TRF1 and TRF2 form distinct shelterin subcomplexes at telomeres

SUMMARY

The shelterin complex protects chromosome ends from aberrant DNA repair and regulates telomerase access to telomeres. Shelterin is composed of six proteins (TRF1, TRF2, TIN2, TPP1, POT1, and RAP1) that can assemble into various subcomplexes in vitro, but their stoichiometry and dynamics in cells remain poorly understood. To quantitatively analyze shelterin function, we generated a panel of human cancer cell lines expressing HaloTagged shelterin proteins from their endogenous loci. We determined both the total and telomeric abundance of each subunit, demonstrating that shelterin proteins are present at telomeres in equal numbers. Using single-molecule live-cell imaging, we showed that TRF1-TIN2-TPP1-POT1 and TRF2-RAP1 form distinct subcomplexes that bind non-overlapping sites on telomeric chromatin. TRF1-TIN2-TPP1-POT1 tightly associates with telomeres, whereas TRF2-RAP1 binds more dynamically, facilitating the recruitment of co-factors that protect chromosome ends. Altogether, our work provides mechanistic insight into shelterin function in telomere maintenance and advances our understanding of telomeric chromatin architecture.

In brief

Using quantitative imaging in live human cancer cells, Janovič et al. demonstrate that shelterin assembles into two distinct subcomplexes at telomeres: a stable TRF1-TIN2-TPP1-POT1 unit and a more dynamic TRF2-RAP1 complex. These findings provide important insights into telomeric architecture and the molecular basis of shelterin-mediated telomere maintenance in cancer cells.

Graphical Abstract

INTRODUCTION

Telomeres are essential for genome stability, serving as protective caps at the ends of eukaryotic chromosomes. Human telomeres are composed of 2–15 kb of double-stranded (ds) repetitive DNA (TTAGGG) followed by a 3′ single-stranded (ss) overhang, 50–300 nucleotides in length.¹ Chromosome ends resemble DNA double-strand breaks and must be protected from inappropriate DNA damage response (DDR) recognition and DNA repair pathway activation. This protection is mediated by the human shelterin complex (Figure 1A), composed of six subunits (TRF1, TRF2, TIN2, TPP1, POT1, and RAP1), which specifically binds to telomeric DNA and counteracts the recognition of telomeres by the DNA repair machinery.^2–4 In addition, shelterin recruits telomerase to telomeres to compensate for incomplete DNA replication of chromosome ends. Telomerase is normally active in stem and germline cells but is silenced in somatic cells.⁵ In contrast, most cancers (~90%) reactivate telomerase to maintain telomere length, supporting their continuous proliferation.^6,7 The remaining 10% of tumors rely on alternative lengthening of telomeres (ALT), a homologous recombination-based process unique to cancer cells.^8,9 How the shelterin complex carries out these distinct functions, balancing the recruitment of telomerase or proteins involved in ALT, with restriction of access for DNA damage repair proteins remains a critical unaddressed question in the field.

Figure 1. Generation and functional characterization of HaloTagged shelterin cell lines.

(A) Model of the shelterin complex bound to telomeric DNA.

(B) CRISPR/Cas9-mediated N-terminal knockin of 3xFLAG-HaloTag with TEV site (top), and schematic of covalent Janelia Fluor (JF) labeling (bottom).

(C) Genomic PCR showing homozygous HaloTag insertion to shelterin genes in U2OS cells.

(D) In-gel fluorescence demonstrating the expression of HaloTagged shelterin proteins in U2OS cells.

(E–H) Western blots of U2OS cell lines expressing HaloTagged (E) TRF1, (F) TRF2, (G) TPP1, and (H) POT1 proteins. Non-specific bands (*); endogenous and tagged proteins (arrows); C1/C2: clones.

(I) IF-FISH of HaloTagged U2OS cell lines: HaloTag labeled with JFX650, 53BP1 IF for DNA damage (TIFs), PNA FISH for telomeres (inset shows colocalization). Scale bars, 10 μm.

(J and K) TIFs analysis of (J) U2OS and (K) HeLa cells (n > 70 cells, Mann-Whitney test).

(L and M) Growth curves of (L) U2OS and (M) HeLa HaloTagged shelterin cell lines.

(N and O) Flow-FISH telomere length analysis over time of (N) U2OS and (O) HeLa HaloTagged shelterin lines.

The shelterin subunits TRF1 and TRF2 are structurally closely related, both forming homodimers mediated by their TRFH domains that specifically bind to double-stranded telomeric DNA via their C-terminal MYB-domains.¹⁰ A major difference lies in a short region of disordered residues at their N terminus: TRF1 has acidic residues, while TRF2 contains basic ones.^11,12 Beyond dimerization, their TRFH domains facilitate interactions with other telomeric proteins. The TRFH domain of TRF1 binds a C-terminal peptide in TIN2, while the same region of TRF2 interacts with the nuclease Apollo and the endonuclease adaptor SLX4 in addition to binding to the C terminus of TIN2.^13–17 TRF2 can also associate with the TRFH domain of TIN2 via a short peptide in the hinge region connecting its TRFH and MYB domains (Figure 1A).^18,19 While TIN2 can associate with TRF2 in two different locations, its recruitment to telomeres in cells is largely mediated by TRF1.^20,21 It is therefore unclear how the interaction between TRF2 and TIN2 observed in vitro contributes to telomere maintenance in cells.^21–23

TRF1 and TRF2 carry out distinct functions at telomeres. TRF2 promotes the formation of telomeric loops (T-loops), where the single-stranded overhang invades the double-stranded telomeric DNA to create a lasso-like structure, which is crucial for preventing the activation of ATM signaling and suppression of non-homologous end joining (NHEJ).^24–28 In contrast, TRF1 facilitates telomere replication by recruiting BLM helicase and suppresses ATR signaling via the recruitment of TIN2, TPP1, and POT1.^21,29–31 POT1 specifically binds to the single-stranded 3′ overhang with its two OB-fold domains, preventing RPA loading.^29,32–34 In addition, POT1 tightly caps the ds-ss telomeric junction by recognizing the 5′-phosphate of the terminal nucleotide using its POT-hole structural feature within the first OB-fold domain.³⁵ Together, these properties of POT1 prevent the activation of ATR signaling at telomeres.^4,29 POT1 forms a heterodimer with TPP1, which in turn associates with the TRFH domain of TIN2 via a short peptide in its C terminus.^18,23,36 In addition to tethering POT1 to TIN2, TPP1 is also critical for telomerase recruitment to telomeres.^5,37 Thus, TIN2 represents the central hub of the shelterin complex connecting proteins associated with double-stranded telomeric DNA (TRF1, TRF2, and RAP1) to proteins bound to the single-stranded overhang (TPP1 and POT1).^15,18,38,39 Finally, RAP1 binds to TRF2 and modulates the TRF2’s interaction with telomeric DNA in vitro^40,41 and protects telomeres by inhibiting the end-joining function of DNA-PK.⁴² The shelterin complex is highly dynamic with extensive conformational variability, which has hindered high-resolution structural analysis of the entire complex, although structures of many individual folded domains of shelterin subunits have been determined.^43–45

In vitro reconstitutions have revealed a wide variety of possible shelterin assemblies, ranging from subcomplexes (e.g., TIN2-TPP1-POT1, TRF1, or TRF2-TIN2-TPP1-POT1) to a fully assembled dimeric complex, containing two copies of all six shelterin proteins.^44–47 Previous work in human cancer cells has suggested that TPP1 and POT1 are less abundant than TRF1, TRF2, and TIN2 and that the number of shelterin proteins associated with telomeres ranges from 65 molecules per telomere for POT1 to 740 molecules per telomere for TIN2. This uneven stoichiometry of shelterin subunits would imply that multiple subcomplexes exist rather than a single six-protein molecular entity.⁴⁸ However, the exact copy number of shelterin proteins at telomeres has not been directly determined, and how it is dynamically regulated to coordinate the different functions of the shelterin complex in telomere maintenance is unknown.

To define shelterin stoichiometry and dynamics at telomeres, we used genome editing⁴⁹ to insert HaloTag into all six shelterin genes in telomerase-positive (HeLa) and ALT (U2OS) cells. This allowed us to directly measure shelterin abundance at telomeres in intact cells. We found that 30–40 copies of each shelterin protein are represented at telomeres in HeLa cells, while the stoichiometry of shelterin at telomeres is altered in U2OS cells. Live-cell single-molecule imaging revealed that TRF2 and RAP1 are dynamically associated with chromosome ends, while the rest of the shelterin components are tightly telomere-bound. Additionally, our results indicate that TRF1 and TRF2 occupy distinct binding sites at telomeres. In total, our findings suggest that shelterin primarily exists in two functionally distinct subcomplexes and provide critical insights into overall telomeric architecture and the molecular mechanism by which shelterin orchestrates multiple essential functions to achieve telomere homeostasis.

RESULTS

Generation and functional characterization of cell lines expressing HaloTagged shelterin components

To analyze shelterin stoichiometry and dynamics in living human cells, we generated a panel of telomerase-positive (HeLa-EM2–11ht,⁵⁰ referred to as HeLa) and ALT (U2OS) cancer cell lines expressing HaloTagged shelterin proteins from their endogenous loci using genome editing.⁴⁹ We introduced a 3xFLAG-Halo-Tag-TEV (referred to as Halo) at the N terminus of TRF1, TRF2, TIN2, TPP1, and POT1 loci (Figures 1A and 1B). For TRF2 and TPP1, we targeted the major isoforms at amino acid positions 42 and 89, respectively. We also edited TRF2 and TIN2 loci in HeLa 1.3 cells, which have longer telomeres (~20 kb) compared to HeLa cells (~4 kb).

We confirmed homozygous insertions by genomic PCR and Sanger sequencing (Figures 1C and S1A), and we validated the protein expression using in-gel fluorescence (Figures 1D and S1B) and western blotting (Figures 1E–1H and S1C–S1E). We successfully generated clonal HaloTagged TPP1 lines in HeLa cells, but they stopped proliferating shortly after clonal isolation (Figures S1D and S1E), suggesting that the N-terminal tag on TPP1 affected telomerase function.^51,52 We also could not validate TIN2 expression by western blotting due to lack of a reliable antibody. However, in-gel fluorescence (Figures 1D and S1B) and genomic PCR (Figures 1C and S1A) demonstrated the expression of Halo-TIN2 at the expected size and targeting of all alleles, respectively. All HaloTagged shelterin proteins, besides Halo-TIN2, were expressed at similar levels to their untagged counterparts, and HaloTagging did not affect the expression of other shelterin components (Figures 1E–1H and S1C–S1E). The introduction of HaloTag at the N terminus allowed us to detect protein isoforms that resulted from alternative splicing downstream of the site of tag insertion. For TRF2, we detected two distinct bands with similar abundance by fluorescence gels in all cell lines (Figures 1D and S1B). Consistent with previous findings, we also observed up to four TIN2 isoforms (Figures 1D and S1B).⁵³

To assure that HaloTagging preserved protein function, we confirmed that the tagged shelterin proteins localized to telomeres and that their expression did not lead to the formation of telomere-dysfunction-induced foci (TIF), marked by 53BP1 (Figures 1I–1K and S1F). We further assessed functionality by monitoring cell growth over 40 days and analyzed telomere length using flow cytometry coupled with telomeric PNA FISH (Flow-FISH). All cell lines had comparable growth rates (Figures 1L and 1M), and telomere length was stable over time in all U2OS cell lines (Figure 1N). In HeLa cells, telomere length was constant in Halo-TRF1, Halo-TRF2, and Halo-TIN2 lines but increased slightly in Halo-POT1 line (Figure 1O). A similar phenotype was observed in cancer cell lines expressing POT1 mutants that alter the DNA binding affinity of POT1,⁵⁴ suggesting that the HaloTag likely alters POT1’s DNA-binding properties. However, HeLa cells expressing Halo-POT1 did not display defect in cell growth (Figure 1M) or telomere protection (Figures 1K and S1F), and HaloTagging POT1 in U2OS cells did not lead to telomere elongation (Figure 1N), indicating that Halo-POT1 interferes with telomerase pathway in HeLa cells. As a second approach, we measured median telomere length in HaloTagged shelterin HeLa cells using interphase Q-FISH and compared it to the parental cell line (Figure S1G). The results were consistent with those obtained using Flow-FISH, showing slight telomere elongation in Halo-TRF1 and Halo-TIN2 and substantial elongation in Halo-POT1 cells. Since telomere lengths remained stable over time in the Halo-TRF1, Halo-TRF2, and Halo-TIN2 lines (Figure 1O), their slight increase in telomere length likely reflects clonal variation rather than altered shelterin function.

Collectively, these results demonstrate that we have successfully generated a panel of telomerase-positive and ALT cancer cell lines expressing HaloTagged shelterin components from their endogenous loci. While the HaloTagged shelterin components are fully functional in chromosome end protection, tagging POT1 in HeLa cells increased telomere length over time—an important caveat when interpreting subsequent results.

Total cellular abundance of shelterin proteins

To quantify the cellular abundance of HaloTagged shelterin proteins, we used an in-gel fluorescence approach previously described.^55,56 HaloTagged proteins were quantitatively labeled with JF657, cell lysates were analyzed by SDS-PAGE, and the fluorescence intensity of the HaloTagged shelterin proteins was compared to recombinant HaloTag standard (Figures 2A and 2B). The standard was supplemented with cell lysate from a known number of U2OS or HeLa cells. Standard curves for both HaloTag and total protein stain signal were linear over a broad dynamic range, which allowed us to quantitatively measure the total cellular abundance of HaloTagged shelterin proteins (Figures S2A and S2B). To minimize the impact of protein composition on SDS-PAGE, we treated lysates with TEV protease to cleave the HaloTag from shelterin proteins, correcting for variation in banding patterns (Figures S2C–S2E). Since TRF2 was tagged after the start codon of its major isoform, which includes an alternative start site adding 41 amino acids to the N terminus, TEV digestion produced two distinct bands containing HaloTag (Figures S2C and S2D). Our results demonstrate that TRF2 is the most abundant shelterin protein in U2OS (~60,000 proteins/cell) and HeLa (~85,000 proteins/cell) cells (Figures 2C and 2D). TRF1 is present at ~10,000–15,000 proteins/cell in both lines (Figures 2C and 2D). In HeLa cells, TIN2 and POT1 were more abundant (~35,000 copies/cell) than in U2OS cells (~10,000 and ~12,000, respectively) (Figures 2C and 2D). TPP1 was expressed at ~20,000 proteins/cell in U2OS cells (Figure 2C), and in HeLa 1.3 cells, both TRF2 and TIN2 were expressed at ~40,000 proteins/cell (Figures S2F and S2G).

Figure 2. Total cellular abundance of HaloTagged shelterin proteins.

(A and B) In-gel fluorescence comparing Halo-shelterin proteins in (A) U2OS and (B) HeLa cells to a known amount of fluorescent 3x-FLAG-HaloTag standard.

(C and D) Quantification of total cellular HaloTagged shelterin abundance in (C) U2OS and (D) HeLa cells (N = 3, mean ± SD).

The shelterin abundance in HeLa cells reported here is comparable to those determined by Takai et al. using quantitative western blotting (Table S1).⁴⁸ While shelterin levels in U2OS cells were similar to HeLa cells, the TIN2 abundance was considerably lower in U2OS cells. Since we could not verify that Halo-TIN2 is expressed at endogenous levels, our measurements may underestimate TIN2 abundance. Altogether, we measured the total cellular abundance of HaloTagged shelterin factors in U2OS, HeLa, and HeLa 1.3 cells and consistently found TRF2 to be the most abundant shelterin subunit.

Shelterin abundance and stoichiometry at telomeres

While the total cellular abundance provides information regarding overall shelterin stoichiometry, it does not report on the shelterin complex composition at telomeres. To determine shelterin copy number at individual telomeres, we carried out fluorescence photobleaching experiments (Figures 3A–3C, Video S1). We quantitatively labeled the HaloTagged shelterin proteins with JFX650, fixed cells, and immediately imaged them using high laser power. Time traces, showing stepwise photobleaching of shelterin telomeric foci, allowed us to measure the intensity of single fluorophores (Figures 3B and 3C), which we used to convert initial fluorescence signals into protein copy numbers. To eliminate contributions from DNA replication or telomere overlap from sister chromatids, cells were synchronized using a double thymidine block at the beginning of S phase (Figure S3A). The number of telomeric foci detected per nucleus were slightly lower than the expected number of chromosome ends in the respective cell line (e.g., ~125 foci compared to 140–150 chromosome ends in HeLa cells; Figures S3B–S3D). The number of foci detected using fluorescent HaloTag labeling in HeLa cells closely matched PNA FISH results (Figure S3E), suggesting that some telomeric signals are not detected because they are below the detection limit or because multiple telomeric signals spatially overlap. U2OS cells showed fewer shelterin telomeric foci (~100 foci; Figure S3D), consistent with increased telomere clustering, for instance due to PML body formation.⁵⁷

Figure 3. Shelterin protein abundance at telomeres.

(A) Representative images of HaloTagged shelterin proteins in HeLa, HeLa 1.3, and U2OS cells labeled with JFX650. Scale bars, 5 μm.

(B) Fluorescence time trace of a Halo-TRF1 signal showing single-step photobleaching events.

(C) Analysis of single fluorophore intensity by Gaussian fitting of photobleaching step histograms. Inset shows combined analysis of multiple traces.

(D and E) Quantification of telomeric protein number in (D) HeLa, HeLa 1.3, and (E) U2OS cells by dividing telomeric intensity by single fluorophore intensity (N = 3, median ± SD; see Figures S3F and S3G for replicate distribution).

In HeLa cells, the telomeric shelterin abundance ranged from 25 to 35 copies for TRF1, TRF2, and TIN2 to around 40 copies for POT1 per telomere (Figure 3D). The increased POT1 telomeric copy number potentially reflects the longer telomeres in Halo-POT1 HeLa cells compared to other HeLa cell lines (Figures 1O and S1G). Therefore, our results indicate that all shelterin factors are present at telomeres in approximately equal numbers. In particular, for TRF1, TRF2, and TIN2, the telomeric protein numbers determined here are substantially lower than previous estimates based on chromatin fractionation (Table S1).48 In HeLa 1.3 cells, TRF2 and TIN2 were detected at ∼40 copies/telomere (Figure 3D), despite a 5-fold increase in telomere length (∼20 kb vs. ∼4 kb). This modest increase suggests that shelterin density does not scale with telomere length, supporting a sparser distribution on longer telomeres.⁴⁸ In U2OS cells, TRF1 and TRF2 telomeric copy numbers were slightly higher (~40 copies/telomere) than in HeLa, while TIN2 was markedly lower (~10 copies/telomere) (Figure 3E). The reduced TIN2 telomeric count likely reflects lower expression of Halo-TIN2 compared to endogenous protein. TPP1 was more abundant (~60 copies/telomere) than TIN2, while ~30 copies of POT1 localized to telomeres in U2OS cells (Figure 3E).

In total, these observations suggest that HeLa telomeres harbor on average ~30 copies of each shelterin factor, while U2OS telomeres show variable stoichiometry, particularly reduced TIN2 levels, which could be an underestimation.

TRF2 and RAP1 display faster dynamics at telomeres than the other shelterin factors

To analyze the biophysical and biochemical properties of shelterin proteins in living cells, we performed live-cell single-molecule imaging using HILO (highly inclined and laminated optical sheet) microscopy⁵⁸ and sparse protein labeling. This approach allowed us to track individual HaloTagged shelterin factors over time in living cells (Figures 4A and 4B, Videos S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, and S13). In all single-molecule videos, we observed mobile and static particles (Videos S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, and S13). To confirm that static molecules were telomere-bound, we first labeled a small fraction of HaloTagged shelterin proteins with JFX650 for single-molecule detection. Immediately afterward, we labeled the remaining HaloTagged shelterin proteins with a high concentration of the spectrally distinct JF549 to visualize bright telomeric foci (Figure 4C). This analysis revealed that almost all static shelterin molecules co-localized with telomeric foci, indicating telomere association (Figure 4D; Videos S14, S15, S16, and S17).

Figure 4. Single-molecule analysis of shelterin protein dynamics.

(A) Illustration of HILO microscopy and sparse labeling for single-molecule live-cell imaging.

(B) Intensity trace of a representative Halo-TRF1 molecule statically associated with telomere in U2OS cells.

(C) Strategy combining sparse single-molecule labeling (magenta) with dense labeling (green) of the same HaloTagged protein to mark telomeres.

(D) Representative HeLa cell showing sparse (JFX650, magenta) and dense (JF549, green) Halo-TRF1 labeling; kymograph of boxed region below. Scale bars, 5 μm.

(E) Two-state model used to analyze single-particle tracking data.

(F–H) Quantification of the fraction of static molecules (top) and diffusion coefficients (bottom) of static HaloTagged shelterin components in (F) U2OS, (G) HeLa, and (H) Halo-TRF1, Halo-TRF2, Halo-RAP1, and Halo-TRF2 in RAP1 knockout U2OS cells based on two-state Spot-On fitting (N = 3, ≥30 cells/replicate, mean ± SD, one-way ANOVA with Tukey’s post-hoc test).

We analyzed single-particle trajectories using two methods: the Spot-On algorithm⁵⁹ (based on mean-square displacement) and the Bayesian HMM (Hidden Markov Model)-based Ex-Track.⁶⁰ Both approaches measured the fraction of immobile molecules, presumably telomere- or chromatin-bound, and the diffusion coefficients of mobile and static states, reflecting their rate of movement (Figure 4E). Both methods require a predefined number of molecular states. The Spot-On fitting of the cumulative distribution functions of single-particle displacements fit well to a two-state model that assumes a freely diffusing and telomere-bound state (Figure S4A). The same fit also closely matched the probability density function of jump sizes (Figures S4B and S4C), indicating that a two-state model is sufficient to describe the experimentally determined step size distributions. In U2OS cells, Spot-On analysis revealed that the majority (>60%) of TRF1, TIN2, TPP1, and POT1 were immobile and thus likely telomere-bound (Figure 4F). In contrast, <30% of TRF2 was static, and most TRF2 molecules were freely diffusing (Figure 4F). In HeLa cells, immobile fractions were lower overall, likely due to shorter telomeres, and TRF2 again showed the lowest static fraction among shelterin proteins (Figure 4G). Similarly, Halo-TRF2 exhibited lower static fraction compared to Halo-TIN2 in HeLa 1.3 cells (Figure S4F).

Analysis of the diffusion rates revealed that static TRF1, TIN2, TPP1, and POT1 had comparable diffusion coefficients (D_bound), while TRF2 was more mobile when telomere-bound (Figures 4F, 4G, and S4F). Notably, the diffusion coefficient of freely diffusing particles (D_free) was similar across all shelterin proteins and cell lines (Figures S4D–S4F). These observations suggest that TRF1, TIN2, TPP1, and POT1 form a complex at telomeres, while TRF2 associates more loosely with chromosome ends.

To verify that the distinct telomeric dynamics of TRF2 were not caused by the HaloTag fusion, we tagged RAP1 at its endogenous locus in U2OS cells (Figures S4G–S4I). In addition, we knocked out RAP1 in Halo-TRF2-expressing U2OS and HeLa cells (Figures S4H and S4I). The total cellular abundance of Halo-RAP1 (~45,000 proteins per cell) was similar to that of Halo-TRF2 in U2OS cells (Figure S4J). Since RAP1 recruitment to telomeres depends on its association with TRF2, we expected RAP1 to have similar dynamics as TRF2. Indeed, RAP1’s immobile fraction and D_bound matched TRF2 and differed significantly from TRF1 (Figures 4H and S4K; Video S18). RAP1 displaying the same dynamics as TRF2 is consistent with their constitutive complex formation and confirms that the diffusion dynamics of TRF2 at telomeres are distinct from those of TRF1, TIN2, TPP1, and POT1. RAP1 knockout had no effect on TRF2 diffusion (Figures 4H and S4K; Video S19), aligning with previous in vitro observations.⁶¹ ExTrack analysis further confirmed both TRF2’s lower immobile fraction and higher telomeric diffusion compared to the other shelterin components (Figures S4L and S4M).

For all single-particle tracking analyses, we included only trajectories from nuclear-localized molecules, identified with DNA staining. TIN2, TPP1, and POT1 showed cytoplasmic signals, while TRF1 and TRF2 were almost exclusively nuclear (Videos S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, and S13). Quantification of the cytoplasmic fraction revealed that ~50% of TIN2, TPP1, and POT1 were found in the cytoplasm in U2OS and HeLa cells (Figure S4N), which is consistent with previous reports.⁶² We therefore adjusted the total protein abundance accordingly to determine nuclear protein levels (Figure S4O). TRF2 was significantly more abundant in the nucleus across all cell lines than other shelterin proteins, which were present in equal amounts (Figure S4O). Together, our single-molecule experiments suggest that the shelterin complex largely exists in two distinct subcomplexes at telomeres with different dynamic properties composed of TRF1-TIN2-TPP1-POT1 and TRF2-RAP1.

TRF2 and RAP1 have lower affinity for telomeres than TRF1 and the remaining shelterin subunits

To analyze the biochemical properties of shelterin components, we determined the residence times of HaloTagged shelterin proteins at telomeres using single-molecule imaging in living cells. Proteins were sparsely labeled with JF657 and imaged under low laser power with a 100 ms exposure at 0.5 or 1.5 s intervals to minimize photobleaching (Figure 5A). Residence times were derived from trajectory length distributions of static particles (Figures 5B and 5C; Video S20). Survival distributions fit well to two-exponential decay functions (Figure S5A), potentially reflecting binding to telomeric vs. non-telomeric chromatin, or distinct telomeric binding sites with different dissociation rates. In U2OS cells, the time constant (average association time) of long-lived binding events was shorter for Halo-TRF2 (62 s), Halo-RAP1 (57 s), and Halo-TRF2 in RAP1 knockout cells (54 s) compared to Halo-TRF1 (93 s), Halo-TIN2 (91 s), Halo-TPP1 (100 s), and Halo-POT1 (104 s) (Figure 5D). In HeLa cells, the time constant for long-lived binding events was comparable across all proteins (~60 s, Figure 5E). Importantly, survival distributions from imaging using 0.5 s intervals closely matched the data generated in 1.5 s intervals, demonstrating that photobleaching is not affecting this analysis (Figures S5B and S5C).

Figure 5. Single-molecule analysis of shelterin residence times at telomeres.

(A) Kymographs of HaloTagged shelterin proteins labeled with JF657, acquired at 1.5 s intervals.

(B and C) Survival distributions of static shelterin molecules in (B) U2OS and (C) HeLa cells from single-particle tracking of immobile molecules.

(D and E) Short-lived (τ₁) and long-lived (τ₂) binding time constants derived from two-exponential fits of survival curves in (D) U2OS and (E) HeLa cells (N = 3, ≥10 cells/replicate, mean ± SD, one-way ANOVA with Tukey’s post-hoc test, all comparisons vs. TRF2).

(F and G) Transition rates from bound to freely diffusing state for HaloTagged shelterin proteins in (F) U2OS and (G) HeLa cells, calculated from high-speed tracking data using ExTrack (N = 3, ≥30 cells/replicate, mean ± SD, one-way ANOVA with Tukey’s post-hoc test).

As a second approach, we used ExTrack to determine transition rates from the bound to freely diffusing state based on fast single-molecule imaging (Figure 4). Halo-TRF2 dissociation rates were 4–5 times higher in U2OS and 2–3 times higher in HeLa cells compared to other shelterin factors (Figures 5F and 5G). The greater sensitivity of ExTrack to short-lived interactions, due to its 100 fps acquisition rate, likely explains the stronger difference relative to residence time analysis. Together, these experiments suggest that TRF2 and RAP1 bind telomeric DNA with lower affinity than the other shelterin proteins analyzed, providing evidence that shelterin forms two distinct subcomplexes at telomeres: TRF1-TIN2-TPP1-POT1 and TRF2-RAP1.

TRF1 and TRF2 form distinct subcomplexes that occupy different telomeric binding sites

To further test the model of two distinct shelterin subcomplexes at telomeres, we examined the recruitment hierarchy of shelterin proteins to telomeres. Previous work by Frescas et al. demonstrated that TIN2 is largely recruited by TRF1,²¹ supporting our model of separate TRF1- and TRF2-based subcomplexes. We first immunopurified endogenously tagged shelterin proteins using a 3xFLAG-tag in U2OS cells and quantified the amount of TRF1 and TRF2 co-purified (Figure 6A). To control for purification efficiency across samples, we supplemented cell lysates with recombinant FLAG-HaloTag protein and fluorescently labeled the mixture with a JF657 during immuno-purification. This approach allowed us to quantify the relative association of TRF1 and TRF2 with HaloTagged TIN2, TPP1, and RAP1. As expected, TRF2 strongly co-purified with Halo-RAP1 (~7-fold more compared to Halo-TIN2 or Halo-TPP1), while TRF1 was ~5-fold more enriched with Halo-TIN2 and Halo-TPP1 than with Halo-RAP1 purifications (Figure 6B).

Figure 6. Recruitment hierarchy of shelterin proteins to telomeres.

(A) Co-IP of endogenous TRF1 and TRF2 from HaloTagged TIN2, TPP1, and RAP1 U2OS cells using FLAG pull-down. Lysates were spiked with recombinant FLAG-HaloTag protein and JF657 HaloTag-ligand prior to incubation with FLAG resin. I, input; FT, flow-through; B, bound.

(B) Quantification of (A), showing relative association of Halo-TIN2, Halo-TPP1, and Halo-RAP1 with TRF1 or TRF2, normalized to Halo-TIN2 (N = 3, mean ± SD, one-way ANOVA with Tukey’s post-hoc test).

(C) In-gel fluorescence (top) and anti-FLAG western blot (bottom) of HeLa Halo-TRF1 and Halo-TRF2 lysates, untreated or treated with Halo-PROTAC for 18 h. Cells were labeled with JFX650 prior to analysis.

(D and E) Representative live-cell images of (D) Halo-TRF1 or (E) Halo-TRF2 HeLa cells transiently expressing mNeonGreen-tagged TIN2, POT1, RAP1, and TRF1/2. Cells were treated ±Halo-PROTAC for 18 h and labeled with JFX650 before imaging. Scale bars, 5 μm.

(F and G) Quantification of the median number of mNeonGreen foci/cell in (F) Halo-TRF1 and (G) Halo-TRF2 cells from (D and E), (N = 2–3, ≥40 cells/replicate, mean ± SD, one-way ANOVA with Tukey’s post-hoc test).

(H and I) Representative live-cell images of (H) Halo-TRF1 or (I) Halo-TRF2 HeLa cells, labeled with JFX650, untransfected or transfected with mNeonGreen-tagged TRF1 or TRF2, respectively. Scale bars, 5 μm.

Next, we used a Halo-PROTAC ligand, which, upon conjugation with the HaloTag, rapidly targets the HaloTag and its fusion partner for proteasomal degradation in living cells (Figure 6C).⁶³ To analyze the telomere recruitment of the other shelterin proteins, we transiently expressed the mNeonGreen fusions of the shelterin factors in HeLa and U2OS cells treated with the Halo-PROTAC ligand. As expected, after Halo-TRF1 degradation, TIN2 and POT1 were displaced from telomeres, while TRF2 recruitment was unaffected (Figures 6D, 6F, S7A, and S7D). RAP1 levels were also reduced after TRF1 degradation (Figures 6D and 6F). In contrast, TRF2 degradation had no detectable effect on TRF1 or TIN2 and only a minor decrease in POT1 recruitment (Figures 6E, 6G, S7B, and S7E). Only RAP1 recruitment was eliminated by TRF2 degradation, consistent with TRF2-RAP1 forming one complex (Figures 6E and 6G). Similar to TRF1 depletion, degradation of TIN2 eliminated POT1 recruitment and reduced RAP1 localization to telomeres (Figures S6A, S6B, S7C, and S7F). These results support a model where TRF1 recruits TIN2-TPP1-POT1, and TRF2 recruits RAP1. Moreover, TIN2 facilitates RAP1 recruitment at endogenous TRF2 levels, but TRF2 overexpression can compensate for TIN2 loss. However, the overexpression of mNeonGreen-RAP1, combined with already high levels of endogenous RAP1 in the nucleus, likely masks small changes in telomere-bound mNeonGreen-RAP1. Specifically, the diffuse nuclear background limits the detection of low-intensity mNeonGreen-RAP1 foci, which is further supported by the lower number of mNeon-Green-RAP1 foci detected compared to other shelterin proteins (Figures 6F and 6G). The dependence of RAP1 recruitment to telomeres on TIN2 was also documented in previous work.⁶⁴ Importantly, we detected fewer mNeonGreen-POT1 foci in U2OS cells expressing Halo-TIN2 compared to the other Halotagged shelterin cell lines (Figures S7D–S7F). Because POT1 recruitment depends on its association with TIN2, this indicates that HaloTagging TIN2 in U2OS cells likely reduces TIN2 expression levels, which in turn limits its loading on telomeres and reduces the number of recruited POT1 molecules.

During these experiments, we observed that overexpression of mNeonGreen-TRF1 displaced Halo-TRF1 from telomeres in control cells not treated with the Halo-PROTAC (Figures 6H and S6E). Similarly, expression of mNeonGreen-TRF2 removed Halo-TRF2 from telomeres (Figures 6I and S6F). However, expression of mNeonGreen-TRF1 did not displace Halo-TRF2 from telomeres unless it was expressed at very high levels (Figures 6E and S6C), and mNeonGreen-TRF2 expression did not affect Halo-TRF1 localization to telomeres (Figures 6D and S6C). These observations suggest that TRF1 and TRF2 occupy distinct binding sites on telomeric DNA.

Since TRF1 and TRF2 can directly bind to telomeric DNA, but also associate with TIN2, it is possible that their indirect recruitment via TIN2 would still allow them to bind to telomeres if all binding sites on telomeric DNA are occupied. To test the contribution of TIN2 to the telomeric recruitment of TRF1 and TRF2, we expressed TRF1^F142A and TRF2^L362R mutants that cannot bind TIN2.^15,18 The expression of mNeonGreen-TRF1^F142A displaced endogenous Halo-TRF1 and Halo-TIN2 from telomeres (Figures S6E and S6G), confirming TIN2-independent telomeric DNA binding. Similar to wild-type TRF1, only very high overexpression of mNeonGreen-TRF1^F142A removed Halo-TRF2 from telomeres (Figure S6D), further supporting that TRF1 and TRF2 occupy distinct binding sites on telomeric chromatin. The expression of mNeonGreen-TRF2^L362R failed to displace Halo-TRF2 from telomeres (Figure S6F), suggesting that TIN2 supports TRF2’s optimal recruitment to telomeres. Additionally, even at high levels, mNeonGreen-TRF2^L362R did not affect Halo-TIN2 signal at telomeres (Figure S6G), reinforcing that TIN2 is primarily recruited by TRF1.

Overall, these observations are consistent with TRF1-TIN2-TPP1-POT1 and TRF2-RAP1 forming distinct sub-complexes, with TRF1 and TRF2 occupying distinct binding sites on telomeric chromatin. However, TRF2-RAP1 requires the interaction with TIN2 in order to accumulate at telomeres.

DISCUSSION

The data presented in this study provide a precise quantification of shelterin composition at telomeres in human cancer cells. Using genome-edited cell lines expressing HaloTagged shelterin components from their endogenous loci, we measured the abundance of shelterin proteins at chromosome ends, defined their diffusion dynamics, and demonstrated that TRF1 and TRF2 occupy distinct binding sites in vivo, despite both binding telomeric repeat DNA in vitro. Moreover, the HaloTagged shelterin cell lines represent a valuable resource for future telomere research.

Overall architecture of telomeres in cancer cells

To define the molecular architecture of telomeres, we determined the total cellular abundance of the shelterin proteins and their copy number at telomeres. Our measurements in HeLa and HeLa 1.3 cells align well with previous observations (Table S1).⁴⁸ While the abundance of shelterin components in U2OS cells was largely comparable to HeLa cells, the level of Halo-TIN2 in U2OS cells was significantly lower than in HeLa cells.⁴⁸ Since we were unable to confirm that Halo-TIN2 is expressed at similar levels as endogenous TIN2, it is possible that Halo-TIN2 is underexpressed in U2OS cells. However, TIN2 abundance was not previously determined in ALT cells. Our Halo-TIN2 cell lines did not have a growth defect, and we did not detect telomere-dysfunction-induced foci, which demonstrates that Halo-TIN2 is fully functional in telomere end protection. While we observed predominantly nuclear localization of TRF1 and TRF2, a substantial fraction (~50%) of TIN2, TPP1, and POT1 was localized in cytoplasm, consistent with the TPP1-POT1 heterodimer being imported into nucleus via TIN2.⁶² Therefore, the core shelterin components TRF1, TIN2, TPP1, and POT1 are present in equimolar amounts in the nucleus, while TRF2 and RAP1 levels are ~5-fold higher.²³

Our fluorescence photobleaching approach allowed us to determine the shelterin copy number at telomeres. In HeLa cells, we observed a median of ~30 copies of each shelterin protein per telomere. Considering a telomere length of 4,000 nucleotides⁶⁵ and a nucleosome spacing of 157 bp ( 27 nucleosomes per telomere),⁶⁶ this suggests that one TRF1 and TRF2 dimer bind every two telomeric nucleosomes (Figure 7). Although U2OS and HeLa 1.3 cells have 4- to 5-fold longer telomeres than HeLa cells, we only observed a modest increase in shelterin abundance (~40 copies per telomere), suggesting a lower shelterin density on long telomeres. This reduced shelterin density could affect telomerase recruitment, since association with an internally bound TPP1 could limit telomerase access to chromosome end. Additionally, the telomeric zinc-finger-associated protein (TZAP), known to load on long telomeres to restrict their length,⁶⁷ may contribute to this regulation. Importantly, in ALT cells (U2OS), shelterin loading onto telomeres could also be influenced by decompacted chromatin.⁶⁸ In U2OS cells, we detected a median of 10 Halo-TIN2 molecules/telomere, substantially lower than TPP1 (~60) and POT1 (~30). As outlined above, Halo-TIN2 may be expressed at lower levels than endogenous TIN2, likely leading to underestimation of its telomeric abundance. Supporting this interpretation, mNeonGreen-POT1 loading onto telomeres was reduced in Halo-TIN2 cells compared to the other U2OS lines.

Figure 7. Model of telomere architecture in HeLa cells with short telomeres.

The shelterin complex associate with telomeres in two distinct subcomplexes: TRF1-TIN2-TPP1-POT1 binds nucleosome entry/exit sites, while TRF2-RAP1 likely binds the linker DNA. TRF2 may transiently associate with TIN2 to stabilize its interaction with the chromosome end. Both sub-complexes are dimeric and present in equal numbers, binding approximately every other nucleosome.

The telomeric copy numbers of the shelterin components determined here are up to 50-fold lower than previously reported values (e.g., 65 for POT1, 720 for TIN2), calculated by measuring the absolute cellular abundance and chromatin-bound fractions (Table S1).⁴⁸ While HaloTag fusion could reduce telomere association, it is unlikely that all shelterin proteins would be impacted equally. Cell-cycle synchronization using thymidine block may also influence shelterin localization. Our live-cell single-molecule imaging indicate that a substantial nuclear fraction (60%–80% in HeLa cells) of the shelterin proteins is not chromatin-bound, suggesting that previous fractionation-based estimates may overestimate telomeric association. Importantly, calculating the telomeric copy number of TRF1 using the nuclear protein level and chromatin-bound fraction (~10,000 molecules per cell * 46% fraction bound/142 telomeres) results in 33 TRF1 molecules per telomere, which closely matches our photobleaching results and increases our confidence in our findings. Finally, the shelterin density on telomeres in HeLa cells determined here aligns well with in vitro estimates of approx. one shelterin per 100 bp proposed by Erdel et al.,⁶¹ although those experiments were carried out in the absence of nucleosomes.

TRF1 and TRF2 form subcomplexes with distinct telomeric binding modes

Many studies have addressed the biochemical basis of shelterin complex formation. When all subunits (TRF1, TRF2, RAP1, TIN2, TPP1, and POT1) are combined, shelterin can form a fully dimeric complex.⁴⁴ Similarly, a 2:2:2 stoichiometry has been observed for TRF1, TIN2, and TPP1.⁴⁷ In contrast, TRF2-RAP1 forms a 2:1 complex with the remaining shelterin subunits, containing a TRF2-RAP1 dimer and a single copy of TIN2, TPP1, and POT1.⁴⁶ In cells, our data, along with previous findings,²¹ demonstrate that TRF1 recruits TIN2 to telomeres, whereas TRF2 does not. Therefore, TIN2 is primarily tethered to telomeres via its interaction with the TRFH domain of TRF1. Consequently, TRF2’s TRFH domain rarely, if ever, interacts with TIN2 at telomeres, as its binding site is already occupied by TRF1. This leaves TRF2’s TRFH domain free to interact with other co-factors, such as Apollo or SLX4, to dynamically recruit proteins critical for chromosome end protection.^13,14,16,17

In HeLa cells, the number of telomere-associated TIN2 and POT1 molecules matches that of TRF1 and TRF2, consistent with a fully dimeric shelterin complex. Our overexpression experiments demonstrate that TRF1 or TRF2 can displace themselves from telomeres but cannot outcompete each other, strongly suggesting their distinct binding modes to telomeric chromatin. Structural studies provide a possible explanation: the MYB domain of TRF1 specifically binds entry or exit sites of telomeric nucleosomes, whereas TRF2 does not bind in similar manner.^47,69 Thus, TRF2 cannot occupy TRF1’s binding sites, explaining its inability to displace TRF1. Furthermore, our results indicate that TRF1 does not require TIN2 for stable recruitment to telomeres. In contrast, TRF2 recruitment to telomeres depends on its interaction with TIN2.^22,23,36,70 Disrupting the TRF2-TIN2 interaction using TRF2^L362R mutant substantially reduced the TRF2’s ability to displace wild-type TRF2 from telomeres, demonstrating that both DNA binding and TIN2 interaction are essential for TRF2 recruitment to telomeres. These observations also suggest that the MYB domain of TRF2 has a lower affinity for telomeric chromatin than that of TRF1, consistent with in vitro experiments.^69,71

Our single-molecule imaging experiments further support the distinct telomeric binding modes for TRF1 and TRF2. We demonstrate that TRF2-RAP1 is more dynamic at telomeres than TRF1-TIN2-TPP1-POT1. This difference is amplified in U2OS cells, where decompacted chromatin⁶⁸ may facilitate more rapid sampling of binding sites by TRF2. We also observed higher dissociation rate from telomeres for TRF2-RAP1 compared to TRF1-TIN2-TPP1-POT1, and shorter residence times for TRF2 in U2OS cells, which may reflect the altered chromatin state of ALT telomeres. Additionally, the distinct shelterin dynamics in HeLa compared to U2OS cells could also be a result of differences in telomere length. U2OS cells have longer telomeres, providing more shelterin binding sites, which could increase residence time by favoring transitions between telomeric binding sites over full dissociation. Finally, ALT-specific features, such as PML bodies, increased DNA damage signaling, and extensive ssDNA overhangs, may each impact shelterin dynamics, which will be important to address in future work.

Altogether, our observations suggest that TRF1-TIN2-TPP1-POT1 form a dimeric complex, which tightly associates with telomeric DNA. In contrast, TRF2-RAP1 complex is more dynamically bound, potentially repositioning rapidly between telomeric binding sites. Consistent with a more dynamic interaction, the affinity of the TIN2’s TRFH domain for the TRF2’s hinge region is in the low micromolar range,¹⁸ which is substantially less stable than the TRF1-TIN2 interaction.¹⁵ TRF2-RAP1 may thus rapidly scan telomeric DNA by forming transient interactions with both telomeric DNA and TIN2 (Figure 7). Such dynamic behavior would allow TRF2 to drive T-loop formation,^{25–28,72,73} recruit Apollo for chromosome end resection,^15,16,74 or to prevent NHEJ at blunt telomeres.^{4,42,73,75,76}

In summary, our results demonstrate that telomeric chromatin in HeLa cells is densely covered with shelterin. The TRF1-TIN2-TPP1-POT1 complex is tightly associated with telomeres to facilitate replication, prevent ATR activation, and to recruit telomerase. In contrast, the TRF2-RAP1 complex dynamically binds to telomeres, allowing it to recruit a variety of co-factors that prevent ATM activation and inhibit NHEJ. In ALT cells, telomeres are more sparsely covered by shelterin with a distinct stoichiometry compared to telomerase-positive cancer cells.

Limitations of the study

Although we thoroughly characterized all HaloTagged shelterin cell lines and used multiple independent clones, we cannot exclude the possibility that N-terminal HaloTag fusions interfere with some aspects of normal shelterin functions like protein stability, binding affinities, or post-translational modifications. In HeLa cells, HaloTagging POT1 resulted in telomere elongation, similar to mutations that reduce POT1’s ssDNA binding affinity. While Halo-POT1 localization to telomeres was unaffected, a reduction in the ssDNA binding affinity may alter its telomeric abundance or dynamics. Additionally, prior studies have shown that N-terminal tagging of TPP1 or TIN2 can disrupt telomerase regulation or mitochondrial localization, respectively.^52,77 Therefore, C-terminal tagging of TIN2, TPP1, and POT1 may potentially better preserve native behavior, though it would exclude protein isoforms. For TIN2, the lack of a high-quality antibody prevented us from comparing Halo-TIN2 to endogenous TIN2 levels. Additionally, shelterin proteins bound to internal telomeric repeats may exhibit different dynamics than those at chromosome ends. While our dual-color single-molecule imaging consistently showed static shelterin molecules colocalizing with telomeres, these telomeres were marked by the same tagged shelterin proteins, which could also be enriched at internal telomeric repeats. Furthermore, our quantification may slightly overestimate the actual telomeric copy numbers of the HaloTagged shelterin proteins due to clustering of chromosome ends within a diffraction-limited signal, particularly in U2OS cells. Finally, free HaloTag or truncated HaloTagged shelterin fragments in the nucleoplasm could affect the analysis of telomere-bound fractions and diffusion coefficients.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jens C. Schmidt (schmi706@msu.edu).

Materials availability

All plasmids generated will be made available on Addgene, and all cell lines and other materials will be made available upon request by Jens C. Schmidt (schmi706@msu.edu).

Data and code availability

• All code used in this study has been deposited on GitHub (https://github.com/SchmidtLabMSU/ShelterinAnalysis).

• All unprocessed data are available upon request from the lead contact.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

In this work, we used these human cancer cell lines: HeLa-EM2–11ht (RRID: CVCL_WN71), HeLa 1.3 (RRID: CVCL_B5DA), and U2OS (RRID: CVCL_0042). U2OS cells were grown in RPMI 1640 media (Gibco) supplemented with 10% FBS (Gibco), 100 U ml⁻¹ penicillin and 100 μg mL⁻¹ streptomycin (Gibco) at 37°C and 5% CO₂. HeLa-EM2–11ht and HeLa 1.3 cels were grown in DMEM media (Gibco) supplemented with 10% FBS, 100 U ml⁻¹ penicillin and 100 μg mL⁻¹ streptomycin at 37°C and 5% CO₂. All experiments were performed using low-passage cell lines. U2OS cells were obtained from ATCC, and cell line authenticity was routinely verified by monitoring cell morphology via microscopy. All lines were regularly monitored for mycoplasma contamination. Although all cell lines used in this study are of female origin, the gender, sex, and ethnicity were not considered variables in the experimental design or results, as this study investigates fundamental molecular mechanism relevant to all cell types.

METHOD DETAILS

Cell lines maintenance

U2OS cells were grown in RPMI 1640 media (Gibco) supplemented with 10% FBS (Gibco), 100 U ml⁻¹ penicillin and 100 μg mL⁻¹ streptomycin (Gibco) at 37°C and 5% CO₂. All generated HeLa cell lines were based on HeLa-EM2–11ht⁵⁰ and were grown in DMEM media (Gibco) supplemented with 10% FBS, 100 U ml⁻¹ penicillin and 100 μg mL⁻¹ streptomycin (Gibco) at 37°C and 5% CO₂.

Generation of HaloTagged cell lines and plasmid construction for genome editing

To insert HaloTag at the 5′ end of all shelterin genes, the CRISPR/Cas9 genome-editing strategy was used with following procedure. All sgRNAs were cloned into BpiI-digested pX330-U6-Chimeric-BB-CBh-hSpCas9 backbone (Addgene #42230, gift from Feng Zhang).⁸¹ Plasmids used as donors for homology-directed repair (HDR) were prepared using Gibson assembly (NEB) with HpaI-digested pFastBacDual backbone, HaloTag insert, left and right homology arms. Homology arms were either PCR amplified from genomic DNA (TRF1, TRF2, TPP1, POT1 genes) or ordered as g-Blocks from IDT (TIN2, RAP1 genes). HaloTag insert consisted of 3xFLAG-HaloTag sequence, followed by a short linker sequence including TEV cleavage site. Sanger sequencing was used to confirm the correct cloning of plasmids. For genome editing, cells were plated into 6-well plate to reach ~70% confluency next day. Cells were transfected with a mixture of 1.25 μg of HDR donor plasmid, 1.25 μg of sgRNA plasmid and 7.5 μL of Lipofectamine 2000 (Invitrogen). After 48–72 h of transfection, cells were fluorescently labeled with HaloTag-ligand JFX650⁷⁸ and single-cell sorted into 96-well plates using FACS. Non-edited cells stained with JFX650 HaloTag-ligand were used as a negative control for acquiring background fluorescence. Single-cell clones were propagated into 24-well plates and screened for homozygous insertion by in-gel fluorescence and genomic PCR followed by Sanger sequencing. To knock-out RAP1 in Halo-TRF2 cell lines, cells were transfected (same protocol as above) with plasmid eSpCas9-ATP1A1-G2-Dual-sgRNA (Addgene #86612, gift from Yannick Doyon) which enabled for marker-free co-selection for NHEJ-based gene editing.⁸² The sgRNA targeting Exon 1 of RAP1 gene was cloned into the BbsI-linearized eSpCas9 vector. Three days post transfection, cells were selected with 0.5 μM ouabain (Sigma-Aldrich) and then single-cell diluted into 96-well plates. Clones containing RAP1 knock-out were screened and validated by western blotting. Cloning of the mNeonGreen fused shelterin proteins was performed by Gibson assembly into pRK2 backbone (original pHTN HaloTag CMV-neo, Promega #G7721). Single-point mutations in the mNeonGreen-TRF1 or mNeonGreen-TRF2 plasmids were introduced using inverse PCR.

Western blotting

Cells grown at 24-well plate were lysed with Laemmli sample buffer (BioRad) and subjected to SDS-PAGE using 4–15% TGX Stain-Free polyacrylamide gels (BioRad). Resolved proteins were transferred to a nitrocellulose membrane using Trans-Blot Turbo system (BioRad). Membrane was blocked with 5% milk in PBS-T (0.05% Tween 20) for 1 h at RT followed by incubation with primary antibody (diluted in 2.5% milk in PBS-T) O/N at 4°C. Then, membrane was washed 3 × 5 min in PBS-T and incubated for 1 h at RT with secondary antibody anti-mouse (Invitrogen #31439) or anti-rabbit (Invitrogen #31460) conjugated with HRP diluted 1:4,000 in 5% milk in PBS-T. Finally, membrane was washed 3 × 10 min in PBS-T and imaged for chemiluminescence using ChemiDoc system (Biorad) and SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific). Signal intensities were analyzed using ImageQuant software (Cytiva). Primary antibodies used in this study: anti-TERF2 (Novus #NB100–56506, 1:1,500 dilution or Novus #NB110–57130, 1:1,500 dilution), anti-TERF2IP (Novus #NB100–292, 1:1,000 dilution), anti-TPP1 (Bethyl #A303–069A, 1:2,000 dilution), anti-POT1 (Proteintech #10581–1-AP, 1:750 dilution), anti-TRF1 (raised in rabbit, 1:1,000 dilution, kind gift from Jan Karlseder), anti-FLAG conjugated with HRP (Sigma-Aldrich, #A8592, 1:2,000 dilution).

Preparation of 6xHis-3xFLAG-HaloTag for in-gel protein abundance measurement

The plasmid carrying 6xHis-3xFLAG-HaloTag was kindly provided by Dr. Youmans and Dr. Cech (University of Colorado Boulder, Boulder, CO, USA). The construct was transformed into BL21-Star (DE3) E. coli (Invitrogen) by heat-shock protocol. Cells were grown in Luria-Bertani media containing 100 μg mL⁻¹ of ampicillin at 37°C and 200 rpm until O.D.₆₀₀ = 0.8, then temperature was lowered to 18.5°C and the expression was induced with 0.5 mM IPTG. Cells were cultured O/N and harvested next morning by centrifugation (4,000 g × 10 min) at RT and pellet was stored at −80°C. All purification steps were carried out at 4°C unless otherwise stated. The bacterial pellet (12 g) was dissolved in 80 mL of lysis buffer (50 mM sodium phosphate pH 8, 500 mM NaCl, 10 mM Imidazole, 5% glycerol, 0.5% Tween 20, 1 mM DTT, 0.5 mM PMSF, 0.5 mg mL⁻¹ lysozyme) and sonicated for 3 min (process time) with 40% amplitude, 1 s pulse ON and 3 s pulse OFF (Fisherbrand Model 505, 0.5-inch tip). The suspension was cleared by centrifugation (19,000 g × 1 h). The supernatant was loaded on a gravity column containing 2.5 mL of Ni Sepharose 6 Fast Flow resin (Cytiva) and incubated for 30 min. Then, column was washed with 3 CV of lysis buffer and protein was eluted with elution buffer (50 mM sodium phosphate pH 7.5, 500 mM NaCl, 300 mM Imidazole, 5% glycerol). Sample was concentrated to 5 mL using Vivaspin 20 10 kDa MWCO (Cytiva), centrifuged (4,000 g × 10 min) and loaded onto HiLoad 16/600 Superdex 75 pg column (GE Healthcare Life Sciences) equilibrated in FPLC buffer (50 mM sodium phosphate pH 7.5, 150 mM NaCl, 5% glycerol). Flow-rate for size-exclusion chromatography was set to 1 mL min⁻¹. Peak fractions according to A₂₈₀ were analyzed by SDS-PAGE, pooled and concentrated to 13.5 mg mL⁻¹ using Vivaspin 20 10 kDa MWCO (Cytiva). Final sample was aliquoted, flash-frozen in LN₂ and stored at −80°C. To fluorescently label the prepared protein with JF657 HaloTag-ligand, the protein was diluted with FPLC buffer to 3 mg mL⁻¹ and mixed with 3-fold molar excess of JF657 HaloTag-ligand (resuspended in 1/10 of a protein volume of FPLC buffer supplemented with 10% DMSO). Mixture was incubated in dark for 3 h at 4°C while slowly rotating. Then, the mixture was centrifuged (15,000 g × 10 min) at 4°C and purified by size-exclusion chromatography (same protocol as before) to remove the excess dye. Peak fractions according to A₂₈₀ were pooled and concentrated using Vivaspin 6 10 kDa MWCO (Cytiva) to 10 μM. The degree of labeling was ~91% (determined by UV-Vis spectroscopy). The extinction coefficients used for calculations were ε₂₈₀ = 67,380 M⁻¹ cm⁻¹ for the 6xHis-3xFLAG-Hal-oTag (analyzed using Expasy ProtParam tool⁸⁶) and ε_max = 137,000 M⁻¹ cm⁻¹ for the JF657 HaloTag-ligand.⁷⁹ The correction factor for JF657 equaled to 0.0267, which was experimentally measured as A₂₈₀/A_max ratio of the ligand. Fluorescently labeled protein was diluted to 10 nM according to A_max and aliquoted by mixing 10 μL of protein +10 μL of 2x Laemmli sample buffer (Biorad). Therefore, each aliquot represented 100 fmol of fluorescently labeled Halo standard. Aliquots were boiled for 10 min at 95°C, flash-frozen in LN₂ and stored at −80°C until use.

In-gel fluorescence for quantification of protein total cellular abundance

Cells (100,000 for U2OS and 150,000 for HeLa) were plated into 24-well plate. Next day, cells were labeled with 250 nM JF657 HaloTag-ligand for 1 h, washed with PBS and incubated with a fresh media for additional 15 min. Then, cells were washed twice with PBS and lysed in 60 μL Laemmli sample buffer (BioRad) and boiled for 10 min at 95°C prior to SDS-PAGE. Prepared Halo standards (section above) were supplemented with a known amount of lysed parental cells (in Laemmli sample buffer), serially diluted, and loaded together with samples on 4–15% TGX Stain-Free polyacrylamide gel (BioRad). In-gel fluorescence was imaged using ChemiDoc system (Biorad) with Cy5.5 filter set, and total protein loading was imaged using Stain free filter with UV activation. To measure the protein abundance, the JF657 signal intensity was related to the standard curve obtained from serially diluted Halo standards. Similarly, the number of loaded cells were related to the standard curve based on Stain free signal of prepared standards. All quantifications were performed using ImageQuant software (Cytiva). For TEV correction, plated cells were labeled with JF657 HaloTag-ligand and washed as above. Then, cells were lysed in 50 μL RIPA buffer (w/o SDS). Samples were split half, supplemented with protease inhibitor cocktail (Sigma-Aldrich, #P8340) and one set of samples were treated with 5 units of TEV protease (NEB, #P8112) for 30 min on ice. After TEV cleavage, treated and untreated samples were mixed with 2x Laemmli sample buffer (Biorad), boiled for 10 min at 95°C and separated by SDS-PAGE. In-gel fluorescence was detected using the same procedure as above and TEV correction factors were obtained as a ratio between the signal intensity of uncleaved vs. cleaved HaloTagged protein normalized to Stain free signal. All experiments were performed in triplicate.

Flow-FISH for measuring relative telomere length

U2OS or HeLa cells were split by trypsinization and 3×10⁶ cells were transferred to a separate tube, washed with PBS and the cell pellet was resuspended in 1.2 mL of PBS. While slowly vortexing, 2.8 mL of 100% EtOH was added dropwise to fix cells (70% EtOH in final volume). Fixed cells were stored in the freezer until use. For the telomeric FISH staining, cells were washed with PBS and resuspended in 100 μg mL⁻¹ of RNase A in PBS followed by incubation at 37°C for 20 min. Then, cells were washed with PBS and permeabilized in PBS +0.05% Triton X-100 for 10 min at RT. After permeabilization, cells were washed with PBS and resuspended in preheated (80°C for 5 min) hybridization buffer (20 mM Tris pH 7.4, 60% formamide, 0.5% blocking solution (Roche #11096176001) and 1:300 dilution of PNA TelC-AF488 probe (PNA Bio #F1004)). Mixture was transferred to PCR tubes and incubated in thermocycler at 80°C for 10 min followed by two-hour incubation at RT in dark. After the hybridization, samples were transferred to 96-well plate and washed 2 × 10 min with preheated (60°C) 2x SSC +0.1% Tween 20 buffer. Then, cells were washed 2 × 5 min with PBS and finally resuspended in 150 μL of PBS and stored at 4°C covered in aluminum foil for the flow cytometry measurement next day. All centrifugation steps were done at 300 g × 3 min at RT with a bucket-swing rotor. Flow cytometry measurement was performed using spectral flow cytometer Cytec Aurora (Cytec). Samples were at 96-well plate. Acquired data were unmixed in SpectroFlo software (Cytec) and further analyzed in FCS Express (De Novo Software).

Immunofluorescence and fluorescence in situ hybridization for TIF analysis

Parental or generated HaloTagged cell lines were plated into 6-well plate (250,000 cells for U2OS and 400,000 cells for HeLa) containing sterilized 18 × 18 mm #1.5 glass coverslips (Corning). The next day, cells were labeled with 250 nM of JFX650 HaloTag-ligand for 30 min in complete media followed by two washes and a 15 min incubation with fresh complete media. After that, cells were washed with PBS and fixed with 4% formaldehyde in PBS for 10 min at RT. After fixation, cells were washed twice with PBS and permeabilized with 0.1% Triton X-100 in PBS for 10 min at RT. Then, cells were washed twice with PBS and blocked with 3% BSA +0.05% Tween 20 in PBS for 45 min at RT followed by the incubation with anti-53BP1 antibody (Novus, #NB100–304) diluted 1:1,000 in blocking buffer for 1 h at RT. Cells were washed 3 × 5 min with PBS-T (0.05% Tween 20) followed by the incubation with secondary anti-rabbit antibody conjugated with Cy3 (Invitrogen, #A10520) diluted 1:500 in blocking buffer for 1 h at RT. Then, cells were washed 3 × 5 min with PBS-T and fixed again with 4% formaldehyde in PBS for 5 min at RT. After fixation, cells were dehydrated with increasing concentration of cold EtOH (70%, 85% and 100%) for 2 min each. Coverslips were then dried and transferred into preheated (80°C) glass slides and incubated with a preheated (80°C for 5 min) hybridization buffer (20 mM Tris pH 7.4, 60% formamide, 0.5% blocking solution (Roche #11096176001) and 1:200 dilution of PNA TelC-AF488 probe (PNA Bio #F1004)) for 10 min at 80°C followed by 2 h incubation at RT in dark. After hybridization, cells were washed 2 × 10 min with preheated (60°C) 2x SSC +0.1% Tween 20 buffer. Then, cells were incubated with Hoechst dye (1:10,000 dilution in 2x SSC) for 2 min at RT to stain nuclei. Finally, cells were washed with 1x SSC and MilliQ water, dried, and mounted onto glass slides with ProLong diamond antifade mountant (Invitrogen #P36970) followed by O/N curing at RT in dark. Samples were imaged the next day using 3i spinning-disc confocal microscope with SoRa modality. Acquired Z-stacks for all channels were processed in Fiji⁸³ to obtain maximum intensity projections, which were used for subsequent automatic high-throughput analysis using custom MATLAB script. Briefly, nuclei channel was segmented using Cellpose machine-learning algorithm⁸⁴ in MATLAB, telomeric and 53BP1 foci were localized using ComDet plugin⁸⁵ in Fiji. Then, the number of TIFs per each segmented cell was determined as a number of colocalization events between 53BP1 and TelC. For each sample, minimum of 100 cells was included into the analysis. Statistical analyses were obtained using the Mann-Whitney test in R Studio.

Microscopy setup for live and fixed cell imaging

Two distinct microscopy systems were used for imaging either live or fixed cells. First, the Olympus IX83 inverted microscope equipped with cellVivo incubation system (temperature, humidity and CO₂ controlled), four laser lines (405, 488, 561 and 640 nm), X-cite Turbo LED excitation source, Olympus UPlanApo-HR 60x/1.5 NA TIRF objective, Twin-cam beamsplitter (Cairn Research) and Hamamatsu Orca-Fusion BT sCMOS or Hamamatsu Orca-Quest qCMOS cameras. The second system was a 3i spinning-disc (Yokogawa) microscope with confocal SoRa modality equipped with incubation chamber (temperature, humidity and CO₂ controlled), four laser lines (445, 488, 561 and 638 nm), Zeiss C PlanApo 63x/1.42 NA objective and Hamamatsu Orca-Fusion BT sCMOS camera. All live-cell imaging experiments were carried out at 37°C, 5% CO₂ and 95% humidity-controlled environment to mimic the conditions used for cell culturing.

Measuring the shelterin abundance at telomeres by fluorescence photobleaching

HaloTagged shelterin cell lines (25,000 U2OS and 45,000 HeLa cells) were plated into 24-well #1.5H glass bottom dish (Celvis #P24–1.5H-N). To synchronize cells at the onset of S-phase, cells were incubated the next day with 2 mM thymidine for ~18 h. Then, cells were washed twice with complete media and incubated for 9 h to complete the DNA replication. Cells were then blocked again with 2 mM thymidine for ~18 h. During the last hour of incubation, cells were labeled with 250 nM HaloTag-JFX650 ligand for 45 min in the presence of 2 mM thymidine. After labeling, cells were washed twice and incubated for additional 15 min in a fresh complete media containing 2 mM thymidine. Then, cells were washed twice and fixed with 4% formaldehyde for 10 min at RT followed by two-time wash with PBS and nuclei staining for 4 min at RT using Hoechst dye (diluted 1:5,000 in PBS). Cells were washed twice with PBS and imaged immediately using the Olympus IX83 microscope (equipped with Orca-Quest camera). Z-stacks of Halo channel (51 planes) were imaged using 640 nm laser line under epi-illumination with these settings: 100% laser power, 40 ms exposure time (HeLa cells) or 65% laser power, 25 ms exposure time (U2OS and HeLa 1.3 cells). Subsequently, the z stack of Hoechst channel was acquired using LED 385 nm excitation light. After collecting Z-stacks for at least 60 cells per sample, the photobleaching time-lapse movies (continuous imaging of one plane capturing 8,000 frames for HeLa or 15,000 frames for U2OS) of Halo-TRF1 foci were acquired using the same exposure settings. Halo-TRF1 cells were chosen due to their specific telomeric shelterin foci formation with minimal nuclear background signal. The analysis for quantifying the number shelterin proteins bound to telomeres was done in two steps using custom written MATLAB scripts. Firstly, the photobleaching movies were used to determine the fluorescence intensity of a single fluorophore. At least 20 Halo-TRF1 foci were cropped and their maximum intensity profiles over time were saved using Fiji. From these traces the background fluorescence (mean intensity over time of nucleoplasm ROI) was subtracted. Then, each photobleaching function was plotted as a histogram where the position of maximum count was detected and corresponding peak was fitted with Gaussian distribution, which represented the fluorescence background level. Next, the closest peak to the back-ground distribution was localized and fit with another Gaussian function, which represented the fluorescence intensity distribution of a single fluorophore. The difference between the mean values of those Gaussian distributions was assigned as a photobleaching step-size, which was used in the next step to extract the number of fluorophores within the foci from its initial foci intensity value. The second step consisted of maximum intensity projection of all acquired Z-stacks with subsequent foci localization using ComDet⁸⁵ plugin in Fiji. Nuclei channel was segmented using Cellpose⁸⁴ machine-learning algorithm in MATLAB. Finally, the fluorescence intensity for each individual shelterin telomeric foci was background subtracted and divided by photobleaching step-size, which allowed to extract the number of HaloTagged shelterin molecules within a respective telomeric foci with single cell resolution in a high-throughput manner.

Flow cytometry

To analyze the cell cycle profile of cells treated with double-thymidine block used for the analysis of shelterin abundance at telomeres, cells were plated in 6-well plate and blocked with 2 mM thymidine in parallel with samples used for the imaging. Then, cells were fixed with 4% formaldehyde, washed twice with PBS, stained with FxCycle PI/RNase Staining Solution (Invitrogen #F10797) and analyzed using BD Accuri C6 cytometer. Signal from more than 100,000 cells was collected and data were further processed and analyzed using FCS Express (De Novo Software).

Single-molecule live-cell imaging

HaloTagged cell lines were plated into 24-well #1.5H glass bottom dish (Celvis #P24–1.5H-N) two days prior to imaging (40,000 cells for U2OS or 65,000 cells for HeLa cell lines). To facilitate single-molecule detection, cells were sparsely labeled with JFX650 HaloTagligand using these conditions: 1 nM for 1 min (TRF1, TIN2, TPP1, POT1) or 1 nM for 30 s (TRF2, RAP1) in complete media at 37°C. Then, cells were washed twice with complete media and incubated for additional 15 min. After incubation, cell nuclei were stained for 2 min with Hoechst dye diluted 1:5,000 in complete media followed by two-times wash. The live-cell imaging was performed using the HILO (Highly Inclined and Laminated Optical Sheet) modality on Olympus IX83 microscope with 640 nm laser line (set to 35% power). The imaging speed was set to 100 Hz (equals to 10 ms time delay). Field of view was adjusted to 512 × 512 px (BT-Fusion camera) or 400 × 400 px with 2 × 2 px binning (Orca-Quest camera). With these settings, 1000 frames were captured for each acquisition. After each recorded movie, one image of Hoechst channel was taken using LED 385 nm light as a nuclear marker for the segmentation of single-molecule signal from nuclei. For combining the sparse and dense labeling of Halo-shelterin proteins to visualize single-molecules together with telomeric foci, the cells were first sparsely labeled with JF657 HaloTag-ligand using these conditions: 2 nM for 2 min (TRF1, TIN2, TPP1, POT1) or 2 nM for 1 min (TRF2, RAP1) in complete media at 37°C followed by washing cells twice with complete media and then densely labeled with 250 nM JF549 HaloTag-ligand⁸⁰ for 15 min. After labeling, cells were washed three-times and incubated for additional 15 min in complete media. The live-cell imaging was performed with the same settings as above with this modification: after acquiring 1000 frames of single-molecule JF657 signal, the z stack using epi-illumination with 488 nm laser line was collected to mark telomeric shelterin foci. z stack was then maximum intensity projected and overlayed with single-molecule movies.

Single-particle tracking (SPT) analysis

For localizing single molecules and linking them into trajectories, the multiple-target tracing algorithm⁸⁷ was used with modified parallel processing version of SLIMfast in MATLAB.⁵⁹ These settings were used to track single particles: exposure time = 10 ms, NA = 1.5, pixel size = 0.1083 μm (BT-Fusion camera) or 0.1533 μm (Orca-Quest camera), emission wavelength = 667 nm (for Halo-JFX650) or 672 nm (for Halo-JF657), D_max = 5 μm² s⁻¹, number of gaps allowed = 2, localization error = 10⁻⁶, deflation loops = 0. Images of nuclei channel were used to generate nuclear masks based on which only nuclear trajectories were segmented and used for analysis. The MATLAB version of Spot-On algorithm⁵⁹ was employed to analyze tracks according to mean-square displacements (MSDs) to derive the diffusion coefficient of tracked single particles and respective fraction of molecules residing in either bound or freely diffusing state (2-state model was used in this analysis). The Spot-On algorithm was used with these parameters: time gap = 10 ms, dZ = 0.700 μm, gaps allowed = 2, time points = 7, jumps to consider = 4, bin width = 0.01 μm, CDF fitting routine, D_free 2-state boundaries = [0.5 10], D_bound 2-state boundaries = [0.0001 0.5]. Alternatively, trajectories were also analyzed using the ExTrack procedure that describes each trajectory as a probability function and derives underlying biophysical properties, including transition rates, based on Hidden-Markov-Modelling (HMM).⁶⁰ The ExTrack analysis was performed in Python with these settings: number of states = 2, minimal trajectory length = 3, dT = 0.01 s, frame length = 6, max distance allowed for consecutive position = 1.0, number of positions per track accepted = [3, 30], localization error type = 1, D_max = 5 μm² s⁻¹, optimization method = ‘Powell’. For extracting the fraction of nuclear HaloTagged proteins, the number of nuclear trajectories (based on nuclear masks generated by Hoechst staining) was divided by the number of total trajectories detected. All experiments were carried out as three independent biological replicates with at least 30 cells for each sample. All statistical analysis were performed using GraphPad Prism by one-way ANOVA with Tukey’s posthoc test.

Residence time imaging and analysis

HaloTagged shelterin cell lines were plated and sparsely labeled with HaloTag-JF657 ligand according to the protocol described in “Single-molecule live-cell imaging”. JF657 was used due to its superior photostability performance.⁷⁹ Long-term single-molecule imaging was performed using HILO modality on Olympus IX83 microscope with 20% laser power (640 nm laser line) and 100 ms exposure time to blur out freely diffusing particles. Time delay between frames was set to either 500 or 1500 ms and a total of 250 frames were collected. This approach allowed to measure residence time with two distinct temporal conditions. To track only static particles, the D_max was set to 0.0075 μm² s⁻¹ and to limit the possibility of chopping long trajectories into several shorter ones, the number of gaps allowed was set to 5 and localization error to 10⁻⁵. Only nuclear single-particle trajectories were used for the analysis (segmented according to nuclear masks generated from Hoechst channel). Additionally, trajectories with a total length of less than 3 were not included into the analysis. Movies with 500 ms time delay were processed by averaging 3 consecutive frames to match the temporal resolution of movies with 1500 ms. The distribution of track length was visualized as a survival distribution obtained as a 1-CDF (Cumulative Density Function of track lengths) in MATLAB and fitted with two-component exponential function to extract the respective rate constants. All experiments were carried out as three independent biological replicates with at least 10 cells for each sample. All statistical analyses were performed using GraphPad Prism by two-way ANOVA with Tukey’s post-hoc test.

Co-immunoprecipitation of shelterin proteins

HaloTagged TIN2, TPP1 or RAP1 U2OS cells (confluent 15 cm² plate) were harvested by trypsinization, and cells were washed with PBS. Each cell pellet was resuspended in 500 μL of cold lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 2 mM MgCl₂, 1 mM DTT, 0.5% Triton X-100) supplemented with protease inhibitor cocktail (Sigma-Aldrich, #P8340) and 250 U of benzonase (Sino Biological, #SSNP01). The lysates were incubated for 1 h at 4°C while rotating. Then, the suspensions were cleared by centrifugation (15,000 g × 20 min) at 4°C. Supernatants were mixed with 120 μL (slurry) anti-FLAG M2 affinity gel (Sigma-Aldrich, #A2220) equilibrated in lysis buffer. Mixtures were supplemented with 1 μM JF657 HaloTag-ligand and 100 fmol recombinant FLAG-Halo protein and incubated for 1 h at 4°C while rotating. Then, samples were washed five times with lysis buffer, beads were resuspended in 75 μL of lysis buffer supplemented with 2 μL FLAG peptide (5 mg/mL) and incubated for 45 min at RT to elute the bound fraction. Then, samples were spun down and filtered through 0.22 μm Spin-X filter (Corning). Input, flow through and bound fractions were subjected to SDS-PAGE and western blotting according to protocol described in Western blotting section. In-gel fluorescence and western blot quantification was performed using ImageQuant software. Specifically, the fluorescence intensity of HaloTagged TIN2, TPP1 and RAP1 samples was divided by fluorescence intensity of recombinant FLAG-Halo protein and obtained value was further divided by either TRF1 or TRF2 western blot signal. Finally, values were normalized to TIN2 sample. Co-IP was carried out as three independent biological replicates. All statistical analysis were performed using GraphPad Prism by one-way ANOVA with Tukey’s posthoc test.

Conditional degradation of HaloTagged proteins for studying shelterin recruitment to telomeres

HaloTagged shelterin cell lines (40,000 U2OS and 65,000 HeLa cells) were plated into 24-well #1.5H glass bottom dish (Celvis #P24–1.5H-N). Next day morning, cells were transfected with 100 ng of plasmids coding for shelterin factors fused with mNeonGreen fluorescent protein using Lipofectamine 2000 (Invitrogen) according to the manufacturer protocol. After ~5 h of transfection, cells were washed twice and incubated with fresh media supplemented with 500 nM Halo-PROTAC-E ligand (Aobious AOB13155) to specifically degrade the endogenous HaloTagged shelterin proteins. After ~18 h of incubation, cells were labeled with 250 nM HaloTag-JFX650 ligand for 45 min with the presence of 500 nM of Halo-PROTAC-E. Then, cells were washed twice, and nuclei were stained using Hoechst dye (1:5,000 dilution) for 4 min followed by washing cells twice and finally keeping cells in fresh complete media supplemented with 500 nM Halo-PROTAC-E ligand for imaging. Live-cell imaging was carried out using 3i spinning-disc confocal microscope with SoRa modality. Acquired Z-stacks were processed in Fiji to obtain maximum intensity projections, which were used for subsequent automatic high-throughput analysis using custom MATLAB script. Briefly, nuclei channel was segmented using Cellpose machine-learning algorithm⁸⁴ in MATLAB, mNeonGreen shelterin foci were localized using ComDet plugin⁸⁵ in Fiji. Then, the number detected foci per cell was quantified.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses were performed using either GraphPad Prism software (one-way ANOVA with Tukey’s post hoc test) or R Studio software (Mann-Whitney test). For all reported p-values, results were considered non-significant (ns) when p > 0.05. Exact p-values for significant results are provided directly in the figures. Each figure legend includes detailed information on sample size, the number of biological or technical replicates, and the specific statistical analysis used.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116178.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-TERF2	Novus	Cat#NB100-56506; RRID: AB_839054
Rabbit polyclonal anti-TERF2	Novus	Cat#NB110-57130; RRID: AB_844199
Rabbit polyclonal anti-TERF2IP	Novus	Cat#NB100-292; RRID: AB_10000825
Rabbit polyclonal anti-TPP1	Bethyl	Cat#A303-069A; RRID: AB_10890369
Rabbit polyclonal anti-POT1	Proteintech	Cat#10581-1-AP; RRID: AB_2284057
Rabbit anti-TERF1	Jan Karlseder	N/A
Mouse monoclonal anti-FLAG (HRP)	Sigma-Aldrich	Cat#A8592; RRID: AB_439702
Rabbit polyclonal anti-53BP1	Novus	Cat#NB100-304; RRID: AB_10003037
Goat polyclonal anti-rabbit (Cy3)	Invitrogen	Cat#A10520; RRID: AB_10563288
Goat polyclonal anti-rabbit (HRP)	Invitrogen	Cat#31460; RRID: AB_228341
Goat polyclonal anti-mouse (HRP)	Invitrogen	Cat#31439; RRID: AB_228292
Bacterial and virus strains
One Shot BL21(DE3) Chemically Competent E. coli	Invitrogen	Cat#C600003
TOP10 Chemically Competent E. coli	Invitrogen	Cat#C404006
Chemicals, peptides, and recombinant proteins
TEV protease	New England Biolabs	Cat#P8112
Lipofectamine 2000	Invitrogen	Cat#11668500
HaloPROTAC-E	AOBIOUS Inc.	Cat#AOB13155
Ouabain	Sigma-Aldrich	Cat#O3125
Thymidine	Sigma-Aldrich	Cat#T1895
PNA TelC-Alexa488	PNA Bio	Cat#F1004
Janelia Fluor JFX650 HaloTag Ligand	Grimm et al.78; Promega	Cat#HT1070
Janelia Fluor 657 HaloTag Ligand	Grimm et al.79	N/A
Janelia Fluor 549 HaloTag Ligand	Grimm et al.80; Promega	Cat#HT1020
Hoechst 33342	Thermo Scientific	Cat#62249
Prolong Diamond Antifade Mountant	Invitrogen	Cat#P36970
Super Nuclease	Sino Biological	Cat#SSNP01
ANTI-FLAG M2 Affinity Gel	Millipore	Cat#A2220
3xFLAG Peptide	Sigma-Aldrich	Cat#F4799
Ni Sepharose 6 Fast Flow histidine-tagged protein purification resin	Cytiva	Cat#17531801
Critical commercial assays
FxCycle PI/RNase Staining Solution	Invitrogen	Cat#F10797
SuperSignal West Femto Maximum Sensitivity Substrate	Thermo Scientific	Cat#34094
4–15% Mini-PROTEAN TGX Stain-Free Protein Gels	Bio-Rad	Cat#4568083
4–15% Criterion TGX Stain-Free Protein Gel	Bio-Rad	Cat#5678084
Trans-Blot Turbo RTA Mini 0.2 μm Nitrocellulose Transfer Kit	Bio-Rad	Cat#1704270
Deposited data
All generated plasmids will be deposited at Addgene		N/A
Photobleaching and residence time analysis algorithms	This study	https://github.com/SchmidtLabMSU/ShelterinAnalysis
Experimental models: Cell lines
Human: HeLa EM2-11ht	Weidenfeld et al.50	RRID: CVCL_WN71
Human: HeLa EM2-11ht, FLAG-HaloTag-TRF1	This study	N/A
Human: HeLa EM2-11ht, FLAG-HaloTag-TRF2	This study	N/A
Human: HeLa EM2-11ht, FLAG-HaloTag-TIN2	This study	N/A
Human: HeLa EM2-11ht, FLAG-HaloTag-POT1	This study	N/A
Human: HeLa EM2-11ht, FLAG-HaloTag-TRF2, RAP1-KO	This study	N/A
Human: HeLa 1.3	Takai et al.48	RRID: CVCL_B5DA
Human: HeLa 1.3, FLAG-HaloTag-TRF2	This study	N/A
Human: HeLa 1.3, FLAG-HaloTag-TIN2	This study	N/A
Human: U2OS	ATCC	Cat#HTB-96; RRID: CVCL_0042
Human: U2OS, FLAG-HaloTag-TRF1	This study	N/A
Human: U2OS, FLAG-HaloTag-TRF2	This study	N/A
Human: U2OS, FLAG-HaloTag-TIN2	This study	N/A
Human: U2OS, FLAG-HaloTag-TPP1	This study	N/A
Human: U2OS, FLAG-HaloTag-POT1	This study	N/A
Human: U2OS, FLAG-HaloTag-RAP1	This study	N/A
Human: U2OS, FLAG-HaloTag-TRF2, RAP1-KO	This study	N/A
Oligonucleotides
All oligonucleotides used in this study can be found in Table S2	This study	N/A
Recombinant DNA
pRK2-mNeonGreen-TRF1 plasmid	This study	N/A
pRK2-mNeonGreen-TRF2 plasmid	This study	N/A
pRK2-mNeonGreen-TIN2 plasmid	This study	N/A
pRK2-mNeonGreen-POT1 plasmid	This study	N/A
pFastBacDual plasmid	Gibco	Cat#10360014
pFastBacDual-FLAG-HaloTag-TRF1 HRD plasmid	This study	N/A
pFastBacDual-FLAG-HaloTag-TRF2 HRD plasmid	This study	N/A
pFastBacDual-FLAG-HaloTag-TIN2 HRD plasmid	This study	N/A
pFastBacDual-FLAG-HaloTag-TPP1 HRD plasmid	This study	N/A
pFastBacDual-FLAG-HaloTag-POT1 HRD plasmid	This study	N/A
pFastBacDual-FLAG-HaloTag-RAP1 HRD plasmid	This study	N/A
pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid	Cong et al.81	Addgene plasmid #42230
pX330-TRF1 plasmid	This study	N/A
pX330-TRF2 plasmid	This study	N/A
pX330-TIN2 plasmid	This study	N/A
pX330-TPP1 plasmid	This study	N/A
pX330-POT1 plasmid	This study	N/A
pX330-RAP1 plasmid	This study	N/A
eSpCas9-ATP1A1-G2-Dual-sgRNA plasmid	Agudelo et al.82	Addgene plasmid #86612
eSpCas9-ATP1A1-G2-Dual-RAP1-KO plasmid	This study	N/A
6xHis-3xFLAG-HaloTag plasmid	Thomas R Cech	N/A
pHTN HaloTag CMV-neo plasmid	Promega	Cat#G7721
Software and algorithms
Fiji 2.16.0	Schindelin et al.83	https://imagej.net/software/fiji/
MATLAB 2024b	Mathworks Inc.	https://www.mathworks.com/
Python 3.11.5	Python Software Foundation	https://www.python.org
Image Quant TL 8.2	Cytiva	https://www.cytivalifesciences.com/en/us/shop/protein-analysis/molecular-imaging-for-proteins/imaging-software/imagequant-tl-analysis-software-p-28619
FCS Express 7 Research	De Novo Software	https://denovosoftware.com/
SpectroFlo v2.2	Cytec	https://cytekbio.com/pages/spectro-flo
R Studio 2022.07.2	Posit Software, PBC	https://posit.co/products/open-source/rstudio/
GraphPad Prism 10.5.0	GraphPad Software	https://www.graphpad.com/
SPT LocAndTrack algorithm	Hansen et al.59	https://github.com/elifesciences-publications/SPT_LocAndTrack
Spot-On algorithm	Hansen et al.59	https://gitlab.com/tjian-darzacq-lab/spot-on-matlab
ExTrack algorithm	Simon et al.60	https://github.com/vanTeeffelenLab/ExTrack
Cellpose 2.0 segmentation algorithm	Stringer et al.84	https://github.com/MouseLand/cellpose
ComDet plugin for Fiji v.0.5.5	Katrukha et al.85	https://imagej.net/plugins/spots-colocalization-comdet
Photobleaching and residence time analysis algorithms	This study	https://github.com/SchmidtLabMSU/ShelterinAnalysis

Highlights.

• Shelterin proteins are equally abundant (30–40 copies) at telomeres in human cancer cells

• Shelterin is present at telomeres predominantly in two subcomplexes

• TRF1-TIN2-TPP1-POT1 binds stably, while TRF2-RAP1 interacts dynamically with telomeres

• TRF1 and TRF2 occupy non-overlapping binding sites on telomeric chromatin