|Home | About | Journals | Submit | Contact Us | Français|
Telomere length homeostasis is an important aspect of telomere biology. Here we show that SUMOylation limits telomere length and targets multiple telomere proteins in Saccharomyces cerevisiae. A main target is Cdc13, which both positively and negatively regulates telomerase and confers end protection. We demonstrate that Cdc13 SUMOylation restrains telomerase functions by promoting Cdc13 interaction with the telomerase inhibitor Stn1, without affecting end protection. Mutation of the Cdc13 SUMOylation site (cdc13-snm) lengthens telomeres and reduces the Stn1 interaction, whereas Cdc13-SUMO fusion has the opposite effects. cdc13-snm's effect on telomere length is epistatic with stn1, but not with yku70, tel1 or est1 alleles, and is suppressed by Stn1 overexpression. Cdc13 SUMOylation peaks in early to mid S phase, prior to its known Cdk1-mediated phosphorylation, and the two modifications act antagonistically, suggesting that the opposite roles of Cdc13 in telomerase regulation can be separated temporally and regulated by distinct modifications.
Telomeres are essential for the stability and complete replication of the genome. In the budding yeast S. cerevisiae, telomeric sequences are composed of ~300 base pairs of TG1–3/CA1–3 repeats followed by short single-stranded 3′ extensions of TG1–3 sequences, called G-overhangs. Telomere DNA is bound by more than a dozen proteins that collaborate in telomere maintenance. Important functions of these proteins include telomere length homeostasis, G-overhang protection, nuclear positioning and transcriptional repression. Each function can be regulated by multiple pathways. In particular, maintenance of a constant telomere length unique to each organism requires a balance of positive and negative regulations of telomerase.
In budding yeast, three positive regulators recruit telomerase to telomeres. These include the essential protein Cdc13 that binds to the Est1 subunit of the telomerase, the DNA end-binding Ku complex and the Tel1 kinase (reviewed in refs. 1,2). Concurrently, three negative regulators restrain telomerase, including the Rif1 and Rif2 proteins, the Pif1 helicase, and Cdc13 when bound to the Stn1 subunit of the Stn1–Ten1 telomerase inhibitory complex (reviewed in refs. 1,2). This multilayered regulatory network possibly confers robustness to telomere length homeostasis.
It is becoming increasingly clear that dynamic protein modifications have important roles in telomere length regulation, possibly by modifying some or all of the telomerase regulators. For instance, Cdk1 is critical for telomere synthesis during late S to G2 phases and is thought to target multiple telomere proteins3–7. One such protein substrate is Cdc13, which, besides regulating telomerase positively and negatively in S phase, also prevents telomere end degradation1,2,8–15. Cdk1 phosphorylates Cdc13 at Thr308 in its Est1 binding domain during late S and G2 phases, leading to enhanced telomerase recruitment at telomeres without affecting end protection6. This regulation explains one facet of Cdc13's activation of telomerase and provides a clue to the conundrum of how Cdc13 exerts opposite regulations of telomerase within the same cell-cycle stage1,2.
Another type of protein modification implicated in telomere metabolism is SUMOylation, which entails the covalent linkage of the small ubiquitin-like modifier (SUMO) to the lysine residues of target proteins16–22. SUMOylation is executed by the sequential action of SUMO E1, E2 and E3 proteins. Most organisms possess one E1, one E2 and multiple E3s, which facilitate conjugation to a wide range of substrates. Like phosphorylation, SUMOylation exerts diverse effects on its targets and is reversible in nature23. It has been shown that mutating SUMO E3s in fission yeast leads to longer telomeres in a telomerase-dependent manner, although it has not been determined which, if any, telomere proteins are directly SUMOylated19.
In this study, we examined how SUMOylation affects telomere length homeostasis in the budding yeast S. cerevisiae. We found that defective SUMOylation in this organism also leads to telomere lengthening. We identified five telomeric proteins as SUMO targets and we demonstrated that SUMOylation of the telomerase regulator Cdc13 limits telomere length by strengthening the Cdc13-Stn1 interaction that is required for telomerase inhibition. Examination of the relationship between SUMOylation and Cdk1-mediated phosphorylation of Cdc13 suggests that the two modifications act antagonistically to promote the opposing roles of Cdc13 in telomere homeostasis.
To know whether SUMOylation contributes to telomere length regulation in budding yeast, we examined ubc9-1, a temperature-sensitive allele affecting the SUMO E2 enzyme Ubc9 (ref. 23), and found that telomeres in ubc9-1 cells were approximately 145 base pairs (bp) longer than those in wild-type cells, even at a semi-permissive temperature of 30 °C (Fig. 1a). Two other ubc9 alleles also resulted in longer telomeres, albeit to a lesser degree than did ubc9-1 (Fig. 1a). These observations suggest that SUMOylation is required for limiting telomere length in budding yeast.
To determine which SUMO E3s are involved in this process, we examined mutants of the nonessential Siz1 and Siz2 E3s and the essential Mms21 E316,24. Telomeres of siz1Δ siz2Δ and siz2Δ mms21-CH double mutants (mms21-CH has mutations in the critical residues of the SUMO E3 domain25) were approximately 60 bp and 90 bp longer than wild-type telomeres, respectively (Fig. 1b). No obvious telomere length alteration was seen in siz1Δ mms21-CH or single SUMO E3 mutants, compared to wild-type cells (Fig. 1b,c). The E3 triple mutant siz1Δ siz2Δ mms21-CH was lethal, preventing the examination of telomere length in those cells. These results suggest that all three SUMO E3s contribute to telomere length regulation with partly redundant roles, consistent with the observation that these E3s share a considerable number of targets (ref. 26 and see below).
Next, we examined whether any of the known telomere proteins are SUMOylated. We tested thirteen proteins: the telomerase subunits, the Ku complex, Cdc13, Stn1, Ten1, Tel1, Rap1, Rif1, Rif2 and Pif1. Each protein was tagged with either TAP (composed of protein A (ProA) and calmodulin-binding peptide) or Myc epitope at its own genomic locus. Protein extracts were made under denaturing conditions to prevent the co-purification of interacting proteins and to preserve SUMOylated forms; each protein was immunoprecipitated and examined by western blot analysis with an antibody specific to the tag and another recognizing the budding yeast SUMO16 (Fig. 1d). Because a low percentage of substrate is typically SUMOylated23, the anti-tag antibody detects only the unmodified form at the exposures shown. Bands preferentially recognized by the anti-SUMO antibody and specifically decreased in SUMO E3 mutants (siz1Δ siz2Δ and mms21) represent SUMOylated forms (Fig. 1d and Supplementary Fig. 1a,b).
In accordance with previous reports, Rap1 and Yku70 were SUMOylated under normal growth conditions16,27. In addition, we identified a previously unknown SUMO substrate, Cdc13, and verified its SUMOylation using a tagged version of SUMO that resulted in a shift of the SUMOylated band (Fig. 1d and Supplementary Fig. 1c).
Because SUMOylation is known to respond to environmental stimuli, we examined the panel of telomere proteins under DNA-damage and high-temperature (37 °C) conditions, which also influence telomere metabolism or telomere proteins28–30. Consistent with our previous report, methylmethanesulfonate (MMS) treatment enhanced Yku70 SUMOylation16. MMS also stimulated the SUMOylation of Rap1 and Cdc13 and allowed clear detection of the SUMOylation of Pif1 and Yku80 (Fig. 1d). Moreover, SUMOylation of four of these proteins (Rap1,Yku70, Cdc13 and Pif1) also increased at 37 °C (Fig. 1e). For comparison, SUMOylation of the recombination protein Rad52 was induced after MMS treatment but not at 37 °C (ref. 31) (Fig. 1e).
It is noteworthy that the SUMOylation patterns of these five telomere proteins vary, with Cdc13 showing mostly a monoSUMO form and the other proteins showing poly- or multiSUMO forms. In addition, protein SUMOylation is differentially dependent on the three SUMO E3s: whereas the Siz proteins were responsible for the bulk of Cdc13, Yku80 and Rap1 SUMOylation, mms21 partly affected that of Pif1 and Yku70 (Supplementary Fig. 1a,b,d; ref. 16). In the case of Cdc13, siz1Δ siz2Δ, but not the single sizΔ mutants, eliminated most of its SUMOylation (Supplementary Fig. 1e). These observations indicate that the three E3s have both unique and overlapping roles in the SUMOylation of telomere proteins, a notion consistent with their effects on telomere length (Fig. 1b,c).
We next examined how SUMO regulates the function of Cdc13. We chose to focus on Cdc13 because it functions solely at telomeres and is SUMOylated in all conditions tested. We first examined the cell cycle stage during which Cdc13 is SUMOylated. Cdc13 SUMOylation was barely detectable in G1 cells, appeared when cells entered S phase, peaked during early to mid S phase and decreased in late S and G2 phases (Fig. 2a). This pattern is different from what is seen in telomerase enrichment and activation at telomeres, which is low in early to mid S phase and peaks in late S phase (refs. 1,2,13 and Supplementary Fig. 2). These observations indicate that SUMOylation may regulate Cdc13 in early to mid S phase before telomerase is activated.
Next, we sought to generate a non-SUMOylatable cdc13 mutant. We identified a SUMOylation consensus sequence ψKXE/D (ψ: bulky hydrophobic residue) within the Stn1-binding domain of Cdc13 that is highly conserved among orthologs in the Saccharomyces clade23,32 (Fig. 2b). Indeed, mutating the lysine to arginine within this consensus sequence abolished Cdc13 SUMOylation in vivo (Fig. 2c). This cdc13-K909R allele (hereafter referred to as cdc13-snm, SUMO no more) affected neither Cdc13 protein levels before or after MMS treatment nor cell cycle progression (Fig. 2d and Supplementary Fig. 3a), indicating that Cdc13-snm retains the essential function of Cdc13 and that SUMOylation does not affect its protein stability. The cdc13-snm allele thus provides a useful genetic tool to assess the role of Cdc13 SUMOylation in telomere metabolism.
We then examined cdc13-snm's effects on telomere length and observed a reproducible 60-bp increase in both W303 and YPH strains (Fig. 3a and Supplementary Fig. 3b). This lengthening was dependent on telomerase, as it did not occur in the absence of telomerase components (tlc1Δ and est1Δ, Fig. 3b and Supplementary Fig. 3c). The epistatic relationship of cdc13-snm with telomerase deletions extended to senescence as well (Fig. 3c and Supplementary Fig. 3d). We also assessed whether cdc13-snm affects end protection but found no detectable changes in G-overhang abundance at either 30 °C or 37 °C in the cdc13-snm strains (Supplementary Fig. 4a). These results demonstrate that cdc13-snm affects telomerase-mediated telomere elongation without affecting end protection.
A logical interpretation of the above observations is that lack of SUMOylation underlies the telomere length defect of cdc13-snm. To further test this idea, we examined the genetic relationship between cdc13-snm and siz1Δ siz2Δ, which abolishes the majority of Cdc13 SUMOylation (Supplementary Fig. 1e). If the above hypothesis is correct, cdc13-snm should not further increase telomere length in siz1Δ siz2Δ cells. Indeed, we found that the telomere length of cdc13-snm siz1Δ siz2Δ cells was similar to that of siz1Δ siz2Δ and of cdc13-snm, and all were approximately 60-bp longer than wild type (Supplementary Fig. 4b). This epistatic relationship supports the above interpretation and indicates that Cdc13 is a main shared substrate for the Siz SUMO E3s in telomere length regulation.
The Lys909 SUMOylation site is located in Cdc13's Stn1-binding domain but not in its Est1, Pol1 or DNA binding domains (refs. 12,32–36 and M. Lei, University of Michigan Medical School, personal communication) (Fig. 2b). We asked whether SUMOylation at Lys909 could alter Cdc13's interaction with Stn1. The Cdc13-Stn1 interaction is difficult to measure: several labs have detected it by yeast two-hybrid (Y2H) assay but not by co-immunoprecipitation of the proteins expressed at endogenous levels6,8,11,15,34,37. Using Y2H assay, we found that when combined with Stn1, wild-type Cdc13 activated the Ade+ reporter to a greater degree than Cdc13-snm (Fig. 4a). Similarly, the Stn1 and Cdc13 pair showed a more robust activation of the His+ and LacZ reporters than the Stn1 and Cdc13-snm pair (Fig. 4a,b). As a control, Cdc13 and Cdc13-snm showed the same level of interaction with Pol1 (Fig. 4b). These results indicate that the cdc13-snm allele specifically weakens the interaction with Stn1. To test if this weakened Stn1 interaction is due to deficient SUMOylation, we examined wild-type Cdc13-Stn1 interaction in Y2H strains lacking the Siz E3 proteins, which are responsible for Cdc13 SUMOylation (Supplementary Fig. 1e). The Cdc13-Stn1 interaction, but not control interactions, was reduced in siz1Δ siz2Δ cells (Supplementary Fig. 4c). Taken together, these results indicate that lack of Cdc13 SUMOylation achieved by either the K909R mutation or by the siz1Δ siz2Δ mutation weakens the Cdc13-Stn1 interaction.
The above results suggest that Cdc13-snm may lead to diminished Stn1-dependent telomerase inhibition. To test this idea, we investigated whether cdc13-snm could suppress mutants defective in other telomerase regulatory pathways. Cells lacking the Ku complex, a positive regulator of telomerase and end protection, have shorter telomeres and are inviable at 37 °C; this temperature sensitivity, due to faulty telomere structures, can be suppressed by the overexpression of telomerase subunits38–42. If cdc13-snm leads to increased telomerase function by weakening Stn1-mediated inhibition, but not by affecting the Ku pathway, then cdc13-snm should suppress kuΔ temperature sensitivity and telomere length defects. Indeed, yku70Δ cdc13-snm and yku80Δ cdc13-snm cells grew at 37 °C and had longer telomeres than kuΔ cells (Fig. 4c and Supplementary Fig. 5a,b). This suppression is unlikely to be the result of rescuing defective end protection in kuΔ cells, as cdc13-snm did not change G-overhang levels in yku70Δ cells (Supplementary Fig. 5c).
We next tested Cdc13-snm's effect in cells lacking Tel1, another positive telomere length regulator1,43, or containing a partially defective Est1, which binds Cdc13 to achieve telomerase recruitment9,11–14. Again, if telomere lengthening in cdc13-snm mutants occurs through misregulation of Stn1, then cdc13-snm should increase telomere length in tel1Δ cells or in cells with defective Est1. As with yku70Δ, cdc13-snm increased telomere length in both types of cells (Fig. 4d,e). These results indicate that cdc13-snm enhances telomerase function independently of Ku and Tel1 and can occur in cells with hypomorphic Est1.
We then conducted epistasis tests with stn1 alleles, which show longer telomeres with increased heterogeneity6,8,10,11,15,37. We used two stn1 alleles: a mild one generated by the addition of a Myc tag (stn1-myc), and stn1ΔC199, which lacks the C-terminal region required for Cdc13 binding and telomerase inhibition6,10,11,15,37. In both cases, cdc13-snm did not further increase telomere length, indicating an epistatic relationship (Fig. 5a,b). We then examined how Stn1 overexpression affects telomere length in cdc13-snm cells. If cdc13-snm weakens the Stn1 interaction, increased levels of Stn1 may compensate for this deficiency and consequently suppress the cdc13-snm defect. Indeed, Stn1 overexpression reduced the telomere length of cdc13-snm cells to approximately wild-type levels (Supplementary Fig. 6). These genetic results support the idea that cdc13-snm influences telomere length by means of the Stn1 pathway.
We also investigated whether Stn1 influences Cdc13 SUMOylation and found that both stn1 alleles led to increased Cdc13 SUMOylation (Fig. 5c). In particular, stn1ΔC199 resulted in both a large increase in mono-SUMOylated Cdc13 and the appearance of polySUMOylated Cdc13. All SUMOylated Cdc13 species in stn1ΔC199 cells depended on the Lys909 residue, as the Cdc13-snm protein was not SUMOylated in these cells (Fig. 5c). The increased level of Cdc13 SUMOylation in stn1 cells was not due to a general enhancement of telomere protein SUMOylation, as stn1ΔC199 did not affect Rap1 and Yku70 SUMOylation (Fig. 5d). It was also not due to increased telomere length, as rif1Δ mutants that also have longer telomeres did not stimulate Cdc13 SUMOylation (Fig. 5e). Nor was the Stn1's effect due to increased G-overhang levels, as yku70Δ cells with excessive G-overhangs had normal levels of Cdc13 SUMOylation (Fig. 5f). Lastly, Cdc13 SUMOylation was unchanged in tel1Δ or est1-myc cells (Fig. 5f), suggesting that defective telomerase function is not sufficient to affect Cdc13 SUMOylation. Therefore, stn1 alleles exert a unique effect on Cdc13 modification, further strengthening the connection between Cdc13 SUMOylation and Stn1.
To analyze the role of Cdc13 SUMOylation from another angle, we used a `constitutively SUMOylated' Cdc13 allele generated by fusing Cdc13 with SUMO. Because Lys909 is only 15 amino acids away from the Cdc13 stop codon, we reasoned that fusing SUMO to the C terminus might resemble a constitutively SUMOylated protein and that this protein and cdc13-snm should express opposite phenotypes. Indeed, the Cdc13-SUMO fusion led to shortened telomeres (Fig. 6a). We then examined the physical interaction between Cdc13-SUMO and Stn1. The Y2H analysis is not suitable for this test, as SUMO can bind a large number of proteins in the assay44,45. We therefore did coimmunoprecipitation using extracts with overexpressed Stn1 because this approach can detect the Cdc13-Stn1 interaction6. We observed a moderate yet reproducible increase in the amount of Cdc13-SUMO co-immunoprecipitated with Stn1, when compared to Cdc13 (Fig. 6b). These results with Cdc13-SUMO further support the notion that SUMO favors Cdc13's interaction with Stn1, leading to stronger inhibition of telomerase function.
Whereas SUMOylation of Cdc13 acts as a negative regulator of telomere length, Cdk1-phosphorylation of Cdc13 at Thr308 (in Cdc13's Est1 binding domain) promotes telomerase recruitment6. We sought to understand the interplay of these two modifications of Cdc13. First, we found that the non-phosphorylatable cdc13-T308A did not affect Cdc13 SUMOylation (Fig. 6c). Conversely, cdc13-snm did not alter the levels of Cdc13-Thr308 phosphorylation (Fig. 6d). Thus, globally assessed, each modification is not a prerequisite for the other, nor are they competitive. Next, telomere length in cells lacking both modifications (cdc13-T308A snm) was examined and found to be between that of cdc13-T308A and cdc13-snm, suggesting that the two modifications act antagonistically and that neither effect is dominant (Fig. 6e). Finally, we examined the combined effect of T308A and SUMO fusion and found that the telomere-shortening effects of CDC13-SUMO and cdc13-T308A were approximately additive in cdc13-T308A-SUMO cells, consistent with the independent effects of the two modifications (Fig. 6a).
Here we demonstrate that SUMOylation limits telomere length by directly modifying telomere proteins in budding yeast. The observation that the abolition of Cdc13 SUMOylation led to longer telomeres, whereas Cdc13-SUMO fusion resulted in shorter telomeres, indicates that SUMO negatively regulates telomere length, in part by targeting Cdc13 (Figs. 3a and and6a).6a). We show that Cdc13 is a main target of Siz SUMO E3s in telomere length regulation, though SUMOylation of other identified substrates probably contribute to the full inhibitory effect of SUMO, because telomere lengthening in cdc13-snm or siz1Δ siz2Δ cells was not as severe as in ubc9-1 cells (Fig. 1 and Supplementary Fig. 4b). It is noteworthy that a similar multisubstrate strategy has been suggested for Cdk1 kinase, with Cdc13 being one of several targets3–7.
As some of the identified SUMO substrates modulate transcriptional silencing and telomere positioning, their SUMOylation could affect these aspects of telomere metabolism. Indeed, Siz2-mediated Ku SUMOylation was found to influence telomere positioning46. Our finding of MMS- and high temperature–induced SUMOylation may suggest that SUMO can also modulate telomerase behavior both at native telomeres and/or at DNA breaks under these conditions (Fig. 1d,e). For example, SUMO may prevent telomere addition at the DNA breaks in favor of DNA repair, similar to the effect of Mec1 checkpoint kinase-mediated phosphorylation of telomere proteins29,30. Our identification of telomere SUMO substrates opens up opportunities for further analysis of SUMO's roles in telomere metabolism under various conditions.
SUMOylation specifically affects the actions of Cdc13 in telomerase regulation but not in end protection, as shown by our finding that cdc13-snm lengthened telomeres in a telomerase-dependent manner without altering end protection in either wild-type or kuΔ cells (Fig. 3 and Supplementary Figs. 3, 4a, 5c). Several lines of evidence support the possibility that Cdc13 SUMOylation affects the Stn1 pathway of telomerase inhibition. Abolition of Cdc13 SUMOylation by cdc13-snm or siz1Δ siz2Δ weakened the Cdc13-Stn1 interaction in Y2H assays (Fig. 4a,b and Supplementary Fig. 4c). Genetically, cdc13-snm was epistatic with stn1 alleles, but not with tel1Δ, kuΔ and hypomorphic est1 alleles, and its telomere defect was suppressed by Stn1 overexpression (Figs. 4c–e, 5a,b and Supplementary Fig. 6). Moreover, Cdc13-SUMO was enriched in the Stn1 immunoprecipitate and led to shorter telomeres (Fig. 6a,b). These results strongly support the idea that SUMO promotes the Cdc13-Stn1 interaction to restrain telomerase function. The specific effect of increased Cdc13 SUMOylation by stn1 alleles also points to a functional interaction between Cdc13 SUMOylation and Stn1 and raises the possibility of an inhibitory feedback mechanism (Fig. 5c). As SUMO does not interact with Stn1 or Ten1 by Y2H, it is unlikely that SUMO enhances the Cdc13-Stn1 interaction by simply binding to these proteins (data not shown). Rather, SUMO may affect other properties of Cdc13 such as conformation or DNA binding. Further investigation of these ideas requires the development of an in vitro assay for the Cdc13-Stn1 interaction.
We showed that SUMOylation and phosphorylation of Cdc13 operate independently and act antagonistically (Fig. 6a,c–e). This result and those of others6 suggest that each role of Cdc13 can be facilitated by a different modification, which helps explain how Cdc13 acts both positively and negatively to regulate telomerase during S phase. Because Cdc13 SUMOylation peaks in early to mid S phase (Fig. 2a), whereas Cdc13 phosphorylation and telomerase recruitment peak in late S to G2 phase (refs. 1,2,13 and Supplementary Fig. 2), these modifications may help to restrict telomere addition to a narrow window within the cell cycle (Fig. 6f). As Cdc13 was also proposed to inhibit telomerase in late S to G2 phase, Cdc13 may exert a dual-phase inhibition of telomerase11. Other models, which are not mutually exclusive, may also help to explain the interplay of the two modifications. For example, because telomerase is preferentially recruited to short telomeres47, phosphorylation and SUMOylation of Cdc13 may predominantly occur at short and long telomere ends, respectively, to promote or restrain telomerase action (Fig. 6f). Future studies will be needed to test these models and to examine the coordination of these modifications with other layers of telomerase regulation.
Because SUMOylation was also found to limit telomere length in fission yeast19, this modification can have an evolutionarily conserved role in inhibiting telomerases. Considering that Cdk1 and Tel1 kinases and their homologs are typically positive regulators of telomerase, SUMOylation and phosphorylation may provide balancing mechanisms in telomere length homeostasis. Further understanding of their interplay could provide new insights into telomere-related human diseases.
Methods and any associated references are available in the online version of the paper at http://www.nature.com/nsmb/.
We thank S. Li (Duke-National University of Singapore Graduate Medical School) and E. Blackburn (University of California, San Francisco) for the Phos-Thr308 antibody and for several strains, as well as acknowledge S. Li's kind help in using these reagents. We thank M. Lei for sharing unpublished results; M. Arneric, B. Luke, M. Hohl and A. Chan for helping to set up the G-overhang and telomere measurement experiments; Zhao lab members, particularly C. Cremona and P. Sarangi, and N. Lue and T. Weinert for comments on the manuscript; as well as V.A. Zakian, in whose lab some of this work was carried out, for discussions on experiments and comments on the manuscript. This work was supported by US National Institutes of Health (NIH) grant R01GM080670 and Bressler Scholars Award (to X.Z.), a fellowship from the Canadian Institutes of Health Research (to I. C.) and NIH grant GM43265 to V.A. Zakian.
Note: Supplementary information is available on the Nature Structural & Molecular Biology website.
AUTHOR CONTRIBUTIONS X.Z. directed the study. X.Z., X.L. and L.E.H. designed the experiments. X.L., L.E.H., Y.Y. and I.C. carried out the experiments. All authors were involved in data analysis. The manuscript was prepared by X.Z. with the assistance of L.E.H. and X.L.
COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.