Mechanism of Lys6 poly-ubiquitin specificity by the L. pneumophila deubiquitinase LotA

SUMMARY

The versatility of ubiquitination to control vast domains of eukaryotic biology is due, in part, to diversification through differently-linked poly-ubiquitin chains. Deciphering signaling roles for some chain types, including those linked via K6, has been stymied by a lack of specificity among the implicated regulatory proteins. Forged through strong evolutionary pressures, pathogenic bacteria have evolved intricate mechanisms to regulate host ubiquitin during infection. Herein, we identify and characterize a deubiquitinase domain of the secreted effector LotA from Legionella pneumophila that specifically regulates K6-linked poly-ubiquitin. We demonstrate the utility of LotA for studying K6 poly-ubiquitin signals. We identify the structural basis of LotA activation and poly-ubiquitin specificity, and describe an essential “adaptive” ubiquitin-binding domain. Without LotA activity during infection, the Legionella-containing vacuole becomes decorated with K6 poly-ubiquitin as well as the AAA ATPase VCP/p97/Cdc48. We propose that LotA’s deubiquitinase activity guards Legionella-containing vacuole components from ubiquitin-dependent extraction.

Graphical Abstract

eTOC:

Warren et al. identify a K6-specific deubiquitinase within the Legionella pneumophila effector LotA and demonstrate its utility as a research tool. Structural and biochemical work explain the mechanism of LotA activation and specificity. During infection, LotA’s deubiquitinase activity guards the Legionella-containing vacuole from K6 poly-ubiquitin and VCP recruitment.

INTRODUCTION

Post-translational regulation with the small protein modifier ubiquitin (Ub) is essential to all eukaryotic life¹. In humans, hundreds of enzymes tightly regulate protein ubiquitination by defining both the target substrate and the type of modification². Unlike binary “on-off” modifications such as phosphorylation or acetylation, the Ub signal can be further diversified through the formation of polymeric Ub (polyUb) chains in which any of the seven Lys residues (e.g., K48- or K63-linked chains) or the amino-terminus (linear/M1-linked chains) are ubiquitinated. Differently-linked polyUb chains code for distinct signaling outcomes, including proteasomal degradation (K48-linked chains), the DNA damage response (K63-linked chains), or innate immune signaling (M1-linked chains)³. Although the existence and importance of all eight possible chain types has been established^4–7, signaling outcomes and regulators for half (K6, K27, K29, and K33) remain ill-defined and have thus been coined “atypical” chains⁸.

Among atypical chain types, K6-linked polyUb has remained particularly enigmatic. Whether K6 polyUb serves as a proteasomal degradation signal remains unclear, as levels of K6 chains show little or no increase upon proteasomal inhibition^5,9. K6 polyUb accumulates following mitochondrial depolarization, UV-induced DNA damage, or inhibition of the AAA ATPase VCP/p97/Cdc48^10–13. Defining the importance of K6 polyUb in these responses, however, has been clouded by the imprecise specificities of the implicated E3 ligases and deubiquitinases (DUBs). Mitophagy, for example, can be regulated by the E3 ligase Parkin and DUBs USP30 and USP8, and while these enzymes can control levels of K6 polyUb, they also target other chain types including K11, K48, and K63^10,14–17. The same is true for the ligase activity of BRCA1 in the DNA damage response^18–20, as well as the ligase activity of HUWE1 in targeting substrates for VCP^12,13,21. Confidently ascribing a signaling function for K6 polyUb, therefore, awaits discovery of new approaches or enzymes that can more specifically measure and/or perturb it. Such was the case with M1-linked polyUb, prior to the discovery of M1-specific regulators²².

PolyUb chain specificity is typically achieved through the concerted action of multiple Ub-binding sites. E2 Ub-conjugating enzymes and E3 ligases achieve linkage specificity through specific coordination of two Ub molecules²³. Within the resulting polyUb chain, Ub molecules are classified as a distal Ub (Ub^dist) that is linked via its C-terminus onto an amino group of the proximal Ub (Ub^prox). For DUBs, specificity is often achieved via Ub-binding sites that oppose the catalytic center, including an S1 site that binds Ub^dist and an S1’ site that binds Ub^prox in a linkage-specific orientation²⁴. Linkage-specific regulators have been identified for M1, K11, K48, and K63 polyUb³, but despite considerable efforts, human enzymes with stringent specificity toward K6 polyUb have thus far evaded detection.

Pathogenic bacteria represent an alternative source of polyUb linkage-specific enzymes. Bacterial pathogens have evolved Ub regulatory enzymes that interfere with human signaling processes during infection, including a number that target specific polyUb chain types²⁵. In fact, NleL from enterohemorrhagic Escherichia coli is the most K6-specific ligase known to-date, assembling a 50:50 mixture of K6 and K48 linkages²⁶. The causative agent of Legionnaires’ disease, Legionella pneumophila, has evolved numerous mechanisms of disrupting or hijacking host Ub signaling. Of the >300 effector proteins secreted into infected cells via the L. pneumophila Dot/Icm type IV secretion system, several dozen regulate host Ub signals²⁷, including some that target specific polyUb chain types²⁸. Many Ub-targeted Legionella effectors localize to the Legionella-containing vacuole (LCV), where they edit the complex “coat” of Ub modifications in order to establish a replicative niche and evade host immune responses²⁹. One such effector, termed Legionella OTU-like protein A (LotA), was reported to contain two catalytic DUB domains with similarity to the human ovarian tumor (OTU) family, as well as DUB activity toward K6, K48, and K63 polyUb³⁰. Whether and how one of LotA’s catalytic OTU domains might be specific toward K6 polyUb, however, remained unknown.

Here we demonstrate that the two catalytic DUB modules of LotA are separable and have distinct activities and specificities. Remarkably, the first OTU domain of LotA encodes perfect specificity toward K6 polyUb and represents a valuable tool for studying this atypical linkage. To understand the molecular basis for K6 specificity, we determined crystal structures for LotA alone and bound to K6 diUb. Detailed biochemical analyses revealed a multi-step mechanism of LotA activation that endows K6 specificity, which incorporates themes of conformational rearrangement and substrate-assisted catalysis observed in human linkage-specific DUBs. In the process of studying LotA, we discovered a class of Ub-binding domain that is adapted across the Legionella Lot class as well as an otherwise unrelated DUB from Orientia tsutsugamushi. During L. pneumophila infection, we find that LotA plays a role in clearing K6 polyUb from the LCV, which prevents downstream recruitment of VCP. We propose that the dual DUB activities of LotA act to guard select components of the LCV from Ub- and VCP-dependent extraction. This work provides a powerful tool for the study of Ub biology and firmly establishes an interesting link for K6 polyUb at the host-pathogen interface.

RESULTS

LotA encodes two catalytic DUB domains with distinct specificities

The Lot class of DUBs encoded by L. pneumophila is currently composed of LotA (lpg2248), LotB (lpg1621), LotC (lpg2529), and LotD (lpg0227) and defined by a ~100 amino acid insertion within the OTU domain. The Lot-class insertion domains are more diverse in primary sequence than the core OTU domains (Fig. S1A–B), suggesting that they may play a role in determining the various polyUb specificities that have been reported^30–35. LotA encodes two Lot-class OTU domains, denoted LotA_N and LotA_M for the N-terminal and middle domains, respectively, followed by the C-terminal LotA_C domain that binds PI(3)P, localizing LotA to the LCV during infection³⁰ (Fig. 1A). As reported previously for shorter polyUb chains³⁰, we observe that full length LotA is capable of hydrolyzing long polyUb chains composed of a range of different linkage types, including notable activity toward K6-, K48-, and K63-linked substrates (Fig. 1B). The sequence relationship between LotA_N and LotA_M is the most distinct observed among the Lot class, with sequence identities of only 19% and 23% within the insertion and OTU domains, respectively (Fig. S1A–B). To determine if LotA_N and LotA_M contribute distinct activities to that observed for full length LotA (Fig. 1B)³⁰, we cloned and expressed the isolated domains for biochemical analyses.

Figure 1: Separation of LotA deubiquitinase activities.

A. Domain architecture of L. pneuomophila LotA, with catalytic triad residues annotated for each OTU domain.

B. Homogeneous assemblies of seven polyUb linkage types were treated with a high concentration (5 μM) of full length LotA for 2 h before the reactions were quenched and visualized by anti-Ub Western blot.

C. Coomassie-stained SDS-PAGE gel showing purified constructs of LotA_N (1–300) and LotA_M (287–543), alongside their catalytically-inactive variants. Reactivity with the Ub-PA activity-based probe is assessed by a shift in mobility.

D. Ub-KG(TAMRA) cleavage assay monitored by fluorescence polarization. Activity was measured using 1 μM LotA_N, 1 μM LotA_M, or 100 nM USP21 as a control DUB.

E-F. Gel-based specificity analysis against all eight canonical diUb linkages. Reactions containing the indicated concentrations of LotA_N (D) and LotA_M (E) were sampled at the indicated timepoints, quenched, and resolved by SDS-PAGE with Coomassie staining.

G-H. Gel-based specificity analysis against seven tetraUb linkages. Reactions containing the indicated concentrations of LotA_N (F) and LotA_M (G) were sampled at the indicated timepoints, quenched, and resolved by SDS-PAGE with Coomassie staining. Note that some tetraUb chains migrate differently based on the linkage site, as observed previously^38,45,46.

I. Homogeneous assemblies of seven polyUb linkage types were treated with a high concentration (5 μM) of full length LotA C303A for 2 h before the reactions were quenched and visualized by anti-Ub Western blot.

J. A homogeneous assembly of K6-linked polyUb was treated with either LotA_N or the full length LotA C303A variant at 25 nM. Reaction samples were collected and visualized by anti-Ub Western blot.

See also Figure S1.

Both the LotA_N and LotA_M isolated constructs reacted with the Ub-Propargylamide (Ub-PA) activity-based probe, indicating that each are competent for DUB activity (Fig. 1C)³⁶. Reactivities of LotA_N and LotA_M with Ub-PA were ablated by mutation of their annotated catalytic cysteine residues, C13 and C303, to alanine (Fig. 1A,C)³⁰. Surprisingly, although both constructs reacted with the monomeric Ub-PA substrate, neither could cleave the monomeric Ub-KG(TAMRA) fluorescent substrate, or the analogous NEDD8, SUMO1, or ISG15 substrates (a discrepancy that will be discussed further below) (Fig. 1D, S1C–E)³⁷. When incubated with each of the eight canonical (iso)peptide-linked diUb substrates, LotA_N demonstrated remarkable specificity for the K6 linkage type (Fig. 1E). In contrast, LotA_M was very inefficient and only showed marginal cleavage of the K6-linked substrate at high concentration after a long incubation period (Fig. 1F). As the polyUb chain length has been reported to affect the activities and/or specificities of some DUBs^38–40, we assembled a panel of seven differently-linked tetraUb substrates (lacking only K27). LotA_N again showed exquisite specificity against the K6 linkage, while LotA_M now demonstrated more robust DUB activity, with a distinct preference toward the K48 linkage and lesser activities toward K6 and K63 tetraUb (Fig. 1G–H). Consistent with observations made with the diUb panel (Fig. 1F), LotA_M activity appeared to stall after the tetraUb and triUb were consumed, leaving a diUb product that was stable over time (Fig. 1H). These results indicate that LotA_N and LotA_M indeed demonstrate distinct DUB properties, with LotA_N showing exquisite K6 specificity while LotA_M shows a preference toward longer K48-linked polyUb that is likely directed by an additional (S2 or S2’) Ub-binding site. Within the context of full length LotA, the specificity and activity of the LotA_N OTU domain are independent of the LotA_M OTU domain, as a full length LotA C303A variant demonstrates the same K6 specificity and rate of cleavage as the isolated LotA_N construct (Fig. 1I–J).

Benchmarking LotA_N as a tool for the study of K6 polyUb

To further characterize the activity of LotA_N toward K6 polyUb, we constructed a fluorescent K6 diUb substrate (see STAR Methods) and monitored cleavage by fluorescence polarization (FP), as reported previously^41–43. Through measuring initial rates of catalysis over a range of substrate concentrations, Michaelis-Menten parameters could be extracted for LotA_N against K6 diUb (Fig. 2A, S2A). Despite a relatively high K_M, LotA_N demonstrated an overall catalytic efficiency that is similar to other linkage-specific OTU DUBs, including the K11-specific Cezanne and the M1-specific OTULIN^42,44.

Figure 2: Application of LotA_N K6 specificity for UbiCRest analysis.

A. Kinetic parameters of LotA_N (1–300) measured by changes in fluorescence polarization of labeled K6 diUb. Initial rates of diUb cleavage were measured over a range of substrate concentrations and fit to a Michaelis-Menten model. Error bars represent standard deviation over three measurements, each made in triplicate.

B. Gel-based specificity analysis against seven tetraUb linkages. Reactions containing a high concentration (5 μM) of LotA_N were sampled at the indicated timepoints, quenched, and resolved by SDS-PAGE with Coomassie staining.

C. Homogeneous assemblies of seven polyUb linkage types were treated with a high concentration (5 μM) of LotA_N for 2 h before the reactions were quenched and visualized by anti-Ub Western blot.

D. UbiCRest analysis of an NleL ligase assembly with 1 μM K6-specific LotA_N, K11-specific Cezanne, K48-specific OTUB1*, K63-specific AMSH*, nonspecific vOTU, or the indicated combinations. Reactions were visualized by anti-Ub Western blot. Cleavage of NleL-assembled polyUb can be observed by a decrease in the “smear” or by a reappearance of monoUb.

E. UbiCRest analysis of a HUWE1 ligase assembly with 1 μM K6-specific LotA_N, K11-specific Cezanne, K48-specific OTUB1*, K63-specific AMSH*, nonspecific vOTU, or the indicated combinations. Reactions were visualized by anti-Ub Western blot. Cleavage of HUWE1-assembled polyUb can be observed by a decrease in the “smear” or by a reappearance of monoUb.

See also Figure S2.

To further benchmark the utility of LotA_N as a tool, we tested if specificity is retained even at high enzyme concentrations. Remarkably, even at very high (5 μM) enzyme concentration against both tetra- and longer polyUb samples, LotA_N exclusively cleaved only K6 linkages (Fig. 2B–C, S2B). To test LotA_N’s utility in releasing K6 linkages from complex, heterogeneous mixtures, we tested it alongside other linkage-specific DUBs in several Ubiquitin Chain Restriction (UbiCRest) analyses^38,45,46.

The HECT-like E3 ligase NleL from enterohemorrhagic Escherichia coli has been reported to assemble mixed K6 and K48 polyUb chains^26,45. Alongside the constitutively-activated K48-specific DUB OTUB1*⁴⁷, LotA_N could indeed confirm this chain architecture in a UbiCRest assay. Treatment of NleL-assembled polyUb with either LotA_N or OTUB1* resulted in a partial collapse of the polyUb smear down to shorter chains such as diUb, as well as a partial return of the monoUb band, while treatment with either the K11-specific Cezanne or the constitutively-activated K63-specific AMSH* had no effect (Fig. 2D)⁴⁷. Co-incubation of the linkage-specific DUBs and comparison to the nonspecific vOTU revealed the abundance of K6 linkages present in the NleL assembly (Fig. 2D)^38,48.

As a further test, we generated polyUb chains with HUWE1, a human HECT E3 ligase that is reported to assemble a mixture of K6-, K11-, and K48-linked chains^13,21,49. Treatment with Cezanne released the largest amount of monoUb, followed by OTUB1*, then LotA_N, and finally AMSH*, which had only a marginal amount of activity (Fig. 2E). DUB co-incubation with or without LotA_N showed a slight difference attributable to K6 linkages, especially at the level of released short polyUb products (Fig. 2E). Combined treatment of LotA_N, Cezanne, OTUB1*, and AMSH* hydrolyzed the assembled polyUb to a similar extent as vOTU, indicating that K11, K48, and K6 linkages form the bulk of HUWE1 products (Fig. 2E). Thus, the high level of activity and specificity exhibited by LotA_N makes it a powerful tool that can be used alongside other linkage-specific enzymes to study K6 polyUb.

Structural analysis of LotA_N

The distinctive sequence and catalytic properties of LotA_N warranted structural analysis for comparison to the other Lot DUBs. We determined a 1.5Å crystal structure of LotA_N residues 1–294 using SAD phasing on bound iodide ions (Fig. 3A, S3A, Table 1). Overall, the LotA_N structure resembled other Lot DUBs, with a papain-like OTU domain and a characteristic α-helical insertion domain at Variable Region 1 (VR-1), but additionally revealed several notable features (Fig. 3A). Firstly, the catalytic triad, comprised of C13, H237, and N239, was misaligned and beyond hydrogen bonding distance in the structure. In fact, two conformations of the C13 side chain were visible in the electron density, whereas no electron density was observed for the H237 imidazole ring (Fig. 3B). Despite this, mutation of C13, H237, and to a lesser extent N239 all severely decreased the ability of LotA_N to cleave the fluorescent K6 diUb substrate (Fig. 3C).

Figure 3: Crystal structure of the LotA_N OTU domain.

A. 1.5Å crystal structure of LotA (1–294) labeled with visible termini, domain architecture, active site, and OTU variable regions (VR1–3) described previously³².

B. Close-up view of the LotA_N active site with 2|Fo|-|Fc| electron density overlaid at 1σ for catalytic triad residues. Alternate conformations of the C13 side chain (a and b) are observed, whereas density for the H237 imidazole ring is absent.

C. Cleavage of fluorescent K6 diUb by the indicated LotA_N active site variants at 10 nM concentration monitored by fluorescence polarization.

D. Structural overlay of LotA_N (blue), LotA_M (PDB 7F9X, orange), LotB (PDB 6KS5, yellow), and LotC (PDB 7BU0, green). Structures were aligned by their core OTU domains to demonstrate variability in the insertion domains.

E. Close-up views of the LotA_M, LotB, LotC (PDB 6YK8), and Ub-bound LotC (PDB 7BU0) active sites. Catalytic triad residues are shown in ball-and-stick representation, with hydrogen bonds indicated by dashed lines.

See also Figure S3.

Table 1:

Data collection and refinement statistics

	LotAN (1–294)	LotAN (1–276) : K6 diUb
Data collection
Wavelength	0.979460	0.979460
Space group	P 41 21 2	P 4 2 2
Cell dimensions
a, b, c (Å)	86.25, 86.25, 92.30	127.28, 127.28, 77.99
α, β, γ (°)	90, 90, 90	90, 90, 90
Resolution (Å)	39.07–1.50 (1.50–1.53)	38.99–2.80 (2.80–2.95)
R merge	0.035 (0.819)	0.257 (1.15)
I/σI	26.6 (2.2)	6.4 (2.0)
Completeness (%)	99.9 (99.7)	99.9 (100)
Redundancy	6.5 (6.5)	6.5 (6.7)
Phasing
Method	SAD	MR
SAD Resolution	1.50	-
SAD Anom completeness	99.6 (99.5)	-
SAD Anom multiplicity	3.4 (3.3)	-
SAD FOM	0.733	-
MR Search Models	-	LotAN, Ub (1UBQ)
Refinement
Resolution (Å)	35.59–1.50	38.99–2.80
No. unique reflections / test set	56168 / 2895	16316 / 822
R work /R free	0.195/0.221	0.207/0.259
No. atoms
Protein	2174	3399
Ligand/ion	19	0
Water	286	59
B-factors
Protein	32.81	47.04
Ligand/ion	54.93	-
Water	39.12	33.63
R.m.s. deviations
Bond lengths (Å)	0.006	0.006
Bond angles (°)	0.77	1.04
Ramachandran statistics
Favored / allowed / outliers	96.7 / 3.3 / 0	97.2 / 2.8 / 0

Values in parentheses are for highest resolution shell.

The core OTU fold of LotA_N closely resembles those of the other Lot-class DUBs (Fig. 3D), as well as the most closely related OTU DUBs from humans, viruses, and bacteria outside of Legionella^{31–34,48,50–52} (Fig. S3B–D). At the putative S1 Ub-binding site, LotA_N more closely follows the other Lot-class DUBs with a large helical domain inserted at VR-1, and minimal structural features at VR-2 and VR-3 compared to other OTU domains³² (Fig. 3D, S3B–D). All of the Lot-class insertion domains structurally resolved thus far exhibit distinguishing adaptions of an underlying helical sub-structure and distinct orientations with respect to the OTU domains (Fig. 3D), a feature that has been proposed to contribute to the various polyUb specificities observed within this class of DUBs. A closer look at the catalytic triad across Lot-class DUBs reveals that, like LotA_N, the active site of LotA_M is also misaligned in the observed crystal structure, whereas the active sites of LotB, LotC, and Ub-bound LotC are all in proper alignment for catalysis (Fig. 3E). This discrepancy in alignment is consistent with our observation that neither LotA_N nor LotA_M could cleave the monomeric Ub-KG(TAMRA) substrate (Fig. 1D), whereas LotB and the related LotD were previously shown to be active in the same assay³². This suggests that LotA_N and LotA_M are distinct from other Lot-class DUBs in their requirement for some form of substrate-assisted catalysis akin to that observed previously for the M1-specific DUB OTULIN⁴².

Mechanism of K6 diUb recognition by LotA_N

To understand the molecular basis of LotA_N activation and K6 polyUb specificity, we determined a 2.8Å crystal structure of LotA_N in complex with K6 diUb by molecular replacement (Fig. 4A, S4A–B, Table 1). In the structure, K6 diUb is bound across the LotA_N active site, sandwiched between the OTU core and the helical insertion domain (Fig. 4A). In order to preserve the K6 diUb while retaining binding to LotA_N, the active site C13A mutation was used. Despite this, the electron density between the LotA_N-bound Ub molecules more closely supports a structural intermediate just after isopeptide cleavage, with no density connecting the ε-amino group of K6 from Ub^prox with the carbonyl carbon of G76 in Ub^dist (Fig. 4B). Rather than cleavage of the K6 linkage, we believe this lack of clear electron density is either due to a shift in register (i.e., Ub^dist and Ub^prox are bound into the S1’ and S1 sites, respectively, of symmetry-related LotA_N molecules) or symmetry averaging across related asymmetric units, as the C-terminus of Ub^prox is in close proximity to the K6 of a symmetry-related Ub^dist (Fig. S4C). The symmetry-related expansion of polyUb chains has been observed in other crystal structures⁴⁷, and in this case suggests that LotA_N exhibits endo-DUB activity that can cleave anywhere within a polyUb chain. Although the catalytic C13 is mutated to alanine, the electron density clearly supports proper alignment and hydrogen bond distance of the remaining H237 and N239 triad members (Fig. 4B), indicating a structural rearrangement upon K6 diUb binding that will be discussed in more detail below.

Figure 4: Crystal structure of the LotA_N OTU domain bound to K6 diUb.

A. 2.8Å crystal structure of LotA_N (1–276, blue) bound to K6 diUb (shades of red). The structure is labeled with visible termini, domain architecture, active site, and OTU variable regions (VR1–3).

B. Close-up view of the K6 diUb-bound LotA_N active site with 2|Fo|-|Fc| electron density overlaid at 1σ for catalytic triad residues. The K6 side chain of Ub^prox and C-terminus of Ub^dist are also shown with overlaid electron density. The LotA_N catalytically inactive C13A variant was used to preserve the diUb linkage.

C. Crystal structures of K6 diUb in isolation (PDB 2XK5) and bound to LotA_N, aligned by their Ub^dist moieties (top). Availability and orientation of common interaction surfaces are shown, with dashed circles indicating the surfaces utilized by either Ub:Ub or LotA_N:Ub interaction.

D. Structural overlay of LotA_N (blue) and LotC (green) bound to their respective distal Ub moieties (red and pink, respectively). The structures are aligned on their core OTU domains, and highlight large differences in Ub orientation, insertion domains, and use of variable regions.

E. Close-up view of the LotA_N S1 site (blue) bound to Ub^dist (red). Interacting residues are shown in ball-and-stick representation, with hydrogen bonds indicated by dashed lines.

F. Cleavage of fluorescent K6 diUb by the indicated LotA_N S1 site variants at 10 nM concentration monitored by fluorescence polarization.

G. Structural overlay of LotA_N (blue) and human USP30 (tan) bound to their respective distal Ub moieties (salmon and purple, respectively). The structures are aligned on their core catalytic domains, and highlight large differences in Ub orientation and features of their S1’ sites.

H. Close-up view of the LotA_N S1’ site (blue) bound to Ub^prox (salmon). Interacting residues are shown in ball-and-stick representation, with hydrogen bonds indicated by dashed lines.

I. Cleavage of fluorescent K6 diUb by the indicated LotA_N S1’ site variants at 10 nM concentration monitored by fluorescence polarization. These data were collected in parallel with those presented in (F), and the WT dataset is shown again for reference.

J. Gel-based LotA_N cleavage assay of K6 diUb variants with indicated Ub^prox mutations. Reactions were quenched at the indicated times and visualized by SDS PAGE with Coomassie staining.

K. Underlying helical domain architecture of the LotA_N (left) and stereotypical (right) adaptive Ub-binding domain (A-UBD), with helices labeled and the Ub-binding α1–2 regions shown in color.

See also Figure S4.

In isolation of binding partners, crystallography and solution NMR data support a predominately closed conformation of K6 diUb, in which the Ub^dist I36 hydrophobic patch is bound to the Ub^prox I44 hydrophobic patch (Fig. 4C)^41,45. In contrast, LotA_N appears to select an open conformation of K6 diUb such that it can interact with the Ub^dist I36 patch as well as the I36 and F4 hydrophobic patches of Ub^prox (Fig. 4C). Selection among heterogeneous K6 diUb conformations appears to be an emerging trend of K6-interacting proteins, as distinct K6 diUb structures have been resolved in complexes with an engineered affimer protein, the DUB USP30, the NZF Ub-binding domain of TAB2, and now also LotA_N (Fig. S4D)^13,16,53.

The interaction with Ub^dist at the LotA_N S1 site appears to be mainly driven through an interface between the inserted helical domain at VR-1 and the I36 patch of Ub^dist (Fig. 4A, C). In contrast, the structure of LotC bound to Ub shows a stark ~180-degree rotation and a primary interface at a nonconserved VR-2 region (Fig. 4D). At the core of the interface, an extended loop region of the LotA_N insertion domain slots L128 into a shallow pocket of Ub^dist formed between I36, L71, and T7 (Fig. 4E). Additional ionic interactions are observed around the periphery, including hydrogen bonds to both R72 and R74 of Ub^dist as the C-terminus enters the LotA_N active site (Fig. 4E). This interface appears to play a minor role in the binding of K6 diUb, as an L128A mutation had minimal effect, and only a severe L128R mutation that introduces a steric clash with the Ub^dist I36 patch significantly disrupted LotA_N cleavage of the fluorescent K6 diUb substrate (Fig. 4F).

The LotA_N S1’ site orients K6 of Ub^prox into the active site through interactions with the Ub I36 and F4 hydrophobic patches (Fig. 4A, C). The human DUB USP30 also recognizes the F4 patch of a K6-linked Ub^prox, but in a very different manner involving a ~100-degree twist (Fig. 4G, S4D). In the LotA_N structure, Ub^prox is sandwiched between the core OTU and insertion domains and makes a single hydrogen bond to R74 of Ub^dist through the carbonyl oxygen of T9 (Fig. 4H). The F4 side chain of Ub^prox makes Van der Waals interactions to the peptide backbone of D10 within the LotA_N OTU core domain (Fig. 4H). At the Ub^prox I36 patch, F120 and L141 from the LotA_N insertion domain interact with I36, while buried within the interface lies a salt bridge between LotA_N R145 and Ub^prox E34. Mutation of these residues in LotA_N had a much more severe impact on K6 diUb cleavage than mutations in the S1 site (Fig. 4F,I). Reciprocal mutations specifically introduced into Ub^prox of K6 diUb (see STAR Methods) showed the strongest cleavage defect with mutation of F4, a modest effect with I36 mutation, and no effect upon mutation of the unutilized I44 site (Fig. 4J).

The LotA_N insertion domain makes key contributions to both the S1 and S1’ Ub-binding sites required for DUB activity. Despite their differences, the core structures of the Lot-class insertion domains follow a related underlying 4-helix topology (Fig. 4K, S4E). By structural comparison using Dali⁵², the insertion domain of LotA_N is the most similar to LotC over other Lot-class DUBs (Fig. S4F). Based on the Ub- and diUb-bound structures of LotC and LotA_N, respectively, it is the distinguishing features of the α1–2 region that dictate interactions to Ub (Fig. S4F–H)³⁴. Modeling and mutagenesis studies implicate the α1–2 regions of LotB and LotA_M in mediating important Ub interactions as well^33,50. Interestingly, Dali searches also detected similarity to the VR-1 region of OtDUB, an effector DUB from Orientia tsutsugamushi that belongs to the CE clan of cysteine proteases⁵⁴ (Fig. S4I). The OtDUB VR-1 region follows the 4-helix structural topology conserved among Lot-class DUBs, and encodes two Ub-binding sites: one at the edge of Helix 3 that forms part of the S1 site to bind Ub^dist (Fig. S4J), and a second located in the α1–2 region that binds an additional Ub with unusually high affinity (Fig. S4K). Unlike the high affinity site in the α1–2 region (Fig. S4K), residues at the edge of Helix 3 in OtDUB VR-1 are conserved in the LotA_N structure (Fig. S4J). Additional binding of Ub at this site in LotA_N, however, would cause steric clash with the Ub^dist observed in our structure (Fig. S4J). Because of its important roles in Lot-class DUBs as well as OtDUB, we propose to assign the name Adaptive Ubiquitin-Binding Domain, or “A-UBD” for short, to this common 4-helix substructure, owing to adaptations of the α1–2 regions that impart distinctive Ub interactions into a common topological scaffold (Fig. 4K, S4E).

Conformational changes leading to LotA_N activation

Comparison of the apo and K6 diUb-bound LotA_N structures revealed several interesting changes. First, the complex structure exhibited a 45-degree swing of the A-UBD with respect to the core OTU domain, resulting in a more closed conformation (Fig. 5A). This swing moves key residues such as L128, F120, and R145 up, forming the S1 and S1’ Ub-binding sites that sandwich K6 diUb against the active site (Fig. 5B, 4E, 4H). Movement within the hinge region is required for LotA_N activity, as a stabilizing A193P hinge mutation reduces cleavage of the fluorescent K6 diUb substrate, while an A193G mutation retains the mobility required for function (Fig. 5C). Analogous hinge regions can be observed between the core OTU and A-UBDs of other Lot-class DUBs (Fig. S5). Comparison of the apo and Ub-bound LotC structures shows a 6-degree swing, in this case toward a more open conformation in the Ub-bound complex (Fig. S5). Thus, it is possible that other Lot-class DUBs may also require mobility of the A-UBD for DUB activity.

Figure 5: Conformational changes of LotA_N during catalysis.

A. Structural overlay of apo (cyan) and K6 diUb-bound (blue) LotA_N, aligned on their core OTU domains. A site for structure-guided mutation within the hinge region is shown.

B. As in (A), following ~90-degree rotation to visualize formation of the LotA_N S1 and S1’ Ub-binding sites upon rotation of the A-UBD.

C. Cleavage of fluorescent K6 diUb by the indicated LotA_N hinge variants at 10 nM concentration monitored by fluorescence polarization.

D. Structural overlay of the active sites from apo (cyan) and K6 diUb-bound (blue) LotA_N, aligned on their core OTU domains. Conformational changes of the Cys and His loops that occur upon Ub binding (shades of red) are shown by red dashed lines. Distances of the Cys and His loop movements are measured by the Cα positions of LotA_N H237 and G11, respectively. The Cα of LotA_N catalytic triad residues are shown as spheres. Ub^prox residues opposing the rearranged LotA_N loops are shown in ball-and-stick representation, with hydrogen bonds indicated as black dashed lines.

E. Gel-based LotA_N cleavage assay of K6 diUb variants with indicated Ub^prox mutations. Reactions were quenched at the indicated times and visualized by SDS PAGE with Coomassie staining.

F. Ub-PA reactivity of LotA_N wild-type alongside the indicated catalytic, S1 site, S1’ site, and hinge mutations. Reactions were quenched at the indicated times and visualized by SDS PAGE with Coomassie staining.

G. Ub-PA reactivity assay combining the indicated LotA_N, Ub-PA, and Ub variants. Left, the ability of each Ub variant to competently bind the LotA_N S1 or S1’ site is listed.

See also Figure S5.

Another dramatic difference between structures is in the active site. As highlighted above, the LotA_N catalytic triad is misaligned in the apo structure but properly aligned for catalysis in the diUb-bound structure (Fig. 3B and 4B). Coincident with this rearrangement, the so-called “Cys loop” preceding the catalytic cysteine is shifted by >4Å toward the remaining triad members. Opposing the Cys loop is F4 of Ub^prox, which appears to select for this active conformation by creating steric conflict with the conformation observed in the apo structure (Fig. 5D). This contact is critical for LotA_N activation, as K6 diUb incorporating a Ub^prox F4A mutation cannot be cleaved (Fig. 4J). The so-called “His loop” preceding the general base histidine is also shifted by >3Å between structures (Fig. 5D). Opposing this loop is H68 of Ub^prox, which makes a hydrogen bond to the backbone carbonyl of E235 in the complex structure (Fig. 5D). This interaction is also important for LotA_N activation, as K6 diUb incorporating a Ub^prox H68A mutation is cleaved less efficiently than wild-type (Fig. 5E). This function of Ub^prox H68 relies on both hydrogen bonding and steric effects, as Ub^prox H68Q or H68F mutants are also cleaved less efficiently (Fig. 5E). Thus, Ub binding at the S1’ site in the correct, K6-specific orientation activates the LotA_N catalytic center through substrate-assisted catalysis, by which F4 and H68 of Ub^prox assist in the proper orientation of the Cys and His loops.

A mechanism of substrate-assisted catalysis is consistent with the observed lack of activity against the monomeric Ub-KG(TAMRA) substrate (Fig. 1D), wherein Ub would presumably bind at the S1 site but the attached KG(TAMRA) moiety occupying the S1’ site could not serve to activate the catalytic triad. LotA_N can, however, react with the monomeric Ub-PA activity-based probe (Fig. 1C). As expected, the Ub-PA reactivity observed in wild-type LotA_N is abolished or severely reduced in the backgrounds of the catalytic C13A or S1-site L128R mutations, respectively (Fig. 5F). Unexpectedly, reactivity was also lost in the S1’-site R145L and the hinge A193P mutants (Fig. 5F). To test if one molecule of Ub-PA is binding at the S1’ site to activate reactivity for a second, we generated an F4A mutant to block any role in activation (see Fig. 4J and 5D). Compared to wild-type Ub-PA, reactivity of the F4A mutant is severely reduced (Fig. 5G). Addition of wild-type but not F4A Ub (lacking the PA warhead) rescues reactivity of F4A Ub-PA by serving as a dedicated “activator” for substrate-assisted catalysis at the S1’ site (Fig. 5G). A prerequisite of substrate-assisted catalysis is movement of the LotA_N A-UBD, as the A193P hinge mutant overrides all Ub-PA reactivity (Fig. 5G).

LotA_N activity is required for restriction of K6 polyUb at the Legionella-containing vacuole

LotA, LotB, and LotC have all been shown to edit Ub modifications deposited onto the LCV^30,31,34. LotA regulates K6, K48, and K63 polyUb following localization via its C-terminal PI(3)P-binding domain³⁰ (Fig. 1A). Whether these signal types are independently regulated by LotA’s two OTU domains, however, had not been tested. We constructed L. pneumophila strains with genomic lotA mutations at either OTU active site (C13S, C303S) for comparison to the wild-type Lp01 and ΔlotA strains, and performed infections in HeLa-FcγRII cells expressing a reporter for K6 polyUb (see STAR Methods). Compared to the wild-type strain, deletion of lotA results in a significant increase in the number of K6-positive LCVs (Fig. 6A–B). While the LotA_M C303S mutant restricts K6 polyUb like wild-type, the LotA_N C13S mutant phenocopies lotA deletion (Fig. 6A–B). These findings are consistent with our in vitro specificity analyses (Fig. 1E–H), and suggest a delineation of roles during infection.

Figure 6: LotA_N restriction of K6 polyUb during L. pneumophila infection.

A. Representative images of HeLa FcγRII cells infected with the indicated L. pneumophila strains at an MOI of 2 for 4 h. Fixed cells were stained for L. pneumophila (red), HA-Ub-K6 (green), and DNA (blue). Arrows indicate the position of a bacterium in each channel. Images were collected at 100x magnification. Scale bars correspond to 10 μm.

B. Quantitation of K6-positive bacteria shown in (A). Infections were performed in triplicate and each value represents scoring from 100 LCVs. Significance was determined using a Welch’s t-test.

C. Quantitation of K6-positive bacteria from HeLa FcγRII cells infected with the indicated L. pneumophila strains at an MOI of 2 for 4 h. Infections were performed in triplicate and each value represents scoring from 100 LCVs. Significance was determined using a Welch’s t-test.

D. Luciferase-expressing L. pneumophila strains were used to infect mouse BMDMs at an MOI of 0.5. Intracellular growth was monitored via bioluminescence changes over time. The mean relative luminescence unit (RLU) and standard deviation of eight independent experiments are shown.

E. Representative images of HeLa FcγRII cells infected with the indicated L. pneumophila strains at an MOI of 2 for 4 h. Fixed cells were stained for L. pneumophila (red), VCP (green), and DNA (blue). Arrows indicate the position of a bacterium in each channel. Images were collected at 100x magnification. Scale bars correspond to 10 μm.

F. Quantitation of VCP-positive bacteria shown in (E), alongside infections with a secretion-deficient ΔdotA strain and strains carrying mutations in the LotA active sites. Infections were performed in triplicate and each value represents scoring from 100 LCVs. Significance was determined using a Welch’s t-test.

See also Figure S6.

As a further test of our structure-guided mechanistic studies, we designed two separate triple-mutant variants of LotA_N: “FRA” combines S1’-site F120A and R145L with the hinge A193P mutation, while “LRA” combines S1-site L128R, S1’-site R145L, and hinge A193P mutations. In vitro, neither variant could cleave K6 diUb, even at high concentration (Fig. S6A). L. pneumophila strains carrying either lotA triple-mutant variant failed to restrict K6 polyUb (Fig. 6C, S6B), demonstrating that the requirements for LotA_N activity in vitro hold true during infection.

The importance of LotA DUB activities for L. pneumophila growth was tested by infecting mouse bone marrow-derived macrophages (BMDMs) with the various mutagenized stains. As reported previously³⁰, deletion of lotA resulted in no significant defect in L. pneumophila intracellular growth in BMDMs, whereas a secretion-deficient ΔdotA strain failed to replicate (Fig. 6D). Strains carrying integrated catalytic mutations in either LotA_N (C13S) or LotA_M (C303S) behaved like wild-type and ΔlotA L. pneumophila, with no apparent growth phenotype (Fig. 6D).

One proposed signaling role for K6 polyUb is as a preferred target for the AAA ATPase VCP/p97/Cdc48, which extracts ubiquitinated targets for degradation. We examined VCP localization following infection with our mutagenized L. pneumophila strains. Infection with wild-type L. pneumophila resulted in low but significant VCP recruitment to the LCV compared with a secretion-deficient ΔdotA strain, as previously observed⁵⁵ (Fig. 6E–F, S6C). Loss of DUB activity through lotA deletion or double C13S C303S mutation starkly increased VCP recruitment (Fig. 6E–F, S6C). Individual LotA_N and LotA_M active site mutants both exhibited significantly more VCP recruitment than wild-type L. pneumophila, indicating nonredundant roles (Fig. 6E–F, S6C). These findings are consistent with a role for K6-, as well as potentially K48- and K63-linked polyUb, in recruiting VCP to the LCV during infection.

DISCUSSION

The identification of K6-specific enzymes in humans has proven difficult^{10,13,16,38,56}. Meanwhile, the first OTU domain of L. pneumophila LotA is extremely specific and highly active toward K6 polyUb. Perhaps coincidentally, the most K6-specific E3 ligase identified to date, NleL, also originates from bacteria²⁶. Just as NleL has become an invaluable tool for studying K6 polyUb⁴⁵, LotA_N will be equally useful, particularly in applications such as UbiCRest analyses. Prior use of OTUD3 as a K6-targeted DUB has required careful consideration of enzyme concentration and secondary activities⁴⁶. We show that LotA_N maintains K6 specificity even at high concentration against long, complex polyUb chains, and will thus be a convenient tool alongside other polyUb-specific DUBs.

The exquisite specificity of LotA_N toward K6 polyUb arises from a blend of regulatory mechanisms (Fig. 7). In our apo structure of LotA_N, the open conformation of the A-UBD is incompatible with polyUb binding and the active site is misaligned. Consistent with inherent flexibility in these regions, the LotA_N Cys loop, His loop, and hinge region all exhibit above average B-factors. Before polyUb binding can occur, flexibility is required in the LotA_N hinge for the A-UBD to swing into position. This in situ formation of Ub-binding sites resembles the human DUB Cezanne binding to K11 diUb⁴⁴. In addition, only binding of a K6-linked Ub^prox would select for the catalytically-competent alignment of the Cys and His loops. The human DUB OTULIN uses a similar mechanism of substrate-assisted catalysis to ensure M1 polyUb specificity⁴². Unlike LotB and LotC, LotA_M also appears to require some form of substrate-assisted catalysis for active site rearrangement, in addition to its requirement for longer polyUb chains.

Figure 7: Model for LotA deubiquitinase activity.

Secreted LotA localizes back to the cytosolic face of the LCV via the PI(3)P-binding LotA_C domain. At the LCV, DUB activity of LotA_M restricts long K48 and K63 polyUb. LotA_N is kept inactive by a flexible A-UBD and inactive arrangement of the catalytic triad. Occupying a closed OTU:A-UBD conformation allows formation of an S1’ Ub-binding site. Binding of a K6-linked Ub into the S1’ site orients the LotA_N catalytic triad, allowing for hydrolysis of K6 polyUb and prevention of VCP recruitment. Created with BioRender.com.

Helical sub-domains inserted into VR-1 of the OTU domain are a defining feature of Lot-class DUBs. While the basic ~4-helix fold is similar, adaptations to the α1–2 region add versatility to the Ub interface, hence our proposed name Adaptive Ubiquitin-Binding Domain, or “A-UBD” (Fig. 4K, S4E). The A-UBD α1–2 regions of LotA_N, LotA_M, LotB, LotC, as well as OtDUB all play important roles in mediating Ub interactions^33,34,50,54. It is remarkable that this domain, now observed in at least 5 cases across two bacterial species, has adopted such distinct Ub interactions. To what extent the A-UBD is incorporated into other regulators of Ub signaling across bacteria, viruses, and eukaryotes will be an interesting area of future research.

During infection, LotA_N and LotA_M regulate distinct polyUb signals on the LCV surface (Fig. 7). The close spacing of the two OTU domains (~20 amino acids), however, raises the interesting possibility that they work cooperatively on a large or particularly complex polyUb signal. Though no apparent growth phenotype was observed for the C13S mutant within LotA_N, functional redundancies among the ~300 secreted L. pneumophila effectors often mask phenotypic differences in single-mutant strains. The DUB activity of LotA_M was previously shown to genetically interact with the SidE family of secreted E3 Ub ligases³⁰. Whether the K6-specific activity of LotA_N functions in concert with other secreted effectors remains to be seen. LotA’s two DUB activities play nonredundant roles in preventing recruitment of the AAA ATPase VCP/p97/Cdc48 to the LCV. Thus, one possible evolutionary rationale for LotA’s two OTU domains with distinct activities and specificities is that they guard different subsets of proteins on the LCV from VCP-dependent extraction and proteasomal degradation. K6 polyUb signals have also been previously tied to the autophagic destruction of mitochondria and bacteria by the E3 ligase Parkin^8,57, raising the possibility that LotA acts together with other L. pneumophila effectors to restrict host autophagic defenses⁵⁸. Whether and how LotA_N might fit into this or other K6 signaling networks, and what additional facets of K6 polyUb biology might be uncovered as a result, will be an exciting area of future work.

Limitations of the study

While we could reliably demonstrate the exquisite specificity of LotA_N, even at high enzyme concentration against long polyUb chains, we cannot predict how additional complications such as Ub acetylation or phosphorylation might impact LotA_N’s activity or specificity. Furthermore, our testing of LotA_N in UbiCRest analyses of complex, heterotypic, and branched polyUb chains was limited to the few Ub conjugation systems known to assemble K6 linkages. Future applications of LotA_N to the study of K6 polyUb will build upon this foundation and establish its reliability as a steadfast research tool.

Studying the role of LotA’s K6 specificity during infection is limited by several parameters. Firstly, the lack of a readily available detection reagent for K6 polyUb necessitated the expression of the exogeneous HA-Ub-K6 construct. While we are confident that HA-Ub-K6 is a reliable reporter of K6 polyUb in this case (owing to the strict specificity of LotA_N), expression levels could influence the amount of K6 polyUb we observe. Our ability to assess the relevance of K6-specific DUB activity on L. pneumophila growth was also hindered by redundancies among the ~300 secreted effectors. The importance of LotA activity could also be more apparent in a different host organism, such as amoebae or ciliated protozoa. Regardless, our findings indicate that the study of LotA in mammalian cells will still yield new insights into the signaling roles for this atypical linkage type.

STAR METHODS

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jonathan N. Pruneda (pruneda@ohsu.edu).

Materials availability

All reagents generated from this study are available from the lead contact upon request.

Data and code availability

• Coordinates and structure factors for the apo and K6 diUb-bound LotA_N structures have been deposited in the Protein Data Bank under accession codes 7UYG and 7UYH, respectively, and will be publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and subject details

TOP10 Escherichia coli (ThermoFisher) were used for all cloning and plasmid propagation. Rosetta (DE3) E. coli (MilliporeSigma) were used for all recombinant protein expression. DH5αλpir E. coli (BEI Resources) were used for allelic exchange. All E. coli strains were grown at 37°C in Luria-Bertani (LB) media containing appropriate antibiotics.

Legionella pneumophila Philadelphia-1 (Lp01) were used as the wild-type strain in our studies⁶⁹. All mutations were introduced into the base L. pneumophila strain by allelic exchange and confirmed by PCR and DNA sequencing⁷⁰. Luciferase-expressing strains were constructed as described previously⁷¹. All L. pneumophila strains were grown at 37°C in liquid N-(2-acetamido)-2-aminoethanesulfonic acid (ACES; MilliporeSigma)-buffered yeast extract (AYE) media or on charcoal-yeast extract (CYE) plates, as described previously⁷².

HeLa-FcγRII cells (a kind gift from K. Arasaki at Tokyo University of Pharmacy and Life Sciences) were grown in Minimum Essential Medium α (MEMα; Gibco) supplemented with 10% fetal bovine serum (FBS) and 400 μg/mL hygromycin (Nacarai)⁵⁹.

Murine bone marrow-derived macrophages were prepared from A/J mice (A/JJmsSlc; Japan SLC) and maintained in RPMI 1640 Medium (GIBCO) supplemented with 5% L cell supernatant and 10% FBS as described previously⁶⁰.

Method details

Cloning and mutagenesis

The lotA gene was cloned from Legionella pneumophila Philadelphia-1 genomic DNA. Initial construct design was performed using Phyre2⁷³ and the LotA_M crystal structure⁵⁰, yielding construct boundaries of 1–300 and 287–543 for LotA_N and LotA_M, respectively. To support crystallization, additional C-terminal truncations of LotA_N led to the 1–294 construct used for determination of the apo LotA_N structure. Based on the apo structure, an additional LotA_N construct encoding amino acids 1–276 was used for determination of the K6 diUb-bound structure. All LotA constructs were cloned into the pOPINB vector, which encodes an N-terminal His tag followed by a 3C protease cleavage site. Cloning and mutagenesis were performed using Phusion DNA Polymerase (New England BioLabs) and TOP10 E. coli (ThermoFisher).

Protein expression and purification

All pOPINB-LotA constructs were expressed and purified similarly. Transformed Rosetta (DE3) E. coli were grown in Luria broth containing 35 μg/mL chloramphenicol and 50 μg/mL kanamycin at 37°C until an optical density (600 nm) of 0.6–0.8. Cultures were cooled to 18°C and induced with 0.2 mM IPTG for an additional 20 hours. Cells were harvested by centrifugation and resuspended in 25 mM Tris, 200 mM NaCl, 2 mM β-mercaptoethanol, pH 8.0 (Buffer A). Following a freeze-thaw cycle, cells were incubated on ice with lysozyme, DNase, and SigmaFAST protease inhibitor cocktail (MilliporeSigma) prior to lysis by sonication. Clarified lysates were applied to HisPur cobalt affinity resin (ThermoFisher) and washed with Buffer A prior to elution with Buffer A containing 300 mM imidazole. Eluted proteins were cleaved with 3C protease overnight during dialysis back to Buffer A at 4°C. Cleaved LotA proteins were concentrated using 10K MWCO Amicon centrifugal filters (MilliporeSigma) and applied to a HiLoad Superdex 75 pg size exclusion chromatography column (Cytiva) equilibrated in 25 mM Tris, 200 mM NaCl, 5 mM DTT, pH 8.0. Fractions were evaluated for purity by SDS PAGE, collected, concentrated, and quantified by absorbance (280 nm) prior to flash freezing and storage at −80°C.

Untagged Ub constructs were expressed from the pET-17b vector. Transformed Rosetta (DE3) Escherichia coli were grown by auto-induction in a modified ZYM-5052 media⁷⁴ containing 35 μg/mL chloramphenicol and 100 μg/mL ampicillin at 37°C for 24–48 h. Cells were harvested by centrifugation, resuspended, and lysed as above for LotA. Clarified lysates were acidified by dropwise addition of 70% perchloric acid to a final concentration of 0.5%. The mixture was allowed to stir on ice for 1–2 h prior to centrifugation. The supernatant was dialyzed against 50 mM sodium acetate, pH 4.5 overnight at 4°C. The protein was applied to a HiPrep SP FF 16/10 cation exchange column (Cytiva), washed with additional 50 mM sodium acetate, pH 4.5, and eluted over a linear gradient to a matched buffer containing 500 mM NaCl. Fractions were evaluated for purity by SDS PAGE, pooled, and dialyzed against 25 mM Tris, 200 mM NaCl, pH 8.0 overnight at 4°C. If necessary, further purification was performed using a HiLoad Superdex 75 pg size exclusion chromatography column (Cytiva) equilibrated in 25 mM Tris, 200 mM NaCl, pH 8.0. Purified Ub was concentrated using 3K MWCO Amicon centrifugal filters (MilliporeSigma), quantified by absorbance (280 nm), and flash frozen for storage at either −20°C or −80°C.

The Ub-PA activity-based probes were prepared using intein chemistry as described previously⁷⁵. Rosetta (DE3) Escherichia coli transformed with pTXB1-Ub(1–75) were grown in Luria broth containing 35 μg/mL chloramphenicol and 100 μg/mL ampicillin at 37°C until an optical density (600 nm) of 0.6–0.8. Cultures were cooled to 18°C and induced with 0.5 mM IPTG for an additional 20 hours. Cells were harvested by centrifugation and resuspended in 20 mM HEPES, 50 mM sodium acetate, 75 mM NaCl, pH 6.5 (Buffer B). After a freeze-thaw cycle, cells were incubated on ice with DNase, SigmaFAST protease inhibitor cocktail (MilliporeSigma), and PMSF prior to lysis by sonication. The clarified lysate was applied to chitin resin (New England BioLabs) and allowed to bind at 37°C for 2 h. The resin was washed with Buffer B, followed by a high salt wash with Buffer B containing 500 mM NaCl, followed by a final wash with Buffer B. The Ub-MesNa was allowed to form and elute from the chitin resin for 42 h at room temperature, following addition of Buffer B containing 100 mM MesNa. The eluted protein was concentrated using 3K MWCO Amicon centrifugal filters (MilliporeSigma) and applied to a HiLoad Superdex 75 pg size exclusion chromatography column (Cytiva) equilibrated in Buffer B. Fractions containing Ub-MesNa were evaluated for purity by SDS PAGE and pooled. To convert Ub-MesNa to Ub-PA, propargylamine hydrochloride was added in 1000-fold excess and, following adjustment of the pH to 8.5, incubated at room temperature for 3 h. The reaction was dialyzed against 25 mM Tris, 200 mM NaCl, pH 8.0 overnight at 4°C. The final product was concentrated using 3K MWCO Amicon centrifugal filters (MilliporeSigma), quantified by absorbance (280 nm), and stored on ice during immediate use.

Assembly of polyUb chains

All diUb and tetraUb chains were assembled and purified according to published methods⁷⁶, with the exception of K27 diUb and K11 tetraUb, which were purchased from UbiQ and R&D Systems, respectively. K6 diUb for fluorescent labeling or for experiments that incorporate Ub^prox mutations were assembled using a “capped” strategy that allows directed formation of K6-linked diUb from dedicated donor and acceptor Ub moieties that are incorporated as Ub^dist and Ub^prox, respectively. The dedicated donor Ub construct incorporated K6R and K48R mutations, but retained an intact C-terminus. The dedicated acceptor Ub construct incorporated a K48R mutation and lacked the C-terminal LRGG motif. Effects of Ub^prox mutation were assessed by incorporating additional mutations into the dedicated acceptor Ub construct. Alternatively, a Ub^prox for fluorophore labeling was incorporated with an acceptor Ub construct that coupled a deletion of G76 with a G75C mutation. Equimolar amounts of donor and acceptor Ub variants were combined with 100 nM UBE1, 600 nM UBE2L3, 10 μM NleL, and 10 mM ATP in buffer containing 40 mM Tris, 10 mM MgCl₂, 0.5 mM DTT, pH 8.5. Reactions were incubated at 37°C for 16 h and subsequently quenched by addition of 10 mM DTT and 50-fold dilution into 50 mM sodium acetate, pH 4.5. The reactions were applied to a HiPrep SP FF 16/10 cation exchange column (Cytiva), washed with additional 50 mM sodium acetate, pH 4.5, and eluted over a linear gradient to a matched buffer containing 500 mM NaCl. Fractions were evaluated for purity by SDS PAGE, pooled, and dialyzed against 25 mM Tris, 200 mM NaCl, pH 8.0 overnight at 4°C. The diUb product was concentrated using 3K MWCO Amicon centrifugal filters (MilliporeSigma), quantified by absorbance (280 nm), flash frozen, and stored at either −20°C or −80°C.

Fluorescent K6 diUb was prepared using the “capped” strategy described above, which allows specific labeling of a cysteine incorporated at position 75 of Ub^prox. Purified K6 diUb was diluted to 50 μM in 25 mM sodium phosphate, 150 mM NaCl, 10 mM DTT, pH 7.4 and incubated at room temperature for 30 min. The diUb was subsequently buffer exchanged into 25 mM sodium phosphate, 150 mM NaCl, pH 7.4 over a PD-10 desalting column (Cytiva). ATTO488-maleimide (ATTO-TEC) was added in 5-fold molar excess to 1 mL of the desalted K6 diUb, and incubated at room temperature in the dark for 18 h. The labeled K6 diUb was again desalted to remove excess fluorophore, quantified by absorbance (488 nm), and flash frozen for storage at −80°C.

Gel-based DUB assays

Gel-based DUB assays were performed according to published methods⁷⁷. DUBs were prepared at 2X concentration in 25 mM Tris, 150 mM NaCl, 10 mM DTT, pH 7.4 and incubated at room temperature for 10 min. Equal volumes of 2X DUB and either 10 μM diUb or 5 μM tetraUb were mixed and incubated at 37°C. Samples were quenched in Laemmli sample buffer containing 0.2 M dithiothreitol (DTT) and analyzed by SDS PAGE.

Ub-PA reactivity assays

Ub-PA reactivity assays were performed according to published methods⁷⁷. DUBs were diluted to 10 μM in 25 mM Tris, 150 mM NaCl, 10 mM DTT, pH 7.4 and incubated at room temperature for 10 min. Equal volumes of 2X DUB and 100 μM Ub-PA were mixed and allowed to react at room temperature for up to 18 h, prior to being quenched in Laemmli sample buffer containing 0.2 M DTT and analyzed by SDS PAGE. For rescue experiments with the F4A Ub-PA variant, 25 μM wild-type or F4A Ub was added to the above reaction.

Fluorescence-based DUB assays

Ub- and Ubl-KG(TAMRA) substrates were prepared as previously described^37,78. Cleavage of the Ub- and Ubl-KG(TAMRA) substrates was monitored by fluorescence polarization according to published methods⁷⁷. Enzymes and substrates were prepared at 2X concentration in 25 mM Tris, 100 mM NaCl, 5 mM β-mercaptoethanol, 0.1 mg/mL BSA, pH 7.4 and mixed 1:1 to initiate the reaction. All substrates were held at 50 nM concentration, while protease concentrations were adjusted according to their specific activities. Fluorescence polarization was recorded using a BMG LabTech ClarioStar plate reader with an excitation wavelength of 540 nm, an LP 566 nm dichroic mirror, and an emission wavelength of 590 nm. Reactions were performed in Greiner 384-well small-volume HiBase microplates with 20 μL reaction volumes. Three technical replicates were performed for each reaction, with additional wells containing KG(TAMRA) and Ub/Ubl-KG(TAMRA) as positive and negative controls, respectively. Fluorescence polarization was recorded once per minute over a 1–2 h time period. Data collected from the positive and negative control wells were used to convert fluorescence polarization measurements to fraction of substrate remaining.

DUB assays using the ATTO488-labeled K6 diUb substrate were performed as above, with minor changes. Fluorescence polarization was monitored using an excitation wavelength of 482 nm, an LP 504 nm dichroic mirror, and an emission wavelength of 530 nm. Assay concentrations were 10 nM LotA_N and 50 nM ATTO488-labeled K6 diUb, except for the LotA_N triple-mutants, which were also assayed at a higher enzyme concentration of 1 μM. ATTO488-labeled monoUb and diUb were used as positive and negative controls, respectively, and used to convert fluorescence polarization measurements to fraction of substrate remaining. To determine Michaelis-Menten parameters, LotA_N concentration was held at 4 nM and initial rates of cleavage were determined from the linear range of data produced over the first 15–25 minutes of the reaction, across a range of substrate concentrations from 100 nM to 50.1 μM. All substrate concentrations maintained a fixed, 100 nM concentration of ATTO488-labeled K6 diUb, while unlabeled K6 diUb (shown to be hydrolyzed equally well) was used for the remainder. Three technical replicates were performed for each reaction, and three separate trials were performed for each K6 diUb concentration. Positive and negative control measurements were used to convert fluorescence polarization data to amount of substrate consumed, which was used to derive initial rates. Prism 9 was used for Michaelis-Menten analysis of the initial rate data.

UbiCRest analysis

Linkage-specific polyUb chains were assembled according to published methods⁷⁶. NleL and HUWE1 chain assemblies were performed at 37°C for 2 h in 25 mM Tris, 150 mM NaCl, 10 mM MgCl₂, 0.5 mM DTT, pH 7.4 using 375 nM UBE1, 1.5 μM Lys-less UBE2L3, 5 mM ATP, and either 3.75 μM NleL (aa 170–782) or 15 μM HUWE1 (aa 3993–4373). Prior to DUB treatment, all polyUb assembly reactions were quenched by addition of 50 mM EDTA and 5 mM DTT. DUBs were diluted to 10 μM in 25 mM Tris, 150 mM NaCl, 10 mM DTT, pH 7.4 and incubated at room temperature for 10 min. Equal volumes of polyUb assembly and DUB were mixed and incubated at 37°C for 2 h, prior to quenching in Laemmli sample buffer containing 0.2 M DTT. Reactions were resolved by SDS PAGE prior to transfer onto PVDF membranes using a Trans-Blot Turbo system (BioRad). Membranes were blocked at room temperature for 30 minutes with TBS-T (Tris-buffered Saline containing 0.1% Tween-20) containing 5% milk, then incubated with primary anti-Ub antibody (MilliporeSigma, MAB1510-I; 1:1,000 dilution) at 4°C overnight. Membranes were washed in TBS-T and incubated with secondary antibody (MilliporeSigma, #12–349; 1:5,000 dilution) at room temperature for 1 hour before performing additional TBS-T washes and detection using Clarity ECL reagent (BioRad).

Protein crystallization and structure determination

LotA_N (1–294), LotA_N (1–276), and K6 diUb were prepared as described above and exchanged into 25 mM Tris, 125 mM NaCl, 5 mM DTT, pH 7.4. LotA_N (1–294) was concentrated to 12 mg/mL and crystallized in hanging drop format with 30% PEG 2K MME, 0.2 M KI, 0.1 M MES pH 6.5 at 20°C in a 1 μL drop with 1:1 protein:precipitant ratio. LotA_N (1–276) was mixed with an equimolar amount of K6 diUb and concentrated using a 3K MWCO Amicon centrifugal filter (MilliporeSigma) to a total protein concentration of 8 mg/mL (determined by Bradford Assay). The protein complex was crystallized in hanging drop format with 25% PEG 3350, 10% ethylene glycol, 0.2 M KSCN, 0.1 M MES pH 5.5 at 20°C in a 1 μL drop with 1:1 protein:precipitant ratio. Apo LotA_N crystals were cryoprotected in mother liquor containing 12.5% glycerol, while complex LotA_N crystals were cryoprotected in mother liquor containing 25% glycerol prior to vitrification.

Diffraction data were collected at the Stanford Synchrotron Radiation Lightsource (SSRL), beamline 9–2. Images were integrated using XDS⁶¹ and scaled using Aimless⁶³. The apo LotA_N (1–294) structure was determined by SAD using the SHELXC/D/E pipeline in CCP4i2^62,64, based on anomalous signal arising from bound iodide ions that were present in the crystallization condition. Automated model building was performed using ARP/wARP⁶⁵, followed by iterative rounds of manual model building in COOT and refinement in PHENIX^67,68. Residues 2–275 could be confidently modeled into the electron density. The complex LotA_N (1–276) structure bound to K6 diUb was determined by molecular replacement with Phaser in CCP4i2, using the apo LotA_N and Ub (PDB 1UBQ) structures^62,66,79. Model building and refinement were performed as above using COOT and PHENIX^67,68. All figures were generated using PyMOL (www.pymol.org).

Legionella strains and culture

L. pneumophila strains used in this study are listed in Table S1. Point mutations in genomic lotA were introduced by a gene knock-in method, whereby mutated lotA genes were integrated into the ΔlotA strain at the endogenous locus³⁰. Plasmids used for the integration were constructed based on pSR47S-ΔlotA³⁰, which contains homologous regions upstream and downstream of the endogenous lotA locus inserted into the pSR47S gene-replacement vector as follows⁸⁰. The linearized pSR47S-ΔlotA was generated by PCR using primers lpg2248KI-1 and lpg2248KI-2. Fragments of the mutated lotA genes were obtained by PCR using primers lpg2248KI-3 and lpg2248KI-4 from pMMB207–3xFLAG-LotA C13S³⁰, 3xFLAG-LotA C303S³⁰, 3xFLAG-LotA C13S/C303S³⁰, pOPINB-LotA F120A/R145L/A193P, and pOPINB-LotA L128R/R145L/A193P. The vector and lotA fragments were ligated using Gibson assembly, and the resulting plasmids were transformed into E. coli DH5αλpir. The allelic exchange was conducted as described previously⁷⁰. The genomic mutations of lotA were validated by PCR and subsequent DNA sequencing. The L. pneumophila strains were grown at 37°C in liquid N-(2-acetamido)-2-aminoethanesulfonic acid (ACES; MilliporeSigma)-buffered yeast extract (AYE) media or on charcoal-yeast extract (CYE) plates, as described previously⁷².

Transfection and infection

HeLa-FcγRII cells were seeded on cover slips in 24-well tissue culture plates at 5 × 10⁴ cells per well 24 h before transfection. If necessary, transfection was performed using Lipofectamine2000 (Invitrogen) and 500 ng of pRK5-HA-Ub-K6 (Addgene, #22900) for 24 h according to the manufacturer’s recommendation. Because of the lysine-to-arginine mutations and the encoded N-terminal epitope tag, HA-Ub-K6 can only be utilized for a) monoubiquitination, b) as a distal capping Ub on the end of a polyUb chain, or c) incorporation into a K6-linked chain. Owing to its stringent requirement for substrate-assisted catalysis, LotA_N would only be able to hydrolyze HA-Ub-K6 that is incorporated into K6-linked chains. For infection, L. pneumophila was cultivated for 24 h in AYE media (with starting A₆₀₀ of 0.2) at 37°C with shaking. The cells were infected with the Legionella strains opsonized with anti-Legionella rat antibody (Scrum, #17457; 1:3,000 dilution) or with anti-Legionella rabbit antibody (BioAcademia, #64–100; 1:3,000 dilution) at a multiplicity of infection (MOI) of 2. After adding bacteria to the cells, the plates were centrifuged at 200 × g to precipitate bacteria onto the layer of cells and were immediately warmed in a 37°C water-bath by floating for 5 min and then placed in a CO₂ incubator at 37°C. At 1 h after infection, the infected cells were washed three times with prewarmed Dulbecco’s Phosphate Buffered Saline (DPBS; MilliporeSigma) to remove the extracellular bacteria and refreshed with prewarmed media, then incubation was resumed at 37°C in a CO₂ incubator. At 4 h after infection, the cells were washed three times with DPBS and fixed with 4% (w/v) paraformaldehyde for 20 min.

Immunofluorescent microscopy

For detection of HA-Ub-K6, the fixed HeLa-FcγRII cells on coverslips were permeabilized by treating with cold methanol for 60 sec and washed three times with DPBS. The samples were blocked with 2% (v/v) goat serum in DPBS and treated with anti-HA rabbit antibody (MBL, #561) with 1:200 dilution for 90 min. After three times washing with DPBS, the samples were treated with anti-Legionella rat antibody (Scrum, #17457; 1:3,000 dilution) for 60 min. After three times washing with DPBS, the samples were treated with secondary fluorescent antibodies, Alexa 488 goat anti-rabbit IgG (Invitrogen, #A11034; 1:500 dilution) and Alexa 568 goat anti-rat IgG (Invitrogen, #A11077; 1:500 dilution), for 40 min. After washing three times with DPBS, the coverslips were rinsed with distilled water and mounted on glass slides with ProLong^™ Diamond Antifade Mountant with DAPI (Invitrogen, P36962). For detection of VCP, the fixed HeLa-FcγRII cells on coverslips were permeabilized and blocked by treating with 0.2% (v/v) Triton X-100 and 2% (v/v) goat serum in DPBS for 20 min. After washing three times in DPBS, the samples were then treated with anti-VCP mouse antibody (Abcam, #ab11433; 1:200 dilution) for 90 min, followed by treatment with anti-Legionella antibody (BioAcademia, #64–100; 1:3,000 dilution) for 60 min. After three additional washes in DPBS, the samples were treated with secondary fluorescent antibodies, Alexa488 goat anti-mouse IgG (Invitrogen, #A411029; 1:500 dilution) and RhodamineRedX anti-rabbit IgG (Invitrogen, #R6394; 1:1,000 dilution), for 40 min. After three final washes in DPBS, the coverslips were rinsed with distilled water and mounted on glass slides with ProLong^™ Diamond Antifade Mountant with DAPI (Invitrogen, P36962). Images were collected with an inverted microscope (TE2000-U; Nikon) equipped with a digital ORCA-ERA camera (Hamamatsu). Images were blinded and quantified by manual counting of immunostained LCVs. In the case of experiments involving HA-Ub-K6, only LCVs contained within cells expressing the HA-Ub-K6 construct were used for counting.

Legionella intracellular growth assay

The luminescent L. pneumophila growth assay was conducted as described previously³⁰. Mouse BMDMs (4 × 10⁴ cells per well) were prepared in a 96-well plate (#353296; Falcon) one day before infection. Cells were infected with L. pneumophila strains expressing luciferase at a MOI of 0.5, then incubated at 37°C under 5% CO₂. Intracellular bacterial growth was monitored by luminescence using a microplate reader (GloMax Explorer; Promega).

Quantification and statistical analysis

Welch’s t tests were performed on all microscopy quantifications using Prism 9 software with data from three independent infection experiments. P values less than 0.05 were considered significant, with annotations as follows: * (p < 0.05), ** (p < 0.01), *** (p < 0.001, **** (p < 0.0001).

Supplementary Material

1

Supplementary Table 1: Bacterial strains, plasmids, and primers used in this study, related to STAR Methods.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Ubiquitin (clone Ubi-1)	MilliporeSigma	Cat#MAB1510-I
Anti-Mouse IgG, HRP conjugate	MilliporeSigma	Cat#12–349
Anti-Legionella	Scrum	Cat#17457
Anti-Legionella	BioAcademia	Cat#64–100
Anti-HA	MBL	Cat#561
Anti-VCP	Abcam	Cat#ab11433
Anti-Rabbit IgG, Alexa488 conjugate	Invitrogen	Cat#A11034
Anti-Rat IgG, Alexa 568 conjugate	Invitrogen	Cat#A11077
Anti-Mouse IgG, Alexa488 conjugate	Invitrogen	Cat#A11029
Anti-Rabbit IgG, RhodamineRedX conjugate	Invitrogen	Cat#R6394
Bacterial strains
See Table S1 for a full list of bacterial strains
Chemicals and recombinant proteins
ATTO488-maleimide	ATTO-TEC	Cat#AD488
K27 diUb	UbiQ	Cat#UbiQ-015
K11 tetraUb	R&D Systems	Cat#UC-45–025
Deposited data
Crystal structure of LotAN	This study	PDB: 7UYG
Crystal structure of LotAN bound to K6 diUb	This study	PDB: 7UYH
Experimental models: Cell lines
Human: HeLa-FcγRII	Laboratory of K. Arasaki	Ref59
Mouse: Bone marrow-derived macrophages	A/J mice	Ref60
Oligonucleotides
See Table S1 for a full list of oligonucleotides
Recombinant DNA
See Table S1 for a full list of recombinant DNA
Software and algorithms
XDS	Ref61	Version Jan 31, 2020
CCP4i2	Ref62	Version 7.1.004
Aimless	Ref63	Version 0.7.4
SHELXC/D/E	Ref64
ARP/wARP	Ref65	Version 8.0
Phaser	Ref66	Version 2.8.3
PHENIX	Ref67	Version 1.17.1
Coot	Ref68	Version 0.9
Pymol	www.pymol.org	Version 2.2.2
Prism	GraphPad	Version 9

Highlights:

• L. pneumophila LotA encodes two functionally-distinct deubiquitinases

• LotA’s specificity for K6-linked poly-ubiquitin makes it a valuable research tool

• Structures reveal that K6 specificity arises from substrate-assisted catalysis

• LotA restricts K6 poly-ubiquitin and VCP recruitment during infection