2.1.

Surface Modification of Glass Coverslips

Borosilicate coverslips ($22\times 22\text{-}\mathrm{mm}$ square, 0.18-mm thick, refractive index $n_D=1.52$) were prepared for nuclei deposition using various methods: soaking in Alconox detergent (10% in water, 2 h), sonication in 1 M KOH for 15 min followed by extensive rinsing with nanopure water and drying under a stream of nitrogen, or plasma cleaning (Harrick PDC-32G, 5 min).

2.2.

Mouse Embryonic Fibroblast Nuclei Preparation

Embryonic mouse fibroblasts (kindly provided by M. A. Kadhim, Oxford Brookes University, UK) were cultured in Dulbecco’s Modified Eagle Medium supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% L-glutamine. Nuclei were prepared by harvesting approximately one hundred million mouse embryonic fibroblast (MEF) cells and osmotically shocking them in fibroblast lysis buffer (12.5 mM Tris pH 7.4, 5 mM KCl, 0.1 mM spermine, 0.25 mM spermidine, 175 mM sucrose). Cell outer membranes were lysed by incubation in 0.2% Nonidet-P40. Nuclei were crosslinked using $100\mu \mathrm{g}/\mathrm{ml}$ psoralen (Trioxalen, Sigma). Nuclei were then lysed in spreading buffer (10 mM Tris pH 7.4, 10 mM ethylenediamine tetra-acetic acid, 0.05% sodium dodecyl sulfate, 1M NaCl, warmed to 37°C). Lysed nuclei solution containing $\sim \mathrm{70,000}$ nuclei was then deposited immediately onto cleaned coverslips. The deposition methods included dropcasting or spin-coating at various speeds (Plas-Labs/Headway Research PWM32; 250, 500, 750, and 1000 rpm with a distance from the center of rotation to the edge of the coverslip of ~10 mm), followed by evaporation, then fixing in methanol ($-20°\mathrm{C}$, 10 min), and then acetone ($-20°\mathrm{C}$, 1 min), followed by dehydration with stepwise ethanol rinses (70%, 95%, and 100% ethanol).

2.3.

Fluorescence In situ Hybridization

MEF-fixed coverslips were rehydrated in phosphate-buffered saline and the sample side placed down on a glass microscope slide with $35\mu \mathrm{l}$ of hybridization buffer: $0.1\mu \mathrm{M}$ peptide nucleic acid (PNA)-Cy5 probe (PNA Bio Inc., Cy5 -5′-CCCTAACCCTAACCCTAA-3′), 70% formamide, 0.5% blocking reagent (Roche, part # 11096176001), and 10 mM Tris pH 7.2. The hybridization chamber was incubated at 80°C in 100% humidity for 10 min and allowed to cool slowly to room temperature overnight in a humid box in the dark. Coverslips were removed from the hybridization chamber and rinsed twice with 70% formamide, 10 mM Tris pH 7.2, followed by three rinses with 100 mM Tris pH 7.2, 150 mM NaCl, and 0.08% Tween-20. Coverslips were dehydrated stepwise with 70%, 95%, and 100% ethanol and stored in the dark for imaging.

2.4.

Preparation of Hybridized Coverslips for Stochastic Optical Reconstruction Microscopy Imaging

Imaging chambers were prepared with hybridized coverslips by placing the coverslips sample-side down onto glass slides (Fisher 12-544-1) and sealing two sides with double-sided tape ($80\text{-}\mu \mathrm{m}$ thickness, 3 M). The oxygen-scavenging imaging buffer (50 mM Tris pH 8, 10 mM NaCl, 10% glucose, 100 mM beta-mercaptoethanol, $40\mu \mathrm{g}/\mathrm{ml}$ catalase, $300\mu \mathrm{g}/\mathrm{ml}$ glucose oxidase) was prepared immediately before use and added to the imaging chamber. The chamber was then sealed on the other two sides with nail polish.

2.5.

Direct Stochastic Optical Reconstruction Microscopy

Direct stochastic optical reconstruction microscopy (dSTORM) was used to acquire SR images of telomeres labeled with the PNA-Cy5 hybridization probe. dSTORM is a single-molecule localization microscopy (SMLM) variant able to reconstruct an SR (subdiffraction limit) optical image from stacks of diffraction-limited images of individual, spatially isolated organic fluorophore labels (in this case Cy5).¹³ Subdiffraction resolution is obtained by reconstructing the image from the centroid positions of the individual fluorophores that can be determined with a precision well below the diffraction limit. In SMLM, the areal density of the emitting labels must be limited so that the individual fluorophores are well resolved at the diffraction limit. With dSTORM, the label is photoswitched between long-lived dark and fluorescent states. With Cy5 this is done using a reducing buffer that reacts with the fluorophore in its triplet state (accessed by intersystem crossing from the first excited singlet state following 637-nm excitation of the singlet ground state) to form a long-lived dark (nonfluorescent) anionic radical. The dark anionic radical has an absorption near 400 nm that is used to return it to its neutral ground singlet state.¹⁴

The optical schematic of the imaging setup is shown in Fig. 1. Imaging of the PNA-Cy5-labeled telomeres was performed using this setup as described in detail elsewhere.¹⁵ The setup consisted of an inverted microscope (Olympus IX-71) modified for through-objective total-internal-reflection evanescent-wave excitation and activation at 637 and 405 nm, respectively. Incidence angles were greater than the critical angle ($\sim 61\mathrm{deg}$). The precise angle was not recorded but was chosen to optimize the signal-to-noise ratio in the image. The excitation beam was homogenized by expansion of the Gaussian excitation laser beam with a telescope and isolation of the central part of this using an iris. No corrections were made for under- and overcritical angle contributions to the images. The evanescent-wave excitation extended only $\sim 100\mathrm{nm}$ into the aqueous buffer substantially reducing the fluorescence background from the imaging buffer. Cy5 fluorescence emission was collected by the same objective (Olympus APON OTIRF, $\times 60$ 1.49 NA), isolated from the scattered excitation light using a 37-nm wide bandpass filter centered at 676 nm (Semrock) and imaged onto an electron-multiplying CCD camera (Princeton Instruments) through the left-hand camera port of the modified inverted microscope base. The camera was used at an electron-multiplying gain setting of 50. Focus ($z$) drift during image acquisition was reduced using a focus-lock system.

The diameters of the overlapped 405-nm activation and 635-nm excitation regions at the cover glass/buffer interface were $\sim 50\mu \mathrm{m}$. Activation and excitation powers were 2 and 4.5 mW, respectively. Computer-controlled shutters in the activation and excitation beam paths were used to control these beams during the activation and excitation (imaging) cycles. Image stacks were collected in groups of 25 successive 100-ms exposure images between activation cycles. To control the areal density of activated (fluorescent) Cy5 labels and to accommodate irreversible photobleaching of the Cy5 labels during the course of image acquisition, the activation time was increased exponentially as a function of cycle number from 50 ms at cycle 1 to $\sim 2500\mathrm{ms}$ at cycle 80. Stacks consisting of $\sim 1000$ to 2000 images (40 to 80 cycles) were collected from each of the examined regions of the samples. Under the excitation conditions used, spots due to single Cy5 molecule fluorescence were comprised of $\sim 100$ detected photons per 100 ms frame. The image stacks were analyzed using the DAOSTORM algorithm designed for spot localization in high spot density images.¹⁶ We used the ImageJ plugin, ThunderStorm,¹⁷ to reconstruct SR images from the DAOSTORM output (spot centroid $x-y$ positions and intensities). Lateral stage drift was corrected in ThunderStorm by cross-correlation of successive SR images reconstructed from successive substacks comprising an entire image stack. Lateral stage drifts during image acquisition were slowly varying and typically less than one camera pixel (106 nm) in $x$ and $y$. We chose to divide our data sets into five bins to capture the slow drift while maintaining good image statistics for the image cross-correlation drift-correction incorporated in Thunderstorm. YOYO-stained SR images of total DNA are shown in Fig. 2. PNA probe-stained telomere fiber images are shown in Fig. 3. The full-width at half maximum of the point-spread function was 420 nm.

Fig. 1

Optical schematic of the setup used for dSTORM imaging of Cy5-PNA-labeled telomeres.

Fig. 2

YOYO1-stained mouse MEF nuclei deposited at various spincasting speeds. (a) 250, (b) 500, (c) 750, and (d) 1000 rpm.

3.1.

Coverslip Cleaning Method

Testing of coverslip cleaning methods demonstrated that sonication in 1 M KOH yielded the best (cleanest and least autofluorescent as visualized microscopically) hydrophilic surface compared to plasma cleaning or Alconox soaking.

3.2.

Sample Deposition

Deposition methods on coverslips included dropcasting or spin-coating at 250, 500, 750, and 1000 rpm. Neither dropcasting nor spin-coating at 250 rpm was enough to spread the DNA, while both 750 and 1000 rpm appeared to overstretch and tear telomere fibers. Coverslips coated at 500 rpm yielded fibers that did not show signs of damage when YOYO-stained and observed microscopically (Fig. 2).

3.3.

Evaluation of Telomere Fiber Images.

Nuclei extracted from MEF cells were deposited onto coverslips at 500 rpm and stained with a PNA telomere probe. In contrast to YOYO1, a general DNA stain, the PNA probe is highly specific to telomere DNA allowing telomeres to be visualized against an $\sim 1500$-fold excess of nontelomeric DNA. Imaging by dSTORM revealed both t-loops and linear fibers, and each showed a wide range of spreading. There were smaller, compact structures and larger, stretched out structures. Few t-loops consisted of a continuous fiber, free of apparent gaps or breaks. In general the more stretched a structure, the more discontinuities it had. Examples of t-loops with different degrees of spreading are shown in Fig. 3.

Fig. 3

Classifying telomere morphology. Images of t-loops, and telomere fibers presumed to be t-loops, were obtained by dSTORM and placed into categories based on morphological characteristics. Scale bars are $5\mu \mathrm{m}$. (a) Small compact loops with few or no gaps. Fiber continuity allows confidence in scoring members of this group as t-loops. (b) Extended loops with numerous gaps. These structures can be scored as t-loops, although with lesser confidence than compact loops, because a chance arrangement of telomere fragments into a well-structured loop would be highly improbable. (c) Diffraction-limited image of the t-loop in panel b. (d and e) Ambiguous “loops” and diagram of alternative interpretations (open versus closed). This group has the general appearance of loops often with breaks in which one or more fragments have been displaced. Confidence in scoring these structures as t-loops is low. (f) Y-shaped structures. The only branch point in a telomere fiber occurs at the stem-loop junction of a t-loop. Therefore Y structures are likely to be broken loops. However, the chance contact between two linear fragments cannot be ruled out, and therefore, Y structures cannot be scored confidently as t-loops.

We classified t-loops into four groups according to degree of spreading and continuity of the fiber. The four groups are (1) small compact loops with few or no gaps [Fig. 3(a)], (2) loops with greater spreading and more numerous gaps [Fig. 3(b)], (3) ambiguous “loops” [Fig. 3(d)], and (4) Y-shaped structures that may be broken loops [Fig. 3(f)]. The first group contained the most convincing examples of t-loops. Members of the last three groups all contained discontinuities, and, therefore, categorization as a loop is somewhat subjective.