*2.2. Sox2HMG Induces Sequential dsDNANANOG Bending Transitions*

Next, we focused on understanding the mechanism of the TF-DNA complex formation. Although the mobility shift assay clearly demonstrates a multistep higher-order Sox2HMG complex formation with the dsDNA (Figure 1b), our ensemble experiments (i.e., fluorescence anisotropy and fEMSA) were not sensitive enough to determine the stoichiometries of respective TF-DNA complexes. To directly observe Sox2HMG-DNANANOG binding steps, we performed single-molecule fluorescence microscopy experiments that provide key advantages over conventional ensemble methods: (1) individual conformational sub-populations that are averaged out in ensemble measurements can be directly detected; and (2) experiments can be performed with extremely low concentrations of the labeled molecule (typically 50–100 pM). The ability to carry out experiments at low biomolecule concentration provides access and resolution for characterizing individual interaction steps in tightly interacting systems.

We utilized the distance-dependence of FRET to characterize Sox2HMG-DNANANOG interaction at single-molecule resolution. smFRET is sensitive to distance changes in the 20–70 Å range [26], and provides the necessary spatial resolution to probe changes in dsDNANANOG conformations as induced by TF binding (estimated end-to-end distance of DNANANOG is 57.4 Å, assuming inter-base axial rise of 3.4 Å [27]). For the smFRET experiments, we labeled DNANANOG with Alexa Fluor 488 and 594 donor-accepter dye-pair (Supplementary Methods). Bursts of fluorescence from donor and acceptor dyes were recorded as dual-labeled *NANOG* promoter DNA passed through the sub-fL observation volume of our custom-built ISS Alba confocal laser microscopy system (described previously [28]). These fluorescence intensities were converted to FRET efficiency (*E*FRET) histograms, providing a scheme for direct visualization of DNA conformational distributions. Without Sox2HMG, the dual-labeled DNA showed a single-peak in its *E*FRET histogram with histogram width typical of smFRET studies of freely diffusing dsDNA molecules [29,30] (Figure 2a; top panel). An NLS fit of the histogram to a Gaussian function yielded *E*FRET value of 0.39 (±0.04). On the basis of this *E*FRET value, we estimate the apparent distance between the two dyes to be approximately 64.6 Å (assuming a Förster distance of 60 Å between Alexa 488/594 dyes [31]). This is consistent with the estimated end-to-end distance of dsDNANANOG, where the slight increase in the apparent distance (compared to the estimated distance) can be attributed to the linkers present in Alexa dyes.

Often, histograms of data collected in diffusion-based smFRET experiments show an additional peak at zero *E*FRET that arise from molecules with active donor(s) and either inactive or absent acceptor [29,30,32–35]. These zero *E*FRET peaks tend to significantly overlap with low *E*FRET peak populations and hamper direct estimation of the position of the non-zero peak(s) [36–39]. Interestingly, our smFRET histograms lack zero *E*FRET peaks (Figure 2a). We attribute this to the absence of dual donor-labeled dsDNA molecules as ensured by sequential labeling of individual DNA strands (Supplementary Methods). Therefore, sequential labeling and purification of individual fluorophore-conjugated oligos prior to duplex formation can be utilized to minimize zero peaks.

DNA bending (also known as DNA looping) is critical for many eukaryotic TF function [40–44]. Accordingly, Sox2 was shown to induce binding-mediated *FGF* (fibroblast growth factor) enhancer bending [18]. We postulate that similar spatially precise bending is induced in Sox2-DNANANOG complexes during gene regulation. To characterize Sox2HMG binding-induced *NANOG* promoter DNA bending, we carried out isothermal smFRET Sox2HMG titration against approximately 100 pM dual-labeled DNA (Figure 2). Our smFRET experiments provide a direct way to distinguish between subtle conformational changes of DNANANOG induced upon Sox2 binding. In our smFRET experiments, we observed a multistep bending transition in the DNA structural landscape (Figure 2a). Initially, DNANANOG undergoes a cooperative bending to a 0.45 (±0.01) *<sup>E</sup>*FRET state that corresponds to 32.1◦ (±1.4◦) apparent bend angle at low Sox2HMG concentrations (≤4 nM) (see Supplementary Methods for the details of FRET-to-apparent-angle conversion). NLS fit of the data yields an estimated *K*<sup>D</sup> of 305 (±39) pM (Figure 2c). Such a tight interaction is unlikely to be driven by higher order

Sox2HMG assemblies and we therefore postulate that this dsDNA conformation (henceforth referred as BI) is induced by binding to single Sox2HMG molecules.

**Figure 2.** smFRET reveals Sox2HMG concentration-dependent multistep bending of DNANANOG. (**a**) *E*FRET histograms of DNANANOG with increasing [Sox2HMG]. (**b**) [Sox2HMG]-*E*FRET contour map color coded based on fractional occupancy of individual DNA conformations. Corresponding DNA conformations are marked on the contour map. (**c**) Sox2 binding isotherm of the U - BI transition as probed by detecting changes in *E*FRET, linked to dsDNA bending transition. The NLS-derived apparent *K*<sup>D</sup> for this binding step is 0.30 (±0.04) nM (binding equation with fixed Hill coefficient of 1). (**d**) dsDNANANOG conformational distributions as modulated by Sox2HMG concentration, determined from NLS fitting of individual smFRET histograms to Gaussian functions.

Our ensemble results suggested that multiple Sox2HMG can form higher order TF-DNA assemblies (Figure 1). To characterize the complex formation, we probed for changes in DNANANOG conformations upon further addition of Sox2 on preformed monomeric Sox2HMG-DNANANOG complexes. With increasing [Sox2HMG], we observe a progressive reduction of the BI population and the emergence of a new population exhibiting higher *E*FRET (~0.68). This higher *E*FRET population corresponds to a DNANANOG apparent bend angle of 70◦ (±2.4◦; henceforth referred to as BII DNA conformation). We infer that this DNA conformation is induced by sequential binding of two individual Sox2 TFs on the dsDNA, where binding of each monomer induces an approximate 32◦ bend at respective binding sites. Our observed apparent bend angle in the ternary complex (two Sox2 monomers and DNA) is similar to the DNA bend angle previously resolved for heterodimeric HMG box TF-DNA complexes [11,17].

Interestingly, an additional transition is visible in our isothermal smFRET titration when additional Sox2HMG is added (i.e., >75 nM [Sox2HMG]). We observe progressive depopulation of the BII bent DNA conformation and coupled emergence of a population at *E*FRET ~0.44 as [Sox2HMG] increases further (henceforth referred as BIII; Figure 2a). We estimate the apparent bend angle for the BIII population to be 30.4◦ (±4.5◦) from the *E*FRET data (Supplementary Methods). A longer fEMSA run also indicates higher-order oligomer formation that is consistent with the formation of BIII population (Figure S2). Mechanistically, Sox family TFs induce DNA bends via FM dipeptide intercalation between two Thymine (T) bases at the minor groove interface [45,46]. Within the *NANOG* composite promoter, three TT pairs are present: two within the two HMG-TF binding sites (Oct/Sox motifs) identified by

Rodda et al. [47] and one in between. We hypothesize that the initial two DNA bends are induced by sequential Sox2 binding to the two high-affinity HMG-TF binding motifs, where each binding induces an apparent 32◦ bend at the sites of interactions (a net 70◦ DNA apparent bend angle in the ternary complex). As Sox2HMG concentration further increases (>75 nM), an additional TF molecule interacts with the DNA at the remaining TT site and induces similar bend albeit at the opposite DNA face. This results in effective reversal of the second bend as evidenced by the increased inter-dye distance (i.e., reduced *E*FRET) at higher [Sox2HMG]. The final bend remains relatively unchanged upon further increase in Sox2 (up to 1 μM; Figure 2d). Overall, our smFRET data directly demonstrates multistep sequential DNA bending transitions dependent on Sox2 concentration.
