Inhibitory protein–protein interactions of the SIRT1 deacetylase are choreographed by post‐translational modification

Abstract

Regulation of SIRT1 activity is vital to energy homeostasis and plays important roles in many diseases. We previously showed that insulin triggers the epigenetic regulator DBC1 to prime SIRT1 for repression by the multifunctional trafficking protein PACS‐2. Here, we show that liver DBC1/PACS‐2 regulates the diurnal inhibition of SIRT1, which is critically important for insulin‐dependent switch in fuel metabolism from fat to glucose oxidation. We present the x‐ray structure of the DBC1 S1‐like domain that binds SIRT1 and an NMR characterization of how the SIRT1 N‐terminal region engages DBC1. This interaction is inhibited by acetylation of K112 of DBC1 and stimulated by the insulin‐dependent phosphorylation of human SIRT1 at S162 and S172, catalyzed sequentially by CK2 and GSK3, resulting in the PACS‐2‐dependent inhibition of nuclear SIRT1 enzymatic activity and translocation of the deacetylase in the cytoplasm. Finally, we discuss how defects in the DBC1/PACS‐2‐controlled SIRT1 inhibitory pathway are associated with disease, including obesity and non‐alcoholic fatty liver disease.

1. INTRODUCTION

Protein–protein interactions (PPIs) involving multi‐domain proteins that contain intrinsically disordered regions (IDRs) are pivotal players in all cellular processes and commonly exhibit dynamic and multivalent features. IDRs and domains linked by IDRs can engage in diverse, sometimes transient, and/or heterogeneous interactions that involve conformational selection (Csermely et al., 2010). Frequently, highly dynamic protein complexes that retain various degrees of disorder are formed, which permits fine‐tuning of interactions in a spatial–temporal fashion with distinct conformations corresponding to different functional outcomes. SIRT1 (silent mating type information regulation 2 homolog 1), a nicotinamide adenine dinucleotide (NAD⁺)‐dependent class III histone deacetylase, is such a multi‐domain protein that links cellular metabolism directly to its enzymatic function (Landry et al., 2000). Many adaptive cellular responses are regulated by SIRT1, involving the post‐translational modification (PTM) of histones (Vaquero et al., 2004; Yuan et al., ²⁰⁰⁹), transcription factors (Brunet et al., 2004; Chen et al., ²⁰⁰⁵; Fu et al., ²⁰⁰⁶; Vaziri et al., ²⁰⁰¹; Wang et al., ²⁰⁰⁶; Yeung et al., ²⁰⁰⁴), and DNA damage repair proteins (Cohen et al., 2004; Jeong et al., ²⁰⁰⁷). Diseases ranging from obesity, diabetes, and cancer to neurodegenerative disorders are all associated with SIRT1 dysregulation (Guarente, 2000). Therefore, elucidating the mechanisms by which SIRT1 exerts its activity and developing ways to control the underlying pathways are of significant medical importance.

In liver and other metabolically active tissues, SIRT1 plays a crucial role in regulating fuel metabolism (Li, 2013). During fasting or caloric restriction when NAD⁺ is elevated, hepatic SIRT1 deacetylates the transcriptional co‐activator PGC‐1α to drive PGC‐1α/PPARα‐dependent expression of genes controlling fatty acid oxidation (FAO), including Cpt1 and Fgf21 (Li et al., 2014; Rodgers et al., ²⁰⁰⁸). Conversely, in response to feeding, insulin triggers inhibition of SIRT1, switching the hepatic transcriptional output signature from FAO genes to the induction of lipogenic gene expression (Wang & Lin, 2021). To allow for these rapid changes in transcriptional programming, SIRT1 enzymatic activity can be acutely regulated through PTMs, PPIs, or a combination of both (Back et al., 2011; Chattopadhyay et al., ²⁰²⁰; Choi et al., ²⁰¹⁷; Conrad et al., ²⁰¹⁶; Kang et al., ²⁰¹¹; Kornberg et al., ²⁰¹⁰; Liu et al., ²⁰¹¹; Nasrin et al., ²⁰⁰⁹; Sasaki et al., ²⁰⁰⁸; Yang et al., ²⁰⁰⁷).

While roles for PTMs in SIRT1 regulation have been intensively investigated, a mechanistic understanding of SIRT1 regulation via PPIs is just beginning to emerge (Kim et al., 2007, 2008; Zhao et al., ²⁰⁰⁸). Deleted in breast cancer‐1 (DBC1) was originally identified as the first negative regulator of SIRT1 (Kim et al., 2008; Zhao et al., ²⁰⁰⁸). Based on a truncation analysis combined with peptide binding studies, it had been suggested that a region of DBC1, believed to be a leucine zipper, binds to the SIRT1 catalytic core and displaces the essential for sirtuin activity (ESA) region of SIRT1 from the catalytic core, inactivating the enzyme (Kang et al., 2011; Kim et al., ²⁰⁰⁸). However, other groups, including ours, have not been able to verify the “leucine zipper”/catalytic domain interaction (Hubbard et al., 2013; Krzysiak et al., ²⁰¹⁸; Li et al., ²⁰⁰⁹; Santos et al., ²⁰¹⁹). Instead, DBC1 primarily binds the SIRT1 N‐terminal region (SIRT1^1‐233, denoted here as SIRT1 NTR) through its S1‐like domain (DBC1^52‐120, denoted here as S1‐L) (Krzysiak et al., 2018). Furthermore, DBC1 does not directly inhibit SIRT1 activity; it facilitates the formation of the inhibitory SIRT1/phosphofurin acidic cluster sorting protein 2 (PACS‐2) complex in response to insulin signaling (Krzysiak et al., 2018). Binding of the DBC1 S1‐L to the SIRT1 NTR exposes a bipartite PACS‐2 interaction site, which includes a three‐helix bundle (SIRT1^183‐233, denoted here as 3HB) that is critical for SIRT1 enzymatic activity. PACS‐2 binding disrupts the folded structure of the 3HB and represses SIRT1 catalytic activity (Krzysiak et al., 2018). A schematic depiction of the domain organization of the three proteins involved in the DBC1, PACS‐2, and SIRT1 protein interaction hub is provided in Figure 1a.

FIGURE 1.

Domain structure of the players in the DBC1, PACS‐2, SIRT1 protein interaction hub and protein constructs used in the present study. (a) Amino acid numbering is for the human proteins and the position of the N‐ and C‐terminal amino acids in the subcomponents are shown by superscripts after the names. For SIRT1 truncations marked with *, the amino acids from K506‐N640 have been replaced with a GGGSGGGS linker as per Dai et al. (2015). The rounded rectangles represent domains or structured elements (gray = structural data verifying the proposed domain is not yet available: LZ = Leucine Zipper, EF = EF Hand, CC = coiled‐coil). Magenta circles represent relevant phosphorylation sites for this manuscript and yellow circles denote acetylation sites. Molecular masses of proteins are shown along the right. (b) Respiratory quotient (RQ = VCO₂/VO₂) expressed as mean during the light and dark cycles over a 24 h period. (c) RT‐qPCR of liver Cpt1α isolated from WT or Pacs2 ^LKO mice that were fed or fasted for 14 h and then re‐fed ad libitum for 4 h. Data are mean + SD. n = 4 mice per group. (d) Isolated WT and Pacs2 ^LKO primary hepatocytes were starved overnight and treated for 6 h with 10 μM WY‐14643 with or without 100 nM insulin. Cpt1α was measured by qRT‐PCR. Data are mean ± SD; n = 3. (e) Primary hepatocytes from WT and Pacs2 ^LKO mice were isolated and subjected to overnight starvation. They were then treated with the PPARα agonist WY‐14643 for 6 h followed by treatment with 10 μM Ex‐527 for 1 h. Fgf21 was measured by qRT‐PCR. Data are mean ± SD; n = 3.

An important piece of information missing from the DBC1‐assisted model of SIRT1 repression by PACS‐2 is a structural understanding of what role the DBC1 S1‐L/SIRT1 interaction plays. In order to provide such critical data, we use a combination of biophysical methods and mouse models of liver metabolism to delineate binding sites on both proteins and propose a mechanistic explanation for the effects of three PTMs: acetylation of K112 on DBC1, as well as phosphorylation of SIRT1 on S162 and S172 by CK2 and GSK3, respectively. We report the x‐ray structure of the DBC1 S1‐L domain as well as an NMR‐based structural characterization of the SIRT1 NTR and its interaction with DBC1 S1‐L. Further, using mouse models and biochemical assays in hepatocytes, we show that insulin triggers the CK2/GSK3‐dependent phosphorylation of SIRT1. The double‐phosphorylated SIRT1 is inhibited by DBC1/PACS‐2 and exported to the cytoplasm. A model describing this hormonally triggered, multi‐layered complex scenario of SIRT1 regulation is presented, as well as how dysregulation of this pathway is associated with obesity and non‐alcoholic fatty liver disease (NAFLD).

2. RESULTS

2.1. Insulin triggers DBC1/PACS‐2 repression of SIRT1 in liver

Considering the central role of hepatic SIRT1 in regulating transcriptional programs that control energy metabolism (Anton et al., 2018), we sought to validate our model of SIRT1 regulation by DBC1/PACS‐2 using liver‐specific Pacs2 knockout mice (Pacs2^LKO, see Figure S1A, B). We housed the mice in metabolic cages to assess the effect of liver Pacs2 loss on the source of metabolic substrate (fat versus carbohydrate) during the transition from the light cycle, when mice are fasted and utilize SIRT1/peroxisome proliferator‐activated receptor alpha (PPARα)‐dependent FAO as their primary fuel source, to the dark cycle, when insulin inhibits SIRT1/PPARα and stimulates the postprandial switch to carbohydrate as the primary fuel source (Figure 1b). Calculation of the respiratory exchange ratio (RER), or the ratio of expired CO₂ to O₂ consumption, which serves as a measure of fuel source, showed that the dark cycle RER was significantly reduced in Pacs2^LKO mice compared to WT mice, suggesting DBC1/PACS‐2 are critically important for postprandial insulin to inhibit SIRT1/PPARα transcriptional activity, and thus the switch from fat to glucose oxidation. Loss of liver PACS‐2 had no effect on RER during the light phase when mice used lipids as fuel or upon feeding or physical activity, suggesting that differences in RER were due to intrinsic changes in fuel selection by the liver (Figure S1C, D). These RER findings were supported by qPCR analysis of isolated hepatocytes, which showed that loss of Pacs2 disrupted the insulin‐dependent repression of SIRT1 target genes. This prevents Cpt1a inhibition and over‐inducing Fgf21, which was reversed by treatment with the small molecule SIRT1 inhibitor, EX‐527 (Figure 1c, e). Together, these findings support the model that insulin triggers liver DBC1/PACS‐2 to repress SIRT1 activity in vivo.

2.2. The DBC1 S1‐L domain contains a distinct SIRT1 binding site

For PACS‐2 to efficiently engage SIRT1 and repress deacetylase activity, the DBC1 S1‐L must first interact with SIRT1 (Krzysiak et al., 2018) (Figure 1a). The DBC1/SIRT1 interaction presumably disrupts intramolecular interactions centered in the SIRT1 NTR (Ghisays et al., 2015), and we previously demonstrated that a SIRT1 truncation lacking amino acids 6‐83 (SIRT1^Δ6‐83) was refractory to both DBC1 S1‐L and full‐length DBC1 binding (Krzysiak et al., 2018). We also demonstrated that the DBC1 S1‐L was necessary and sufficient for the interaction with the SIRT1 NTR in cell culture (Krzysiak et al., 2018). Therefore, to study the initiation of PACS‐2‐mediated SIRT1 inhibition, the interaction of the DBC1‐S1‐L with the SIRT1 NTR was examined using NMR spectroscopy. To this end, we employed uniformly (u)¹⁵N,¹³C‐labeled S1‐L domain and completed backbone resonance assignments (98% completeness). As can be appreciated, uniform intensity and well‐separated resonances are observed in the ¹H,¹⁵N HSQC spectrum of this domain (Figures 2a and S2A), indicative of a stably folded protein. Comparison of the ¹H,¹⁵N HSQC spectra of uniformly (u)¹⁵N‐labeled S1‐L in the absence and presence of natural abundance SIRT1 NTR revealed S1‐L resonances that are affected by SIRT1 NTR binding. The resulting chemical shift perturbations (CSPs) are associated with amino acids throughout the primary sequence (Figure 2a, b), and the largest CSPs comprise three clusters: F58‐G60, V96 and K97, and K112‐S117. K _d values extracted from the binding isotherms revealed that S1‐L binds weakly to the SIRT1 NTR (K _d = 240 ± 20 μM) (Figure 2c, Table 1).

FIGURE 2.

Identification of the SIRT1 binding region on the DBC1 S1‐L domain. (a) Superposition of ¹H,¹⁵N HSQC spectra of the DBC1 S1‐L in the presence (●) or absence (●) of 300 μM SIRT1 NTR. (b) Combined ¹H,¹⁵N CSPs for DBC1 S1‐L in the presence of 500 μM SIRT1 NTR. The line denotes the average CSP + 1 standard deviation. (c) Individual binding isotherms for six DBC1 S1‐L resonances that experience CSPs in the presence of the SIRT1 NTR. Global fitting yields a K _d value of 240 ± 20 μM for the interaction. (d) Ribbon representation of the DBC1 S1‐L x‐ray structure, with amino acid positions shown in magenta whose resonances exhibit significant CSPs in the presence of the SIRT1 NTR. The position of K112 is marked by a black dot.

TABLE 1.

Binding affinities between DBC1/SIRT1 measured by NMR.

Monitored	Titrant	Affinity
SIRT1183‐233	DBC1S1‐L	1000 ± 300 μM
SIRT1183‐233 M218A	DBC1S1‐L	340 ± 30 μM
SIRT1183‐233 T219A	DBC1S1‐L	320 ± 40 μM
SIRT1141‐233	DBC1S1‐L	330 ± 50 μM
SIRT1109‐233	DBC1S1‐L	310 ± 60 μM
SIRT11‐233	DBC1S1‐L	290 ± 40 μM
DBC1S1‐L	SIRT11‐233	240 ± 20 μM
DBC1S1‐L K112Q	SIRT11‐233	290 ± 20 μM
DBC1S1‐L	SIRT11‐233 S162D	220 ± 10 μM
DBC1S1‐L	SIRT11‐233 S172D	140 ± 10 μM
DBC1S1‐L	SIRT11‐233 S162D/S172D	80 ± 10 μM
DBC1S1‐L K112Q	SIRT11‐233 S162D/S172D	210 ± 20 μM
DBC1S1‐L	SIRT1141‐655*	420 ± 150 μM a
DBC1S1‐L	SIRT11‐655*	20 ± 10 μM a
DBC1S1‐L	SIRT1	8 ± 2 μM a

Note: *Same designation as Figure 1.

^a Denotes apparent affinity due to loss of signal.

To locate the interaction contacts on the 3D structure of the S1‐L domain, the crystallization of protein was pursued, and crystals belonging to the P2₁2₁2₁ space group were obtained. The x‐ray structure was solved at 2 Å resolution by molecular replacement using a Rosetta (Raman et al., 2009; Song et al., ²⁰¹³)‐derived search model, based on the NMR structure of another S1 domain, the first cold shock domain of the human upstream of N‐Ras protein (PDB ID 1WFQ; Goroncy et al., ²⁰¹⁰). The asymmetric unit contains two molecules that are essentially identical, with a pairwise all‐atom r.m.s.d. of 1.52 Å (Table S1). Small differences in the two molecules are seen in the conformation of the loop that connects β‐strands β4 and β5, with low electron density at its center residues ¹⁰¹NPG¹⁰³. Extensive loop structural variability has generally been described for all S1 domains (Deryusheva et al., 2019). In agreement with the well‐dispersed ¹H,¹⁵N HSQC spectrum, DBC1 S1‐L is an overall β‐sheet protein, closely resembling other S1 cold shock/RNA binding domains (Hollmann et al., 2020; Schnier et al., ¹⁹⁸²) (Figure 2d). Mapping residues whose amide resonances experienced the largest CSPs in the ¹H,¹⁵N HSQC spectrum of u¹⁵N‐labeled S1‐L upon SIRT1 NTR binding onto the S1‐L crystal structure revealed that the binding interface is confined to only one side of the β‐barrel, comprising the β1, β4, and β5 strands (Figure 2d).

2.3. Modulation of the DBC1/SIRT1 interaction by DBC1 K112 acetylation

Localization of K112 in the SIRT1 binding interface on the S1‐L was of particular interest because it is one of two acetylation sites, K112 and K215 (Zheng et al., 2013), in the N‐terminus of DBC1. Acetylation of K215 disrupts a nuclear localization signal, impeding nuclear entry of DBC1, and converting it into a SIRT1 substrate (Hubbard et al., 2013; Zheng et al., ²⁰¹³ and Figure S2B). In contrast, acetylation of K112 does not transform DBC1 into a SIRT1 substrate but reduces the DBC1/SIRT1 interaction as evidenced by co‐immunoprecipitation experiments (Hubbard et al., 2013; Zheng et al., ²⁰¹³) (also see Figure 3a inset). We hypothesized that acetylation of K112 might physically interfere with SIRT1 binding, and evaluated an acetylation mimetic, K112Q (S1‐L^K112Q) for SIRT1 binding. The ¹H,¹⁵N HSQC spectrum of this variant was not significantly changed from WT (Figure 3a), indicating that the 3D structure of the domain remained intact. Surprisingly, addition of natural abundance SIRT1 NTR to u¹⁵N‐labeled S1‐L^K112Q yielded a binding affinity comparable to that of WT DBC1 S1‐L (Table 1). However, the details of S1‐L^K112Q/SIRT1 NTR interaction were distinctly different from the WT S1‐L/SIRT1 NTR interaction: fewer CSPs were observed (Figure 3b), and the size of the CSPs was smaller (Figure S2C). For example, K97 and K112Q exhibited CSPs that are only 22% and 30% of the values seen for WT S1‐L/SIRT1 NTR binding and were not larger than the average observed CSP. This suggests that the equilibrium between the apo‐conformation and the bound conformation is shifted toward the apo‐conformation for the S1‐L^K112Q acetylation mimetic, implying that the S1‐L/SIRT1 NTR binding is mediated by conformational selection.

FIGURE 3.

The effect of the S1‐L^K112Q acetylation mimetic on SIRT1 binding. (a) Superposition of the ¹H,¹⁵N HSQC spectra of S1‐L (●) and S1‐L^K112Q (●) and co‐immunoprecipitation of SIRT1‐flag with ha‐tagged DBC1 or DBC1^K112Q (inset). (b) Superposition of the ¹H,¹⁵N HSQC spectra of S1‐L^K112Q in the presence (●) or absence (●) of 500 μM SIRT1 NTR. (c) Superposition of the ¹H,¹⁵N HSQC spectra of S1‐L in the presence (●) or absence (●) of 150 μM SIRT1^1‐655*. (d) Superposition of the ¹H,¹⁵N HSQC spectra of DBC1 S1‐L^K112Q in the presence (●) or absence (●) of 150 μM SIRT1^1‐655*.

A more pronounced effect of the K112Q acetylmimetic was detected using different length SIRT1 truncations. The affinity of S1‐L for SIRT1 increases with increasing size of SIRT1. The upper estimate for the affinity of S1‐L for full‐length SIRT1 (Table 1: K _d = 8 ± 2 μM) is an order of magnitude tighter than for the NTR alone. Enhanced affinity, compared to the NTR, is also observed for a protein segment that contains the SIRT1 NTR and the ESA joined to the catalytic domain by a flexible linker (SIRT1^{1‐655 *} see Figures 1a, S3A, S3B, and Table 1: K _d = 20 ± 10 μM). For the SIRT1 catalytic domain alone (SIRT1^{233‐655 *}), only very small CSPs were observed in the presence of S1‐L (Figure S3C), suggesting that binding competent conformers are enriched/stabilized in the conformational ensemble of the NTR by the addition of the catalytic domain. Addition of 150 μM natural abundance SIRT1^1‐655* to u¹⁵N‐labeled WT S1‐L broadens nearly all resonances beyond detection (Figure 3c), demonstrating that a large complex is formed. However, at this concentration, S1‐L^K112Q does not display a spectrum with broadened resonances in the presence of SIRT1^1‐655* (Figure 3d), indicating that S1‐L^K112Q, unlike WT S1‐L, is unable to engage in higher affinity interactions with larger SIRT1 polypeptides. Taken together, the above data suggest that acetylated S1‐L is less competent to interact with SIRT1.

2.4. Identification of the DBC1 binding region on SIRT1

To determine which regions in the SIRT1 NTR are involved in S1‐L binding, reciprocal NMR titration experiments were carried out. Specifically, a series of ¹H,¹⁵N HSQC spectra of u¹⁵N‐labeled SIRT1 NTR was recorded in the presence of increasing amounts of natural abundance S1‐L (Figure 4a). In good agreement with the above described S1‐L titration data, weak μM affinity was observed (Figure 4a inset, Table 1: K _d = 280 ± 40 μM) with CSPs detected in both the unstructured portion of the NTR (amino acids 1–182) as well as the structured 3HB region (amino acids 183–233, see Figure 4a–c). CSPs were identified for resonances belonging to amino acids A6‐A83, which are required for the S1‐L/SIRT1 interaction (Krzysiak et al., 2018). To determine whether any residues in this region are involved in direct contacts with S1‐L, ¹H,¹⁵N HSQC spectra of u¹⁵N‐labeled SIRT1^1‐124 were collected in the absence and presence of natural abundance S1‐L. Resonances corresponding to R39 and A73‐R77 exhibited CSPs upon addition of increasing amounts of S1‐L, suggesting that the region from A6‐A83 physically interacts with S1‐L (Figure 3d), although the affinity was too weak to permit the determination of a K _d.

FIGURE 4.

Mapping of DBC1 S1‐L binding on the SIRT1 NTR. (a) Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1 NTR in the presence (●) or absence (●) of 100 μM DBC1 S1‐L and individual binding isotherms for seven resonances. Global fitting yielded a K _d value of 290 ± 40 μM. (b) Combined ¹H,¹⁵N CSPs for the 3HB region of the SIRT1 NTR in the presence of 500 μM DBC1 S1‐L. (c) Amino acid sequence of the SIRT1 NTR with residues that exhibit amide resonance CSPs in the presence of S1‐L highlighted by magenta boxes. In (b) and (c), the individual helices of the 3HB are indicated by brown bars above the sequence. (d) Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1^1‐124 in the presence (●) or absence (●) of 500 μM DBC1 S1‐L. The enlarged region depicts resonances from A73‐R77. (e) Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1 3HB in the presence (●) or absence (●) of 500 μM DBC1 S1‐L. In the inset, individual binding isotherms for four resonances are shown. Global fitting yields a K _d value of 1000 ± 300 μM.

The most pronounced CSPs were observed for resonances associated with helix‐3 of the SIRT1 3HB (Figures 1a, 4a, b), a region present in the SIRT1^Δ6‐83 truncation that is refractory to S1‐L binding (Krzysiak et al., 2018). To determine whether the 3HB directly contacts S1‐L or is simply allosterically responsive to S1‐L binding at another site within the IDR spanning amino acids 6 to 83, u¹⁵N‐labeled SIRT1 3HB was used in titration experiments with natural abundance S1‐L. The ¹H,¹⁵N HSQC spectrum of the isolated 3HB exhibited CSPs in the presence of S1‐L, indicating a direct interaction between these two domains, although the binding affinity was ~four‐fold weaker (Figure 4e and Table 1). These data imply that S1‐L makes direct contact with at least two areas in the SIRT1 NTR, the structured 3HB as well as the IDR from A6‐A83.

In the IDR of the SIRT1 NTR, most of the resonances that exhibited CSPs in the presence of S1‐L are associated with amino acids E130‐T177 (Figure 4c). To dissect the contribution of this region to S1‐L binding, natural abundance S1‐L was employed in titrations with progressively longer u¹⁵N‐labeled SIRT1 polypeptides, beginning with the 3HB and extending N‐terminally into the IDR (Figure S4). ¹H,¹⁵N HSQC spectra‐derived binding isotherms revealed that the SIRT1^141‐233 polypeptide exhibited a comparable apparent affinity for S1‐L as seen for the entire NTR (Table 1). This suggests that the observed affinity of the S1‐L for the SIRT1 NTR is not additive between the A6‐A83 region and the 3HB. Instead, these data suggest that the 3HB primarily drives the physical interaction between the two polypeptides and the region spanning G141‐I182 enhances this primary S1‐L/3HB interaction.

Since the A6‐A83 segment of the SIRT1 NTR is responsive to DBC1 S1‐L binding (Figure 4d) and is required for interaction with full‐length SIRT1 (Krzysiak et al., 2018), we questioned whether this region is responsible for the 10‐fold higher affinity observed between the DBC1 S1‐L domain and SIRT1^{1‐655 *} or full‐length SIRT1 (Figure 3c and Table 1). Indeed, a version of SIRT1^{1‐655 *} lacking amino acids 1‐140 (SIRT1^{141‐655 *}) exhibits an affinity for S1‐L comparable to the NTR alone (Figure S4 and Table 1). Therefore, we posit that the SIRT1 NTR contains two regions that physically interact with the DBC1 S1‐L. The 3HB governs low affinity interactions that are modulated by the region from E130‐T177 while A6‐A83 drives higher affinity interactions but requires the presence of the catalytic domain to perform this function.

2.5. DBC1 modulates the conformational ensemble of the SIRT1 NTR

Although the SIRT1 NTR is an unstructured IDR, transient intramolecular interactions within the SIRT1 NTR are possible and have been reported to play an important role in SIRT1 regulation. In particular, the two halves (amino acids 1‐124 and 109‐233) of the SIRT1 NTR were suggested to interact, thereby shielding potential binding sites for repressive molecules (Ghisays et al., 2015). Since the DBC1 S1‐L is required to facilitate binding of a repressive molecule, PACS‐2 (Krzysiak et al., 2018), thereby inhibiting SIRT1 enzymatic activity (Krzysiak et al., 2018), we explored whether S1‐L binding affects how the SIRT1 NTR is presented to other interactors. Since chemical shifts are like fingerprints for a particular structure, we compared the ¹H,¹⁵N HSQC spectrum of the complete SIRT1 NTR (SIRT1^1‐233) and the corresponding spectra of the two halves of the NTR, SIRT1^1‐124, and SIRT1^109‐233 (Figure 5a). Any differences may provide clues as to whether potential interactions are present, even transiently. Resonances that exhibit different chemical shifts, compared to the isolated halves, were identified throughout the entire SIRT1 NTR spectrum, with the largest number of affected resonances associated with amino acids G30‐Q90 (Figure 5b). These CSPs were confirmed using u¹⁵N‐labeled SIRT1^1‐54 and SIRT1^1‐84 to which natural abundance SIRT1^109‐233 was added and vice versa (Figures 5c–e, and S5). Addition of SIRT1^109‐233 to u¹⁵N‐labeled SIRT1^1‐54 identified CSPs in a basic cluster from S26‐R46, and addition to the longer u¹⁵N‐labeled SIRT1^1‐84 revealed additional CSPs in the region spanning amino acids A71 to R77. The reciprocal experiment, adding SIRT1^1‐54 to u¹⁵N‐labeled SIRT1^109‐233, identified CSPs in the large acidic cluster sequence stretch, ¹²²DEDDDDEGEEEEE¹³⁴. Gratifyingly, all these regions align with those identified from the comparison of the ¹H,¹⁵N HSQC spectrum of the complete NTR with those of the individual halves. These data support previously proposed intramolecular interaction within the IDR region of the SIRT1 NTR (Ghisays et al., 2015) and allowed us to identify specific amino acids involved in it. Furthermore, residues that were identified as sensing the interaction within the SIRT1 NTR and those that are responsive to S1‐L binding possess A71‐R77 as a common stretch (compare Figures 4c and 5b), indicating that S1‐L binding influences the conformational ensemble of the IDR in the SIRT1 NTR. Together, these data indicate the presence of a network of weak, transient, but important interactions within the SIRT1 NTR and support the increasingly accepted notion that IDR regions in proteins can play important roles in PPIs.

FIGURE 5.

Intra‐molecular interactions in the unstructured region of the SIRT1 NTR mapped by NMR. (a) Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1 NTR (●) and that of SIRT1^1‐124 (●) (left) and SIRT1^109‐233 (●) (right). (b) Amino acid sequence of the SIRT1 NTR with residues whose amide resonance exhibit larger than average CSPs in the presence of SIRT1^1‐124 or SIRT1^109‐233 are highlighted by magenta boxes. Helices of the 3HB are indicated by brown bars above the sequence. (c) Combined amide CSPs of SIRT1^1‐54 in the presence of 500 μM SIRT1^109‐233. (d) Combined amide CSPs of SIRT1^1‐84 in the presence of 500 μM SIRT1^109‐233. (e) Combined amide CSPs of SIRT1^109‐233 in the presence of 500 μM SIRT1^1‐54.

2.6. Double phosphorylation by CK2 and GSK3β enhances the DBC1/SIRT1 interaction

In the unstructured portion of the SIRT1 NTR, many resonances that display CSPs in the presence of the S1‐L belong to amino acids neighboring S162 and S172 (Figure 4c), a region that enhances the S1‐L/SIRT1 interaction, based on our ¹H,¹⁵N HSQC titration studies (Table 1). Both serines have been reported to be phosphorylated, and their phosphorylation plays a role in the insulin‐stimulated, DBC1/PACS‐2‐mediated (pS162) or the obesity‐induced nuclear export‐mediated (pS172) models of SIRT1 enzymatic repression (Choi et al., 2017; Krzysiak et al., ²⁰¹⁸; Sasaki et al., ²⁰⁰⁸). Interestingly, insulin treatment of fasted mouse hepatocytes induces phosphorylation of S164 (equivalent to human pS172), suggesting that DBC1/PACS‐2‐mediated repression of SIRT1 may involve phosphorylation at both S162 and the obesity‐associated S172 in the SIRT1 NTR (Figure 6a). We therefore investigated whether singly phosphorylated or doubly phosphorylated SIRT1, mediated by CK2 (Patwardhan & Miller, 2007), impacts the DBC1/SIRT1 interaction. Single‐site phosphomimetic variants, SIRT1^{1‐233 S162D} or SIRT1^{1‐233 S172D}, or the double‐phosphomimetic, SIRT1^{1‐233 S162D/S172D} were titrated into u¹⁵N‐labeled S1‐L, and the interaction was assessed by ¹H,¹⁵N HSQC spectroscopy. The S162D substitution did not affect the affinity of S1‐L for the SIRT1 NTR, while the S172D substitution enhanced the affinity by 40%. Interestingly, the S162D/S172D double substitution synergistically enhanced the affinity by a factor of 3 (Table 1: compare SIRT1^1‐233 K _d = 240 ± 20 μM with SIRT1^{1‐233 S162D/S172D} K _d = 80 ± 10 μM). These observations were confirmed in the cellular context using HCT116 cells transfected with a plasmid expressing HA‐tagged DBC1 alone or together with plasmids expressing flag‐tagged wild‐type (WT) SIRT1, SIRT1^S162D/S172D (SIRT1 DD) or SIRT1^S162A/S172A (SIRT1 AA) (Figure 6b). IP of SIRT1 (flag) confirmed a weaker interaction with SIRT1 AA, compared to WT SIRT1 or SIRT1 DD.

FIGURE 6.

Phosphorylation of SIRT1 on S162 and S172 impacts the 3HB. (a) Primary WT mouse hepatocytes expressing SIRT1‐FLAG were starved overnight and then treated with or without 100 nM insulin for 30 min. FLAG‐tagged SIRT1 was immunoprecipitated and pS¹⁶⁴‐SIRT1 was detected by western blot. (b) Immunoprecipitation of DBC1‐ha, co‐expressed with flag‐tagged SIRT1, SIRT1^S162D/S172D or SIRT1^S162A/S172A in HCT116 cells detected by western blotting (HA). Data are average ± SD, n = 3. (c) Superposition of ¹H,¹⁵N HSQC spectra of u¹⁵N labeled SIRT1 NTR prior to (●) and following incubation (●) with 100 nM CK2 (top), 500 nM GSK3β (middle), or both CK2 and GSK3β (bottom) for 6 h. (d) Quantification of resonance intensities for S162 and S172. (e) Isolated mouse primary hepatocytes co‐expressing SIRT1‐flag and PACS‐2‐ha were starved for 16 h and then treated with 10 μM TBB or 0.5 μM CHIR98014 for 1 h followed by 10 nM insulin treatment for 30 min. SIRT1‐flag was immunoprecipitated and bound PACS‐2‐ha was detected by western blot. Data are average ± SD, n = 3. (f) Amino acid sequence of the SIRT1 NTR with residues whose amide resonance experience larger than average CSPs upon S‐D substitution or CK2/GSK3β phosphorylation highlighted by magenta boxes. Helices of the 3HB are indicated by brown bars above the sequence.

To verify that CK2 is capable of phosphorylating both S162 and S172, we directly monitored the in vitro phosphorylation of both serines in the SIRT1 NTR. After 6 h of treatment with recombinant CK2, the S162 amide resonance in the ¹H,¹⁵N HSQC spectrum disappeared and new resonances emerged, providing direct evidence for CK2 phosphorylation of S162 (Figure 6c top, d). Unexpectedly, the resonance corresponding to S172 remained unchanged, suggesting that another kinase, possibly acting downstream of CK2, may phosphorylate S172. Since CK2 can prime phosphorylation by GSK3β at nearby sites (Cohen & Frame, 2001), we repeated the experiment using recombinant GSK3β. Incubation of the u¹⁵N‐lableled SIRT1 NTR with GSK3 for 12 h did not affect the S162 amide resonance but reduced the intensity of the S172 amide resonance by ~27% (Figure 6c middle, d). When the SIRT1 NTR was incubated with both CK2 and GSK3β, the S172 amide resonance was reduced by ~36% after 6 h and ~ 64% after 12 h (Figure 6c bottom, d). The importance of SIRT1 phosphorylation by CK2/GSK3 for the interaction with DBC1/PACS‐2 and SIRT1 was confirmed in primary mouse hepatocytes. Insulin treatment of starved hepatocytes stimulated the mTORC2‐dependent phosphorylation of Akt at S473, as expected, and also induced phosphorylation of SIRT1 at S164 (equivalent to human S172 and see Krzysiak et al., ²⁰¹⁸; Betz et al., ²⁰¹³). By contrast, preincubation of the cells with either the CK2 inhibitor TBB or the GSK3 inhibitor CHIR98014 blocked the insulin‐dependent interaction between SIRT1 and PACS‐2 but, as expected, had no effect on Akt S473 phosphorylation (Figure 6e). Together, these data suggest that insulin‐dependent phosphorylation of S162 (mouse S154) by CK2 primes phosphorylation of S172 (mouse S164) by GSK3, which triggers interaction between SIRT1 and DBC1/PACS‐2.

2.7. Phosphorylation of S162/172 primes the SIRT1 3HB for interaction with the S1‐L

To clarify how SIRT1 phosphorylation at S162/S172 may impact S1‐L binding to the SIRT1 NTR, the ¹H,¹⁵N HSQC spectrum of WT SIRT1 NTR was compared with those of the S162A/S172A and S162D/S172D variants. For both substituted NTRs, CSPs were detected in resonances corresponding to amino acids surrounding the substituted residues. However, the phosphomimetic mutations affected the amide resonances of other amino acids in the unstructured parts of the NTR and in helix‐1 of the 3HB (Figure 6f), observed most prominently on the L192 amide resonance (Figure S6). A similar pattern of CSPs was observed between the phosphomimetic variant and the tandem CK2/GSK3 phosphorylated NTR, with the phosphorylated NTR displaying a more wide‐spread effect on the 3HB resonances (Figure 6f, middle, and bottom). Importantly, resonances experiencing CSPs due to the S‐D change or phosphorylation in the NTR are those that experience CSPs in the presence of the S1‐L (compare last 3 lines of sequence in Figures 4c and 6f), suggesting that PTM and S1‐L binding influence equivalent regions of the NTR.

In the x‐ray structures of human SIRT1 (4ZZH (Dai et al., 2015) and 5BTR (Cao et al., 2015)), the L192 side chain in the 3HB is in van der Waals contact with the M218 side chain (Figure 7a). In response to S1‐L binding, the M218 amide resonance displays the largest CSP (Figure 4b) of any SIRT1 NTR resonance, suggesting that the L192/M218 contact is altered. However, phosphomimicry/phosphorylation does not affect the M218 resonance, implying that the chemical modification of S162/S172, while enough to create a different chemical environment for the L192 amide group, does not inherently alter the L192/M218 interaction. Therefore, the effect of the modification likely primes the 3HB for a more efficient interaction with S1‐L. If a repositioning of side chains at the top of helix‐3 of the 3HB is important for S1‐L/SIRT1 complex formation, then other changes at these positions may also facilitate S1‐L/SIRT1 interaction. To test this hypothesis, we used an alanine variant at position 218. The 3HB with a M218A substitution binds stronger to the S1‐L, with an apparent affinity comparable to that of the complete SIRT1 NTR (Figure 7b top, Table 1). A similar finding was noted upon substituting another nearby, surface‐exposed amino acid, T219, for alanine (Figure 7b bottom, Table 1). Together, the alanine substitution data indicate that side chains at the top of helix‐3 of the 3HB are important for the interaction between the SIRT1 NTR and S1‐L. Addition of amino acids E151‐T177 to the 3HB increases the affinity for S1‐L by 4‐fold, with an additional increase in affinity by phosphorylation of this region (Table 1: compare SIRT1^183‐233, SIRT1^141‐233, and SIRT1^{1‐233 S162D S172D}). Thus, we posit that the region comprising amino acids E151‐T177, especially when phosphorylated, changes the important stabilizing contact L192/M218 in the 3HB and allows engagement of SIRT1 with the DBC1 S1L domain.

FIGURE 7.

Structure of the 3HB and the influence of alanine substitution on the DBC1 S1L binding. (a) Backbone ribbon representation of the 3HB in the x‐ray structure (PDB ID 4ZZH) (Dai et al., 2015). The side‐chains of L192 and M218 are shown in stick representation. (b) Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1 3HB, and (top) the M218A or (bottom) T219A variant in the presence (●) or absence (●) of 450 μM DBC1 S1‐L. Binding isotherms for individual resonances are shown in the insets. Global fitting yielded K _d values of 340 ± 30 μM and K _d = 320 ± 40 μM, respectively.

2.8. Insulin triggers DBC1/PACS‐2 to export CK2/GSK3‐phosphorylated SIRT1 to the cytoplasm

In mice challenged with a high‐fat diet, stable phosphorylation of SIRT1 at S164 (S172 in humans) sequesters the deacetylase to the cytoplasm, which underpins the obese phenotype and fatty liver (Kemper et al., 2013). As metabolic insulin signaling also induces phosphorylation of S172 (mouse S164) and the DBC1/PACS‐2‐mediated inhibition of SIRT1, we asked whether insulin‐dependent repression of SIRT1 activity similarly involves PACS‐2‐mediated export of the deacetylase to the cytoplasm. In support of this hypothesis, a 30‐min treatment with insulin caused SIRT1 to redistribute from the nucleus to the cytoplasm in hepatocytes from WT mice but not Pacs2^LKO mice (Figure 8a). Additional experiments with SIRT1 phosphorylation‐state mutants suggested that the CK2/GSK3 modified NTR was required for the PACS‐2‐dependent export of SIRT1 from the nucleus; adenovirus‐expressed phosphomimic mSIRT1^S164D localized in the cytoplasm of WT hepatocytes but remained in the nucleus in Pacs2^LKO hepatocytes (Figure 8b). By contrast, non‐phosphorylatable mSIRT1^S164A remained in the nucleus in both WT and Pacs2^LKO hepatocytes. As expected, mSIRT1 also concentrated in the nucleus, independent of Pacs2 status, since the cells were not exposed to insulin to trigger CK2/GSK3 phosphorylation at S154/S164. Together, these results suggest that insulin directs transcriptional reprogramming in liver via triggering the CK2‐ and GSK3‐dependent phosphorylation of chromatin‐associated SIRT1. In turn, DBC1 binds and remodels phosphorylated SIRT1, which permits PACS‐2 to engage the deacetylase and sequester it in the cytoplasm (Figure 8c, d).

FIGURE 8.

PACS‐2 facilitates SIRT1 cytoplasmic shuttling. (a) Primary WT and Pacs2 ^LKO hepatocytes were left untreated (fed) or serum‐starved overnight and then treated with or without 100 nM insulin for 30 min. Cells were fractionated into cytosolic and nuclear fractions and the nucleocytoplasmic distribution of SIRT1 was determined by western blotting. (b) Primary WT and Pacs2 ^LKO hepatocytes were transduced with adenoviruses expressing WT SIRT1, SIRT1^S164D or SIRT1^S164A. After 24 h, the cells were fractionated into cytosolic and nuclear fractions and the nucleocytoplasmic distribution of SIRT1 was determined by western blotting. Data are mean + SD. n = 3. (c) Active SIRT1 is primed for interaction with the DBC1 S1‐L by the sequential phosphorylation at S162 and S172 by CK2 and GSK3β, respectively. DBC1 binding alters the SIRT1 NTR, priming SIRT1 for interaction with PACS‐2 without disrupting the intramolecular communication network within the N‐terminus. PACS‐2 binding disrupts the 3HB and inactivates SIRT1. (d) Acetylation of K215 on DBC1 promotes SIRT1's nuclear retention (Hubbard et al., 2013; Zheng et al., ²⁰¹³). Two phosphorylation events prime SIRT1 for interaction with DBC1. In response to metabolic cues, triggered by insulin, phosphorylation of hS162/mS154 by CK2 leads to phosphorylation on hS172/mS164 by GSK3β, enhancing the DBC1 S1‐L interaction with the SIRT1 NTR. PACS‐2, although predominantly cytoplasmic, shuttles between the cytoplasm and the nucleus (Atkins et al., 2014), requiring phosphorylation on S437 to interact with SIRT1 (Atkins et al., 2014). Following DBC1‐mediated exposure of the PACS‐2 bipartite binding site on SIRT1, PACS‐2^pS437 engages SIRT1 and disrupts the structure of the 3HB (Krzysiak et al., 2018) thereby lowering SIRT1 enzymatic activity. Subsequently, the complex containing PACS‐2/SIRT1 translocates from the nucleus to the cytoplasm.

3. DISCUSSION

Early studies identified DBC1 as a negative regulator of SIRT1 activity (Kim et al., 2008; Kokkola et al., ²⁰¹⁴). In contrast to the initial assumption that SIRT1 repression is directly caused by DBC1 binding, involving a putative “leucine zipper” region of DBC1 (Kang et al., 2011; Kim et al., ²⁰⁰⁸), we showed that the formation of a DBC1/SIRT1 complex enables the multifunctional sorting protein PACS‐2 to engage SIRT1, by conformational selection, i.e. binding to destabilized/ ill‐structured binding competent conformers of the 3HB region in the NTR of SIRT1, thereby causing inhibition of the enzyme (Krzysiak et al., 2018). This finding is confirmed and further detailed in the studies presented here. Our structural mapping data clearly delineates the region on the DBC1 S1‐L domain involved in the DBC1/SIRT1 interaction, namely a single face of the β‐barrel where K112 resides (Figure 2d). This structural data agrees well with previous work that implicated acetylation of DBC1 at residue K112 in SIRT1 activity (Zheng et al., 2013). Since any experimental evidence for the presence of a leucine zipper structure is lacking or that this putative leucine zipper region directly contacts SIRT1, our data support the involvement of K112 acetylation, since the DBC1/SIRT1 binding interface will be disrupted by this modification. Further experimental evidence for this notion was obtained by ¹H,¹⁵N HSQC spectroscopy using the phosphorylated SIRT1 NTR as well as a longer SIRT1 protein construct, comprising the NTR and the catalytic domain with a fused ESA region (Table 1, Figure 3c, d). Thus, our data shows unambiguously that the DBC1/SIRT1 interaction is primarily driven by the DBC1 S1‐L domain, although further involvement of other regions in SIRT1 may also contribute to binding.

The effect of DBC1 S1‐L binding is felt throughout the SIRT1 NTR since this region is essentially an unstructured IDR. It is known that IDRs can retain conformational disorder, even in the bound state, which has been proposed to be favorable since it allows for easy regulation of the interaction (Tompa & Fuxreiter, 2008). Using a divide and conquer approach, we were able to identify two regions, A6‐A83 and the 3HB, that are directly influenced by DBC1 S1‐L binding (Figure 4). We also showed that phosphorylation of S162/S172 facilitates DBC1 S1‐L binding (Figure 6b), most likely by influencing the conformational ensemble of the unstructured region of the NTR to drive interaction between the DBC1 S1‐L and SIRT1 3HB. Our data provide direct evidence for a multimodal interaction between these two proteins (Hubbard et al., 2013; Zheng et al., ²⁰¹³). In our ¹H,¹⁵N HSQC analysis of the DBC1 S1‐L/SIRT1 NTR communication, we identified a weak ~250 μM interaction driven by the 3HB (Figure 2c, Table 1) which is not affected by the S1‐L^K112Q acetyl‐mimetic (Figure 3, Table 1). When DBC1 K112 is acetylated, SIRT1 may utilize acetylated K215 (Hubbard et al., 2013; Zheng et al., ²⁰¹³) as a substrate without activating a repressive mechanism, while in the non‐acetylated state, the more stable complex may possibly exhibit a conformation that enhances binding to PACS‐2 and thereby facilitate repression. This is supported by our data of DBC1 binding to SIRT1 polypeptides, minimally comprising the SIRT1 NTR and the catalytic domain/ESA region, which resulted in an ~10‐fold tighter interaction (Figure 3, Table 1). This requires the unstructured A6‐A83 region (Figure S4, Table 1) and is no longer seen with the S1‐L^K112Q acetylmimetic (Figure 3, Table 1). Loss of the higher affinity interaction explains the inability of SIRT1^Δ6‐83 to interact with DBC1 as well as why acetylation of K112 was found to disrupt interaction with SIRT1 in cell culture (Hubbard et al., 2013; Zheng et al., ²⁰¹³).

We also provide direct NMR‐derived experimental data for the earlier proposal that in an active SIRT1 molecule, the NTR is folded back upon itself (Ghisays et al., 2015). A network of transient but functionally important interactions within the unstructured component of the SIRT1 NTR was identified by comparing chemical shifts of individual halves of the SIRT1 NTR with the intact SIRT1 NTR (Figure 5). Five regions with resonances exhibiting different chemical shifts are observed: A2‐L9, S26‐S47, A65‐A86, Y121‐E134, and the 3HB. The S26‐S47 region contains the basic sequence ³⁴RKRPRR³⁹, interacting with the acidic sequence ¹²²DEDDDDEGEEEEE¹³⁴ via electrostatic contacts. The A71‐R77 region engages in intramolecular interactions within the SIRT1 NTR as well as intermolecularly with the DBC1 S1‐L domain (Figures 4c and 5b). These interactions contribute to the proposed shield within the SIRT1 NTR that interferes with PACS‐2 binding (Krzysiak et al., 2018).

From a broader biological perspective, our data reveal the complex interplay between members of the DBC1/PACS‐2/SIRT1 metabolic hub that modulates liver gene expression. Under fasting conditions, nuclear SIRT1 deacetylates PGC‐1α (Villar‐Pazos et al., 2023), which drives PGC‐1α/PPARα‐induction of FAO gene expression. Under fed conditions, insulin triggers both the CK2/GSK3‐dependent phosphorylation of SIRT1 at S162 and S172, and the mTORC2/Akt‐dependent phosphorylation of PACS‐2 at S437 (Aslan et al., 2009; Atkins et al., ²⁰¹⁴; Krzysiak et al., ²⁰¹⁸); the two signaling pathways converge to promote the DBC1/pS437‐PACS‐2‐dependent inhibition and cytoplasmic sequestration of ppS162,172‐SIRT1, which promote the switch in metabolic fuel source from fat to glucose oxidation. Our study supports the emerging role of the PACS family members, PACS‐1 and PACS‐2, as key in vivo regulators of histone deacetylases. In the nucleus, PACS‐2 binds SIRT1 to modulate acetylation of p53 and PGC‐1α, whereas PACS‐1 binds HDAC2 and HDAC3 to support genome stability (Atkins et al., 2014; Krzysiak et al., ²⁰¹⁸; Mani et al., ²⁰²⁰). Cytoplasmic PACS‐1 modulates HDAC6 activity to control microtubule dynamics, and a recurrent disease mutation in PACS1 dysregulates HDAC6, leading to neural deficits and intellectual disability (Villar‐Pazos et al., 2023).

Our determination that the insulin‐stimulated, DBC1/PACS‐2‐dependent inhibition of SIRT1 requires the sequential phosphorylation of the SIRT1 NTR, first by CK2 at S162 followed by GSK3 at S172, is consistent with earlier findings that GSK3 kinases frequently require a priming phosphorylation of their target substrate at a nearby site by a separate kinase, such as CK2 (Doble & Woodgett, 2003; Picton et al., ¹⁹⁸²). GSK3 kinases are emerging as key regulators of energy metabolism beyond glycogen storage. For example, GSK3β phosphorylates PGC‐1α, targeting the transcriptional co‐activator for proteasomal degradation (Anderson et al., 2008). In this way, the opposing actions of SIRT1 (deacetylation) and GSK3β (phosphorylation) on PGC‐1α regulate the speed and duration of PCG‐1α‐dependent gene expression. Interestingly, GSK3α‐dependent phosphorylation of PPARα switches transcriptional output from genes directing FAO to transcriptional programs driving fat accumulation and obesity‐associated lipotoxicity (Nakamura et al., 2019). This finding is consistent with our earlier report that persistent phosphorylation at SIRT1 S172 is associated with NAFLD and directly correlates with disease severity (Choi et al., 2017). Identification of the GSK3 isoform that phosphorylates SIRT1 S172, as well as the phosphatase that dephosphorylates this site to reactivate SIRT1, will provide important insight into the cellular machinery that mediates insulin‐ and DBC1/PACS‐2 controlled liver homeostasis and how dysregulation of this metabolic hub causes disease.

4. CONCLUSIONS

We determined the x‐ray structure of the DBC1 S1‐L domain, identified the binding site for SIRT1 on this domain, and delineated specific regions in SIRT1 that are important in the DBC1/PACS‐2 cooperative repression of SIRT1 enzymatic activity. Multiple PTMs modulate the specifics of this protein–protein hub, especially phosphorylation of SIRT1 by CK2 and GSK3, and their respective contributions were validated in mouse models and hepatocyte cultures. As such, this regulatory hub is another example for the involvement of IDRs in binding and transfer between protein partners. Overall, the underlying mechanistic understanding provided by our data further establishes the DBC1/PACS‐2 axis as a major regulator of SIRT1 activity and may offer guidance for designing therapeutic interventions for metabolic disorders, including obesity and NAFLD.

5. MATERIALS METHODS

5.1. DBC1/SIRT1 cloning and protein purification

The cloning and protein expression of SIRT1^183‐233, SIRT1^141‐233, SIRT1^1‐233, and DBC1^52‐120 has been previously described (Krzysiak et al., 2018). Proteins unique to this study, SIRT1^1‐54, SIRT1^1‐84, SIRT1^1‐124, and SIRT1^109‐133 were cloned using Gibson assembly into the same modified pET41 vector and expressed and purified in the same manner. All mutant proteins were generated using non‐overlapping site‐directed mutagenesis. Briefly, all DNA constructs were used to transform our custom E. coli cell line termed Rosetta* (Krzysiak et al., 2018). For preparing isotopically labeled samples, cells were grown at 37°C in modified M9 medium, supplemented with ¹⁵N‐ammonium chloride or ¹⁵N‐ammonium chloride and ¹³C‐glucose to an OD₆₀₀ = 0.6, induced with 400 μM IPTG and grown overnight at 18°C for 16–18 h. Cells were harvested by centrifugation (4600 × g; 10 min; 4°C) and resuspended in 20 mM HEPES pH 7.0, 500 mM NaCl, and 5 mM DTT for lysis using a microfluidizer (Microfluidics). DNase (80 μg/mL) and RNase (64 μg/mL) were added to the lysate and the reaction was incubated at 4°C with stirring for 30 min. The lysate was clarified by centrifugation (38,000 × g/30 min/4°C) and applied to a GSTrap column (Cytiva). Bound protein was eluted with 20 mM HEPES pH 8.0, 500 mM NaCl, 40 mM reduced glutathione, and 0.5 mM TCEP. GST‐tagged protein was separated from any contaminating proteins by gel filtration. The protein was digested with TEV protease and cleaned up further using anion exchange followed by a final gel filtration step. In the case of SIRT1^1‐54, which does not contain any amino acids contributing to an extinction coefficient, following TEV treatment, the digested protein was separated from both TEV and GST using IMAC (5 mL HisTrap column Cytiva). The protein remains in the flowthrough. Flowthrough was concentrated using Centriprep YM3 concentrators (Millipore Sigma) and subjected to gel filtration, Superdex 75 16/60 column (GE Life Sciences). The elution volume was found to be between 76 and 89 mL using SDS‐PAGE. To determine the concentration, a standard curve of the integrated peak volume at A₂₁₀ was generated by injecting known concentrations of a peptide, SMAEVMQHPSEGS (Genscript), over a Superdex 75 10/300 column. At least three injection volumes of an unknown SIRT1^1‐54 sample were averaged to determine the sample concentration.

SIRT1 was cloned as previously described by Atkins et al. (2014) and was purified as previously described in Krzysiak et al. (2018). using IMAC following the same general procedure described above for the GST‐tagged proteins.

SIRT1^1‐655*, SIRT1^233‐655*, and SIRT1^141‐655* were cloned into a pET41 vector with substantial modifications. The GST has been replaced with a N‐terminally 6xhis‐tagged thioredoxin from P. furiosis that has had all naturally occurring cysteines converted to serines (Canali et al., 2014). The thioredoxin is connected to the target protein through a GSTSGSGTS linker followed by a TEV site. Purification of the thioredoxin‐tagged proteins followed the SIRT1 scheme of IMAC, gel filtration, and TEV cleavage, but finished with the negative IMAC procedure of SIRT1^1‐54 followed by gel filtration.

For all proteins, the final gel filtration buffer was that of the reported conditions for a particular NMR experiment. Final protein purity was estimated at >98% by SDS‐PAGE.

5.2. Kinase cloning and protein purification

For the CK2 holoenzyme, human CK2α 1‐335 and CK2β D55A/E57A were separately cloned into a NdeI/XhoI linearized pET41a backbone using Gibson assembly. Clonal transformation, cell growth, protein expression, and cell lysis followed using the same procedure as listed for DBC1/SIRT1 proteins. Cells expressing each subunit were grown separately and equal amounts of each cell pellet were combined 1:1 prior to lysis. The cell lysate was passed over a HiTrap Heparin HP column (Cytiva) equilibrate in 20 mM HEPES pH 7.5/0.5 mM TCEP. Bound protein was eluted using a linear gradient from 400 mM‐1 M NaCl. Eluted protein was further purified and buffer exchanged using a Superdex 200 26/60 gel filtration column (Cytiva) equilibrated in 20 mM HEPES pH 7.0/100 mM /0.5 mM TCEP.

Human GSK3β 26‐420 was cloned into a pET41a backbone where the GST was replaced with a 6xHistag separated from the cloning site by a TEV cleavage sequence. Clonal transformation, cell growth, protein expression, and cell lysis were followed using the same procedure as listed for DBC1/SIRT1 proteins. Protein purification followed the same procedure listed above for SIRT1^1‐655*. Mass spec analysis indicated that GSK3β was purified as a mixture of unphosphorylated, singly phosphorylated, and doubly phosphorylated species.

5.3. Protein crystallization

DBC1^52‐120 was concentrated to 500 μM in HEPES pH 7.5, 100 mM NaCl, 0.5 mM TCEP and mixed 1:1 with mother liquor using the microbatch technique using a 70:30 mix of light paraffin oil (EMD PX0047‐1) and polydimethylsiloxane (Millipore Sigma 46,319). Crystals used for structure determination were grown under oil on Nunclan Delta Surface plates (Thermo Fisher 136,528) by replicating the microbatch conditions using 0.1 M sodium acetate pH 4.6/2 M ammonium sulfate (Molecular Dimensions JCSG Screen 1 #35). Crystals were cryoprotected by soaking for a few seconds in the crystallization solution supplemented with 35% glycerol prior to flash‐cooling (−180°C).

Diffraction data were collected in‐house using a Rigaku FR‐E super bright rotating anode equipped with a Raxis HTC detector. The structure was solved via molecular replacement utilizing a Rosetta (Raman et al., 2009; Song et al., ²⁰¹³) derived model of a S1 domain as the structural probe in PHASER (McCoy et al., 2007). The structure was iteratively refined by alternating between manual rebuilding of the structure in COOT (Emsley & Cowtan, 2004) and structural refinement with the PHENIX software package (Adams et al., 2010). There were two molecules in the asymmetric unit. Both molecules lacked density for amino acids 52‐55 but residues 118‐120 were also missing in the A subunit and there was reduced electron density around ¹⁰⁰NPG¹⁰³. MOLPROBITY (Davis et al., 2007) evaluation indicates that all residues in the electron density lie in the favored and allowed regions of the Ramachandran plot. The structural statistics can be found in Supplemental Table 1. PYMOL was used to generate all structural figures (Schrodinger, 2015).

5.4. NMR spectroscopy

All 2D ¹H,¹⁵N HSQC spectra of proteins were recorded at 25°C or 37°C using Bruker AVANCE 800 MHz or 900 MHz NMR spectrometers equipped with a z‐axis gradient cryoprobes in 20 mM HEPES pH 7.0 or 7.5, 100 mM NaCl, 0.5 mM TCEP. For titration experiments, either 30 or 50 μM of u¹⁵N‐labeled protein was combined with increasing amounts of unlabeled titrant as individual samples. Combined chemical shift changes were calculated according to the equation:

$${\mathrm{\Delta }}^1{H}^{15}\mathrm{N}\text{chemical shift}=\surd \left(\mathrm{\Delta }{H}^2+{\left(\mathrm{\Delta }N/6.51\right)}^2\right),$$

where ΔH and ΔN represent the change in the proton or nitrogen chemical shifts, respectively. The K _d was determined by plotting the Δ¹H¹⁵N chemical shifts as a function of the natural abundance titrant and fitting the data using the quadratic equation. Specifically, in the cases of u¹⁵N‐labeled S1‐L titrated with either SIRT1^1‐655* or SIRT1, the titration resulted in peak broadening and not the detection of CSPs. As such, an apparent ‡K _d was determined by fitting the decrease in the total peak height of all S1‐L backbone amide resonances as a function of natural abundance titrant to the equation:

$$\mathrm{SPH}=1-(0.{\text{5A}}^{*}\left(\left(30+x+‡K_{\mathrm{d}}\right)-\left(\left(\surd \left[{\left(30+x+‡K_{\mathrm{d}}\right)}^2-\left(4^{*}{30}^{*}x\right)\right]\right)\right)\right),$$

where the remaining percentage of the starting total backbone amide peak height (SPH) is represented by an adjusted form of the quadratic equation where (A) is the final amplitude of the combined amide peak heights, (x) represents titrant concentration, (‡K _d) is the apparent dissociation constant, and (30) represents the μM concentration of u¹⁵N‐labeled S1‐L used in the experiments.

Backbone chemical shift assignments for SIRT1^1‐54, SIRT1^1‐84, SIRT1^1‐124, SIRT1^109‐233, SIRT1^141‐233, and DBC1^52‐120 were obtained using standard 2D HSQC and 3D HNCACB, HN(CO)CACB, HNCA, HN(CO)CA, HNCO and HN(CA)CO experiments, recorded on a Bruker AVANCE 800 or 900 MHz NMR spectrometers equipped with a z‐axis gradient cryoprobe at 25°C. Non‐Uniform Sampling was used during data collection using 20% of the data points and 32, 16, 16, 16, 8, and 32 scans, respectively. Spectra were reconstructed using the SMILE algorithm (Ying et al., 2017), processed using NMRPipe (Delaglio et al., 1995), and analyzed using CARA (Keller, 2004) and NMRFAM‐Sparky (Lee et al., 2015). The chemical shifts for the backbone atoms in SIRT1^1‐54, SIRT1^1‐84, SIRT1^1‐124, SIRT1^109‐233, and SIRT1^141‐233 have been deposited in the BMRB. The SIRT1 NTR ¹H,¹⁵N HSQC spectrum was assigned through compilation of the backbone assignments obtained for SIRT1^1‐54, SIRT1^1‐84, SIRT1^1‐124, SIRT1^109‐233, SIRT1^141‐233, and SIRT1^183‐233 (Figure S7).

5.5. Cells, animals and chemicals

HCT116 cells were cultured in DMEM containing 10% fetal bovine serum (FBS). Mouse primary hepatocytes were isolated from 10‐to 12‐week‐old male C57BL/6 WT and Pacs2^LKO mice by two‐step perfusion with calcium and magnesium‐free Hanks' salt solution followed by a Leibovitz's L‐15 medium (Thermo Fisher, #11415064) containing Liberase™ (Roche, #5401127001). Hepatocytes were plated in William E (Thermo Fisher, #32551) containing 10% FBS, dexamethasone, and insulin, and adhered for 4 h on a collagen‐coated dish (Thermo Fisher, CB40236). After adhesion, cells were incubated in William E containing 0.2% FBS before the indicated treatments. Chemicals were purchased as follows: Insulin (Sigma‐Aldrich, I0516), TBB (Selleckchem, S5265), CHIR 98014 (Tocris, 6695), Ex‐527 (Tocris, 2780), WY‐14643 (Sigma‐Aldrich, C7081).

5.6. Experimental animals and metabolic studies

The Institutional Animal Care and Use Committee approved all animal studies. All mice used in this study were on the C57BL/6J background. Pacs‐2^flox6/7 mice (C57BL/6NTac‐Pacs2tm1a(EUCOMM)Hmgu/IcsOrl) were obtained from INFRAFRONTIER European Mutant Mouse Archive (EMMA). C57BL/6J, FLPo (Gt(ROSA)26Sortm2(FLP*)Sor), and Alb‐Cre (B6.Cg‐Speer6‐ps1 ^{Tg(Alb‐cre)21Mgn} /J) mice were purchased from Jackson labs. Homozygous Pacs‐2^flox6/7 mice were crossed with FLPo mice, which removed the lacZ and Neo cassettes and restored Pacs‐2 expression. Homozygous offspring were then crossed with Alb‐Cre mice to generate Pacs2 ^LKO mice. Male C57BL/6 WT and Pacs2 ^LKO mice were allowed ad libitum access to standard chow diet and water. They were kept in temperature‐controlled, filter‐sterilized animal quarters under a 12 h light:12 h dark cycle. For fasting and re‐feeding studies, mice were randomly assigned under the following conditions: ad libitum fed, after a 16 h fasting, or after a 16 h fasting followed by re‐feeding for 4 h. Metabolic cage studies were performed using Sable Systems Promethion Multiplexed Metabolic cage system in order to measure activity, feeding, and oxygen consumption (VO₂) and carbon dioxide production (VCO₂) for indirect calorimetry and calculation of energy expenditure and the RER. RER is the ratio of VCO₂ to VO₂ and serves as an index of relative whole‐body metabolic substrate use, where a value of 1.0 indicates carbohydrate oxidation and 0.7 indicates lipid oxidation. Mice were housed in the metabolic cage system for 72 h where the first 24 h was considered acclimation and the subsequent 48 h was used for analysis.

5.7. Adenoviral infection, plasmid transfection, and insulin treatment

Mouse primary hepatocytes were infected with adenoviruses expressing WT mSIRT1, mSIRT1^S164A, or mSIRT1^S164D for 24 h (Choi et al., 2017). According to the manufacturer's protocol, the transient transfection with plasmids expressing WT hSIRT1‐flag, hSIRT1^S162D/S172D‐flag, hSIRT1^S162A/S172A‐flag, DBC1‐ha, DBC1^K112Q‐ha and PACS‐2‐ha was performed in HCT116 cells and hepatocytes by using the Lipofectamine 2000 (Invitrogen, 11,668,019). Insulin stimulation was conducted by subjecting HCT116 cells and hepatocytes to 16 h of starvation, followed by treatment with 10 nM insulin for 30 min. To examine the PGC‐1α/PPARα‐dependent induction of Fgf21 after insulin stimulation, hepatocytes were starved overnight and subsequently treated with 10 μM Ex‐527 for 1 h. This was followed by a 6 h treatment with 10 μM WY‐14643, both with and without 100 nM insulin. Furthermore, to explore the CK2 and GSK3β‐dependent PACS‐2/SIRT1 interaction, hepatocytes were starved for 16 h and treated with 10 μM TBB or 0.5 μM CHIR98014 for 1 h. Subsequently, the cells were subjected to 10 nM insulin treatment for 30 min.

5.8. Immunoprecipitation and nuclear fractionation

For immunoprecipitation, HCT116 cells and hepatocytes were harvested in lysis buffer (50 mM Tris HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 1% NP‐40, 10% glycerol) freshly supplemented with protease inhibitors (0.5 mM PMSF, 0.1 mM aprotinin, E‐64 and leupeptin) and phosphatase inhibitors (1 mM Na₃VO₄ and 20 mM NaF). For nuclear fractionation, hepatocytes were lysed in buffer A (50 mM Tris‐HCl pH 7.9, 10 mM KCl, 1 mM EDTA, 0.2% NP‐40, 10% glycerol) and centrifuged at 6000 rpm for 3 min at 4°C. Pellet was washed with buffer A and lysed with buffer B (400 mM NaCl, 1% NP‐40, 20% glycerol, 20 mM HEPES pH 7.9, 10 mM KCl, 1 mM EDTA) for 20 min at 4°C. The insoluble and soluble nuclear fractions were separated by centrifugation at 14,000 rpm for 10 min. Western blots were probed with the following antibodies: actin (Millipore, MAB1501), p‐Akt (Cell Signaling Technology, #9271), flag (Sigma‐Aldrich, F7425), HA (Cell Signaling Technology, #3724), Lamin A/C (Cell Signaling Technology, #4777), PACS‐2 (Simmen et al., 2005), p‐SIRT1 at S164 (Choi et al., 2017), SIRT1 (Cell Signaling Technology, #2028), α/β tubulin (Cell Signaling Technology, #2148).

5.9. RNA isolation and quantitative real‐time PCR (qPCR)

RNA was isolated from the liver using RNeasy (QIAGEN, #74134). RNA was reverse transcribed using the SuperScript IV First‐Strand cDNA Synthesis Kit (Invitrogen, # 18091050). qPCR was performed in a StepOne Real‐Time PCR System with the Power SYBR Green PCR Master Mix (Applied Biosystems, #4368706). The primer sequences are as follows: Fgf21 forward 5′‐CCTCTAGGTTTCTTTGCCAACAG‐3′; Fgf21 reverse 5′‐AAGCTGCAGGCCTCAGGAT‐3′; Cpt1α forward 5′‐GCAGTCGACTCACCTTTCCT‐3′; Cpt1α reverse 5′‐ATTTCTCAAAGTCAAACAGTTCCA‐3′; Cyclophilin forward 5′‐GGAGATGGCACAGGAGGAA‐3′; Cyclophilin reverse 5′‐GCCCGTAGTGCTTCAGCTT‐3′. Melting curve analysis determined the specificity of amplification. Relative changes in gene expression were expressed as the fold change using the 2^−ΔΔCT . Gene expression was normalized to the expression of the control cyclophilin gene.

AUTHOR CONTRIBUTIONS

Troy C. Krzysiak: Conceptualization; writing – original draft; investigation; writing – review and editing. You‐Jin Choi: Conceptualization; investigation; writing – review and editing. Yong Joon Kim: Investigation. Yunhan Yang: Investigation. Christopher DeHaven: Investigation. Lariah Thompson: Investigation. Ryan Ponticelli: Investigation. Mara M. Mermigos: Investigation. Laurel Thomas: Investigation. Andrea Marquez: Investigation. Ian Sipula: Investigation. Jongsook Kim Kemper: Resources. Michael Jurczak: Supervision; funding acquisition; conceptualization; writing – review and editing. Gary Thomas: Conceptualization; funding acquisition; writing – review and editing; supervision; project administration. Angela M. Gronenborn: Conceptualization; funding acquisition; writing – review and editing; project administration; supervision.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing financial interests.

Supporting information

Figure S1: Characterization of PACS‐2 liver‐specific knockout mice. (A) Left: Schematic of the PACS‐2 liver‐specific knockout mouse. PACS‐2^flox6/7 mice were crossed with Alb‐cre mice to generate PACS‐2 liver‐specific knockout mice (PACS2^LKO). Right: PCR analysis of ear or liver genomic DNA using Ef/Er primer to identify offspring carrying floxed Pacs2 as determined by the presence of the 654 bp PCR product. PACS2^LKO mice were identified by PCR of genomic DNA with primers Ef/L3r, which amplify a 524 bp PCR product, demonstrating deletion of Pacs2 exons 6 and 7 specifically in liver. (B) Western blot of PACS‐2 expression in liver or brain isolated from WT or PACS2^LKO mice. (C) Feeding data expressed as the total grams of food consumed per kg lean mass during the light and dark cycles or total time spent in the metabolic cages. (D) Total activity expressed as meters traveled or moved during the light and dark cycles or total time spent in the metabolic cages. Data are the mean ± SEM for n = 6 per group. Data were compared by student's t‐test.

Figure S2: DBC1 S1‐L and SIRT1 NTR induced S1‐L^K112Q CSPs. (A) ¹H,¹⁵N HSQC spectrum of DBC1‐S1‐L. Resonance assignments are given by residue name and number. (B) Confocal analysis of GFP‐tagged DBC1 or DBC1^K215Q expressed in U2OS cells. The GFP signal in the nucleus versus the cytoplasm was quantified. Data are mean ± SEM, n = 35. (C) Amide CSPs for the S1‐L (■) and S1‐L^K112Q (■) in the presence of 500 μM SIRT1 NTR. (D) Amide CSPs for the S1‐L (■) and S1‐L^K112Q (■) in the presence of 277 or 324 μM SIRT1^{1‐233 S162D/S172D}, respectively. In (c and d) the arrows point out resonances corresponding to residues in the SIRT1 binding interface.

Figure S3: Interactions of the S1‐L with the SIRT1 catalytic domain. (A) Superposition of ¹H,¹⁵N HSQC spectra of DBC1 S1‐L in the presence (●) or absence (●) of 500 μM SIRT1^233‐655*. (B) Combined amide CSPs of the DBC1 S1‐L in the presence of 500 μM SIRT1 NTR (■) or 500 μM SIRT1 catalytic domain with a linked ESA region (■). (C) Superposition of ¹H,¹⁵N HSQC spectra of the SIRT1 catalytic domain with a linked ESA region in the presence (●) or absence (●) of 500 μM DBC1 S1‐L.

Figure S4: DBC1 S1‐L binding by SIRT1^141‐233, SIRT1^109‐233, and SIRT1^141‐655*. (A) Superposition of ¹H,¹⁵N HSQC spectra of 30 μM SIRT1^141‐233 in the presence (●) or absence (●) of 100 μM DBC1 S1‐L. (B) Superposition of ¹H,¹⁵N HSQC spectra of 30 μM SIRT1^109‐233 in the presence (●) or absence (●) of 100 μM DBC1 S1‐L. (C) Superposition of ¹H,¹⁵N HSQC spectra of 30 μM DBC1 S1‐L in the presence (●) or absence (●) of 160 μM SIRT1^{141‐655 *}. The insets in all spectra display individual binding isotherms for four resonances. Global fitting yielded K _d values of 330 ± 50 μM; 310 ± 60 μM and *K _d = 420 ± 150 μM, respectively.

Figure S5: Interaction between the two halves of the SIRT1 NTR. (A) Superposition of ¹H,¹⁵N HSQC spectra of SIRT1^1‐54 in the presence (●) or absence (●) of 500 μM SIRT1^109‐233. (B) Superposition of ¹H,¹⁵N HSQC spectra of SIRT1^1‐84 in the presence (●) or absence (●) of 500 μM SIRT1^109‐233. (C) Superposition of ¹H,¹⁵N HSQC spectra of SIRT1^109‐233 in the presence (●) or absence (●) of 500 μM SIRT1^1‐54. The insets in (A), (C) display binding isotherms for four resonances. Global fitting yielded K _d values of 310 ± 40 μM and K _d = 260 ± 30 μM, respectively.

Figure S6: Modification of SIRT1 S162 and S172. Superposition of the ¹H,¹⁵N HSQC spectra of the SIRT1 NTR and SIRT1^{1‐233 S162A/S172A} (top), SIRT1^{1‐233 S162D/S172D} (middle), or SIRT1 NTR treated with CK2/GSK3β (bottom) for 6 h. The boxed regions are enlarged in the insets highlighting the L192 resonance.

Figure S7: Amide resonance assignments for the complete SIRT1 NTR. Assignments from all SIRT1 sub‐constructs were combined and nearly complete resonance assignments by residue name and number are labeled on the ¹H,¹⁵N HSQC spectrum of the SIRT1 NTR. The four insets provide enlargements of the center region of the spectrum. Each inset enlarges a slice from top to bottom of the crowed middle region from top to bottom.

Supplemental Table 1: Crystallographic Statistics for DBC1 52‐120.

Krzysiak TC, Choi Y‐J, Kim YJ, Yang Y, DeHaven C, Thompson L, et al. Inhibitory protein–protein interactions of the SIRT1 deacetylase are choreographed by post‐translational modification. Protein Science. 2024;33(4):e4938. 10.1002/pro.4938

DATA AVAILABILITY STATEMENT

Backbone ¹H, ¹⁵N, and ¹³C chemical shifts for the SIRT1^1‐54 (BMRB ID: 52178), SIRT1^1‐84 (BMRB ID: 52181), SIRT1^1‐124 (BMRB ID: 52182), SIRT1^141‐233 (BMRB ID: 52183), SIRT1^109‐233 (BMRB ID: 52184) and DBC1^52‐120 (BMRB ID: 52186) have been deposited in the BMRB, http://www.bmrb.wisc.edu. The DBC1 S1‐L crystal structure has been deposited in the PDB with code 8EZ6.