Extreme mechanical diversity of human telomeric DNArevealed by fluorescence-force spectroscopyJaba Mitraa, Monika A. Makurathb,c, Thuy T. M. Ngob, Alice Troitskaiab, Yann R. Chemlab,d,e, and Taekjip Had,e,f,g,h,i,1

aDepartment of Materials Science and Engineering, University of Illinois at Urbana–Champaign, Urbana, IL 61801; bCenter for Biophysics andComputational Biology, University of Illinois at Urbana–Champaign, Urbana, IL 61801; cDepartment of Molecular and Integrative Physiology, University ofIllinois at Urbana–Champaign, Urbana, IL 61801; dDepartment of Physics, University of Illinois at Urbana–Champaign, Urbana, IL 61801; eCenter for thePhysics of Living Cells, University of Illinois at Urbana–Champaign, Urbana, IL 61801; fDepartment of Biophysics, Johns Hopkins University, Baltimore, MD21218; gDepartment of Biophysics and Biophysical Chemistry, Johns Hopkins University, Baltimore, MD 21218; hDepartment of Biomedical Engineering,Johns Hopkins University, Baltimore, MD 21218; and iHoward Hughes Medical Institute, Johns Hopkins University, Baltimore, MD 21218

Edited by Steven M. Block, Stanford University, Stanford, CA, and approved March 12, 2019 (received for review September 16, 2018)

G-quadruplexes (GQs) can adopt diverse structures and arefunctionally implicated in transcription, replication, translation,and maintenance of telomere. Their conformational diversity underphysiological levels of mechanical stress, however, is poorly un-derstood. We used single-molecule fluorescence-force spectroscopythat combines fluorescence resonance energy transfer with opticaltweezers to measure human telomeric sequences under tension.Abrupt GQ unfolding with K+ in solution occurred at as many asfour discrete levels of force. Added to an ultrastable state and agradually unfolding state, there were six mechanically distinct struc-tures. Extreme mechanical diversity was also observed with Na+,although GQs were mechanically weaker. Our ability to detect smallconformational changes at low forces enabled the determination ofrefolding forces of about 2 pN. Refolding was rapid and stochasti-cally redistributed molecules to mechanically distinct states. A singleguanine-to-thymine substitution mutant required much higher ionconcentrations to display GQ-like unfolding and refolded via inter-mediates, contrary to the wild type. Contradicting an earlier pro-posal, truncation to three hexanucleotide repeats resulted in asingle-stranded DNA-like mechanical behavior under all conditions,indicating that at least four repeats are required to form mechan-ically stable structures.

G-quadruplex | single-molecule biophysics | fluorescence resonance energytransfer | optical tweezers

Certain guanine-rich DNA and RNAs self-associate and foldinto intra- or intermolecular four-stranded structures called

G-quadruplexes (GQs) (1–3). GQ-forming sequences are abun-dant in the genome, especially in telomeres, mitotic and mei-otic double-stranded break sites, promoters, and transcriptionalstart sites, where they modulate DNA replication and tran-scription, recombination, and other cellular processes (4, 5). Forexample, human telomeres have 100- to 200-nt-long repeats ofd[TTAGGG] at the 3′ end (6). Owing to incomplete replication,the hexanucleotide repeats erode during mitosis, culminating incellular aging and apoptosis (7). Malignant cells exhibit highlevels of telomerase activity, whereby they can circumventtelomere-programmed cell death (8). The telomerase activity invitro is modulated by the GQs (9, 10), and GQ-stabilizing agentsinterfere with telomere homeostasis and induce growth arrestand apoptosis in cultured cells (11). Hence, GQs are widelystudied as potential cancer therapeutic targets (12) and un-derstanding the structural diversity and conformational dynamicsof GQs is desirable. In addition, because forces on the order ofpiconewtons are generated during transcription and replication(13), understanding the GQ’s dynamics under tension is neces-sary for comprehending its in vivo conformations and responsesto enzymes that act on DNA in general and telomeric DNA inparticular (14, 15).The basic unit of a GQ is a G-quartet formed via cyclic

Hoogsten hydrogen bonding between four guanines. The planarquartets stack up on top of each other to form GQs. GQs are

highly polymorphic, and at least six different structures (parallel,antiparallel, hybrid, etc.; SI Appendix, Fig. S1) have been reported,differing in the relative orientation of the DNA strands connectingstrings of Gs. GQ structures are modulated by the type ofmonovalent cation and show lower thermodynamic stability inNa+ compared with K+ (16, 17).We now know detailed biophysical and structural properties of

GQs thanks to tools such as NMR, CD, and X-ray crystallog-raphy. Although powerful, these methods have not yet been ableto resolve multiple coexisting GQ structures. Single-moleculestudies are advantageous because of the ability to characterizeindividual subpopulations in a heterogeneous sample (18–21).For example, an early smFRET (single-molecule FRET) study ofhuman telomeric GQ molecules revealed four folded states thatdiffer in FRET efficiencies or in lifetimes and two unfolded statesof different lifetimes (22), and GQ conformational diversity in theabsence of mechanical stress has been well documented in othersmFRET studies (23–30). Force-based single-molecule measure-ments showed that human telomeric GQmolecules in 100 mMK+

unfold at ∼20 pN (31) with a hint of three folded states that differin lifetimes (32). Thus far, purely mechanical studies have notexamined GQ dynamics in Na+, and the only available data onforce-dependent GQ dynamics in Na+ are from Long et al. (33),who used magnetic tweezers combined with smFRET to probe

Significance

G-quadruplex (GQ)–forming G4 DNA, made of four repeats ofguanines, is implicated in many important processes. TelomericGQs are being studied extensively as anticancer drug targets.Because many nucleic acid-processing enzymes can exertpiconewton forces and can be functionally regulated by ten-sion on DNA, we used a single-molecule instrument combiningfluorescence with mechanical manipulation to study the dy-namics of human telomeric DNA under physiological levels oftension. We observed extreme mechanical diversity with atleast six different interconverting species that differ in me-chanical responses. Such extreme mechanical diversity mayinduce differential interactions with helicases and polymerasesand modulate processes such as replication and transcription.Our comparative analysis further revealed the effects of se-quence changes and ionic conditions.

Author contributions: J.M., T.T.M.N., and T.H. designed research; J.M., M.A.M., and A.T.performed research; Y.R.C. contributed new reagents/analytic tools; J.M. and M.A.M.analyzed data; and J.M., M.A.M., Y.R.C., and T.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence should be addressed. Email: tjha@jhu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1815162116/-/DCSupplemental.

Published online April 3, 2019.

8350–8359 | PNAS | April 23, 2019 | vol. 116 | no. 17 www.pnas.org/cgi/doi/10.1073/pnas.1815162116

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

GQ folding/unfolding kinetics in the force range of 1 and 8 pN.Therefore, up until now there is no side-by-side comparison ofmechanical behaviors of GQs in Na+ vs. K+.In this study, we performed fluorescence-force spectroscopy

(34–36) by applying tension to a human telomeric GQ using anoptical trap and probing its conformational transitions viasmFRET. The previous fluorescence-force analysis by Long et al.(33) was performed under constant tension, precluding analysisin K+. As we show below, the telomeric GQ in K+ displays alarge hysteresis such that at a constant force one cannot fre-quently observe folding and unfolding transitions. By varying theforce through multiple cycles of stretching and relaxation, wecould directly detect GQ unfolding and folding reactions in K+

solutions. Complete unfolding of GQs in 100 mM K+ concen-tration occurred stochastically over a wide range of force, from2 to 25 pN, and with some remaining folded with up to 28 pN offorce. GQs in Na+ showed an overall similar diversity but weremechanically weaker. A single-point mutant that replaced aguanine in the central quartet and a poly-dT construct furthervalidated our findings. It has been proposed that GQs areformed via intermediate G-triplex structures, involving three ofthe four G-rich hexanucleotide repeats (37–39). However,truncation to three repeats revealed ssDNA-like mechanicalbehavior under all conditions tested, suggesting that at least fourstretches of guanines are necessary for the formation of GQ-likestructures that can resist forces above 1 pN.

ResultssmFRET Analysis of Human Telomeric DNA at Zero Force. Our con-struct, hTel22, included a 22-nt human telomeric DNA sequence[GGG(TTAGGG)3T] flanked by duplex DNA stems serving as atether to the surface (5′ side) and the λ-DNA (3′ side) (Fig. 1B).CD spectroscopy of the 22-nt sequence in 100 mM K+ showed astrong positive peak at ∼290 nm with a shoulder at ∼265 nm anda weaker negative peak at ∼235 nm, consistent with hybrid-typeGQs (Fig. 1A and SI Appendix, Fig. S2A) (40). In contrast, CDspectrum in 100 mM Na+ showed evidence of basket-type anti-parallel GQs with positive peaks at ∼295 nm and 246 nm and anegative peak at ∼265 nm (SI Appendix, Fig. S2B) (41). Tominimize the potential influence of an adjacent duplex stem onGQ folding, we introduced (dT)17 between the telomeric se-quence and the 18 nt on the 3′ side complementary to the 3′ endof the λ-DNA bridge (Fig. 1B). We measured the zero-forcedistribution of FRET efficiency E between a donor (Cy3) andan acceptor (Cy5) placed adjacent to the ends of the GQ-forming sequence over 0 to 100 mM K+ or Na+ concentrations(Fig. 1C and SI Appendix, Fig. S3A). The peak at E = 0 repre-sents molecules with inactive or missing acceptors. Withoutadded cations, we observed a peak at E ∼ 0.28. An additionalpeak at E ∼ 0.81 appeared at 2 mM K+, and it increased inpopulation at the expense of the low E population with in-creasing K+ concentration. In addition, at higher ionic concen-trations, the low FRET peak shifted to higher E values due to thescreening of electrostatic repulsion between backbone phos-phates (42). We assigned the populations with low and high Evalues to unfolded and folded GQ conformations, respectively(22). A similar behavior was observed with Na+ but higher Na+

concentrations were required to induce the same effect (SI Ap-pendix, Fig. S3A), likely due to the inferior coordination by thesmaller Na+ compared with K+ (17). These observations areconsistent with the previously reported behavior of fluorescentlylabeled human telomeric DNA in smFRET measurements (43),suggesting that the duplex DNA stems built into our constructfor optical trapping do not perturb GQ folding significantly. In100 mM K+ or Na+, these different states are stable becausesingle-molecule time traces show no transitions within our tem-poral resolution of 30 ms during our observation time up to 150 s(Fig. 1D and SI Appendix, Fig. S3B).

Conformational Dynamics of hTel22 Under Tension. For fluorescence-force spectroscopy, we pulled the surface-tethered hTel22 con-struct through a λ-DNA tethered to the other end using an opti-cally trapped bead (Fig. 2A) (34, 36). We applied force bytranslating the microscope stage at a speed of 455 nm/s, typicallyincreasing from ∼0.3 pN to ∼28 pN in ∼6.5 s (average loading rate∼4.5 pN/s), followed by relaxation at the same speed while fluo-rescence intensities of the donor and the acceptor were recorded.Unfolding and refolding were characterized by E changes (Fig.

2B and SI Appendix, Fig. S4A). In 100 mM K+, most molecules ina folded state showed stable E values of ∼0.88 up to a certainforce, followed by a sudden drop in E to ∼0.15 due to unfolding(Fig. 2 B and F). Further force increases resulted in gradualincreases in the extension of the unfolded hTel22, as evidencedby a gradual decrease in E from ∼0.15 to ∼0.05. Refolding wasdetected as a sudden increase in E during relaxation (Fig. 2 Band F). We also observed an ultrastable state in 39% (150 out of381) of stretching–relaxation cycles where no change in E couldbe detected on application of force up to ∼28 pN, and somemolecules transitioned into and out of an ultrastable state be-tween stretching/relaxation cycles [Fig. 2 B and C (cycle 3), Fig. 2F, v, and SI Appendix, Fig. S4A]. In addition, some stretchingcycles (11%, 43 cycles) could be characterized by gradual yetincomplete unfolding whereby a final mid-E value ∼0.4 wasreached (Fig. 2 F, vi and SI Appendix, Fig. S4 A, 2 and 3). Thecorresponding refolding events retraced the unfolding pathway.

Fig. 1. Conformational analysis of human telomeric repeats (hTel22). (A) CDspectrum in 100 mM K+. (B) Schematic diagram of hTel22 construct forsmFRET. The 5′ extension to the 22-nt-long human telomeric repeat GGG(TTAGGG)3T was annealed to a 18-nt-long biotinylated strand and immo-bilized on a PEG-passivated quartz surface through biotin–neutravidin in-teraction. The 5′ end of the biotinylated strand is labeled with Cy5(acceptor). The main G4 strand is labeled with Cy3 (donor) at the 3′ end of thetelomeric repeat and is followed by (dT)17 and an 18-nt extension that isannealed to a 30-nt-long λ-bridge. (C) E histograms as a function of K+ con-centration. (D) Representative single-molecule time traces of donor and ac-ceptor intensities and corresponding E in 100 mM K+ (30-ms integration time).

Mitra et al. PNAS | April 23, 2019 | vol. 116 | no. 17 | 8351

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

We defined the unfolding force, funfold, as the force corre-sponding to the transition midpoint, and similarly for therefolding force, frefold, during relaxation. frefold, typically 1 to 5 pN(Fig. 2 E and F and SI Appendix, Fig. S4C), was in most caseslower than funfold, typically 2 to 28 pN (Fig. 2 D and F and SIAppendix, Fig. S4C), and such hysteresis suggests nonequilibriummeasurement conditions. Subsequent stretching/relaxation cyclesyielded similar initial and final E values yet different funfold andfrefold (Fig. 2B and SI Appendix, Fig. S4A). Fig. 2C and SI Ap-pendix, Fig. S4B illustrate cycle-to-cycle variability in funfold forthe same single DNA molecule (∼2.5, 17, and 25 pN in cycles 1,2, and 4, respectively, in Fig. 2C and ∼25, 1.5, and 17 pN in cycles1, 2, and 3, respectively, in SI Appendix, Fig. S4B). Therefore, anunfolded molecule can refold to a state of different mechanicalstability, suggesting that the diversity in mechanical stability isnot due to chemical differences, for example caused by DNAsynthesis errors or damage.Interestingly, funfold histogram clustered around four force

values (∼3, 11, 18, and 25 pN, n = 188; Fig. 2D). The Dudko–Szabo model of rupture force distribution was used to fit the

funfold histogram to assign the average force value of each cluster(44, 45). Including the ultrastable state and the state that showedgradual, partial unfolding, our data suggest that there are at leastsix mechanically different G4 conformations that are indistin-guishable based on the zero-force E value. Fig. 2 F, i–vi showsexamples of these six different mechanical species.In all, force-induced hTel22 unfolding was predominantly

characterized by an abrupt decrease in E which is likely due tocooperative disruption of a GQ structure. To further test thisinterpretation, we measured (dT)22, which has the same length ashTel22 but is unstructured (42). In the absence of force, weobserved a single peak at E ∼ 0.5 in 100 mM K+ (Fig. 2G, Inset).E values decreased gradually upon stretching and increased uponsubsequent relaxation, retracing the same E vs. force curve (Fig.2G). The stark contrast in behaviors of hTel22 and (dT)22 isanother piece of evidence that the abrupt E changes are due tothe folding and unfolding of GQ structures. We also observedsimilar E vs. force characteristics for the minor population withE ∼ 0.5 of hTel22 in 100 mM K+ (Fig. 2H), suggesting that itrepresents an unfolded population.

Fig. 2. Conformational dynamics of hTel22 under tension in 100 mM K+. (A) The G4 strand is annealed to a biotinylated strand and immobilized on aneutravidin-coated quartz surface. The other end is connected to a 1-μm-diameter, optically trapped bead through a λ-DNA. The G4 construct was stretchedto ∼28 pN in 6.5 s. FRET was measured between Cy3 (donor) and Cy5 (acceptor) as a function of force. (Inset) Unfolding (black) and refolding (red). (B) Arepresentative single-molecule time trace (20-ms integration time) of donor and acceptor intensities and corresponding E. Stretching and relaxation occurredat a stage speed of 455 nm/s. (C) funfold and frefold in different stretching cycles from a single molecule shown in B. (D and E) Distributions of funfold (D) andfrefold (E). The red and black curves in D, respectively, denote individual and overall rupture force distributions predicted using the Dudko–Szabo model (44,45). Only those cycles showing complete unfolding were included (n = 188). (F) E vs. force example curves, one representative from each of the four peaks infunfold distribution (i–iv), one representative of a ultrastable state (v), and one representative of gradual partial unfolding/folding without hysteresis (vi). (Gand H) E vs. force curves of (dT)22 (G) and hTel22 midE population (H). Corresponding E histograms are shown in insets. Error bars represent SEs.

8352 | www.pnas.org/cgi/doi/10.1073/pnas.1815162116 Mitra et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

Inherent Mechanical Heterogeneity Confirmed Using Dual OpticalTraps. To further corroborate the extreme diversity in mechanicalstability of GQ observed from fluorescence-force spectroscopy, weperformed purely mechanical measurements of the same 22-nt se-quence using dual optical traps (Fig. 3A). We stretched the DNA to∼30 pN over ∼3.5 s (average loading rate ∼8.5 pN/s) by moving oneof the trapped beads away from the other at the speed of 100 nm/s,followed by relaxation at the same speed, and detected unfoldingevents as a sudden force drop in the force-extension curve (Fig. 3Band SI Appendix, Fig. S5A). Fig. 3B shows four consecutivestretching and relaxation cycles of the same molecule in 100 mMK+.Some unfolding events could not be directly detected but force-extension curves showed that unfolding did occur [Fig. 3B (cycle 3)and SI Appendix, Fig. S5B], likely because unfolding at forces <3 pN(51 of 345 cycles) are undetectable due to noise. An ultrastable statewas also observed where unfolding could not be detected even at themaximum applied force (SI Appendix, Fig. S5B). funfold for such anevent was recorded as the maximum applied force.We observed a wide range of funfold values across multiple

stretching cycles of a given molecule (Fig. 3C), and this was true formost of the 71 molecules we examined (Fig. 3D). funfold broadlyranged from <3 pN to above 30 pN, and its histogram suggests forceclusters at ∼11, 22, 29, and 36 pN (Fig. 3E). Thus, we could re-capitulate the funfold clusters at ∼11, 18, and 25 pN, observed with the

fluorescence-force spectroscopy assay. Most refolding occurred atforces <3 pN (335 of 345) and could not be directly detected with theassay (Fig. 3F), highlighting the superior ability of fluorescence-forcespectroscopy to detect small conformational changes at low forces.Coefficient of variation of funfold calculated from within each

single molecule and averaged over the entire population is similar tothat calculated from the entire population (345 unfolding eventsfrom 71 molecules), indicating that the heterogeneous unfoldingproperties are intrinsic (Fig. 3H). Indeed, funfold values in consecutivestretching cycles from a single molecule showed weak or no corre-lation (correlation coefficient, r = 0.18), further indicating that uponunfolding a GQ molecule loses memory of its prior conformation(Fig. 3G). Additionally, ∼9% (36 events) of the pulling cyclesrecorded lacked a distinct force rip but showed gradual transitions toand from the unfolded states upon stretching and relaxation, re-spectively (SI Appendix, Fig. S5B). Such events may correspond tothe gradual and incomplete unfolding observed in fluorescence-force spectroscopy. Overall, GQ unfolding using the dual trap as-say showed a wide spectrum of mechanical stability, substantiatingour observations with fluorescence-force spectroscopy.

Mechanical Diversity Across Different Ionic Conditions. Fluorescence-force spectroscopy data acquired at a lower concentration of60 mM K+ also displayed mechanical diversity (SI Appendix, Fig.

Fig. 3. Mechanical unfolding of hTel22 in 100 mM K+. (A) Dual optical trap experimental scheme. The G4 strand is sandwiched between two 1.5- and 1.7-kb-long dsDNA spacers to the left and right, respectively. A (dT17) sequence is inserted between the spacers and the GQ-forming sequence to minimize anyinfluence of base stacking on mechanical unfolding. The spacers are linked to beads (diameter ∼1 μm) via digoxigenin–anti-digoxigenin and biotin–streptavidinlinkages, respectively. The G4 construct was stretched to ∼30 pN in 3.5 s. (B) Representative force vs. extension response from a single molecule in four consecutivestretching/relaxation cycles (100 mM K+). Black and red curves represent GQ unfolding and refolding, respectively. The folded and unfolded WLC fits are re-spectively denoted by black and red dashed lines. Black and red arrows indicate unfolding and refolding events, respectively. (C) funfold vs. cycle number for themolecule shown in B. (D) funfold values obtained from 345 cycles from 71 molecules are widely distributed within each molecule. Low force unfolding below thedetection limit of ∼3 pN, marked with black dashed line, is shown as a gray circle at 0 pN. The red dashed lines indicate the force limit (∼28 pN) achievable withthe fluorescence-force spectroscopy assay. (E and F) Distributions of funfold (D) and frefold (E) for 345 cycles from 71 molecules. Only those cycles showing completeunfolding were included (n = 277). The red and black curves, respectively, denote individual and overall the rupture force distributions predicted using theDudko–Szabo model (44, 45). (G) funfold across consecutive pulling cycles shows weak or no correlation. (H) Coefficient of variation for funfold calculated for theentire data set and within individual molecules (averaged over all molecules).

Mitra et al. PNAS | April 23, 2019 | vol. 116 | no. 17 | 8353

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

S6 A and B) although with slightly reduced mechanical stability [seeslight shift of funfold (SI Appendix, Fig. S6C) and frefold (SI Appendix,Fig. S6D) distributions to lower values compared with 100 mMK+].Sixty-two percent (97 out of 156 cycles) showed abrupt unfoldingvia transitions from E ∼ 0.85 to E ∼ 0.15, 29% (45 cycles) showedultrastabilty, and 8% (14 cycles) showed gradual partial unfolding.We also observed at least four apparent peaks in funfold distributionat ∼3, 10, 17, and 23 pN (SI Appendix, Fig. S6C).Next, we examined hTel22 responses to force in 100 mM Na+.

Without force, we observed a major population centered at E ∼0.87 (Fig. 4A). Unfolding occurred over a wide range of forces(Fig. 4B and SI Appendix, Fig. S8), and funfold varied betweenstretching cycles within the same DNA molecule (Fig. 4B). Bothdistributions of funfold (Fig. 4C) and frefold (Fig. 4D) shifted to lowervalues compared with those in 100 mM K+, indicating G4 struc-tures are mechanically weaker in Na+. Ultrastable states were alsoobserved but with a reduced probability of 25% (53 out of251 cycles) compared with 39% in 100 mM K+ and 29% in 60 mMK+. We observed three distinct peaks in funfold histogram, sug-gesting high mechanical diversity also in Na+ solution (Fig. 4C).We did not observe gradual partial unfolding in Na+ solution.The same analysis performed in 10 mM Na+, close to the mam-

malian intracellular concentration of Na+ (46), showed a qualita-tively different behavior even though CD spectra suggested theformation of basket-type antiparallel GQs in both 10 mM and100 mM Na+ solutions (SI Appendix, Fig. S2B) (41). In the absenceof force, hTel22 predominantly adopted an unfolded confirmation,with a peak centered at E ∼ 0.33, but we also observed a secondminor population with a peak at E ∼ 0.77, suggestive of a foldedstate (Fig. 4E) that interconverts with the unfolded population (SIAppendix, Fig. S3B). Under tension, a single DNA molecule initiallyin a low E state can transition into a high E state upon relaxationand transition back to the initial low E state upon relaxation duringthe next cycle (Fig. 4F). Each stretching–relaxation cycle can becategorized into transitions from low to high E state (22 out of

103 cycles, ∼21%, Fig. 4 G, 1), high to low E state (17 cycles,∼17%, Fig. 4 G, 2), and remaining in high E states (five cycles,∼5%, Fig. 4 G, 3) or low E states (59 cycles, ∼57%, Fig. 4 G, 4).When a molecule stays in the low E state during a cycle, changes inEwere gradual in both stretching and relaxation periods (Fig. 4G, 4).Interestingly, all molecules in the folded state at the beginning of acycle unfolded at a force of ∼1.5 pN and all molecules that foldedback upon relaxation did so at a similar force. Lack of hysteresissuggests that conformational equilibrium is reached much morerapidly under low ionic conditions. Folding/unfolding at such lowforces would have been invisible in purely mechanical measurements.

hTel22 Mutant Takes on GQ-Like Structures only in Very High IonicConditions. To further ascertain that the peculiar mechanicalcharacteristics are indeed due to GQ structures, we introduced apoint mutation, wherein G8 from the 5′ end, which is located inthe middle quartet, is replaced with T [hTel22mut, GGG(TTAGGG)TTAGTG(TTAGGG)T (22)]. In the absence offorce, unlike hTel22 for which folding could be initiated below10 mM K+, hTel22mut displayed unstructured ssDNA-like be-havior up to ∼100 mM K+, as observed by gradual shift of a peakof low E to higher values, without adding a second populationwith high E (Fig. 5A). Only above 100 mM K+ could we observefolded conformations alongside the unfolded, ultimately result-ing in a distinct high E population at 1 M K+ (Fig. 5A). Weobserved similar behaviors also with Na+ (SI Appendix, Fig. S11A).CD spectroscopy showed a weak positive peak at ∼292 nm in100 mM K+, suggestive of an onset of GQ formation, and in-creasing K+ to 1 M generated a strong positive peak at ∼295 nmwith a shoulder at ∼275 nm and a negative peak at ∼260 nm (SIAppendix, Fig. S10A), indicative of antiparallel GQs (41).In response to increasing and decreasing forces, hTel22mut in

100 mM K+/Na+ showed a gradual decrease and increase in E,respectively, and E vs. force curves were identical betweenstretching and relaxation periods, consistent with unstructured

Fig. 4. Extreme diversity of hTel22 in Na+. hTel22 behavior in 100 mM (A–D) and 10 mM Na+ (E–G). (A) E histogram in 100 mM Na+ in the absence of force. (B)A representative E time trace during two stretching/relaxation cycles. (C and D) Distributions of funfold and frefold (n = 197 for both). The red and black curves inC, respectively, represent individual and overall rupture force distributions predicted using the Dudko–Szabo model (44, 45). (E) E histogram of GQs in 10 mMNa+ in the absence of force. (F) A representative E time trace during four stretching/relaxation cycles. (G) E vs. force example curves. B and F were acquired at20-ms integration time. Error bars represent SEs.

8354 | www.pnas.org/cgi/doi/10.1073/pnas.1815162116 Mitra et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

ssDNA (Fig. 5B and SI Appendix, Figs. S10B and S11B).Therefore, the highly diverse mechanical behavior of the wild-type sequence under the same conditions is due to the bona fideGQ structures.GQ-like behavior with abrupt transitions could only be ob-

served in very high ionic conditions (Fig. 5C). In 1 M K+, 42%(56 out of 134 cycles) showed a single step unfolding (Fig. 5C,cycles 1 and 2), 50% showed unfolding via intermediates (Fig.5C, cycle 3), and 8% displayed ultrastability. funfold changedacross different stretching–relaxation cycles (Fig. 5C and SIAppendix, Fig. S10C) and its distribution can be resolved intodiscrete groups centered at force values of ∼7, 12, and 17 pN(Fig. 5D). Refolding occurred at lower forces (1 to 5 pN; Fig. 5E)and, distinctly from hTel22, occurred in multiple steps via in-termediates of different E values (Fig. 5C and SI Appendix, Fig.S10D). In 1 M Na+, hTel22mut unfolded at lower forces and didnot show an ultrastable state (SI Appendix, Fig. S11).

Conformational Dynamics of Three Repeats of Human TelomericSequence. It has been hypothesized that GQs fold via interme-diates such as G-triplexes (37, 47). G-triplexes are made throughstacking of G-triads. In a G-triad, a central guanine interacts withtwo other guanines by Hoogsteen-like hydrogen bonds (48). Itwas further proposed that G-triplexes formed from three humantelomeric hexanucleotide repeats are stable and have ruptureforces ranging between 31 pN and 35 pN (49). This proposal,however, is difficult to reconcile with our present observationthat even a single base substitution eliminates GQ-like me-chanical features from human telometric repeats in physiologicalionic conditions (Fig. 5). Therefore, we analyzed a sequencecontaining three telomeric repeats [hTel16: GGG(TTAGGG)2T].In the absence of force, hTel16 showed a single dominant

population with low E. As the concentrations of Na+/K+ in-creased, the peak gradually shifted to higher values, consistentwith compaction of unstructured ssDNA (Fig. 6A and SI Ap-pendix, Fig. S13A). Single-molecule time traces did not show anydiscernible fluctuations beyond noise (30-ms resolution) (Fig. 6Band SI Appendix, Fig. S13B). The CD spectrum showed a mainpositive peak at ∼255 nm in 100 mM K+, similar to that what wasobserved in the absence of monovalent cations, except for anadditional positive peak at ∼284 nm (SI Appendix, Fig. S12).

Because a recent CD analysis suggested that Ca2+ is more ef-fective in stabilizing G-triplexes than other ions (49), we alsoperformed Ca2+ titration. Although Ca2+ caused gradual DNAcompaction as observed through gradual increase in E to∼0.85 in 2 mM Ca2+, no additional population emerged (SIAppendix, Fig. S14 A and B).In contrast to the previous optical trap study that showed

evidence of discrete unfolding events under increasing tension

Fig. 5. Single-point mutant of human telomeric repeats (hTel22mut) exhibits GQ-like behavior in 1 M K+. (A) E histograms hTel22mut as a function of K+

concentration in the absence of force. Red arrow indicates initiation of secondary structure formation at 150 mM K+. (B and C) Representative E time traces(20-ms integration time) during stretching/relaxation in 100 mM (B) and 1 M K+ (C). (D and E) Distributions of funfold and frefold of molecules showing abruptunfolding transitions (n = 56). The black and red curves, respectively, denote an overall and individual rupture force distributions predicted using the Dudko–Szabo model (44, 45). Data presented in C–E were collected in 1 M K+.

Fig. 6. G-triplexes behave like ssDNA. (A) E histograms of hTel16 as afunction of K+ concentration. (B) A representative E time trace in 100 mM K+

(30 ms integration time). Average E vs. force curves of hTel16 (C) and (dT)16(D) in 100 mM K+. Error bars represent SEs.

Mitra et al. PNAS | April 23, 2019 | vol. 116 | no. 17 | 8355

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

(49), hTel16 showed gradual decreases in E with increasing forceand retraced the same E vs. force curve during relaxation (Fig. 6C).We observed similar mechanical responses from an unstructuredDNA of the same length, (dT)16 (Fig. 6D). Furthermore, we ob-served the same mechanical responses of hTel16 across differentcationic concentrations used in a previous report (49): 100 mM K+

(Fig. 6C, Top), 100 mM Na+ (SI Appendix, Fig. S13C), 2 mM Ca2+

(SI Appendix, Fig. S14D), and a mixture of 2 mM Ca2+ and100 mM K+ (SI Appendix, Fig. S14E). The end-to-end distancechange upon proposed G-triplex unfolding is about 5 nm (49), wellwithin the detectable limit of smFRET. Even if the mechanicalunfolding of a potential G-triplex occurs at lower forces, it wouldhave been detected because our readout is FRET. Therefore, wecan rule out the formation of a triplex structure with discreteunfolding responses to force.

DiscussionConformational diversity of G4 DNA is well-documented (40,50–56). The end-to-end distance of our human telomeric DNAconstruct (hTel22) for different GQ conformations ranges be-tween 1 and 2 nm and hence cannot be readily resolved solely viaFRET, given the flexible linkers connecting the fluorophores tothe DNA (SI Appendix, Fig. S1). Even though they have in-distinguishable zero-force FRET efficiencies, we detected asmany as six different species that are distinct in their mechanicalunfolding properties. Such extreme diversity in mechanical sta-bility has not been observed before and demonstrates the in-terchangeability and dynamic equilibrium between differentfolded conformations of GQ even under the same solutioncondition (16, 57).The advantages of fluorescence-force spectroscopy became

clear through our investigation. GQ refolding that occurs at lowforces can be directly detected and the refolding forces could beaccurately measured. This is because of our unique ability tomeasure small conformational changes at arbitrary low forces(34) because our readout of smFRET does not have reducedsensitivity at low forces, unlike in purely mechanical measure-ments. In addition, because smFRET reports on local structuralchanges surrounding the labeling locations, our approach is lessprone to potential artifacts arising from surface interactions withthe DNA that are outside the region of interest. This may explainwhy we did not see abrupt unfolding signatures from the threerepeats of hexanucleotides.The abrupt unfolding of GQs observed in the majority of

events suggests that mechanical perturbation of a few HoogstenH-bonds leads to a cooperative destabilization of stacks of G-quartets. The gradual and partial unfolding behavior observedfrom a minor population, however, is suggestive of a morecompliant structure, held together by local interactions whichgradually get disrupted with the increasing force. One possibleexplanation is mutual slippage observed in computational sim-ulations of parallel GQs where one guanine repeat slips relativeto the other repeats one nucleotide at a time, disrupting indi-

vidual G-quartets one at a time (58). If instead GQ unfolds viathe separation of one guanine repeat from the rest, as proposedfor antiparallel or hybrid forms of GQs (58), we would get acooperative disruption of all G-quartets, leading to the abruptunfolding behavior. Several studies have postulated GQunfolding via intermediates such as G-triplexes and hairpins (37,59). Such GQ unfolding intermediates must be shorter-lived thanour time resolution of 20 ms because they could not be capturedfor most molecules with our present assay, except possibly for theG-to-T mutant in very high ionic conditions.The distribution of forces at which abrupt unfolding occurs

informs on the underlying unfolding energy landscape. Assuminga single transition state toward unfolding, three parameters candescribe the landscape: the distance to the transition state Δx‡,the dwell time at zero-force τu(0), and the apparent free energyof activation ΔG‡ (44, 45). Human telomeric GQ structures havebeen identified as parallel, basket-type, and (3 + 1) hybrid (forms1, 2, and 3) in K+ solutions (40, 52–55) and three-layered basket-type, (2 + 2) antiparallel, and so on, in Na+ solutions (50, 51) (SIAppendix, Fig. S1). Several clusters observed in funfold distributionmay be due to different structures with different Δx‡, τu(0), andΔG‡ values. Using the phenomenological model proposed byDudko et al. (44, 45), we estimated Δx‡, τu(0), and ΔG‡s for theindividual force clusters observed with hTel22 in 100 mMK+. Δx‡

was estimated as ∼6, 3.1, 1.8, and 2.8 nm for the funfold peaks at∼3, 11, 18, and 25 pN, respectively (Fig. 7A). Interestingly, thefunfold clusters at around 11 and 18 pN have similar τu(0) and ΔG‡

values but differ in Δx‡, suggesting conformationally distinctspecies (Fig. 7 B and C). However, although funfold clusters ataround 10 and 25 pN have similar Δx‡ values, the latter hadsignificantly higher τu(0) and ΔG‡ (Fig. 7 B and C), consistentwith short- and long-lived species reported at zero force (22).Because unfolding is a stochastic process, we considered thepossibility that the ultrastable state that resisted unfolding up toour maximum force value of 28 pN may not be a separate statebut is due to some pulling cycles of the funfold = 25 pN populationthat did not unfold. However, we estimated that only about 12%of the funfold = 25 pN population would resist unfolding under theexperimental condition, comprising only 2% of all pulling cycles.Hence, the ultrastable population which not only persists formultiple pulling cycles but also constitutes a large fraction ofmechanical response (∼39%) is likely to represent at least onemechanically distinct species (44, 45). The Δx‡ values of differentfunfold populations were preserved between 100 mM and 60 mMK+, indicating that they indeed represent distinct GQ structures(SI Appendix, Fig. S15A), whereas τu(0) and ΔG‡ values wereslightly different, possibly due to the difference in ionic strength.We attempted to assign the funfold clusters to known structuresbut so far have not succeeded, largely because we do not knowthe structural features of the corresponding transition states.Similar analysis with the dual optical trap funfold distribution

yielded Δx‡ = ∼2.6, 1.4, and 2.4 nm for the funfold peaks at ∼11,22, and 29 pN respectively, which agrees reasonably well with the

Fig. 7. Values of (A) Δx‡, (B) τu(0), and (C) ΔG‡s estimated for the funfold clusters observed from fluorescence-force spectroscopy data of hTel22 in 100 mM K+.

8356 | www.pnas.org/cgi/doi/10.1073/pnas.1815162116 Mitra et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

fluorescence-force measurements (SI Appendix, Fig. S15B). Theslight differences in funfold may be attributed to differences in theloading rates (SI Appendix, Fig. S15C). In fact, the funfold valuespredicted with the free energy parameters derived from thefluorescence-force measurements, but at the loading rate of theoptical trap assay, have broader funfold peaks but similar forcevalues (SI Appendix, Fig. S15D). Additionally, we observed asubpopulation at funfold ∼36 pN (Δx‡ = 1.2 nm, ΔG‡ = 12.4 kBT),which may partially account for ultrastable state in fluorescence-force measurements which capped the force at 28 pN.We found that unfolding occurs at lower forces in Na+ com-

pared K+. This observation does not follow automatically fromthe known difference in thermodynamic stability which is lowerin Na+ (60) because Δx‡ could be potentially smaller for struc-tural ensembles populated in Na+, making them more resistantto force. Our data show that this is clearly not the case, that is,Δx‡ is not systematically smaller for structures in Na+ (Δx‡ =∼5.2, 3.7, and 1.7 for funfold of ∼5, 13, and 25 pN, respectively; SIAppendix, Fig. S16) compared with those in K+.Different GQ conformations arise from differences in the

arrangement patterns of syn and anti guanosines. Because cat-ions are stably bound to a folded GQ molecule, rearrangementof the syn and anti guanosine pattern occurs only after substantialunfolding. As the unfolded G-strands are stretched further, amolecule loses memory of its prior state and on subsequentrefolding may lead to a different syn and anti guanosine pattern,potentially explaining why there is no correlation between me-chanical stabilities of GQ structures obtained in consecutivestretching–relaxation cycles. The folding of GQ under physiolog-ically relevant K+ concentrations can take about 40 ms (61) butinterconversion among GQ conformations may take as long as∼4,000 s in the absence of force (37). This can explain the per-sistence of ultrastable GQ species over multiple pulling cycles.A mechanically weak GQ species with an average funfold of

∼2 pN was observed under physiological 100 mM K+ concen-tration. We speculate that this population represents a misfoldedstructure with two quartets instead of three. A canonical threeG-quartet structure can mechanically unfold and stochasticallyrefold into a two G-quartet topology upon force relaxation.Structural studies have proposed a basket-type antiparallel to-pology for a two G-quartet telomeric GQ in K+

, attributed tofaulty stacking between quartets (54). However, such a structurecannot be resolved from a GQ population solely via FRET (SIAppendix, Fig. S1), and the CD signature of such species isanalogous to that exhibited by hybrid-type GQs and hence is notdistinguishable (62). Because two G-quartet telomeric GQs canform easily and can persist for several minutes (63), in the timescale of our measurements, which take a few seconds per me-chanical cycle, a misfolded GQ might not transition to a fullyfolded GQ topology before it is mechanically unfolded. Thepossibility of such species has been elaborated under physio-logical conditions by theoretical as well as experimental studies(58). Also, Galer et al. (64) have shown the predominance oftwo-quartet GQs under acidic pH, which may reduce mechanicalstrength of GQs in hypoxic cells.The GQ folding landscape is extremely rugged. Different

conformations vary with respect to their folding rates such thatmisfolded two-quartet GQs are fastest and parallel GQs areslowest to form (63). Irrespective of the salt concentrations, GQrefolded rapidly on the time scale of a second or shorter once wereduced the force. On the surface, this seems to contradict thenotion that three-quartet GQ folding is slow. Upon force re-laxation, G4 DNA can stochastically refold into a conformationaldistribution that initially is not in thermodynamic equilibrium andmay rearrange over time to form perfectly aligned, enthalpicallyfavorable GQ conformations. We could not test this model ex-perimentally due to photobleaching that limits the observationtime window.

Extreme conformational and mechanical diversity of GQs canresult in differential binding of proteins under physiologicallyrelevant levels of tension, which in turn might affect the equi-librium conformations of telomeric DNA within a cell. Fur-thermore, the unfolding force of GQs, especially in 100 mM K+,is often higher than stall force of motor proteins including DNAand RNA polymerases (13, 65) and may present roadblocks forreplication or transcription. Mechanical heterogeneity may alsocause differential interactions with polymerases and helicasesand thereby regulate replication and transcription.

Materials and MethodsDNA Constructs. All DNA oligonucleotides were purchased from IntegratedDNA Technologies. The G4 strand sequence used is 5′ TGGCGACGGCAGC-GAGGCGGG(TTAGGG)3T/Cy3/T17TCG GGAGCGGACGCACGG 3′, where thetelomeric repeat motif is bold-faced. For the point mutation studies, GGG(TTAGGG)3T, referred to as hTel22, was replaced with GGG(TTAGGG)TTAGTG(TTA GGG), referred to as hTel22mut, where the underline denotesmutation from G to T. A truncated GQ strand hTel16, where GGG(TTAGGG)3Twas replaced with GGG(TTAGGG)2T, was also used in our experiments. Acomplementary stem strand of sequence 3′/Biotin/ACCGCTGCCGTCGCTCCG/Cy5/5′ was used to immobilize the G4 onto the slide surface. A second com-plementary strand (λ-bridge) of sequence 3′ AGCCCTCGCCTGCGTGCCTC-CAGCGGCGGG 5′was used to bridge the G4 strand with λ-DNA. The 12-nt COSsite could be further annealed with a digoxigenin-labeled strand of sequence3′AGGTCGCCGCCC/dig/5′. The G4 strands were first annealed with the com-plementary stem in 1.1:1 ratio in a buffer containing 50 mM NaCl and 10 mMTris·HCl, pH 8, at 95 °C for 5 min, followed by slow cooling to room temper-ature. The λ-bridge was further added to the above mixture in the ratio of1.5:1 to the complementary stem strand and incubated with rotation at roomtemperature for an hour. The G4-construct(s) thus generated were used forsingle-molecule experiments by total internal reflection microscopy (TIRF).

For integrated smFRET–optical tweezers assay, the construct generatedabove was annealed to λ-DNA (New England Biolabs) and the dig-labeledstrand. For annealing, λ-DNA (16 nM) was first heated in the presence of120 mM Na+ at 80 °C for 10 min and then placed on ice for 5 min. The GQ-construct(s) and BSA were added to the λ-DNA at a final concentration of8 nM and 0.1 mg/mL, respectively. The mixture was incubated/rotated atroom temperature for 2 to 3 h. Finally, the dig-strand was added to a con-centration of 200 nM and then incubated with rotation at room tempera-ture for 1 h.

Sample Assembly. To eliminate nonspecific surface binding, PEG (a mixture ofmPEG-SVA and biotin-PEG-SVA; Laysan Bio)-coated coverslips and quartzslides/glass slides were used in our experiments (18). The slides and coverslipswere further assembled to form imaging chambers. For TIRF experiments,50 pM G4-construct(s) were immobilized on the surface via biotin–neu-travidin interaction. Finally, imaging buffer [20 mM Tris·HCl, pH 8, 0.8% wt/volD-glucose (Sigma), 165 U/mL glucose oxidase (Sigma), 2,170 U/mL catalase(Roche), 3 mM Trolox (Sigma), and a predetermined amount of NaCl/KCl]was added for data acquisition.

For integrated smFRET–optical tweezers experiments, the imagingchamber was incubated in blocking buffer [10 mM Tris·HCl, pH 8, 50 mMNaCl, 1 mg/mL BSA (NEB), and 1 mg/mL tRNA (Ambion)] for 1 h. The DNAconstructs were then diluted to 10 pM and immobilized on the surface viabiotin–neutravidin interaction. Next, 1 μM anti-digoxigenin–coated poly-styrene beads (Polysciences) diluted in a buffer containing 10 mM Tris·HCl,pH 8, and 50 mM NaCl was added to the imaging chamber and incubated for30 min. Finally, imaging buffer [50 mM Tris·HCl, pH 8, 0.8% wt/vol D-glucose(Sigma), 0.5 mg/mL BSA (NEB), 165 U/mL glucose oxidase (Sigma), 2,170 U/mLcatalase (Roche), 3 mM Trolox (Sigma), and a predetermined amount ofNaCl/KCl] was added for data acquisition.

Fluorescence-Force (smFRET-Force) Spectroscopy. An integrated fluorescence-optical trap instrument was recently developed in our laboratory to studyconformational changes of biomolecular systems under tension (34, 36). Anoptical trap was formed by an infrared laser (1,064 nm, 800 mW, EXLSR-1064-800-CDRH; Spectra-Physics) through the back port of the microscope(Olympus) on the sample plane with a 100× immersion objective (Olympus).The piezoelectric stage holding the microscope slide was translated, to applytension on the sample tethers. The applied force was read out via positiondetection of the tethered beads using a quadrant photodiode (UDT/SPOT/9DMI). A detailed calibration procedure has been described previously (34,36). A confocal excitation laser [532 nm, 30 mW (maximum average power);

Mitra et al. PNAS | April 23, 2019 | vol. 116 | no. 17 | 8357

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

World StarTech] was focused on the sample through the side port of themicroscope. A piezo-controlled steering mirror (S-334K.2SL; Physik Instru-ment) scanned the sample with the excitation laser. The fluorescenceemission was filtered from infrared laser by a band-pass filter (HQ580/60 m;Chroma) and excitation by a dichroic mirror (HQ680/60 m; Chroma) andsubsequently detected by two avalanche photodiodes.

Data Acquisition. For smFRET imaging in the absence of force, prism-type totalinternal reflection microscopy, with 532-nm laser excitation and back-illuminated electron-multiplying charge-coupled device camera (iXON;Andor Technology) was used (18). Fluorescence data were obtained with theimaging buffer [20 mM Tris·HCl, pH 8, 0.8%wt/vol D-glucose (Sigma), 165 U/mLglucose oxidase (Sigma), 2,170 U/mL catalase (Roche), 3 mM Trolox (Sigma),and a predetermined amount of NaCl/KCl]. Throughout the experiments,smFRET efficiency E was estimated using IA/(IA + ID), where IA and ID are thedonor and acceptor intensities, respectively, after background subtractionand cross-talk correction. FRET efficiency histograms were constructed byaveraging the first 10 data points of each molecule’s time trace.

A detailed data acquisition procedure pertaining to integrated single-molecule fluorescence-force spectroscopy has been described by Hohnget al. (34). In summary, after trapping a tethered bead, its origin was de-termined by stretching the tether along opposite directions along x and yaxes. The trapped bead was then separated from its origin by 14 μm. Theconfocal laser was then used to scan and locate the fluorescence spot on thetether. All unfolding experiments were performed by translating the mi-croscope stage at a speed of 455 nm/s between 14 μm and 16.8 to 17.2 μm.Fluorescence emission from the tether molecule was detected concurrent tothe stage movement, 20 ms after each step in the stage movement.Fluorescence-force data were obtained with the imaging buffer [50 mM Tris·HCl, pH 8, 0.8% wt/vol D-glucose (Sigma), 0.5 mg/mL BSA (NEB), 165 U/mLglucose oxidase (Sigma), 217 U/mL catalase (Roche), 3 mM Trolox (Sigma),and a predetermined amount of NaCl/KCl].

The funfold and frefolds were cast into histograms with bin width approxi-mations from kernel density estimation (66). The funfold histograms wereanalyzed and the free energy parameters such as the transition distances tounfolding and so on were predicted via global fitting to the Dudko–Szabomodel. These derived parameters [Δx‡, τu(0), and ΔG‡] were insensitive to thescaling factor, ν, described in the model. We used parameters at ν = 1/2 forboth the fluorescence-force and dual optical traps assays, to reconstruct theforce profiles (44, 45).

DualOptical Traps Assay. For dual optical trapsmeasurements, the hTel22 sequence,flanked by poly-dT17 on either side, was sandwiched between two long stretchesof dsDNA [left handle (LH), 1.5 kb and right handle (RH), 1.7 kb]. The left and right

handles were amplified from pBR322 plasmid (NEB) and λ-phage DNA (NEB),respectively. The handles were synthesized using primers L1 (5′/biotin/TGAAGTGGTGGCCTAACTACG) and L2 (5′ CAAGCCTATGCCTACAGCAT) for the LH andR1 (5′/digoxigenin/GGGCAAACCAAGACAGCTAA) and R2 (CGTTTTCCCGAAAAGCCAGAA) for the RH. The LH and RH were functionalized with 5′ biotin and5′ digoxigenin to ensure formation of strong attachments to streptavidin- andantidigoxigenin-coated polystyrene beads, respectively. Short ssDNA over-hangs are made by digestion of the LH and RH with restriction endonucleases(NEB). To prevent the LH from self-ligating and to increase the yield of thefinal construct, the 5′ phosphate is removed from the LH. The final constructwas synthesized by ligation of the LH, RH, and hTel22 insert in 1:1:1 ratio atroom temperature for 1 h, followed by gel purification. The detailed PCR,enzymatic digestion, dephosphorylation, and ligation protocols have beendescribed by Whitley et al. (67).

The DNA construct was incubated with streptavidin-coated polystyrenemicrospheres at room temperature for 1 h. All optical trapping experimentswere done in a custom-made microfluidic chamber. The pulling curves andconstant force experiments are performed in the central channel in buffercontaining 50 mM Tris·HCl, pH 8, 0.8% wt/vol D-glucose (Sigma), 0.5 mg/mLBSA (NEB), 165 U/mL glucose oxidase (Sigma), 217 U/mL catalase (Roche),3 mM Trolox (Sigma), and 100 mM KCl. The DNA- and antidigoxigenin-coated microspheres were shunted from the outer to the central channelthrough thin glass capillaries. The DNA construct was stretched between twooptically trapped microspheres at a loading velocity of 100 nm/s and eval-uated by fitting its force vs. extension curve to an extensible worm-like-chain(WLC) model [dsDNA persistence length of 50 nm, ssDNA persistence lengthof 1 nm, stretch modulus of 1,000 pN, and distance per base pair of 0.34 nm(dsDNA) and 0.59 nm (ssDNA)]. The molecules that did not fit to the WLCmodel were excluded from the analysis.

CD Spectroscopy. CD spectra of the human telomeric oligonucleotides wererecorded on an Aviv-420 spectropolarimeter, using a quartz cell of 1-mmoptical path length. The oligonucleotides were diluted to 20 μM in abuffer containing 20 mM Tris, pH 8, and an appropriate concentration ofK+/Na+ ions. Before each measurement, the oligonucleotides were heated to90 °C for ∼5 min and slowly cooled to room temperature, to avoid formationof intermolecular structures. An average of three scans were recorded be-tween 220 and 320 nm at room temperature. The spectra were corrected forbaseline and signal contributions from the buffer.

ACKNOWLEDGMENTS. This work was supported by National ScienceFoundation Grant PHY-1430124 (to T.H. and Y.R.C.) and National Institutesof Health Grants GM122569 (to T.H.) and GM120353 (to Y.R.C.). T.H. is anInvestigator with the Howard Hughes Medical Institute.

1. Gellert M, Lipsett MN, Davies DR (1962) Helix formation by guanylic acid. Proc Natl

Acad Sci USA 48:2013–2018.2. Sen D, Gilbert W (1988) Formation of parallel four-stranded complexes by guanine-

rich motifs in DNA and its implications for meiosis. Nature 334:364–366.3. Gilbert DE, Feigon J (1999) Multistranded DNA structures. Curr Opin Struct Biol 9:

305–314.4. Bochman ML, Paeschke K, Zakian VA (2012) DNA secondary structures: Stability and

function of G-quadruplex structures. Nat Rev Genet 13:770–780.5. Rhodes D, Lipps HJ (2015) G-quadruplexes and their regulatory roles in biology.

Nucleic Acids Res 43:8627–8637.6. Moyzis RK, et al. (1988) A highly conserved repetitive DNA sequence, (TTAGGG)n,

present at the telomeres of human chromosomes. Proc Natl Acad Sci USA 85:

6622–6626.7. Harley CB, Futcher AB, Greider CW (1990) Telomeres shorten during ageing of human

fibroblasts. Nature 345:458–460.8. Bodnar AG, et al. (1998) Extension of life-span by introduction of telomerase into

normal human cells. Science 279:349–352.9. Oganesian L, Moon IK, Bryan TM, Jarstfer MB (2006) Extension of G-quadruplex DNA

by ciliate telomerase. EMBO J 25:1148–1159.10. Zahler AM, Williamson JR, Cech TR, Prescott DM (1991) Inhibition of telomerase by G-

quartet DNA structures. Nature 350:718–720.11. Rezler EM, et al. (2005) Telomestatin and diseleno sapphyrin bind selectively to two

different forms of the human telomeric G-quadruplex structure. J Am Chem Soc 127:

9439–9447.12. Balasubramanian S, Hurley LH, Neidle S (2011) Targeting G-quadruplexes in gene

promoters: A novel anticancer strategy? Nat Rev Drug Discov 10:261–275.13. Yin H, et al. (1995) Transcription against an applied force. Science 270:1653–1657.14. Bochman ML, Sabouri N, Zakian VA (2010) Unwinding the functions of the Pif1 family

helicases. DNA Repair (Amst) 9:237–249.15. Li JH, et al. (2016) Pif1 is a force-regulated helicase. Nucleic Acids Res 44:4330–4339.16. Burge S, Parkinson GN, Hazel P, Todd AK, Neidle S (2006) Quadruplex DNA: Sequence,

topology and structure. Nucleic Acids Res 34:5402–5415.

17. Lane AN, Chaires JB, Gray RD, Trent JO (2008) Stability and kinetics of G-quadruplexstructures. Nucleic Acids Res 36:5482–5515.

18. Roy R, Hohng S, Ha T (2008) A practical guide to single-molecule FRET. Nat Methods 5:

507–516.19. Joo C, Balci H, Ishitsuka Y, Buranachai C, Ha T (2008) Advances in single-molecule

fluorescence methods for molecular biology. Annu Rev Biochem 77:51–76.20. Juette MF, et al. (2014) The bright future of single-molecule fluorescence imaging.

Curr Opin Chem Biol 20:103–111.21. Lerner E, et al. (2018) Toward dynamic structural biology: Two decades of single-

molecule Förster resonance energy transfer. Science 359:eaan1133.22. Lee JY, Okumus B, Kim DS, Ha T (2005) Extreme conformational diversity in human

telomeric DNA. Proc Natl Acad Sci USA 102:18938–18943.23. Maleki P, Budhathoki JB, Roy WA, Balci H (2017) A practical guide to studying G-

quadruplex structures using single-molecule FRET. Mol Genet Genomics 292:483–498.24. Tippana R, Xiao W, Myong S (2014) G-quadruplex conformation and dynamics are

determined by loop length and sequence. Nucleic Acids Res 42:8106–8114.25. Ying L, Green JJ, Li H, Klenerman D, Balasubramanian S (2003) Studies on the struc-

ture and dynamics of the human telomeric G quadruplex by single-molecule fluo-

rescence resonance energy transfer. Proc Natl Acad Sci USA 100:14629–14634.26. Noer SL, et al. (2016) Folding dynamics and conformational heterogeneity of human

telomeric G-quadruplex structures in Na+ solutions by single molecule FRET micros-

copy. Nucleic Acids Res 44:464–471.27. Aznauryan M, Søndergaard S, Noer SL, Schiøtt B, Birkedal V (2016) A direct view of

the complex multi-pathway folding of telomeric G-quadruplexes. Nucleic Acids Res

44:11024–11032.28. Ray S, et al. (2013) RPA-mediated unfolding of systematically varying G-quadruplex

structures. Biophys J 104:2235–2245.29. Long X, Stone MD (2013) Kinetic partitioning modulates human telomere DNA G-

quadruplex structural polymorphism. PLoS One 8:e83420.30. Okamoto K, Sannohe Y, Mashimo T, Sugiyama H, Terazima M (2008) G-quadruplex

structures of human telomere DNA examined by single molecule FRET and BrG-sub-stitution. Bioorg Med Chem 16:6873–6879.

8358 | www.pnas.org/cgi/doi/10.1073/pnas.1815162116 Mitra et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0

31. Koirala D, et al. (2011) A single-molecule platform for investigation of interactionsbetween G-quadruplexes and small-molecule ligands. Nat Chem 3:782–787.

32. You H, et al. (2014) Dynamics and stability of polymorphic human telomeric G-quadruplex under tension. Nucleic Acids Res 42:8789–8795.

33. Long X, Parks JW, Bagshaw CR, Stone MD (2013) Mechanical unfolding of humantelomere G-quadruplex DNA probed by integrated fluorescence and magnetictweezers spectroscopy. Nucleic Acids Res 41:2746–2755.

34. Hohng S, et al. (2007) Fluorescence-force spectroscopy maps two-dimensional re-action landscape of the holliday junction. Science 318:279–283.

35. Ngo TTM, Zhang Q, Zhou R, Yodh JG, Ha T (2015) Asymmetric unwrapping of nu-cleosomes under tension directed by DNA local flexibility. Cell 160:1135–1144.

36. Zhou R, Schlierf M, Ha T (2010) Force-fluorescence spectroscopy at the single-moleculelevel. Methods Enzymol 475:405–426.

37. Gray RD, Trent JO, Chaires JB (2014) Folding and unfolding pathways of the humantelomeric G-quadruplex. J Mol Biol 426:1629–1650.

38. Koirala D, et al. (2012) Intramolecular folding in three tandem guanine repeats ofhuman telomeric DNA. Chem Commun (Camb) 48:2006–2008.

39. Mashimo T, Yagi H, Sannohe Y, Rajendran A, Sugiyama H (2010) Folding pathways ofhuman telomeric type-1 and type-2 G-quadruplex structures. J Am Chem Soc 132:14910–14918.

40. Ambrus A, et al. (2006) Human telomeric sequence forms a hybrid-type intra-molecular G-quadruplex structure with mixed parallel/antiparallel strands in potas-sium solution. Nucleic Acids Res 34:2723–2735.

41. Dapi�c V, et al. (2003) Biophysical and biological properties of quadruplex oligo-deoxyribonucleotides. Nucleic Acids Res 31:2097–2107.

42. Murphy MC, Rasnik I, Cheng W, Lohman TM, Ha T (2004) Probing single-strandedDNA conformational flexibility using fluorescence spectroscopy. Biophys J 86:2530–2537.

43. Hwang H, et al. (2014) Telomeric overhang length determines structural dynamicsand accessibility to telomerase and ALT-associated proteins. Structure 22:842–853.

44. Dudko OK, Hummer G, Szabo A (2006) Intrinsic rates and activation free energiesfrom single-molecule pulling experiments. Phys Rev Lett 96:108101.

45. Dudko OK, Hummer G, Szabo A (2008) Theory, analysis, and interpretation of single-molecule force spectroscopy experiments. Proc Natl Acad Sci USA 105:15755–15760.

46. Lodish H, et al. (2000) Intracellular ion environment and membrane electric potential.Molecular Cell Biology (Freeman, New York), 4th Ed, Section 15.4.

47. Li W, Hou XM, Wang PY, Xi XG, Li M (2013) Direct measurement of sequential foldingpathway and energy landscape of human telomeric G-quadruplex structures. J AmChem Soc 135:6423–6426.

48. Limongelli V, et al. (2013) The G-triplex DNA. Angew Chem Int Ed Engl 52:2269–2273.49. Jiang HX, et al. (2015) Divalent cations and molecular crowding buffers stabilize G-

triplex at physiologically relevant temperatures. Sci Rep 5:9255–9265.50. Lim KW, Ng VC, Martín-Pintado N, Heddi B, Phan AT (2013) Structure of the human

telomere in Na+ solution: An antiparallel (2+2) G-quadruplex scaffold reveals addi-tional diversity. Nucleic Acids Res 41:10556–10562.

51. Wang Y, Patel DJ (1993) Solution structure of the human telomeric repeat d[AG3

(T2AG3)3] G-tetraplex. Structure 1:263–282.52. Dai J, Carver M, Punchihewa C, Jones RA, Yang D (2007) Structure of the hybrid-2 type

intramolecular human telomeric G-quadruplex in K+ solution: Insights into structure

polymorphism of the human telomeric sequence. Nucleic Acids Res 35:4927–4940.53. Dai J, et al. (2007) Structure of the intramolecular human telomeric G-quadruplex in

potassium solution: A novel adenine triple formation. Nucleic Acids Res 35:

2440–2450.54. Lim KW, et al. (2009) Structure of the human telomere in K+ solution: A stable basket-

type G-quadruplex with only two G-tetrad layers. J Am Chem Soc 131:4301–4309.55. Parkinson GN, Lee MP, Neidle S (2002) Crystal structure of parallel quadruplexes from

human telomeric DNA. Nature 417:876–880.56. Zhang Z, Dai J, Veliath E, Jones RA, Yang D (2010) Structure of a two-G-tetrad in-

tramolecular G-quadruplex formed by a variant human telomeric sequence in K+

solution: Insights into the interconversion of human telomeric G-quadruplex struc-

tures. Nucleic Acids Res 38:1009–1021.57. Phan AT, Patel DJ (2003) Two-repeat human telomeric d(TAGGGTTAGGGT) sequence

forms interconverting parallel and antiparallel G-quadruplexes in solution: Distinct

topologies, thermodynamic properties, and folding/unfolding kinetics. J Am Chem

Soc 125:15021–15027.58. Stadlbauer P, Krepl M, Cheatham TE, 3rd, Koca J, Sponer J (2013) Structural dynamics

of possible late-stage intermediates in folding of quadruplex DNA studied by mo-

lecular simulations. Nucleic Acids Res 41:7128–7143.59. Gray RD, Buscaglia R, Chaires JB (2012) Populated intermediates in the thermal un-

folding of the human telomeric quadruplex. J Am Chem Soc 134:16834–16844.60. Bhattacharyya D, Mirihana Arachchilage G, Basu S (2016) Metal cations in G-

quadruplex folding and stability. Front Chem 4:38.61. Zhang AY, Balasubramanian S (2012) The kinetics and folding pathways of intra-

molecular G-quadruplex nucleic acids. J Am Chem Soc 134:19297–19308.62. Karsisiotis AI, et al. (2011) Topological characterization of nucleic acid G-quadruplexes

by UV absorption and circular dichroism. Angew Chem Int Ed Engl 50:10645–10648.63. Marchand A, Gabelica V (2016) Folding and misfolding pathways of G-quadruplex

DNA. Nucleic Acids Res 44:10999–11012.64. Galer P, Wang B, �Sket P, Plavec J (2016) Reversible pH switch of two-quartet G-

quadruplexes formed by human telomere. Angew Chem Int Ed Engl 55:1993–1997.65. Herbert KM, Greenleaf WJ, Block SM (2008) Single-molecule studies of RNA poly-

merase: Motoring along. Annu Rev Biochem 77:149–176.66. Silverman BW (1989) Density Estimation for Statistics and Data Analysis (Chapman &

Hall/CRC, London).67. Whitley KD, Comstock MJ, Chemla YR (2017) High-resolution “Fleezers”: Dual-trap

optical tweezers combined with single-molecule fluorescence detection. Methods

Mol Biol 1486:183–256.

Mitra et al. PNAS | April 23, 2019 | vol. 116 | no. 17 | 8359

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

, 202

0