Roles of the N domain of the AAA+ Lon protease in substrate recognition, allosteric regulation and chaperone activity



Degron binding regulates the activities of the AAA+ Lon protease in addition to targeting proteins for degradation. The sul20 degron from the cell-division inhibitor SulA is shown here to bind to the N domain of Escherichia coli Lon, and the recognition site is identified by cross-linking and scanning for mutations that prevent sul20-peptide binding. These N-domain mutations limit the rates of proteolysis of model sul20-tagged substrates and ATP hydrolysis by an allosteric mechanism. Lon inactivation of SulA in vivo requires binding to the N domain and robust ATP hydrolysis but does not require degradation or translocation into the proteolytic chamber. Lon-mediated relief of proteotoxic stress and protein aggregation in vivo can also occur without degradation but is not dependent on robust ATP hydrolysis. In combination, these results demonstrate that Lon can function as a protease or a chaperone and reveal that some of its ATP-dependent biological activities do not require translocation.


An Escherichia coli cell contains more than 4000 different proteins, with wide variations in copy numbers. Under conditions that result in protein misfolding, about half of cytosolic protein degradation in E. coli is dependent on the AAA+ Lon protease (Chung and Goldberg, 1981). Lon appears to recognize some substrates by binding to degrons consisting largely of hydrophobic residues that are exposed as a consequence of unfolding or misfolding (Gur and Sauer, 2008). Lon also degrades natively folded proteins, including the SulA cell-division inhibitor, which contains an exposed C-terminal degron that is recognized by Lon (Higashitani et al., 1997; Gur et al., 2012). Lon proteases are present in most bacteria, in archaea, and in the endosymbiotic organelles of eukaryotes (Gur, 2013). Lon is necessary for rapid cell-cycle progression or pathogenicity in some bacteria, knock-down of mitochondrial Lon kills lymphoma cells, and overexpression of Lon increases fungal lifespan (Wright et al., 1996; Robertson et al., 2000; Ingmer and Brøndsted, 2009; Luce and Osiewacz, 2009; Bernstein et al., 2012; Breidenstein et al., 2012; Gora et al., 2013).

Like other AAA+ proteases, Lon sequesters its proteolytic active sites within a chamber, uses a hexameric ring and ATP hydrolysis to unfold and translocate proteins through a narrow axial pore into this chamber, and recognizes substrates predominantly by binding to degrons or peptide tags (Cha et al., 2010; Sauer and Baker, 2011). Unlike many AAA+ proteases, however, the AAA+ ATPase module and protease domain of Lon are part of a single polypeptide, and degron binding regulates Lon ATPase and protease activity in addition to serving a recognition function. For example, when otherwise identical proteins are tagged with either the sul20 or β20 degron, which correspond respectively to the C-terminal 20 residues of SulA and a β-galactosidase sequence buried in the native protein, the maximal rate of E. coli Lon degradation can differ by fivefold or more (Higashitani et al., 1997; Ishii and Amano, 2001; Gur and Sauer, 2008; 2009). These results suggest that degron binding shifts Lon into conformations with higher or lower protease activity.

In addition to its AAA+ module and peptidase domain, E. coli Lon contains a family-specific N domain that is necessary but not sufficient for hexamerization and has been proposed to be involved in substrate binding (Roudiak and Shrader, 1998; Ebel et al., 1999; Rudyak and Shrader, 2000; Melnikov et al., 2008; Adam et al., 2012). Consistently, N-domain mutations or truncations result in defects in Lon activity in vitro (Cheng et al., 2012), but substrate binding to the N domain has not been directly demonstrated. Crystal structures (in non-native oligomeric states) are known for regions of the N domain, and recent electron-microscopy studies suggest that formation of Lon dodecamers is also mediated by the N domain (Li et al., 2005; 2010; Duman and Löwe, 2010; Vieux et al., 2013).

Here, we show that the sul20 degron binds to the N domain of E. coli Lon and identify mutations that define this binding site. Degron binding to this site is not required for proteolysis of sul20-tagged substrates in vitro but enhances degradation by allosterically activating protease activity. Using additional Lon mutations that affect ATP hydrolysis, translocation and proteolysis, we also probe the requirements for SulA inactivation and suppression of proteotoxic stress in vivo. SulA inactivation requires binding to the sul20 binding site in the N domain and ATP hydrolysis but does not require translocation or proteolysis. Lon-mediated relief of proteotoxic stress and protein aggregation can also occur without protein degradation but does not require robust ATP hydrolysis or a functional sul20 binding site in the N domain. In combination, our results show that E. coli Lon can function as a protease or as a chaperone, reveal that some biological activities do not require translocation through the axial pore, and support a model in which substrate binding to multiple sites on the Lon enzyme can alter its conformation and biological activities.


The Lon N domain binds the sul20 degron

We initially sought to test if the sul20 degron binds to a site in the N domain of E. coli Lon. However, N-domain fragments do not form stable hexamers (Li et al., 2010), raising potential problems if substrate binding requires hexamerization or if interactions with hydrophobic surfaces normally buried in subunit–subunit interfaces create spurious non-specific binding. To circumvent these problems, we fused the Lon N domain to E. coli ClpXΔN, an AAA+ enzyme that forms stable ring hexamers. Chimera307 contained the entire Lon N domain (residues 1–307; Fig. 1A) fused to ClpXΔN, whereas chimera211 contained the first 211 residues of Lon, which included a globular region of the N domain but not an extended helical region (see Fig. 2B). In addition, chimera211 contained disulphide bonds between the subunits of ClpXΔN, which have been shown to stabilize functional covalent hexamers (Glynn et al., 2012). Both chimeras supported degradation of an ssrA-tagged substrate in the presence of ClpP, the proteolytic partner of ClpX (Fig. 1B). As ClpXΔN hexamerization is required for functional interaction with ClpP (Stinson et al., 2013), these results confirm that both fusion proteins can assemble into active hexamers.

Figure 1.

The N domain of Lon binds the sul20 degron.

A. Domain organization of E. coli Lon and Lon-ClpXΔN chimeras.

B. In combination with E. coli ClpP, chimera211 or chimera307 supported degradation of CM-titinI27-ssrA as assayed by SDS-PAGE. No substrate degradation was observed using ClpP alone. Pyruvate kinase (pk) was present for ATP regeneration.

C. Binding of a fluorescein-labelled sul20 peptide (200 nM) to LonS679A, chimera211, chimera307 or ClpXΔN was assayed by changes in fluorescence anisotropy (excitation 494 nm; emission 521 nm). Values are means ± SEM (n ≥ 2) after subtraction of the anisotropy of the free peptide. Solid lines are fits to a hyperbolic equation with fitted KDs of 4.7 ± 0.2 μM (LonS679A), 3.4 ± 0.4 μM (chimera211) and 1.5 ± 0.7 μM (chimera307).

D. Binding of 160 nM fluorescent sul20 peptide to 6 μM LonS679A hexamer or binding of 80 nM fluorescent sul20 peptide to 6 μM chimera211 hexamer was assayed by anisotropy in the presence or absence of the specified concentrations of non-fluorescent sul20 peptide or an unrelated control peptide (KREHGAANDENYCLAA).

Figure 2.

Identification of the sul20 binding site in the Lon N domain.

A. Binding of fluorescein-labelled sul20 peptide to LonS679A or variants containing multiple alanine-substitution mutations was assayed by fluorescence anisotropy as described in the Fig. 1 legend. For clarity, error bars (± SEM; n = 3) are only shown for the mutants with severe defects in sul20 binding.

B. The positions of residues 14–15 and 33–38, where mutations cause the largest defects in sul20 binding, are coloured red in a surface representation of the crystal structure of part of the N domain (3LJC.pdb; Li et al., 2010).

C. Lon33–35 (3 μM hexamer equivalents) behaved as a mixture of hexamers and dodecamers in sedimentation velocity analytical ultracentrifugation performed at 20°C and 16 000 r.p.m. in 50 mM HEPES (pH 7.5), 150 mM NaCl, 0.01 mM EDTA, 0.1 mM Tris (2-carboxyethyl) phosphine, 1 mM MgCl2 and 0.1 mM ATPγS.

As assayed by changes in fluorescence anisotropy, chimera307, chimera211 and LonS679A, a variant with the active-site proteolytic serine mutated to alanine (Amerik et al., 1991), bound to a fluorescently labelled sul20 peptide with KDs of ∼ 2–5 μM, whereas ClpXΔN alone did not bind this peptide (Fig. 1C). The binding of chimera211 and LonS679A to the fluorescent peptide was inhibited by an unlabelled sul20 peptide but not by a control peptide (Fig. 1D). Thus, the sul20 peptide binds specifically to a site contained within the first 211 residues of the N domain of E. coli Lon.

Mapping the sul20 binding site

We attached a UV-activatable cross-linker that contained a biotin and cleavable disulphide to a sul20 ‘bait’ peptide. Following incubation of the bait peptide with Lon, we activated cross-linking by UV irradiation, reduced the disulphide to remove the sul20 portion of the cross-linked moiety, cleaved with trypsin, and enriched for biotinylated peptides. Mass spectrometry identified a peptide with a mass (2022.8 Da) very close to that expected for Lon residues 100–113 plus the biotin label (2022.9 Da), suggesting that the sul20 binding site was within 14 Å (the linker length between the cross-linker and ‘bait’) of this peptide. Next, we performed alanine-scanning mutagenesis of solvent-exposed residues within 14 Å of residues 100–113 in the crystal structure of an N-domain fragment (3LJC.pdb; Li et al., 2010). We mutated blocks of two or three residues, purified the variants, and assayed for defects in sul20-peptide binding by fluorescence anisotropy. Mutating residues 14–15, 33–35 or 36–38 to alanines resulted in substantial loss of binding (Fig. 2A). All of these residues were close in the 3LJC crystal structure, suggesting that all of these mutations affect the same binding site (Fig. 2B). We focused further studies on the R33A/E34A/K35A variant, henceforth called Lon33–35, which had one of the largest defects in sul20 binding. Like the wild-type enzyme (Vieux et al., 2013), Lon33–35 sedimented as a mixture of hexamers and dodecamers in analytical-centrifugation experiments (Fig. 2C).

An allosteric role for the sul20 binding site in the N domain

The sul20-binding surface in the N domain could be essential for efficient high-affinity tethering of sul20-tagged substrates to Lon prior to transfer to the translocation and degradation machinery or might function to enhance proteolytic activity of these substrates by an allosteric mechanism (Gur and Sauer, 2009). To distinguish between these possibilities, we assayed Lon33–35 degradation of a set of model substrates that are degraded by wild-type Lon. These substrates included sul20- and β20-tagged derivatives of cp6-SFGFP, a readily degraded circularly permuted variant of superfolder GFP, FITC-casein, a fluorescent β20 peptide, and sul20-tagged native and unfolded variants of the titinI27 domain (Gur and Sauer, 2008; 2009; Wohlever et al., 2013).

We initially assayed rates of Lon33–35 and wild-type Lon degradation of different concentrations of cp6-SFGFP-sul20 by monitoring changes in native fluorescence. Lon33–35 degraded this substrate with a Vmax ∼ 20-fold lower than wild-type Lon but with a similar KM (Fig. 3A; Table 1). This result supports an allosteric model in which sul20-degron binding to the 33–35 site in the wild-type N domain enhances proteolysis. It does not support a model in which the function of degron binding to the 33–35 sites is to increase the local concentration of a substrate that is then degraded directly by Lon, as this tethering model predicts an increase in the KM for Lon33–35 degradation of cp6-SFGFP-sul20 but no change in Vmax. We found that Lon33–35 degraded β20-cp6-SFGFP and FITC-casein with KM and Vmax values similar to wild-type Lon (Fig. 3B and C; Table 1), establishing that the 33–35 substitutions do not cause general defects in Lon degradation of all substrates. In combination, these results and experiments reported below support the existence of two types of binding sites for the sul20 degron; the N-domain sites defined by the 33–35 substitutions, which appear to mediate allosteric activation of protease activity upon sul20 binding, and a distinct site or sites responsible for determining the KM for cp6-SFGFP-sul20 degradation.

Figure 3.

Substrate degradation and ATP hydrolysis by Lon33–35 and wild-type Lon.

A. Degradation of different concentrations of cp6-SFGFP-sul20 assayed by fluorescence (excitation 467 nm; emission 511 nm). Values are means ± SEM (n ≥ 2).

B. Degradation of different concentrations of β20-cp6-SFGFP assayed by fluorescence (excitation 467 nm; emission 511 nm). Values are means ± SEM (n ≥ 2).

C. Degradation of different concentrations of FITC-casein (type II; Sigma) assayed by fluorescence (excitation 490 nm; emission 525 nm). Values are means ± SEM (n = 3).

D. Degradation of F-β20-Q peptide (2 μM) was assayed by fluorescence (excitation 320 nm; emission 422 nm) in the presence of increasing concentrations of titinI27-sul20. Values are averages (n = 5).

E. ATP hydrolysis by wild-type Lon or Lon33–35 was assayed in the presence of difference concentrations of cp6-SFGFP-sul20.

In (A)–(C), lines are fits to the Hill form of the Michaelis–Menten equation {rate = Vmax/[1 + (KM/[S])n]; see Table 1 for fitted parameters}. In (E), lines are fits to basal + amp/[1 + (Kapp/[S])n], where Vmax = basal + amp (see Table 1 for fitted parameters). Lon or Lon33–35 concentrations were 0.3 μM (hexamer equivalents) in (A)–(C) and 0.15 μM (hexamer equivalents) in (D)–(E). The wild-type Lon data in (A)–(B) was taken from Wohlever et al. (2013).

Table 1. Steady-state kinetic parameters
SubstrateLon variantProtein degradationATP hydrolysis
Vmax min−1 enz−1KM (μM)Hill constantVmax min−1 enz−1Kapp (μM)Hill constant
  1. Errors are from non-linear-least-squares fitting. The wild-type data for cp6-SFGFP-sul20 and cp6-SFGFP-β20 are from Wohlever et al. (2013). nd; not determined. Kapp values represent the substrate concentration required for 50% stimulation of ATP hydrolysis. Basal rates of ATP hydrolysis in the absence of protein substrate were 10 ± 1 min−1 enz−1 (wild-type Lon), 10 ± 1 min−1 enz−1 (Lon33–35) and 13 ± 3 min−1 enz−1 (LonY398A). Enzyme concentrations were calculated as hexamer equivalents.
cp6-SFGFP-sul20Wild type3.6 ± 0.313 ± 21.2 ± 0.1217 ± 169.3 ± 1.31.2 ± 0.1
33–350.15 ± 0.0113 ± 21.2 ± 0.484 ± 133 ± 0.81.5 ± 0.6
β20-cp6-SFGFPWild type2.3 ± 0.321 ± 6.01.2 ± 0.2143 ± 52.6 ± 0.21.7 ± 0.1
33–351.6 ± 0.114 ± 1.01.2 ± 0.2ndndnd
FITC-caseinWild type4.4 ± 0.416 ± 4.01.2 ± 0.1ndndnd
33–357.0 ± 2.031 ± 230.9 ± 2ndndnd
CM-titinI27-sul20Wild type17.1 ± 0.712 ± 1.01.4 ± 0.2200 ± 200.6 ± 0.11.3 ± 0.4
33–3510.5 ± 0.470 ± 5.01.4 ± 0.1270 ± 7013 ± 100.7 ± 0.2
Y398Andndnd23 ± 12.3 ± 0.53 ± 1.8
titinI27-sul20Wild type2.0 ± 0.129 ± 3.01.5 ± 0.2118 ± 51.0 ± 0.11.1 ± 0.1
33–350.8 ± 0.380 ± 700.9 ± 0.279 ± 83.5 ± 0.91.2 ± 0.3
CM-titinI27-β20Wild type5.5 ± 0.118 ± 1.01.4 ± 0.1174 ± 71.6 ± 0.20.9 ± 0.1
33–35ndndnd250 ± 505.0 ± 3.00.7 ± 0.2
Y398Andndnd25 ± 21.4 ± 0.41 ± 0.3

At low concentrations, some Lon substrates can activate degradation of other substrates with different degrons (Gur and Sauer, 2009). To test for an effect of the 33–35 mutations in this transactivation assay, we monitored cleavage of a fluorescent β20 peptide by Lon33–35 or wild-type Lon as a function of increasing concentrations of a non-fluorescent sul20-tagged protein, titinI27-sul20. As shown in Fig. 3D, low concentrations of titinI27-sul20 activated β20 cleavage by wild-type Lon, as expected for allosteric activation, whereas higher concentrations resulted in decreased cleavage, as expected for substrate competition for general proteolytic machinery. In contrast, titinI27-sul20 activated Lon33–35 β20 cleavage to a much smaller extent, supporting a role for sul20 binding to the 33–35 sites in the N domain in allosteric activation of protease activity.

Lon33–35 hydrolysed ATP at a rate ∼ 2.5 slower than wild-type Lon in the presence of saturating cp6-SFGFP-sul20 (Fig. 3E; Table 1). Modest decreases in the rate of ATP hydrolysis result in very slow degradation of GFP substrates by ClpXP (Martin et al., 2008a; Nager et al., 2011), and reduced ATPase stimulation probably amplifies the Vmax defect for Lon33–35 degradation of cp6-SFGFP-sul20. Indeed, compared to wild type, Lon33–35 showed smaller but still substantial Vmax defects in degrading native titinI27-sul20 and denatured CM-titinI27-sul20 (Table 1), with the latter variant being unfolded by carboxymethylation of cysteines normally buried in the protein core. For these substrates, KM for Lon33–35 degradation was also increased compared to wild-type Lon (Table 1). Thus, the 33–35 mutations alter the kinetics of degradation of sul20-tagged substrates in ways that depend on properties of the tagged protein. Indeed, the KM for sul20-tagged substrates by wild-type Lon can also vary substantially depending on the model substrate (Wohlever et al., 2013). No simple model allowed quantitative fitting of all of these experimental results, an unsurprising outcome given the complexity of this system.

Requirements for SulA inactivation in vivo

Inactivation of SulA by Lon is required for resumption of growth following repair of UV-induced DNA damage (Gottesman et al., 1981). To test the importance of different Lon activities in SulA inactivation, we used mutations that prevent or diminish specific biochemical functions. One mutation (S679A) prevents proteolysis by inactivating the peptidase active sites. We also used Lon variants expected to be defective in substrate translocation as a consequence of the Y398A mutation in the axial pore, variants expected to be defective in ATP hydrolysis as a consequence of the E424Q mutation in the Walker-B motif, and variants containing the 33–35 mutations. For these studies, wild-type Lon or mutant variants were cloned into low-copy plasmids, which were transformed into E. coli strains lacking the chromosomal lon gene. Following UV irradiation, plasmids expressing wild-type Lon, LonS679A or LonY389A/S679A rescued growth (Fig. 4A). Western blots revealed that these Lon variants were expressed at levels similar to or slightly lower than chromosomal Lon (Fig. 4B). Overexpression of LonS679A was previously shown to inactivate SulA in vivo (Van Melderen and Gottesman, 1999). Our results confirm that Lon degradation is not required for SulA inactivation, even when the enzyme is expressed at cellular levels roughly comparable to normal cellular Lon levels. Our results also suggest that efficient translocation into the proteolytic chamber is unnecessary for inactivation, as LonY389A/S679A should be translocation defective. In contrast, the empty vector and plasmids expressing Lon33–35, LonE424Q and LonE424Q/S679A failed to rescue growth (Fig. 4A), even though these variants were also expressed at levels similar to the other variants (Fig. 4B). Thus, Lon inhibition of SulA appears to require binding to the N-domain site defined by the 33–35 mutations as well as an activity affected by the E424Q mutation but not by the Y398A mutation.

Figure 4.

Inactivation of SulA.

A. To monitor Lon-mediated inactivation of SulA, E. coli W3110 Δlon::kanR cells transformed with low-copy pBAD33 plasmids expressing Lon or Lon variants were irradiated with UV light, dilutions were spotted onto an LB agar plate, and the plate was incubated overnight at room temperature.

B. As assayed by SDS-PAGE and Western blotting with anti-Lon antibodies, chromosomal Lon in E. coli strain W3110 and Lon variants expressed from pBAD33 plasmids in strain W3110 Δlon::kanR were present at roughly similar intracellular levels following UV irradiation and 1 h of growth at room temperature.

C. Binding of LonY398A (KD = 2.6 ± 0.1 μM) and LonE424Q (KD = 0.9 ± 0.1 μM) to the fluorescent sul20 peptide (200 nM). See Fig. 1 legend for conditions.

D. ATP hydrolysis by wild-type Lon, LonY398A and LonE424Q (0.15 μM hexamer equivalents) was assayed in the absence of substrate or in the presence of CM-titinI27-sul20 (20 μM) or CM-titinI27-β20 (20 μM).

E. LonY398A (0.6 μM hexamer equivalents) did not detectably degrade CM-titinI27-sul20 (10 μM) or CM-titinI27-β20 (10 μM) as assayed by SDS-PAGE.

F. LonE424Q (0.6 μM hexamer equivalents) degraded CM-titinI27-sul20 (10 μM) or CM-titinI27-β20 (10 μM) very slowly.

To confirm that LonY398A and LonE424Q were not defective in binding the sul20 degron and had the expected biochemical phenotypes, we purified both proteins and characterized their activities in vitro. Both mutants bound the sul20 peptide (Fig. 4C), with LonE424Q having slightly higher affinity. Thus, the inability of LonE424Q and LonE424Q/S679A to inactivate SulA in vivo is not correlated with a binding defect. As expected, LonE424Q displayed reduced rates of both basal and substrate-stimulated ATP hydrolysis (Fig. 4D). This reduced activity probably accounts for its inability to inactivate SulA. In contrast, LonY398A had basal ATP-hydrolysis activity similar to wild-type Lon, but displayed substantially reduced rates of ATP hydrolysis in the presence of sul20- or β20-tagged substrates (Fig. 4D). Thus, a high level of substrate stimulation of ATPase activity does not appear to be important for SulA inactivation. LonY398A did not degrade the unfolded CM-titinI27-sul20 or CM-titinI27-β20 substrates (Fig. 4E), a property consistent with the expected defect in translocation of substrates. Indeed, mutations analogous to Y398A in the axial pores of other AAA+ proteases also prevent substrate translocation and degradation (Siddiqui et al., 2004; Hinnerwisch et al., 2005; Park et al., 2005; Martin et al., 2008b). LonE424Q translocated and degraded CM-titinI27-sul20 and CM-titinI27-β20 at a very slow rate (Fig. 4F), suggesting that its very slow rate of ATP hydrolysis allows a correspondingly slow rate of proteolysis.

Suppression of proteotoxic stress occurs without degradation

We also tested the ability of different Lon variants to support growth of cells subjected to proteotoxic stress by high temperature, the absence of the chromosomal ClpXP and Lon proteases, and low levels of the DnaJ and DnaK chaperones (Tomoyasu et al., 2001). In this background at 42°C, expression of wild-type Lon from a low-copy plasmid allowed robust growth, as did expression of Lon33–35, LonE424Q, Lon33–35/S679A and LonE424Q/S679A (Fig. 5A). Thus, degradation, robust ATP hydrolysis, and binding of client proteins to the sul20 site in the N domain are not required for Lon's ability to suppress proteotoxic stress. Expression of LonS679A provided partial rescue, whereas expression of LonY398A/S679A gave no rescue (Fig. 5A). The latter result suggests that substrate translocation is required for suppression of proteotoxic stress, although the inactivity of LonY398A/S679A could potentially arise from another defect conferred by the Y398A mutation (see Discussion). The plasmid-expressed Lon variants used in these assays were again expressed at levels similar to each other and to Lon expressed from its normal chromosomal location (Fig. 5B). We note that LonS679A was partially active, whereas Lon33–35/S679A and LonE424Q/S679A were as active as wild-type Lon. These results suggest that the S679A mutation destabilizes the Lon conformation that is active in relieving stress, whereas the 33–35 and E424Q mutations stabilize this conformation. Indeed, LonE424Q bound the sul20 peptide approximately fivefold more tightly than LonS679A (Figs 1C and 4C), which could also reflect differential stabilization of different Lon conformations by these mutations. Notably, the Lon mutants that were active in inhibition of SulA were largely inactive in suppressing proteotoxic stress and vice versa (Figs 4A and 5A), emphasizing that these activities occur by distinct mechanisms.

Figure 5.

Suppression of proteotoxic stress and aggregation in vivo.

A. Lon-mediated rescue of cells from proteotoxic stress caused by growth at 42°C, under expression of DnaK and DnaJ, and deletion of the chromosomal genes for lon, clpX and clpP. E. coli strain BB7357 was transformed with pBAD33 plasmids expressing wild-type Lon or Lon variants, dilutions were spotted onto LB agar plates, and plates were incubated at 42°C for ∼ 12 h.

B. Expression levels of mutants used in (A) were determined by Western blotting and densitometry and normalized to the average value of Lon in E. coli strain W3110. Values are averages ± SD (n ≥ 2).

C. Following growth for 1 h at 42°C, aggregated proteins were purified from strain BB7357 transformed with an empty vector or with pBAD33 variants expressing wild-type Lon, Lon33–35/S697A or LonE424Q/S697A, and were visualized after SDS-PAGE by Coomassie-blue staining. The molecular weights of proteins in the SeeBlue* Plus2 standard mixture are shown (Life Technologies).

Under the conditions of the proteotoxic-stress assay, ∼ 1.5% of intracellular proteins are recovered as insoluble aggregates both in Δlon and in Δlon ΔclpXP cells, which led to the proposal that Lon degrades misfolded proteins that accumulate when DnaK and DnaJ are limiting (Tomoyasu et al., 2001; Rosen et al., 2002). To test this model, we purified aggregated proteins from cells without Lon and cells expressing wild-type Lon, Lon33–35/S679A or LonE424Q/S679A. As monitored by SDS-PAGE and straining with Coomassie Blue, the levels of aggregated protein were comparable in cells expressing wild-type Lon and both proteolytically inactive double mutants and were substantially lower than the levels in cells with no Lon (Fig. 5C). Thus, Lon can suppress aggregation by a mechanism that does not require proteolysis, robust ATP hydrolysis, or the 33–35 N-domain sul20 binding site.


The biochemical and genetic experiments described here provide evidence that the N domain of E. coli Lon binds the sul20 degron and this binding regulates proteolysis allosterically. A set of spatially adjacent residues in the N domain is required for efficient degradation of sul20-tagged substrates. For example, the 33–35 substitutions in the N domain dramatically weaken binding to a sul20 peptide, reduce Vmax for degradation of sul20-tagged substrates, and prevent Lon relief of SulA inhibition of cell division following DNA damage. Lon33–35 degrades some model substrates with steady-state kinetics similar to wild-type Lon, demonstrating that the 33–35 site is only required for the binding and/or efficient degradation of a subset of substrates. These mutations also prevent efficient transactivation of cleavage of a β20 peptide by a sul20-tagged protein, and alter stimulation of ATP hydrolysis by sul20-tagged substrates. These results support a model in which binding of the sul20 degron to the 33–35 site causes allosteric changes in the conformation of wild-type Lon that stimulate ATP hydrolysis and proteolysis. Although SulA is restricted to γ proteobacteria, Lon residues 33–35 are highly conserved across proteobacteria and Gram-positive bacteria, suggesting that this site serves to bind substrates with sul20-related degrons in these organisms.

The sul20 degrons of some cp6-SFGFP-sul20 molecules are proteolytically clipped by Lon without global degradation, implying that the sul20 tag is the first segment of the substrate to pass through the axial pore and enter the degradation chamber (Wohlever et al., 2013). However, we find that the sul20 degron also binds to a distinct site in the N domain of Lon. Lon33–35 degrades cp6-SFGFP-sul20 with the same KM as wild-type Lon, a fact inconsistent with a model in which the sul20 degron of this substrate initially binds to the N domain and is subsequently transferred to the axial pore. Rather, it appears that the sul20 degrons of some substrate molecules bind to the N domain and allosterically activate proteolysis of other substrate molecules whose degrons are independently engaged by the pore. Indeed, this model is supported by transactivation experiments reported here and previously (Gur and Sauer, 2009). We also find that the Y398A mutation, which truncates a highly conserved aromatic side-chain in the axial pore, prevents degradation of unfolded substrates bearing the sul20 or β20 degrons, dramatically reduces the maximal level of stimulation of ATP hydrolysis by these substrates, but does not impair binding of the sul20 degron to the N domain. These results are consistent with independent binding of sul20 degrons on different substrate molecules to the N domain and to the axial pore. Studies with other AAA+ proteases show that mutations corresponding to Y398A prevent or greatly slow the rate of substrate translocation (Siddiqui et al., 2004; Hinnerwisch et al., 2005; Park et al., 2005; Martin et al., 2008b). Thus, both allosteric activation via degron binding to the N domain and interaction of a translocating segment of polypeptide with the axial-pore loops appear to be required for normal co-ordination of substrate binding, ATP hydrolysis and substrate translocation by Lon.

Escherichia coli SulA, which inhibits cell division by binding to FtsZ (for review, see Löwe et al., 2004), can be inactivated by Lon in the absence of degradation. Van Melderen and Gottesman (1999) showed that overproduction of proteolytically inactive LonS679A rescued cell growth after UV irradiation, but LonS679A with a Walker-A mutation was not active in this assay. Because the Walker-A motif is required for ATP binding and hydrolysis, they proposed that LonS679A inactivates SulA by unfolding and translocating it into the inert proteolytic chamber of the mutant enzyme. Our results using LonE424Q, a Walker-B mutant, show that a robust level of ATP hydrolysis is required to inactivate SulA. However, LonY398A, which appears to be defective in substrate translocation, is fully active in inhibiting SulA in vivo. Together, these results suggest that an activity, which depends upon robust ATP hydrolysis but not upon efficient substrate translocation, is required for degradation-independent inhibition of SulA by Lon. The sul20 degron of SulA is not involved in FtsZ binding and should be accessible to Lon in the complex, based upon the crystal structure of Pseudomonas aeruginosa FtsZ●SulA (Higashitani et al., 1997; Cordell et al., 2003). It is possible, therefore, that Lon binds the sul20 degrons of one or more SulA molecules bound to FtsZ, with ATP hydrolysis then driving conformational changes in the N domains that allow Lon to strip SulA from FtsZ and prevent its rebinding. Indeed, regions of the Lon N domain have been shown to undergo nucleotide-dependent motions (Cheng et al., 2012).

Gur and Sauer (2009) found that the β20 degron, which is hydrophobic and likely to be similar to degrons in most misfolded proteins, stabilizes a Lon conformation with high ATPase activity and little or possibly no protease activity. In this conformation, they proposed that Lon unfolds β20-tagged substrates by translocating them through the axial pore but then releases them without degradation, giving misfolded proteins a chance to refold correctly. Although this model has not been ruled out, our SulA-inactivation results suggest that Lon can perform mechanical functions that require robust ATP hydrolysis but are not coupled to substrate translocation or degradation. Thus, the β20-stabilized conformation of Lon that hydrolyses ATP rapidly may be one in which translocation-independent conformational changes can be coupled to protein remodelling reactions.

The Lon enzymes from yeast, mammals and certain bacteria appear to mediate chaperone activity that is independent of proteolysis (Rep et al., 1996; Hori et al., 2002; Lee et al., 2004; Coleman et al., 2009). Indeed, in E. coli, we find that the protease-defective Lon33–35/S679A and LonE424Q/S679A mutants suppress proteotoxic stress and protein aggregation as well as the wild-type Lon enzyme. Thus, proteolytically inactive Lon variants can function as chaperones, and wild-type Lon may also prevent aggregation of misfolded proteins by a degradation-independent mechanism. Many chaperones functions by binding to exposed hydrophobic patches in client proteins (Fenton et al., 1994; Vabulas et al., 2010), and such binding could also be important for Lon suppression of aggregation. Consistent with this model, the E424Q mutation dramatically reduces ATP hydrolysis but does not alter chaperone activity. If simple binding of misfolded proteins to Lon suppresses their aggregation, then the inactivity of LonY398A/S679A in the stress assay is likely to arise from perturbations in the binding of specific substrates in or near the axial pore, as homologous mutations in the pores of ClpX and ClpA reduce binding of specific degrons (Siddiqui et al., 2004; Hinnerwisch et al., 2005; Martin et al., 2008b). Alternatively, the low level of ATP hydrolysis mediated by LonE424Q may be sufficient to remodel misfolded proteins, allowing them to refold properly rather than aggregating. In this model, the inactivity of LonY398A/S679A could also be explained by its translocation defect. Because Lon33–35/S679A is fully active in the proteotoxic-stress assay, binding of the Lon N domain to sul20-like degrons does not appear to be important in suppressing protein aggregation.

Our present results and previous studies (Gur and Sauer, 2009) provide evidence for three distinct activities of E. coli Lon. These activities are likely to correspond to different conformations of the hexamer and/or dodecamer, with the binding of certain degrons stabilizing specific conformations. The proteolytic activity of Lon requires ATP-fuelled translocation into the degradation chamber. Both hexamers and dodecamers have proteolytic activity, albeit with different substrate profiles (Vieux et al., 2013). Another activity, which mediates degradation-independent inactivation of SulA by variants such as LonY398A/S679A, appears to involve remodelling that requires robust ATP hydrolysis but is translocation independent. A third activity suppresses aggregation of misfolded proteins by a degradation-independent mechanism that does not require robust ATP hydrolysis and may or may not involve translocation. The second and third activities are likely to correspond to different enzyme conformations, but whether hexamers or dodecamers have both activities remains to be determined.

The family-specific N domains of the ClpX and ClpA AAA+ unfoldases serve as binding platforms for some substrates and adaptor proteins but can be deleted without compromising hexamer formation or robust degradation of certain substrates by ClpXP and ClpAP (Sauer and Baker, 2011). In contrast, the N domain of Lon is more highly integrated into overall enzyme architecture and function, including substrate binding, hexamer and dodecamer formation, and allosteric control of ATP hydrolysis and protease activity. Understanding in structural and dynamic terms how the N domain accomplishes these tasks is an important future challenge.

Experimental procedures

Protein cloning, expression and purification

Variants of E. coli Lon were cloned into pBAD33 or a variant in which the chloramphenicol resistance marker of pBAD33 was replaced with an ampicillin resistance marker from pSH21. E. coli ClpXΔN and chimeras were cloned into a HTUA vector and contained an N-terminal His6 tag followed by a TEV protease site. Chimera307 contained Lon residues 1–307, a two residue scar (EL, resulting from cloning into a SacI restriction site), and ClpXΔN (residues 62–424 of wild-type ClpX). Chimera211 contained Lon residues 1–211, a GSSG linker, the EL dipeptide and ClpXΔN. In addition, chimera211 contained the C39S Lon mutation and C169S ClpX mutation to remove exposed cysteines and the ClpX T66C and P388C mutations to form inter-subunit disulphide bonds to stabilize hexamer formation (Glynn et al., 2012). ClpP was cloned into a pET22b vector with a His6 tag on the C-terminus. TitinI27 variants were cloned into a pSH21 vector with an N-terminal His6 tag. β20-cp6-SFGFP and cp6-SFGFP-sul20 were cloned into a pCOLADuet1 vector with an N-terminal His6 tag followed by a PreScission protease site. Mutations were generated either by QuickChange PCR (Stratagene) or by standard PCR techniques.

Lon was overexpressed with minor modifications from a method described previously (Wohlever et al., 2013). Briefly, cells were grown at 37°C until OD600 = 1.0, induced with 0.2% arabinose at 37°C for 3.5 h, harvested, and resuspended in LBA buffer [100 mM potassium phosphate (pH 6.5), 1 mM DTT, 1 mM EDTA and 10% glycerol] to a final volume of 20 ml. Cells were incubated with lysozyme before sonication, and the crude cell lysate was cleared by high-speed centrifugation. The cleared lysate was incubated on ice for 20 min with 2 μl of benzonase (250 U ml−1, Sigma) and then bound to P11 phosphocellulose resin (Whatman) equilibrated in LBA buffer. This resin was washed twice with LBA buffer and once with LBA buffer plus 100 mM potassium phosphate (pH 6.5). Lon was eluted from the P11 resin using LBA buffer plus 300 mM potassium phosphate (pH 6.5). The eluant was filtered to remove phosphocellulose, polyethyleneimine (PEI) was added to a final concentration of 0.12% to precipitate nucleic acids, additional phosphocellulose was added to remove excess PEI, and the mixture was filtered, concentrated and chromatographed on an S200 column (GE Healthcare) equilibrated in 50 mM HEPES (pH 7.5), 2 M NaCl and 1 mM DTT. Peak fractions from this column were pooled, buffer exchanged into storage buffer [50 mM HEPES (pH 7.5), 150 mM NaCl, 10 μM EDTA, 1 mM DTT and 10% glycerol] and frozen at −80°C.

Escherichia coli ClpP, cp6-SFGFP-sul20, β20-cp6-SFGFP and titinI27 variants were expressed, purified and carboxymethylated (if necessary) as described (Kenniston et al., 2003; Gur and Sauer, 2009; Glynn et al., 2012; Wohlever et al., 2013). For 35S-labelling, cells were grown in a rich defined medium lacking methionine (TekNova) until OD600 = 0.6, and 35S-methionine (Perkin-Elmer) was added after 20 min of induction with 1 mM IPTG. 35S-labelled proteins were purified through the Ni-NTA step and then mixed at a 1:19 ratio with purified unlabelled substrate.

Cells expressing ClpXΔN and Lon-ClpXΔN chimeras were grown until OD600 = 1.0, induced with 1 mM IPTG for 3.5 h at room temperature, harvested, resuspended in lysis buffer [25 mM HEPES (pH 7.5), 400 mM NaCl, 100 mM KCl, 20 mM imidazole, 10% glycerol and 10 mM 2-mercaptoethanol] to a total volume of 20 ml, and lysed by incubation with lysozyme and sonication. Following lysis, 2 μl of benzonase (250 U ml−1, Sigma) and PMSF (final concentration 1 mM) were added, the lysate was cleared by high-speed centrifugation, and the supernatant was bound to Ni-NTA resin equilibrated in lysis buffer. The resin was washed with 30 ml of lysis buffer and eluted with lysis buffer plus 250 mM imidazole. For ClpXΔN and chimera307, the eluant was chromatographed on S300 column (GE Healthcare) equilibrated in 50 mM Tris (pH 8.0), 300 mM KCl, 1 mM DTT and 10% glycerol. Appropriate fractions were pooled, concentrated and frozen at −80°C. After elution of chimera211 from the Ni-NTA resin, the protein was buffer exchanged into low-salt buffer [25 mM HEPES (pH 7.5), 50 mM KCl, 5 mM MgCl2, 1 mM DTT and 10% glycerol], incubated with 1 μM TEV protease for 90 min at room temperature to remove the His6 tag, chromatographed on an S300 column as described above, and treated with copper phenanthroline to catalyse disulphide-bond formation between subunits as described (Glynn et al., 2012). The disulphide-bonded chimera211 was purified on a Superose 6 column, concentrated and frozen at −80°C as described above, except in a buffer lacking DTT. Anti-Lon antibodies used for Western blots were a gift from the Baker lab (MIT).


Peptides were synthesized, purified by reverse-phase HPLC, and masses were verified by mass spectrometry. The F-β20-Q peptide (sequence Z-QLRSLNGEWRFAWFPAPEAV-nY-A, where Z is a para-aminobenzoic acid fluorophore and nY is a nitrotyrosine quencher) was dissolved in dimethylsulphoxide and concentration was determined by absorbance (ε381 = 2200 M−1 cm−1). The sul20 peptide (sequence ASSHATRQLSGLKIHSNLYH) was dissolved in 50 mM HEPES (pH 7.5) and concentration was measured by absorbance (ε280 = 1490). The sul20 peptide with an N-terminal fluorescein was dissolved in 25 mM Tris (pH 8.0) and concentration was determined by absorbance (ε495 = 83 397 M−1 cm−1).

Biochemical assays

Unless noted, biochemical assays were performed in 25 mM Tris (pH 8.0), 100 mM KCl, 10 mM MgCl2, at 37°C using enzyme concentrations calculated for hexamer equivalents. Kinetic and anisotropy assays were performed in a SpectraMax M5 plate reader using 384-well clear plates (Corning) for absorbance assays and 96-well flat bottom ½-area plates (Corning) for fluorescence assays. ATPase assays contained supplemental 5 mM DTT, 2 mM ATP, lactate dehydrogenase (10 U ml−1), and an ATP regeneration system [rabbit muscle pyruvate kinase (Sigma, 10 U ml−1), 20 mM phosphoenolpyruvate (Sigma)]. The rate of ATP hydrolysis was measured by monitoring changes in absorbance at 340 nm, and reactions were initiated by the addition of MgCl2 that had been pre-warmed to 37°C. Degradation assays contained supplemental 1 mM DTT, an ATP regeneration system and 2 mM ATP, which was used to initiate the reaction. Fluorescent substrates were incubated in the plate reader until the fluorescence was constant prior to initiation of degradation. For degradation assays monitored by SDS-PAGE, 10 μl aliquots were taken at specified time points and mixed with 3.3 μl of 4× loading buffer [8% SDS, 250 mM Tris (pH 6.8), 40% glycerol, 160 mM DTT, and bromophenol blue]. The rate of degradation of 35S-labelled titinI27-sul20 variants was determined by measuring the amount of soluble radioactive products following precipitation with ice-cold trichloroacetic acid (Gottesman et al., 1998). The binding of fluorescent sul20 peptide to LonS679A and variants was measured in the presence of 1 mM ATPγS to prevent substrate translocation; fluorescence anisotropy values were corrected for G-factor and scattering and fitted to a hyperbolic equation to determine a KD value. Sedimentation-velocity ultracentrifugation was performed as described, except using proteolytically active Lon33–35 (Vieux et al., 2013).

Cross-linking and mass spectrometry

Reactions were performed in the dark until the photo-activation step. The sul20 peptide [1 mM in 50 mM HEPES (pH 7.5), 150 mM NaCl, 10 μM EDTA, 10% glycerol] was incubated with 1 mM Sulfo-SBED (Pierce) for 30 min at room temperature, precipitated material was removed by centrifugation, and unreacted cross-linker was removed by dialysis using a 2 kDa MWCO membrane. The cross-linker-modified sul20 peptide (200 μM) was incubated with 10 μM LonS679A (hexamer equivalents), 1 mM ATPγS and 1 mM MgCl2 at room temperature for 5 min. Cross-linking was initiated by UV irradiation (365 nm) with a handheld lamp at a distance of 2 cm for 15 min. To reduce the disulphide bond linking the sul20 peptide to the cross-linker and Lon, 100 mM 2-mercaptoethanol was added and the reaction was incubated at room temperature for 1 h. Free sul20 peptide was removed by two consecutive microbio spin columns (Bio-Rad). Labelling of Lon was verified by Western blotting with an anti-biotin antibody. The modified Lon protein was digested with sequencing grade trypsin (Roche) using a 1:100 enzyme:substrate ratio at 37°C for 14 h, and cleavage was quenched with 1 mM TLCK. Biotinylated peptides were enriched by passage over a Monomeric Avidin Resin (Pierce) and were eluted from this column with 100 mM glycine buffer (pH 2.8). Samples were loaded onto a reverse phase protein trap, which was desalted on-line and eluted isocratically, and then analysed by nanospray LC-MS using a QSTAR Elite quadrupole-time-of-flight mass spectrometer. Deconvolution of the electrospray data to generate molecular-weight spectra was performed with the BioAnalyst software included with the QSTAR Elite data system.

Biological assays

For assays of SulA inactivation in vivo, E. coli strain W3110 Δlon::kanR was transformed with pBAD33 or plasmid variants (camR) expressing wild-type Lon or Lon mutants. Cells were grown in Luria–Bertani (LB) broth to an OD600 of 0.9–1.3, diluted into fresh LB broth to an OD600 of 0.25, and 10-fold serial dilutions were prepared. Ten microlitres of each dilution was spotted onto an LB-agar plate containing 25 μg ml−1 kanamycin and 10 μg ml−1 chloramphenicol. The plate was exposed to 254 nm UV light from a handheld lamp at a distance of 5 cm for 10 s and was incubated overnight in the dark at 37°C.

Escherichia coli strain BB7357 expresses the DnaJ and DnaK chaperones from an IPTG-inducible promoter and lacks the chromosomal lon, clpX and clpP genes (Tomoyasu et al., 2001). At 42°C, BB7357 grows poorly when IPTG levels are low but an otherwise isogenic lon+ strain grows well under these conditions. For proteotoxic-rescue assays, we transformed strain BB7357 with a pBAD33 variant encoding ampicillin resistance (empty vector) or derivatives expressing wild-type Lon or Lon mutants. Cells were grown at 30°C in LB broth plus 1 mM IPTG until late-log phase, diluted to a final OD600 of 0.1, and serial fivefold dilutions were prepared in LB broth. Small aliquots of each dilution were then spotted onto LB agar plates containing 25 μM IPTG, 25 μg ml−1 kanamycin, 10 μg ml−1 chloramphenicol and 100 μg ml−1 ampicillin, and the plates were incubated overnight at 30°C or 42°C. For assays of protein aggregation, cells were grown at 30°C in LB broth plus 100 μM IPTG to mid-log phase, the culture was split in two, grown for an additional 60 min at 30 or 42°C, and then harvested. Aggregated proteins were purified by the method described by Tomoyasu et al. (2001), separated by SDS-PAGE, and visualized by staining with Coomassie Blue.


We thank B. Bukau, I. Levchenko, I. Papayannopoulos, E. Vieux, and members of the Sauer and Baker labs for reagents and helpful discussions, and D. Pheasant (MIT Biophysical Instrumentation Center) for help with the AUC experiments. T.A.B. is an employee of the Howard Hughes Medical Institute. This work was supported by N.I.H. Grant AI-16982 and by an N.S.F. Graduate Research Fellowship to M.L.W.