Structural and functional characterization of the role of acetylation on the interactions of the human Atg8-family proteins with the autophagy receptor TP53INP2/DOR

ABSTRACT

The Atg8-family proteins (MAP1LC3/LC3A, LC3B, LC3C, GABARAP, GABARAPL1 and GABARAPL2) play a pivotal role in macroautophagy/autophagy through their ability to help form autophagosomes. Although autophagosomes form in the cytoplasm, nuclear levels of the Atg8-family proteins are significant. Recently, the nuclear/cytoplasmic shuttling of LC3B was shown to require deacetylation of two Lys residues (K49 and K51 in LC3B), which are conserved in Atg8-family proteins. To exit the nucleus, deacetylated LC3B must bind TP53INP2/DOR (tumor protein p53 inducible nuclear protein 2) through interaction with the LC3-interacting region (LIR) of TP53INP2 (TP53INP2LIR). To examine their selectivity for TP53INP2 and the role of the conserved Lys residues in Atg8-family proteins, we prepared the six human Atg8-family proteins and acetylated variants of LC3A and GABARAP for biophysical and structural characterization of their interactions with the TP53INP2LIR. Isothermal titration calorimetry (ITC) experiments demonstrate that this LIR binds preferentially to GABARAP subfamily proteins, and that only acetylation of the second Lys residue reduces binding to GABARAP and LC3A. Crystal structures of complexes with GABARAP and LC3A (acetylated and deacetylated) define a β-sheet in the TP53INP2LIR that determines the GABARAP selectivity and establishes the importance of acetylation at the second Lys. The in vitro results were confirmed in cells using acetyl-mimetic variants of GABARAP and LC3A to examine nuclear/cytoplasmic shuttling and colocalization with TP53INP2. Together, the results demonstrate that TP53INP2 shows selectivity to the GABARAP subfamily and acetylation at the second Lys of GABARAP and LC3A disrupts key interactions with TP53INP2 required for their nuclear/cytoplasmic shuttling.

Introduction

To date, there are over forty known ATG (autophagy related) genes and within this group twenty Atg proteins are considered to constitute the core machinery for the process including the ubiquitin-like protein Atg8 [1]. In yeast there is a single Atg8 protein, whereas the Atg8 protein family in humans is composed of six orthologs of Atg8, which can be divided into two subfamilies. The LC3 subfamily of human Atg8-family proteins consists of LC3A, LC3B and LC3C whereas the GABARAP subfamily consists of GABARAP (GABA type A receptor-associated protein), GABARAPL1 (GABA type A receptor associated protein like 1), and GABARAPL2 (GABA type A receptor associated protein like 2) [2]. These six human Atg8-family proteins share both sequence and structural homology, which includes a conserved glycine residue near the C-terminal end of the protein that is conjugated to phosphatidylethanolamine (PE) as well as a common fold that includes a ubiquitin-like (UBL) domain [3,4]. Despite their structurally similarity, the LC3 and GABARAP subfamilies appear to perform different functions during autophagosome formation, with the LC3 subfamily playing a key role in autophagosome elongation, whereas the GABARAP subfamily proteins are required for autophagosome vesicle closure [5]. However, precise individual functions for the individual human Atg8-family proteins remain less clearly defined.

Although autophagy was initially thought to be a nonselective bulk sequestration of macromolecules, it is now known that the Atg8-family proteins play a key role in several selective forms of autophagy [6,7]. More specifically, it was determined that most of these selected autophagy receptors contain an LC3-interacting region (LIR) in their sequence that enables them to bind to one or several of the six human Atg8-family proteins [5,8]. The core consensus sequence for LIRs comprises four amino acid residues, (W/F/Y)₀-X₁-X₂-(L/I/V)₃, where X can be any amino acid [5,8]. Structural studies show that the different Atg8-family proteins bind to the LIRs in similar manners, but that certain autophagy receptors have the capacity to preferentially target select members [9,10]. These studies demonstrate that the aromatic residue in the 0 position and the hydrophobic residue in position 3 of the LIR motif bind respectively to hydrophobic pocket 1 (HP1) and HP2 on the surface of all Atg8-family proteins. Although the interactions with the two HPs is the main driving force for complex formation, the presence of neighboring negatively charged amino acids (Asp and Glu) in the LIR contributes to electrostatic interactions with positively charged residues in the Atg8-family proteins [5,9–13].

Although autophagosomes form in the cytoplasm, human Atg8-family proteins are abundantly present in the nucleus and select autophagy signals trigger their shuttling to the cytoplasm [14–18]. Recently, it was demonstrated that nuclear LC3B is acetylated by the histone acetylase EP300/p300 (E1A binding protein p300) on two conserved Lys residues (K49 and K51 in LC3-subfamily proteins and K46 and K48 in GABARAP-subfamily proteins) [14,15]. In response to starvation, LC3B must be first deacetylated on both of these Lys residues by the histone deacetylase SIRT1 (sirtuin 1) in order for it to exit the nucleus and enter the cytoplasm. This translocation also requires that LC3B binds to the transcriptional regulator protein TP53INP2/DOR (tumor protein p53 inducible nuclear protein 2), which contains a LIR motif [19,20]. In complex with TP53INP2, LC3B enters the cytoplasm, where it is conjugates with PE prior to attachment onto phagophores [14,15].

Given that acetylation and deacetylation of LC3B at two conserved Lys residues (K49 and K51) has been shown to play an essential role in regulating its binding to TP53INP2 [14–16,18], we were interested in determining whether the six human Atg8-family proteins all have similar binding affinities for TP53INP2 protein and whether acetylation/deacetylation of other human Atg8-family proteins played a role in their binding to TP53INP2 as well as in their translocation from the nucleus to the cytoplasm during autophagy. Using a combination of biophysical, structural, and cellular studies, we demonstrated that the LIR of TP53INP2 binds with almost 100-fold higher specificity to the GABARAP subfamily and that site-specific acetylation at the second conserved Lys residue of LC3A (K51) and GABARAP (K48) significantly reduced the binding to TP53INP2, whereas site-specific acetylation at the first conserved Lys of LC3A (K49) and GABARAP (K46) had little to no effect. Structural studies of TP53INP2 in complex with GABARPAP and LC3A showed that several of the residues of GABARAP that form key interactions with the LIR of TP53INP2 (TP53INP2LIR; residues 31 to 43 of human TP53INP2) are not conserved in the LC3A and that acetylation at the second conserved Lys of GABARAP and LC3A disrupted key interactions involving W35 of the TP53INP2LIR. The in vitro studies were supported by cell-based studies using acetyl-mimetic variants of GABARAP, which demonstrated that the second conserved Lys significantly inhibited nuclear to cytoplasm shuttling as well as colocalization with TP53INP2 during autophagy. Taken together, these results provide important insight into the specificity of the six human Atg8 proteins for TP53INP2 as well as the role that the acetylation status of the second conserved Lys residues play in the interaction of TP53INP2 with GABARAP and LC3A in regulating their shuttling from the nucleus to participate in autophagosome formation in the cytoplasm.

Results

The TP53INP2LIR shows selectivity for the GABARAP subfamily

Previous cellular studies show that the TP53INP2 protein plays a critical role in the regulation of the autophagy process through its ability to interact with deacetylated Atg8-family proteins and thereby control their shuttling from the nucleus to the cytoplasm. Although these studies demonstrated that the interaction between TP53INP2 and the Atg8-family proteins require the LIR of TP53INP2, there are no detailed structural studies comparing the binding specificity of the TP53INP2LIR to the six different human Atg8-family proteins and the role that acetylation of the two conserved Lys residues has in the interaction [14,15]. To initiate these studies, we first examined the binding specificity of the TP53INP2LIR for the six different human Atg8 proteins. In the case of the TP53INP2, three previous studies have attempted to examine its binding to various members of the Atg8-family proteins, but these three different studies gave extremely varying results [21–23]. In all the previous studies, either the Atg8 proteins, the TP53INP2 portion and/or both were prepared as fusion products, tagged and/or attached to a surface, and these modifications could explain the discrepancies between the different results.

To help clarify whether the TP53INP2LIR binds selectively to a specific Atg8-family protein, we purified the six human Atg8 proteins free of tags and determined their dissociation constants (K_D) for binding to a purified peptide containing the TP53INP2LIR (residues 31 to 43 of human TP53INP2) in solution using isothermal titration calorimetry (ITC) experiments. The ITC experiments (Figure 1) revealed that the TP53INP2LIR peptide bound with significantly higher affinity to the three GARAPAP subfamily proteins under the experimental conditions (20 mM phosphate pH 6.5 and 50 mM NaCl), and the highest affinity was observed with GABARAP (K_D = 0.0033 ± 0.0007 μM), followed by GABARAPL1 (K_D = 0.024 ± 0.002 μM) and then with GARARAPL2 (K_D = 0.067 ± 0.015 μM) Figure 1(B). In contrast, LC3A (K_D = 0.43 ± 0.07 μM), LC3C (K_D = 0.62 ± 0.16 μM) and LC3B (K_D = 2.9 ± 1.5 μM) all bound to the TP53INP2LIR peptide with significantly lower affinity than the three members of the GABARAP subfamily and in particular all three LC3-subfamily proteins bound over two-orders of magnitude weaker than GABARAP Figure 1(B).

Figure 1.

The TP53INP2LIR displays binding selectivity to the GABARAP subfamily. (A) a representative ITC thermogram demonstrating the binding between GABARAP and the TP53INP2LIR; (B) comparison of the K_D values (in μM) obtained in ITC studies with the six human Atg8-family proteins and the TP53INP2LIR. All ITC experiments in (B) were performed at 25°C in 20 mM phosphate buffer pH 6.5, 50 mM NaCl and 1 mM dithiothreitol. The ITC experiments were fit to a single binding site model with 1:1 stoichiometry and the standard deviations for the K_D and N-values (stoichiometry) were based on duplicate measurements or more.

Given the high affinity observed with the GABARAP subfamily in the ITC experiments in buffer containing 50 mM NaCl, the ITC experiments were repeated with the GABARAP subfamily and LC3A under identical buffer conditions, but in the presence of 150 mM NaCl (Table 1). Again, all three members of the GABARAP family members bound with significantly higher affinities [GABARAP (K_D = 0.020 ± 0.002 μM), GABARAPL1 (K_D = 0.033 ± 0.001 μM) and GARARAPL2 (K_D = 0.230 ± 0.05 μM)] than LC3A (K_D = 1.7 ± 0.1 μM), and binding to GABARAP was again almost two orders of magnitude stronger than the binding to LC3A (85-fold). These ITC results indicate that the TP53INP2LIR bound preferentially to the GABARAP subfamily in comparison to the LC3 subfamily and are most consistent with the previous results using time-resolved fluorescence energy resonance transfer/FRET experiments [22].

Table 1.

ITC results in 150 mM NaCl.

Atg8-family proteins	TP53INP2LIR KD (µM)	N-value (stoichiometry)
GABARAP	0.020 ± 0.002	1.185 ± 0.005
GABARAPL1	0.033 ± 0.001	1.05 ± 0.000
GABARAPL2	0.23 ± 0.05	1.29 ± 0.000
GABARAP[K48Ac]	0.23 ± 0.08	1.02 ± 0.000
GABARAP[K46Ac]	0.047 ± 0.003	1.16 ± 0.00
LC3A	1.7 ± 0.1	1.19 ± 0.02
LC3A[K51Ac]	22 ± 2.5	0.821 ± 0.012
LC3A[K49Ac]	3.7 ± 0.3	0.956 ± 0.001

All experiments were conducted at 25°C in 20 mM phosphate buffer pH 6.5, 150 mM NaCl and 1 mM dithiothreitol (DTT). The ITC experiments were fit to a single binding site model with 1:1 stoichiometry and the standard deviations for the KD values and the N-values (stoichiometry) were based on two or more measurements.

GABARAP displaces LC3B from a complex with the TP53INP2LIR

To confirm that the TP53INP2LIR preferentially bound to GABARAP in comparison to LC3B, nuclear magnetic resonance (NMR) competition experiments were performed. To initiate these experiments, we first added an equal concentration of the unlabeled TP53INP2LIR peptide (0.20 mM) to a sample containing¹⁵N-labeled LC3B (0.20 mM). As expected, we observed significant changes for several¹H and¹⁵N chemical shifts in the¹H-¹⁵N HSQC spectra of the labeled LC3B Figure 2(A), and the signals from LC3B that exhibited the most significant chemical shift changes (Figure 2(B)) when mapped onto the structure of LC3B Figure 2(C) were located around the consensus LIR-binding region present in the LC3-family proteins. We then added an equimolar amount of unlabeled GABARAP (0.20 mM) to the¹⁵N-LC3B-TP53INP2LIR complex and observed that the ¹H and¹⁵N resonances of LC3B, which shifted upon complex formation returned to the same chemical shift positions of the free form of LC3B Figure 2(D). Taken together with the ITC studies, the NMR results demonstrated that the TP53INP2LIR preferentially bound to GABARP in comparison with LC3B.

Figure 2.

GABARAP displaces the TP53INP2LIR from binding to LC3B at equimolar concentrations. (A) overlay of 2D ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled LC3B in the free form (black) and in the presence of an equimolar concentration (0.2 mM) of unlabeled TP53INP2LIR (blue); (B) histogram of the variation in chemical shifts [Δδ (ppm)] observed in the¹H-¹⁵N HSQC spectra of ¹⁵N-labeled LC3B upon the formation of the LC3B-TP53INP2LIR complex. The chemical shifts were calculated with the formula Δδ=[(0.17ΔN_H)²+(ΔH_N)²]^½, and are given in parts per million. The dashed line represents the value for significant chemical shift changes for a given amino acid position. (C) cartoon representation of the structure of LC3B (yellow; PDB: 3VTU) highlighting the residues (magenta) that undergo significant chemical changes (Δδ > 0.25) upon addition of the TP53INP2LIR based on the histogram shown in panel (B); (D) overlay of 2D ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled LC3B in the free form (black), in the presence of an equimolar concentration (0.2 mM) of unlabeled TP53INP2LIR (blue) and in the presence of equimolar concentration (0.2 mM) of both unlabeled TP53INP2LIR and unlabeled GABARAP (red). All NMR experiments were carried out at 298 K in 20 mM sodium phosphate, pH 6.5, 50 mM NaCl and 90% H₂O, 10% D₂O.

Acetylation of LC3A and GABARAP inhibits binding to TP53INP2LIR

Recently, it has been shown that LC3B present in the nucleus is acetylated on two conserved Lys residues (K49 and K51 in LC3A) by EP300 [14,15,24], and that these Lys residues must be de-acetylated by SIRT1 for the LC3B protein to interact with TP53INP2 and exit the nucleus during autophagy [14,15]. The two Lys residues in LC3A (K49 and K51) and their equivalents in GABARAP (K46 and K48), are on the surface of the protein adjacent to two hydrophobic pockets (HP1 and HP2) that play a key role in the interactions with autophagy regulatory proteins containing LIR motifs such as TP53INP2 [25–27]. However, the exact role that acetylation at the individual Lys residues plays in the interactions of these Atg8-family proteins with TP53INP2 has not been determined either biophysically or structurally. To test the role of acetylation at the individual Lys positions, we expressed and purified variant proteins of GABARAP site-specifically acetylated at either K46 (GABARAP[K46Ac]) or K48 (GABARAP[K48Ac]) as an example for the GABARAP subfamily as well as variant proteins of LC3A site-specifically acetylated at either K49 (LC3A[K49Ac]) or K51 (LC3A[K51Ac]).

These variant proteins were generated using a bacterial system that allows for site selective incorporation of acetyl-Lys [28,29]. ITC experiments were then performed in 20 mM phosphate pH 6.5 buffer containing 150 mM NaCl to determine the K_D values for the binding of the acetylated variants of GABARAP and LC3A to the TP53INP2LIR as described above for the wild-type proteins (Table 1). In the case of GABARAP, acetylation at K48 decreased the binding affinity to the TP53INP2LIR by over 11-fold (K_D = 0.23 ± 0.08 µM) in comparison to the wild-type GABARAP (K_D = 0.020 ± 0.001 µM), whereas acetylation at K46 decreased binding by just slightly over 2-fold (K_D = 0.047 ± 0.003 µM). Similarly, with LC3A, acetylation at K51 decreased the binding affinity to the TP53INP2LIR by 13-fold (K_D = 22 ± 2.5 µM) in comparison to the wild-type LC3A (K_D = 1.7 ± 0.1 µM), whereas as acetylation at K49 only decreased binding by slightly more than 2-fold (K_D = 3.7 ± 0.3 µM). Taken together, the results demonstrated that acetylation of the second conserved Lys residue (K51 in LC3A and K48 in GABARAP) of GABARAP and LC3A significantly reduced the binding to the TP53INP2LIR whereas acetylation of the first conserved Lys residue (K49 in LC3A and K46 in GABARAP) only slightly reduced the affinity. This suggests that deacetylation of the second conserved Lys residue in LC3A and GABARAP by SIRT1 was important for their shuttling from the nucleus to the cytoplasm.

Structures of TP53INP2LIR in complex with GABARAP

To determine the binding preference of the TP53INP2LIR for the Atg8-family proteins at the atomic level, attempts were made to determine crystal structures of the TP53INP2LIR in complex with all six human Atg8 proteins. After extensive screening of different crystallization conditions, co-crystals were obtained for a TP53INP2LIR-GABARAP complex, and these crystals diffracted at 2.1Å resolution (Table S1). In complex with the TP53INP2LIR, GABARAP adopted a typical Atg8 fold Figure 3(A), like a recent structure of a TP53INP2LIR-GABARAP fusion protein. As seen in the structure of the TP53INP2LIR-GABARAP fusion protein [21], the TP53INP2LIR also formed a short two stranded β-sheet in the co-crystal complex with GABARAP Figure 3(A). In the co-crystals, this β-sheet conformation was stabilized by two intermolecular interactions between the TP53INP2LIR and GABARAP as well as one intramolecular interaction in the TP53INP2LIR. The intramolecular interaction was a hydrophobic contact between the side chains of V32 and I37 of the TP53INP2LIR, whereas the intermolecular interactions were a V32-L55 (TP53INP2LIR-GABARAP, respectively) hydrophobic contact and an E31-K46 ionic interaction Figure 3(B). The TP53INP2LIR in the co-crystal differed from what is observed in the structure with the TP53INP2LIR-GABARAP fusion protein. In the structure with the fusion protein, there is no V32-L55 interaction (7.7Å in fusion complex versus 3.9Å in co-crystal complex) due to constraints the fusion protein imposes (Figure S1).

Figure 3.

Co-crystal structure of TP53INP2LIR-GABARAP complex. (A) cartoon representation of the structure of the TP53INP2LIR-GABARAP complex highlighting the β-sheet conformation of the TP53INP2LIR. In this representation, GABARAP is colored in cyan and the TP53INP2LIR in magenta and the N terminus (N) and C terminus (C) of both are labeled. (B) Close-up and metrics (in Å) of key amino acid residues that stabilize the β-sheet conformation of the TP53INP2LIR in complex with GABARAP. The interface residues involved are shown in stick representation along with the distances of key intramolecular contacts (indicated with dashed black lines) between V32-I37 of the TP53INP2LIR as well as key intermolecular contacts between V32-L55 and E31-K46 (TP53INP2LIR-GABARAP, respectively). (C) Close-up and metrics (dashed black lines in Å) of the TP53INP2LIR-GABARAP complex highlighting the side chains of residues in HP1 of GABARAP (cyan) that make either hydrophobic (Y5, I21, P30, I32, L50, F104), hydrogen bond (E17) or cation-π (K48) interactions with the side chain of W35 of TP53INP2LIR (magenta). (D) Close-up and metrics (dashed black lines in Å) of the TP53INP2LIR-GABARAP complex highlighting the side chains of residues in HP2 (Y49, V51, P52, L55, F60 and L63) of GABARAP (cyan) that interact with I38 of TP53INP2LIR (magenta). (E) Close-up and metrics (dashed black lines in Å) of the TP53INP2LIR-GABARAP complex highlighting additional key hydrophobic interactions involving amino acids in HP1 of GABARAP (cyan) and residues in the TP53INP2LIR (magenta). These hydrophobic contacts are between V32-L50, L36-K46, L36-Y49, L40-P52, L40-L55 and P41-L55 (TP53INP2LIR-GABARAP, respectively). (F) Close-up and metrics (dashed black lines in Å) of the TP53INP2LIR-GABARAP complex highlighting electrostatic interactions between E31-K46 as well as between D39-R28 (TP53INP2LIR-GABARAP, respectively).

The predominant interactions at the binding interface of the TP53INP2LIR-GABARAP complex involved the side chains of W35 and I38 of TP53INP2, which are the residues located in the zero and third position of the core LIR motif Figure 3(C,D). These residues inserted into the hydrophobic pocket 1 (HP1) and HP2 of GABARAP respectively, and this is consistent with most other previously determined structures of LIR motifs in complex between Atg8-family proteins [9,10,23,27,30–32]. More specifically, the indole ring of W35 from the TP53INP2LIR made hydrophobic contacts with the side chains of Y5 (4.7Å), I21 (3.8Å), P30 (4.5Å), I32 (3.6Å), L50 (4.8Å) and F104 (3.5Å) of GABARAP in HP1 Figure 3(C), whereas the side chain of I38 of the TP53INP2LIR made hydrophobic contacts with Y49 (4Å), V51 (4.8Å), P52 (3.7Å), L55 (4.3Å), F60 (5Å) and L63 (3.9Å) of GABARAP in HP2 Figure 3(D). Furthermore, the indole ring of W35 also formed a hydrogen bond and cation-π interaction with E17 (3Å) and K48 (3.8Å) of GABARAP, respectively Figure 3(C). However, the binding interface is also stabilized by additional interactions with several other amino acids before, within and after the core of the LIR motif of TP53INP2. Additional hydrophobic interactions occurred between the side chains of V32-L50 (3.9 Å; TP53INP2LIR-GABARAP, respectively), L36-K46 (4.3Å), L36-Y49 (4Å), L40-P52 (3.9Å), L40-L55 (3.9Å) and P41-L55 (3.5Å) Figure 3(E), whereas electrostatic interactions were present between the sidechains of E31-K46 (3Å) and D39-R28 (4.9Å) Figure 3(F).

Overall, the interface is similar in the co-crystal and the fusion proteins. However, the TP53INP2LIR in the co-crystals contained two additional residues at the C-terminal (P41 and D42), and there is a P41-L55 contact (3.9 Å) in the co-crystal Figure 3(E). In addition, the TP53INP2LIR-GABARAP fusion construct contained two additional amino acids on the N-terminal (E29 and E30), and in the structure with the fusion protein E29 formed an electrostatic contact with R67 (3.9 Å) of GABARAP (Figure S1B) [21]. In total, the binding interface of the co-crystal structure of the TP53INP2LIR-GABARAP complex involved 16 residues of GABARP and 9 residues of the TP53INP2LIR. Consistent with the ITC results, all 16 residues in GABARAP at the interface are identical in GABARAPL1 and this explained the similar binding affinities. In contrast, only 13 of the 16 binding residues of GABARAP are conserved in GABARAPL2. The three residues that are not identical are Y5, L55 and L63 of GABARAP, which correspond to F5, I55 and I63 in GABARAPL2. Thus, these three differences could explain the 10-fold lower binding affinity of GABARAPL2 for the TP53INP2LIR.

Effect of GABARAP K46 and K48 acetylation on binding to TP53INP2LIR at the atomic level

To examine the binding difference observed in the ITC studies with the individually acetylated forms of GABARAP (K46 and K48) at the atomic level, we determined the structure of co-crystals of a TP53INP2LIR-GABARAP[K48Ac] complex and a TP53INP2LIR-GABARAP[K46Ac] complex that diffracted at 2Å and 1.6Å resolution, respectively (Table S1). In the structure of the TP53INP2LIR-GABARAP[K46Ac] complex, the GABARAP[K46Ac] looked virtually identical to wild-type GABARAP except for K46 as expected, but the TP53INP2LIR bound in an extended conformation as opposed to the β-sheet conformation observed in the complex with the wild-type GABARAP Figure 4(A), S2A. In contrast in the structure of the TP53INP2LIR-GABARAP[K48Ac] complex, both the GABARAP[K48Ac] and the TP53INP2LIR looked virtually identical as in the complex between the wild-type GABARAP and the TP53INP2LIR Figure 4(B), S2B.

Figure 4.

Co-crystal structures of GABARAP acetylated at either K46 or K48 in complex with the TP53INP2LIR. (A) cartoon representation of the structure of the TP53INP2LIR-GABARAP[K46Ac] complex highlighting the linear conformation of the TP53INP2LIR (magenta) when complexed with GABARAP[K46Ac] (TV blue). The N terminus (N) and C terminus (C) of GABARAP[K46Ac] and the TP53INP2LIR are labeled. (B) cartoon representation of the structure of the TP53INP2LIR-GABARAP[K48Ac] complex highlighting the β-sheet conformation of TP53INP2LIR (magenta) in a complex with GABARAP[K48Ac] (sky blue). The N terminus (N) and C terminus (C) of GABARAP[K48Ac] and the TP53INP2LIR are labeled. (C) Close-up view and metrics (dashed black lines in Å) of amino acid residues from HP1 of GABARAP[K46Ac] (TV blue) that make either hydrophobic (Y5, I21, P30, I32, L50, F104), hydrogen bond (E17) or cation-π (K48) interactions with the side chain of W35 of TP53INP2LIR (magenta). (D) Close-up view and metrics (dashed black lines in Å) of amino acid residues from HP2 (Y49, V51, P52, L55, F60 and L63) of GABARAP[K46Ac] (TV blue) that interact with I38 of TP53INP2LIR (magenta). (E) Close up and metrics (dashed black lines in Å) highlighting electrostatic interaction between R28 of the TP53INP2LIR and D39 of GABARAP[K46Ac]. GABARAP[K46Ac] (TV blue) and TP53INP2LIR (magenta) are shown as cartoons and the interacting amino acids as stick conformations. (F) Close-up view and metrics (dashed black lines in Å) of amino acid residues from HP1 of GABARAP[K48Ac] (sky blue) that form hydrophobic (Y5, I21, P30, I32, L50, F104) and hydrogen bond (E17) interactions with the side chain of W35 of the TP53INP2LIR (magenta). Acetylation of K48 of GABARAP disrupts the cation-π interaction between W35 of the TP53INP2LIR as well as an electrostatic interaction with E17 of the TP53INP2LIR. (G) Close-up view and metrics (dashed black lines in Å) of amino acid residues from HP2 (Y49, V51, P52, L55, F60 and L63) of GABARAP[K48Ac] (sky blue) that interact with I38 of the TP53INP2LIR (magenta). The side chain of residues involved in the interactions are shown in stick representation along with the distances of key interactions. (H) Close-up of W35 of the TP53INP2LIR in complex with GABARAP[K48Ac]. GABARAP[K48Ac] (sky blue) and TP53INP2LIR (magenta) are shown as cartoons and the amino acids of interest as stick conformations. In the complex, acetylation of K48 of GABARAP disrupts the cation-π interaction between W35 of TP53INP2LIR and K48 of GABARAP as well as the electrostatic interaction with E19 of the TP53INP2LIR.

Despite the difference in the structure of the TP53INP2LIR, the interactions observed at the interface of the complex between the TP53INP2LIR and the GABARAP[K46Ac] variant in the co-crystals were overall like the interactions observed at the interface with wild-type GABARAP as expected from the results of the ITC experiments Figure 4(C,D). For example, the indole ring of W35 from the TP53INP2LIR made similar contacts with the side chains of I21 (3.9Å versus 3.8Å in wild-type GABARAP), P30 (3.5Å versus 4.5Å), I32 (5.8Å versus 3.6Å), L50 (4.8Å versus 4.8Å), F104 (3.7Å versus 3.5Å), E17 (3.4Å versus 3.0Å) and K48 (2.9Å versus 3.8Å) of HP1 in GABARAP[K46Ac] Figure 4(C). However, there are two interactions that were noticeably absent in the structure with the K46Ac variant in comparison to the wild-type GABARAP and they were between V32-L55 (TP53INP2-GABARAP[K46Ac], respectively) and E31-K46. The ionic interaction between E31-K46 was no longer present due to the absence of a positive charge on K46 due to the presence of the acetyl moiety, and this likely explained the loss of the β-sheet conformation in the TP53INP2LIR as well as the loss of the V32-L55 interaction. Interestingly, this appeared to be partially compensated by a more pronounced ionic interaction between R28 of GABARAP[K46Ac] and D39 of the TP53INP2LIR (2.9Å in GABARAP[K46Ac] versus 4.9Å in wild-type GABARAP) Figure 4(E), S2C-D.

In contrast, acetylation at K48 disrupted many key interactions involving W35 in the TP53INP2LIR-GABARAP[K48Ac] complex Figure 4(F-H). For instance, the cation-π interaction between W35-K48Ac (TP53INP2LIR-GABARAP[K48Ac]) and the electrostatic interaction between E17-K48Ac were not present because of the missing positive charge that resulted from the acetylation of K48 Figure 5(H), S2E-F. These two changes explained why acetylation at K48 significantly reduced the binding of GABARAP to the TP53INP2LIR and supports the ITC results.

Figure 5.

Structural characterization of the TP53INP2LIR-LC3A complex and comparison to the binding interface with the TP53INP2LIR-GABARAP complex. (A) cartoon representation of the structure of the TP53INP2LIR-LC3A complex highlighting the β-sheet conformation of the TP53INP2LIR (magenta) when bound to LC3A (forest green). The N terminus (N) and C terminus (C) of LC3A and the TP53INP2LIR are labeled (B) Close-up and metrics (in Å) of key amino acid residues that stabilize the β-sheet conformation of TP53INP2LIR when bound to LC3A. The residues involved are shown in stick representation. (C) Close-up view and metrics (in Å) of amino acid residues from HP1 of LC3A (forest green) that make either hydrophobic (F7, I23, P32, I34, L53 and F108), hydrogen bond (E19) or cation-π (K51) interactions with the side chain of W35 of the TP53INP2LIR (magenta). (D) Close-up view and metrics (in Å) of amino acid residues from HP2 (F52, V54, P55, V58, L63 and I66) of LC3A (forest green) that interact with I38 of TP53INP2LIR (magenta). The side chain of residues involved in interactions are shown in stick representation along with the distances of key interactions. (E) left Panels: Close-up view and metrics (in Å) of W35 of the TP53INP2LIR interacting with either F7 of LC3A (upper left) or the corresponding Y5 of GABARAP (lower left). center Panels: Close-up view and metrics (in Å) of I38 of TP53INP2LIR interacting with either L63 of LC3A (upper center) or the corresponding F60 of GABARAP (lower center). right Panels: Close-up view and metrics (in Å) of V32 of TP53INP2LIR interacting with either H27 of LC3A (upper right) or the corresponding Y25 of GABARAP (lower right). In all figures in panel (E), LC3A (green), GABARAP (cyan) and the TP53INP2LIR (magenta) are drawn as cartoons, the side chains of the key amino acids from each protein are in stick representations and the distances are indicated in Å with black dash lines.

Structure of TP53INP2LIR in complex with LC3A

To examine the difference in binding affinities observed between the GABARAP and LC3 subfamilies for the TP53INP2LIR at the atomic level, we next attempted to obtain co-crystals of a TP53INP2LIR-LC3A complex. Although we were not able to obtain co-crystals, we obtained crystals of a TP53INP2LIR-LC3A fusion protein that diffracted at 2.1Å resolution (Table S2). As observed in co-crystals with GABARAP, TP53INP2LIR formed a short two stranded β-sheet in the structure obtained when fused with LC3A, that is stabilized by hydrophobic interactions involving the side chains of V32 and I37 of TP53INP2LIR with residues in LC3A as well as an electrostatic interaction between E31 of the TP53INP2LIR and K49 of LC3A Figure 5(A,B).

As seen in complex with GABARAP, the predominant interactions at the interface formed between the side chain of W35 of TP53INP2 with residues in HP1 of LC3A (F7, I23, P32, I34, L53 and F108) and between the side chain of I38 of TP53INP2 with residues in the HP2 of LC3A (F52, V54, P55, V58, L63 and I66 Figure 5(C,D). However, there were several differences in comparing the two binding interfaces that helped explain the drastic difference in binding affinities of GABARAP and LC3A for the TP53INP2LIR by ITC (~85-fold). In particular, residues within HP1 and HP2 that are not conserved between the two Atg8-family proteins displayed significantly different interactions with the side chains of W35 and I38. For example, in HP1, F7 of LC3A is positioned 6.7Å from the aromatic ring of W35 in TP53INP2 in the TP53INP2LIR-LC3A structure, whereas the corresponding Y5 of GABARAP is located 4.7Å from W35 of TP53INP2 in TP53INP2LIR-GABARAP complex Figure 5(E); left panels. Likewise, in HP2, L63 of LC3A are positioned 6.1Å from the side chain of I38 in TP53INP2 in the TP53INP2LIR-LC3A structure, whereas the corresponding F60 of GABARAP is located 5Å from I38 of TP53INP2 in TP53INP2LIR-GABARAP complex Figure 5(E); center panels.

The second major difference between the two complexes involves interaction involving the side chains of V32 and I37 of TP53INP2. In complex with GABARAP, these residues of TP53INP2 were part of a network of hydrophobic interactions that enabled V32 to interact with Y25 of GABARAP (4.2Å). In contrast, in the LC3A complex the interaction was not observed between V32 of TP53INP2 and H27 of LC3A (6.3Å), which corresponds to Y25 in GABARAP Figure 5(E); right panels. However, we cannot rule out that this difference might be in part due to restrictions of the TP53INP2-LC3A fusion protein versus the co-crystals obtained for TP53INP2 with GABARAP. Taken together, these differences indicated that several non-conserved residues before, within and after the core LIR motif were important for the higher affinity binding observed with the GABARAP subfamily for the TP53INP2LIR when compared with the LC3 subfamily.

Effect acetylation at K49 and K51 on binding of LC3A to the TP53INP2LIR at the atomic level

To reveal the molecular interactions controlling the effect of acetylation at the individual acetylated Lys residues of LC3A (K49 and K51) on the interaction with the TP53INP2LIR, we prepared fusion proteins where the TP53INP2LIR was attached the N-terminal of LC3A[K49Ac] and LC3A[K51Ac] as done with the wild-type LC3A and obtained crystals that diffracted at 1.85 Å and 1.9 Å, respectively Table S2 and Figure 6(A,B). As seen with LC3A, the TP53INP2LIR formed a short two stranded β-sheet in both complexes, which was stabilized in part by an electrostatic interaction between E31 of TP53INP2 and K49 of LC3A[K51Ac]. Interestingly, in TP53INP2LIR-LC3A[K49Ac], it was not possible to form the E31-K49 electrostatic interaction due to the acetylation of K49, but we observed an alternative electrostatic interaction between E31 of TP53INP2 and R70 of LC3A[K49Ac] (Figure S3A-B). This interaction helped maintain the β-sheet conformation of the TP53INP2LIR and this in part accounted for the comparable binding affinity with the wild-type LC3A in the ITC experiments. In addition, the indole ring of W35 from the TP53INP2LIR made similar interactions in HP1 with the side chains of I23 (3.2Å versus 3.6Å in wild-type LC3A), P32 (3.8Å versus 3.7Å), I34 (4.8Å versus 5.8Å), L53 (4.7Å versus 4.2Å), F108 (3.4Å versus 4.2Å), E19 (3.7Å versus 3.7Å) and K51(4.6Å versus 4.3Å) of LC3A^49Ac Figure 6(C).

Figure 6.

Structures of the TP53INP2LIR-LC3A[K49Ac] and TP53INP2LIR-LC3A[K49Ac] complexes. (A and B) cartoon representation of the TP53INP2LIR (magenta) in complex with LC3A[K49Ac] (A; lemon) and LC3A[K51Ac] (B; lime). The TP53INP2LIR adopts a β-sheet conformation in both complexes. The N terminus (N) and C terminus (C) of the TP53INP2LIR as well as LC3A[K49Ac] and LC3A[K51Ac] are labeled; (C) Close-up and metrics (in Å) of the TP53INP2LIR-LC3A[K49Ac] complex highlighting the side chains of amino acids residues in HP1 of LC3A[K49Ac] (lemon) that make either hydrophobic (F7, I23, P32, I34, L53 and F108), hydrogen bond (E19) or cation-π (K51) interactions with the side chain of W35 of the TP53INP2LIR (magenta). The side chain of residues involved in interactions are shown in stick representation along with the distances of key interactions in Å as black dashed lines. (D) Close-up of W35 of the TP53INP2LIR in the TP53INP2LIR-LC3A[K51Ac] complex. LC3A[K51Ac] (lime) and the TP53INP2LIR (magenta) are shown as cartoons and the amino acids of interest as stick conformations. In the complex, acetylation of K51 of LC3A disrupts the cation-π interaction with W35 of the TP53INP2LIR as well as its electrostatic interaction with E19 of the TP53INP2LIR.

In contrast, acetylation at K51 disrupted key interactions involving the side chain of W35 of the TP53INP2LIR due to the loss of the positive charge. This included the cation-π interaction between W35-K51Ac (TP53INP2-LC3A[K51Ac], respectively) as well as the electrostatic interaction between E19-K51Ac Figure 6(D), S3C-D. Together these differences explained the over 10-fold decrease in the binding affinity upon acetylation of LC3A at K51 observed in the ITC studies and indicated that the acetylation of LC3A at the K51 position plays a key role in binding to TP53INP2LIR.

Effect of lysine acetylation on cellular localization of GABARAP and LC3A

We next explored the role that acetylation at the individual Lys residues had on the nuclear localization of GABARAP and LC3A and their participation in foci formation in the cytoplasm during autophagy. It is known that the acetylated form of LC3B localizes in the nucleus and that it must be deacetylated by SIRT1 on both Lys residues (K49 and K51 in LC3B) to exit the nucleus and form foci in the cytoplasm [14,15]. To test the effect of the acetylation on the individual Lys residues, vectors for expressing wild-type green fluorescent protein (GFP)-tagged GABARAP and LC3A were prepared. The GABARAP variants were GABARAP^K46Q, GABARAP^K48Q and GABARAP^K46,48Q, whereas the LC3A variants were LC3A^K49Q, LC3A^K51Q and LC3A^K49,51Q.

To first verify that the K-to-Q substitutions mimic acetylation, we performed ITC studies with the purified GABARAP^K46Q, GABARAP^K48Q, LC3A^K49Q and LC3A^K51Q variants and the TP53INP2LIR. The ITC experiments demonstrated that there are significant changes in binding to the TP53INP2LIR when compared to the wild-type proteins for a GABARAP^K48Q and the LC3A^K51Q variants, but only minor differences with the GABARAP^K46Q and the LC3A^K49Q variants (Table S3). These results were very similar to what we observed with the acetylated variants binding to the TP53INP2LIR and support the fact that the K-to-Q variants can function as acetylation mimics in cell-based experiments.

The cell-based experiments were performed in U2OS cells and when expressed in U2OS cells, the wild-type, and all K-to-Q variants of GABARAP and LC3A displayed predominantly nuclear localization under full medium conditions, which is consistent with previous results with LC3B (Figure 7, S4). Upon starvation, the wild-type GFP-GABARAP and GFP-LC3A as well as their variants containing the K-to-Q substitution at the first conserved Lys residue (GABARAP^K46Q and LC3A^K49Q) formed puncta in the cytoplasm, accompanied by the redistribution of the proteins from the nucleus to the cytoplasm (Figure 7, S4). In sharp contrast, the K-to-Q variants of GFP-GABARAP and GFP-LC3A containing the substitution at either the second conserved Lys residue (K48Q in GABARAP and K51Q in LC3A) or at both of the conserved Lys residues (K46Q and K48Q in GABARAP and K49Q and K51Q in LC3A) formed significantly fewer puncti in the cytoplasm and were predominantly present in the nucleus even following starvation (Figure 7, S4). In agreement with the structural and ITC data, the results demonstrated that deacetylation at the second Lys residue of both GABARAP (K48) and LC3A (K51) was crucial for their binding to TP53INP2 and their shuttling into the cytoplasm to participate in the formation of the autophagosome membrane during starvation induced autophagy.

Figure 7.

Subcellular localization of GABARAP is dependent on the acetylation status of the second conserved Lys. U2OS cells were transfected with either 2 μg of GFP-GABARAP or 2 μg of one of its three K-to-Q variants (GFP-GABARAP^K46Q, GFP-GABARAP^K48Q and GFP-GABARAP^K46,48Q). The cells were grown in either full medium (top images of group) or subjected to starvation (bottom images of group) by incubation in media deficient serum at 37°C for 18–24 h and imaged (left panel). The autophagic foci and nuclear staining intensity (DAPI: Hoechst 33342) for each group of cells is calculated for those grown in full media (FM) and starved (ST) cells (right panel). The number of foci per cells (green bars on left side) and the nuclear staining intensity (blue bars on right side) of the GABARAP proteins were calculated based on 50 cells. Images were analyzed with ImageJ/Fiji software and imported via the Bio-Formats plugin. For nuclear staining, intensity region of interest was selected using Hoechst/DAPI image and mean intensity of GFP signal was extracted. For cytoplasmic foci, mask image of GFP signal (background subtracted to visualize individual foci) was created and number of foci was counted using the analyse particles function of the ImageJ software. The distance for each image is indicated by the black bar at the bottom of the figure. Scale bar: 10 μm.

The LIR of WDFY3/ALFY binds similar to the K-to-Q variants of Atg8 proteins as the TP53INP2LIR

Since, like the TP53INP2LIR, the LIR of the WDFY3/ALFY (WD repeat and FYVE domain containing 3) preferentially binds to the GABARAP subfamily and forms a β-sheet structure when bound to GABARAP [33], we were interested to determine the role that acetylation of the two conserved Lys residues played on the binding of the LIR of WDFY3 (WDFY3LIR) to GABARAP and LC3A. For these experiments, we used the K-to-Q variants of GABARAP (K46Q and K48Q) and LC3A (K49Q and K51Q) as acetyl-mimetics and determined their dissociation constants (K_D) using ITC experiments (Table S4) for binding to a purified peptide containing the WDFY3LIR (residues 3341–3355 of human WDFY3). In the case of GABARAP and the two variants, the K48Q substitution decreased the binding affinity to the WDFY3LIR by slightly over 20-fold (K_D = 0.45 ± 0.01 µM) in comparison to wild-type GABARAP (K_D = 0.022 ± 0.002 µM), whereas the K46Q substitution decreased the binding by only 2.8-fold (K_D = 0.062 ± 0.002 µM). With LC3A and the two variants, the K51Q substitution decreased the binding affinity to the WDFY3LIR by almost 24-fold (K_D = 170 ± 15 µM) in comparison to wild-type LC3A (K_D = 7.1 ± 0.6 µM), whereas as acetylation at K49 decreased binding by only 3-fold (K_D = 22 ± 3.4 µM).

Taken together, the results demonstrated that the acetyl-mimetic substitution at the second conserved Lys residue of GABARAP and LC3A significantly reduced the binding to the WDFY3LIR whereas acetylation of the first conserved Lys residue only slightly reduced the affinity. These results for the WDFY3LIR binding to the acetyl-mimetic variants were very similar to the results we obtained with the TP53INP2LIR binding to either the acetylated and acetyl-mimetic variants of GABARAP and LC3A. Taken together, this suggested that the key role of deacetylation of the second conserved Lys residue as compared to deacetylation of the first conserved Lys residue was not specific for binding to the TP53INP2LIR.

Effect of lysine acetylation on colocalization of GABARAP and TP53INP2

TP53INP2 has been previously shown to mediate the nuclear export of LC3B following its deacetylation by SIRT1 under conditions of starvation. However, the role of TP53INP2 in the shuttling of GABARAP from the nucleus to the cytoplasm and the role that deacetylation of Lys residues in GABARAP play on its binding to TP53INP2 in cells has not been examined [14,15]. Given that TP53INP2 displayed selectivity for the GABARAP-subfamily proteins over the LC3-subfamily proteins in our ITC studies, we were interested to determine whether TP53INP2 shuttled GABARAP from the nucleus in response to starvation and verify that deacetylation of K48 plays a key role.

U2OS cells were co-transfected with a vector expressing red fluorescent protein (RFP)-tagged TP53INP2 and a second vector expressing either wild-type GFP-GABARAP or one of the three K-to-Q variants of GFP-GABARAP (GFP-GABARAP^K46Q, GFP-GABARAP^K48Q and GFP-GABARAP^K46,48Q). Under full medium conditions, both RFP-TP53INP2 and either GFP-GABARAP, GFP-GABARAP^K46Q, GFP-GABARAP^K48Q or GFP-GABARAP^K46,48Q were predominantly distributed in the nucleus, and not colocalized to cytoplasmic foci (Figure 8). Upon starvation, GFP-GABARAP and GFP-GABARAP^K46Q showed almost a complete cytoplasmic distribution and colocalization with RFP-TP53INP2 in foci (Figure 8). In contrast, cells expressing RFP-TP53INP2 with either GFP-GABARAP^K48Q or GFP-GABARAP^K46,48Q again showed predominantly a nuclear distribution for both proteins. These results were consistent with the ITC results and demonstrated that deacetylation of the second conserved Lys residue of GABARAP (K48) played a key role in its interaction with TP53INP2 in cells and this interaction was important for its exit from the nucleus and participation in autophagosomes in the cytoplasm.

Figure 8.

Colocalization with TP53INP2 is dependent on the acetylation status of the second conserved Lys of GABARAP. U2OS cells were co-transfected with 1 μg of RFP-tagged TP53INP2 and either 1 μg of wild-type GFP-GABARAP or 1 μg of one its three K-to-Q variants (GFP-GABARAP^K46Q, GFP-GABARAP^K48Q or GFP-GABARAP^K46,48Q). The cells were either grown in full media (top images) or subjected to starvation by incubation in media deficient serum at 37°C for 18–24 h (bottom images) and imaged (left panel). GABARAP and variants are colored green (GFP), TP53INP2 in red (RFP) and the nucleus blue (DAPI: Hoechst 33342). Autophagic foci indicative of the presence of both RFP-TP53INP2 and GFP-GABARAP (WT or K-to-Q variant) for each group of cells were calculated for full media (FM) and starved (ST) cells (right panel). The number of foci per cells (green bars in right panel) were calculated for 20 cells. Images were analyzed with ImageJ/Fiji software and imported via the Bio-Formats plugin. For the cytoplasmic foci, the number of yellow foci (containing both RFP-TP53INP2 and GFP-GABARAP or K-to-Q variant) was counted by hand. The distance for each image is indicated by the black bar at the bottom of the figure. Scale bar: 10 μm.

Given that the RFP-TP53INP2 was predominantly found in the nucleus when we co-transfected with the GFP-GABARAP^K48Q even during starvation, we next attempted to determine the cellular location of LC3A in the cells expressing the GFP-GABARAP^K48Q. To perform these experiments, we co-transfected RFP-LC3A with either GFP-GABARAP, GFP-GABARAP^K48Q or GFP-GABARAP^K46Q and the colocalization was evaluated in full medium or starved U2OS cells (Figure S5). In these experiments, we observed that when co-transfected with either GFP-GABARAP or GFP-GABARAP^K46Q, the RFP-LC3A was predominantly found in the nucleus of full media cells and predominantly colocalized with GFP-GABARAP o GFP-GABARAP^K46Q in cytoplasmic foci in starved cells. In contrast when RFP-LCA is co-transfected with the GFP-GABARAP^K48Q, the RFP-LC3A and the GFP-GABARAP^K48Q were both found predominantly in the nucleus in full media cells as well as in starved cells.

Based on these results, it appears that the GFP-GABARAP^K48Q variant sequestered with another factor that competes for its binding to TP53INP2 and prevented it from leaving the nucleus. Given the fact that LC3A was also not able to exit the nucleus, this factor which was sequestered together with the GABARAP^K48Q variant was a rate-limiting factor for the nuclear/cytoplasmic shuttling of GABARAP and it also played an important role in shuttling of LC3A as well.

Discussion

Growing evidence indicates that selective binding to the LIR motif of autophagy factors plays a key role in defining the individual functions of the six human Atg8 proteins [2,3,23]. As observed with the TP53INP2LIR, the LIRs of ANK3/ANG (ankyrin 3), FYCO1 (FYVE and coiled-coil domain autophagy adaptor 1), PLEKHM1 (pleckstrin homology and RUN domain containing M1), PCM1 (pericentriolar material 1), ULK1 (unc-51 like autophagy activating kinase 1), GABRG2/γ2-GABAAR (gamma-aminobutyric acid type A receptor subunit gamma2), WDFY3 and RETREG1/FAM134B (reticulophagy regulator 1) preferentially bind to the GABARAP subfamily in comparison to the LC3 subfamily [9,10,23,33–38]. These previously identified GABARAP subfamily specific LIRs are classified into two subgroups based on their structures bound to GABARAP: the LIR of ULK1 has been coined the GABARAP-interaction motif/GIM or [W/F]-[V/I]-X2-V motif due to the presence of a Val residue following the aromatic residue in the core [23,33,36], whereas the remaining LIRs contain a C-terminal extension following the core. Of these GABARAP subfamily specific LIRs, the LIR of WDFY3 is the only other example that forms a β-sheet like structure when bound to GABARAP as observed in the structure of the TP53INP2LIR bound to GABARAP [33].

In the TP53INP2LIR, the β-sheet is formed by an intramolecular contact between V32 and I37 in the TP53INP2LIR and an intermolecular interaction between V32 of the TP53INP2LIR and Leu55 in GABARAP. In this β-sheet conformation, V32 is positioned to form a network of hydrophobic interactions with residues in GABARAP that are not conserved in LC3A. In the complex between WDFY3 and GABARAP, the β-sheet formed by the WDFY3LIR is stabilized by a cation-π interaction between Lys3343 and Phe3348 [33]. Interestingly, Lys3343 and Phe3348 of WDFY3 are in the same positions relative to the core of the LIR as V32 and I37 in the TP53INP2LIR. In addition, both LIRs contain a Gly residue in the position immediately before the aromatic residue of the core LIR motif that helps with the formation of the β-sheet.

These similarities between the TP53INP2LIR and WDFY3LIR suggest that formation of the β-sheet helps to position these LIRs to selectively recognize members of the GABARAP subfamily in comparison to members of the LC3 subfamily. In the case of the TP53INP2LIR, V32 is positioned to make contacts with several residue in HP1 (Y5) and HP2 (L55 and F60) that are conserved in GABARAP subfamily, but not in the LC3 subfamily (For alignment highlighting the differences see Figure S6). Thus, TP53INP2 and WDFY3 appear to make up a third group of LIR motifs that specifically target the GABARAP subfamily by combining the formation of a β-sheet conformation with a C-terminal extension. This is further supported by our results demonstrating that like the TP53INP2LIR, only the acetyl-mimetic substitution at the second conserved Lys residue of GABARAP and LC3A significantly reduced their binding to the WDFY3LIR.

It was previously shown that for TP53INP2 to bind to LC3B in the nucleus, both of conserved Lys residues of LC3B (K49 and K51) must be deacetylated by SIRT1 (See Figure 9). This interaction was also shown to be essential for the shuttling of LC3B from nucleus to cytoplasm for autophagosome biogenesis [14,15]. In contrast to what is observed with LC3B, our results with K-to-Q variants of LC3A and GABARAP showed that only deacetylation of the second conserved Lys residue in LC3A (K51) and GABARAP (K48) is required for their interaction with TP53INP2 as well as their subsequent shuttling from the nucleus to the cytoplasm (Figure 9). The cellular studies are supported by the ITC and structural studies of the TP53INP2LIR in complex with GABARAP and LC3A in both the deacetylated and individual acetylated forms.

Figure 9.

Cartoon highlighting the role of Lys acetylation in regulating the nuclear/cytoplasm shuttling of different Atg8-family proteins. During conditions of full medium nutrient status in the cell, the Atg8-family proteins are predominantly located in the nucleus where they are acetylated on two conserved Lys residues (top of image above the straight black dashed line). GABARAP (left scheme) is acetylated at K46 and K48, whereas LC3A (middle scheme) and LC3B (right scheme) are acetylated at the corresponding positions K49 and K51. In response to starvation (below the dashed line), the Atg8 proteins in the nucleus are deacetylated by SIRT1. In the case of GABARAP (left scheme) and LC3A (middle scheme), deacetylation by SIRT1 only at the second Lys (K48 in GABARAP and K51 in LC3A) is required for its binding to TP53INP2 through its LIR motif. In contrast, a previous study has shown that LC3B (right scheme) must be deacetylated on both Lys residues prior to binding to the LIR motif of TP53INP2 [14,15]. Following their binding to TP53INP2, the Atg8-family proteins in complex with TP53INP2 exit the nucleus and enter the cytoplasm (below the solid curved black line) where they are subsequently modified by PE prior to entry into phagophores. Of note, the K48 or K46 and K48 acetylated form of GABARAP inhibits the ability of LC3A to bind TP53INP2 (red bar) and exit the nucleus.

Taken together, our results demonstrated that the TP53INP2LIR preferentially binds to GABARAP subfamily proteins, and that GABARAP and LC3A colocalization with TP53INP2 in the cytoplasm under starvation conditions was dependent on the second conserved Lys residue. The higher affinity for GABARAP-subfamily proteins suggests that during early stages of autophagy induction in response to nutrient deprivation that TP53INP2 would preferentially associate with the GABARAP-subfamily proteins before LC3A, LC3B and LC3C. Such a mechanism would be consistent with previous studies showing that GABARAP-subfamily proteins are crucial for phagophore biogenesis at early stages of autophagy whereas, the LC3-subfamily proteins appear to be more important for expansion of phagophores during later stages [39]. The shuttling mechanism of the different human Atg8-family proteins appears to regulate their relative abundance between the nucleus and cytoplasm based on their acetylation status, and this would provide a mechanism for controlling the timing and initiation of autophagy in response to starvation that would be in part regulated by SIRT1. This is consistent with previous studies showing that overexpression of SIRT1 led to LC3B redistributing into the cytoplasm and the formation of LC3B-containing puncta, whereas addition of the SIRT1 specific inhibitor EX-527 prevents starvation-induced nuclear to cytoplasmic redistribution of LC3B by inhibiting the deacetylation of both of the conserved Lys residues [14,15]. In the case of GABARAP and LC3A, our studies indicate that deacetylation at only the second Lys is required for their binding to TP53INP2 and subcellular redistribution under starvation condition as opposed to both Lys residues as observed with LC3B (Figure 9). This difference in acetylation status may also help provide a mechanism for controlling the timing for the different Atg8-family proteins exiting the nucleus with TP53INP2 or it might suggest that other factors besides TP53INP2 could play an important role in the nuclear/cytoplasmic shuttling of Atg8 proteins depending on their acetylation status.

Materials and methods

Reagents

Chemicals, buffers and purification materials were obtained as follows: QuikChange II XL (Stratagene 200,523), ¹⁵N-ammonium chloride (Sigma-Aldrich 299,251), sodium ampicillin (BioShop, Amp201), isopropyl-b-D-thiogalactopyranoside (IPTG; Inalco, 1758–1400), Glutathione Sepharose 4B resin (GSH; Cytiva, 17-0756-04), High-Performance SP-Sepharose (Cytiva 17,108,701), High Performance Q-Sepharose (Cytiva 17,101,401), Superose 12 10/300GL column (Cytiva 29,036,225), streptomycin sulfate (US Biological, S7975), kanamycin sulfate (Fisher, BP906–5), nicotinamide (Sigma-Aldrich 72,340), Nε-acetyl-lysine (ApexBio, A7231), Chelating-Sepharose resin (Cytiva 17,057,501), nickel sulfate hexahydrate (Sigma-Aldrich, N4882), imidazole (Sigma-Aldrich, I202), dithiothreitol (DTT; Inalco, 1758–9030), 2-(N-morpholino)ethanesulfonic acid (MES; Sigma-Aldrich, M3671), Dulbecco’s modified Eagle’s medium (Wisent, 319–065 CL), fetal bovine serum (Wisent, 090–110), jetPRIME (Polyplus 101,000,046), 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen, D1306) and Hoechst 33342 (Invitrogen; H3570). Oligonucleotides were purchased from Integrated DNA technologies. Vectors used were obtained from the following sources pGEX-2T vector (Cytiva 28,954,653); pCDF-pyIT and pBK-AcKRS3 (Dr. Jason Chin; Cambridge University), pet16B-TEV (DNASU repository, HsCD00671538), pEGFP-C1 and RFP (Dr. Jimmy Dikeakos; Foigu Ontario University). The following bacterial cells were acquired as follows BL21(DE3) (Novagen; 69450), TOPP2 (Stratagene; gift from Dr. Jack Greenblatt, University of Toronto). The U2OS cells were purchased (American Type Culture Collection/ATCC), HTB-96)

Expression vectors

The DNA sequences encoding for for the human Atg8-family proteins (LC3A, residues 1–121; LC3B, residues 1–125; LC3C, residues 8–125; GABARAP, residues 1–117; GABARAPL1, residues 1–117; and GABARAPL2, residues 1–117), TP53INP2LIR (residues 31–43 of human TP53INP2), WDFY3LIR (residues 3341–3355 of human WDFY3) and TP53INP2LIR-LC3A (residues 31–43 of human TP53INP2 fused to residues 2–121 of LC3A) were synthesized as oligonucleotides (Integrated DNA Technologies) with flanking BamHI and EcoRI restriction enzyme sites and cloned into a modified pGEX-2T vector with a Tobacco Etch Virus (TEV) protease cleavage site replacing the original thrombin cut site [40]. All K-to-Q variants of LC3A and GABARAP were generated by site directed mutagenesis starting from the appropriate wild-type construct. For preparation of the acetyl-Lys variants, the codon of the specific Lys residue (K46 or 48 of GABARAP and K49 or K51 of LC3A) to be substituted by acetyl-Lys was replaced by an amber codon (UAG) and cloned into the pCDF-pyIT vector, along with a pBK-AcKRS3 vector [28,29]. These two vectors were kindly provided by Dr. Jason Chin (Cambridge University). For expression of His-tagged TEV protease, a pET16B-TEV vector was used. For expressing GFP-GABARAP and GFP-LC3A in cells, the sequences encoding for human GABARAP and human LC3A were inserted into the pEGFP-C1 vector (kindly provided by Dr. Jimmy Dikeakos; Western Ontario University) using the BglII/EcoRI restriction sites. The Gln variants of GFP-GABARAP (K46Q, K48Q and K46,48Q) and GFP-LC3A (K49Q, K51Q and K49,51Q) and were generated using site-directed mutagenesis with QuikChange II XL to the manufacturer’s instructions. For the RFP-TP53INP2 and RFP-LC3A expression, the sequence encoding the human TP53INP2 and human LC3A were inserted into an RFP vector (kindly provided by Dr. Jimmy Dikeakos, Western Ontario University) using the BglII/EcoRI as restriction sites. All constructs were verified by DNA sequencing.

Peptides and protein expression and purification

LC3A, LC3B, LC3C, GABARAP, GABARAPL1, GABARAPL2, TP53INP2LIR-LC3A, TP53INP2LIR, WDFY3LIR as well as the K-to-Q variants of GABARAP (K48 and K46) and LC3A (K49 and K51) were expressed as GST-fusion proteins in E. coli host strain TOPP2. Cells were grown in Luria Broth (LB) medium or minimal media with¹⁵N-ammonium chloride as the only nitrogen source (for¹⁵N-labeled LC3B) supplemented with 100 mg/mL ampicillin overnight at 37°C. The next day, cells were diluted in LB plus antibiotics and grown until reaching O.D.600 between 0.6–0.8. Protein expression was induced with 0.7 mM IPTG for 4 h at 30°C. Cells were pelleted by centrifugation (15 min at 12,000 × g) and resuspended in lysis buffer (20 mM Tris-HCl pH 7.4, 1 M NaCl, 0.2 mM EDTA, 1 mM DTT) and lysed using a French Press. The resulting suspension was centrifuged (1 h, 105,000 × g, 4°C) and the supernatant incubated (1 h, 4°C) with GSH resin. The GSH resin was then rinsed three times with TEV buffer (20 mM phosphate buffer pH 7.4, 125 mM NaCl, 5 mM DTT) and incubated overnight at room temperature with in-house prepared TEV protease. The next day, the proteins were further purified by ion exchange chromatography using either a High-Performance SP-Sepharose column (LC3A, LC3B, LC3C GABARAP, GABARAPL1, GABARAPL2, TP53INP2LIR-LC3A as well as the GABARAP and LC3A K-to-Q variants) or a High-Performance Q-Sepharose column (TP53INP2LIR and WDFY3LIR). For the ion-exchange chromatography, the columns were equilibrated in 20 mM sodium phosphate buffer pH 6.5 with 1 mM DTT (buffer A) and the proteins eluted with a gradient of 20 mM sodium phosphate buffer pH 6.5 with 1 M NaCl and 1 mM DTT (buffer B). The fractions from the ion exchange columns were then dialyzed into 20 mM sodium phosphate buffer pH 6.5 with 50 mM NaCl and 1 mM DTT and purified by gel filtration chromatography over a Superose 12 10/300GL column and the final purified proteins stored at −80°C prior to usage in ITC and crystallography experiments.

Expression and purification of acetylated variants of GABARAP and LC3A

For the expression of the acetylated forms of LC3A (K49 and K51), TP53INP2LIR-LC3A (K49 and K51) and GABARAP (K46 and K48), BL21(DE3) cells were transformed with modified pCDF PylT plasmid carrying the Open Reading Frame (ORF) for either LC3A, TP53INP2LIR-LC3A or GABARAP with amber codons incorporated at specific sites for incorporation of acetyl-Lys as well as with the pBK-AcKRS3 plasmid (courtesy of Dr. Jason Chin, Cambridge University) for the expression of the tRNA synthetase specific for the incorporation of Nε-acetyl-Lys at the amber codons [28,29]. Cells were grown at 37°C in LB medium supplemented with 50 mg/mL streptomycin and 50 mg/mL kanamycin, diluted 1:4 after overnight incubation. Cells were then supplemented with 20 mM nicotinamide and 10 mM Nε-acetyl-Lys. Protein expression was induced for 4 h at 37°C with 0.7 mM IPTG. The cells were centrifuged and resuspended in lysis buffer (20 mM Tris buffer, pH 7.4, 1 M NaCl and 50 mM imidazole), lysed in a French press and centrifuged at 105,000 × g for 1 h at 4°C. The resulting supernatant was incubated with a Chelating-Sepharose resin charged with nickel for 1 h at 4°C, washed with lysis buffer, and the protein eluted from the resin with a 20 mM Tris buffer pH 7.4 with 1 M NaCl and 500 mM imidazole. The eluted protein was then dialyzed against TEV buffer and the His-tag removed by incubating overnight in 100 units of in-house prepared TEV protease. The proteins were then purified by ion exchange chromatography using a High-Performance SP-Sepharose column equilibrated in 20 mM sodium phosphate buffer pH 6.5 with 1 mM DTT (buffer A) and the proteins eluted with a gradient of 20 mM sodium phosphate buffer pH 6.5 with 1 M NaCl and 1 mM DTT (buffer B). The fractions from the ion exchange columns were then dialyzed into 20 mM sodium phosphate buffer pH 6.5 with 50 mM NaCl and 1 mM DTT and further purified by gel filtration chromatography over a Sepharose 12 10/300GL column and the final purified proteins stored at −80°C prior to usage in ITC and crystallography experiments. The identities were confirmed by MALDI-TOF mass spectrometry facility in the Chemistry Department at the University of Montreal.

ITC experiments

Proteins were dialyzed overnight at room temperature into 20 mM phosphate buffer pH 6.5 with either 50 mM or 150 mM NaCl. Each titration consisted of 25 injections of 12 μL with a spacing of 240 sec between the injection. The Atg8-family proteins were placed in the vessel and the LIR peptides were placed in the syringe for each titration and the molar ratio of the protein concentration to peptide was always 1:10. The protein concentrations ranged between 10–50 μM and the peptide concentrations ranged from 100–500 μM. The protein and peptide concentrations were determined by UV absorbance at 280 nm. ITC measurements were performed at 25°C using a VP-ITC calorimeter (MicroCal). Data were analyzed using Origin Software (Origin; Version 7) and all experiments fit the single binding site model with 1:1 stoichiometry. Standard deviations in K_D values and N-values (stoichiometry) were determined from duplicate measurements or more.

NMR competition experiments

For the NMR competition experiments, an initial¹H-¹⁵N-HSQC experiment was recorded of 0.2 mM of¹⁵N-labeled LC3B in the presence of an equimolar ratio of unlabeled TP53INP2LIR (0.2 mM). Next, an equimolar equivalent of unlabeled GABARAP (0.2 mM) was added to the 1:1 complex of¹⁵N-LC3B-TP53INP2LIR (0.2 mM) and a second¹H-¹⁵N-HSQC experiment was recorded on the three-component mixture in a 1:1:1 ratio. To map the binding site of the TP53INP2LIR on the structure of the LC3B protein, the backbone assignment of LC3B from the Biological Magnetic Resonance Data Bank (accession numbers 26881) was used as a reference to determine the amino acids of LC3B that undergo a significant chemical shift change (Δδ (ppm) + 1 standard deviation; Δδ = [(0.17ΔN_H)² + (ΔH_N)²]^1/2} following the addition of the TP53INP2LIR. These residues were then mapped on the structure of LC3B, PDB: 3VTU. All NMR experiments were carried out at 298 K on a 4-channel Bruker BioSpin 700-MHz spectrometer in 20 mM sodium phosphate (pH 6.5), 50 mM NaCl and 90% H₂O, 10% D₂O and the NMR data were processed with NMRPipe/NMRDraw [41] and analyzed with CcpNMR [42].

Crystallization and data collection

LC3A, TP53INP2LIR-LC3A, TP53INP2LIR-LC3A[K49Ac], GABARAP, GABARAP[K46Ac] and GABARAP[K48Ac] proteins were concentrated to a final concentration between 150–700 μM in 25 mM MES, 50 mM NaCl and 1 mM DTT buffer. For crystallization of GABARAP and the acetylated GABARAP variants with the TP53INP2LIR peptide, the lyophilized TP53INP2LIR peptide was added to yield a protein:peptide ratio of 1:2. The crystals of all the protein:peptide complexes and fusion proteins were generated using the hanging drop vapor diffusion method. The selected crystals were cryoprotected in a mother liquor containing 20% glycerol. Diffraction data collected using Pilatus3S 6 M at beamline 08-ID of the Canadian Light Source (CLS) or either a Pilatus3S 6 M detector or at the beamline IDB7 of the Macromolecular Cornell High Energy Synchrotron Source (MacCHESS). Datasets were indexed, integrated, and scaled using HKL2000 (HKL Research, Inc.).

Structure determination and refinement

Initial phases were obtained by molecular replacement using the previously determined crystal structure of LC3A (PDB:3WAL) or GABARAP (PDB:1GNU) as a search template. Phases were improved by iterative cycles of model building with Coot [43,44] and refinement with PHENIX [6,45]. Test data sets were randomly selected from the observed reflections prior to refinement. The figures were prepared with PyMOL (Schrödinger, Version 2.5).

Cellular localization and colocalization studies

For cellular localization and colocalization studies, U2OS cells (American Type Culture Collection, HTB-96) were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, 1% L-glutamine, and 1% penicillin/streptomycin under 5% CO₂ at 37°C. For the subcellular distribution experiments, transient transfections of 2 μg of GFP-LC3A, GFP-GABARAP and the K-to-Q variants were performed using jetPRIME®️ according to the manufacturer’s instructions (Polyplus). Nuclei were counterstained with 1 μg/mL of Hoechst 33342 for live cell imaging and nucleus were stained with 300 nM of DAPI for fixed permeabilized cells. For colocalization assay, cells were co-transfected with either 1 μg of RFP-TP53INP2 or 1 μg RFP-LC3A as well as with either 1 μg of GFP-GABARP or one of the K-to-Q variants. In all experiments, transfection reagent was removed after 6 h incubation time and cells were allowed to recover for a minimum of 16 h prior to any experiment. To induce autophagy, cells were washed three times with pre-warmed PBS then subjected to starvation by incubation in serum deficient media for 18–24 h under 5% CO₂ at 37°C. Following the incubation period, the cells were fixed in PBS containing 4% paraformaldehyde for 15 min at room temperature and then stored in PBS at 4°C until images were acquired. Images were acquired on scanning confocal microscopes (Zeiss Axio-Observer Z1 and ZEISS LSM 800). Images were analyzed with ImageJ/Fiji software (Fiji; ImageJ version 1.53t free software). For each biological transfection, several areas were randomly selected and imaged. For image analysis with ImageJ, images were imported via the Bio-Formats plugin (Open Microscopy Environments; Version 7.2.0). For nuclear staining, intensity region of interest was selected using Hoechst/DAPI image and mean intensity of either the GFP signal or RFP signal was extracted for 50 cells. For cytoplasmic foci in subcellular localization studies, mask image of GFP signal (background subtracted to visualize individual foci) was created and number of foci was counted using the ImageJ analyze particle’s function for 50 cells and correct identification of foci was verified. For the analyze particle function, the criteria used was 0.2 μM²-100 μM² for particle size with no restriction to circularity. For cytoplasmic foci in colocalization studies, the number of yellow foci (those foci containing both RFP-TP53INP2 or GFP-GABARARP or K-to-Q variant) in each cell was counted by hand for 20 different cells taken from duplicate or more transfection experiments.

Statistical analysis

For the ITC binding experiments, the standard deviations for the K_D and N-values (stoichiometry) were determined from duplicate measurements or more. For the structures of the complexes determined by X-ray crystallography, statistics for the final models were obtained with PHENIX [45,46] and Molprobity [47]. For the nuclear/cytoplasmic distribution experiments, the standard deviations for the number of foci and the nuclear staining intensity (GFP or RFP) were determined based on analysis of 50 cells taken from duplicate or more transfection experiments. For the colocalization experiments, the standard deviations for the number of foci (containing both GFP and RFP) were determined based on analysis of 20 cells taken from duplicate or more transfection experiments.

Funding Statement

The work was supported by the Canadian Institutes of Health Research [130414 and 179806]; Canadian Institutes of Health Research [166104]; Defense Threat Reduction Agency [HDTRA1-22-1-0009].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Abbreviations

DAPI:

4′,6-diamidino-2-phenylindole;

DTT:

dithiothreitol;

EDTA:

ethylenediaminetetraacetic acid;

EP300/p300:

E1A binding protein p300;

FM:

full medium;

GABARAP:

GABA type A receptor-associated protein;

GABARAPL1:

GABA type A receptor associated protein like 1;

GABARAPL2:

GABA type A receptor associated protein like 2;

GFP:

green fluorescent protein;

HP:

hydrophobic pocket;

IPTG:

isopropyl-b-D-thiogalactopyranoside;

ITC:

isothermal titration colorimetry;

LIR:

LC3-interacting region;

MAP1LC3/LC3:

microtubule associated protein 1 light chain 3;

NMR:

nuclear magnetic resonance;

PE:

phosphatidylethanolamine;

RFP:

red fluorescent protein;

SIRT1:

sirtuin1;

ST:

starvation;

TEV:

tobacco etch virus;

TP53INP2/DOR:

tumor protein p53 inducible nuclear protein 2;

UBL:

ubiquitin like;

WDFY3/ALFY:

WD repeat and FYVE domain containing 3;

WT:

wild-type.

Data availability statement

The atomic coordinates and structure factors have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/). The PDB Accession codes are: 8T31 for the TP53INP2LIR-GABARAP complex, 8T32 for the TP53INP2LIR-GABARAP[K48Ac] complex, 8T33 for the TP53INP2LIR-GABARAP[K46Ac] complex, 8T4T for the TP53INP2LIR-LC3A complex, 8T35 for the TP53INP2LIR-LC3A[K51Ac] complex, and 8T36 for the TP53INP2LIR-LC3A[K49Ac] complex.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15548627.2024.2353443.