Article

Nir Hecht1, Ofi

0022-2836/© 2018 Elsevi

Proteasome accessory factor A(PafA) transferase activity makessense in the light of its homology withglutamine synthetase

r Regev2, Daniel Dovrat 1,

Amir Aharoni 1, 2 and Eyal Gur1, 2

1 - Department of Life Sciences, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel2 - The National Institute for Biotechnology in the Negev, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel

Correspondence to Eyal Gur: gure@bgu.ac.ilhttps://doi.org/10.1016/j.jmb.2018.01.009Edited by J. Buchner

Abstract

The Pup-proteasome system (PPS) is a prokaryotic tagging and degradation system analogous in function tothe ubiquitin-proteasome system (UPS). Like ubiquitin, Pup is conjugated to proteins, tagging them forproteasomal degradation. However, in the PPS, a single Pup-ligase, PafA, conjugates Pup to a wide variety ofproteins. PafA couples ATP hydrolysis to formation of an isopeptide bond between Pup and a protein lysine viaa mechanism similar to that used by glutamine synthetase (GS) to generate glutamine from ammonia andglutamate. GS can also transfer the glutamylmoiety from glutamine to a hydroxyl amine in anATP-independentmanner. Recently, the ability of PafA to transfer Pup from one protein to another was demonstrated. Here, wereport that such PafA activity mechanistically resembles the transferase activity of GS. Both PafA and GStransferase activities are ATP-independent and proceed in two catalytic steps. In the first step catalyzed byPafA, an inorganic phosphate is used by the enzyme to depupylate a Pup donor, while forming an acylphosphate Pup intermediate. The second step consists of Pup conjugation to the new protein, alongside therelease of an inorganic phosphate. Detailed experimental analysis, combined with kinetic modeling of PafAtransferase activity, allowed us to correctly predict the kinetics and magnitude of Pup transfer between twotargets, and analyze the effects of their affinity to PafA on the efficiency of transfer. By deciphering themechanism of the PafA transferase reaction in kinetic detail, this work provides in-depth mechanisticunderstanding of PafA, a key PPS enzyme.

© 2018 Elsevier Ltd. All rights reserved.

Introduction

Intracellular protein degradation is essential forproper cell function by contributing to many biologicalprocesses and modulating favorable responses toexternal stimuli [1,2]. At the same time, intracellularprotein degradation must be tightly regulated andproper balance of protein turnover must be main-tained so as to avoid deleterious consequences [3].In eukaryotic cells, regulated protein degradation isconducted primarily by the ubiquitin-proteasomesystem (UPS) [4–7], while in most prokaryotes thisis carried out by a number of proteases that aresimpler and smaller than the 26S proteasome, andwhich usually bind their targets directly [1]. Nonethe-less, post-translational tagging of proteins for degra-dation does occur in some bacterial speciesbelonging to the phyla Actinobacteria and Nitrospira

er Ltd. All rights reserved.

through the actions of the Pup-proteasome system(PPS) [8]. In this system, Pup, a prokaryotic ubiquitin-like protein, is covalently conjugated to lysine sidechains of target proteins, thereby tagging them fordegradation by the proteasome [9].Pup and ubiquitin are functional analogs, rather than

homologous proteins. Unlike ubiquitin, Pup is unstruc-tured. Moreover, poly-Pup chains are only rarelyassembled on pupylated (tagged) proteins in vivo,again in contrast to ubiquitin [10–12]. InMycobacteriumtuberculosis, where the PPS was initially discovered,Pup is a 64 amino acid-containing protein that endswith a di-glycine motif followed by a C-terminalglutamine [9]. The enzyme Dop (Deamidase of Pup)is responsible for deamidation of this glutamine, thusconverting it into a glutamate [13]. The γ-carboxylate ofthis C-terminal glutamate is conjugated to the aminegroup of target lysines by a single ligating enzyme,

J Mol Biol (2018) 430, 668–681

669PafA transferase activity

PafA (Proteasome accessory factor A) [9]. Theconjugation of Pup to protein targets by PafA isthermodynamically favored, as PafA couples isopep-tide bond formation, an uphill reaction on its own, toATP hydrolysis [14]. PafA is the sole Pup ligase, andbeing extremely promiscuous in target binding, it canconjugate Pup to hundreds of protein targets [15,16].Finally, pupylation is reversible, asDop can depupylatepupylated proteins, namely de-conjugate Pup from analready pupylated target via hydrolysis of the isopep-tide bond that links Pup to the target protein [17,18].The two activities catalyzed by Dop obey the sameenzymatic mechanism. Given that deamidation isessentially hydrolysis of the isopeptide bond betweenPup and the ammonia moiety, PupQ deamidation canbe viewed as ammonia depupylation.Despite catalyzing different biochemical reactions,

Dop and PafA are homologous enzymes [13,19]. InM. tuberculosis, PafA andDop share 32% identity and50% similarity, with both belonging to the glutaminesynthetase (GS) superfamily [13,19]. As such, bothPafA and Dop retained the GS active site fold [20](Fig. 1a) and rely on a mechanism that involves theformation of a phosphorylated glutamyl intermediate[14,19,21]. Specifically, both PafA and GS follow atwo-step mechanism in which ATP is initially used tophosphorylate the carboxylate of a γ-glutamyl group,forming an acyl phosphate intermediate, followed bythe downhill displacement of the phosphate by anamine group (Fig. 1b) [14]. In the case of PafA, it hasbeen shown that the acyl phosphate Pup intermediateis relatively stable, as it can be isolated and detectedusing radiolabeled ATP (ATP-[γ-32P]) in the absenceof a pupylation target [14]. In vivo, the lysine side chainon a target protein provides the amine nucleophile forPafA, whereas ammonia serves as a nucleophilein the case of GS. The similarity between the twoenzymes was further emphasized by in vitro experi-ments demonstrating how free ammonia can serveas a PafA substrate, as is also true for GS [12]. Inthis manner, PafA can catalyze the condensationof ammonia and a molecule of Pup presenting aC-terminal glutamate (PupE) to form a Pup with aC-terminal glutamine (PupQ).As a key enzyme in nitrogen metabolism and with

numerous homologs across the tree of life, GS hasbeen extensively researched. It is well establishedthat GS can catalyze three related reactions, namely,the ‘biosynthetic’ reaction (or simply, the forwardreaction), the reverse reaction and the transferasereaction [22,23] (Fig. 1b). The reaction catalyzedhighly depends on whether the nucleotide being usedisATPorADP,whether themetal ion isMn2+ orMg2+,and whether the nucleophile is an ammonium ion,hydroxylamine or water [22]. For example, thetransferase reaction is preferred when Mn2+, ADPand inorganic phosphate are present and hydroxyl-amine serves as the nucleophile. In the first step of thetransferase reaction, a phosphate displaces the

ammonia on the γ-carbon of glutamine, leading tothe formation of anacyl phosphate intermediate. In thesecond step, hydroxylamine displaces the phosphate(Fig. 1b). Hence, in the presence of ADP and Pi,GS catalyzes the transfer of a glutamyl group fromone substrate, ammonia, to another, hydroxylamine[22,23].Recently, it was reported that PafA possesses a

previously unrecognized “transpupylation”activity [24],namely the ability to transfer Pup from one pupylationtarget (Pup donor) to another (Pup recipient). It wassuggested that this ability is of regulatory importanceas it enables PafA to recycle Pup, while shaping theidentity and composition of the pupylome (the intra-cellular pool of pupylated proteins). In describing thisnewly observed PafA activity, a three-stepmechanismwas proposed. In the first step, depupylation of apupylated donor yields PupE and a detagged donor.Next, pupylation of a new recipient occurs via theestablished two-step ATP-dependent pupylationmechanism, as depicted in Fig. 1b. Surprisingly, thismodel revealed an unexpected deviation from theanalogous mechanism employed by GS, despite theestablished structural and mechanistic similaritiesbetween the two enzymes. Specifically, the use of amechanism by PafA like the one employed by GS incatalyzing the transferase reaction (Fig. 1b) wouldpredict Pup transfer between targets, as was indeedconsidered [24]. Moreover, the involvement of a GS-basedmechanism is consistent with the inclusion of aninorganic phosphate in the in vitro reactions reported inthis study [24], and could explain howPafAemploysaninorganic phosphate molecule to displace one pupyla-tion target by another through the formation of an acylphosphate Pup intermediate. However, as ATPseemed to be required for the transfer of Pup betweentwo pupylation targets [24], the parsimonious expla-nation, reflecting the homology of PafA to GS, wasdismissed.In this present study, we provide strong evidence for

rejecting the mechanism previously offered by Zhanget al. [24] for transpupylation by PafA, and insteadprovide evidence for a mechanism analogous to thatemployed in the transferase reaction catalyzed by GS(Fig. 1b). Specifically, we show that an inorganicphosphate, is absolutely essential in the PafA-catalyzed reaction, as it is used in the first step of thereaction to depupylate the donor through the formationof an acyl phosphate Pup intermediate. Next, an ATP-independent nucleophilic attack by a recipient lysineresults in the pupylation of this residue. We propose amodel that accounts for the transfer kinetics of Pupfrom donor to recipient, highlighting the effects of PafAaffinity for the donor and recipient on the efficiency ofthe transferase reaction. The model presented heredraws power from the established homology betweenPafA and GS, an observation which cannot beignored, and provides a better mechanistic under-standing of PafA, the sole Pup ligase.

670 PafA transferase activity

Results

Pup transfer between targets by PafA isATP-independent

PupQ can be viewed as pupylated ammonia, thesmallest possible PafA target. With this in mind, wefirst tested whether PafA can transfer Pup fromammonia to a protein recipient. For this, we usedMycobacterium smegmatis PafA, as it is easilypurified, routinely used in our lab, and shares 97%similarity with the M. tuberculosis PafA; Pup is 98%similar in both species. PanB, a bona fide pupylationtarget [25], was used as a recipient protein. We thustested the transfer of Pup from ammonia to PanB inthe presence of ADP and inorganic phosphate, givenhow the equivalent GS reaction, the transferasereaction (Fig. 1b), is best supported by ADP andinorganic phosphate [22,23]. We found that PanBpupylation indeed occurred under these conditions,albeit much more slowly than did PanB pupylationvia the forward route using PupE and ATP (Fig. 2a).Importantly, the ability of ADP to support the transferof Pup from ammonia to PanB depended on thepresence of an inorganic phosphate (Fig. 2a). To testwhether ATP can also support Pup transfer to PanB,ATP was used instead of ADP. (Fig. 2b). In thisreaction, pyruvate kinase (PK) and phosphoenolpyr-uvate (PEP) were included as an ATP regenerationsystem, in order to phosphorylate potential ADPcontamination that may be present in our ATP stock.The results indicated that ATP was not able tosupport Pup transfer, in contrast to a previous report[24]. In a control reaction, PanB was efficientlypupylated when PupE was used, indicating that theaddition of PK and PEP did not inhibit PafA.

An acyl phosphate Pup intermediate is formedduring the PafA transferase reaction

For PafA to transfer Pup from a donor to a recipient,depupylation of the donor must first take place. Todetermine which nucleotides best support this firststep of the reaction, we used pupylated 5-FAM lysine(Pup-Fl) [26] (Fig. 2c), as its depupylation can beeasily and quantitatively followed by fluorescenceanisotropy. Upon Pup-Fl depupylation, the freedomof the 5-FAM lysine moiety to rotate in solutionincreases, which translates into a measurable de-crease in its fluorescence anisotropy. In this study,Pup-Fl depupylation by PafA was monitored usingATP, ADP and AMP, with or without inorganicphosphate. Whereas AMP failed to support thereaction, ADPsupported a six-fold faster depupylationrate than recorded using ATP (Fig. 2d&e). Possibly,the slow rate observed with ATP results from ADPcontaminations in the ATP stock. It should be notedthat in a previously reported study, PafA-catalyzed

depupylation of a fluorescent Pup derivative was notdetected [24]. We attribute the apparent discrepancybetween our findings and those of this previous studyto differences in the sensitivity of the depupylationassaysused in eachcase. In the presenceofADPandinorganic phosphate, Pup-Fl depupylation catalyzedby Dop occurred some 370-fold faster than whencatalyzed by PafA (Fig. 2d). In contrast, Dopdepupylated bona fide substrates much slower, andat a rate comparable to that of PafA (Fig. 3a). Indeed,while Pup-Fl depupylation by Dop was completed in5 min (Fig. 3b), more than an hour was required forPup-PanB and Pup-FabD depupylation by both Dopand PafA. Pup-IdeR depupylation was even slower(Fig. 3a), indicating that depupylation by both en-zymes is sensitive to the identity of the conjugatedtarget. These results are consistent with our previousindication that Dop interaction with the isopeptidebond is limited by large conjugated targets, but not by5-FAM-Lys or conjugated ammonia [27].Importantly, the presence of an inorganic phos-

phate was absolutely necessary to support Pup-Fldepupylation by PafA, regardless of the nucleotideused (Fig. 2d&e). This, combined with the finding thatADP best supports the reaction, makes perfect sensein view of the homology between PafA and GS, andthe known mechanisms of the GS-catalyzed reverseand transferase reactions (Fig. 1b). We, therefore,hypothesized that PafA could use inorganic phos-phate as a nucleophile to depupylate a pupylateddonor (e.g., PupQ and Pup-Fl), leading to theformation of an acyl phosphate Pup intermediate. Totest this possibility, depupylation reactions containinga mixture of 32P and non-labeled inorganic phos-phate, ADP, PupQ and PafA were organized. Aliquotswere removed at intervals and separated by SDS-PAGE, followed by Coomassie staining and autora-diography. PafA-dependent formation of a radioactiveprotein was detected, which migrated in SDS-PAGEasdid PupQ (Fig. 4a). This result strongly supports ourhypothesis, and indicates that donor depupylation byPafA proceeds through the formation of an acylphosphate Pup intermediate. These findings furtherindicate that the recently reported transpupylaseactivity of PafA is analogous to GS transferaseactivity, and follows similar mechanistic principles.

Intermediate formation is the rate-limiting stepof the transferase reaction

When Pup-Fl depupylation by PafA in the presenceof ADP and inorganic phosphate was monitored, theaddition of the Pup recipient PanB did not change therate of the reaction (Fig. 2d&e). As steady state rateswere measured, these results indicate that the rate ofPup transfer from 5-FAM lysine to PanB is limited bythe formation of the acyl phosphate Pup intermediate.These results further indicate that in the absence of arecipient, the intermediate does not remain bound to

(a)

(b)

Fig. 1. PafA belongs to the GS superfamily. (a) The crystal structure of PafA (purple) in complex with PupE (red) and ATP(PDB 4BJR) is shown on the left. The cradle-shaped β-sheet-based active site is shown in dark purple. On the right, a zoom-inview of the PafA active site is shown super-imposed on the structures of YbdK (blue; PDB 1R8G), gamma-glutamylcysteinesynthetase (gamma-GCS; green; PDB 1VA6) and GS (yellow; PDB 1F52), all members of the GS superfamily. Selectedconservedcatalytic residuesare shown. * denotes that theAsn in the protein used for crystallizationwas replacedby the nativeresidue (Asp) in Chimera using the rotamer library tool [36]. ** denote that this Asp is not shown, as a neighboringmonomer inGS contributes this residue to the active site. (b) Both PafA (top) and GS (bottom) employ a two-step catalytic mechanism,which involves the formation of an acyl phosphate intermediate. In the upper diagram, R is an amine-presenting PafA target.According to thismechanism, in addition to the forward reaction,GS can also catalyze a reaction in the reverse direction and atransferase reaction, where hydroxylamine can serve as recipient for the glutamyl group.

671PafA transferase activity

the enzyme.Otherwise, eachenzymemoleculewouldhave been able to catalyze only a single turnover,despite the excess of substrate in these reactions. To

further explore this issue, Pup-Fl depupylation inthe presence of ADP and inorganic phosphatewas monitored using the anisotropy-based assay

(a)

(c)

(e)

(d)

(b)

Fig. 2. ADP and inorganic phosphate best support PafA-mediated Pup transfer between targets. (a) Pup transfer to PanB(10 μM) mediated by PafA (1 μM), using PupQ (20 μM) as donor and ADP (5 mM) was assayed over a 5 h period with orwithout addition of inorganic phosphate (50 mM). Aliquots were removed at the indicated time points for SDS-PAGE analysisfollowed by Coomassie brilliant blue staining. For comparison, a standard pupylation reaction containing PupE (20 μM) andATP (5 mM) is also shown. (b) as in a, except ATP (0.2 mM) was used as the sole nucleotide. PK (10 U/mL) and PEP (2 mM)were added for ADP depletion and ATP regeneration. (c) Pup-Fl contains a pupylated 5-FAM lysine, namely a lysine (blue)linked in a peptide bond to 5-carboxyfluorescein (red). (d) Steady state rates of Pup-Fl (4 μM) depupylation byDop (1 μM; redbar) and PafA (1 μM; blue bars) in the presence of the indicated nucleotides (5 mM each) with or without inorganic phosphate(50 mM).WhenPanBwas included, its final concentration was 10 μM.Averages and standard deviations of three repeats arepresented. The inset shows representative curves of 5-FAM lysine fluorescence anisotropy as a function of time, in the sameconditions as used to determine the steady state rates presented in the bar graph. (e) End point (5 h) aliquots were removedfrom the reactions described in d for SDS-PAGE analysis followed by Coomassie brilliant blue (CBB) staining (left) and in-gelfluorescence detection (right).

672 PafA transferase activity

employed above at increasing PafA concentrations.At higher PafA concentrations, higher initial Pup-Flanisotropy resulted, owing to Pup-Fl binding by theenzyme (Fig. 4b). Nevertheless, the anisotropy in allreactions, other than that lacking enzyme, eventually

approached the same value, a value attributed to non-conjugated 5-FAM lysine anisotropy (Fig. 4b). Thisindicates that, given enough time, Pup-Fl depupyla-tion occurs to a similar extent, regardless of theenzyme concentration. It, therefore, appears that

PanB

PafA

PafA Dop

Pup-PanB

Dop

FabD

PafA

Pup-FabD

Dop

PafA

IdeR

Pup-IdeR

Dop

0 1 211

2 0 1 2 h

0 5 10 300 5 10 30 min

1

2

51

39

28

51

39

28

51

39

28

19

PupE

Pup-Fl

Dop

5-FAM Lys

in-gel

e

CBB

28

39

51

19

14

(a) (b)

Fig. 3. Depupylation of protein tar-gets byPafA andDop. (a)Depupylationof Pup-PanB (5 μM), Pup-FabD (5 μM)and Pup-IdeR (contaminated with~equimolar non-conjugated IdeR;10 μM total) by PafA and Dop (1 μMeach) was compared in vitro in reactionbuffers containing ADP (5 mM) and

inorganic phosphate (50 mM). Samples were removed at the indicated time points for SDS-PAGE analysis followed byCoomassie brilliant blue (CBB) staining. (b) Pup-Fl (4 μM)depupylation byDop (1 μM) under the same conditions andusing thesame enzyme stock as in the experiment described in panel “a”. Following electrophoresis, in-gel fluorescence was detected(left panel) followed by CBB staining (right panel).

673PafA transferase activity

upon catalyzing Pup-Fl depupylation in the presenceof ADP and inorganic phosphate, PafA is not“clogged” with the acyl phosphate Pup intermediate,even in the absence of a recipient. In an experimentwhere PupQ served as donor, formation of theintermediate was not followed by PupE formation(Fig. 4a). Thiswas readily detectable, given howPupQ

and PupE differentlymigrate in SDS-PAGE [13,15,28](Fig. 2a). As such, rather than being spontaneouslyhydrolyzed, the intermediate is instead stable and canleave the enzyme, thus enabling multiple turnovers tooccur, even in the absence of a recipient.We next asked whether the rate of intermediate

formation under the conditions used here for monitor-ing the transferase activity of PafA was limited by theavailability of inorganic phosphate. To this end, initialrates of Pup-Fl depupylation were measured in thepresence of ADP and increasing concentrations ofinorganic phosphate. A cooperative binding modelbest fit the data, yielding an apparent binding constant(Kapp) of 6 ± 0.6 mM and a Hill coefficient of 1.6 ± 0.2(Fig. 4c). These results indicate that in the transferaseand depupylation assays depicted in Figs. 2 and 4a,PafA was saturated with inorganic phosphate, as thephosphate concentrations used (50 mM) greatlyexceeded the apparent binding constant. In a similarmanner, the affinity of PafA to ADPwas determined tobe 51 ± 2 μM (Fig. 4d), a value much lower than theconcentration used in our experiments. Finally, todetermine the catalytic rate (kcat) of intermediateformation by PafA in the presence of ADP andinorganic phosphate, aMichaelis–Menten experimentwas performed in which increasing Pup-Fl concentra-tions and saturating phosphate concentrations wereused to measure initial depupylation rates. Here aswell, a cooperativebindingmodel fit the datawith aK0.5of 0.90 ± 0.07 μM, a Hill coefficient of 2.4 ± 0.6 and akcat of 0.023 ± 0.001 min−1 (Fig. 4e). These findingsindicate that PafA catalyzes Pup-Fl depupylation with

a catalytic rate that is 40–50-fold slower than thecatalytic rate previouslymeasured for target pupylationby PafA (~1 min−1) [14,29]. These results, togetherwith those found in previous reports [29], furtherindicate that PafA binds Pup-Fl and PupE with similaraffinity, and are consistentwith a recent report showingthat a pupylated protein and PupE bind PafA equallywell [30].

The transferase reaction reaches an equilibriumdictated by donor and recipient affinity to PafA

To become a Pup donor, a molecule first needs tobe a pupylation target, namely to accommodate thePafA target-binding site. The same holds true for amolecule to become a recipient. Therefore, pupylationtargets can serve both as Pup donors (whenpupylated) and as recipients. Accordingly, IdeR, abona fide pupylation target, could receive Pup fromPup-Fl (Fig. 5a), while pupylated IdeR acted as adonor in a parallel reaction in which PanB served as arecipient (Fig. 5b). Not surprisingly, PafA, itself a bonafide pupylation target [16,31], was also pupylated inthis reaction, albeit to a lower extent (Fig. 5b). Aspupylated targets bind PafA as well as Pup [30](Fig. 4e), one would expect different pupylated targetsto possess similar propensities to serve as donors. Onthe other hand, based on differential affinity to PafA ofvarious recipients, the identity of a recipient isexpected to greatly affect the efficiency of transfer.Specifically, once the intermediate has formed, theefficiency of transfer is predicted to be determined bythe affinity of the recipient to PafA, with higher affinitypupylation targets expected to be better Pup recipi-ents. It is also expected that when a transferasereaction is conducted in vitro, a depupylated donormay again become a recipient and compete with thattarget protein initially added as recipient. Here,pupylation of the latter will depend on its concentration

(a)

(b)

(d)

(c)

(e)

0 0.5 1.5 3 0 0.5 1.5 3 0 0.5 1.5 3 0 0.5 1.5 3 h

+PafA

CBB phosphorimage

-PafA +PafA -PafA

75

60

45

35

25

15

10

Pup*

PafA

time (h)

0 0.5 1.0 1.5 2.0

Flu

orescence a

nis

otr

opy

0.08

0.10

0.12

0.14

0.160

[PafA] ( M)

1

2

3

4

Pup-F

l depupyla

tion r

ate

(m

in-1 p

er P

afA

)

0.00

0.01

0.02

0.03

[Pi] (mM)

Kapp

= 6 ± 0.6 mM

n = 1.6 ± 0.2

0 10 20 30 40 50

K0.5

= 0.90 ± 0.07 M

kcat

= 0.023 ± 0.001 min-1/PafA

n = 2.4 ± 0.6

Pup-F

l depupyla

tion r

ate

(m

in-1 p

er P

afA

)

Kapp

= 51 ± 2 M

n = 1.00 ± 0.04

[ADP] ( M)

0 200 400 600

0.00

0.01

0.02

0.03

Pup-F

l depupyla

tion r

ate

(m

in-1 p

er P

afA

)

[Pup-Fl] ( M)

0 2 4 6 8

0.00

0.01

0.02

0.03

Fig. 4. PafA can use inorganic phosphate to catalyze the formation of an acyl phosphate Pup intermediate. (a) PupQ

(20 μM) was incubated in reaction buffer together with PafA (4 μM), ADP (5 mM) and inorganic phosphate (10 mM)containing 50 μCi 32P–phosphate. Aliquots were removed at the indicated time points for SDS-PAGE followed byCoomassie brilliant blue (CBB) staining (left) and autoradiography (right). Pup⁎ denotes PupQ and phosphorylated Pup.(b) Depupylation of Pup-Fl (4 μM) over time was measured at increasing PafA concentrations, as indicated, in thepresence of ADP (5 mM) and inorganic phosphate (50 mM). (c) Steady state rates of Pup-Fl (4 μM) depupylation by PafA(1 μM) in the presence of ADP (5 mM) and increasing concentrations of inorganic phosphate. Averages and standarddeviations of three repeats are presented. The solid curve was generated by fitting a cooperative binding model (the Hillequation), where n is the Hill coefficient. (d) As in C, except that the inorganic phosphate concentration was kept constant(50 mM) and ADP was titrated. (e) As in C, except that the inorganic phosphate concentration was kept constant (50 mM)and Pup-Fl was titrated.

674 PafA transferase activity

and affinity to PafA, relative to that of the depupylateddonor.This scenario is schematically represented in the

diagram presented in Fig. 5c. According to thisdiagram, either of two different pupylated proteins,Pup-A and Pup-B, can lead to the formation of the

intermediate Pup-Pi with the same efficiency. The rateof Pup-Pi formation in this case is dictated by theconcentrations of Pup-A and Pup-B, by Kp, the affinityof Pup to PafA (1 μM [29,30] (Fig. 4e)), and by k1, thecatalytic rate of depupylation by PafA (0.023 min−1

(Fig. 4e andTable 1)). The intermediate can then react

675PafA transferase activity

with either A or B at a rate that depends on the affinityof these targets to PafA (KA and KB, respectively), ontheir concentrations, andon k2, the catalytic rate of thisstep of the reaction. In fact, this step is identical to thesecond step of a standard pupylation reaction, and

(a)

(c)

(d)

(f)

Fig. 5 (Legend o

was previously measured to occur with a catalytic rateof ~1 min−1 [14] (Table 1). Under these conditions,competition for PafA binding among pupylated andnon-pupylated targets exists, as all accommodate thePafA target-binding site. Likewise, pupylated targets

(e)

(g)

(b)

n next page)

Table 1. Kinetic parameters associated with PafA catalysis.

Parameter Numeric value Description Reference

k1 0.023 ± 0.001 min−1 The catalytic rate of intermediate formation by PafA This workk2 0.95 ± 0.1 min−1 The catalytic rate of Pup transfer from the acyl phosphate Pup

intermediate to a target[14]

KP ~ 1 μM The affinity constants of free Pup and pupylated targets to PafA [14,29,30,this work]KPanB 21 ± 2 μM The apparent affinity constant of PanB to PafA [29]KIdeR 220 ± 50 μM The apparent affinity constant of IdeR to PafA [12]Kammonia 90 ± 4 mM The apparent affinity of ammonia to PafA [28]KLys 22 ± 10 mM The Km of lysine to PafA [14]

676 PafA transferase activity

and the intermediate accommodate the Pup-bindingsite on PafA and, therefore, also compete for enzymebinding.To test the validity of this depiction of the PafA

transferase reaction, we chose to model it using thekinetic parameters listed in Table 1, and to thencompare the results of such modeling with experi-mental results (Materials and methods and Supple-mentary File 1). In the first simulation, we set theconcentrations of the various components to matchthose included in the experiment depicted in Fig. 5a.The Km for 5-FAM lysine was set at 22 mM, the valuepreviously measured for free lysine [14], and the Kmfor IdeR was set at 220 μM [12] (SupplementaryFile 1). Remarkably, this simulation yielded kineticsthat were in good agreement with the resultspresented in Fig. 5a (Fig. 5d). The simulation alsodemonstrated that the reaction approaches equilib-rium, with the rate of IdeR pupylation balancing therate of Pup-IdeR depupylation. In contrast, Pup-Fl isalmost completely depupylated, owing to the verylow affinity of 5-FAM-Lys to PafA. The moderateaffinity of IdeR to PafA allows substantial accumu-lation of the intermediate.Next, we simulated a transferase reaction in which

Pup is delivered from Pup-IdeR to PanB, as in Fig. 5b,using a Km of 21 μM for PanB [29] (SupplementaryFile 1). Again, the simulation yielded transferasekinetics that were in good agreement with theexperimental results presented in Fig. 5b (Fig. 5e).Minor inconsistencies between the simulation andexperimental data inevitably exist, since, for instance,in the real reaction, PafA is also pupylated, given how

Fig. 5. The transferase reaction reaches an equilibrium deter(10 μM) was incubated with Pup-Fl (10 μM) and PafA (1 μMphosphate (50 mM). Aliquots were removed at the indicated timstaining. (b) As in a, except that mixture of pupylated and non-puinstead of IdeR and Pup-Fl. The asterisk denotes an unidedemonstrated [12]. (c) A scheme representing Pup transfer byformation of an acyl phosphate Pup intermediate (Pup-Pi). The r(0.023 min−1/PafA; this study), and k2 is the catalytic rate of isopeconstant between PafA and Pup (~1 μM [29]),KA is the binding cbetween PafA and B. (d) Simulation of Pup transfer by PafAand 10 μM IdeR. The rate and binding constants used are listed iPup-IdeR, 5 μM IdeR and 10 μM PanB. (f) As in d, assumingX corresponds to a target protein. (g) As in d, assuming initial co

PafA itself can serve as a pupylation target [16,31](Fig. 5b). Importantly, the simulation demonstrateshow equilibrium is reached in this reaction, with therates of Pup-IdeR and Pup-PanB depupylation beingequal to the rate of their formation, reflecting thereversibility of the transferase reaction (Fig. 5e). In thereactions presented inFig. 5a&b, and in the respectivesimulations (Fig. 5d&e), only a fraction of the recipientis found pupylated at equilibrium. For a recipient to bepupylated to a higher extent, its affinity to PafA mustbe relatively high, while the donor must have pooraffinity to PafA. For instance, when PupQ is used, thedonor is ammonia, a poor PafA target, with anapparent affinity of 90 mM [28] (Table 1). Supposingthat the recipient is a protein that binds PafA aswell asdoes Pup (Km = 1 μM [29]), about 80% of this highaffinity PafA target is expected to become pupylated(Fig. 5f). To date, however, no target that binds PafAbetter than PanB has been identified. Accordingly,PanB pupylation based on Pup transfer from PupQ toPanB occurred to a lower extent both when simulatedand when experimentally assessed (Figs. 2a and 5g).

Discussion

In this work, we investigated the molecular mech-anism that enables PafA to transfer Pup from onetarget to another. By characterizing this mechanism inkinetic detail, we not only gained comprehensiveunderstanding of this activity, but also obtained theability to model and predict the kinetics of Pup transferbetween targets of known affinities to PafA.Moreover,

mined by the affinity between PafA and its targets. (a) IdeR) in reaction buffer containing ADP (5 mM) and inorganice points for SDS-PAGE followed by Coomassie brilliant bluepylated IdeR (10 μM total) and PanB (10 μM) were includedntified band, potentially a pupylated Pup as previouslyPafA from target A to target B and vice versa through theate constant k1 is the catalytic rate of intermediate formationptide bond formation (~1 min−1/PafA [14]).Kp is the bindingonstant between PafA and A, and KB is the binding constant(1 μM), assuming initial concentrations of 10 μM Pup-Fl

n Table 1 (e) As in d, assuming initial concentrations of 5 μMinitial concentrations of 10 μM PupQ and 10 μM X, wherencentrations of 10 μM PupQ and 10 μM PanB.

677PafA transferase activity

the agreement of such modeling with experimentalresults strengthened the validity of the model pro-posed here as describing the mechanism of Puptransfer by PafA.Although the transferase activity of PafA was

previously identified [24], the mechanism proposed

(a)

(b)

O

O- + ATP

C-term

glutamate

Pu Pp up

Pup

Pup- -glutamy

phosphate

pupylated targe

+

NH2-R2

R2R2-NH2

OPO3

2-

O

O

Pup

+ADP + Pi

target

Pup

PupE +

ATP +

target

reaction pr

Gib

bs free e

nergy

4

3

3

Fig. 6. The reactions catalyzed by PafA and Dop. (a) SchemPafA reactions, all involving the formation and decomposition ois currently putative, and would be expected to occur considercoordinate diagram for the pupylation and transferase reactioreaction catalyzed by Dop. The numbers in blue correspondhydrolysis yields 7.3 kcal/mol [34], peptide bond hydrolysis yie11.8 kcal/mol [34]. The peak heights were set arbitrarily, as thunknown. The peaks in the transferase reaction are symmetric,step of the pupylation reaction is identical to the second step osame energetic profile. As the first step of the pupylation reactioreaction [14], both steps should have a similar activation energystep of the transferase reaction, and, therefore, has the samreaction was set to have the lowest activation energy, to accouwith transpupylation.

to account for Pup transfer between targets isinconsistent both with our results and with thehomology between PafA and GS. Specifically, weshowed that ATP hydrolysis is not required for Puptransfer byPafAnor isPupE formedduring the reaction(Fig. 6a). Rather, the enzyme utilizes inorganic

Pu p

l

pupylated target 1

t 2

AD +P ADP + Pi

+ ADP + Pi

NH2-R1

R1-NH2

R1-NH2

-NH2

O

+ADP + Pi

PupE + target

ADP + Pi +

Pup target

Pi

ogression

transferase

.5

.6

.7

pupylation

depupylation

atic representation of the forward, reverse and transferasef an acyl phosphate Pup intermediate. The reverse reactionably slower than the transferase reaction. (b) A schematicns catalyzed by PafA, and the depupylation/deamidationto ΔG°‘ values in units of kcal/mol, assuming that ATP

lds 3.6 kcal/mol [37], and acyl phosphate hydrolysis yieldse activation energy of each enzyme-catalyzed reaction isas this reaction is identical from each direction. The secondf the transferase reaction. Therefore, both steps have then has a rate only slightly slower than the second step of this. The first step of depupylation by Dop is identical to the firste energetic profile. The second step of the depupylationnt for the 370-fold faster rate of depupylation, as compared

678 PafA transferase activity

phosphate to facilitate Pup transfer between targets, inperfect analogywith thewell characterized transferaseactivity of GS (Figs. 1b & 6a). Accordingly, the parallelPafA activity should be regarded as transferaseactivity, rather than ‘transpupylation’, as recentlysuggested [24].Still, caution should be exercised when generaliz-

ing in vitro results, as even under conditions thatsupport the transferase activity of PafA, this reactionoccurs with a catalytic rate about 40–50-fold slowerthan the catalytic rate of the forward pupylationreaction. In vivo, the PafA-mediated transferasereaction is expected to be disfavored because of,for instance, ATP concentrations being higher thanADP concentrations (the [ATP]:[ADP] ratio is 10:3)[32,33]. It is thus safe to presume that the activity ofthe forward reaction greatly exceeds the transferaseactivity of the protein. As such, the prevalence, andeven the physiological importance, of PafA transfer-ase activity in vivo is doubtful.While pupylation is a downhill reaction, no net

free energy is released in the transferase reaction(ΔG°‘= 0), as this reaction essentially involves thereplacement of one isopeptide bond by anotherthrough the formation and breakage of an acylphosphate bond (illustrated in Fig. 6b). As a result,the transferase reaction is fully reversible,meaning thatonly a fraction of Pup transfer occurs with most bonafide targets. A coordinate diagramof the pupylationandtransferase reactions (Fig. 6b) also demonstrates howthe energetic differences between these reactions lie inthe intermediate formation step. Owing to ATPhydrolysis, the activation energy of intermediateformation during pupylation is 3.7 kcal/mol lower thanduring transferase activity (Fig. 6b). This is consistentwith themeasured40–50-fold difference in the catalyticrates of both reactions. There is, however, room topresume that alongside the forward and transferasereactions, PafA can also catalyze, in principle, thereverse reaction (Fig. 6a). However, this reaction,which includes ATP formation, would be non-spontaneous and highly thermodynamically unfavor-able under standard conditions.When deamidation/depupylation is considered, the

difference between PafA-mediated and Dop-mediatedcatalysis is seen in the second step of the reaction. ForDop, the second step consists of PupE formationthrough hydrolysis of the acyl phosphate bond, yieldinga free energy (ΔG°‘) of about 11.8 kcal/mol [34](Fig. 6b). This renders the entire depupylation reactioncatalyzedbyDopenergetically favorable (ΔG°‘b 0). Fordeamidation/depupylation by PafA, this mechanisticstep does not exist. Rather, a phosphorylated Pup anda depupylated target are the end products of an uphillreaction (ΔG°‘= 8.2 kcal/mol; Fig. 6b). Therefore, asPafA catalyzes deamidation/depupylation involving theformation of an acyl phosphate Pup, the reaction islikely to gobackwardsevenbefore the target leaves theenzyme, leading to the observed slow rates of net

deamidation/depupylation by PafA. In contrast, follow-ing Dop-catalyzed intermediate formation, the forwardsecond step, namely acyl phosphate hydrolysis, isenergetically more favorable than is going back andpupylating the target. These differences between thetwo enzymes translate to the 370-fold difference inPup-Fl depupylation rates measured here (Fig. 2d).These findings stand in apparent disagreement withthe slow protein depupylation by Dop (Fig. 3 and [24]).However, this likely reflects the poor affinity Doppresents towards many pupylated proteins, as op-posed to its high affinity to PupQ, which is in thenanomolar range, as we recently showed [27]. Theseaffinity differences allow Dop to facilitate proteinpupylation by efficiently catalyzing PupQ deamidationwithout canceling PafA activity via extensive depupyla-tion of tagged cellular proteins. In accordance, unfold-ing of a pupylated protein by the proteasome regulatorysubunit can greatly enhance the rate of Dop-catalyzeddepupylation [18]. Froma catalytic standpoint, it seemsthat Dop, a PafA homolog, acquired the ability tohydrolyze the acyl phosphate bond of the Pupintermediate over the course of evolution, making itan efficient deamidase/depupylase, as compared toPafA.AsPafA is amember of theGSsuperfamily, thiswork

relied heavily on the accumulated structural andbiochemical data on GS (reviewed in [22]), as well ason the proven mechanistic similarity between the twoenzymes [14,19,20]. In addition to analyzing themechanism of PafA-mediated catalysis, this workalso exemplifies the importance and validity of ad-dressing biochemical functions from the evolutionarycontext. It is, therefore, only appropriate to concludethis paper with a quote attributed to TheodosiusDobzhansky: "Nothing in biology, (molecular andorganismic) makes sense except in the light ofevolution" (1964, 1973).

Materials and methods

Protein expression and purification

Recombinant M. smegmatis PafA, PanB and Pupvariants were purified as described [29], Dop andPup-Fl were purified as described [26], as was IdeR[12]. N-terminal polyhistidine tagged M. tuberculosisFabD that presents arginine substitutions of lysines35, 122, and 291 was cloned following the sameprotocol used for IdeR purification [12].For generation and purification of pupylated PanB,

IdeR and FabD, aCorynebacterium glutamicum PafA(cgPafA) was used that presents a N-terminal poly-histidine tag followed by a TEV protease sequence.cgPafA was purified using the same protocol that wasused for purification of M. smegmatis PafA, exceptfollowing elution from the Ni++-NTA beads, the

679PafA transferase activity

imidazole in the buffer was removed via a bufferexchange step using a PD10 column (GE Health-care), and the TEV protease was added at aTEV/PafA ratio of 1:100 (w/w). Following a 6 hincubation, the protein solution was loaded on a pre-washedNi++-NTAcolumn, and the cgPafA-containingflow-through was collected and loaded on a Super-dex200 column (GE Healthcare) pre-washed with abuffer containing 50 mM Hepes pH 7.5, 500 mMNaCl and 1 mM DTT.PanB, IdeR and FabD were expressed and purified

as described above. However, following elution fromthe Ni++-NTA beads, the buffers were exchangedusing PD10 columns (GE Healthcare) into pupylationbuffers. For IdeRandFabD, a pupylation buffer lackingglycerol was used. Next, cg-PafA and PupE wereadded to a final concentration of 2.5, and 200 μM,respectively. Following a 6 h incubation at 30 °C,standard Ni++-NTA purifications were performed toremove cgPafA and PupE, as these proteins lack apolyhistidine tag. The eluted pupylated proteins werefurther purified by size-exclusion chromatographyusing a Superdex200 column (GE Healthcare) pre-washed with a buffer containing 25 mMHepes pH 7.5and 300 mM NaCl. For PanB, glycerol (10% v/v) wasincluded.

In vitro assays

The buffer used for all reactions contained 50 mMHepes (pH 7.5), 20 mM MgCl2, 150 mM KCl, 1 mMDTT and 10% (v/v) glycerol. Sodium phosphate buffer(pH 7.5), where added, was a source of inorganicphosphate. For phosphorimaging-based identificationof the acyl phosphate Pup intermediate H3[

32P]O4(40 nM; 50 μCi; 17 Ci/μmol; Perkin Elmer) were usedwith 10 mM of Na2PO4 (pH 7.5). Samples wereanalyzed by electrophoresis on a 12%polyacrylamideBis-Tris gel followed by Coomassie brilliant bluestaining. The gel was dried, exposed to a storagephosphor screen (GE Healthcare) and scanned usinga PhosphoImager (FujiFilm). Fluorescence anisotro-py measurements and in-gel fluorescence detectionwere carried out as described [26].When titration assays were performed, a coopera-

tive binding model was fitted to the data using theequation v = Vmax·[S]

n/(Kn + [S]n), where v is thereaction rate, Vmax is the maximal reaction rate underenzyme saturation, [S] is the substrate concentration,K is substrate concentration required for half maximalrate, and n is the Hill coefficient.

Kinetic modeling

Modeling the kinetics of Pup transfer from onetarget (A) to a second target (B) was based on theassumption that the change in concentration ofpupylated A (PA) over time equals the rate of itspupylation (vA pupyl.) minus the rate of PA depupyla-

tion (vPA depup.). Based on the transferase modeldescribed in this study (summarized in the diagrampresented Fig. 5c), the following equation was usedto describe the change of PA concentration overtime:

d PA½ �d t

¼ vA pupyl:−vPA depup:

¼ E½ �k2P0½ �

P0½ � þ K P 1þ PA½ � þ PB½ �K P

� �

� A½ �A½ � þ K A 1þ B½ �

K Bþ PA½ � þ PB½ �

K P

� �

− E½ �k1PA½ �

PA½ � þ K P 1þ P0½ � þ PB½ �K P

� �

ð1Þ

In this equation, E is PafA, P′ is the acyl phosphatePup intermediate and PB is pupylated B. The rateand binding constants are described in Table 1 andin the text. The pupylation rate is represented byMichelis-Menten expressions that include competi-tion of P′with PA and PB, and of A with B, PA and PBfor PafA binding. The depupylation rate is represent-ed by a Michelis-Menten expression that includescompetition of PA with P′ and PB for PafA binding.Similarly, the change in PB concentration over timeis described by:

d PB½ �d t

¼ vB pupyl:−vPB depup:

¼ E½ �k2P0½ �

P0½ � þ K P 1þ PA½ � þ PB½ �K P

� �

� B½ �A½ � þ K B 1þ A½ �

K Aþ PA½ � þ PB½ �

K P

� �

− E½ �k1PB½ �

PB½ � þ K P 1þ P0½ � þ PA½ �K P

� �

ð2Þ

Based on a mass conservation assumption, it wasdetermined that:

A½ � ¼ A0½ � þ PA0½ �− PA½ � ð3Þ

B½ � ¼ B0½ � þ PB0½ �− PB½ � ð4Þ

P0½ � ¼ PA0½ �− PA½ � þ PB0½ �− PB½ � ð5Þ

680 PafA transferase activity

Using Eqs. (1)-(5), the concentrations of [PA], [PB],[A], [B] and [P′] over time were simulated usingMicrosoft Excel (Sup File 1). Initial parameters forPA, PB, A, B and P′ were set in lines 1 and 2 ofthe Excel spreadsheet (Sup. File 1). Next, a “Time”column was assigned from 0 to 1176 with increments(dt) of 4. Although these time increments can besmaller or slightly bigger and still yield similarmodelingaccuracy, setting the time increments to values thatare too big would have resulted in critical modelingerrors. At each time point, PA and PB concentrationswere calculated by multiplying Eqs. (1) and (2),respectively, with dt, and summing the product withthe value of the previous time point.

Structural alignment

Molecular graphics and analyses were performedwith the UCSF Chimera package [35].

Acknowledgements

This work was supported by Israel Science Foun-dation grant 588/14. We thank Maayan Korman, ShaiSchlussel and Nimrod Zilberberg for critical review ofthe manuscript.

Appendix A. Supplementary data

Supplementary data to this article can be foundonline at https://doi.org/10.1016/j.jmb.2018.01.009.

Received 16 August 2017;Received in revised form 16 January 2018;

Accepted 16 January 2018Available online 2 February 2018

Keywords:Glutamine synthetase;

PafA;Pupylation;

Transferase;Transpupylation

References

[1] E. Gur, D. Biran, E.Z. Ron, Regulated proteolysis in gram-negative bacteria-how and when? Nat. Rev. Microbiol. 9(2011) 839–848.

[2] M.P. Hinault, A. Ben-Zvi, P. Goloubinoff, Chaperones andproteases: cellular fold-controlling factors of proteins inneurodegenerative diseases and aging, J. Mol. Neurosci.30 (2011) 249–265.

[3] U. Jenal, R. Hengge-Aronis, Regulation by proteolysis inbacterial cells, Curr. Opin. Microbiol. 6 (2003) 163–172.

[4] T. Ravid, M. Hochstrasser, Diversity of degradation signals inthe ubiquitin-proteasome system, Nat. Rev. Mol. Cell Biol. 9(2008) 679–690.

[5] R. Yau, M. Rape, The increasing complexity of the ubiquitincode, Nat. Cell Biol. 18 (2016) 579–586.

[6] G.A. Collins, A.L. Goldberg, The logic of the 26S proteasome,Cell 169 (2017) 792–806.

[7] I. Dikic, Proteasomal and Autophagic degradation systems,Annu. Rev. Biochem. 86 (2017) 193–224.

[8] F. Striebel, F. Imkamp, D. Özcelik, E. Weber-Ban, Pupylationas a signal for proteasomal degradation in bacteria, Biochim.Biophys. Acta 1843 (2014) 103–113.

[9] M.J. Pearce, J. Mintseris, J. Ferreyra, S.P. Gygi, K.H. Darwin,Ubiquitin-like protein involved in the proteasome pathwayof Mycobacterium tuberculosis, Science 322 (2008)1104–1107.

[10] X. Chen, W.C. Solomon, Y. Kang, F. Cerda-Maira, K.H.Darwin, K.J. Walters, Prokaryotic ubiquitin-like protein pup isintrinsically disordered, J. Mol. Biol. 392 (2009) 208–217.

[11] S. Liao, Q. Shang, X. Zhang, J. Zhang, C. Xu, X. Tu, Pup, aprokaryotic ubiquitin-like protein, is an intrinsically disorderedprotein, Biochem. J. 422 (2009) 207–215.

[12] O. Regev, Z. Roth, M. Korman, I. Khalaila, E. Gur, A kineticmodel for the prevalence of mono- over poly-pupylation,FEBS J. 282 (2015) 4176–4186.

[13] F. Striebel, F. Imkamp, M. Sutter, M. Steiner, M. Mamedov, E.Weber-Ban, Bacterial ubiquitin-like modifier pup is deamidatedand conjugated to substrates by distinct but homologousenzymes, Nat. Struct. Mol. Biol. 16 (2009) 647–651.

[14] E. Guth, M. Thommen, E. Weber-Ban, Mycobacterialubiquitin-like protein ligase PafA follows a two-step reactionpathway with a phosphorylated pup intermediate, J. Biol.Chem. 286 (2011) 4412–4419.

[15] F.A. Cerda-Maira, M.J. Pearce, M. Fuortes, W.R. Bishai, S.R.Hubbard, K.H. Darwin, Molecular analysis of the prokaryoticubiquitin-like protein (pup) conjugation pathway in Mycobac-terium tuberculosis, Mol. Microbiol. 77 (2010) 1123–1135.

[16] Y. Elharar, Z. Roth, I. Hermelin, A. Moon, G. Peretz, Y.Shenkerman, M. Vishkautzan, I. Khalaila, E. Gur, Survival ofmycobacteria depends on proteasome-mediated aminoacid recycling under nutrient limitation, EMBO J. 33 (2014)1802–1814.

[17] K.E. Burns, F.A. Cerda-Maira, T. Wang, H. Li, W.R. Bishai, K.H. Darwin, "Depupylation" of prokaryotic ubiquitin-like proteinfrom mycobacterial proteasome substrates, Mol. Cell 39(2010) 821–827.

[18] F. Imkamp, F. Striebel, M. Sutter, D. Özcelik, N. Zimmermann,P. Sander, E. Weber-Ban, Dop functions as a depupylase inthe prokaryotic ubiquitin-like modification pathway, EMBORep. 11 (2010) 791–797.

[19] L.M. Iyer, A.M. Burroughs, L. Aravind, Unraveling thebiochemistry and provenance of pupylation: a prokaryoticanalog of ubiquitination, Biol. Direct 3 (2008) 45.

[20] D. Özcelik, J. Barandun, N. Schmitz, M. Sutter, E. Guth, F.F.Damberger, F.H. Allain, N. Ban, E. Weber-Ban, Structures ofpup ligase PafA and depupylase Dop from the prokaryoticubiquitin-like modification pathway, Nat. Commun. 3 (2012)1014.

[21] M. Bolten, C. Vahlensieck, C. Lipp, M. Leibundgut, N. Ban, E.Weber-Ban, Depupylase Dop requires inorganic phosphatein the active site for catalysis, J. Biol. Chem. 292 (2017)4044–4053.

681PafA transferase activity

[22] D. Eisenberg, H.S. Gill, G.M. Pfluegl, S.H. Rotstein,Structure-function relationships of glutamine synthetases,Biochim. Biophys. Acta 1477 (2000) 122–145.

[23] L. Levintow, A. Meister, Reversibility of the enzymaticsynthesis of glutamine, J. Biol. Chem. 209 (1954) 265–280.

[24] S. Zhang, K.E. Burns-Huang, G.V. Janssen, H. Li, H. Ovaa, L.Hedstrom, K.H. Darwin, Mycobacterium tuberculosis protea-some accessory factor a (PafA) can transfer prokaryoticubiquitin-like protein (pup) between substrates, MBio 8 (2017)e00122–17.

[25] M.J. Pearce, P. Arora, R.A. Festa, S.M. Butler-Wu, R.S.Gokhale, K.H. Darwin, Identification of substrates of theMycobacterium tuberculosis proteasome, EMBO J. 25 (2006)5423–5432.

[26] N. Hecht, E. Gur, Development of a fluorescence anisotropy-based assay for Dop, the first enzyme in the pupylationpathway, Anal. Biochem. 485 (2015) 97–101.

[27] Y. Elharar, Z. Roth, N. Hecht, R. Rotkopf, I. Khalaila, E. Gur,Posttranslational regulation of coordinated enzyme activitiesin the pup-proteasome system, Proc. Natl. Acad. Sci. U. S. A.113 (2016) E1605–1614.

[28] O. Regev, M. Korman, N. Hecht, Z. Roth, N. Forer, R.Zarivach, E. Gur, An extended loop of the pup ligase, PafA,Mediates Interaction with Protein Targets, J. Mol. Biol. 428(2016) 4143–4153.

[29] N. Ofer, N. Forer, M. Korman, M. Vishkautzan, I. Khalaila, E.Gur, Allosteric transitions direct protein tagging by PafA, theprokaryotic ubiquitin-like protein (pup) ligase, J. Biol. Chem.288 (2013) 11287–11293.

[30] J. Barandun, F.F. Damberger, C.L. Delley, J. Laederach, F.H.Allain, E. Weber-Ban, Prokaryotic ubiquitin-like proteinremains intrinsically disordered when covalently attached toproteasomal target proteins, BMC Struct. Biol. 17 (2017) 1.

[31] X. Chen, C. Li, L. Wang, Y. Liu, C. Li, J. Zhang, Themechanism of Mycobacterium smegmatis PafA self-Pupyla-tion, PLoS One 11 (2016), e0151021.

[32] B.D. Bennett, E.H. Kimball, M. Gao, R. Osterhout, S.J. VanDien, J.D. Rabinowitz, Absolute metabolite concentrationsand implied enzyme active site occupancy in Escherichiacoli, Nat. Chem. Biol. 5 (2009) 593–599.

[33] Q.H. Tran, G. Unden, Changes in the proton potential and thecellular energetics of Escherichia coli during growth byaerobic and anaerobic respiration or by fermentation, Eur. J.Biochem. 251 (1998) 538–543.

[34] J.M. Berg, J.L. Tymoczko, L. Stryer, Biochemistry, Section16.1, Glycolysis Is an Energy-Conversion Pathway in ManyOrganisms, 5th editionW H Freeman, New York, 2002.

[35] E.F. Pettersen, T.D. Goddard, C.C. Huang, G.S. Couch, D.M.Greenblatt, E.C. Meng, T.E. Ferrin, UCSF chimera–avisualization system for exploratory research and analysis,J. Comput. Chem. 25 (2004) 1605–1612.

[36] R.L. Dunbrack Jr., Rotamer libraries in the 21st century, Curr.Opin. Struct. Biol. 12 (2002) 431–440.

[37] R.B. Martin, Free energies and equilibria of peptide bondhydrolysis and formation, Biopolymers 45 (1997) 351–353.