 (67 ± 2)	48 ± 0.3 (77 ± 5)	47 ± 1 (95 ± 1)	78 ± 1
Step rate, bp/s	83 ± 2 (91 ± 2)	82 ± 1 (114 ± 5)	87 ± 1 (107 ± 2)	96 ± 2
Shuttling frequency, % of N‡	85 (28)	53 (17)	39 (17)	N.D.
Shuttling frequency at the end of DNA hairpin, % of N*	85 (N.D.)	57 (N.D.)	39 (N.D.)	N.D.
Shuttling duration, s‡	13 ± 3 (2.4 ± 0.3)	2.0 ± 0.2 (2.4 ± 0.8)	2.3 ± 0.6 (2.8 ± 1)	N.D.
Median run length, bp	104 ± 4 (78 ± 1)	143 ± 1 (113 ± 2)	157 ± 1 (237 ± 11)	47 ± 1
Total number of complete unwinding events, N*	20 (N.D.)	69 (N.D.)	100 (N.D.)	N.D.
Total number of unwinding events, N	95 (141)	76 (87)	100 (30)	193
ssDNA binding and ATP-driven translocation
Kd, dT54, nM§	34 ± 4	38 ± 3	91 ± 18	>3,000
Occluded site size, nt per protein monomer,¶	28 ± 2	27 ± 3	27 ± 2	19 ± 3
Translocative ATPase (ktrans), s−1#	32 ± 1	37 ± 1	54 ± 1	47 ± 2

N.D., not determined. Additional mechanistic parameters are shown in Table S1.

^†Magnetic-tweezers data obtained with hairpin DNA substrate (gapped DNA data in parentheses) (Fig. 1 and Figs. S1–S4). Reported errors are SDs of fits.

^‡The reported shuttling frequency (duration) includes shuttling events occurring during unwinding and at the end of the hairpin DNA substrate, and indicates strand-switching frequency (duration) for gapped DNA in parentheses.

^§From fluorescence anisotropy titrations (Fig. S5A).

^¶From oligo-dT length dependence of the apparent DNA dissociation constant during ATPase activity (Fig. S5C), based on Eqs. S1 and S2.

^#From oligo-dT length dependence of ATPase k_cat (Fig. S5D), based on Eqs. S1 and S3.

Fig. S2.

Quantification of single-molecule measurements of DNA unwinding. (A) The mean rate was calculated from a linear fit between the initial unwinding position and the maximum unwinding position before the enzyme dissociated or the hairpin rezipped (gray line). The step rate corresponds to the unwinding rate between pauses obtained from fitting the trajectories with a step-finding algorithm (black line). (B) Mean rate, step rate, and run length distribution of hairpin DNA unwinding by RecQ constructs. The run length distributions correspond to the maximum unwound position for each unwinding event plotted as normalized cumulative distributions to estimate processivity. Gaussian fits to rate histograms (black lines) and median positions of run length distributions (dashed lines) are shown. The error bars indicate SDs (square root of the number of events in each bar).

The mean hairpin unwinding rate (i.e., between the initial and maximum unwinding positions) of RecQ was severalfold slower than that of the mutant constructs (Fig. 1D, Fig. S2, and Table 1). This large difference resulted from frequent, long-duration, pauses as the step rates (i.e., between pauses) for all constructs were comparable (Fig. 1D and Table 1). The marked reduction of the median run length of RecQ-dWH compared with the WHD-containing constructs demonstrates that the WHD significantly contributes to the processivity of unwinding (Fig. S2 and Table 1; see also below).

Helicase pausing appeared to be more frequent at certain positions on the hairpin. For example, many pauses occurred near the 100-bp unwinding position (Fig. 1 and Fig. S1). This behavior may reflect sequence-dependent unwinding and/or pause kinetics, as has been observed for other superfamily (SF) 1 and 2 helicases including hepatitis virus NS3 helicase (18), XPD helicase (19), and RecBCD helicase (20). Additional experiments will be required to characterize the sequence dependence of RecQ unwinding and pausing, and the role of the HRDC domain in these processes.

The RecQ HRDC Domain Binds to the Displaced DNA Strand During Unwinding.

The above results led us to suspect that the frequent pausing by RecQ is caused by binding of the HRDC domain to one of the nascent ssDNA strands during hairpin unwinding, consistent with the ssDNA-binding capability of the isolated HRDC domain (13). As both ssDNA strands are under tension in the hairpin substrate (Fig. 1B), the binding of the HRDC domain to either strand could transiently stall the enzyme. To determine the ssDNA-binding geometry of HRDC-dependent pausing, we measured RecQ unwinding on a gapped dsDNA substrate (Fig. 1B, Inset). Here, unlike in the hairpin substrate, the displaced ssDNA segment (i.e., the one complementary to the tracking strand) is not under tension. Thus, if the HRDC domain bound to the displaced ssDNA, it would not hinder progression of the enzyme. On the other hand, significant pausing similar to that observed with the DNA hairpin would occur if the HRDC bound the tracking ssDNA strand or the dsDNA ahead of the enzyme. We observed no significant pauses during unwinding of the gapped DNA substrate by RecQ (Fig. 1E and Fig. S3). Moreover, the mean unwinding rates of RecQ, RecQ*, and RecQ-dH on the gapped DNA substrate were comparable, indicating that pausing and shuttling during unwinding of the hairpin substrate is mediated by HRDC binding to the displaced ssDNA strand (Fig. 1F, Fig. S4, and Table 1). Nonetheless, RecQ displays an overall less efficient helicase activity than RecQ-dH (Table 1), suggesting that the HRDC inherently regulates RecQ core unwinding activity (see below). This may be important to maintain the balance between exonuclease such as RecJ and helicase unwinding or to naturally prevent long ssDNA exposure during DNA resection at the initial DNA recombination step (21, 22).

Fig. S3.

Additional single-molecule records of gapped DNA unwinding. Traces are shown for RecQ, RecQ*, and RecQ-dH. Experiments were performed as in Fig. 1 B and E (Materials and Methods). Individual traces are displaced along the y axis for clarity. (Scale bar: 100 bp.)

Fig. S4.

Mean rate and step rate distribution of gapped DNA unwinding by RecQ constructs. Gaussian fits to rate histograms (black lines) are shown. The error bars indicate SDs.

All RecQ constructs displayed occasional strand-switching events before dissociating from the gapped DNA substrate (Table 1). This behavior is distinct from the shuttling observed with the WT enzyme on the hairpin substrate, and is clearly independent of the HRDC domain, as it was observed for all constructs.

During ssDNA Translocation, the WHD Contributes to DNA Binding, Whereas the HRDC Domain Suppresses the ATPase Activity.

To further dissect the mechanistic roles of the WHD and HRDC domains, we evaluated their effects on DNA binding and DNA-activated ATP hydrolysis using ssDNA (oligo-dT) substrates. Fluorescence anisotropy titrations of 3′–Cy3-labeled dT₅₄ showed that the WHD markedly, whereas the HRDC domain moderately, contributes to the ssDNA binding affinity of RecQ in the absence of ATP (Fig. S5A and Table 1). As previously shown (23–25), the ssDNA binding affinity and the occluded site size of helicases on ssDNA during ATP hydrolysis can be deduced from the ssDNA concentration dependence of their ATPase activity (Fig. S5B). We found that the occluded site size of RecQ on ssDNA (per protein monomer) is not influenced by HRDC deletion, but it is significantly reduced by WHD deletion (Fig. S5C and Table 1).

Fig. S5.

Role of the WHD and HRDC domains in DNA binding and ATPase activity. (A) To further assess the influence of the WHD and HRDC domains on the unwinding activity, we investigated their contribution to the binding affinity of RecQ for a homopolymer ssDNA (dT₅₄). We monitored the fluorescence anisotropy of 15 nM Cy3-labeled dT₅₄ upon titration with RecQ constructs in the absence of nucleotides. Solid lines show hyperbolic fits. Asterisks indicate the location of fluorescent label on DNA. The dT₅₄ affinity of RecQ constructs decreased in the order RecQ > RecQ* > RecQ-dH >> RecQ-dWH, with the latter construct showing no detectable DNA binding (Table 1). This finding reflects the dominant role of the WHD in the overall interaction of RecQ with ssDNA. (B) To correlate the DNA-restructuring activities of the constructs with their ATP hydrolysis activity, we determined their steady-state ATPase kinetic parameters in the absence (Table S1) and presence (B) of ssDNA substrates. The DNA-free (basal) ATPase activity of RecQ constructs was largely unaffected by the mutations (Table S1). All RecQ constructs showed marked activation of their ATPase activity by dT₇₂ substrate (10 nM enzyme, 1 mM ATP, PK-LDH–coupled assay). Solid lines show fits using a quadratic binding equation (Eq. S1) from which k_cat (ATPase catalytic constant) and K_d,app (apparent DNA dissociation constant during ATPase activity) values were determined. The k_cat values of RecQ proteins lacking the HRDC domain (dH and dWH constructs) were significantly higher than those of the corresponding HRDC-containing constructs, whereas RecQ* showed RecQ-like k_cat values (Table S1) (16, 17). These findings suggest that the HRDC domain—independent of its ssDNA interaction—hinders ATP hydrolysis by the enzymes. In recent crystal structures of BLM helicase (PDB ID codes 4O3M and 4CGD), the HRDC domain interacts with the motor core both in the absence and in the presence of 3′-tailed DNA, and the interaction was proposed to hinder the ATPase activity (26). (C) As determined from the oligo-dT concentration dependence of the ATPase activity (examples for dT₇₂ are shown in B) (Eq. S1), the apparent ssDNA dissociation constants of RecQ constructs during ATP hydrolysis (K_d,app) decreased exponentially with oligo-dT length until reaching a lower plateau value. B shows the oligo-dT length dependence of the natural logarithm of K_d,app for RecQ constructs, respectively. This dependence allowed the determination of the occluded site sizes of constructs on ssDNA by fitting the model described in Eq. S2 (solid lines show best fits) (23–25). Determined parameters are listed in Table 1 and in the description of Eq. S2. The occluded site sizes (per protein monomer) of RecQ, RecQ*, and RecQ-dH were comparable, indicating that the HRDC domain does not contribute significantly to ssDNA binding (Table 1). In contrast, the occluded site size for RecQ-dWH decreased by ∼8 nt (Table 1). (D) Previous studies showed that the ssDNA length dependence of steady-state ATPase activity (k_cat) provides information about the ssDNA translocation mechanism (23, 24, 53, 54). For all RecQ constructs, k_cat depended on oligo-dT length. However, this dependence was less pronounced for proteins lacking the HRDC domain (dH and dWH constructs). Analysis of the data based on our previous work (23, 24, 55) (Eq. S3, Table 1, and Table S1) revealed that the rate constant of dissociation from DNA ends (k_off,end; Table S1) is slightly elevated for constructs lacking the HRDC domain. Interestingly, the ATPase activity at the end of DNA molecules (k_ATP,end; Table S1), where further translocation does not occur, is increased markedly for constructs lacking the HRDC domain.

The DNA-free ATPase activity (k_basal) of RecQ was unaffected by the WHD and/or HRDC domains, indicating an intact motor core for all constructs (Table S1). However, the ssDNA-activated ATPase activity (k_cat) was suppressed by the HRDC domain (Fig. S6 B and D) in line with our previous report (17), without influencing the ATP binding affinity to the motor core (K_ATP, Table S1). Furthermore, ATPase suppression was not related to the ssDNA binding ability of the RecQ HRDC domain as RecQ and RecQ* showed similar k_cat values (Fig. S6 B and D, and Table 1). These results suggest that the HRDC domain of RecQ suppresses ATP hydrolysis via interactions with the motor core. Similar HRDC-dependent suppression of the ATPase activity has been shown for human BLM (26).

Fig. S6.

Additional DL4 unwinding experiments. (A) Cy3 fluorometric image of a 12% native PAGE electrophoretogram of DNA species involved in DL4 disruption assays (model shown in Fig. 2A). DNA species (D1 + D3 + D4) and (D1 + D3) never appeared during enzymatic disruption of DL4. (B) Electrophoretogram of the DL4 unwinding experiments shown in Fig. 2B. (C) Example kinetic profile of DL4 unwinding by RecQ in the absence of ATP. Here, the concentration of all DNA species remained constant during the 300-s time course, indicating that all transitions in other experiments resulted from ATP-dependent helicase activity. (D) Example kinetic profiles of DL4 unwinding by RecQ in the absence of ssDNA trap strand. After stopping the reaction, 3 µM trap strand was added to the mixture to prevent reannealing of unwound DNA molecules. As expected, the omission of the ssDNA trap strand markedly accelerated the decomposition of all multistranded DNA species (Fig. 2B). (E) Example kinetic profiles of DL4, DL3, and 3T unwinding by RecQ in the presence of 15 µM trap strand (five times higher than in Fig. 2B). Solid lines represent best fits of the kinetic model shown in Fig. 2A. Elevation of trap strand concentration had no effect on the determined apparent unwinding (k_U) and rebinding (k_R) rate constants, but increased the fraction of the nonproductive DLN pathway (Table S1). This indicates that, when added in a large excess, the ssDNA trap strand accelerates dissociation of the enzyme from DNA substrates during unwinding. A similar effect was observed previously for ssDNA-mimicking polyanions (heparin and dextran sulfate) that were used as protein trap in experiments measuring the translocation kinetics of RecQ helicases (17, 24, 54). DNA species are color coded as in Fig. 2A. Determined parameters are listed in Table S1. DNA substrates are described in Table S2.

The WHD and HRDC Domains Enable Directed Processing of D-Loop Structures.

In addition to its roles in dsDNA unwinding and ATPase modulation, the HRDC domain has been implicated in the processing of DNA substrates containing an invading DNA strand (12), for example, D-loop structures formed during recombination. To decipher the mechanism of D-loop processing and to dissect the roles of the WHD and HRDC domains in the process, we developed a kinetic assay to monitor the enzymatic processing of three related fluorescently labeled DNA substrates (Fig. 2 and Figs. S6 and S7). (i) We recorded the time courses of formation and disappearance of all intermediates during the unwinding of a four-stranded D-loop–like substrate (DL4) (Fig. 2, Fig. S6 A and B, and Tables S1 and S2). In separate experiments, we monitored the unwinding of (ii) a three-stranded D-loop-like substrate (DL3), and (iii) a two-stranded, 3′-tailed DNA structure (3T) (Fig. 2). (DL3 and 3T are constituents of DL4, as depicted in Fig. 2A.) In the unwinding experiments, RecQ constructs premixed with labeled DNA substrates (DL4, DL3, or 3T) were mixed with excess ATP and a large excess of unlabeled ssDNA trap (with a sequence identical to that of the labeled invading strand) to inhibit DNA reannealing and enzyme rebinding to labeled substrates after dissociation. We note that the DL4 preparations contained fractions of the DL3 and 3T species (Fig. 2 and Fig. S6 A and B). However, the initial heterogeneity of DL4 preparations turned out to be useful as the fraction of each DNA species could be accurately determined throughout the time courses—which, in conjunction with separate DL3 and 3T unwinding experiments, facilitated robust determination of the kinetic mechanism by global fitting analysis of all reactions (see below).

Fig. 2.

The WHD and HRDC domains enable directed processing of D-loop structures. (A) Pathways of processing of D-loop–like (four-stranded DL4, black; three-stranded DL3, red) and 3′-tailed (3T, blue) DNA substrates. DNA-bound helicases are shown as green drops pointing in the direction of unwinding (3′–5′ on the tracked DNA strand). Black stars represent the Cy3 fluorescent label at the 5′ end of the invading DNA strand. Enzyme–DNA complexes are initially distributed into different configurations (DLI, DLE, DLI′, DLN; 3TE, 3TN; as described in the text). DLN and 3TN represent all nonproductive unwinding runs starting from any possible enzyme–DNA configuration (white drops; see text). Unwinding (occurring at rate constant k_U) leads to the indicated DNA products. Slow rebinding of enzyme to these DNA products (inhibited by excess ssDNA trap strand, occurring at k_R for DL4 and DL3, or k_R′ for 3T) leads to reformation of enzyme–DNA configurations (with the same distribution as for the initial complexes; reformation and redistribution together are indicated by downward arrows labeled k_R and k_R′). (B–E) Kinetic profiles of DL4, DL3, and 3T unwinding by RecQ (B), RecQ* (C), RecQ-dH (D), and RecQ-dWH (E) [100 nM enzyme, except that 3 μM RecQ-dWH was used in E, 30 nM labeled DNA, 3 µM ssDNA trap strand; means ± SE (n = 5 for DL4, n = 3 for DL3 and 3T) are shown; see also Figs. S6 and S7]. Color code of DNA species is as in A. Solid lines show global best fits to all unwinding data (DL4, DL3, and 3T) of each enzyme variant. (F) Distributions of enzyme-DL (DL4 or DL3) configurations resulting from global fits shown in B–E. Determined parameters are listed in Table S1. DNA substrates are described in Table S2.

Fig. S7.

Fluorescence titrations of DNA substrates with RecQ constructs show strong binding to D-loop structures. (A) Fluorescence anisotropy titration of 10 nM fluorescein-labeled ss54-FLU DNA substrate with increasing amounts of RecQ constructs in the absence of ATP. Solid lines show hyperbolic fits. Determined K_d values are listed in Table S1. (B–D) Competitive fluorescence anisotropy titrations using increasing concentrations 3T (blue), DL3 (red), or DL4 (black) DNA substrates to compete with ss54-FLU for binding to RecQ constructs. ss54-FLU signal was monitored. The concentrations of helicase constructs and ss54-FLU were kept constant at the values shown in the panels. Solid lines in B–D are best fits based on Eqs. S4 and S5. Determined K_d values are listed in Table S1. (E) In competitive titrations, RecQ constructs showed weak apparent binding affinities to 3T (B–D). Therefore, we performed direct titrations of these constructs with 3T, monitoring the change in the intrinsic fluorescence of RecQ constructs (100 nM) upon DNA binding (16). Solid lines show fits using a quadratic binding equation described in ref. 23. Determined K_d values reflected strong binding to 3T by all RecQ constructs, in line with their high affinity to other DNA substrates (Table S1). The 3T binding affinity was slightly decreased by the Y555A point mutation or the deletion of the HRDC domain. These results show that the weak efficiency of 3T competition for binding sites on RecQ constructs (B–D) was not due to weak binding of 3T to RecQ constructs, but probably arose from simultaneous binding of 3T and ss54-FLU to the same RecQ molecule. Strong binding to the DL4, DL3, and 3T species by all WHD-containing constructs, revealed in the experiments shown in this figure, confirmed that practically all DNA molecules were initially enzyme-bound in the unwinding experiments of Fig. 2 and Fig. S6. Furthermore, the rebinding rate constants (k_R for DL4 and DL3, k_R′ for 3T) determined from global fits shown in Fig. 2 were in the same range for all helicase constructs (except for RecQ-dWH), further suggesting that binding to these DNA structures is not markedly influenced by the mutations (Table S1).

Table S2.

Oligonucleotides used in solution biophysical experiments

Name	Sequence, 5′–3′
D1	GACGCTGCCGAATTCTACCAGTGCCTTGCTAGGACATCTTTGCCCACCTGCAGGTTCACCC
D2	GGGTGAACCTGCAGGTGGGCGGCTGCTCATCGTAGGTTAGTTGGTAGAATTCGGCAGCGTC
D3*	Cy3-TAAGAGCAAGATGTTCTATAAAAGATGTCCTAGCAAGGCAC
D4	TATAGAACATCTTGCTCTTA
ss54-FLU	TCCTTTTGATAAGAGGTCATTTTTGCGGATGGCTTAGAGCTTAATTGCGCAACG-Fluorescein
DL4 trap strand	Same as D3 but without Cy3 label
3T	D3 + D4
DL3	D1 + D2 + D3
DL4	D1 + D2 + D3 + D4

dsDNA-forming complementary regions are in bold.

^*Fluorescein-labeled version (instead of Cy3 label) of D3 was used in Fig. S8.

In general, the disappearance of DL4 was accompanied by a transient accumulation of 3T, which was eventually converted to ssDNA; whereas in the separately performed DL3 and 3T unwinding experiments, both substrates showed biphasic disappearance profiles (Fig. 2 B–E). The simplest kinetic model accounting for these behaviors involves multiple D-loop processing pathways dictated by different enzyme–DNA binding configurations and the 3′–5′ unwinding directionality of RecQ (Fig. 2A). In the model, a fraction of initial enzyme–DL4 and enzyme–DL3 complexes (DLI in Fig. 2A) is oriented for disruption of the invasion, producing 3T from DL4 or ssDNA from DL3, as the enzyme tracks the strand to which the invading strand is annealed. In a second fraction (DLE), the enzyme is oriented “outward” from the invasion and thus will produce DL3 from DL4—or, in the case of the DL3 substrate, leave the substrate intact. In a third fraction (DLI′), the enzyme tracks the invading strand starting from its 3′ end, thus producing ssDNA from either DL4 or DL3. In the remainder of the cases (represented by the DLN fraction), no unwinding of DL4 or DL3 occurs, either because the enzyme unwinds one of the dsDNA arms flanking the invasion (which will then rapidly reanneal), or due to premature termination of unwinding initiated from any of the possible initial binding configurations. Unwinding of the 3T substrate occurs from a single binding configuration (3TE), whereas 3TN represents the fraction of enzyme–3T complexes that are not unwound.

For parsimony, successful unwinding of any of the dsDNA segments in the DNA substrates was modeled to occur with a single rate constant k_U, as the lengths of these segments were similar (21-bp invasion, 20-bp other segments). k_U was also used as a lower bound for the rate constant of nonproductive unwinding events (represented in the DLN and 3TN fractions). After termination of a single unwinding run, slow rebinding of the enzyme to the DNA substrates (hindered but not completely inhibited by the ssDNA trap strand) was modeled to occur at rate constant k_R for DL4 and DL3, or k_R′ for 3T (k_R, k_R′ << k_U). The initial partitioning of enzyme–substrate complexes, as well as the four rate constants of the model, were determined by global kinetic fitting of the DL4, DL3, and 3T unwinding reactions of individual RecQ constructs (Fig. 2 B–E and Table S1). In the best-fit models, the DLE fraction was 0 for all assessed constructs, indicating that this pathway does not occur in RecQ reactions (Table S1). Apart from DLE, the omission of any further element of the model resulted in significant deterioration of the fits, indicating that the model comprising the DLI, DLI′, and DLN pathways represents the simplest plausible description of the reactions.

As depicted in Fig. 2A, the DLI and DLI′ fractions represent disruption of the invading strand in D-loop–like substrates, either starting from the branch point (DLI) or the 3′ end of the invading strand (DLI′). In vivo, analogous D-loop disruption results in termination of HR initiation. Conversely, the DLN (and DLE) fractions represent maintenance of the invading strand. As mentioned above, DLN represents the sum of short, unproductive unwinding runs (in any orientation) plus unwinding events targeted at either of the flanking dsDNA “arms” of the D-loop–like substrates (bottom left and right arms in the DL4 and DL3 structures drawn in Fig. 2A). All of these DLN events are followed by substrate DNA reannealing and are thus not detected in our assay (Fig. 2A). In the case of enzymes that are highly processive and bind D-loop–like substrates with high affinity (such as RecQ* and RecQ-dH; Fig. 1C and Fig. S7 C and D), short unwinding runs are rare and DLN will thus dominantly represent unwinding of dsDNA arms. Analogous unwinding events in D-loops in vivo are likely to extend and stabilize the invasion and, thus, propagate HR, either by moving the invasion branch point away from the 3′ end of the invading strand (unwinding of the bottom left dsDNA arm in Fig. 2A), or by liberating template for the DNA synthetic extension of the 3′ end of the invading strand (unwinding of the bottom right arm in Fig. 2A).

The decrease in the disrupted fraction (DLI+DLI′) observed for the HRDC and combined WHD-HRDC deletion constructs reflects the contribution of both domains to the disruption of invasions by RecQ (Fig. 2F). Furthermore, the HRDC domain of RecQ confers a strong bias toward the DLI configuration, and this effect is largely independent of the ssDNA binding capability of this domain (RecQ, RecQ*, and RecQ-dH data in Fig. 2F). In experiments measuring the affinity of RecQ enzyme constructs and isolated (WT and Y555A mutant) HRDC domains for the DNA substrates, we ruled out the possibility that the HRDC domain specifically binds branched DNA structures independent of its ssDNA binding capability (Figs. S7 and S8).

Fig. S8.

Fluorescence anisotropy titrations of DNA substrates with isolated RecQ HRDC domains. The binding of isolated WT (A) or Y555A point mutant (HRDC*; B) HRDC domain to 10 nM fluorescein-labeled DL4 (black), DL3 (red), or ssDNA (D3, gray) substrates was monitored in fluorescence anisotropy titration experiments. DNA substrates are described in Table S2. Fits using the Hill equation (solid lines) revealed noncooperative binding of the WT HRDC domain to DL3 and ssDNA (Hill coefficients were 0.95 ± 0.14 and 0.81 ± 0.11, respectively) with moderate binding affinities (K_d = 0.67 ± 0.08 and 0.85 ± 0.12 μM, respectively). Importantly, the weak binding of WT HRDC to DL4 substrate (K_d >> 10 μM) compared with DL3 and ssDNA indicates that the HRDC domain dominantly interacts with flexible ssDNA regions and does not recognize specific features of the D-loop structure. This hypothesis is further supported by the undetectable binding of HRDC*, in which the Y555A point mutation abolishes ssDNA binding (13), to DL3 and DL4.

Figs. S6–S8 and Table S1 show results of control experiments verifying the kinetic analysis of the D-loop DNA disruption experiments.

Activities Mediated by the HRDC Domain Suppress Illegitimate Recombination in Vivo.

To assess the physiological effects of HRDC and WHD mutations, we created E. coli strains (based on MG1655 used as WT) lacking RecQ (ΔrecQ), or expressing the characterized point mutant (recQ*) or truncated forms (recQ-dH, recQ-dWH) in place of the WT protein (Table S3). The growth curves of WT, recQ*, and recQ-dH strains were similar, whereas ΔrecQ and recQ-dWH showed slightly slower growth at the end of log phase (Fig. 3A). We characterized the genome damage tolerance of the strains via UV irradiation and nitrofurantoin (NIT) survival assays (Fig. 3 B and C, and Table S4). UV irradiation induces the formation of pyrimidine dimers, whereas NIT introduces interstrand cross-links in DNA (27, 28). Both forms of DNA damage can halt DNA replication if unrepaired. With the exception of recQ-dWH, we found no significant effects of recQ mutations on the UV sensitivity of E. coli strains (Fig. 3B and Table S4).

Table S3.

E. coli strains used in this study

Strain	Genotype	Source
YmeI	supE supF	Ikeda Laboratory§
WL95	supE supF metB trpR hsdR tonA (P2)	Ikeda Laboratory§
HI1165	λ cI857	Ikeda Laboratory§
MG1655	F– λ– ilvG– rfb-50 rph-1	CGSC†
MK1830	MG1655 ΔrecQ	Our laboratory
MK555	MG1655 recQ* (Y555A point mutant)	Our laboratory
MK1239	MG1655 recQ-dWH (1239-1830 bp deleted)‡	Our laboratory
MK1569	MG1655 recQ-dH (1569-1830 bp deleted)	Our laboratory
MK1655λ	MG1655 λcI857	Our laboratory
MK1830λ	MG1655 ΔrecQ λcI857	Our laboratory
MK555λ	MG1655 recQ* λcI857	Our laboratory
MK1239λ	MG1655 recQ-dWH λcI857‡	Our laboratory
MK1569λ	MG1655 recQ-dH λcI857	Our laboratory

^†Cell Genetic Stock Center (cgsc.biology.yale.edu).

^‡For technical reasons, encoded protein was truncated 1 aa earlier (at amino acid 413) than the in vitro-characterized RecQ-dWH protein.

^§Dr. Hideo Ikeda, Center for Basic Research, Kitasato Institute, Tokyo.

Fig. 3.

Shuttling and directed D-loop disruption activities are required to suppress illegitimate recombination in E. coli. (A) Growth curves of MG1655 (WT), ΔrecQ, recQ*, recQ-dH, and recQ-dWH E. coli strains. (B and C) Dose-dependent survival curves of E. coli strains exposed to UV irradiation (B) or NIT (C). Solid lines in B and C show fits to averaged data based on a standard dose–response model (Eq. S6). (D) Frequencies of illegitimate recombination (IR) in E. coli strains (log-transformed values), as determined using the λ phage Spi^– assay, with or without UV irradiation (50 J/m²). Asterisks indicate significant difference from WT values (P < 0.05, one-way ANOVA followed by Tukey test). The table shows in vitro activities of RecQ constructs expressed by the individual strains, as determined in biophysical experiments. +, activity present; 0, activity absent; (0), weak activity. Statistics of parameters determined for individual datasets are shown in Table S4.

Table S4.

In vivo effects of recQ mutations

Parameter	WT (MG1655)	ΔrecQ	recQ*	recQ-dH	recQ-dWH
Log LD50, UV irradiation, J/m2†	1.4 ± 0.1 (2.4 ± 0.5)	1.2 ± 0.1 (2.5 ± 0.3)	1.3 ± 0.2 (2.1 ± 0.5)	1.3 ± 0.1 (2.2 ± 0.2)	1.0 ± 0.1* (2.6 ± 0.6)
Log LD50, nitrofurantoin, µg/mL†	0.57 ± 0.02 (19 ± 2)‡	0.02 ± 0.10* (1.5 ± 0.4)**	0.56 ± 0.02 (8.9 ± 0.9)*	0.47 ± 0.05 (9.9 ± 2.5)*	−0.08 ± 0.04* (2.0 ± 0.3)**
Log Spi– frequency§	−9.33 ± 0.15 (–7.58 ± 0.16)	−8.71 ± 0.20* (–6.91 ± 0.13)	−8.79 ± 0.13 (–6.75 ± 0.18)*	−8.60 ± 0.08* (–6.33 ± 0.28)*	−8.55 ± 0.14* (–7.09 ± 0.21)

Significant difference from WT values is indicated by asterisks (one-way ANOVA, Tukey post hoc test, P < 0.05). Double asterisk (**) indicates further significant difference from values with single asterisk (*).

^†Values in parentheses are Hill slopes (Fig. 3 B and C, and Eq. S6). Means ± SE for n = 8–15 are shown.

Deletion of the recQ gene has been shown to sensitize E. coli cells to DNA-damaging agents including UV radiation and NIT (22, 29). UV radiation generates pyrimidine dimers, whereas NIT treatment causes interstrand cross-links in DNA (27, 28). These lesions are generally repaired by different repair pathways (27, 56). However, if left unrepaired, the lesions can cause replication fork stalling, leading to the formation of DSBs or ss gaps in the DNA. Although the RecBCD pathway is the major DSB repair pathway in E. coli (57), the significant effects of recQ deletion on genome damage sensitivity were proposed to imply that the RecF pathway (dependent on RecQ helicase activity) plays a role in DNA repair (58). HR-mediated replication-coupled repair requires resection at DSBs to generate ssDNA overhangs for RecA-catalyzed strand exchange. Resection is catalyzed by the RecBCD complex; however, RecQ along with RecJ nuclease can also resect DSBs or nascent lagging strands at stalled replication forks (37). Interestingly, we found no significant effects of recQ mutations on UV tolerance (with the exception of recQ-dWH). However, the lack of RecQ helicase (ΔrecQ) or the expression of the unwinding-deficient RecQ-dWH variant in place of WT RecQ caused significant increase in NIT sensitivity of E. coli cells, whereas RecQ* and RecQ-dH restored tolerance of these agents to WT levels (Fig. 3 B and C). Together with our biophysical results, these findings imply that the core dsDNA unwinding activity of RecQ is sufficient for resection of DSB ends and the processing of ss gaps during replication-coupled DNA repair, whereas the geometry-dependent complex activities become crucial during HR quality control.

^‡The highly cooperative nature of NIT survival of WT E. coli cells indicates that the repair of interstrand cross-links occurs in a steady state, in response to the formation of these lesions. The rate of lesion formation is likely proportional to NIT concentration. Cells can tolerate NIT up to a threshold concentration, but above this threshold there occurs a sudden collapse of genome integrity (Fig. 3C). Interestingly, the apparent cooperativity of the NIT effect was significantly lower in recQ* and recQ-dH strains than in WT, and further significantly lowered in the ΔrecQ and recQ-dWH strains.

^§See Fig. 3D. Values in parentheses indicate results obtained upon UV light irradiation (50 J/m²). [Note that irradiation conditions in Spi^– assays were different from those applied in UV survival experiments (Materials and Methods).] Means ± SE for n = 10–19 are shown.

Whereas E. coli RecQ is involved in IR suppression under stress-free conditions, the complete loss of RecQ activities has broader and more robust effects when cells suffer elevated levels of DNA damage (29, 31, 59). Both bacterial and eukaryotic RecQ helicases have been implicated in the processing of late replication intermediates (LRIs), catenated DNA structures that arise at converging replication forks (3, 39). E. coli RecQ and topoisomerase III are involved in an alternative LRI processing pathway distinct from the main pathway driven by type II topoisomerases (39, 60). The RecQ HRDC domain is dispensable for LRI processing (39). Accordingly, we did not see marked effects of recQ mutations on the growth of E. coli strains under normal conditions (Fig. 3A).

Nevertheless, our IR frequency measurements suggest that the action of WT RecQ helicase suppresses illegitimate recombination, whereas the unwinding-deficient RecQ-dWH mutant is unable to suppress IRF below the level characteristic of the absence of the enzyme. Importantly, however, RecQ*, which can efficiently remove strand invasions but lacks the shuttling behavior, resulted in increased UV-induced IRF. This effect was even more pronounced for RecQ-dH, which can efficiently unwind DNA but is deficient for both the shuttling and oriented D-loop disruption activities, underscoring the cellular relevance of these activities characterized in vitro.

Besides the effects of RecQ mutations on the shuttling and D-loop disruption activities of the enzyme (Figs. 1 and 2), one must consider the possibility that the elevated IR frequencies observed in RecQ mutant strains may at least partially arise from the abolishment of interactions between RecQ and its partner proteins. However, E. coli ssDNA binding protein (SSB), the only in vitro verified RecQ interaction partner, interacts dominantly with the WHD of RecQ (61, 62). Thus, the recQ* and recQ-dH mutations are unlikely to influence the RecQ–SSB interaction. In general, the Y555A point mutation in RecQ* is alone unlikely to significantly affect any other possible protein–protein interactions. These findings strongly suggest that the elevated UV-induced IR frequencies in the recQ* and recQ-dH strains arise from the altered DNA-restructuring activities of the mutant enzymes. Moreover, the BLM helicase-core construct was able to partially complement the elevated IR frequency of RecQ-deficient E. coli cells (63). As BLM is unlikely to be able to form specific interactions with the protein partners of RecQ, this finding suggests that the DNA-restructuring activities of BLM alone can markedly influence IR processes in E. coli.

An antibiotic sensitivity screen of single-gene knockout E. coli strains showed that the ΔrecQ mutation selectively increased NIT sensitivity (29). We found that NIT exerts a toxic effect even on WT, recQ*, and recQ-dH cells with a very high apparent cooperativity (Fig. 3C and Table S4). Consistent with a previous study (29), we found significantly increased NIT sensitivity for ΔrecQ compared with WT. A similar increased NIT sensitivity was observed for recQ-dWH (Fig. 3C and Table S4).

The Spi^– λ phage assay allows quantification of illegitimate recombination frequencies (IRFs) (30, 31). Upon lytic phase induction, λ phage is normally excised from the genome of lysogenic strains by site-specific recombination. IR events produce phage variants that lack the red and gam genes and therefore, unlike the WT phage, are able to form plaques on P2 lysogenic E. coli strains (Spi^– phenotype). We found that, in the absence of UV irradiation, all mutant strains showed elevated IRF values compared with WT (Fig. 3D and Table S4). UV irradiation markedly increased the IRF in WT, consistent with previous findings (22, 31). The UV-induced IRF values of ΔrecQ, recQ-dWH, and recQ* were elevated compared with that of WT, whereas recQ-dH exhibited an even higher IRF value (Fig. 3D and Table S4). Reported differences are statistically significant (P < 0.05) based on two-tailed t test statistics. More stringent ANOVA analysis of log-transformed IRF values (performed based on log-normal distribution of IRF), followed by the Tukey post hoc test, demonstrated that the UV-free IRF values of ΔrecQ, recQ-dWH, and recQ-dH, and the UV-induced IRF of recQ* and recQ-dH were significantly higher than the corresponding WT values (P < 0.05) (Fig. 3D). These results show that in vivo IR suppression by RecQ is compromised by the abolition of activities conferred by the HRDC and WHD domains.

Discussion

Single-molecule and ensemble measurements revealed a striking DNA geometry-dependent unwinding behavior for E. coli RecQ (Figs. 1 and 2). Unwinding of gapped DNA, similar to DNA resection, was comparable for all RecQ constructs. Unwinding of hairpin DNA, similar in some respects to disruption of homologous paired DNA, was dramatically different for the RecQ constructs in which the HRDC domain was mutated or deleted. Unwinding behavior at the end of the DNA hairpin, although not obviously analogous to physiological DNA structures, provided additional mechanistic insights into the functions of the specific domains and their interactions. Based on these observations, we propose a mechanistic model in which shuttling and directed D-loop processing activities arise from an interplay between the dynamic interactions of the HRDC domain and the substrate preference of the WHD (Fig. 4).

Fig. 4.

DNA geometry-dependent activities of RecQ helicases and their contribution to recombination quality control. (A) Model illustrating key differences between DNA geometries (hairpin DNA, Left; gapped DNA, Right) and shuttling events occurring during unwinding and upon reaching the tip of the hairpin DNA substrate. During dsDNA unwinding (state 1), the WHD (blue oval) is bound to the dsDNA segment ahead of the progressing helicase core (green drop), whereas the HRDC domain (orange circle) can form transient interactions with the displaced ssDNA strand, thereby inducing shuttling on hairpin DNA (state 3). At the hairpin end (Figs. S1 and S9), additional shuttling modalities can be observed. On gapped DNA, HRDC–ssDNA interactions do not hinder the progression of the helicase core, which unwinds DNA processively without shuttling or pausing. (B) RecQ helicase preferentially binds to D-loops in an orientation favoring invasion disruption (left panels, configuration “DLI” in Fig. 2A, facilitated by HRDC–motor core interactions). During DNA unwinding, HRDC-dependent pausing and shuttling activities hinder the disruption of stable, long invasions occurring during legitimate recombination, thereby enabling HR progression (upper row). However, short homologous regions, characteristic of illegitimate recombination, are unwound rapidly, thus reducing the probability of HRDC-induced pausing and shuttling. This mechanism thus aids quality control by selective disruption of illegitimate invasions.

In the proposed model, the WHD acts as a processivity factor through its stabilizing interactions with dsDNA ahead of the progressing helicase core (state 1 in Fig. 4A). This role is consistent with the decrease in processivity and DNA binding that accompanies deletion of the WHD (RecQ-dWH; Fig. 1; Figs. S1, S2, and S5; Table 1; Tables S1 and S5). Accordingly, a WHD is present in many nucleic acid binding proteins in which it is typically implicated in binding ds nucleic acids (6). In line with the diversity of their cellular roles, RecQ family members show considerable variation regarding the presence and functional role of the WHD and HRDC domains (32). A β-hairpin within the WHD is an essential DNA unwinding wedge in human RECQ1 helicase (33). However, the corresponding region of E. coli RecQ is dispensable for unwinding activity (33), and the entire WHD of human BLM was found to be dispensable for DNA binding and unwinding (34). Our results with RecQ-dWH show that the WHD is dispensable for individual unwinding steps per se, but that it facilitates processive dsDNA unwinding (Figs. 1 and 2, Fig. S2, Table 1, and Table S1). Consistent with this interpretation, deletion of the WHD drastically reduced the affinity of RecQ for various DNA substrates (Figs. S5 and S7, Table 1, and Table S1).

Table S5.

Experimental evidence for models shown in Fig. S9

Element of model	Experimental result	Figures
HRDC-independent resetting (Fig. S9, pathway A)	Slow reannealing events in MT hairpin unwinding traces (RecQ, RecQ*, RecQ-dH)	Fig. 1C and Fig. S1
HRDC-tethered backsliding (Fig. S9, pathway B)	Limited-extent shuttling in MT hairpin unwinding traces (RecQ, not in RecQ*, RecQ-dH)	Fig. 1C and Fig. S1
Untethered sliding (Fig. S9, pathway C)	Long, rapid reannealing events in MT hairpin unwinding traces (RecQ, RecQ*, RecQ-dH)	Fig. 1C and Fig. S1
HRDC–ssDNA interaction (Fig. 4 and Fig. S9)	Reduced shuttling in gapped DNA unwinding traces, compared with hairpin DNA unwinding (largest effect in RecQ)	Fig. 1E and Figs. S1 and S3
	Isolated HRDC (but not HRDC*) binding to ssDNA in solution	Fig. S8
HRDC–motor core interaction (Fig. 4B and Fig. S9, pathway C)	ATPase suppression by HRDC domain, independent of HRDC–ssDNA interaction (RecQ, RecQ*; not in RecQ-dH, RecQ-dWH)	Fig. S5 B and D
	HRDC-dependent orientational preference for invasion disruption, independent of HRDC–DNA interaction (RecQ, RecQ*; not in RecQ-dH)	Fig. 2 B–D and F
WHD–dsDNA interaction (Fig. 4A and Fig. S9)	Decreased unwinding processivity of RecQ-dWH compared with WHD-containing constructs	Fig. 1C and Figs. S1A and S2
	Decreased D-loop disruption and unwinding efficiency of RecQ-dWH compared with WHD-containing constructs	Fig. 2 B–F
WHD–ssDNA interaction (Fig. S9, states A1, B1, C1)	Decreased ssDNA binding affinity and site size of RecQ-dWH compared with WHD-containing constructs	Figs. S5 and S7

During dsDNA unwinding (states 1–2 in Fig. 4A), the HRDC domain may interact transiently with the displaced ssDNA strand (arrows in state 1). This results in frequent and long pauses that are often accompanied by short-extent, long-duration shuttling of RecQ on hairpin DNA (state 3 in Fig. 4A). This behavior is pronounced for RecQ at the end of the DNA hairpin, along with other shuttling modalities (Figs. S1 and S9 for a possible mechanistic scheme for different shuttling modalities). Shuttling, and in particular short-extent, long-duration shuttling, was significantly reduced in the HRDC point (RecQ*) and deletion (RecQ-dH) mutants, and during gapped DNA unwinding by all constructs (Fig. 1, Fig. S1, and Table 1). Whereas the HRDC domain is dispensable for strong DNA binding (Figs. S2, S5, and S7, Table 1, and Tables S1 and S5), its transient interactions with ssDNA reduce the processivity and mean rate of hairpin DNA unwinding (Fig. 1, Fig. S2, and Table 1). In addition, the HRDC domain suppresses the ATPase activity of E. coli RecQ (Fig. S5 and Table 1), most likely through interactions with the motor core, as has been observed in recent BLM crystal structures (26, 35).

Fig. S9.

Possible mechanisms for different shuttling modalities performed by RecQ helicases at the tip of the hairpin DNA substrate (Fig. 1). Upon reaching the hairpin end (tip represented by red star, Fig. 4A), three modalities of shuttling activity were observed (Fig. S1). WT RecQ exhibited all three modalities, among which rapid and limited shuttling events were most frequent. In RecQ*, longer rapid reannealing events were most frequent. In the HRDC-deleted construct (RecQ-dH), shuttling occurs mostly through slow reannealing events (Fig. S1 and Table S1). Based on our observations, we suggest that these shuttling modalities are related to different mechanisms, governed by the interactions of the HRDC domain with ssDNA and/or with the helicase core. Below, we provide a possible explanation of these mechanisms that is consistent with experimental observations. All pathways (A, B, and C) start at the hairpin end (Fig. 4A) and include transient enzyme–ssDNA configurations (A1, B1, and C1) in which the WHD interacts with ssDNA ahead of the motor core. WHD binding to ssDNA in RecQ is supported by the ∼8-nt decrease in the occluded site size of RecQ-dWH on ssDNA (Fig. S6E and Table 1) and the decrease in ssDNA binding affinity of RecQ-dWH compared with RecQ constructs containing an intact WHD (Figs. S5C and S7, Table 1, and Table S1). A, HRDC-independent resetting: If the HRDC domain does not interact with either ssDNA or the motor core upon reaching the hairpin tip (state A1), the enzyme may rapidly dissociate from ssDNA (as observed frequently for HRDC-mutant constructs in Fig. 1C) or continue to translocate past the tip (state A2). Thus, reannealing of the hairpin can occur behind the translocating motor core, at a rate similar to the ssDNA translocation rate of the enzyme (state A2) (9, 47). These steps are observed as slow reannealing events occurring after full unwinding of the hairpin (Fig. 1C and Fig. S1). Slow reannealing events are frequently followed by unwinding, suggesting that the helicase core switches to the opposite ssDNA arm of the hairpin to reengage the reformed dsDNA segment (states A3–A5). These events were observed frequently for RecQ-dH and occasionally for RecQ, supporting the proposal that ssDNA translocation past the hairpin tip is favored in the absence of HRDC interactions. This HRDC-independent resetting process is equivalent to strand switching observed on dsDNA substrates. Strand switching may be driven by the inherent preference of the WHD for dsDNA, or possibly the ssDNA–dsDNA junction, over ssDNA. Dissociation of the WHD from ssDNA would preferentially lead to recapture of the reannealed dsDNA segment by the WHD (analogous to state A3). The resulting, likely unfavorable, configuration of the enzyme–DNA complex may trigger dissociation of the core from ssDNA (states A3–A4), followed by reestablishment of the original, favorable unwinding configuration (state A5, Fig. 4A). This leads to a new cycle of unwinding of the dsDNA adjacent to the end of the hairpin or, in vivo, in a locally base-paired DNA region. These resetting events are distinct from position-independent stochastic strand switching observed occasionally for gapped DNA substrates, as strand switching at the hairpin tip occurs usually after a fixed extent of reannealing, as shown by the results with RecQ-dH (Fig. S1). B, HRDC-tethered shuttling: If the HRDC domain binds to the displaced DNA strand when the helicase reaches the hairpin tip, translocation past the hairpin tip will be inhibited (state B1). The HRDC-tethered helicase may thus slide backward along—or dissociate and rebind to—the tracked ssDNA strand (state B2). The extent of HRDC-tethered backsliding is probably limited by the length and flexibility of the linker region connecting the WHD and HRDC domains. During backsliding, limited reannealing of the hairpin occurs before the helicase reengages. Through this process, the helicase can repetitively unwind the reformed hairpin (states B3–B4). Supporting this proposition, rapid short reannealing events at the hairpin tip, observed for RecQ, were markedly less frequent in RecQ* and RecQ-dH constructs (Fig. S1 and Table S1). This tethered shuttling was not limited to the end of DNA hairpin but also occurred throughout DNA hairpin unwinding by RecQ (Fig. 1C). Importantly, such behavior was absent in case of the gapped DNA substrate, supporting the proposed model in which the HRDC domain binds to the displaced DNA strand. Similar short repetitive unwinding was also observed for BLM previously in single-molecule FRET experiments during unwinding of a splayed arm dsDNA molecule and in magnetic-tweezers measurements of hairpin DNA unwinding (10, 11). This behavior was proposed to result from strand switching and backsliding of the helicase after limited unwinding (10, 11). The narrow distribution of the extent of the observed repetitive unwinding (10, 11) is readily explained by our HRDC-tethered backsliding model. C, Untethered sliding: We propose that, if the HRDC domain binds to the helicase core at the hairpin end (state C1), the helicase is prone to performing untethered sliding (state C2) on the tracked ssDNA strand, represented by longer rapid incomplete reannealing events. This pathway was observed for all constructs, but the relative frequency of untethered sliding was the highest for RecQ*. In RecQ*, the HRDC domain binds ssDNA with a markedly reduced affinity due to the point mutation (Fig. S8B). However, it is plausible that the HRDC (both WT and point mutant) interacts with the helicase core, as was recently found for BLM (26, 35). The dynamic interaction of the HRDC domain with the motor core suppresses the ATPase activity of both BLM (26) and RecQ (Fig. S5 B and D). Furthermore, recently, we showed that the presence of the HRDC domain, independent of its ssDNA binding ability, decreases the ssDNA translocation rate and processivity of RecQ (17). The preference of the HRDC domain for binding to the motor core—rather than to ssDNA—in the presence of pure ssDNA substrates is further supported by the findings that the abolition of the HRDC–ssDNA interaction (in RecQ*) or deletion of the entire domain (in RecQ-dH) resulted only in a moderate reduction in the DNA affinity of the enzymes, and did not affect their occluded site size on ssDNA (Fig. S5C, Table 1, and Table S1). Based on these considerations, it is possible that binding of the HRDC domain to the helicase core hinders processive ssDNA translocation. In this state, the enzyme may either rapidly dissociate from ssDNA, or the interaction between the helicase and the tracked DNA strand can be maintained through the WHD (as proposed in ref. 11 for human BLM) and/or the helicase core. Sliding toward the 3′ end of the tracked strand will result in rapid reannealing of the hairpin. The helicase can then reengage and repetitively unwind the hairpin (state C3). We note that our data do not exclude the possibility of forward sliding toward the 5′ end of the tracked strand. This would result in reannealing of the hairpin behind the enzyme, forming a state similar to state A2. Nonetheless, rapid reannealing longer than 50 bp followed by unwinding, particularly for RecQ-dH, was not considered as untethered sliding because these sliding events cannot be distinguished from the dissociation of the helicase at the junction followed by unwinding of a second helicase bound at a distal site on the ssDNA. Consistent with this scenario, the probability of rapid, large-extent (>50 bp) but incomplete reannealing followed by unwinding events increased with increasing RecQ-dH concentration, whereas shuttling duration was independent of RecQ-dH concentration (Fig. S12). Our models involving distinct shuttling modalities provide a broader context for previous results on the shuttling behavior of other RecQ helicases (9–11). The HRDC-tethered backsliding mechanism, which we propose based on RecQ domain engineering combined with analysis of DNA geometry-dependent shuttling behavior on hairpin and gapped DNA substrates (Figs. 1 and 4, and Figs. S1 and S3), readily explains the narrow distribution of the extent of repetitive unwinding events observed for HRDC domain-containing BLM (10, 11) and At RECQ2 (9) enzymes. Furthermore, our results indicate the role of the WHD in maintaining the interaction of the enzymes with DNA during resetting and backsliding events, implied by the behavior of a WHD point mutant BLM construct (11). Embracing all of the above observations, our model suggests how the WHD and HRDC domains can mediate DNA structure-dependent complex unwinding events leading to repeated unwinding of short dsDNA regions. The predicted domain movements and interactions can be tested in further studies. Experimental evidence for the models shown is summarized in Table S5.

In addition to inducing shuttling during dsDNA hairpin unwinding, we found that the HRDC domain increased the efficiency of D-loop disruption, that is, oriented unwinding of the invading ssDNA strand, even when its ssDNA binding capability was abolished (by the Y555A mutation in RecQ*) (Fig. 2, Figs. S5 and S7, Table 1, and Table S1). These results (Fig. 2F) suggest that efficient D-loop disruption is achieved through the initial binding of the helicase in an orientation that leads to unwinding of the invading strand (configuration DLI in Fig. 2A; Fig. 4B), rather than unwinding the intact dsDNA segments, which would promote invasion (configurations DLE and DLN in Fig. 2A). Our results showed that the isolated HRDC domain interacts with the branched region of D-loops by binding ssDNA rather than a structure specific interaction (Fig. S8). These findings raise the possibility that the HRDC–motor core interaction stabilizes the helicase in a configuration specific for invasion disruption (36) (Fig. 4B).

Taken together, our results show that the conserved RecQ domain architecture supports sensing of DNA geometry during DNA-restructuring reactions and effects complex responses to enzyme-induced changes in substrate geometry. Below, we propose how these mechanisms may support recombination quality control.

Bacterial and eukaryotic RecQ helicases participate in several DNA repair and replication restart pathways in which they perform distinct functions (5). In E. coli, RecQ functions as a canonical helicase unwinding dsDNA in the RecF HR pathway and during replication restart (37, 38). However, RecQ also disrupts D-loops and, in conjunction with topoisomerase III, resolves converging replication forks (8, 39). Consistent with this functional diversity, we found that E. coli strains lacking RecQ helicase activity (ΔrecQ, recQ-dWH) exhibited a modest decrease in DNA damage tolerance, whereas strains expressing RecQ helicase variants with altered or deleted HRDC domains (recQ*, recQ-dH) exhibited WT-like damage tolerance (Fig. 3 B and C, and Table S4).

A cardinal function of RecQ helicases is the suppression of potentially harmful IR, that is, recombination occurring between nonallelic DNA segments (1), which can be initiated from locally base-paired intermediates (31). RecQ-catalyzed disruption of these structures effectively suppresses IR (31). Accordingly, we found that ΔrecQ and recQ-dWH E. coli strains showed elevated IRF compared with WT (Fig. 3D). Strikingly, we found that replacing WT RecQ with the HRDC-deletion construct (recQ-dH strain) markedly elevated IRF to levels well above those in the ΔrecQ-null strain (Fig. 3D). This finding, combined with the fact that the recQ-dH strain shows a WT-like DNA damage repair phenotype (Fig. 3 A–C), indicates that the HRDC domain exerts control over the helicase core to promote disruption of IR despite the apparent tendency of the HRDC-less enzyme to enhance IR. Interestingly, the recQ* strain also shows higher IRF than WT (Fig. 3D) even though the efficiency of in vitro D-loop disruption by RecQ* was comparable to that of RecQ (Fig. 2F). This suggests that processive dsDNA unwinding facilitated by the WHD without the HRDC-induced pausing and shuttling appears to promote IR, likely due to the generation of ssDNA segments that can lead to illegitimate base pairing combined with the elevated probability of extending rather than disrupting the resulting short D-loop–like structures (Figs. 2 and 4B) (22). Conversely, the HRDC-mediated increase in the precision of HR events likely arises from the ability of RecQ to preferentially and specifically reverse ssDNA strand invasion of duplex DNA and disrupt D-loop–like structures (Figs. 2 and 4B), which are early recombination intermediates. In this scenario, RecQ preferentially disrupts illegitimate, D-loop–like structures that can form spontaneously between short homologous regions, similar to those in our D-loop–like structures that were preferentially disrupted by RecQ (Figs. 2 and 4B). Short homologous regions are unwound rapidly as HRDC-mediated shuttling is infrequent in this length range (Fig. 4B). In contrast, legitimate invasions are longer, more stable, and/or specifically stabilized by additional factors. During unwinding of long invasions, HRDC-mediated pausing and shuttling occur with a high probability and halt D-loop disruption for extended periods of time (shuttling duration in Table S1). This mechanism may create an opportunity for organization of protein complexes to extend the D-loop and promote HR (Fig. 4C). Thus, based on our model, both HRDC-mediated shuttling and directed D-loop disruption jointly contribute to quality control of HR via selective disruption of IR intermediates (Fig. 3D and Table S4). This model is supported by the marked increase in IRF for the recQ-dH E. coli strain expressing RecQ lacking the HRDC domain (Fig. 3D). The lack of the HRDC-mediated shuttling activity, together with the lack of directed disruption of strand invasions, renders RecQ-dH much less effective in the disruption of short base-paired intermediates, and may in fact promote strand invasion in vivo.

Taken together, these findings underscore the importance of the auxiliary DNA binding domains in sensing and response to DNA geometry during DNA-processing reactions. In addition to providing a mechanistic explanation of the observed behavior of RecQ, our model has broad implications for nucleic acid-based motor enzymes more generally. (i) The presence of a dsDNA-binding domain attached to the helicase core, that is, the WHD in RecQ enzymes, is sufficient to elicit a complex, repetitive unwinding behavior involving strand switching. Strand switching is a common feature among many SF1 and SF2 helicases containing at least one auxiliary dsDNA-binding domain (40–42). An additional ssDNA-binding element, that is, the HRDC domain, adds further complexity to repetitive unwinding via dynamic, multipartite interactions. (ii) The model implies large-scale reorientation of protein domains and/or changes in DNA conformation. Such changes have been demonstrated for several SF1 helicases (43, 44), and recent studies suggest that the mobility of the WH and HRDC domains is a general feature of RecQ helicases (26, 33, 35, 45, 46).

Materials and Methods

Magnetic-Tweezers Experiments.

The DNA hairpin substrate was attached to the flow cell surface and to a 2.8-µm magnetic bead via a 1-kbp dsDNA handle and 60 nt of poly-dT, respectively. Attachment was performed as follows. The 0.3 nM DNA was incubated with 32 ng of anti-digoxigenin in 50 µL of 1× PBS, pH 7.4 (Invitrogen), for 1 h at room temperature. This mixture was introduced into a sample cell coated with a low concentration of stuck beads and incubated overnight at 4 °C. Unbound DNA was washed out with 200 µL of wash buffer [PBS supplemented with 0.04% (vol/vol) Tween 20 and 0.3% (wt/vol) BSA]. Twenty microliters of 20× dilution of streptavidin-coated magnetic beads (MyOne; Invitrogen) were then introduced in wash buffer and allowed to tether for 1 h and washed with 1 mL of wash buffer. Once a proper DNA substrate was found based on contour length and extension change due to force-dependent dsDNA opening, the chamber was washed with 200 µL of RecQ buffer [30 mM Tris, pH 8, 50 mM NaCl, 5 mM MgCl₂, 0.3% (wt/vol) BSA, 0.04% (vol/vol) Tween 20, and 1 mM DTT]. After washing, RecQ constructs were added at a concentration of 50 pM (except for RecQ-dWH, 500 pM) in 200 µL of RecQ buffer supplemented with 1 mM ATP. Measurements with the DNA hairpin were performed at a constant force of 8 pN under which the hairpin did not open spontaneously. In the presence of RecQ helicases and saturating ATP (1 mM), unwinding activity was monitored in real time by tracking the 3D position of a tethered bead at 60 or 200 Hz using image analysis software, simultaneously tracking a stuck bead for drift correction. Conversion of the measured change in extension to the number of DNA base pairs unzipped was done based on the worm-like chain model of ssDNA (47). Measurements with the gapped DNA substrate were performed at a constant force of 15 pN. The change in extension associated with the transition from dsDNA to ssDNA provided a measure of helicase unwinding activity. Typical unwinding records contained multiple unwinding events defined as the opening of the DNA hairpin by the helicase (when the extension exceeded 3 SDs of the baseline Brownian noise measured when no unwinding occurs) to the full reannealing of DNA hairpin back to the baseline (Fig. 1C). The total number of events for individual enzymes are reported in Table 1. Trajectories of the bead extension as a function of time were analyzed by fitting with a t test-based step-finding algorithm to obtain the mean unwinding rate, the step unwinding rate between pauses, the pause positions, and the pause durations (48). The mean rate was calculated by linear fits ranging from the initial unwinding position to the maximum unwinding position before the helicase dissociated or the hairpin rezipped. The step rate corresponds to the unwinding rate between pauses obtained from fitting the trajectories with a step-finding algorithm.

DNA substrate preparation and additional controls for magnetic tweezers experiments can be found in SI Materials and Methods and Figs. S10–S12.

Fig. S10.

No unwinding activity is detected on intact duplex DNA at high RecQ-dH concentration. There was no observable change in DNA extension over ∼30 min when intact DNA (the gapped DNA construct without the 37-nt ssDNA gap) was used as a substrate with 1 nM RecQ-dH under otherwise identical conditions as in the gapped DNA unwinding measurements (Fig. 1 E and F, and Figs. S3 and S4).

Fig. S12.

Shuttling, sliding, and unwinding initiation rate as a function of RecQ-dH concentration. (A) The probability of rapid, large (>50 bp) but incomplete reannealing (left axis, blue dots) increases as a function of RecQ-dH concentration. This suggests that large incomplete reannealing events can arise from dissociation of the active RecQ-dH from the hairpin ssDNA–dsDNA junction in which the hairpin is prevented from completely reannealing due to an additional RecQ-dH molecule bound at a distal site on the ssDNA. The unwinding initiation frequency expressed in hertz (right axis, green squares) increases roughly linearly with increasing RecQ-dH concentration, but remains relatively low even at 500 pM RecQ. The low initiation frequency indicates that, over the entire examined concentration range, RecQ-dH binds the DNA substrate with a markedly longer wait time (>30 s) than the time required to fully unwind the 150-bp hairpin (∼3 s). Together, these two results suggest that the binding of multiple helicases acting in concert is rare, but that additional RecQ-dH enzymes can bind the distal end of the exposed ssDNA. Binding the distal end of the ssDNA is favored because this region is unwound first by the helicase and is therefore exposed for a greater amount of time during subsequent unwinding. (B) The average duration of shuttling events by RecQ-dH was independent of the concentration of the helicase. This result suggests that shuttling reflects the activity of a single RecQ-dH enzyme rather than the interaction of multiple RecQ-dH enzymes that dissociate and permit short-range reannealing up to the location of the next RecQ-dH molecule. If the latter scenario occurred, then the duration of shuttling events would be expected to increase as more RecQ-dH bound behind the lead enzyme. The constant duration of shuttling events over a 10-fold increase in RecQ-dH concentration suggests that shuttling reflects the dynamics of a single RecQ-dH molecule.

Solution Kinetic DNA Unwinding Experiments.

DNA substrates (30 nM; final reaction concentrations stated) were incubated with excess enzyme (100 nM, unless otherwise stated) at 4 °C for 10 min, and then at 37 °C for 3 min in Buffer H (30 mM Tris⋅HCl, pH 7.5, 100 mM KCl, 1 mM DTT, 50 µg/mL BSA, 20 mM creatine phosphate, 20 µg/mL creatine kinase). Reactions were started at 37 °C by mixing the DNA–enzyme complex with Buffer H containing ATP (3 mM), MgCl₂ (3 mM), and ssDNA trap strand (3 µM, unless otherwise stated) to inhibit enzyme rebinding to DNA. Reactions were stopped manually after preset reaction times by the addition of EDTA (40 mM final, pH 8.1) and loading dye [10 mM Tris⋅HCl, pH 7.5, 40 mM EDTA, 60% (vol/vol) glycerol, 0.075% (wt/vol) Orange G, 0.83% (wt/vol) SDS]. Mixtures were incubated at 37 °C for additional 3 min. Samples were then loaded on 12% (wt/vol) nondenaturing polyacrylamide gels in TBE buffer (89 mM Tris⋅HCl, pH 7.5, 89 mM boric acid, 20 mM EDTA). Electrophoresis was carried out at 4 °C. Cy3-labeled DNA was detected by using a Typhoon TRIO+ Variable Mode Imager (Amersham Biosciences). The intensities of bands corresponding to the DNA substrate and unwinding products were quantified by densitometry.

Cell Growth and Survival Assays.

Growth curves were recorded by inoculating aliquots of overnight E. coli cultures into Luria–Bertani (LB) medium to an OD₆₀₀ of 0.01, followed by incubation at 37 °C with aeration and monitoring their OD₆₀₀ value. In UV irradiation survival assays, cells were grown to 4 × 10⁸ cells per mL at 37 °C in LB medium. Aliquots of cultures were diluted 10⁶-fold, spread onto LB plates, and irradiated with different UV doses (at 254 nm) using an UVP CX-2000 cross-linker. Following irradiation and overnight growth at 37 °C, colonies were counted to quantify survival. NIT survival assays were performed as described for UV survival assays, except that UV irradiation was omitted and cells were spread and grown on LB plates containing different concentration of NIT. Dimethylformamide (used as solvent in NIT stock solutions) was applied in control plates at a concentration identical to NIT-containing plates (0.06%).

Spi^– Phage Assay for Illegitimate Recombination.

The frequency of illegitimate recombination of λ phage was measured using the λ Spi^– assay developed by Ikeda and coworkers (31). For lysogenization, WT and recQ mutant E. coli cells were grown to 7 × 10⁸ cells per mL at 30 °C in T-broth [1% bacto tryptone (Difco), 0.5% NaCl, 10 mM MgSO₄, 0.2% maltose]. λcI857 phage, isolated from HI1165 cells after heat shock induction of lytic phase, was added to WT and recQ mutant E. coli cells (at multiplicity of infection of 2) and incubated for 30 min at 30 °C. Cells were then spread onto T-plates [T-broth with 1.2% (wt/vol) agar] and incubated at 30 °C overnight. Single colonies were picked, amplified, and checked for lysogeny by heat shock induction of lytic phase [42 °C for 15 min, followed by shaking (250 rpm, Innova 44 shaker incubator, New Brunswick Co., Edison, NJ) at 37 °C and monitoring decrease in OD₆₀₀]. To analyze the formation of λ Spi^– phages, lysogenic strains (Table S3) were grown to 4 × 10⁸ cells per mL in λYP broth [1% bacto tryptone (Difco), 0.1% yeast extract (Oxoid), 0.25% NaCl, 0.15% NaHPO₄, 0.018% MgSO₄]. When applicable, 5 mL of the culture was irradiated at room temperature with UV light at a dose of 50 J/m² using a UVP CX-2000 cross-linker. To enter the lytic phase, λcI857 prophage was heat induced by incubation of the culture at 42 °C for 15 min. The culture was then incubated at 37 °C for 2 h with aeration and centrifuged to isolate the phages from cell debris. The total phage titer and the titer of λ Spi^– phages was determined by phage infection of YmeI and WL95 (P2 lysogen) E. coli cells, respectively (Table S3). Mixtures of (diluted) aliquots of the phage suspension and YmeI or WL95 cells were incubated at room temperature for 30 min, mixed with λ top agar (1% bacto tryptone, 0.5% NaCl, 0.4% agar; for YmeI) or λ trypticase top agar [1% trypticase peptone (Difco), 0.5% NaCl, 0.4% agar; for WL95] and spread onto 1.2% (wt/vol) agar plates (otherwise identical to the respective top agar). Plaques were counted after overnight incubation at 37 °C. The burst size, calculated by dividing the titer of total phages by the titer of infective centers, was in the range of 18–82. The IRF was obtained by dividing the titer of λ Spi^– phages by the titer of total phage.

Data Analysis.

Means ± SE values are reported in the paper, unless otherwise specified. Sample sizes (n) are given for (i) number of observed events in magnetic-tweezers experiments, (ii) number of ensemble in vitro measurements performed using independent protein preparations (biological replicates; n = 3, unless otherwise specified), (iii) number of experiments performed using independent bacterial colonies (biological replicates). Data analysis was performed using OriginLab 8.0 (Microcal) and GelQuant Pro software (DNR Bio Imaging). Magnetic-tweezers data traces were analyzed using a custom step-finding program written in Igor Pro (WaveMetrics) (49). Global-fitting kinetic analysis was performed using KinTek Global Kinetic Explorer 4.0, based on mass action rate equations accounting for all steps depicted in Fig. 2A, and the initial fractions of the DNA species shown at zero time in Fig. 2 B–E and Fig. S6E (50).

Further descriptions of materials and methods can be found in SI Materials and Methods.

SI Materials and Methods

Reagents.

All reagents were from Sigma-Aldrich, unless otherwise stated. ATP was from Roche Applied Science. For concentration determination, ε₂₆₀ values of 8,400 M⁻¹⋅cm⁻¹⋅nt⁻¹ and 10,300 M⁻¹⋅cm⁻¹⋅nt⁻¹ were used for oligo-dT and for nonhomopolymeric oligonucleotides, respectively. DNA concentrations are expressed as those of oligonucleotide or polynucleotide molecules (as opposed to those of constituent nucleotide units), unless otherwise stated.

Cloning, Protein Expression, and Purification.

Coding regions for RecQ-dH (comprising amino acids 1–523), RecQ-dWH (amino acids 1–414), and the HRDC domain (amino acids 523–609) were amplified by PCR using the pTXB3/RecQ (24) plasmid as template, and subcloned between the NcoI and SapI sites of pTXB3. The DNA constructs for RecQ* and HRDC* (harboring the Y555A point mutation) were created using QuikChange (Agilent Technologies). All constructs were verified by DNA sequencing.

RecQ was expressed and purified based on ref. 24. RecQ*, RecQ-dH, and RecQ-dWH were expressed and purified as was RecQ, with the following modifications. Fractions containing expressed protein were dialyzed against MonoQ buffer [50 mM Tris⋅HCl, pH 7.5, 1 mM DTT, 10% (vol/vol) glycerol] after elution from the chitin column. Samples were then loaded onto a MonoQ anion exchange column (GE Healthcare). Proteins were eluted by a linear gradient of 0–1 M NaCl in MonoQ buffer. Fractions containing the protein of interest were dialyzed against MonoQ buffer supplemented with 200 mM NaCl. Isolated HRDC and HRDC* domains were expressed and purified as RecQ constructs, with the following modifications. After elution from the chitin column, the sample containing the expressed HRDC domain was concentrated to 3 mL and further purified on a HiPrep 16/60 S-100 Sephacryl (GE Healthcare) gel filtration column equilibrated with a buffer containing 50 mM Tris⋅HCl, pH 8, 0.5 M NaCl, and 1 mM EDTA. Protein purity was checked by SDS/PAGE. Concentrations of purified proteins were measured using the Bradford method. The proper folding of RecQ constructs was verified by CD spectroscopy. Purified proteins were flash-frozen and stored in liquid N₂ in 20-µL droplets.

DNA Substrate Preparation.

For hairpin DNA substrate, a 174-bp DNA hairpin with a ∼1.0-kb DNA handle was generated first by PCR of pKZ1 plasmid between positions 4550 and 368 using forward primer (position 368) containing a BsaI restriction enzyme digestion site and reverse primer (position 4550) 5′-labeled with digoxigenin. pKZ1 was made by ligating a 44-bp dsDNA segment containing two BbvCI restriction sites separated by 37 nt into pET28b plasmid at the BamHI restriction site. PCR product was digested with restriction enzymes Nt.BbvCI (New England Biolabs) and BsaI-HF (New England Biolabs) for 16 h at 37 °C to create two nicks at positions 1093 and 1129, near the middle of the DNA template and the 5′ overhang at the end of dsDNA, respectively. After digestion, the 37-nt oligonucleotide was removed by heating the DNA to 80 °C for 20 min and cooling down gradually to 4 °C in the presence of excess complementary 37-nt oligonucleotide, thus producing a 37-nt ssDNA region. To complete the 174-bp DNA hairpin, the digestion product was ligated to 12-bp DNA with a dT₄ loop and a 5′ overhang complementary to that of the digestion product, and a 87-bp oligomer that consists of a 33-nt region that is complementary to the ssDNA region of the digestion product and 3′ biotin-labeled poly-dT.

For gapped DNA substrate comprising 2.5 kbp of dsDNA with a 37-nt gap, 2.5 kbp of dsDNA was generated by PCR of pKZ1 between positions 4550 and 1677 using 5′-biotin (position 1677)– and 5′-digoxigenin (position 4550)–labeled primers. The procedure to make a gap in the substrate between positions 1093 and 1129 was similar to that for the hairpin DNA substrate except that Nb.BbvCI was used to nick rather than Nt.BbvCI.

Both single-molecule DNA constructs contain a ssDNA region adjoining the DNA duplex, providing a specific high-affinity binding site for a single RecQ molecule (15). Nonetheless, it was recently demonstrated that, at high concentrations, in excess of 20 nM, RecQ can initiate unwinding on intact dsDNA (51). To ensure that the single-molecule measurements were insensitive to this secondary unwinding pathway, we verified that intact dsDNA was not unwound in our experimental configuration by 1 nM RecQ, a 20-fold higher concentration than that used in the single-molecule measurements (Fig. S10). The specific binding site allowed us to make measurements at low RecQ concentrations (50 pM) and, based on the concentration independence of the mean unwinding rate, appears to limit the binding of multiple interacting RecQ molecules up to a concentration of at least 500 pM (Fig. S11). Furthermore, the low unwinding initiation frequencies and the helicase concentration independence of the average duration of shuttling events reflected that the observed DNA-processing events dominantly resulted from the action of a single helicase molecule (Fig. S12).

Fig. S11.

Mean unwinding rate as a function of RecQ-dH concentration. The mean hairpin DNA unwinding rate was independent of RecQ-dH concentration over a 10-fold concentration range (50–500 pM RecQ-dH).

3T, DL3, and DL4 DNA substrates were generated as follows. Equimolar amounts of the applicable oligonucleotides (Table S2) were mixed in a buffer comprising 10 mM Tris⋅HCl, pH 7.5, and 50 mM NaCl. Samples were heated to 100 °C and were left to cool down to room temperature overnight. The annealed DNA mixtures were purified on a MonoQ anion exchange column using a 0.01–1 M NaCl gradient for elution. Eluted fractions were analyzed by PAGE. Fractions containing the desired DNA structures were desalted by using an Amicon Ultra centrifuge filter (Millipore). DNA substrates were aliquoted and stored at –80 °C.

Fluorescence Anisotropy Titrations.

These measurements were carried out in SF50 buffer (50 mM Tris⋅HCl, pH 7.5, 50 mM NaCl, 1 mM DTT, 5 mM MgCl₂, and 50 μg/mL BSA). Ten nanomolar ss54-FLU (3′–fluorescein-labeled ss54, Table S2), fluorescein-labeled DL4, DL3 (Table S2), or 3′-Cy3-dT₅₄ was titrated with increasing concentrations of protein at 37 °C (ss54-FLU, Fig. S7A) or 25 °C (Figs. S5A and S8). In case of competitive titration experiments (Fig. S7 B–D), Cy3-labeled DL4, DL3, or 3T substrates were used to compete with ss54-FLU binding to the given helicase construct, whereas ss54-FLU signal was monitored at 37 °C. Fluorescence anisotropy was measured in a Synergy H4 Hybrid Multi-Mode Microplate Reader (BioTek).

Intrinsic Protein Fluorescence Titrations.

These measurements were carried out in BSA-free SF50 buffer at 37 °C. One hundred nanomolar concentration of the given RecQ construct was titrated with 3T substrate (Table S2). Aromatic RecQ residues were excited at 280 nm (1-nm bandwidth), and the fluorescence emission spectrum (300–400 nm, 4-nm bandwidth) of each titration point was recorded in a SPEX Fluoromax spectrofluorometer. Spectra were corrected for the inner filter effect of DNA, which was determined by titration of 400 nM NATA (N-acetyl-l-tryptophanamide) with 3T.

ATPase Measurements.

Steady-state ATPase experiments were carried out in SF50 buffer by using a pyruvate kinase-lactate dehydrogenase (PK-LDH)-coupled assay (14 U/mL PK, 20 U/mL LDH, 1 mM ATP, 1 mM phosphoenol pyruvate, 200 μM NADH) at 25 °C. Time courses of NADH absorbance (ε₃₄₀ = 6,220 M⁻¹⋅cm⁻¹) were followed in a Shimadzu UV-2101PC spectrophotometer.

Bacterial Strains.

recQ mutant strains [derivatives of MG1655, denoted as wild type (WT)] were constructed using the Red/ET Recombination Kit (GeneBridges) (Table S3). MK1830 is a recQ knockout mutant (ΔrecQ). In place of the WT recQ gene, the MK555 strain possesses the Y555A point mutant (recQ*) gene. MK1569 and MK1239 strains possess recQ genes in which the regions 1569–1830 bp (recQ-dH) and 1239–1830 bp (recQ-dWH) have been deleted, respectively. These strains express RecQ proteins ending at amino acids 523 (lacking the HRDC domain) and 413 (lacking the WHD and the HRDC domain), respectively.

SI Equations

Eq. S1: Quadratic Equation Used for Analysis of the DNA Concentration Dependence of the Steady-State ATPase Activity (Fig. S5B).

$$\text{ATPase}\hspace{0.17em}\text{activity}=k_{\text{basal}}+k_{\text{cat}}\hspace{0.17em}\frac{c_{\text{E}}+n\cdot c_{\text{DNA}}+K_{\text{d,app}}-\sqrt{{\left(c_{\text{E}}+n\cdot c_{\text{DNA}}+K_{\text{d,app}}\right)}^2-4c_{\text{E}}\cdot n\cdot c_{\text{DNA}}}}{2c_{\text{E}}},$$ [S1]

where k_basal is the ATPase activity of the helicase in the absence of DNA, k_cat is the catalytic constant of the ATPase activity at saturating DNA concentration, c_E is enzyme concentration, n is the binding stoichiometry (moles of helicase/moles of DNA), c_DNA is the concentration of DNA, and K_d,app is the apparent dissociation constant of the helicase–DNA interaction during ATP hydrolysis. This equation is based on assuming equilibration of DNA-bound and DNA-free enzyme molecules during ATPase activity, characterized by a quadratic binding curve derived from the law of mass action as described in ref. 23.

Eq. S2: Equation Used for Analysis of the Oligo-dT Length Dependence of Steady-State ATPase K_d,app Values (Fig. S5C).

K_d,app values at each oligo-dT length were determined using quadratic fits (Eq. S1) to the DNA concentration dependence of steady-state ATPase activities (Fig. S5B). The standard change in Gibbs free energy (ΔG⁰) at each oligo-dT length (L) was calculated as ∆G⁰ (L) = RT ln K_d,app (L), where R is the universal gas constant, and T is the absolute temperature.

The equation below is based on the consideration that ΔG⁰ upon ssDNA binding to RecQ is proportional to the length of the ssDNA segment interacting with the enzyme. Thus, the absolute value of ΔG⁰ will linearly increase with oligo-dT length (L) until L reaches the occluded site size of the enzyme (b). At this length, specific interactions formed between the enzyme and ssDNA will determine ΔG⁰_spec. Another term (ΔG⁰_nonspec) is introduced to account for the ΔG⁰ value extrapolated to L = 0. The value of ΔG⁰ will remain constant in the regime where L > b:

$$\mathrm{\Delta }{G}^0\left(L\right)=\mathrm{\Delta }{G^0}_{\text{spec}}\frac{\left(L+\hspace{0.5em}b\right)-\left|L-\hspace{0.5em}b\right|}{2b}+\mathrm{\Delta }{G^0}_{\text{nonspec}}.$$ [S2]

Fitted values of b are listed in Table 1. ΔG⁰_spec values (in RT/nt units) were as follows: –13.6 ± 0.9, –13.0 ± 1.6, –10.0 ± 0.9, and –6.5 ± 1.5 for RecQ, RecQ*, RecQ-dH, and RecQ-dWH, respectively. ΔG⁰_nonspec values (in RT units) were –7.3 ± 0.9, –7.0 ± 1.6, –8.9 ± 0.8, and –7.8 ± 1.5, for RecQ, RecQ*, RecQ-dH, and RecQ-dWH, respectively.

Eq. S3: Equation Used for Analysis of the Oligo-dT Length Dependence of Steady-State ATPase k_cat Values (Fig. S5D).

This equation is based on considerations described in our earlier work (23, 24). Briefly, the overall steady-state catalytic constant of ATP hydrolysis (k_cat) will be determined by a weighted average of the ATP hydrolysis rate constant during translocation (k_ATP,trans) and that at the 5′ end of the DNA substrate (k_ATP,end). Other parameters influencing k_cat are the length of the ssDNA substrate (L, expressed in nucleotides), the occluded site size of the protein on ssDNA (b, expressed in nucleotides), the macroscopic coupling between ATP hydrolysis and translocation (s, expressed in nucleotides translocated per ATP hydrolyzed), and the rate constant of enzyme dissociation from the 5′-end of ssDNA (k_off,end):

$$k_{\text{cat}}={(\frac{\frac{L-b}{2}\cdot \frac{1}{k_{\text{ATP},\text{trans}}}\hspace{0.17em}+\hspace{0.17em}\frac{k_{\text{ATP},\text{end}}}{k_{\text{off},\text{end}}}\cdot \frac{1}{k_{\text{ATP},\text{end}}}}{\frac{L-b}{2s}\hspace{0.17em}+\hspace{0.17em}\frac{k_{\text{ATP},\text{end}}}{k_{\text{off},\text{end}}}}\hspace{0.17em})}^{-1}.$$ [S3]

Eqs. S4 and S5: Equations Used for Analysis of Competitive Fluorescence Anisotropy Titrations (Fig. S7 B–D).

In these experiments, the observed anisotropy (A) of ss54-FLU will be determined by the following:

$$A=f\cdot A_{\mathrm{helicase}\mathrm{.ss}54}+\left(1-f\right)\cdot A_{\mathrm{ss}54},$$ [S4]

where A_{helicase.ss54} and A_ss54 are the anisotropy values of the helicase.ss54-FLU complex and that of free ss54-FLU, respectively; and f is the [helicase.ss54-FLU]/[ss54-FLU]_total ratio.

According to ref. 52, the concentration of the helicase.ss54-FLU complex at a given total competitor (3T, DL3, or DL4) concentration ([competitor]_tot) will thus be the following:

$$\left[\mathrm{helicase}\mathrm{.ss}54\right]=\frac{\left[\mathrm{ss}54\right]\cdot (2\cdot \sqrt{a^2-3b}\cdot \cos \left(\mathrm{\Theta }/3\right)-a)}{3\cdot K_{\mathrm{ss}54}+\left(2\cdot \sqrt{a^2-3b}\cdot \cos \left(\mathrm{\Theta }/3\right)-a\right)},$$ [S5]

where

$$a=K_{\text{ss}54}+\hspace{0.5em}{K}_{\text{competitor}}+\hspace{0.5em}{\left[\mathrm{ss}54\right]}_{\text{tot}}+{\left[\mathrm{competitor}\right]}_{\text{tot}}-{\left[\mathrm{helicase}\right]}_{\text{tot}},b=K_{\text{ss}54}\cdot \hspace{0.5em}{K}_{\text{competitor}}+\hspace{0.17em}\hspace{0.5em}{K}_{\text{ss}54}\cdot \left({\left[\mathrm{competitor}\right]}_{\text{tot}}-\hspace{0.5em}{\left[\mathrm{helicase}\right]}_{\text{tot}}\right)+\hspace{0.17em}{K}_{\text{competitor}}\cdot \left({\left[\mathrm{ss}54\right]}_{\text{tot}}-\hspace{0.5em}{\left[\mathrm{helicase}\right]}_{\text{tot}}\right),\mathrm{\Theta }\hspace{0.5em}=\hspace{0.5em}\arccos \hspace{0.17em}\frac{-2\cdot \hspace{0.5em}{a}^3+\hspace{0.5em}9\cdot a\cdot \hspace{0.5em}b-27\cdot c\hspace{0.17em}}{2\cdot \sqrt{{\left(a^2-\hspace{0.17em}\hspace{0.5em}3\cdot b\right)}^3}},$$

$$c\hspace{0.17em}\hspace{0.5em}=\hspace{0.17em}-{\left[\mathrm{helicase}\right]}_{\text{tot}}\cdot K_{\text{ss}54}\cdot K_{\text{competitor}},$$

K_ss54 is the dissociation constant of the helicase–ss54-FLU interaction, and K_competitor is the dissociation constant of the helicase–competitor interaction. To reduce the number of floating parameters, A_ss54, A_{helicase.ss54}, and K_ss54 were determined in separate experimental sets and were fixed during fitting.

Eq. S6: Equation Used to Fit NIT and UV Dose-Dependent Survival Experiments (Fig. 3 B and C).

$$\text{Relative}\hspace{0.5em}\text{survival}=A_0+\frac{\hspace{0.5em}{A}_{\max }-A_0}{1+{10}^{\left(\log \hspace{0.5em}\mathrm{L}{\mathrm{D}}_{50}\hspace{0.5em}-\hspace{0.5em}x\right)\cdot p}}.$$ [S6]

To determine the effective concentration of NIT and the effective UV dose that decreases relative survival to 50% (LD₅₀), we fitted our data using a standard dose–response model. A₀ and A_max represent relative survival values at infinite and zero effector doses, respectively. x represents the log-transformed dose. p is the Hill slope coefficient, which reflects the steepness of the dose–response curve. (Absolute values of p are listed in Table S4.)