POT1 recruits and regulates CST–Polα/primase at human telomeres

Summary

Telomere maintenance requires extension of the G-rich telomeric repeat strand by telomerase and fill-in synthesis of the C-rich strand by Polα/primase. At telomeres, Polα/primase is bound to Ctc1/Stn1/Ten1 (CST), a single-stranded DNA-binding complex. Like mutations in telomerase, mutations affecting CST–Polα/primase result in pathological telomere shortening and cause a telomere biology disorder, Coats plus (CP). We determined cryogenic electron microscopy structures of human CST bound to the shelterin heterodimer POT1/TPP1 that reveal how CST is recruited to telomeres by POT1. Our findings suggest that POT1 hinge phosphorylation is required for CST recruitment, and the complex is formed through conserved interactions involving several residues mutated in CP. Our structural and biochemical data suggest that phosphorylated POT1 holds CST–Polα/primase in an inactive auto-inhibited state until telomerase has extended the telomere ends. We propose that dephosphorylation of POT1 releases CST–Polα/primase into an active state that completes telomere replication through fill-in synthesis.

Graphical Abstract

In Brief

Cryo-EM structures of CST–POT1/TPP1 complexes reveal the molecular basis of how the shelterin subunit POT1 recruits CST–Polα/primase to telomeres to solve the end-replication problem. POT1 phosphorylation regulates the activity of CST–Polα/primase, providing a mechanism for the temporal control of telomeric C-strand maintenance in humans.

Introduction

Telomeres prevent the inappropriate activation of the DNA damage response at the ends of linear chromosomes. Human telomeric DNA is comprised of double-stranded (ds) 5’-TTAGGG-3’ repeats that terminate in a single-stranded (ss) 3’ overhang. This DNA protects chromosome ends through its interaction with the six-subunit shelterin complex (reviewed by de Lange¹). With each cell cycle, incomplete DNA synthesis and nucleolytic resection at telomere ends results in attrition of the telomeric DNA that can lead to excessive telomere shortening if unchecked (reviewed by Cai and de Lange²). Telomere maintenance is critical for the long-term proliferation of human cells and involves two distinct DNA synthesizing enzymes: telomerase, which elongates the G-rich strand, and CST–Polα/primase, which maintains the C-rich repeats through fill-in synthesis (reviewed by Cai and de Lange² and Lim and Cech³).

Loss of CST leads to progressive shortening of the C-rich strand at telomeres duplicated by lagging-strand DNA synthesis. This loss of telomeric DNA is caused by the inability of the replisome to generate Okazaki fragments close to the end of a linear DNA⁴. This lagging-end replication problem is solved by CST–Polα/primase, not telomerase. CST–Polα/primase also counteracts C-strand shortening at telomeres duplicated by leading-strand DNA synthesis, where the blunt-ended leading-strand products are resected to recreate the 3’ overhang. In this context, CST–Polα/primase guards against C-strand loss resulting from hyper-resection^5–8. Whereas much is known about how telomerase is recruited and regulated at telomere ends^{9, 10}, the mechanism of CST–Polα/primase recruitment, regulation, and action is not well understood.

CST is a heterotrimeric ssDNA-binding protein that shares structural homology with the major eukaryotic ssDNA-binding protein, Replication Protein A (RPA)^11–13. However, it is likely that CST evolved independently in an archaeal ancestor and became specialized for telomere maintenance in eukaryotes¹⁴ (reviewed by Cai and de Lange²). In metazoans, the large Ctc1 subunit of CST contains two sets of tandem oligosaccharide/oligonucleotide-binding (OB) folds¹³ that are connected by a three-helix bundle, termed the Acidic Rpa1 OB-binding Domain-Like three-helix bundle (Ctc1^ARODL) based on similarity to its archaeal counterpart^{2, 14}. Ctc1^OB-A/B/C/D and the Ctc1^ARODL comprise the metazoan-specific N-terminal domain of the protein, whereas Ctc1^OB-E/F/G are ancestral and mediate ssDNA binding and formation of the Ctc1/Stn1/Ten1 heterotrimer. Stn1, comprising an N-terminal OB fold (Stn1^N) and C-terminal tandem winged helix-turn-helix (wHTH) domains (Stn1^C), connects the single OB fold of Ten1 to Ctc1¹³.

Recent cryo-EM structures of CST–Polα/primase revealed how CST interacts with Polα/primase in two distinct conformations. In these structures, Polα/primase either adopts a compact, auto-inhibited state¹⁵ or an extended state compatible with enzymatic activity¹⁶. CST binds the auto-inhibited Polα/primase using the Ctc1 N-terminal domain, and this interaction is conserved in metazoans¹⁵. The inactive complex was proposed to represent a Recruitment Complex (RC) that describes the conformation of CST–Polα/primase as it is brought to the telomere prior to fill-in. In the Pre-Initiation Complex (PIC) in which Polα/primase is primed for activation, CST uses its ancestral RPA-like regions (Ctc1^OB-E/F/G, Stn1, and Ten1) to stabilize and stimulate the extended Polα/primase¹⁶. The cryo-EM structures indicate that a significant conformational change is required to transition from the RC to PIC, but how this change is regulated is unclear.

The importance of CST–Polα/primase in the maintenance of telomeres is illustrated by CP, a rare pediatric condition characterized by the ophthalmological disorder Coats disease plus associated systemic disorders primarily affecting the brain, bone, and gastrointestinal tract^17–25. CP is caused by hypomorphic mutations in Ctc1 and Stn1 that reveal their phenotype only when the second copy of the gene is non-functional. In addition, CP can be caused by homozygosity for a hypomorphic mutation in Stn1 or POT1^{23, 26}.

It has remained unclear how CST–Polα/primase is recruited to telomeres. Human TPP1 readily interacts with Ctc1 and was therefore thought to play a primary role in recruitment of human CST²⁷. However, at mouse telomeres, it is one of the two POT1 proteins (mPOT1b) that interacts with CST and mutations in mPOT1b diminish CST recruitment and lead to lack of C-strand fill-in at telomeres⁵. Whereas human POT1 has not been shown to interact with CST in co-immunoprecipitation (co-IP) experiments, it does interact with Ctc1 in a yeast two-hybrid assay²⁸ (but also see Miyake et al.¹²). Furthermore, the finding that the POT1 CP mutation (S322L) confers a telomere phenotype consistent with CST deficiency argues that POT1 contributes to CST recruitment²³. An initially confusing result indicated that removal of TPP1 or POT1 from human telomeres led to an increase, rather than a decrease, of CST loading²⁸. However, this result is explained by the activation of ATR signaling at deprotected telomeres lacking POT1, which leads to 53BP1- and Shieldin-dependent loading of CST^{29, 30}.

Here, we present structural and biochemical data revealing that human CST primarily interacts with POT1. This interaction is dependent on POT1 phosphorylation, reconciling the negative co-IP data. We identified four amino acids in mPOT1b that, when inserted into the human protein, result in POT1 phosphorylation in both human cells and insect cells. With the phosphorylated form of POT1, we reconstituted CST–POT1/TPP1 complexes in the presence and absence of telomeric ssDNA and determined their structures using cryo-EM. The structures suggest that phosphorylation of the POT1 hinge region drives charge complementarity between the two proteins. Two other interfaces form independently from the mPOT1b insertion and confirm that human POT1 is the primary recruiter of CST. Several CP mutations map to or near the primary interface between Ctc1^OB-D, Ctc1^ARODL, and POT1^OB-3. The POT1 N-terminal OB folds interact with Ctc1’s ssDNA-binding site, directly competing with CST binding to ssDNA and Polα/primase in the active PIC conformation, whereas the major interface on Ctc1 for Polα/primase binding in the auto-inhibited RC conformation is unobstructed. In vitro C-strand synthesis assays reveal that POT1/TPP1 inhibits CST–Polα/primase activity when bound to CST, supporting the structural analysis. Together, our data point toward a phosphorylation-dependent switch in POT1 that controls its interaction with CST and can regulate the activity of CST–Polα/primase at the telomere. This provides a molecular basis for how shelterin recruits and regulates CST–Polα/primase and reveals the mechanism of mutations in CST and POT1 that cause telomere dysfunction in CP.

Results

Reconstitution of a stable CST–POT1/TPP1 complex

To understand how POT1/TPP1 recruits CST, we examined how TPP1, the established CST interactor, binds. Co-IP data indicated that a region within the C-terminal 20 residues of TPP1 is necessary for its interaction with CST (Fig. S1A). This region of TPP1 constitutes its TIN2-Interaction Domain (TID) and is required to anchor the POT1/TPP1 heterodimer to the telomeric dsDNA-binding proteins in shelterin³¹. Indeed, co-IP experiments showed that TIN2 competes with CST for TPP1 binding (Fig. S1B). AlphaFold-Multimer³² modeling of the TID bound to Ctc1 predicted that TPP1 binds to OB-B of Ctc1 using the same amino acids used for binding to TIN2, consistent with the finding that TPP1 can interact with either TIN2 or Ctc1 but not both (Fig. S1B–D). Since its association with TIN2 is required for TPP1/POT1 to localize to telomeres^{33, 34}, we consider it unlikely that the TPP1^TID is involved in the recruitment of CST. As predicted, a recent report confirmed that loss of the interaction between TPP1^TID and Ctc1^OB-B did not disrupt fill-in by CST–Polα/primase³⁵. We therefore omitted the C-terminus (aa 252–458), including the TID, from TPP1 to limit sample heterogeneity from this largely disordered region.

Interestingly, although human POT1 does not form a stable complex with CST in co-IP experiments, mPOT1b readily bound to human CST (Fig. 1A–C). This finding suggested that mPOT1b can constitutively bind CST whereas the binding of human POT1 to CST may be regulated, perhaps by post-translational modifications. To determine which region in mPOT1b promotes the CST interaction, chimeric proteins containing swaps between mPOT1b and human POT1 regions were assayed for CST binding in co-IP experiments (Fig. 1B). When swapped with the human POT1 hinge (aa 299–344), the mPOT1b hinge (aa 300–350) conferred CST interaction (Fig. 1A, C). Further truncations and sequence comparison identified four residues in mPOT1b (ESDL, aa 323–326, Fig. 1B–C) that, when inserted into the human POT1 hinge (between aa 320–321, Fig. 1B–C), conferred robust interaction with CST (Fig. 1A–C). Furthermore, whereas there was no detectable interaction between purified wild-type (WT) POT1/TPP1 and CST, POT1(ESDL)/TPP1 formed a complex with CST that was detectable by fluorescence-detection size-exclusion chromatography (FSEC, Fig. 1D–E). POT1(ESDL)/TPP1 formed a complex with CST in the presence or absence of ssDNA, but addition of a telomeric [GGTTAG]₃ ssDNA allowed complex formation at lower protein concentrations (Fig. 1E; Fig. S2A).

Figure 1. Identification of residues promoting the POT1–CST interaction.

See also Figure S1 and Figure S2.

(A) Domain schematics and design of constructs containing chimeric swaps between human POT1 (red) and mouse mPOT1b (salmon). HMb: Human/Mouse POT1b swap; OB: oligosaccharide/oligonucleotide-binding domain; HJRL: Holliday junction resolvase-like domain. The interaction with human CST is indicated as shown in (C).

(B) Sequence alignment of the region within the POT1 hinge in POT1, mPOT1b, and POT1(ESDL) showing the location of the ESDL insertion.

(C) Co-IPs of Myc-tagged POT1/mPOT1b chimeras and FLAG-tagged Ctc1 from 293T cells transfected with Myc-POT1 (and variants), FLAG-TPP1, FLAG-Ctc1, FLAG-Stn1, and FLAG-Ten1. Western blots of Myc IPs were probed with anti-Myc and anti-FLAG antibodies to detect POT1 and Ctc1, respectively. POT1 construct details and names are provided in (A).

(D) Coomassie-stained SDS-PAGE gels and cartoon schematics of the purified POT1/TPP1 and CST proteins used for fluorescent-detection size-exclusion chromatography (FSEC) analysis.

(E) FSEC analysis of the interaction between CST and POT1(WT)/GFP-TPP1 (black) or POT1(ESDL)/GFP-TPP1 (blue) in the absence (top) or presence (bottom) of telomeric ssDNA. Traces without CST are shown as dashed lines and color key is the same for both panels. RFU: relative fluorescence units.

We reconstituted two complexes of CST–POT1(ESDL)/TPP1. With the addition of the 18-nt [GGTTAG]₃ ssDNA, we could reconstitute a stable, native complex containing full-length CST, full-length POT1(ESDL), and TPP1 (aa 1–251; Fig. 2A; Fig. S2B). Without ssDNA, the complex dissociated at the low concentrations (< 200 nM) used for negative-stain and cryo-EM (Fig. S2A). To increase the local concentration of CST relative to POT1(ESDL)/TPP1 for structural studies, we fused aa 1–403 of TPP1 to the N-terminus of Stn1, retaining 162 aa of the disordered serine-rich linker of TPP1 to allow for flexibility between the two polypeptides (Fig. S2C). This fusion allowed for purification of an apo CST–POT1(ESDL)/TPP1 complex in the absence of ssDNA (Fig. S2D).

Figure 2. Cryo-EM structures of CST–POT1(ESDL)/TPP1 complexes.

See also Figure S3 and Figure S4.

(A) Domain organization of CST and the shelterin subunits POT1(ESDL) and TPP1. Regions of TPP1 not observed in the cryo-EM maps are shown at 50% opacity. Dashed lines indicate regions not included in the expression construct. *The TPP1 serine-rich linker was included in the apo complex but not the ssDNA-bound complex. See also Fig. S2C. ARODL: Acidic Rpa1 OB-binding Domain-Like 3-helix bundle; wH: winged helix-turn-helix domain; RD: recruitment domain, TID: TIN2-interacting domain. For other abbreviations see Fig. 1.

(B) Cryo-EM density map (contour threshold 0.15) of apo CST–POT1(ESDL)/TPP1 at 3.9-Å resolution (see also Fig. S3).

(C) Cryo-EM density map (contour threshold 0.15) of ssDNA-bound CST–POT1(ESDL)/TPP1 at 4.3-Å resolution (see also Fig. S4).

(D-E) Atomic models for the apo (D) and ssDNA-bound (E) CST–POT1(ESDL)/TPP1 complexes, respectively. See also Video S1.

Cryo-EM structures of CST–POT1/TPP1

Negative-stain EM analysis showed additional density bound to CST, attributable to the addition of POT1(ESDL)/TPP1 in the two complexes (Fig. S2E). We determined cryo-EM structures of apo and ssDNA-bound CST–POT1(ESDL)/TPP1 at overall resolutions of 3.9 Å and 4.3 Å, respectively (Fig. 2; Fig. S3–4; Video S1; Table S1). Into these density maps, we could unambiguously place models of the individual subunits predicted by AlphaFold2^{36, 37} or previously determined by X-ray crystallography or cryo-EM as a starting point to build the models (see Table S1). The map quality of the apo CST–POT1(ESDL)/TPP1 reconstruction was higher (Fig. S3H), so we primarily used that map for model building. The resulting atomic model from the apo complex was then used as a starting model to build the ssDNA-bound complex. We observed continuous density for the POT1 flexible hinge and were able to build its Cα backbone de novo, generating an experimental structure of full-length POT1(ESDL) (Fig. 2D–E, Fig. S3–4; Video S1). The POT1 C-terminus, consisting of POT1^OB-3 and POT1^HJRL (Fig. 2A), is bound by the POT1-recruitment domain of TPP1 (TPP1^RD), consistent with previous crystal structures (PDB 5H65³⁸ and PDB 5UN7³⁹) (Fig. 2D–E). TPP1^RD does not interact directly with CST in either structure. The N-terminal OB fold of TPP1 (TPP1^OB) was not resolved, presumably due to flexible attachment, and TPP1^OB was indeed dispensable in the interaction of POT1(ESDL)/TPP1 with CST (Fig. S5A).

POT1(ESDL) is held in a single conformation stretched along the entire length of Ctc1 and buries a total of 5,246 Ų of solvent-accessible surface area (Fig. 2D–E). The structures of the apo and ssDNA-bound complexes had identical overall interactions between POT1(ESDL) and CST, but the addition of ssDNA allowed visualization of the POT1^OB-1 domain, which binds ssDNA in conjunction with POT1^OB-2 (Fig. 2C, E). Because the region of the map containing the POT1 N-terminal OB-folds was at lower resolution (Fig. S4H), we could only dock in the existing crystal structure of POT1 bound to 5’-TTAGGGTTAG-3’ (PDB 1XJV⁴⁰) and could not assign a register to the DNA or observe additional nucleotides of the [GGTTAG]₃ oligonucleotide. Without ssDNA, POT1^OB-1 is flexibly attached and thus remains unresolved in the map of the apo complex (Fig. 2B; Fig. S3).

Phosphorylation of POT1 mediates the interaction with CST

The hinge region of POT1(ESDL) containing the ESDL insertion sits in a positively charged cleft of Ctc1^OB-D (Fig. 3A; Fig. S5B). We term this region of POT1 the Ctc1 Cleft-Interacting Region (CCIR, aa 309–321). Although there is connectivity in the cryo-EM map, we could not unambiguously resolve the side chain positions, suggesting this interface is formed by weaker interactions (Fig. 3A; Fig. S5B). Notably, the ESDL insertion does not make specific contacts with Ctc1, suggesting that the observed complex is not the result of an artificial interaction of this mouse/human chimera. The POT1 CCIR contains aspartate residues that complement the basic cleft of Ctc1 as well as an abundance of serine residues, suggesting that this interaction may be regulated by phosphorylation of the CCIR to drive charge complementarity between the two proteins.

Figure 3. Phosphorylation-dependent interaction of POT1(ESDL) with CST.

See also Figure S5.

(A) Interaction between the POT1 CCIR and Ctc1. (Left) The black box indicates the region of the apo CST–POT1(ESDL)/TPP1 model shown in the right panels. (Right) Zoomed-in view showing surface electrostatic potential analysis of the POT1(ESDL) CCIR interacting with Ctc1^OB-D. The POT1(ESDL) CCIR is shown in cartoon (left) and surface (right) representation with electrostatic potential colored from red (−10 k_BT/e) to white (0 k_BT/e) to blue (+10 k_BT/e) (see also Fig. S5B). The ESDL insertion is colored salmon as in Fig. 1A, and residue numbering corresponds to the WT human POT1 sequence.

(B) FSEC analysis of the phosphorylation-dependent interaction between CST and POT1(WT)/GFP-TPP1 (left, black) and POT1(ESDL)/GFP-TPP1 (right, blue) in the presence of telomeric ssDNA (see also Fig. S5C–D). Traces without CST are shown as dashed lines, and traces containing POT1/TPP1 that have been treated with λ protein phosphatase (λPP) are shown in red. RFU: relative fluorescence units. The chromatograms for the WT mock and ESDL mock samples are the same as in Fig. 1E. They are reproduced here as the controls for the phosphatase experiment, which was performed simultaneously.

(C) Representative MP histograms showing the phosphorylation-dependent interaction between CST and POT1(WT)/GFP-TPP1 (left, black) and POT1(ESDL)/GFP-TPP1 (right, blue) in the presence of telomeric ssDNA (see also Fig. S5E). Experiments with λPP-treated POT1/TPP1 experiments are shown in translucent red. Peak maxima are indicated along with cartoons (peak 1: free POT1/TPP1, peak 2: free CST, peak 3: CST–POT1/TPP1) that correspond to complexes of that molecular mass to within 10%.

(D) Kinase assay with HeLa nuclear extract and filter-bound peptides corresponding to the CCIR regions of POT1, mPOT1b, mPOT1a, and POT1(ESDL). Phosphorylation is monitored using incorporation of γ-³²P-ATP.

(E) Quantification of MP results (Fig. S5E). Proportion bound POT1/TPP1 corresponds to the ratio of total counts in peak 3 (CST–POT1/TPP1 complex) divided by total counts of POT1/TPP1 (peak 1 and peak 3). Error bars represent the S.D. from three technical replicate experiments. POT1 construct details and names are provided in (H).

(F - G) Co-IPs of Myc-tagged POT1 protein variants and FLAG-tagged Ctc1 from 293T cells co-transfected with Myc-POT1, FLAG-TPP1, FLAG-Ctc1, FLAG-Stn1, and FLAG-Ten1. Western blots of Myc IPs were probed with anti-Myc and anti-FLAG antibodies to detect POT1 and Ctc1, respectively. Short and long exposures of the anti-FLAG probing for Ctc1 are shown. POT1 construct details and names are provided in (H).

(H) Sequences of phosphomimetic, alanine, and CP CCIR mutants. The interaction with CST is indicated as shown in (E-G). The “+” signifies and interaction matching that of the positive control, POT1(ESDL), and “−” signifies no interaction. “++” represents greater interaction than the positive control and “+/−“ and “−/+” indicate slightly weaker interactions in decreasing strength, respectively.

To test the role of phosphorylation, we treated the POT1/TPP1 proteins with λ protein phosphatase (λPP; Fig. S5C). The CST–POT1(ESDL)/TPP1 interaction was severely diminished by dephosphorylation of POT1(ESDL)/TPP1 in both FSEC and mass photometry (MP) analysis of complex formation (Fig. 3B–C; Fig. S5D–E). The propensity of the ESDL insertion to enhance phosphorylation was shown in a kinase assay in which HeLa cell nuclear extract was incubated with nitrocellulose membrane-bound synthetic peptides (Fig. 3D). Human POT1 and mPOT1b peptides containing the analogous CCIRs were phosphorylated more than the corresponding peptide from mPOT1a, consistent with the inability of mPOT1a to bind CST⁵, and the POT1(ESDL) peptide was phosphorylated to the greatest degree of the four (Fig. 3D). Together, this suggests that the ESDL insertion primarily functions to enhance the phosphorylation of human POT1 rather than to directly interact with CST.

Mass spectrometry efforts to determine the sites of phosphorylation in the Ser-rich CCIR were not informative, as the relevant peptide was not well-represented despite multiple trials with different proteases. To validate the structure and further characterize the phosphorylation events involved in complex formation, we generated phosphomimetic CCIR mutants (S-to-D) in POT1 and POT1(ESDL) as well as S-to-A mutants (Fig. 3E–H). MP was used to quantify binding efficiency by measuring the proportion of CST-bound POT1/TPP1 (MP peak at ~390 kDa) to free POT1/TPP1 (MP peak at ~140 kDa; Fig. 3C, E; Fig. S5E). In contrast to the sharp complex peak (σ ≤ 40) observed with POT1(ESDL)/TPP1, the CST–POT1/TPP1 peaks with CCIR phosphomimetic changes were broader (σ > 40) but still symmetric and centered at the same mass value (Fig. S5E). The broader peaks suggest that the phosphomimetic mutant complexes may be slightly more heterogeneous and less stable than the phosphorylated POT1(ESDL)/TPP1. The quantification by total peak counts (area under the Gaussian curve) masks this subtle difference, but importantly, the phosphomimetic CCIR mutants conferred λPP-resistant CST binding compared to both wild-type POT1/TPP1 and dephosphorylated POT1(ESDL)/TPP1 (Fig. 3E; Fig. S5E). Phosphomimetic substitutions at Ser317, Ser318, Ser320, and Ser322 increased the proportion of CST-bound POT1/TPP1 to almost the same level as observed with the ESDL insertion, and the interactions with the phosphomimetic proteins were resistant to phosphatase treatment (Fig. 3E; Fig. S5E). Similarly, in co-IP experiments that queried the interaction of POT1 with Ctc1, phosphomimetic CCIR substitutions of Ser317, Ser318, and Ser320 led to a robust interaction close to that seen with the ESDL version of POT1 (Fig. 3F). Greater interaction over that of POT1(ESDL) was observed with a combination of the phosphomimetic substitutions and the ESDL, suggesting that the insertion adds additional charge to provide the increased stability seen with the purified proteins (Fig. 3E–F, H). However, the in vitro and in vivo assays gave disparate results regarding the Ser322 position. In the co-IPs, the S322D substitution diminished the interaction despite the phosphomimetic substitutions at the other positions whereas in vitro, the S322D substitution was required for optimal binding of POT1/TPP1 to CST. The reason for this discrepancy remains to be determined. Nonetheless, both assays point to the critical role of serine phosphorylation in the CCIR of POT1 for Ctc1 binding. This conclusion was further substantiated by converting CCIR residues Ser317, Ser318, and Ser320 to alanine in POT1(ESDL), which abolished the ability of POT1(ESDL) to form a complex with Ctc1 in co-IP experiments (Fig. 3G–H).

POT1^OB-3 forms the primary interface with Ctc1

Human POT1 directly interacts with Ctc1 at two sites independent from the POT1^CCIR and the ESDL insertion (Fig. 2D–E). The primary interface contains POT1^OB-3, Ctc1^ARODL, and Ctc1^OB-D and buries 2,012 Ų of solvent-accessible surface area (Fig. 4A–B; Fig. S6A). This interface is formed by the C-terminal region of the POT1 hinge (aa 322–347), which folds back on POT1^OB-3 to create the surface to which Ctc1 binds. Part of this region (aa 332–347) has previously been crystallized³⁹, but in another crystal structure it was too flexible to be resolved³⁸ (Fig. 4A). In our structure, this region of the hinge is pinned to POT1^OB-3 at two sites (sites (i) and (ii)) of intramolecular interactions that limit its flexibility (Fig. 4A–B). POT1 Ser322 makes the furthest N-terminal self-interaction with POT1^OB-3 with Arg367 at site (i) to create a linchpin that locks aa 322–341 in place (Fig. 4B; Fig. S6A). Ser322 is mutated to leucine in CP²³, which would disrupt the hydrogen-bond interaction. Indeed, mutation of the corresponding Ser326 in POT1(ESDL) to alanine or leucine severely attenuates CST binding (Fig. 4C–D). This interaction is conserved, as the corresponding S328L mutation in mPOT1b abrogates CST binding (Fig. S6B). The S322A substitution also abolishes binding of the phosphomimetic CCIR mutants (Fig. 4C), consistent with the linchpin role of Ser322. At site (ii), aa 335–337 of the hinge interact with the C-terminus of POT1 (aa 627–629) in an anti-parallel β-strand configuration, as previously observed³⁹ (Fig. 4B; Fig. S6A). In the context of the full complex, these two points of contact secure the otherwise flexible hinge region against POT1^OB-3 (Fig. 4A–B).

Figure 4. CP mutations map to the primary Ctc1–POT1 interface.

See also Figure S6.

(A) Comparison of CST-bound POT1^OB-3/TPP1 (this study, colored) to unbound POT1^OB-3/TPP1 (PDB 5H65/5UN7^{38, 39}, grayscale).

(B) Close-up views of sites of interest at or near the interface between POT1 and Ctc1 (see also Fig. S6A). Residues colored in bright red are mutated in CP. (i) Self-interaction between POT1 Ser322 and Arg367. Hydrogen bonds are shown as yellow dashed lines. CP mutation S322L is predicted to disrupt this linchpin interaction. (ii) Self-interaction between POT1 hinge and C-terminus. (iii) Position of Gly503 in the hydrophobic core of Ctc1^ARODL predicts a destabilizing effect of the G503R CP mutation. (iv) Primary hydrophobic interface between POT1 hinge, POT1^OB-3, and Ctc1^ARODL. CP mutation H484P is predicted to disrupt the hydrophobic stacking interactions with POT1 Pro603 and Ctc1 His488 and Pro483. (v) Salt bridge between POT1 hinge residue Glu325 and Ctc1^OB-D residue Arg624.

(C) Co-IPs of Myc-tagged POT1 protein variants and FLAG-tagged Ctc1 from 293T cells transfected with Myc-POT1, FLAG-TPP1, FLAG-Ctc1, FLAG-Stn1, and FLAG-Ten1. Western blots of Myc IPs were probed with anti-Myc and anti-FLAG antibodies to detect POT1 and Ctc1, respectively. Short and long exposures of the anti-FLAG probing for Ctc1 are shown.

(D) Co-IPs of Myc-tagged POT1(ESDL) and FLAG-tagged Ctc1 WT and CP variants from 293T cells transfected with Myc-POT1, FLAG-TPP1, FLAG-Ctc1, FLAG-Stn1, and FLAG-Ten1. POT1(ESDL) CP contains the S326L mutation, analogous to S322L in POT1^CP (see (E)). Western blots of Myc IPs were probed with anti-Myc and anti-FLAG antibodies to detect POT1 and Ctc1, respectively. Short and long exposures of the anti-FLAG probing for Ctc1 are shown.

(E) Sequences of CCIR mutants and their interaction with CST as shown in (C-D).

Our structure also provides insight into the Ctc1 G503R CP mutation^{17, 41, 42}. Although Gly503 is not at the Ctc1–POT1 interface, it is in the hydrophobic core of Ctc1^ARODL at site (iii). A G-to-R substitution would destabilize Ctc1^ARODL, which forms the primary interaction with POT1 (Fig. 4B; Fig. S6A). A loop in Ctc1^ARODL, which interacts with POT1, is the location of another CP mutation, H484P²⁴. At site (iv), His484 participates in a network of hydrophobic stacking interactions with Ctc1 Pro483, Ctc1 His488, and POT1 Pro603. This interface is stabilized by hydrophobic packing of the Ctc1^ARODL loop against the POT1 tail (aa 630–634), POT1^OB-3, and its hinge (Fig. 4B). Both the G503R and H484P substitutions in Ctc1 abrogate binding to POT1(ESDL) (Fig. 4D).

The end of the POT1 tail (632-DVI-634 in humans) was previously implicated in CST binding by mPOT1b⁵. Substitution of the sequence in mPOT1b (DII) to the mPOT1a sequence (NVV) diminished the interaction with CST⁵, and our structure explains the importance of the bulky hydrophobic residues at this site. The structure also shows that POT1 Asp632 is at an appropriate distance to hydrogen bond with Ctc1 His489, which would further stabilize the complex (Fig. 4B; Fig. S6A). Finally, there is an additional salt bridge between POT1 Glu325 and Ctc1 Arg624 at site (v), which are in the POT1 hinge and Ctc1^OB-D, respectively (Fig. 4B; Fig. S6A).

POT1 blocks the CST ssDNA anchor site

The second OB fold of POT1, POT1^OB-2, also interacts with CST (Fig. 2C, E). The resolution of the CST–POT1^OB-2 interface was in the 6-Å range (Fig. S3H; Fig. S4H) but still allowed for visualization of the docked structures. POT1^OB-2 sits between Ctc1 and the C-terminal lobe of Stn1 (Stn1^C) at Ctc1’s ssDNA anchor site¹³ (Fig. 5A). The POT1 ssDNA-binding interface is facing outwards and is bound to the [GGTTAG]₃ ssDNA (Fig. 2E). Helix αD of POT1 directly superposes with the ssDNA observed in the cryo-EM structure of CST¹³ and the N-terminal part of the POT1 hinge runs through the ssDNA anchor site towards the basic cleft in Ctc1^OB-D (Fig. 5A; Fig. 3A). Stn1^C, which can bind Ctc1 in multiple configurations, is in a position consistent with the structure of monomeric CST in the absence of ssDNA¹³ (Fig. 5A). There was additional low-resolution density of a flexible loop in Ctc1 (aa 909–927), which has not been resolved in other structures of CST, and we used poly-alanine stubs to indicate its presence. The low resolution indicates that POT1^OB-2 may only weakly interact with this loop (Fig. 5A).

Figure 5. POT1^OB-2 interacts with the CST ssDNA anchor site, precluding formation of the CST–Polα/primase PIC.

See also Figure S7.

(A) Interaction between POT1^OB-2 and Ctc1 at the CST ssDNA anchor site. (Left) The black box indicates the region of the apo CST–POT1(ESDL)/TPP1 model shown in the right panels. (Right) Zoomed-in views of the interface. The first panel shows the ssDNA-bound CST structure (PDB 6W6W¹³, gray with DNA colored cyan). The second panel shows the same view of the CST–POT1(ESDL)/TPP1 structure (this study, colored). Ctc1 aa 909–927 (density map shown; contour threshold 0.15) are modeled as poly-alanine stubs, and no ssDNA is bound to Ctc1. The two structures are superposed in panel three.

(B) Negative-stain EM 2D averages of ssDNA-bound CST–POT1(ESDL)/TPP1 (left) showing three major conformations depicted by the cartoon schematics (right). 2D averages are flipped or rotated by 90° increments from Fig. S2E into similar orientations for ease of comparison and are sorted by number of particles per class from the most populous class at top left to the least populous class at bottom right. The 2D averages that correspond to each cartoon state are indicated with black, gray, or white circles. The scale bar represents 333 Å (See also Fig. S2E).

(C) The CST–POT1(ESDL)/TPP1–ssDNA complex (left) and its superposition with the CST–Polα/primase recruitment complex (RC, PDB 7U5C¹⁵, middle) and pre-initiation complex (PIC, PDB 8D0B¹⁶, right). The CST–POT1(ESDL)/TPP1–ssDNA complex is shown as an opaque surface, and CST–Polα/primase complexes are shown in cartoon representation with transparent surfaces. Orthogonal views show that POT1(ESDL)/TPP1 binding does not obstruct the major interface of the RC, but POT1^OB-1/2 and Stn1^C would interfere with binding of the POLA1 catalytic core to the CST ssDNA anchor site in the PIC (see also Fig. S7A–B).

(D) Cartoon schematics showing how CST–POT1/TPP1 is incompatible with PIC formation but could form RC-like complexes in both monomeric and dimeric forms.

The particles of the ssDNA-bound CST–POT1(ESDL)/TPP1 dataset were more heterogenous than those in the apo dataset; a subset of particles appeared to be missing density for the POT1 N-terminal OB-folds and Stn1^C (Fig. S2E; Fig. S4D). Loss of this density suggests that ssDNA may be bound to Ctc1, in contrast to the major conformation described above. POT1 has a greater affinity for telomeric ssDNA than CST^{43, 44} (Fig. S6C) and is thus expected to outcompete CST for access to the ssDNA in limiting conditions. The CST–POT1(ESDL)/TPP1 fusion complex bound the telomeric [GGTTAG]₃ ssDNA with the same high affinity (K_D,app ~100 pM) as POT1(ESDL)/TPP1, indicating that the presence of CST did not change the ssDNA-binding affinity of POT1 (Fig. S6C). However, the reconstitution was performed with an excess of [GGTTAG]₃ that could result in both proteins being bound to DNA. We could not obtain a high-resolution map for the conformation in which both CST and POT1 were bound to ssDNA, likely because monomeric ssDNA-bound CST is flexible¹³. The negative-stain EM data raised the possibility that this conformation is an intermediate in the dimerization of CST–POT1/TPP1 (Fig. 5B). CST can dimerize in the presence of ssDNA¹³, and CST dimers bound by POT1/TPP1 are observed in negative-stain EM 2D averages (Fig. 5B; Fig. S2E). CST dimerization requires displacement of POT1^OB-2 and Stn1^C13, but the POT1 C-terminus appears to be still attached by the POT1^OB-3 and CCIR interactions. The dimer interface observed for CST–POT1/TPP1 appears similar to that of CST alone, though we note that the position of the POT1 C-terminus interaction would block the previously observed downstream tetramerization and decamerization of CST¹³. The CST–POT1(ESDL)/TPP1 dimers, which were a minor fraction of the particles observed by negative-stain EM (~6%), could not be found in the cryo-EM dataset (Fig. S4), indicating that they may not withstand the vitrification process.

POT1/TPP1 can recruit CST–Polα/primase in the RC but not in the PIC state

The ssDNA anchor site of CST is critical for formation of the CST–Polα/primase PIC in which the POLA1 catalytic core contacts the telomeric ssDNA template and is stabilized by Stn1^C, which is in a different position relative to the monomeric and DNA-bound states of CST¹⁶. POT1/TPP1 binding to CST is incompatible with Polα/primase binding in a PIC-like conformation when POT1^OB-1, POT1^OB-2, and Stn1^C are engaged (Fig. 5C; Fig. S7A–B; Video S2). Therefore, we propose that POT1/TPP1 does not bind CST–Polα/primase in the PIC conformation. In contrast, the major interface between Ctc1 and Polα/primase in the auto-inhibited RC-like conformation is orthogonal to the POT1/TPP1 interface and is unobstructed, thus allowing for the formation of a POT1/TPP1-bound RC-like complex (Fig. 5C–D; Fig. S7A–B; Video S3). There is a minor clash between PRIM2 and POT1^HJRL, but the cryo-EM structure of the RC was determined in cross-linking conditions, limiting the flexibility of the complex¹⁵. This clash could be alleviated by the flexibility of CST relative to Polα/primase observed in multi-body refinement analysis⁴⁵ of the CST–Polα/primase RC complex (Fig. S7C; Videos S4). Structural heterogeneity of CST–Polα/primase in an RC-like state was previously observed and proposed to be partially attributed to binding of the POLA2 N-terminal domain (POLA2^NTD), which is attached by a flexible linker to POLA2¹⁵. Indeed, AlphaFold-Multimer predicts an interaction between the POLA2^NTD and Ctc1 (Fig. S7D–E), supporting the existence of flexible RC-like states observed in low-resolution cryo-EM maps¹⁵.

We propose that the OB-2-disengaged state, when telomeric ssDNA is not limiting, represents a transient intermediate to the formation of a dimer of CST–POT1/TPP1 (Fig. 5B). A CST–POT1/TPP1 dimer is likely to form at the telomere in the context of shelterin, which is fully dimeric and hence contains two copies of POT1/TPP1 each capable of binding to CST⁴⁶. The high local concentrations of two CST molecules brought together by a shelterin dimer would favor dimerization. CST dimerization would also prevent PIC formation¹⁶ and is compatible with the RC¹⁵ (Fig. 5D).

POT1 bound to CST inhibits CST–Polα/primase C-strand synthesis in vitro

Our structural model predicts that the purified POT1(ESDL), which is competent for CST binding compared to POT1(WT), will inhibit C-strand synthesis by CST–Polα/primase by limiting formation of an active PIC. Previous experiments showed that POT1(WT)/TPP1 produced in E. coli did not have a strong inhibitory effect on CST–Polα/primase activity⁴⁷, but our data indicate that this protein complex is unable to interact with CST. Thus, we used our insect cell-derived POT1(WT)/TPP1 and POT1(ESDL)/TPP1 to directly assay the effect of POT1 binding on inhibition of active CST–Polα/primase. As previously described, when CST–Polα/primase co-purified from HEK293T cells was incubated with a synthetic telomeric G-strand template (9xTEL or 15xTEL, representing 9x or 15x of ss TTAGGG repeats), a characteristic 6-nt repeat pattern of products was produced⁴⁷ (Fig. 6A–B). Extension on the 9xTEL template was only inhibited by POT1(WT)/TPP1 at high concentrations (10-fold excess over template and CST–Polα/primase) as previously observed⁴⁷. On the other hand, POT1(ESDL)/TPP1 was a strong inhibitor of the C-strand synthesis reaction with an IC₅₀ 7-fold lower than that of the WT protein. The preferential inhibition of C-strand synthesis by POT1(ESDL)/TPP1 was also clear on the 15xTEL template, although higher concentrations were required on the longer template containing more binding sites (Fig. 6A–B).

Figure 6. CST binding enhances POT1/TPP1 inhibition of telomeric C-strand synthesis.

See also Figure S8.

(A) DNA synthesis by 50 nM CST–Polα/primase on 50 nM ssDNA templates consisting of 9 or 15 telomeric TTAGGG repeats (9xTEL and 15xTEL, respectively). Products labeled with α-³²P-dATP. LC3: an oligonucleotide loading control.

(B) Quantification of the data in (A), normalized to reactions without added POT1/TPP1. IC₅₀ +/− uncertainty of the fit to the data points. An independent repeat of the 9xTEL experiment gave equivalent results (Fig. S8A–C).

(C) FSEC analysis of the interaction between CST and POT1(ESDL ΔOB1)/GFP-TPP1 in the absence of telomeric ssDNA. The trace without CST is shown as a dashed line. RFU: relative fluorescence units.

(D) Design of pre-primed templates to monitor primer extension by Polα. RNA sequence is shown in red.

(E) CST–Polα extension of pre-primed templates. M: telomerase reaction products as size markers. The two panels have equal exposure, but one-ninth as much dT42–2xTEL product as 9xTEL product was loaded to compensate for the higher incorporation with the former template. LC1, LC2, LC3: three oligonucleotide loading controls.

(F) Quantification of the experiment shown in (E) (left). IC₅₀ values include replicate experiments on separate days, with errors encompassing the range of values obtained.

(G) Quantification of the experiment shown in (E) (right). Only partial inhibition occurred at the highest POT1/TPP1 concentrations, so IC₅₀ values were not calculated.

To test the role of the ssDNA-binding activity of POT1, we generated a POT1(ESDL ΔOB1) construct missing POT1^OB-1 (aa 1–145), rendering it unable to bind DNA⁴⁸. Consistent with our cryo-EM structures showing that POT1^OB-1 does not directly interact with CST (Fig. 2), deletion of this domain had no effect on complex formation between POT1(ESDL ΔOB1)/TPP1 and CST (Fig. 6C). POT1(ESDL ΔOB1)/TPP1 only weakly inhibited C-strand synthesis (Fig. S8A–C).

To isolate the effect of POT1 on CST and Polα in a primase-independent setting, we designed pre-primed DNA templates (Fig. 6D; Fig. S8D). The RNA primer length and sequence were as previously determined⁴⁷, and the RNA primer was made as a chimera with a 12-nt DNA to specify its binding site on the DNA template by base pairing. Extension by CST–Polα/primase no longer required ribonucleotides, confirming that the reaction was now independent of primase activity (Fig. S8E). Polα extension of the primer on the pre-primed telomeric DNA template was again inhibited most strongly by POT1(ESDL)/TPP1, with an IC₅₀ 4–5-fold lower than for POT1(WT)/TPP1 (Fig. 6E–F). Notably, the IC₅₀ for Polα extension was considerably higher (weaker inhibition) than for reactions initiated by primase (Fig. 6B), consistent with greater inhibition occurring at the initiation step. POT1(ESDL ΔOB1)/TPP1 was too weak of an inhibitor to measure IC₅₀ (Fig. 6E–F), suggesting that telomeric DNA binding is important for inhibition by POT1/TPP1.

To confirm that DNA binding was important for POT1/TPP1 inhibition of C-strand synthesis, we substituted the telomeric repeats of the pre-primed template with poly(dT) (Fig. 6D). This pre-primed template resulted in much stronger synthesis due to the lack of inhibitory G-quadruplex folding⁴⁷. On the poly(dT) template, which is poorly bound by both POT1/TPP1 and CST^{40, 44, 49}, all three POT1/TPP1 complexes were similarly weak inhibitors of DNA synthesis (Fig. 6E; Fig. 6G). Thus, telomeric DNA binding, in addition to CST binding, contributes strongly to the inhibition of C-strand synthesis by POT1/TPP1. The enhanced inhibition of POT1(ESDL)/TPP1 also occurred with 10-fold higher DNA template to CST–Polα/primase ratios, where re-priming is repressed and the extension is limited by the inherent processivity of the polymerase (Fig. S8F–G). In summary, the markedly greater inhibition of C-strand synthesis by POT1(ESDL)/TPP1 in vitro provides functional support for the structure-based model where POT1 prevents formation of the CST–Polα/primase PIC.

Discussion

Decades of research have revealed how shelterin prevents inappropriate DNA damage signaling and DNA repair at telomeres¹. Similarly, extensive studies have established the role of shelterin in the recruitment and regulation of telomerase, the enzyme that solves part of the end-replication problem by extending the G-rich telomeric repeat strand⁹. The data presented here illuminate the third function of shelterin, which is to promote and regulate the synthesis of the C-rich telomeric repeats by CST–Polα/primase, the second enzyme needed to solve the end-replication problem⁴ (Fig. 7A).

Figure 7. Model for telomere maintenance by shelterin and CST–Polα/primase.

(A) Shelterin-mediated telomere protection and maintenance. Shelterin recruits and regulates two telomere maintenance enzymes: telomerase and CST–Polα/primase. Adapted from².

(B) Model for the recruitment and regulation of CST–Polα/primase by POT1. CST–Polα/primase is recruited to telomeres in the auto-inhibited RC-like state by phosphorylated POT1, preventing activation of CST–Polα/primase during 5’ end resection and telomerase action following canonical DNA replication. After telomerase has elongated telomeres, a proposed switch results in dephosphorylation of POT1 and subsequent release of CST–Polα/primase to the telomeric ssDNA. When released, CST–Polα/primase can readily form the PIC and execute fill-in synthesis, the final step of telomere replication.

The structures of CST bound to POT1/TPP1 provide insights into the mechanism by which POT1 recruits and regulates CST–Polα/primase (Fig. 7B). The complex formed by POT1/TPP1 and CST is compatible with the binding of CST to Polα/primase in the auto-inhibited state, which we previously proposed to be the recruitment state of CST–Polα/primase. In contrast, the structure of CST bound to POT1/TPP1 is not compatible with the active state of the fill-in machinery, and our biochemical data confirm that POT1/TPP1-binding inhibits the ability of CST–Polα/primase to form an active fill-in enzyme. Therefore, CST–Polα/primase must undergo a major conformational switch at telomeres and its release from the POT1/TPP1 subunits of shelterin is required for this switch. The data further reveal the critical role of phosphorylation of human POT1 in regulating the recruitment and release of CST–Polα/primase. Our structural and biochemical data suggest that phosphorylation in a short Ser-rich part of the POT1 hinge (the CCIR) promotes insertion of the CCIR into a basic cleft of Ctc1, thereby allowing complex formation and recruitment of CST–Polα/primase to telomeres (Fig. 7B).

We propose that phosphorylated POT1 positions CST–Polα/primase at telomeres and holds the complex in an inactive state until the replication machinery has completed DNA synthesis, 5’ end-resection has taken place, and telomerase has extended the leading- and lagging-end telomere termini. At that point, we imagine that dephosphorylation of POT1 releases CST–Polα/primase so that the enzyme can transition into the active state and execute fill-in synthesis as the last event in the telomere end processing reactions (Fig. 7B). This scenario is compatible with data obtained on the order of processing events at human telomeres (^{6, 50, 51}; reviewed by Cai and de Lange²).

The ability of POT1 to hold CST–Polα/primase in an inactive state until telomerase has initiated G-strand extension may be needed to ensure that CST does not prematurely inhibit telomerase. CST can block telomerase from reinitiating G-strand synthesis in vitro⁵² and loss of CST leads to inappropriate telomere extension by telomerase in vivo²⁸. This ability of CST to terminate telomerase-mediated telomere extension requires its ssDNA-binding activity⁵². Since POT1 blocks the CST ssDNA anchor site, we predict that POT1-bound CST is an ineffective telomerase inhibitor. Release of CST–Polα/primase from POT1 would therefore both initiate fill-in synthesis and terminate further telomere extension by telomerase.

Human POT1 has at least two critical functions. First, it represses the activation of ATR signaling at telomeres by blocking the binding of RPA to the ssDNA^53–55 and by preventing the Rad17/9-1-1 complex from engaging the telomeric 5’ end⁵⁶. Second, as shown here, POT1 promotes the maintenance of telomeric DNA through its ability to recruit CST–Polα/primase. In most metazoan shelterin complexes, a single POT1 must execute both functions⁵⁷. In contrast, a rodent lineage that includes mice and rats has duplicated the POT1 locus and evolved two distinct POT1 proteins, one of which is dedicated to ATR repression (POT1a) and the other to CST–Polα/primase recruitment (POT1b). Interestingly, mouse POT1b brings CST–Polα/primase to telomeres throughout the cell cycle⁵ whereas in human cells, CST peaks at telomeres in late S/G2²⁸. Perhaps the presence of POT1a allows POT1b to permanently associate with CST whereas the ancestral POT1 needs to be regulated by phosphorylation to toggle between a CST-bound form and a form that can block ATR signaling.

Our results show how a subset of CP mutations affect telomeric CST recruitment, explaining the POT1 S322L genetic data and providing further evidence for the causative link between telomere maintenance defects and CP pathogenesis. Four CP mutations map to regions involved in CST–Polα/primase recruitment: Ctc1 V665G maps to the primary interface observed in the RC, and POT1 S322L, Ctc1 H484P, and Ctc1 G503R affect CST–POT1 complex formation. Furthermore, CP mutations in Ctc1^OB-B are numerous and potentially affect the predicted interactions between Ctc1 and the POLA2^NTD, though further experiments are needed to confirm the AlphaFold-Multimer predictions. Together, these CP mutations highlight CST–Polα/primase recruitment as a critical step in normal telomere maintenance. Once the kinase(s) responsible for regulation of this step is identified, it is possible that additional patient mutations will be discovered within the kinase or its regulatory factors.

These findings open another avenue for research on the regulation of telomere maintenance. Whereas telomere elongation by telomerase has been studied for more than 30 years, examination of the equally important synthesis of the telomeric C-strand has lagged (pun intended). We expect that the identification of the kinases/phosphatases that control CST recruitment and its subsequent release will reveal how the cell cycle controls the critical last step in telomere maintenance.

Limitations of the study

Here, we propose a model for the regulated recruitment of CST–Polα/primase by the shelterin subunit POT1 based on our structural and biochemical data. As discussed above, further in vivo work will be required to determine the kinase, phosphatase, and exact cell cycle timing of this process. Because the kinase was not known, we did not formally show native phosphorylation of human POT1 in this study, though our findings are highly suggestive. Instead, we used a modified POT1(ESDL) construct to increase the stability of the CST–POT1/TPP1 complex for structural studies. The phosphomimetic mutations in POT1 confer similar binding, but they are slightly weaker interactors. We speculate that this could be a feature, not a bug, of the human system compared to that in the mouse, as discussed above. To confirm this, additional biochemical analysis using a natively phosphorylated human POT1 is required (once the kinase is identified). Similarly, the TPP1^TID was truncated to limit structural heterogeneity, but given the proximity of many CP mutations to the Ctc1^OB-B–TPP1^TID interface, it will be interesting to continue to explore the functional role of this interaction. Finally, although the in vitro C-strand synthesis data are highly suggestive of our model, the structural information provided here are limited to CST and POT1. Additional study is needed to define the structure of the proposed POT1–CST–Polα/primase ternary recruitment complex.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Titia de Lange (delange@rockefeller.edu).

Materials Availability

Unique reagents generated in this study are available from the lead contact with a completed Materials Transfer Agreement.

Data and Code Availability

• The cryo-EM maps have been deposited at the Electron Microscopy Data Bank under accession codes EMD-40659 (apo CST–POT1/TPP1 complex) and EMD-40660 (ssDNA–CST–POT1/TPP1 complex), and the coordinates have been deposited in the Protein Data Bank under accession codes PDB 8SOJ and PDB 8SOK, respectively. Motion-corrected micrographs have been deposited in the Electron Microscopy Public Image Archive⁵⁹ under accession codes EMPIAR-12046 and EMPIAR-12047, respectively. This study also uses data previously reported² and deposited under the accession code EMPIAR-11131. The data are publicly available as of the date of publication.

• This paper does not report original code.

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

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

All cell lines and source materials used in this study are listed in the key resources table.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal Anti-DYKDDDDK Tag (D6W5B) antibody	Cell Signaling Technology	Cat#14793; RRID:AB_2217020
Mouse monoclonal Anti-c-Myc (9E10) antibody	Sigma-Aldrich	Cat#M4439; RRID:AB_439694
Mouse monoclonal Anti-Strep-tag antibody	IBA Lifesciences	Cat#2-1507-001; RRID:AB_513133
Mouse monoclonal Anti-HA.11 antibody (16B12)	Covance	Cat#MMS-101R; RRID:AB_291262
Bacterial and Virus Strains
E. coli: MAX Efficiency™ DH10Bac Competent Cells	Thermo Fisher	Cat#10361012
E. coli: One Shot™ TOP10 Chemically Competent E. coli	Thermo Fisher	Cat#C404003
Chemicals, Peptides, and Recombinant Proteins
Cellfectin® II Reagent	Thermo Fisher	Cat#10362100
DMEM	Corning	Cat#10-013-CV
HyClone Bovine Calf Serum (BCS)	Cytiva	Cat#SH30073.03
L-Glutamine (200 mM)	Gibco	Cat#25030081
Penicillin-Streptomycin (10,000 U/mL)	Gibco	Cat#15140122
MEM Non-Essential Amino Acids Solution (100X)	Gibco	Cat#11140050
Phosphate-Buffered Saline, 1X without calcium and magnesium, pH 7.4 ± 0.1	Corning	Cat#21-040-CM
cOmplete™, EDTA-free Protease Inhibitor Cocktail	Roche	Cat#11873580001
PhosSTOP™	Roche	Cat#4906845001
Pierce™ Anti-c-Myc Magnetic Beads	Thermo Fisher	Cat#88842
Benzonase® Nuclease	Millipore Sigma	Cat#E1014
NuPAGE™ LDS Sample Buffer (4X)	Thermo Fisher	Cat#NP0007
Biotinylated peptides	Kinexus Bioinformatics Co.	N/A
ATP, [γ-32P]	Perkin Elmer	Cat#BLU502Z
Strep-Tactin® Superflow® high capacity resin	IBA Lifesciences	Cat#2-1208-025
Lambda protein phosphatase (expressed and purified in-house from E. coli)	N/A	UNIPROT: P03772
Fos-Choline-8, Fluorinated, Anagrade	Anatrace	Cat#F300F
T4 Polynucleotide Kinase	New England Biolabs	Cat#M0201
Hybond®-XL hybridization membranes	GE Healthcare	Cat#GERPN2020S
Deposited Data
Cryo-EM data of CST–Polα/primase recruitment complex	Cai et al., 2022	EMPIAR-11131
Cryo-EM map of apo CST–POT1/TPP1 complex	This paper	EMD-40659
Cryo-EM map of ssDNA–CST–POT1/TPP1 complex	This paper	EMD-40660
Coordinates for apo CST–POT1/TPP1 complex	This paper	PDB8SOJ
Coordinates for ssDNA–CST–POT1/TPP1 complex	This paper	PDB8SOK
Motion-corrected micrographs for apo CST–POT1/TPP1 complex dataset	This paper	EMPIAR-12046
Motion-corrected micrographs for ssDNA–CST–POT1/TPP1 complex dataset	This paper	EMPIAR-12047
Experimental Models: Cell Lines
HEK 293T/17	ATCC	Cat#CRL-11268; RRID:CVCL_1926
Sf9	Thermo Fisher	Cat#11496015; RRID:CVCL_0549
Tni	Expression Systems	Cat#94-002F; RRID:CVCL_RY32
HEK 293T	ATCC	Cat#CRL-3216; RRID:CVCL_0063
HeLa1.3	Takai et al., 2010	RRID:CVCL_B5DA
Oligonucleotides
d[GGTTAG]3 oligonucleotide	Thermo Fisher	N/A
d[TTAGGG]9 9xTEL oligonucleotide	IDT	N/A
d[TTAGGG]15 15xTEL oligonucleotide	IDT	N/A
d[T]42[TTAGGG]2CTGGAACCTAGT top strand for dT42-2xTEL pre-primed template	IDT	N/A
[TTAGGG]9CTGGAACCTAGT top strand for 9xTEL pre-primed template	IDT	N/A
d[ACTAGGTTCCAGCCCT]r[AACCCUAA] bottom strand for pre-primed template	IDT	N/A
Recombinant DNA
See Table S1 for a list of recombinant DNA constructs	N/A	N/A
Software and Algorithms
EMAN2 v2.06	Tang et al., 2007	https://blake.bcm.edu/emanwiki/EMAN2 RRID:SCR_016867
RELION-3.1	Zivanov et al., 2018	http://www2.mrc-lmb.cam.ac.uk/relion RRID:SCR_016274
SerialEM	Mastronarde, 2005	http://bio3d.colorado.edu/SerialEM/ RRID:SCR_017293
cryoSPARC v4.1.1	Punjani et al., 2017	https://cryosparc.com/ RRID:SCR_016501
PHENIX v1.20.1	Adams et al., 2010 Liebschner et al., 2019	https://www.phenix-online.org/ RRID:SCR_014224
3DFSC	Tan et al., 2017	https://3dfsc.salk.edu/
UCSF Chimera v1.13.1	Pettersen et al., 2004	https://www.cgl.ucsf.edu/chimera/ RRID:SCR_004097
UCSF ChimeraX v1.5	Pettersen et al., 2021	https://www.cgl.ucsf.edu/chimerax/ RRID:SCR_015872
AlphaFold Protein Structure Database	Jumper et al., 2021 Varadi et al., 2021	https://alphafold.ebi.ac.uk/ RRID:SCR_023662
Coot v0.9.8.7	Emsley et al., 2010	http://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/ RRID:SCR_014222
ISOLDE v1.5	Croll, 2018	https://tristanic.github.io/isolde/
AlphaFold-Multimer v1.0	Evans et al., 2022	https://github.com/google-deepmind/alphafold
COSMIC2	Cianfrocco et al., 2017	https://cosmic-cryoem.org/
AcquireMP v2022R1	Refeyn	https://www.refeyn.com/
DiscoverMP v2022R1	Refeyn	https://www.refeyn.com/
ImageJ	Schneider et al., 2012	https://imagej.net/ RRID:SCR_003070
GraphPad Prism v9.5.1	GraphPad	http://www.graphpad.com/ RRID:SCR_002798
ImageQuant	GE Life Sciences	http://www.gelifesciences.com/webapp/wcs/stores/servlet/catalog/en/GELifeSciences-us/products/AlternativeProductStructure_16016/29000605 RRID:SCR_014246
KaleidaGraph	Synergy Software	http://www.synergy.com/ RRID:SCR_014980
PyMOL v2.5.4	Schrödinger	http://www.pymol.org/ RRID:SCR_000305
Adobe Illustrator	Adobe	http://www.adobe.com/products/illustrator.html RRID:SCR_010279
Gautomatch v0.53	MRC LMB (Kai Zhang)	https://www2.mrc-lmb.cam.ac.uk/download/gautomatch-053/
CTFFIND-4	Rohou and Grigorieff, 2015	http://grigoriefflab.janelia.org/ctffind4 RRID:SCR_016732
MotionCor2	Zheng et al., 2017	https://emcore.ucsf.edu/cryoem-software RRID:SCR_016499
Other
Graphene Oxide on Quantifoil® R1.2/1.3 Cu, 400 mesh TEM grid	Electron Microscopy Sciences	Cat#GOQ400R1213Cu100
Amersham™ Typhoon™ NIR	Cytiva	Cat#29238583
HiLoad Superdex 200 16/600 PG	Cytiva	Cat#28989335
Superose 6 Increase 10/300 GL	Cytiva	Cat#29091596
HiTrap Heparin HP 1 mL	Cytiva	Cat#17040601
SW 41 Ti Swinging-Bucket Rotor	Beckman Coulter	Cat#331362

METHOD DETAILS

DNA construct generation

The DNA constructs used in this study are listed in Table S2. For recombinant insect cell expression, the biGBac⁶⁰ vector system was used. All protein sequences are numbered according to the canonical sequence entry in UniProt. Recombinant bacmids were generated from the plasmids in Table S2 using MAX Efficiency DH10Bac competent cells (Thermo Fisher Scientific) and transfected into Sf9 insect cells (Thermo Fisher Scientific) with Cellfectin II Reagent (Thermo Fisher Scientific) to generate a P1 baculovirus stock. P1 baculovirus was amplified in adherent Sf9 insect cells to generate P2 and P3 stocks, and the P3 virus was used to infect Tni suspension insect cell culture (Expression Systems) for protein expression.

Co-immunoprecipitation

293T cells were grown in DMEM with 10% (v/v) bovine calf serum (HyClone), 2 mM L-glutamine (Gibco), 100 U/mL penicillin (Gibco), 0.1 mg/mL streptomycin (Gibco), and 0.1 mM nonessential amino acids (Gibco). 2×10⁶ 293T cells were plated in a 10-cm dish at 20–24 h prior to transfection by calcium-phosphate precipitation. Transfections were done using plasmid DNA as indicated such that total DNA did not exceed 20 μg per plate. Each experiment combined two identical 10-cm plates. At 16 h post-transfection, the medium was changed, and 24 h later, cells were harvested by flushing with 1x phosphate-buffered saline (PBS; Corning), frozen in liquid nitrogen (LN₂), and resuspended in 1 mL lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 10% (v/v) glycerol, 0.1% (v/v) NP-40, cOmplete EDTA-free protease inhibitor cocktail (Roche), and PhosSTOP phosphatase inhibitor mix (Roche)), and incubated on ice for 30 m. After centrifugation at 16,000 rcf for 10 m at 4°C, input samples were taken from the supernatant. 20 μL of equilibrated Pierce Anti-c-Myc magnetic beads slurry (Thermo Fisher) and 1 μL benzonase (Millipore Sigma) were added to the supernatant, and samples were incubated with end-over-end rotation for 2–3 h at 4°C. Beads were washed 3 times at 4°C with lysis buffer and immunoprecipitated proteins were eluted with 100 μL of 1x NuPAGE LDS buffer (Thermo Fisher). Samples were boiled for 5 m before separation on SDS-PAGE.

Peptide phosphorylation assay

The phosphorylation assay was performed as previously described⁶¹ with some modifications. Biotinylated peptides of the POT1 protein CCIR regions (POT1 aa 313–327; mPOT1b aa 319–333; mPOT1a aa 319–333; POT1(ESDL) aa 313–331) were produced by Kinexus Bioinformatics Co. HeLa cells cultured in the same medium as the 293T cells were used for isolation of nuclear extract⁶¹. 6×10⁷ cells were trypsinized, rinsed with cold PBS, and washed twice with 2 mL cytoplasm buffer (10 mM Tris-HCl pH 7, 10 mM NaCl, 3 mM MgCl₂, 30 mM sucrose, and 0.5% NP-40) on ice. After centrifugation, the pellet was rinsed twice with 2 mL of CaCl₂ buffer (10 mM Tris-HCl pH 7, 10 mM NaCl, 3 mM MgCl₂, 30 mM sucrose, and 100 μM CaCl₂). The pellet was resuspended in 1 mL buffer C (20 mM Tris-HCl pH 7.9, 20% glycerol, 100 mM KCl, and 0.2 mM EDTA), and centrifuged at 20,000 rcf for 30 m at 4°C. The supernatant was collected, stored at −80°C, and used as nuclear extract for the kinase assay. The nitrocellulose membrane containing the spotted peptide array was rinsed once with reaction buffer (20 mM Tris-HCl pH 7.9, 50 mM KCl, 10 mM MgCl₂, 1 mM DTT, cOmplete EDTA-free protease inhibitor cocktail, PhosSTOP) and subsequently incubated in 4 mL of reaction buffer containing 50 μM ATP, 50 μCi/μmol [γ-³²P] ATP, and nuclear extract (16 μg protein) for 45 m at 30°C. EDTA (15 mM final concentration) was added to stop the reaction, and the membrane was rinsed 10 times with cold 1 M NaCl. The membrane was rinsed once with wash buffer (4 M guanidine, 1% (v/v) SDS, 0.5% 2-mercaptoethanol (β-ME)) and exposed for radiography (Typhoon Biomolecular Imager, GE Healthcare).

Protein expression and purification

50 mL of P3 baculovirus was used per 500 mL of Tni culture, infected at a cell density of 2×10⁶ cells/mL. Infected cells were grown in spinner flasks at 150 rpm for 72 h at 27°C. Cells were harvested by centrifugation (500 rcf) and transferred to a syringe before flash freezing droplets in LN₂.

CST was purified as previously described¹⁵. For CST, frozen pellets were lysed by cryogenic milling (Retsch) and the cryo-milled powder was resuspended in a buffer containing 20 mM Tris pH 8.0, 500 mM NaCl, 15 mM β-ME, 20 mM imidazole, 10% glycerol, and 1 mM phenylmethylsulfonyl fluoride (PMSF), supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche). The lysate was cleared by centrifugation at 40,000 rcf for 1 h at 4°C. Supernatants were incubated with end-over-end rotation for 1 h at 4°C with Ni-NTA resin (Invitrogen) equilibrated with a buffer containing 20 mM Tris pH 8.0, 500 mM NaCl, 15 mM β-ME, 20 mM imidazole, and 5% glycerol and subsequently washed with 20–50 column volumes (CV) of the same buffer. Bound protein was eluted in a buffer containing 20 mM HEPES (pH 7.5), 150 mM NaCl, 0.5 mM tris(2-carboxyethyl)phosphine (TCEP), 250 mM imidazole, and 5% glycerol. After elution from the Ni-NTA resin, His₆-MBP-PreSc-Ctc1/Stn1/Ten1 was loaded onto a HiTrap Heparin HP column (Cytiva) equilibrated with a buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.5 mM TCEP, and 5% glycerol and eluted with a linear gradient of NaCl concentration to 1 M. Protein-containing fractions were loaded onto a HiLoad Superdex 200 16/600 PG column (Cytiva) equilibrated with a buffer containing 20 mM HEPES pH 7.5, 300 mM NaCl, 0.5 mM TCEP, and 5% glycerol. The His₆-MBP tag of Ctc1 was only cleaved off for the negative-stain EM analysis of the CST–POT1(ESDL)/TPP1–ssDNA complex (Fig. S2E) but left uncleaved for all other experiments to issues with aggregation and poor vitrification. For tag cleavage, the eluate from the Ni-NTA resin was incubated with rhinovirus 3C protease overnight at 4°C to remove the His₆-MBP tag prior to Heparin-affinity and size-exclusion chromatography as described above.

For the POT1/TPP1 proteins, frozen pellets were lysed by cryogenic milling (Retsch) and the cryo-milled powder was resuspended in a buffer containing 50 mM HEPES pH 7.5, 350 mM NaCl, 5 mM β-ME, 5% glycerol, 0.05% (v/v) Tween-20, and 1 mM PMSF, supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche). The lysate was cleared by centrifugation at 40,000 rcf for 1 h at 4°C. Supernatants were incubated with end-over-end rotation for 1 h at 4°C with Strep-Tactin Superflow high-capacity resin (IBA Lifesciences) equilibrated with gel-filtration buffer (20 mM HEPES pH 7.5, 300 mM NaCl, 0.1 mM TCEP, 0.05% Tween-20, and 5% glycerol) and subsequently washed with 20–50 CV of the same buffer. Bound protein was eluted in the same buffer supplemented with 10 mM d-desthiobiotin, concentrated, and loaded onto a HiLoad Superdex 200 16/600 PG column (Cytiva) equilibrated with the same gel-filtration buffer.

The fusion CST–POT1/TPP1 complex was purified similarly, but after elution from the Strep-Tactin resin, it was concentrated to 500 μL and loaded on an 11-mL linear 10–30% glycerol gradient. Ultracentrifugation was carried out at 41,000 rpm in an SW 41 Ti rotor for 18 h at 4°C. 500 μL fractions were manually collected and protein-containing fractions were identified by SDS-PAGE.

Protein-containing fractions were concentrated, flash frozen in LN₂, and stored as aliquots at −80°C. Protein concentrations were measured on a NanoDrop-1000 spectrophotometer.

Fluorescence-detection size-exclusion chromatography analysis

Protein–protein interaction experiments were performed on a Superose 6 Increase gel-filtration column (Cytiva) equilibrated in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, and 0.1 mM TCEP. Eluate from the column was passed through an RF-20A fluorescence detector (Shimadzu) operated with excitation and emission wavelengths of 280 nm and 340 nm, respectively. Purified proteins were mixed to a final volume of 120 μL and incubated on ice for at least 30 mutes prior to injection of 100 μL on the column. GFP-tagged POT1/TPP1 complexes were added at the specified concentrations, and [GGTTAG]₃ ssDNA (when present) and CST were added at a twofold molar excess. Complex formation was evaluated by comparing the mobility on the gel-filtration column of POT1/TPP1 alone versus when mixed with CST.

For the dephosphorylation experiments, 10 μM POT1/TPP1 was incubated with 10 μM λ protein phosphatase (λPP, expressed and purified in-house from E. coli) and 1 mM MnCl₂ in the FSEC gel-filtration buffer. For mock treated samples, the λPP was omitted and replaced with buffer. Samples were incubated for 8–16 h at 4°C. An aliquot from each sample was taken for SDS-PAGE analysis before proteins were diluted for the FSEC experiments described above.

Reconstitution of the ssDNA–CST–POT1/TPP1 complex

Purified His₆-MBP-Ctc1/Stn1/Ten1 and His₆-POT1(ESDL)/TwinStrep-GFP-TPP1 were mixed with [GGTTAG]₃ ssDNA in 150 μL at final concentrations of 6 μM, 6 μM, and 9 μM, respectively, in a buffer containing 20 mM HEPES pH 7.5, 0.5 mM TCEP pH 7.5, and 1% glycerol. Based on the volume of the input proteins (which were purified at 300 mM NaCl), the NaCl was adjusted to a final concentration of 150 mM. The protein components were mixed first and incubated on ice for 1 h prior to addition of the ssDNA. The protein and ssDNA were incubated on ice for 2 h prior to loading on an 11-mL linear 10–30% glycerol gradient. Ultracentrifugation was carried out at 41,000 rpm in an SW 41 Ti rotor for 18 h at 4°C. 500 μL fractions were manually collected and protein- and DNA-containing fractions were identified by SDS-PAGE and native PAGE.

Negative-stain EM sample preparation, data collection, and image processing

Protein samples for negative-stain EM (3.5 μL drops, in a concentration range of 0.01–0.05 mg/mL) were adsorbed to glow-discharged carbon-coated copper grids with a collodion film, washed with three drops of deionized water, and stained with two drops of freshly prepared 0.7% (w/v) uranyl formate. Samples were imaged at room temperature using a Phillips CM10 electron microscope equipped with a tungsten filament and operated at an acceleration voltage of 80 kV. The magnification used for the CST-only samples corresponds to a calibrated pixel size of 3.5 Å. The magnification used for the CST–POT1(ESDL)/TPP1 samples corresponds to a calibrated pixel size of 2.6 Å. Particles were auto-picked using the Swarm picker (CST-only) or Gauss picker (CST–POT1(ESDL)/TPP1) in EMAN2⁶². Particle extraction and 2D classification were performed in RELION-3.1⁶³.

Cryo-EM sample preparation and data collection

Apo CST–POT1(ESDL)/TPP1 complex was frozen at a concentration of 0.075 mg/mL, corresponding to 0.2 μM of GFP-tagged protein, in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, and 0.1 mM TCEP. The CST–POT1(ESDL)/TPP1–ssDNA complex was prone to aggregation following vitrification at the concentration used for the apo complex and faced challenges with inconsistent ice thickness across the grid. Thus, the CST–POT1(ESDL)/TPP1–ssDNA complex was diluted to 0.02 mg/mL, or 0.05 μM of GFP-tagged protein. The addition of 0.75 mM fluorinated Fos-choline-8 (Anatrace) resulted in even ice thickness.

3.5 μL of the samples were applied to Quantifoil R1.2/1.3 mesh Cu400 holey carbon grids covered with a graphene oxide support layer (EMS) that were glow-discharged for 5 s at 40 mA in an EMS100X glow discharge unit (EMS) and then blotted for 0.5–1 s (Blot Force −2; Wait Time 20 s; Drain Time 0 s) and plunge frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific) operated at 4°C and 100% humidity. Cryo-EM imaging was performed in the Evelyn Gruss-Lipper Cryo-EM Resource Center at The Rockefeller University using SerialEM⁶⁴. Data collection parameters are summarized in Table S1.

For both complexes, data were collected on a 300-kV Titan Krios electron microscope equipped with a Cs corrector at a nominal magnification of ×105,000, corresponding to a calibrated pixel size of 0.86 Å (image dimensions of 5,760×4,092 px) on the specimen level. Images were collected using a slit width of 20 eV on the GIF Quantum energy filter (Gatan) and a defocus range from −1 to −2.5 μm with a K3 direct electron detector (Gatan) in super-resolution counting mode. Three movie stacks were recorded per hole in a 3×3 matrix of holes acquired using beam-image shift with a maximum shift of 3.5 μm. Exposures of 1.2 s were dose-fractionated into 40 frames (30 ms per frame) with an exposure rate of 31 electrons/pixel/s (approximately 1.25 electrons per Ų per frame), resulting in a total electron exposure of 50.3 electrons per Ų.

Cryo-EM data processing

For all datasets, movie stacks were motion-corrected with the RELION-3.1⁶³ implementation of MotionCor2⁶⁵, and motion-corrected micrographs were manually inspected and curated (graphene oxide coverage of grids was inconsistent) prior to CTF parameter estimation with CTFFIND-4⁶⁶ implemented in RELION-3.1.

To generate templates and an initial model for the apo complex, particles were automatically picked without a template using Gautomatch (v0.53) and extracted (352 px box, binned 4-fold to 88 px) in RELION-3.1. After 2D classification and cleanup to remove contaminating ice particles, the cleaned particle stack was imported into cryoSPARC (v4.1.1) and 2D classification was performed. The best 2D classes (containing ~150,000 particles) were used to generate an initial model that was used as a 3D reference. The motion-corrected micrographs were then imported into cryoSPARC⁶⁷, and CTF parameter estimation was performed with CTFFIND-4⁶⁶ implemented in cryoSPARC. Particles were automatically picked using the Template Picker with the initial model as a 3D reference. Particles were extracted, Fourier-cropped from a 352 px box to a 256 px box to speed up image processing and processed using the described image-processing pipeline depicted in Fig. S3. Three rounds of heterogeneous refinement were performed with one good reference and three noise “decoy” references to filter out junk particles. After the number of particles in the junk classes dropped to 15% or below, the good particle stack was further classified using heterogeneous refinement into four classes using a single reference. The 372,930 particles in the best class were re-extracted at full-size (352 px box). There was additional heterogeneity, particularly in the region of Stn1^C and POT1^OB-1/2, so further unsupervised heterogeneous refinement was used to identify particles with fully intact Stn1^C and POT1^OB-1/2. The final stack of 132,356 particles was refined and sharpened to a nominal resolution of 3.6 Å using the Non-uniform Refinement job in cryoSPARC and the local resolution of the map was estimated in cryoSPARC. The reported 3.9-Å resolution was calculated with Phenix (phenix.validation_cryoem) using the gold-standard Fourier shell correlation (FSC) between half-maps (Fig. S3I). FSC plots and sphericity values⁵⁸ of the reconstruction were calculated using the 3D-FSC server (https://3dfsc.salk.edu/).

For the ssDNA-bound complex, the motion-corrected micrographs were imported into cryoSPARC, and CTF parameter estimation was performed with the Patch CTF Estimation job. The apo complex map was used as a 3D template for particle picking with the Template Picker job. Particles were extracted, Fourier-cropped from a 352 px box to a 128 px box to speed up image processing and processed using the described image-processing pipeline depicted in Fig. S4. The 2D-class averages of the extracted particles and negative-stain EM analysis of the complex (Fig. 4) revealed heterogeneity. This heterogeneity could be divided into three categories of particles: fully intact CST–POT1/TPP1 complexes that resembled the apo complex and contained bound Stn1^C and POT1^OB-1/2, half complexes in which Stn1^C and POT1^OB-1/2 were disengaged but the POT1 C-terminus was bound, and particles containing CST without POT1/TPP1. Thus, three 3D references were generated from the apo complex by segmenting the map in UCSF Chimera⁶⁸. The references are shown in Fig. S4. Two noise decoy references were used in addition to the three real references to perform iterative heterogeneous refinement until the noise classes contained fewer than 10% of the particles. An additional heterogeneous refinement with the three different references was performed prior to re-extracting the particles with Stn1^C and POT1^OB-1/2 bound at full size. There was still some heterogeneity in that region, so the map was segmented to generate a mask around the Ctc1 C-terminus, Stn1, Ten1, and POT1^OB-1/2. This mask was used to perform focused 3D classification into four classes without alignment in cryoSPARC. Two classes contained density for both Stn1^C and POT1^OB-1/2 and were pooled into a final stack of 76,359 particles. The other two classes were missing density for one of the components. The final particle stack was refined and sharpened to a nominal resolution of 4.0 Å using the Non-uniform Refinement job in cryoSPARC and the local resolution of the map was estimated in cryoSPARC. The reported 4.3-Å resolution was calculated with Phenix (phenix.validation_cryoem) using the gold-standard FSC between half-maps (Fig. S4I). FSC plots and sphericity values⁵⁸ of the reconstruction were calculated using the 3D-FSC server (https://3dfsc.salk.edu/).

We also attempted to continue processing the complex in which ssDNA was presumably bound to CST where only the C-terminus of POT1 is bound to CST (Fig. S4D). Focused 3D classification revealed that there were few particles truly missing the Stn1^C and POT1^OB-1/2 as cartooned in Fig. 5B, but the association was weak. Neither heterogeneous refinement nor focused 3D classification without alignment were able to further classify the particles or align to higher-resolution features. This is likely due to the intrinsic flexibility of CST, for which a high-resolution structure of a monomeric complex without other interacting factors could not be determined¹³. It appears that when bound, the Stn1^C and POT1^OB-1/2 interactions stabilized CST to allow for high-resolution structure determination.

Model building and refinement

The map of the apo complex was used for atomic model building, as it had a higher overall resolution and better quality. The AlphaFold2^{36, 37} database model of Ctc1 (AF-Q2NJK3) was merged with the cryo-EM structure of CST (PDB 6W6W¹³) as the starting model for CST. The X-ray crystal structures of POT1^OB-3/HJRL/TPP1^RD (PDB 5H65³⁸) and the POT1 OB-folds bound to ssDNA (PDB 1XJV⁴⁰) were used as starting models for the POT1/TPP1 components. CST, POT1^OB-2, and POT1^OB-3/HJRL/TPP1^RD were first rigid-body docked into the map using UCSF Chimera. The model was then manually inspected in Coot⁶⁹, where additional density of the hinge region of POT1 was observed. There was clear density for aa 322–347 that allowed unambiguous manual modeling of the hinge self-interactions with POT1^OB-3. The N-terminal region of the hinge (aa 299–321 and the ESDL insertion) showed connected Cα backbone density, but many side-chain placements were ambiguous. Because this region of the hinge makes a weak interaction that appears to be mediated primarily by charge complementarity rather than specific interactions, the exact side-chain positions were not critical to the analysis. Thus, we first modeled the hinge Cα backbone into the density and observed that it spanned the distance between the preceding and following regions of the chain. Then, we assigned the sequence in register with the rest of the protein. The key reason for modeling the side-chains for residues 308–321 was to show the abundance of serine and aspartate residues in this region. Residues 301–307 were modeled as poly-alanine stubs, as this part of the chain runs in a groove of Ctc1 and we did not want to over- or mis-interpret specific interactions. The resolution of POT1^OB-2 was not high, so POT1^OB-1 was only flexibly fitted into the density using ISOLDE⁷⁰ after rigid-body docking. For Ctc1, the AlphaFold2 model fit well overall into the density and required only few manual adjustments. One region of Ctc1 (aa 909–927) appeared to contact POT1 weakly. This region of Ctc1 has not been resolved in previous experimental structures and AlphaFold2 predicts this loop with low confidence. There was clear density near the predicted position of the loop, so it was approximated with poly-alanine stubs. The map regions containing Stn1 and Ten1 were also at lower resolution, so they were docked in and fitted similarly to POT1^OB-2. Iterative real-space refinement in Phenix (phenix.real_space_refine) and manual model adjustment in ISOLDE was used to correct the most egregious geometry errors, though some errors could not be fixed in a way supported by the experimental map. The model and maps were validated in Phenix (phenix.validation_cryoem)^{71, 72}.

For the ssDNA-bound complex, the apo complex model was used as the starting model. POT1^OB-2 from the apo complex was replaced with a rigid-body docking of POT1^OB-1/2 and a 5’-TTAGGGTTAG-3’ DNA from the X-ray crystal structure (PDB 1XJV⁴⁰). Because the overall resolution of the map of the ssDNA-bound complex was lower, but the overall interaction and map density was similar between the apo and ssDNA-bound complex, we treated the apo model as relatively correct. The apo model was fitted to the ssDNA-bound complex map using iterative real-space refinement in Phenix (phenix.real_space_refine) and some manual model adjustment in ISOLDE. The resolution of POT1^OB-1/2 was low, so it was modeled by flexible fitting of the crystal structure and some attempts were made to correct its geometry in ISOLDE. ssDNA was clearly present, but the resolution was too low to assign the register or see if more than 10 nt were bound. Thus, only the 5’-TTAGGGTTAG-3’ from the crystal structure was retained in the final model. The model and maps were validated in Phenix (phenix.validation_cryoem).

For PDB deposition, the POT1 ESDL insertion was notated using insertion code as an insertion after residue 320 (E – 320A; S – 320B; D – 320C; L – 320D) to maintain the canonical numbering of the human POT1 sequence. The TPP1–Stn1 fusion in the apo complex required deposition as a single chain, so the numbering corresponds to the fusion. However, individual mappings to TPP1 and Stn1 within the chain are annotated in the deposition.

Multi-body refinement analysis of CST–Polα/primase RC flexibility

The dataset used for RC analysis was the same as previously published¹⁵ (EMPIAR-11131) and reprocessed in RELION-3.1 with the same particle coordinates and particles extracted. The final stack used for multi-body refinement contained 203,045 particles (300 px box; 1.08 Å/px). Masks for Polα/primase (body 1) and CST (body 2) were generated by segmentation of the RC map in UCSF Chimera and mask creation in RELION-3.1. Multi-body refinement was performed with default parameters⁴⁵. Angular/translational priors of 10°/2px and 20°/5px were used for Polα/primase and CST, respectively. Volume series movies were generated in UCSF ChimeraX⁷³.

AlphaFold-Multimer complex structure prediction

AlphaFold-Multimer (v1.0)³² was used on the COSMIC2 server⁷⁴ with default settings (full_dbs). For the Ctc1^OB-A/B/C–TPP1^TID complex, the input sequence file contained Ctc1 aa 1–410 and TPP1 aa 361–458. For the Ctc1–POLA2 complex, the input sequence file contained Ctc1 aa 1–1217 and POLA2 aa 1–158. Flexible regions with low predicted confidence were not shown in the figures.

Mass photometry

Purified POT1/TPP1 (5 μM) and CST (7.5 μM) proteins were mixed with 7.5 μM [GGTTAG]₃ in MP buffer (20 mM HEPES pH 7.5, 150 mM NaCl, and 0.1 mM TCEP pH 7.5, filtered 3x through 0.22 μm syringe filter) and incubated for 1 h at 4°C. Where indicated, λPP treatment of POT1/TPP1 proteins was performed as described above prior to incubation with CST. Immediately prior to measurement, samples were diluted to 100 nM POT1/TPP1 in MP buffer, then 2 μL was added to 8 μL MP buffer in a sealed well on a glass slide for a final concentration of 20 nM POT1/TPP1 and 1.5-fold excess CST and ssDNA. Data were collected using a OneMP mass photometer (Refeyn) calibrated with bovine serum albumin (66 kDa), beta amylase (224 kDa), and thyroglobulin (660 kDa). Movies were acquired for 6,000 frames (60 s) using AcquireMP software (v2022R1) and default settings. Final protein concentrations were empirically determined to achieve ~50 binding events per second. Raw event measurements were converted to frequency distributions using DiscoverMP software (v2022R1) or Prism 9 (GraphPad) and a bin size of 10 kDa. Gaussian curves were autofitted with DiscoverMP to provide total peak counts for quantification.

Double-filter DNA-binding assays

A modified double-filter DNA-binding assay was used to calculate apparent binding constants (K_D,app)⁴⁴. Binding buffer contained 50 mM HEPES-KOH pH 7.5, 150 mM KCl, 0.5 mM TCEP, 0.1 mg/mL bovine serum albumin, and 15% glycerol. The [GGTTAG]₃ ssDNA oligonucleotide (Thermo Fisher) was 5’-end-labeled using [γ-³²P]ATP with T4 polynucleotide kinase (NEB). Free nucleotides were removed using Micro Bio-Spin 6 columns (BioRad) and the labeled oligonucleotide was diluted in binding buffer to achieve a final concentration of <10 pM in the experiment. The maximum concentration of 10 pM was calculated with the assumption of no product loss on the spin column. Experiments were performed in a 96-well vacuum manifold dot blot apparatus (GE Healthcare) with nitrocellulose, Hybond-XL filters (charged nylon; GE Healthcare), and 3MM chromatography filter paper (GE Healthcare). All filters were presoaked in 50 mM HEPES pH 7.5 prior to assembly in the apparatus. Proteins were serially diluted to reach the concentrations indicated, mixed with the final <10 pM labeled oligonucleotide in a final volume of 220 μL, and incubated in a 96-well plate on a cold block in ice for 1 h. 100 μL binding buffer was applied to the apparatus to wash the membranes and briefly vacuumed without allowing the membrane to completely dry. The vacuum line was removed prior to loading 100 μL protein-ssDNA mix per well with a multi-channel pipette. Two technical replicates to control for pipetting error were performed per experiment by loading two wells per condition from the same protein-ssDNA mix. After incubation for 1 m, vacuum was applied and the wells were washed three times with 100 μL binding buffer. Filters were air dried and wrapped in plastic wrap before exposing to phosphor imaging screens for 10 days (GE Healthcare). Screens were scanned on an Amersham Typhoon imager and data were quantified using ImageJ⁷⁵ and GraphPad Prism 9.

C-Strand synthesis assay

CST–Polα/primase was expressed in HEK293T/17 cells (ATCC, CRL-11268, LOT: 63696280) transfected with plasmids encoding the three subunits of CST and purified by sequential pull-down with antibodies to the FLAG tag on Ctc1 and the HA tag on Ten1, as described⁴⁷. The reported protein concentrations refer to CST; the endogenous Polα/primase, which co-immunopurifies with the CST, is present at about 20% stoichiometry. DNA templates and chimeric RNA-DNA primers were purchased from IDT. C-strand synthesis reactions proceeded for 60 m at 30°C in 50 mM Tris-HCl pH 8.0, 1 mM MgCl₂, 1 mM spermidine, 5 mM β-ME, 50 mM KCl, 50 mM NaCl, 0.2 mM each ATP, CTP and UTP, 0.5 mM each dCTP and TTP, 0.01 mM dATP, and 0.17 mM ³²P-dATP (3,000 Ci/mmole, Revvity, formerly Perkin Elmer). Products were separated by electrophoresis on a 10% acrylamide/7 M urea/1x TBE gel. ³²P-labeled loading controls L1 (18 nt), L2 (16 nt) and L3 (15 nt) were added with the stop buffer (3.6 M NH₄Ac, 0.2 mg/mL glycogen). Gels were dried and scanned with a phosphorimager (Cytiva), signal intensities were determined (ImageQuant), normalized to the signal intensity in the absence of added POT1/TPP1, and data fit to a sigmoidal binding curve (KaleidaGraph). Reactions with pre-primed templates were performed similarly, with equal concentrations of the template and primer first pre-incubated for 10 m at room temperature and then stored at −20°C. Prior to the reaction the pre-primed templates were thawed at room temperature and added to the same reaction mixture as above except with no ribonucleotides. Products from the pre-primed substrates were analyzed as described above for C-strand synthesis.

Structure analysis and visualization

The apo CST–POT1(ESDL)/TPP1 complex was used for all interaction analysis, as the model and map quality was superior to that of the ssDNA-bound complex. The additional POT1^OB-1 and telomeric DNA in the ssDNA-bound complex did not interact with CST. UCSF Chimera, UCSF ChimeraX, and PyMOL (Schrödinger) were used to analyze the maps and models. Figures were prepared using UCSF ChimeraX and Adobe Illustrator. Maps in ChimeraX were prepared with the “hide dust” limit set to 10 and the contour thresholds as specified in the figure legends.

QUANTIFICATION AND STATISTICAL ANALYSIS

Sample sizes for the cryo-EM datasets were determined by the need to obtain meaningful structures and the availability of cryo-EM time. For the apo CST-POT1/TPP1 dataset, 15,201 movie stacks were collected, and 132,356 particles were used for the final reconstruction, which was sufficient to yield a 3.9-Å resolution structure. For the ssDNA-CST-POT1/TPP1 dataset, 31,914 movie stacks were collected, and 76,359 particles were used for the final reconstruction, which was sufficient to yield a 4.3-Å resolution structure. Micrographs clearly suffering from astigmatism, image drift, poor graphene oxide coverage, ice contamination, and/or cubic ice formation were excluded during the micrograph curation step in all cryo-EM datasets analyzed. Particles in 2D classes showing no secondary structural features and in 3D classes showing unsatisfactory structural features were excluded from the final reconstructions in all datasets analyzed.

For the high-resolution structures obtained in this study, small negative-stain EM datasets were first collected on an 80-kV Phillips CM10 electron microscope. Cryo-EM data collection on a 300-kV Titan Krios electron microscope resulted in structures that matched the negative-stain EM 2D averages obtained. Each biochemical reconstitution was performed at least three times and was reproducible.

All other statistical analyses and technical replicates are described in the methods and figure legends for the corresponding experiments.

Supplementary Material

1. Figure S1. Analysis of the CST–TPP1 interaction, related to Figure 1.

(A) Co-IPs of Myc-tagged TPP1 proteins with the indicated C-terminal truncations with FLAG-tagged Ctc1, Stn1, and Ten1 from co-transfected 293T cells showing that the C-terminus of TPP1 is required for the CST–TPP1 interaction. Western blots of Myc IPs were probed with anti-Myc and anti-FLAG antibodies.

(B) Co-IPs of Myc-tagged TPP1 with HA-tagged TIN2 and FLAG-tagged Ctc1, Stn1, and Ten1 from co-transfected 293T cells showing that TIN2 competes with CST for TPP1 interaction. Western blots of Myc IPs were probed with anti-Myc, anti-FLAG, and anti-HA antibodies.

(C) Predicted aligned error (PAE) plots of the top four (of 5) ranked AlphaFold-Multimer³² models for Ctc1–TPP1. Green arrowheads indicate high confidence in the position prediction of TPP1 relative to Ctc1. All 5 top-ranked models predicted the interaction with high confidence.

(D) (Left) AlphaFold-Multimer model of Ctc1 OB-folds A, B, and C bound to TPP1^TID compared to the crystal structure of TRF2/TIN2/TPP1^TID (PDB 5XYF³¹). TPP1^TID is predicted to bind to Ctc1 and TIN2 using the same peptide, providing a structural basis for their competition. (Right) Domain schematics of Ctc1, TPP1, and TIN2 with gray shadows showing the overlapping interaction site on TPP1^TID. OB: oligosaccharide/oligonucleotide-binding domain; ARODL: Acidic Rpa1 OB-binding Domain-Like 3-helix bundle; RD: recruitment domain, TID: TIN2-interacting domain; TRFH: telomere repeat factor homology domain.

2. Figure S2. Reconstitution of CST–POT1(ESDL)/TPP1 complexes, related to Figure 1.

(A) FSEC analysis of the binding between increasing concentrations of CST and POT1(ESDL)/TPP1 in the absence of telomeric ssDNA showing the concentration dependence of the interaction. RFU: relative fluorescence units.

(B) Native glycerol gradient analysis of ssDNA-bound CST–POT1(ESDL)/TPP1. Pooled fractions are indicated with an asterisk. (Top) Coomassie-stained SDS-PAGE gel (4–12% Bis-Tris gel run in MOPS-SDS buffer, Invitrogen) of glycerol gradient fractions. (Bottom) SYBR Gold-stained native PAGE (4–20% TBE gel run in 0.5x TB buffer, Invitrogen).

(C) Domain architecture of components used to reconstitute an apo CST–POT1(ESDL)/TPP1 complex. TPP1 is fused to the N-terminus of Stn1 and retains part of the TPP1 serine-rich linker.

(D) Coomassie-stained SDS-PAGE gel (4–12% Bis-Tris gel run in MOPS-SDS buffer, Invitrogen) showing the purified CST–POT1(ESDL)/TPP1 fusion complex used for structural analysis.

(E) Negative-stain EM analysis of CST and CST–POT1(ESDL)/TPP1 complexes. (Left) Representative negative-stain EM micrographs. (Middle) Top 25 reference-free 2D-class averages (sorted by number of particles per class from most populous class at top left to least populous class at bottom right) of each complex. (Right) Enlarged views of selected 2D averages to show CST features. Additional density attributable to the addition of POT1(ESDL)/TPP1 is indicated with red arrowheads.

3. Figure S3. Cryo-EM image-processing pipeline for apo CST–POT1(ESDL)/TPP1 complex, related to Figure 2.

(A) Representative motion-corrected micrograph.

(B) Enlarged view of the area marked in (A) with selected particles circled.

(C) Representative 2D-class averages show high-resolution features and different orientations.

(D) Cryo-EM image-processing pipeline used for the apo CST–POT1(ESDL)/TPP1 complex, including supervised 3D classification with noise decoy classes. Several rounds of heterogeneous refinement were used to select for particles with well-resolved Stn1^C and POT1^OB1/2 (Table S1).

(E) Final map (contour threshold 0.2) of the apo CST–POT1(ESDL)/TPP1 complex with POT1(ESDL)/TPP1 colored as in Fig. 2A for reference.

(F) Directional FSC plots and sphericity values⁵⁸ of the reconstruction calculated using the 3D-FSC server (https://3dfsc.salk.edu/).

(G) Plot of the angular distribution of particles in the final reconstruction.

(H) Local resolution estimates for the map (contour threshold 0.2) of apo CST–POT1(ESDL)/TPP1.

(I) Gold-standard (blue) and model-vs-map (red) FSC curves for the reconstruction of apo CST–POT1(ESDL)/TPP1. The map resolution was estimated using the gold-standard FSC and a cut-off criterion of 0.143. The similarity between the resolution based on the gold-standard FSC (0.143 cut-off) and the resolution estimate based on the model-vs-map FSC (0.5 cut-off) suggests no substantial over-fitting.

(J) Cryo-EM map densities for each subunit indicating quality of fit for the apo CST–POT1(ESDL)/TPP1 model. Contour thresholds for Ctc1, Stn1, Ten1, POT1, and TPP1 are 0.1, 0.065, 0.15, 0.1, and 0.05, respectively.

4. Figure S4. Cryo-EM image-processing pipeline for ssDNA-bound CST–POT1(ESDL)/TPP1 complex, related to Figure 2.

(A) Representative motion-corrected micrograph.

(B) Enlarged view of the area marked in (A) with selected particles circled.

(C) Representative 2D-class averages show high-resolution features and different orientations.

(D) Cryo-EM image-processing pipeline used for the ssDNA-bound CST–POT1(ESDL)/TPP1 complex, including supervised 3D classification with noise decoy classes. Focused 3D classification with a mask was used to select for particles with well-resolved Stn1^C and POT1^OB1/2 (Table S1).

(E) Final map (contour threshold 0.15) of the ssDNA-bound CST–POT1(ESDL)/TPP1 complex with ssDNA–POT1(ESDL)/TPP1 colored as in Fig. 2A for reference.

(F) Directional FSC plots and sphericity values⁵⁸ of the reconstruction calculated using the 3D-FSC server (https://3dfsc.salk.edu/).

(G) Plot of the angular distribution of particles in the final reconstruction.

(H) Local resolution estimates for the map (contour threshold 0.15) of the ssDNA-bound CST–POT1(ESDL)/TPP1.

(I) Gold-standard (blue) and model-vs-map (red) FSC curves for the reconstruction of CST–POT1(ESDL)/TPP1. The map resolution was estimated using the gold-standard FSC and a cut-off criterion of 0.143. The similarity between the resolution estimate based on the gold-standard FSC (0.143 cut-off) and the resolution estimate based on the model-vs-map FSC (0.5 cut-off) suggests no substantial over-fitting.

(J) Cryo-EM map densities for each subunit indicating quality of fit for the model of ssDNA-bound CST–POT1(ESDL)/TPP1. Contour thresholds for Ctc1, Stn1, Ten1, POT1, and TPP1 are 0.075, 0.14, 0.1, 0.1, and 0.1, respectively.

5. Figure S5. Analysis of the TPP1 OB-fold and POT1 CCIR, related to Figure 3.

(A) Deletion of the TPP1 OB-fold does not affect the CST–POT1(ESDL)/TPP1 interaction. Protein used and FSEC analysis of CST–POT1(ESDL)/TPP1(ΔOB) interaction in the absence (top) and presence (bottom) of telomeric ssDNA. Traces without CST are shown as dashed lines. RFU: relative fluorescence units.

(B) Electrostatic surface from Fig. 3A with map density (contour threshold 0.1) of hinge shown as mesh.

(C) Coomassie-stained SDS-PAGE gel (4–12% Bis-Tris gel run in MOPS-SDS buffer, Invitrogen) showing the purified POT1/TPP1 proteins (untreated or λ-treated) used for FSEC and MP analysis. Descriptions of the POT1 variant sequences are given in Fig. 3E.

(D) FSEC analysis of the phosphorylation-dependent interaction between CST and POT1(WT)/GFP-TPP1 (top, black) and POT1(ESDL)/GFP-TPP1 (bottom, blue) in the absence of telomeric ssDNA (see also Fig. 3B). Traces without CST are shown as dashed lines, and traces containing POT1/TPP1 that have been treated with λ protein phosphatase (λPP) are shown in red. RFU: relative fluorescence units. The chromatograms for the WT and ESDL mock samples are shown in Fig. 1E and are reproduced here as the controls for the phosphatase experiment, which was performed simultaneously.

(E) MP histograms used for the quantification shown in Fig. 3F. Each panel shows histograms from three technical replicates with Gaussian curves auto-fitted by DiscoverMP software. Inset tables show σ (standard deviation of values within peak) and and total counts per peak for each replicate. Total peak counts were used to quantify the proportion of bound POT1/TPP1, the ratio of total counts in peak 3 (CST–POT1/TPP1 complex) divided by total counts of POT1/TPP1 (peak 1 and peak 3).

6. Figure S6. POT1^OB-3–Ctc1 interaction analysis and ssDNA binding, related to Figure 4.

(A) As in Fig. 4B but individual panels include the cryo-EM map density for the apo CST–POT1(ESDL)/TPP1 complex shown as mesh (contour thresholds 0.3, 0.3, 0.7, 0.5, and 0.5, respectively).

(B) Co-IPs of Strep-tagged POT1 (or mPOT1b) protein variants and FLAG-tagged Ctc1, Stn1, and Ten1 from co-transfected 293T cells showing that the corresponding S328L CP mutation in mPOT1b disrupts the CST interaction. Western blots of Strep IPs were probed with anti-Strep and anti-FLAG antibodies.

(C) Double-filter DNA-binding assays showing high affinity [GGTTAG]₃ binding by POT1/TPP1 alone and in complex with CST. Left: Representative membrane scans. The nitrocellulose membrane is labeled as “Bound” and the Hybond-XL membrane is labeled as “Free.” Right: Quantitation of binding data. Data were fitted using the “one site – specific binding” model on GraphPad Prism 9. Error bars represent SEM from two technical replicate experiments.

7. Figure S7. Comparison of the CST–POT1(ESDL)/TPP1 structure with those of CST–Polα/primase, related to Figure 5.

(A) Multiple views of CST–Polα/primase in RC (left) and PIC (right) conformations. CST–Polα/primase structures are shown in cartoon representation with a transparent surface.

(B) Superposition of the ssDNA-bound CST–POT1(ESDL)/TPP1 structure with the structures of the RC (left, see also Video S3) and PIC (right, see also Video S2) complexes showing additional views compared to Fig. 5C. The clash between the POT1^HJRL and PRIM2 is indicated.

(C) Multi-body analysis of CST–Polα/primase in the RC conformation. Polα/primase was designated body 1 and CST was designated body 2 with the corresponding masks shown. Histograms of the projections of the relative orientations onto the corresponding components show a unimodal distribution, consistent with continuous flexibility rather than discrete states. The first three principal components account for 61% of the variance in the data. Reconstructed maps from the extreme ends are shown in red and blue for each of the first three principal components with arrows indicating the direction of motion (see also Video S4).

(D) PAE plots of the top two (of 5) ranked AlphaFold-Multimer³² models for Ctc1/POLA2. Green arrowheads indicate high confidence in the position prediction of POLA2 relative to Ctc1. All 5 top-ranked models predicted the interaction with high confidence.

(E) AlphaFold-Multimer model of Ctc1 bound to POLA2^NTD. Residues with known CP mutations are shown as spheres. Red spheres indicate mutations previously shown to disrupt Polα/primase association⁴¹; black spheres indicate mutations with unknown mechanism.

8. Figure S8. CST-binding enhances POT1/TPP1 inhibition of telomeric C-strand synthesis, related to Figure 6.

(A) Inhibition of C-strand synthesis by POT1(ESDL)/TPP1 is enhanced by its DNA-binding domain. Reactions of 50 nM CST–Polα/primase with 50 nM single-stranded 9xTEL DNA template. LC1, LC2, LC3: oligonucleotide loading controls.

(B) Quantification of the initial extension products, bands 21–23.

(C) Quantification of all products, bands 21–40, gave similar IC₅₀ values as in (B).

(D) Design of pre-primed templates to monitor primer extension by Polα. Red sequence is RNA.

(E) Validation of the pre-primed template reactions by performing reactions in the absence of ribonucleotides. Most reaction products were dependent on both template and primer. The 9xTEL template products (which were formed in the absence of the primer strand) were longer than the template, indicating that they resulted from self-priming of the DNA template.

(F) CST–Polα extension of pre-primed templates. M: telomerase reaction products as size markers. Stronger inhibition of C-strand synthesis by POT1(ESDL)/TPP1 persists under conditions that limit re-priming (50 nM template and 5 nM CST–Polα/primase, a ten-fold higher template/enzyme ratio than used in Fig. 6E).

(G) Quantification of the data in (F) (left), normalized to reactions without added POT1/TPP1. IC₅₀ values with error bars indicating uncertainty of the fit to the data points.

(H) Quantification of the data in (F) (right), normalized to reactions without added POT1/TPP1. IC₅₀ values were not calculated because inhibition was incomplete at highest POT1/TPP1 concentrations.

9. Video S1. Cryo-EM map and model of CST–POT1(ESDL)/TPP1 complexes, related to Figure 2.

360° rotation of the cryo-EM map and model from apo and ssDNA-bound CST–POT1(ESDL)/TPP1. Colors are the same as in Fig. 2.

11. Video S2. Superposition of ssDNA-bound CST–POT1(ESDL)/TPP1 with CST–Polα/primase pre-initiation complex, related to Figure 5.

Rotation movie showing superposition of CST–POT1(ESDL)/TPP1–ssDNA complex (surface representation; solid colors as in Fig. 2) with CST–Polα/primase PIC (surface representation at 50% opacity; cartoon model shown underneath). Colors of Polα/primase are the same as in¹⁵. PRIM1-salmon; PRIM2-light orange; POLΑ1-light green (Ctc1-recognition loop-lime green); POLA2-yellow.

12. Video S3. Superposition of ssDNA-bound CST–POT1(ESDL)/TPP1 with CST–Polα/primase recruitment complex, related to Figure 5.

Rotation movie showing superposition of CST–POT1(ESDL)/TPP1–ssDNA complex (surface representation; solid colors as in Fig. 2) with CST–Polα/primase RC (surface representation at 50% opacity; cartoon model shown underneath). Colors of Polα/primase are the same as in¹⁵. PRIM1-salmon; PRIM2-light orange; POLΑ1-light green (Ctc1-recognition loop-lime green); POLA2-yellow.

13. Video S4. Multi-body refinement movies of the first three principal motions contributing to CST-Polα/primase RC flexibility, related to Figure 5 and Figure S7.

Principal components 1, 2, and 3, respectively.

Highlights.

• Cryo-EM structures reveal how the shelterin subunit POT1 binds to Ctc1 in CST

• The structures reveal the mechanism of Coats plus syndrome mutations in Ctc1 and POT1

• CST recruitment to shelterin requires phosphorylation of POT1

• Activation of CST–Polα/primase for C-strand synthesis involves its release from POT1