Biophysical Investigation of RNA ⋅ DNA : DNA Triple Helix and RNA : DNA Heteroduplex Formation by the lncRNAs MEG3 and Fendrr

Long non‐coding RNAs (lncRNAs) are important regulators of gene expression and can associate with DNA as RNA : DNA heteroduplexes or RNA ⋅ DNA : DNA triple helix structures. Here, we review in vitro biochemical and biophysical experiments including electromobility shift assays (EMSA), circular dichroism (CD) spectroscopy, thermal melting analysis, microscale thermophoresis (MST), single‐molecule Förster resonance energy transfer (smFRET) and nuclear magnetic resonance (NMR) spectroscopy to investigate RNA ⋅ DNA : DNA triple helix and RNA : DNA heteroduplex formation. We present the investigations of the antiparallel triplex‐forming lncRNA MEG3 targeting the gene TGFB2 and the parallel triplex‐forming lncRNA Fendrr with its target gene Emp2. The thermodynamic properties of these oligonucleotides lead to concentration‐dependent heterogeneous mixtures, where a DNA duplex, an RNA : DNA heteroduplex and an RNA ⋅ DNA : DNA triplex coexist and their relative populations are modulated in a temperature‐dependent manner. The in vitro data provide a reliable readout of triplex structures, as RNA ⋅ DNA : DNA triplexes show distinct features compared to DNA duplexes and RNA : DNA heteroduplexes. Our experimental results can be used to validate computationally predicted triple helix formation between novel disease‐relevant lncRNAs and their DNA target genes.


Introduction
Long non-coding RNAs (lncRNAs) are physiologically important non-coding RNAs.They are longer than 200 nucleotides.LncRNAs have been reported to regulate fundamental cellular processes including gene transcription, chromatin accessibility, translation and RNA-protein-DNA complex assembly. [1,2]ncRNAs can bind to genomic DNA, involving both, RNA : DNA heteroduplex and RNA • DNA : DNA triplex formation. [3,4]The cellular functions of RNA • DNA : DNA triplexes and their regulatory mechanisms have been described to be operative at the levels of chromatin modifications, DNA repair, transcriptional regulation as well as post-transcriptional RNA processing. [3,4]In general, triple helical structures have also been identified in several naturally occurring RNAs, including the S-adenosylmethionine (SAM)-bound riboswitch, [5,6] the U2/U6 small nuclear RNAs (snRNAs) triplex with pre-mRNA [7] and the human telomerase RNA pseudoknot. [8]Furthermore, microRNAs (miR-NAs) have been shown to regulate transcription through miRNA • DNA : DNA triplexes, which are further stabilized by Argonaute proteins. [9]n RNA • DNA : DNA triplex formation, the single-stranded RNA binds to the major groove of the double-stranded DNA.This RNA • DNA : DNA interaction is stabilized by Hoogsteen or reverse Hoogsteen hydrogen bonds.Which type of Hoogsteen interaction is favored depends on the orientation of the RNA strand.Antiparallel triplexes are formed at neutral pH and in the presence of multivalent cations such as Mg 2 + by reverse Hoogsteen hydrogen bond formation, resulting in U • AT, G • GC and C + • GC triplets.Of these three interaction patterns, only the formation of a C + • GC triplex requires slightly acidic environment to favor the required protonation of cytosines (Figure 1). [3,11,12]The fact that only one but not all of the interactions require an acidic pH has led to the misconception that parallel triplex formation occurs only at low pH.RNA • DNA : DNA triplex formation was reported for several lncRNAs.Examples include the lncRNA KHPS1 that activates the expression of the sphingosine kinase 1 (SPHK1) [13] and the Hypoxia-inducible factor 1-alpha Antisense RNA 1 (HIF1α-AS1) which controls gene expression of Ephrin receptor A2 (EPHA2) and Adrenomedullin (ADM). [14]The triplex-forming lncRNA maternally expressed gene 3 (MEG3) was reported to target the TGF-β pathway genes TGFBR1, TGFB2, and SMAD2, in a transacting manner through its GA-rich triplex forming region. [15]The Foxf1-adjacent non-coding developmental regulatory RNA (Fendrr), another triplex-forming lncRNA, is an example of triplex formation associated with a large number of different functional annotations.It was initially described to be essential for heart, lung and body wall development [16][17][18] and reported to interact with PRC2 and TrxG/MLL (Trithorax group/MLL protein complex).[21][22] The triplex-forming element of Fendrr transcripts regulates certain genes in developing mouse lungs, such as epithelial membrane protein gene 2 (Emp2), which has a triplex target site in its promoter. [23]All these examples document that triplex formation plays important roles in various physiological processes or diseases. [2]Hence, a better understanding of the formation, the mechanism and the structural features is central, also for exploiting triplex formation in therapeutic approaches.
In addition to duplex and triplex formation, the process of R-loop formation during replication is another essential process involving three oligonucleotide strands.During R-loop formation, the hydrogen bonds between two complementary genomic DNA strands are broken and RNA : DNA heteroduplexes are formed.This structure involves the co-localization of three oligonucleotide strands and is an essential process during replication and transcription. [24,25]X-ray diffraction and nuclear magnetic resonance (NMR) studies indicated that RNA : DNA heteroduplexes can adopt both conformations, the more compact A-form, with C3'-endo sugar conformation and the less compact B-form, with a C2'-endo sugar conformation. [26][29] Predicting and identifying triplexes has been challenging, despite being different from R-loop formation.Concerning prediction, several bioinformatics tools, including TriplexAligner, [4] Triplexator, [30] and Triplex Domain Finder [31] have been developed and can be used to predict sequences with high propensity to form triple helices.These sequences notably include promoter regions of different genes.However, experimental demonstration of triplex formation remained challenging and requires evidence obtained by different experimental approaches.Chen et al. recently showed how experimental data can be reconciled with simulation data by using WAXS-driven (wide-angle X-ray scattering) molecular dynamics simulations. [32]Furthermore, Zielinski et al. revealed the formation of RNA duplexes prior to the triplex formation by time-resolved WAXS data, highlighting the dynamic nature of RNA structure. [33]To date, the DNA • D-NA : DNA triplex formation has been demonstrated in vivo through the use of monoclonal antibodies labeled as "triplexspecific", including Jel318 and Jel466 [34,35] or with additional  [10] ) immunofluorescence staining, [35] as well as fluorescent molecules such as thiazole orange. [36,37]Although these methods are interesting, they lack sequence-specificity so they can only be used in combination with approaches focusing on the triplex formation of an individual RNA • DNA : DNA assembly.In our research, we combine a set of in vitro biochemical and biophysical methods to distinguish between DNA duplex, RNA : DNA heteroduplex and both, parallel and antiparallel RNA • DNA : DNA triplexes and determine their relative stabilities to gain more insights in the triplex formation.These methods include electromobility shift assays (EMSA), single-molecule Förster resonance energy transfer (smFRET), microscale thermophoresis (MST), circular dichroism (CD) and NMR spectroscopies.

RNA and DNA Preparation and Hybridization
RNA and DNA constructs (Table 1), except those used for smFRET experiments, were purchased from Horizon Discovery (USA).All samples were purified by HPLC.RNA samples containing 2'-ACE protecting groups were deprotected and desalted according to the manufacturer's protocol before HPLC purification.After HPLC purification, DNA samples were precipitated with isopropanol and RNA samples with LiOCl 4 .DNA samples were desalted using NAP™ columns (GE Healthcare), lyophilized and reconstituted in ddH 2 O. RNA samples were extensively desalted and concentrated using centrifugal concentrators (Sartorius AG).RNA and DNA constructs for smFRET experiments were purchased from Integrated DNA Technology (IDT) (USA).Constructs with amino modifications were fluorescently labeled with either Cy3 or Cy5.All constructs were purified by HPLC. [38]Labeled RNA was also used for MST experiments.Briefly, 30 nmol of the desired construct were precipitated with ethanol.The pellet was resuspended in 20 μL 100 mM NaHCO 3 and mixed with the dye resuspended in 20 μL DMSO and incubated for 90 min at room temperature.After ethanol precipitation, the labeled construct was purified by HPLC.The purified construct was precipitated and resuspended in 100 μL ddH 2 O.
For the hybridization, the complementary DNA single strands were first combined and incubated for 5 minutes at 95 °C, in hybridization buffer (25 mM HEPES (pH 7.4), 50 mM NaCl, 10 mM MgCl 2 for antiparallel triplex constructs; 25 mM KPi (pH 5.0) or 20 mM LiOAc (pH 5.0), 10 mM MgCl 2 for parallel triplex constructs) and then cooled to room temperature.The same protocol was performed for the hairpin DNA.RNA • D-NA : DNA triplexes were formed by adding RNA and hybridizing the RNA strand to a DNA duplex or hairpin at 60 °C for 1 h followed by slow cooling to room temperature. [39]RNA : DNA heteroduplexes were formed by adding the RNA to the CT-rich DNA strand (anti-parallel triplex constructs) or the GA-rich DNA strand (parallel triplex constructs) in the hybridization buffer.The mixtures were incubated at 60 °C for 1 h followed by slow cooling to room temperature. [39]

Synthesis and Choice of Constructs
To perform in vitro investigations of triplex-forming lncRNAs and their target DNAs, construct design is important.First, • GC base triplets depending on their sequence specificity and RNA length. [40]They determined an optimal RNA length between 19 and 27 nucleotides (nt) with up to two mismatches to form a stable RNA • DNA : DNA triplex. [40]Excessively long constructs should be avoided to prevent repetitive sequences and intramolecular interactions.Previous data showed that high concentrations of RNA are required to form RNA • DNA : DNA triplexes with a DNA duplex. [14,23]To stabilize the triplex formation and allow the use of less RNA, we designed DNA hairpin constructs in which both DNA strands were linked by a 5 nt thymidine linker.Besides constructs synthesized by solid-state synthesis, in vitro transcription can be used to prepare RNA.This requires a DNA template containing the sequence of the RNA of interest as well as the T7 promotor and allows the preparation of isotopelabeled RNA constructs, necessary for a range of NMR experiments. [41]However, transcription of triplex-forming RNAs proved to be challenging since the constructs contain a high GC-value, which often results in low transcription efficiency and difficult purification.

Characteristic Features in Biochemical and Biophysical Methods of RNA • DNA : DNA Triplexes
To establish a set of biochemical and biophysical experiments including EMSA, thermal melting assays, FRET, CD and NMR spectroscopies for triplex characterization, we investigated RNA • DNA : DNA triplex, DNA duplex and RNA : DNA heteroduplex motifs.In principle, EMSA experiments can be performed with either unlabeled, fluorescently labeled or radiolabeled samples, at low concentrations, allowing to distinguish between DNA duplex and RNA • DNA : DNA triplex (Figure 2A).With increasing RNA concentration, triplex-formation is favored and the DNA bands shift. [13,42]Biophysical experiments provide information about the structure of duplexes, heteroduplexes, parallel and antiparallel triplexes.Temperature-dependent CD spectroscopy experiments can be performed with samples of low μM concentrations.CD spectra can be used to distinguish between secondary structures of proteins and oligonucleotides.Not only the differentiation between DNA and RNA • DNA : DNA triplex is possible, but also heteroduplexes and different Gquadruplex polymorphs can be discriminated (Figure 2B).Specific properties of CD spectra of antiparallel RNA • DNA : DNA triplexes are a small positive peak at 220 nm, two negative peaks at 210 nm and 240 nm and a blue-shift of the peak at 270 nm. [14,43,44]For parallel triplexes a negative peak at 210 and 240 nm and a positive peak at 220 nm and an enhanced peak at 280 nm are prominent. [43,45]Thermal melting assays provide thermodynamic information from the melting temperatures.
The characteristic feature of an RNA • DNA : DNA triplex is the biphasic melting transition with the first melting point corresponding to the melting of the Hoogsteen hydrogen bonds that stabilize the RNA • DNA : DNA triplex.The second melting point reports on the melting of the Watson-Crick base pairing (Figure 2C).The triplex samples contain heterogeneous mixtures with different structural motifs that form depending on the temperature.Triplexes are stable at lower temperatures compared to double-stranded DNA or RNA : DNA heteroduplexes (Figure 2D).The stability of structural motifs is highly dependent on the choice of DNA and RNA sequence.If a DNA duplex is used instead of a stabilized hairpin DNA, the melting temperatures of the different states will vary.
To further investigate triplex formation, MST experiments can be carried out to determine K D values.Mondal et al. used  this method by labeling one DNA strand with a fluorophore. [42]n our studies, we switched the label onto the RNA and determined K D values that lie in the higher nM to low μM range for the triplex and in the lower nM range for heteroduplex formation (Figure 2E).
smFRET experiments can be used to distinguish between the triplex and the heteroduplex.The resulting FRET efficiencies depend highly on the construct design with the positioning of the fluorophores forming the FRET pair due to its distance dependence. [46]Our constructs result in high FRET efficiencies for triplexes and medium to low FRET efficiencies for heteroduplexes (Figure 2F).[49][50] Not only the structural motifs but also the different base pairing types, such as Watson-Crick base pairs and Hoogsteen base pairs, can be differentiated.These molecular signatures lead to different chemical shifts for reporter signals in the NMR spectra.While aromatic hydrogens of the molecules resonate in signals between 5-10 parts per million (ppm), Hoogsteen base pairs can be observed between 10 and 12 ppm and Watson-Crick base pairs between 12 and 15 ppm.These chemical shift differences allow differentiation between DNA, RNA, heteroduplexes and RNA • DNA : DNA triplexes, as well as G-quadruplexes.Specific for parallel triplexes at lower pH are base-paired C + signals around 15.5 ppm (Figure 2G). [12]This pool of techniques provides a reliable basis for in vitro characterization of RNA • DNA : DNA triplexes.
In EMSA, the running behavior in the 15 % PAGE of the MEG3 • TGFB2_GA-rich : TGFB2_CT-rich triplex is slower compared to the DNA duplex (Figure S1A).Additionally, the triplex formation of MEG3 RNA by binding to TGFB2 DNA duplex could be confirmed by CD spectra (Figure 3B).Furthermore, we performed thermal melting assays and obtained different melting temperatures for the aforementioned duplexes and triplexes (Figure 3C, Table 2).The second melting transition is comparable to the melting temperature of the RNA : DNA heteroduplex (Figure 3C), leading to the conclusion that the strands form a heterogeneous mixture containing both the triplex and the heteroduplex.Further, 1 H-1D NMR spectra of TGFB2 DNA duplex, MEG3 : TGFB2_CT-rich heteroduplex and MEG3 • TGFB2_GA-rich : TGFB2_CT-rich triplex were obtained at different temperatures between 278 K and 308 K. Triplex formation was validated by the 1 H-1D NMR imino signals observed in the Hoogsteen spectral region between 9 and 12 ppm using 3 eq of MEG3 lncRNA (Figure 3D, Supplementary Figure S1B).

Characterization of a Stabilized Two-Stranded Antiparallel Triplex by the Example MEG3 • TGFB2_CTGA
To reduce the amount of RNA necessary for triplex hybridization, an intramolecular DNA hairpin was designed as a DNA : DNA duplex platform (Table 1, Figure 4A).This approach stabilizes RNA • DNA : DNA triplex formation and inhibits RNA : D-DNA heteroduplex formation.Using the DNA hairpin construct   S2A).The CD spectrum of the stabilized MEG3 • TGFB2_CTGA triplex showed characteristic features compared to the TGFB2_CTGA hairpin DNA with a positive peak at 220 nm, two negative peaks at 210 nm and 240 nm and the blue-shifted peak at 270 nm (Figure 4B).Our thermal melting temperature studies showed an increase in the temperature of the first melting point of the MEG3 • TGFB2_ CTGA triplex compared to the triplex formed with doublestranded (Figure 4C, Table 2).MST measurements were performed with constant MEG3-Cy5 RNA and a titration series of The higher K D value corresponds to the triplex and the lower K D value to the heteroduplex formation, as this value is comparable to the K D (MEG3 : TGFB2_CT-rich heteroduplex) = 79 � 33 nM (Figure 4D, middle).The stabilizing effect of the TGFB2_CTGA hairpin construct is confirmed by the decreased K D (MEG3 • TGFB2_CTGA triplex) = 847 � 378 nM (Figure 4D, bottom).The formation of the antiparallel triplex was further characterized by smFRET experiments using the T-loop construct TGFB2_GACT_sm.smFRET constructs were designed to expect a high FRET state in close proximity to the FRET pair Cy3 and Cy5 (Figure S2B).The results show that besides the donor only signal (E FRET ~0), the presence of 50 nM MEG3-Cy5 RNA reveals a high FRET state (Figure 4E, E FRET = 0.72) corresponding to the formed triplex.This was validated by experiments with TGFB2_CT_sm as the control construct, which showed a moderate FRET state (Figure 4F, E FRET = 0.40).The FRET efficiency at 0.76 potentially could rise due to a formation of a RNA • RNA : DNA because of the high RNA concentration.The formation of RNA:RNA : DNA triplexes is however not further addressed in the current study.The experiments conducted with the control construct successfully served their purpose in identifying the FRET efficiency of the heteroduplex.We confirmed the findings of the smFRET experiments, using this method on another known antiparallel triplex formed by lncRNA HIF1α-AS1 and EPHA2 [14] with a similar construct design (Figure S3).Further, triplex formation between the TGFB2_CTGA hairpin DNA and MEG3 RNA was validated by 1 H-1D NMR spectra of the TGFB2_CTGA DNA hairpin and the MEG3 • TGFB2_ CTGA triplex.Here, we observed new imino signals in the Hoogsteen spectral region between 9 and 12 ppm using only 1 eq of MEG3 RNA at temperatures between 278 and 308 K (Figure 4G, Figure S2C).

Characterization of Stabilized Two-Stranded Parallel Triplex by the Example Fendrr • Emp2_GACT
We further studied the lncRNA Fendrr that forms a parallel triplex.Ali and Rogala et al. previously identified Emp2 as a target gene of Fendrr, containing a triplex target site.Through Hoogsteen base pairs, Fendrr interacts with the CT-rich strand of the Emp2 duplex, forming a parallel triplex with an excess of 5 eq RNA. [23]Here, we report triplex formation using an intramolecular DNA hairpin.The stabilization allows triplex formation with only 2 eq of RNA.Furthermore an RNA : DNA heteroduplex can be formed by RNA interaction with the GArich DNA strand via Watson-Crick and Wobble base pairing (Table 1, Figure 5A).
The band shift upon addition of Fendrr RNA to the Emp2_ CTGA DNA hairpin confirmed the triplex formation with EMSA (Figure S4A).Typical CD features of a triplex were observed (Figure 5B).Moreover, DNA hairpin, RNA : DNA heteroduplex and RNA • DNA : DNA triplex can be distinguished.We also applied thermal melting analysis to the three different motifs and obtained the specific melting temperatures.(Figure 5C, Table 2).By MST, the following K D values were determined: K D (Fendrr • Emp2_GACT triplex) = 589 � 138 nM, K D (Fendrr:Emp2_ GA-rich heteroduplex) = 16 � 12 nM and K D (Fendrr • Emp2_ GACTtriplex) = 497 � 70 nM (Figure 5D).The formation of the parallel triplex was also confirmed by smFRET experiments.Using the hairpin construct Emp2_CTGA_sm, the DNA duplex was stabilized, and a similar labeling strategy was used as for the antiparallel triplexes (Figure S4B).The addition of the Fendrr-Cy5 RNA revealed a high FRET state (Figure 5E, E FRET = 0.78) corresponding to the formed triplex, as the FRET pair Cy3 and Cy5 are in close proximity.The triplex formation was validated by experiments with the Emp2_GA_sm as the control construct representing the heteroduplex formation, which showed a low FRET state (Figure 5F, E FRET = 0.20) and was not found in the previously described histogram of the triplex.The FRET histograms shown represent the red filtered results (Unfiltered results can be found in Figures S4C and D).The pH dependence was controlled by changing the pH to 7 (Figure S4E).This leads to nearly complete elimination of the high FRET state indicating that triplexes are not formed under this condition.In the next step, we performed 1 H-1D NMR spectra at various temperatures.The different aforementioned motifs showed specific signals, most prominent for the triplex formation are new signals in the non-canonical spectral region between 9 ppm and 12 ppm caused by Hoogsteen base pairing.Besides, new signals between 15.5 and 16 ppm arise that are specific for base paired C + (Figure 5G, Figure S4F and S5). [12,51]he described spectroscopic signatures that were used to distinguish DNA, heteroduplexes and triplexes, both parallel and antiparallel can be found in Table 3.

Conclusions
Recent studies investigating the interaction of DNA duplexes with RNA single strands have mainly focused on triplex formation, but have neglected the fact that RNA : DNA heteroduplexes are also formed in triplex samples.Our results show that triplex samples consist of heterogeneous mixtures, including DNA duplexes, RNA : DNA heteroduplexes and RNA • DNA : D-DNA triplexes.We present a variety of biochemical and biophysical methods to validate antiparallel and parallel triplex formation, which will provide better insight into the formation, structure and stability.A simple, fast and cost-efficient approach to detect triplex formation in vitro is the native gel shift assay, which shows a band shift when the triplex is formed compared to the DNA duplex.Typically, EMSAs are performed with low sample concentrations in the nanomolar range, which can also be used with unlabeled or radiolabeled oligonucleotides.However, this approach remains imprecise concerning structural information, demanding the use of additional biophysical methods. [13,52]n particular, the combination of CD spectroscopy and thermal melting assays can be used for detection of triplex formation as a time-saving method with sample concentrations in the low micromolar range between 5 and 10 μM, providing more accurate results than EMSAs for triplex validation.CD spectra not only distinguish between the different motifs of Gquadruplexes, proteins, DNA duplexes, RNA : DNA heteroduplexes and RNA • DNA : DNA triplexes, but also differentiate between antiparallel and parallel triplexes.CD ellipticity charac-teristics for antiparallel triplexes include two prominent negative peaks at 210 and 240 nm, a small positive peak at 220 nm and a blue shift of the peak at 280 nm to 270 nm. [43,44]For parallel triplexes, two prominent negative peaks at 210 and 240 nm, a transition at 260 nm and a large positive peak at 280 nm are observed. [45]Furthermore, thermal, [14,23] UV [44,53] and fluorescence [54] melting assays determine triplex stability by melting at a constant heating rate.The melting curves showed significantly different features, allowing a direct comparison of DNA duplex and RNA : DNA heteroduplex and different RNA • D-NA : DNA triplex motifs.We observed an increased melting temperature of the hairpin DNA compared to the DNA duplex of approximately 14 °C for TGFB2 constructs, resulting from stabilization effects due to the thymidine linker insertion.Previously, the RNA : DNA heteroduplex formation has been shown to be thermodynamically stable compared to DNA duplexes, [27] depending on the length and composition of the oligonucleotide strands with decreasing thermal stability for short oligonucleotides and including continuous high A : T/U stretches within the sequences. [29]Consistent with this, we have shown an increased thermal stability of MEG3 : TGFB2 RNA : DNA heteroduplexes, as their melting temperature is 7 °C higher than that of TGFB2 DNA duplexes.The biphasic melting transition is the major characteristic of triplex formation.The first melting temperature corresponds to the dissociation of the RNA third strand by melting of the weaker Hoogsteen hydrogen bonds.The higher melting temperature is usually described as the melting of the DNA duplex, but to some extent the second transition could also result from the RNA : DNA heteroduplex as it could be formed by strong Watson-Crick hydrogen bonds. [44,53,55]e determined the K D values of RNA : DNA heteroduplex and RNA • DNA : DNA triplex by MST, which is a fast and costefficient method.MST detects changes in the motion of molecules in a temperature gradient. [56]It is necessary to label one construct with a fluorescent dye, as the fluorescence intensity is used as a readout. [57]MST reveals K D values for the RNA : DNA heteroduplex in the low nanomolar range between 15 and 150 nM and K D values for the RNA • DNA : DNA triplex in the mid nanomolar to low micromolar range between 500 nM and 2.2 μM.Our results clearly state the stabilization of the triplex by the hairpin constructs as the K D value is lowered, for both the antiparallel and parallel triplex.MST experiments showed a higher stability of the parallel triplex compared to the antiparallel triplex.This can be explained by the number of formed Hoogsteen base pairs formed (parallel: 24; antiparallel 10) and further by the results of the studies by James et al. which showed that alternating C + • GC and T • AT triplets form the most stable triplexes. [58]Previously, the formation of a DNA triplex has been observed using smFRET experiments by Lee et al.. [36] We employed smFRET, to distinguish between RNA : D-DNA heteroduplexes and RNA • DNA : DNA triplexes, which is feasible at very low concentrations (pM-nM) due to the distance dependence. [59]However, the distance dependence also limits this tool in the size of the samples used.However, by changing the location of the label or changing the FRET pair, there is a lot of flexibility for the experimental design. [46]Another limitation of this method is a high background during the measurements due to the free dissociation of the labeled sample in solution.Nevertheless, measurements with labeled sample concentrations of 50 nM were possible without any difficulties.The resulting FRET efficiencies with our labeling strategy resulted in medium to low FRET states for the RNA : DNA heteroduplexes and high FRET states for the RNA • DNA : DNA triplexes.Additionally, our smFRET studies demonstrate that a slightly acidic pH value is required for parallel triplex formation, at least in the presence of C + • GC triplets, as described by Keppler and Fox. [60]he most powerful method for triplex validation is NMR spectroscopy.It is the most time-consuming method and requires sample concentrations in a range of 50 to 500 μM, but also yields the most accurate results.By NMR spectroscopy sequence specific base pairings of the studied triplexes can be determined by distinguishing between Watson-Crick and Hoogsteen base pairs as previously shown by Rajagopal and Feigon, as well as de los Santos et al. [12,61] A discrimination between homoduplex, heteroduplex, and triplex-formation is In this review, we have combined the different biophysical tools to report on both, the parallel and antiparallel triplex forming lncRNAs in vitro.Each method can be applied individually, but with different levels of accuracy.Thus, a combination of these techniques facilitates the validation and structural elucidation of triplexes, as well as the distinction between them and heteroduplexes, which were mostly overlooked in previous studies.To decipher the biological functions and molecular mechanisms of triplex formation between lncRNAs and their DNA targets, it is important to gain insight into their structural features using various biochemical and biophysical techniques, including EMSAs, FRET, CD and NMR spectroscopy, together with molecular modeling approaches.

Experimental Section Electrophoretic Mobility Shift Assay
EMSAs were performed using native RNA PAGE, respectively.Samples were prepared containing 50 % of glycerol at a sample concentration of 0.5 μM.Native RNA PAGE gels were prepared using 15 % (v/v) polyacrylamide and TA buffer (50 mM Tris/acetate, 50 mM sodium acetate (pH 8.3)).Bands were separated at constant power (< 1 W) for 6-9 h at 4 °C, stained with GelRed® (Biotium, USA) and visualized using the Gel Doc XR + (Bio-Rad, USA) gel documentation imager.

Microscale Thermophoresis
For all microscale thermophoresis (MST) measurements, the RNA was 5'-labeled with the fluorescent dye Cy5.For triplex measurements, the DNA was folded as described above in the corresponding hybridization buffer for antiparallel or parallel triplexes.A 1 : 1 dilution series of the DNA in the hybridization buffer was prepared.RNA in the appropriate hybridization buffer was added to each sample in the dilution series and hybridized as described above.Finally, the samples were supplemented with 0.4 mg/mL BSA (NEB), 0.05 % Tween 20 and RNasin (Promega, Germany).The final RNA concentration in each sample was 30 nM.Samples were transferred to monolith capillaries (NanoTemper).Measurements were performed between 22 and 25 °C with an excitation power of 40 % and MST power of 20 % to 40 % using a microscale thermophoresis device (NanoTemper Monolith NT.115).Each sample was scanned at least three times and analyzed using MO.Affinity Analysis with T-Jump evaluation.

Single-Molecule FRET
For smFRET experiments, samples were hybridized as described above with 1 μM DNA in the appropriate buffer for parallel or antiparallel triplex measurements.For construct immobilization, the glass surface was coated with biotinylated BSA and streptavidin.The sample was diluted to the appropriate concentration for immobilization, ranging from 15 to 50 nM.The diluted sample (30 μL) was flowed over the slide and incubated for 2 min.The slide was washed with the corresponding hybridization buffer (25 mM HEPES (pH 7.4), 50 mM NaCl, 10 mM MgCl 2 for antiparallel triplex constructs, 20 mM LiOAc, 10 mM MgCl 2 (pH 5.0) for parallel constructs).to remove the unbound sample.The appropriate hybridization buffer was saturated with Trolox (6-hydroxy-2,5,7,8tetramethylchroman-2-carboxylic acid) and additionally contained 5 mM protocatechuic acid and 50 nM protocatechuate-3,4-dioxygenase as an oxygen scavenging system and 0 to 50 nM of the corresponding RNA was applied to the slide.Measurements were started after an equilibration time of 5 min.
A custom-built prism-type TIRF microscope was used for smFRET experiments [62] using an Olympus IX73 microscope and a UPLSAPO 60XW, NA 1.2 water immersion objective with a bandpass filter (Semrock FF01-515/588/700-25).The Cy3 dye was excited by a 532 nm green laser (Laser Quantum) with laser powers between 12 and 30 mW.The Cy5 dye was excited by a 637 nm red (Coherent) diode laser at 15 mW to filter out red signals.The emission path was separated by 643 nm dichroic mirrors (Chroma, H 643 LPXR).Fluorescence emission was detected with a 100 ms integration time using an EMCCD camera (Photometrics Evolve 512).50 frame movies were recorded, consisting of 20 frames with green laser excitation and 20 frames with red laser excitation.The red laser excitation was used to filter out molecules without a Cy5 dye.Fluorescence time traces were extracted and analyzed using MATLAB (The MathWorks, Inc.) scripts.

CD Spectroscopy and Thermal Melting Curve Analysis
Circular dichroism spectra were acquired on a Jasco J-810 spectropolarimeter.Measurements were recorded from 210 to 320 nm at 25 °C using a 1 mm path-length quartz cuvette.CD spectra were recorded on 8 μM samples of each DNA heteroduplex and RNA • DNA : DNA triplex (25 mM HEPES (pH 7.4), 50 mM NaCl, 10 mM MgCl 2 for antiparallel triplex constructs, 25 mM KPi (pH 5.0) for parallel triplex constructs).Spectra were acquired with eight accumulation scans at a speed of 50 nm/min and the data were baseline corrected and smoothed with Savitzky-Golay filters (15 fit).Observed ellipticities, recorded in millidegrees (mdeg), were converted to molar ellipticity [θ] = deg×cm 2 ×dmol À 1 .Melting curves were acquired at constant wavelength using a temperature rate of 1 °C/min in the range of 5 °C to 95 °C.All data were evaluated using Microsoft Excel 2016 and SigmaPlot 12.5.All melting temperature data were converted to normalized ellipticity and evaluated using SigmaPlot according to the following equation: [14] f ¼ a

Supporting Information
Additional supporting Figures can be found in the Supporting Information.

Figure 1 .
Figure 1.Base triplets involved in the parallel and antiparallel triplex formation.RNA bases are indicated in red, DNA bases are indicated in black.The antiparallel base triplets include A • AT, G • GC and U • AT and the parallel base triplets are U • AT, G • GC and C + • GC.The notation of dots (•) indicates the Hoogsteen interactions whereas the pairing of DNA strands occurs through Watson-Crick interactions.(Adapted from Bacolla et al.[10] )

Table 1 .
All DNA and RNA oligonucleotides used for EMSA, CD, melting curve analysis, 1 H-1D NMR and smFRET.

Table 2 .
Melting points of all systems examined.

Table 3 .
Spectroscopic signatures of the different motifs.MST and smFRT signatures were be determined (n.d.) for duplex DNA and hairpin DNA alone.for both antiparallel and parallel triplexes.Both triplex motifs show new signals between 10 and 12 ppm caused by Hoogsteen base pairing.Besides this for parallel triplexes additional signals are visible in a range of 15.5 to 16 ppm, which are dedicated to the base-paired C + under the acidic conditions. possible