The E3 ubiquitin ligase RNF8 stabilizes TPP1 to promote telomere end protection

RNF8 mediates the DNA damage response at dysfunctional telomeres

The importance of RNF8 for the recruitment of DNA damage factors to sites of genomic DSBs prompted us to examine whether RNF8 is required to initiate a DDR at uncapped telomeres. To examine whether RNF8 is required to localize DDR proteins to dysfunctional telomeres, uncapped telomeres were generated in both Rnf8 competent and Rnf8^−/− SV40LT immortalized mouse embryonic fibroblasts (MEFs) by removing mTRF2 with retrovirus-mediated short hairpin RNA against mTRF2 (shTrf2), or by expressing the dominant-negative mutant mTPP1^ΔRD to deplete the endogenous mTPP1-mPOT1a/b complex^3,4,6,7. Removal of mTRF2 or mTPP1-mPOT1a/b from Rnf8^+/+ MEFs initiated equally robust DDR at telomeres, manifested as 53BP1 and γ-H2AX positive dysfunctional telomere-induced DNA damage foci (TIFs) (Fig. 1a, b). In Rnf8^−/− MEFs devoid of mTRF2 or mTPP1-mPOT1a/b, γ-H2AX-positive TIFs, but not ~3BP1-positive TIFs, accumulated on dysfunctional telomeres (Fig. 1a–c). Compared with Rnf8^+/+ MEFs, in Rnf8^−/− MEFs γ-H2AX-positive TIF formation was reduced ~3-fold at telomeres devoid of either mTRF2 or mTPP1-mPOT1a/b (Fig. 1c). These results suggest that RNF8 is absolutely required for recruitment of 53BP1 to uncapped telomeres. In contrast, RNF8 is required for efficient phosphorylation of γ-H2AX to uncapped telomeres.

We next examined whether RNF8 localizes to uncapped telomeres. RNF8 contains an N-terminal forkhead-associated (FHA) domain which is required for its localization to damaged DNA by interacting with phosphorylated MDC1^12,13,15. RNF8 also contains a C-terminal RING-finger domain responsible for its ubiquitin ligase activity^16,17. We found that wild-type FLAG-mRNF8 readily localized to dysfunctional telomeres lacking mTRF2 or mTPP1-mPOT1a/b (Fig. 1d, e, f). Similarly, the mRNF8^C406S RING domain mutant also localized to uncapped telomeres (Fig. 1d, e, f). In contrast, the mRNF8^R42A FHA domain mutant failed to localize to uncapped telomeres (Fig. 1d, e, f), supporting the notion that this domain is critically important for localization of RNF8 to sites of DNA damage.

Reconstitution of wild-type mRNF8 into Rnf8^−/− MEFs in the setting of mTPP1-mPOT1a/b deficiency completely restored γ-H2AX-positive TIF formation (Supplementary Fig. 1a, b). Since the mRNF8^C406S RING-finger domain mutant readily localizes to uncapped telomeres, we asked whether this domain is required for TIF formation. Reconstitution of Rnf8^−/− MEFs with the mRNF8^C406S mutant failed to bring γ-H2AX-positive TIF levels back to those observed in Rnf8^−/− MEFs reconstituted with wild-type mRNF8 (Supplementary Fig. 1a, b). Taken together, these results suggest that localization of mRNF8 to uncapped telomeres requires an intact FHA domain. In contrast, the RNF8 E3 ubiquitin ligase function is required for efficient recruitment of DNA damage checkpoint proteins to dysfunctional telomeres.

RNF8 is required for telomere end protection

We asked whether RNF8 is required for the repair of dysfunctional telomeres. We have recently shown that dysfunctional telomeres engage two distinct DNA repair pathways: removal of mTRF2 elicits the classical non-homologous end joining (C-NHEJ) reaction to generate long chains of end-to-end chromosome fusions with robust telomere signals at fusion sites, while removal of mTPP1-mPOT1a/b engages an alternative-NHEJ (A-NHEJ) repair pathway that produces fused chromosome without telomeres at fusion sites⁷. Metaphase spreads of Rnf8^+/+ and Rnf8^−/− MEFs were generated and telomere-PNA FISH used to examine the status of chromosome ends. While Rnf8^+/+ MEFs displayed only rare p-p arm chromosomal aberrations involving ~6% of all chromosomes examined, Rnf8^−/− cells displayed elevated spontaneous chromosome aberrations, including p-p fusions, q-q fusions, p-q fusions, rings and tri- or quadri-radials (Fig. 2a, b). A ~20-fold increase in telomere signal-free chromosome ends were observed in metaphases from Rnf8^−/− MEFs, corresponding to the ~20% decrease in total telomere length revealed by TRF Southern analysis (Fig. 2a, Supplementary Fig. 2a, b). These results suggest that telomere end protection is compromised in the absence of RNF8.

In contrast to the chromosome fusions observed upon mTRF2 depletion, chromosome fusions observed in the setting of mRNF8 deficiency resembled those observed when mTPP1-mPOT1a/b was depleted from telomeres⁷. Indeed, removal of mTPP1-mPOT1a/b from telomeres of Rnf8^−/− MEFs increased the number of chromosome fusions ~2.5-fold, to involve ~30% of all chromosome ends (Fig. 2c, d). These results suggest that in the absence of mRNF8, telomeres are rendered dysfunctional and are repaired via the A-NHEJ pathway. To directly assess whether the C-NHEJ repair pathway is functional in the absence of mRNF8, we removed mTRF2 from telomeres of Rnf8^−/− MEFs. End-to-end chromosome fusions decreased ~10 fold in Rnf8^−/− MEFs devoid of mTRF2, and characteristic long trains of fused chromosomes with telomeres at fusion sites were not observed (compare Fig. 2e, left and right panels). The small number of chromosome fusions that remain were indistinguishable from those observed upon mTPP1-mPOT1a/b depletion (Fig. 2e, f), suggesting that they too are repaired via ANHEJ. This result suggests that C-NHEJ is abrogated in Rnf8^−/− MEFs, in agreement with our data demonstrating the inability of 53BP1 to localize to uncapped telomeres in the absence of mRNF8, and the requirement for 53BP1 in promoting C-NHEJ-mediated repair of mTRF2-depleted telomeres⁷.

To assess which domains of RNF8 are required to mediate A-NHEJ of uncapped telomeres, we expressed mRNF8 wild-type or the mRNF8^C406S RING-finger domain mutant in Rnf8^−/− MEFs. Reconstitution of Rnf8^−/− MEFs devoid of mTPP1-mPOT1a/b with wild-type Rnf8, but not with the mRNF8^C406S RING domain mutant, reduced the number of chromosome fusions ~6-fold, suggesting that a functional mRNF8 RING domain is required to repress A-NHEJ mediated chromosome fusions (Supplementary Fig. 3a, b). Taken together, our data suggest that the E3 ubiquitin ligase function of mRNF8 is required to repress A-NHEJ at telomeres. In addition, mRNF8 is also required for C-NHEJ-mediated chromosome fusions by recruiting 53BP1 to telomeres devoid of mTRF2.

RNF8 is required for the accumulation of the mTPP1-mPOT1a/b complex at telomeres

The observation that the A-NHEJ pathway is activated at telomeres of Rnf8^−/− MEFs suggests that the mTPP1-mPOT1a/b complex might be functionally compromised in the absence of mRNF8. We hypothesize that accumulation of mTPP1-mPOT1a/b at telomeres is reduced in Rnf8^−/− MEFs, resulting in telomere deprotection and failure to repress A-NHEJ-mediated chromosome fusions. In support of this notion, Western analysis revealed that expression of endogenous (Fig. 3a, lanes 1 and 3) and over expressed (lanes 5 and 7) mTPP1 was reduced ~3-fold in Rnf8^−/− MEFs, while the expression of endogenous mTRF2 was not appreciably affected. To confirm that RNF8 levels influenced endogenous TPP1 protein expression, we depleted RNF8 in HeLa cells using two independent siRNAs. Endogenous hTPP1 protein levels were markedly reduced in hRNF8-depleted HeLa cells due to a decrease in protein half-life, while hTRF2 protein levels remain unaltered (Fig. 3b and Supplementary Fig. 4a). Since Tpp1 mRNA levels were unaffected in both Rnf8^−/− MEFs and in RNF8 depleted HeLa cells, these results suggest that hRNF8 regulates hTPP1 at the posttranslational level (Supplementary Fig. 4b, c).

We next performed indirect immunofluoresence and telomere-PNA FISH on Rnf8^+/+ and Rnf8^−/− MEFs to examine the level of endogenous mTPP1 at telomeres. Compared to Rnf8^+/+ MEFs, Rnf8^−/− MEFs displayed a ~5-fold reduction of endogenous mTPP1 at telomeres (Fig. 3c, d). Examination of epitope-tagged shelterin components revealed that while hTRF1, hTRF2, and hRAP1 were found in similar quantities at telomeres of both Rnf8^+/+ and Rnf8^−/− MEFs, hTIN2, hTPP1 and mPOT1a accumulation at telomeres were all reduced ~3-fold in Rnf8^−/− MEFs (Supplementary Fig. 5a–c). This result is in agreement with previous observations that hTPP1 expression is critical for the localization of both hPOT1 and hTIN2 to telomeres¹⁸.

Finally, we asked whether the ubiquitin ligase activity of RNF8 is required for accumulation of TPP1 at telomeres. While expression of wild-type mRNF8 in Rnf8^−/− MEFs resulted in robust accumulation of mTPP1 at telomeres, the mRNF8^C406S RING domain mutant was unable to restore mTPP1 expression at telomeres (Fig. 3e, Supplementary Fig. 6). Taken together, these results suggest that stable accumulation of mTPP1 at telomeres requires the ubiquitin ligase activity of mRNF8.

Loss of RNF8 results in increased TPP1 turnover by proteasome-mediated degradation

The ubiquitin/26S proteasome pathways are involved in protein degradation. We addressed the possibility that the decreased mTPP1 protein levels observed in Rnf8^−/− MEFs is due to increased proteasome-mediated turnover of mTPP1. We examined the levels of endogenous mTPP1, as well as exogenously expressed HA-mTPP1, in both Rnf8^+/+ and Rnf8^−/− MEFs in the presence or absence of the proteasome inhibitor MG132. MG132 treatment substantially increased endogenous mTPP1 protein levels in Rnf8^−/− MEFs to levels observed in Rnf8^+/+ cells, suggesting that mRNF8 protects mTPP1 from proteasome-dependent degradation (Fig. 4a, b). Exogenously expressed FLAG-hTIN2 was also stabilized by the addition of MG132 in Rnf8^−/− MEFs, consistent with previous data suggesting that hTPP1 stabilizes hTIN2 at telomeres¹⁸. (Fig. 4c). In contrast, the addition of MG132 did not substantially influence the exogenous expression of FLAG -hTRF1, FLAG -hTRF2, HA-hRAP1 and Myc-mPOT1a in Rnf8^−/− MEFs (Fig. 4c, d). Since MG132 treatment stabilized mTPP1 protein levels in the absence of RNF8, we next asked whether stabilized mTPP1 are observed at telomeres of Rnf8^−/− MEFs treated with MG132. While treatment with MG132 resulted in rapid depletion of the nuclear ubiquitin pool in IR-treated HeLa cells¹⁵, the absence of exogenous DNA damaging agents resulted in the slow depletion of the nuclear ubiquitin pool from MG132-treated Rnf8^+/+ MEFs, resulting in the retention of the majority of conjugated ubiquitin in the nucleus and mTPP1 at telomeres (Fig. 4e, f, Supplementary Fig. 7a, b). Surprisingly, both endogenous and exogenously expressed mTPP1 were unable to localize to telomeres in Rnf8^−/− MEFs treated with MG132, instead displaying diffuse cytoplasmic staining (Fig. 4e, f). Taken together, these results suggest that functional RNF8 is required for TPP1 to localize to telomeres and for the maintenance of TPP1 protein stability by preventing its proteosome-mediated degradation in the cytoplasm.

Physical interaction between RNF8 and TPP1 stabilizes TPP1 through ubiquitylation

To determine the mechanism through which RNF8 regulates TPP1 stability, we asked whether RNF8 physically interacts with TPP1. We co-transfected 293T cells with either FLAG-mRNF8 and HA-hTPP1 or HA-mRNF8 and Myc-hTPP1. Whole-cell extracts were subjected to reciprocal immunoprecipitation with anti-HA antibody. FLAG-mRNF8 was found to co-precipitate with HA-hTPP1, while Myc-hTPP1 associated with HA-mRNF8, suggesting a specific interaction between RNF8 and TPP1 in vivo (Fig. 5a, Supplementary Fig. 8). The finding that TPP1 interacts physically with RNF8 led us to hypothesize that RNF8 might mediate TPP1 ubiquitylation. To examine whether TPP1 ubiquitylation is directly and specifically mediated by RNF8, we carried out an in vitro ubiquitylation assay with purified wild type mRNF8 and the catalytic dead mRNF8^C406S RING domain mutant in presence of purified E1 ubiquitin-activating enzyme (Uba1) and its E2 (Ubc13 and Mms2) together with ATP ubiquitin^19,20. Western blotting with anti-mTPP1 and anti-HA antibodies suggests that RNF8 polyubiquitylates TPP1 in an ATP-dependent manner (Fig. 5b). In contrast, the mRNF8^C406S mutant failed to ubiquitylate mTPP1, suggesting that the RNF8 RING domain is essential for TPP1 ubiquitylation (Fig. 5b; compare lanes 1 and 4). Omission of any component required for ubiquitin conjugation completely abrogates TPP1 ubiquitylation (Fig. 5b and data not shown). The mRNF8^C406S mutation did not alter the ability of mRNF8^C406S to bind to mTPP1 (data not shown). Collectively, these results suggest that RNF8 directly and specifically ubiquitylate TPP1 through its E3 ligase function.

TPP1 ubiquitylation by RNF8 was further validated by an in vivo ubiquitylation assay using a HeLa-His/biotin-ubiquitin (HeLa UB) cell line stably expressing tandem 6× His-biotin-tagged ubiquitin²¹. This system offers the advantage of purifying ubiquitylated proteins under fully denatured conditions, thus reducing non-specific interactions. In order to perform this assay, we stably expressed HA-mTPP1 in HeLa UB cell lines and treated the cells with biotin overnight. Ubiquitylated protein was then purified with strepavidin-agarose beads. Several HA-positive bands were observed, were observed, S.</author></authors></contributors><auth-address>Department of Genetics, MD Anderson Cancific, since they were not detected in the absence of His-biotin-tagged ubiquitin expression or in HeLa UB cell lines transfected with empty vector (Fig. 5c). As further proof of specificity, we treated HeLa UB cells stably expressing mTPP1 with a shRNA against mTPP1. mTPP1 polyubiquitylation was undetectable in mTPP1 depleted cells, demonstrating the specificity of mTPP1 polyubiquitylation in vivo (Fig. 5c, lane 4).

Ubiquitin can form polymers in vivo through any of its seven lysyl residues²². To analyze the class of polyubiquitin chains supported by RNF8's E3 ligase activity on TPP1, we co-transfected HeLa cells with plasmids encoding HA-mTPP1 and His-tagged expression vectors for ubiquitin species in which all lysines were mutated to arginines except the ones indicated. This system enables the in vivo synthesis of polyubiquitin chains containing defined isopeptide linkages. The expressed His-tagged ubiquitin was then pulled down from the cell lysates with Ni²⁺-NTA agarose beads under denatured conditions. Subsequent anti-HA immunoblotting enabled detection of ubiquitylated HA-mTPP1. We found that wild-type mTPP1 undergoes ubiquitylation with both lysine 48 and lysine 63 linked polyubiquitin chains (Fig. 5d, lanes 3, 5, 6). Other polyubiquitin chains, including K6, K11, K27, K29, and K33 were not conjugated to mTPP1 (see below). We next asked if the K63 and K48 polyubiquitin chains on mTPP1 were specifically generated by endogenous hRNF8. To accomplish this, we transfected HeLa cells with His-tagged K48 and K63 ubiquitin species and treated them with siRNA against RNF8. Both K48 and K63 polyubiquitin chains were completely abolished in hRNF8-depleted HeLa cells (Fig. 5d, lanes 7, 8). In addition, treatment with MG132 in HeLa cells depleted of endogenous hRNF8 restored K48 but not K63 polyubiquitin chains on mTPP1 (Supplementary Fig. 9, lanes 8, 9). Therefore, our data suggest that TPP1 can be ubiquitylated by RNF8 through K63 polyubiquitin chain, while K48 polyubiquitin chains of TPP1 are likely synthesized by an unknown ubiquitin ligase.

To further examine the role of RNF8-mediated K63 polyubiquitin on TPP1 stability and telomere metabolism, we probed immunoprecipitated nuclear and cytoplasmic extracts isolated from Rnf8^+/+ or Rnf8^−/− MEFs expressing mTPP1 with linkage-specific antibodies that specifically discriminate ubiquitin chains formed by either K63 or K48²³. mTPP1 ubiquitylated species were only detected by the anti- K63 but not with the anti-K48-linkage antibody in the nuclear fraction isolated from Rnf8^+/+ MEFs (Supplementary Fig. 10a, b). The presence of K63-linked ubiquitin chains was not detected in Rnf8^−/− MEFs, strongly suggesting that RNF8 is required to generate K63 chain on TPP1(Supplementary Fig. 10a, b).

Finally, we utilized a RNF8 RING domain mutant (hRNF8^I405A) specifically impaired to perform K48-based but not K63-based polyubiquitylation reaction²⁴. In contrast to the hRNF8^ΔRING or the mRNF8 RING^R42A mutants, hRNF8^I405A efficiently targeted mTPP1 to telomeres in Rnf8^−/− MEFs (Supplementary Fig. 11). Reconstitution of hRNF8^I405A into Rnf8^−/− MEFs in the setting of mTRF2 deficiency completely restored 53BP1-positive TIF formation (Supplementary Fig. 12a) and C-NHEJ-mediated end-to-end chromosome fusions (Supplementary Fig. 12b, c). Taken together, these results suggest that RNF8-dependent K63-linked, but not K48-linked ubiquitin chain formation on TPP1 is required to promote TPP1 stability and function at telomeres.

RNF8 can accommodate on its RING finger domain with several E2 enzymes to synthesize either K48 or K63 polyubiquitin chains on its substrates¹⁶. UBC13 is the E2 conjugating enzyme that exclusively promotes K63 linked ubiquitylation^25,26. Our data suggest that RNF8 can promote in vivo synthesis of K63, but not K48 polyubiquitin chain on TPP1 (Figure 5d, Supplementary Figure 9). Therefore, we hypothesized that RNF8 physically interacts with UBC13 to generate K63-linked polyubiquitin chains to stabilize TPP1 and prevent its degradation. To support our hypothesis, we performed in vitro ubiquitylation assays with purified hTPP1, hRNF8, Ubc13, Mms2, Uba1 along with wild type or mutant ubiquitins. Both the wild type and the K63 ubiquitins were efficiently conjugated to hTPP1, suggesting that the K63 isopeptides is the preferred linkage for hTPP1 ubiquitylation by hRNF8 together with UBC13 (Fig. 5e). To further confirm that K63 ubiquitylation of hTPP1 is dependent upon UBC13, we used three different siRNA to knockdown UBC13. As expected, knockdown of UBC13 resulted in the loss of hTPP1 expression, further solidify the notion that hRNF8 coordinates with UBC13 to conjugate K63 polyubiquitin chains on hTPP1 to prevent its degradation (Supplementary Fig. 13).

A conserved lysine 233 residue of TPP1 is required for RNF8-mediated TPP1 ubiquitylation and localization to telomeres

Our data demonstrate that RNF8 directly ubiquitylates and stabilizes TPP1 at telomeres. We next sought to determine the lysine residues critical for TPP1 ubiquitylation and its localization to telomeres. Alignment of TPP1 sequences across several vertebrate species revealed three highly conserved lysine residues: TPP1^K170, TPP1^K232 and TPP1^K233 all reside within the OB fold, while TPP1^K467 resides in the S/T domain (Fig. 6a). To test the roles of these lysine residues on hTPP1 ubiquitylation, we mutated them individually or together into arginines and trasfected these FLAG-hTPP1 mutants into HeLa cells for in vivo ubiquitylation assays. We also generated purified mutant TPP1 proteins for in vitro ubiquitylation assays. All mutants expressed at similar levels in HeLa cells, suggesting that lysine to arginine mutations did not impact upon protein stability (Supplementary Fig. 14a, lower panel). Western analysis revealed that mRNF8-mediated hTPP1 in vitro ubiquitylation was reduced ~2.5-fold in the hTPP1^K233R mutant as compared to hTPP1^FL (Fig. 6b). Analysis of the hTPP1^K232R,K233R double mutant revealed that it is completely unable to be ubiquitylated in vivo (Supplementary Fig. 14a, upper panel). In contrast, hTPP1^K170R and hTPP1^K476R mutants did not alter mRNF8-mediated hTPP1 ubiquitylation, while uqiquylation was only slightly reduced in the hTPP1^K232R mutant. These results suggest that lysine 233 is required for mRNF8-mediated ubiquitylation.

Since the ubiquitylation status of TPP1 might influence its localization to telomeres, we examined the telomere localization of the hTPP1 lysine mutants. FLAG-hTPP1^WT, FLAG-hTPP1^K170R, FLAG-hTPP1^K232R and FLAG-hTPP1^K476R were all visualized on telomeres (Fig. 6c, d). In contrast, MEFs transfected with FLAG-TPP1^K233R and the double mutant FLAG-hTPP1^K232R/K232R displayed only diffused nuclear staining, again reinforcing the observation that lysine K233 is the critical lysine residue for the accumulation of TPP1 at telomeres (Fig. 6c, d). The failure of the FLAG-hTPP1^K233R mutant to localize to telomeres suggests that failure to ubiquitylate TPP1 at lysine 233 results in either an inability of this mutant to be recruited to telomeres, or an inability to be retained there. Retention of endogenous TPP1 at telomeres requires the presence of nuclear ubiquitin, since the addition of MG132 depletes TPP1 from telomeres (Supplemental Fig. 7a, b). This result suggests that ubiquitylation of TPP1 by RNF8 is important for its retention at telomeres.

Since POT1 recruitment to telomeres depends upon its interaction with TPP1^3,27,28, we next asked whether localization of POT1-TPP1 complex to telomeres requires TPP1 ubiquitylation. To address this question, we co-transfected Myc-mPOT1a with either FLAG-hTPP1^WT or FLAG-tagged hTPP1 mutants into wild-type MEFs. Myc-mPOT1a failed to localize to telomeres only in the presence of FLAG-hTPP1^K233R (Supplementary Fig. 14b and data not shown). Co-IP experiments revealed that all hTPP1 mutants were able to interact efficiently to mPOT1a in vitro (Supplementary Fig. 14c). Metaphases isolated from wild-type MEFs expressing hTPP1^K233R also displayed ~3-fold increase in chromosome fusions, likely due to its ability to titrate endogenous mPOT1a/b away from telomeres (Supplementary Fig. 14d). These results suggest that retention of the POT1-TPP1 complex at telomeres requires RNF8-mediated ubiquitylation of TPP1 at lysine 233.

Materials and Methods