Identification of a Class of Protein ADP-Ribosylating Sirtuins in Microbial Pathogens

Abstract

Sirtuins are an ancient family of NAD(+)-dependent deacylases connected with the regulation of fundamental cellular processes including metabolic homeostasis and genome integrity. We show the existence of a hitherto unrecognized class of sirtuins, found predominantly in microbial pathogens. In contrast to earlier described classes, these sirtuins exhibit robust protein ADP-ribosylation activity. In our model organisms, Staphylococcus aureus and Streptococcus pyogenes, the activity is dependent on prior lipoylation of the target protein and can be reversed by a sirtuin-associated macrodomain protein. Together, our data describe a sirtuin-dependent reversible protein ADP-ribosylation system and establish a crosstalk between lipoylation and mono-ADP-ribosylation. We propose that these posttranslational modifications modulate microbial virulence by regulating the response to host-derived reactive oxygen species.

Graphical abstract

Figure 1

Macrodomain-Associated Sirtuins Form a Distinct Class within the Sirtuin Family (A) Unrooted phylogram illustrating the relationship between the known sirtuin classes (I–IV and U) and sirtuins found associated with macrodomains. The latter form a distinct class among the sirtuins (highlighted in red). Human sirtuins (SIRT1–7) are highlighted in blue and sirtuins used in this study in orange. Schematic genome arrangements of the macrodomain-linked sirtuins are given underneath the tree. Details regarding class M-specific motifs can be found in Figure S1 and Table S1. (B) Schematic overview and organization of the extended SirTM operons. In addition to the sirtuin/macrodomain, the operon encodes a glycine cleavage system H-like (GcvH-L) protein and lipoate protein ligase A (LplA2). The operon is commonly associated with the ORFs of homologs of the Old Yellow Enzyme (OYE) and bacterial luciferase-like monooxygenase (LLM) family (i, ii, and v). Less frequently, association with only one of the two ORFs can be observed (iii, iv, and vi). The encoded proteins are indicated on top of the schemes and the gene loci of S. aureus (sav0322–sav0327) and S. pyogenes (spyM50865–spyM50870) are indicated underneath the corresponding schemes. (C) Scheme of the functional relationship between the extended operon components.

Figure 2

The LplA2/GcvH-L Pair Resembles Its Canonical Homologs (A) Wild-type, but not mutant LplA2, can lipoylate the GcvH-L protein. The mutations were chosen in analogy to the canonical LplA/GcvH pair of E. coli where they interfere both with the lipoate adenylation and the subsequent lipoate transfer (Fujiwara et al., 2010). (B) Mutations of lipoylation motif residues within GcvH-L impair the lipoate transfer reaction. Mutation of Lys56 interferes with lipoyl attachment, whereas Glu53 is important for recognition by LplA2 (Fujiwara et al., 1991, 2010). Control reactions were carried out using the canonical LplA/GcvH pairs of S. aureus (SauGcvH, SauLplA1) and E. coli (EcoGcvH, EcoLplA). (C) SauGcvH-L and SpyGcvH-L were expressed in the presence and absence of lipoic acid supplementation. In addition, protein synthesis of some samples was interrupted by supplementation with kanamycin 1 hr prior to culture harvesting. The effect of the additives on GcvH-L lipoylation was assessed by immunoblot. For further characterization of the LplA2/GcvH-L pair, see Figure S2.

Figure 3

Class M Sirtuins Lack Deacylase Activity (A) Delipoylation assay performed with SauSirTM on in vitro and in vivo lipoylated SauGcvH-L. The in vitro lipoylation was carried out for 30 min prior to addition of SirTM and NAD⁺. Control samples with in vivo lipoylated SauGcvH-L were treated as “in vitro” samples, however, without addition of SauLplA2 and LA. (B) Delipoylation and debiotinylation assay performed on S. aureus cell extracts. For lipoylation only, dihydrolipoamide S-acetyltransferase (DLAT) could be detected, whereas BCCP, biotin protein ligase (BPL), and pyruvate carboxylase (PC) could be identified as biotinylated. For assays on E. coli cell lysates, see Figure S3A. (C) Delipoylation and debiotinylation assays were performed as in (B), but using human 293T cell extract. For lipoylation only DLAT could be detected, whereas PC, 3-methylcrotonyl-CoA carboxylase (MCC) and propionyl-CoA carboxylase (PCC) could be identified as biotinylated. (D) Debiotinylation assay using recombinant, in vitro modified BCCP as substrate. Free biotin was removed from the initial biotinylation reaction by passing it twice over a desalting column. (E) Deacetylation activity of SirTMs was tested on BL21(DE3)ΔCobB lysates. Human SIRT2 (isoform 2) and SauCobB were used as positive controls. (F) Deacetylase activities of SauSirTM and SpySirTM against a p53-derived peptide were assessed using the SIRT-Glo assay (Promega). Data are background corrected means ± SD of triplicate measurements. (G) Deacetylase activity of sirtuins (compare to F) was tested against penta-acetylated histone H3 (modified residues: K4, K9, K14, K18, and K23).

Figure 4

Crosstalk between Lipoylation and ADP-Ribosylation: The ADP-Ribosylation of GcvH-L by SirTM Is Lipoylation-Dependent (A) Mono-ADP-ribosylation assays of operon and associated proteins were performed with GST-tagged target proteins derived from S. aureus (indicated on top). Proteins were incubated in the presence and absence of SauSirTM. In addition to the Apo-proteins, in vivo lipoylated GcvH-L was tested (right). To control for self-modification of SauSirTM the enzyme was incubated in the absence of target proteins (cntr). The lipoylated protein observed in the LLM samples corresponds by apparent molecular mass to co-purified E. coli DLST. (B) MARylation assays were performed with WT and lipoylation-deficient mutants of GcvH-L. The lipoylation reaction was carried out for 30 min prior to addition of SirTM and NAD⁺. (C) Cross-MARylation assays performed using canonical GcvHs from S. aureus and E. coli as putative target proteins. (D) Wild-type, but not the catalytic mutant N118A, of SpySirTM modifies lipoylated GcvH-L in a time-dependent manner. (E) Comparison of the ADP-ribosyl transferase activities of human SIRT4ΔMTS and T. brucei TmSir2 with SpySirTM. The activities were assessed on previously described and herein identified substrates: glutamate dehydrogenase (GDH), histone H1 (H1), and GcvH-L for SIRT4, TmSir2 and SpySirTM, respectively. Deacetylase activity of TmSir2 was controlled using a p53-derived peptide as substrate (Figure S3D). (F) ADP-ribosyl hydrolase assays were performed with the operon macrodomains (SpyMacro and SauMacro). Control reactions were carried out with the closely related macrodomain proteins MacroD1 (human) and YmdB (E. coli) or in the absence of a macrodomain protein (cntr). (G) Identification of the Macro reaction product by TCL. Protein bound ³²P-ADPr is immobile under the TLC condition (origin), whereas released ³²P-ADPr co-migrates with the PARP1/PARG (human) control. (H) GST-pulldown assay using GST-fused SauGcvH-L as bait. Prior to pull-down SauGcvH-L was either lipoylated or Lpa treated to define the modification status. The molar ratio of proteins (GcvH-L/GST:operonal) was 1:1. Recombinant GST was used as negative control and binding was monitored by IB. (I) ADP-ribosylation band-shift assay performed with (de)modified SauGcvH-L. SauGcvH-L was incubated consecutively with LplA2, SirTM, and Macro under standard assay conditions (lipoylation 30 min, MARylation 1 hr, and de-MARylation 1 hr). Samples were taken after each reaction step and untreated SauGcvH-L served as control. Under the electrophoretic condition, lipoylation leads to an increased migration (►), while MARylation results in a discrete retention (‡). MARylation was confirmed by its reversibility by Macro treatment. See also Figure S4.

Figure 5

Structural Insights into ADP-Ribosylation by SpySirTM (A) Ribbon representation of the overall structure of SpySirTM in complex with NAD⁺ (yellow). The structure follows a typical sirtuin fold comprised of a Rossmann fold domain (purple) and a small bipartite zinc-coordinating domain (light red) tethered together by four loop regions (red). The zinc ion is indicated in blue and the unresolved loop region as a dotted line (light blue). (B) Comparison of the binding sites of SirTM Apo (purple) in complex with ADPr (light red) or NAD⁺ (red). The view corresponds to (A). (C) The activity of selected SpySirTM mutants was tested in a MARylation assay. SpySirTM WT in the presence and absence of GcvH-L lipoylation were used as controls. For detailed information about the residues see Figure S1 and Table S1. (D) Delipoylation assays were performed with in vivo lipoylated SauGcvH-L (see also Figure 2C). The protein was MARylated with SauSirTM and subsequently delipoylation was performed using WT or catalytically impaired (S259A) lipoamidase. (E) MARylation assays were performed using selected GcvH-L mutants. To distinguish the effects of E24A and D27A short (S) and extended (E) autoradiographic exposures are shown. (F) Comparison of residues involved in lipoyl attachment (K56) and MARylation (E24 and D27). SpyGcvH-L structure is shown in black and canonical GcvH of cattle (PDB: 3KLR), pea (PDB: 1DMX), and M. tuberculosis (PDB: 3HGB) in orange, green, and yellow, respectively. Residue numbers for GcvH-L are given. See also Figures S5 and S6.

Figure 6

SirTM Function Is Linked to Oxidative Stress Responses (A) Wild-type C. albicans and mfs1Δ mutants were inoculated onto plates with or without 5 mM H₂O₂ and growth assayed after 2 days of incubation at 30°C. Four independent mfs1 homozygous deletion clones (labeled #1–#4) were tested. Moderate protection from oxidative stress was observed for all four mutant clones. (B) The extent of growth inhibition by H₂O₂ was quantified in 96-well plate-based broth assays (see Supplemental Experimental Procedures). Following a 24 hr incubation at 37°C, percentage of growth was calculated relative to control (no H₂O₂). The experiment was performed independently twice, in biological duplicates (two independent colonies per strain). A similar trend was observed in both experiments. Shown are averages of growth inhibition ± SD for one of the experiments. See also Figure S7.