Mechanism of substrate unfolding and translocation by the regulatory particle of the proteasome from Methanocaldococcus jannaschii

Abstract

In the archaebacterium Methanocaldococcus jannaschii (M. jannaschii), the proteasomal regulatory particle (RP), a homohexameric complex of proteasome-activating nucleotidase (PAN), is responsible for target protein recognition, followed by unfolding and translocation of the bound protein into the core particle (CP) for degradation. Guided by structure-based mutagenesis, we identify amino acids and structural motifs that are critical for PAN function. Key residues line the axial channel of PAN, defining the apparent pathway of substrate translocation. Subcomplex II of PAN, comprising the ATPase domain, associates with the CP, and drives ATP-dependent unfolding of the substrate protein, whereas the distal subcomplex I forms the entry port of the substrate translocation channel. A linker between subcomplexes I and II was shown to be essential for PAN function, implying functional and perhaps mechanical coupling between these domains. Sequence conservation suggests that the principles of PAN function are likely to apply to the proteasomal RP of eukaryotes.

The 26S proteasome of eukaryotes contains a 20S core particle (CP) and a 19S regulatory particle (RP)^1–7. Functional and mechanistic investigations of the eukaryotic CP have been greatly facilitated by its atomic structures^8–10. In contrast, such studies of the eukaryotic RP have been hampered by a lack of structural information. The PAN complex from the archaebacterium M. jannaschii exhibits considerable structural homology to the eukaryotic RP and performs a similar set of essential functions – substrate binding, unfolding and translocation^11–13; yet its simple homohexameric nature offers a major advantage over the eukaryotic RP for structural studies. In a companion report¹⁴, we show that the PAN complex consists of subcomplex I, which comprises a stable hexamer of the CC-OB domain, and subcomplex II, which is formed by the ATPase domain. We then determined the crystal structures of subcomplex I and the ATPase domain, the latter allowing for structural modeling of subcomplex II¹⁴. The advent of structural information on the M. jannaschii RP provides a powerful tool for systematic investigation on the mechanisms of proteasomal function. To assist these studies, we first characterized the function of the PAN complex.

Recycling of the PAN complex in unfolding reactions

M. jannaschii lives at an ambient temperature of above 80 °C, which is poorly suited to in vitro studies. Previous investigations of the unfolding activity of the PAN complex were carried out at 45 °C^11,15–18. To more closely mimic the growth temperature of M. jannaschii, we modified the unfolding assay for the recombinant PAN complex to be performed at 65 °C, using GFP-ssrA as the substrate protein. The 11-residue ssrA peptide^18–20 is appended to the C-terminus of GFP, and serves as a degradation signal. The selection of GFP as the substrate was based on the observation that GFP is extremely heat-stable, with a melting temperature of 83°C²¹. In addition, the unfolding of GFP is accompanied by disappearance of fluorescence, which can be quantitatively monitored in real time over the 30-minute time course of the assay. In the absence of PAN, GFP exhibited a gradual decrease in fluorescence (Fig. 1a, black dotted line); this loss of fluorescence is reversible and likely reflects minor conformational changes in response to the elevated temperature. At the end point, approximately 66% of the initial fluorescence remained.

Figure 1. Functional characterization of the PAN complex.

(a) The PAN complex fully unfolds and releases the globular protein GFP-ssrA in the absence of the 20S CP. (b) The presence of 20S CP greatly accelerates the unfolding of substrate. (c) The 20S CP does not increase the rate of ATP hydrolysis by the PAN complex during unfolding reactions. (d) Improved substrate unfolding by the 20S CP is negated by inhibition of the protease activity. Curves shown in (b) and (c) are representative of at least three independent experiments with standard diviation (SD) of less than 5%. All values reported in this manuscript are means ± SD of at least three independent experiments.

PAN is highly active at 65°C, capable of unfolding approximately 94% of GFP-ssrA with PAN in 6-fold molar excess (Fig. 1a). This compares favorably to the previously reported 50% unfolding of GFP-ssrA achieved in a 45°C assay¹⁸. We then assayed unfolding with substrate in excess, in order to test whether PAN can be recycled under these conditions by release of the unfolded substrate protein into solution. We used a fixed quantity of GFP-ssrA in the unfolding reaction, while altering the molar ratio of PAN hexamer to GFP. If the PAN hexamer and GFP are incubated in a 1:2 molar ratio, the percentage of folded GFP should not fall below 50%, assuming that the PAN complex is not recycled. Contrary to this scenario, approximately 33% of the folded GFP remained (Fig. 1a, blue line). Similarly, the molar ratio 1:4 of the PAN hexamer over GFP should give rise to at least 75% folded GFP under the non-recycling scenario; yet the measured value was only 42% (Fig. 1a, orange line). This analysis indicates that the PAN complex is recycled in unfolding reactions in the absence of the 20S CP, and that the reaction product, unfolded GFP, can be released from the PAN complex.

Acceleration of unfolding by 20S CP

To investigate the impact of the 20S CP on the unfolding reaction by the PAN complex, we used fixed molar quantities of GFP-ssrA and the PAN complex, altered the molar ratio of 20S CP over PAN, and monitored the time course of fluorescence change (Fig. 1b). The substrate GFP-ssrA used in these reactions was in 20-fold molar excess over the PAN complex. In the absence of the 20S CP, approximately 62% of the folded substrate remained at the end point (Fig. 1b, black line). These data confirm that the unfolded substrate is released from the PAN complex. Increasing amounts of the 20S CP resulted in progressively higher rates of unfolding. For example, an equi-molar amount of 20S CP over the PAN hexamer led to about 18% of the folded GFP at the end point (Fig. 1b, red line). Thus, an equi-molar amount of 20S CP increased the bulk unfolding activity of the PAN complex by approximately 2.2 fold [(100% − 18%)/(100% − 62%)].

One potential explanation for the accelerated unfolding is that the 20S CP increases the rate of ATP hydrolysis by the PAN complex during unfolding. This hypothesis, although attractive, proved to be untrue. Rather than increasing the rate of ATP hydrolysis, the presence of an equi-molar amount of 20S CP actually decreased it by approximately 10% (Fig. 1c). Another explanation for accelerated unfolding by the PAN complex is that substrate degradation by the 20S CP facilitates unfolding and translocation of GFP-ssrA, perhaps through accelerated substrate release from the PAN complex. This hypothesis predicts that, if substrate degradation is blocked, unfolding and translocation will also be negatively affected. Supporting this notion, incubation of increasing concentrations of the protease inhibitor with the reactions led to progressive decrease of the unfolding activity of the PAN complex (Fig. 1d). At 3.25 mM inhibitor concentration, substrate degradation by the 20S was inhibited by 96% (Fig. 1d, inset), and the unfolding activity actually fell below that of the PAN complex alone by about 20%. Thus, jamming the protease activity of the 20S CP negatively affects substrate unfolding and translocation by the PAN complex.

Structural motifs of the PAN complex

Together, our biochemical and functional analyses demonstrate that the PAN complex fully unfolds and releases the globular protein GFP-ssrA in the absence of the 20S CP. The presence of 20S CP greatly accelerates the unfolding of substrate, without increasing the rate of ATP hydrolysis by the PAN complex. This effect is negated by inhibition of the protease activity, suggesting that active 20S CP facilitates substrate release. These results and associated assays lay the foundation for structure-guided mutational analysis.

Elucidation of the mechanisms of RP function requires understanding the biochemical roles of individual amino acids and structural motifs of PAN in the processes of substrate binding, unfolding and translocation. Because surface-exposed amino acids are more likely to play a functional role, and also to avoid perturbation of the overall structure of PAN, we carefully examined the structure and selected 30 exposed residues for functional characterization. These amino acids map to 4 distinct locations of the RP: axial channel of subcomplex I, coiled-coils of subcomplex I, interface between subcomplexes I and II, and axial channel of subcomplex II (Supplementary Fig. 1). These residues were individually mutated to Ala or in some cases other amino acids, and all mutant proteins were purified to homogeneity. In no case did the mutations lead to spurious in vitro degradation of PAN or destabilization of the PAN oligomer into monomers or other chromatographically distinct subassemblies (data not shown). The functional impact of these mutations was evaluated using three biochemical assays: ATPase activity, unfolding, and degradation. Analysis of these results reveals significant insights into the mechanisms of the archaeal and eukaryotic RP.

The axial channel of subcomplex I

The shape of subcomplex I resembles a funnel, with the bottom half of the funnel constituting an axial channel (Fig. 2a). In AAA proteins that unfold substrate for degradation^1,2,4,5, substrate is thought to be translocated through this axial channel. However, the axial channel of subcomplex I is unique to PAN and the proteasome, as it is formed by an OB domain¹⁴. The OB domain has three structural components: a distal ring formed by the L45 loop (Asn138-Gln139-Gln140-Thr141), a proximal ring by the L23 loop (Ser111-Thr-112-Gly113), and hydrophobic inner surface between the two rings (Pro91, Ile93, Pro114, Pro116, Leu137, and Leu142) (Fig. 2a). The proximal ring, with a surface diameter of 13 Å, is considerably narrower than the distal ring. The rigid structure and stable nature of subcomplex I further suggests that unfolding of substrate is likely a result of forced translocation through the narrow axial channel. To examine this hypothesis, we evaluated the functional consequences of mutations that target amino acids lining the axial channel.

Figure 2. Role of the axial channel of subcomplex I in the function of the PAN complex.

(a) A schematic diagram of the structure of subcomplex I, with the axial channel highlighted. (b) Impact of mutations of the axial channel on unfolding and degradation activities of the PAN complex.

Gly113, which allows ample freedom for main chain conformation, is located at the constriction point of the proximal ring (Fig. 2a). Because it has no side chain, the substitution of Gly113 with any amino acid is expected to markedly reduce the diameter of the proximal ring. Strikingly, the mutations G113A and G113W nearly crippled the unfolding and substrate degradation activity of PAN (Fig. 2b). By contrast, mutations in the distal ring had much less pronounced impact (Fig. 2b). These observations demonstrate that the proximal ring plays an essential role in facilitating the unfolding and degradation of substrate proteins. The size and location of the proximal ring support the notion that substitutions resulting in narrowing of the proximal ring can block substrate passage through the axial channel, leading to loss of substrate unfolding and degradation.

Interestingly, mutation of residues in the hydrophobic surface also had appreciable consequence on PAN function. Mutation of Pro114 led to reduction of substrate unfolding and degradation by as much as 70–80%, whereas mutations of other residues also had varying degrees of impact on these activities (Fig. 2b). These observations suggest that the hydrophobic surface may play an important role in facilitating unfolding and translocation of substrate proteins, perhaps through the substrate-recognizing propensity common among OB domains. The location and mutational phenotype of the hydrophobic surface relative to the proximal ring are fully consistent with the suggested path of substrate unfolding and translocation.

The coiled-coils of subcomplex I

The upper half of the funnel-shaped subcomplex I consists of three coiled-coils, which appear as a tripod extending out of the compact OB domain hexamer (Fig. 3a). In striking contrast to the hydrophobic inner surface of the axial channel, the coiled-coils contain a number of polar and charged amino acids, which point into the inside of the funnel. Importantly, this general structural feature appears to be conserved in the Rpt subunits (Fig. 3a), suggesting a conserved function. Surprisingly, mutation of some of these residues to Ala, including R86A, R88A, and R81A/R82A, led to enhanced activity of substrate unfolding and degradation (Fig. 3b). For example, compared to the WT protein, the double mutation R81A/R82A led to a 2.3-fold increase in substrate unfolding and 32.8 percent increase in degradation. The increased activity of unfolding is likely attributable to altered interactions between the coiled-coils and the substrate protein. In one scenario, the positively charged residues of the coiled-coils may transiently bind to the negatively charged ssrA peptide, thus competing with its entry into the nearby translocation channel. In this case, mutations that reduce the charge density of the coiled-coil region would accelerate entry of the ssrA peptide into the channel, thus facilitating unfolding and degradation of GFP-ssrA.

Figure 3. Role of the coiled coils of subcomplex I in the function of the PAN complex.

(a) A schematic diagram of the structure of subcomplex I, with the coiled coils highlighted. (b) Impact of mutations of the coiled-coils on unfolding and degradation activities of the PAN complex. (c) Removal of negative charges in the ssrA tag resulted in accelerated unfolding rates.

The axial channel of subcomplex II

Structure determination of the PAN ATPase domain and its comparison with HslU led to the modeling of subcomplex II¹⁴, which is predicted to contain an axial channel with two rings (Fig. 4a). The distal ring is formed by residues from the highly conserved Ar-Φ loop, Phe244-Ile245-Gly246, and the proximal ring is composed of Thr284-Gly285-Gly286. Studies of HslU, and, subsequently, other ATPase components of ATP-dependent proteases, as well as ClpB, have led to the model that ATP-driven movement of Ar-Φ loop within the substrate translocation channel may provide a power stroke for substrate unfolding. [probably best to ref 1/7/22 here] Supporting this notion, several mutations of the distal ring, F244A, I245A, I245W, and G246A, led to dramatic defects in the unfolding and degradative activities of PAN (Fig. 4b). Of particular note is the observation that mutation of Ile245 to Ala, a smaller amino acid, or to Trp, a bulkier residue, both led to complete loss of function. These data suggest that not only the size but also the specific conformation of the distal ring (Ar-Φ loop) are essential for unfolding and translocation. Thus, the protein-unfolding activity of PAN appears to be comparable to that of ClpA, ClpX, and several other AAA ATPases^6,22.

Figure 4. Role of the axial channel of subcomplex II in the function of the PAN complex.

(a) A schematic diagram of the modeled structure of subcomplex II, with the axial channel highlighted. (b) Impact of mutations of the axial channel on unfolding and degradation activities of the PAN complex. Error bars represent mean +/− SD of at least three independent experiments.

Interestingly, the ATPases of the eukaryotic proteasome (the Rpt proteins) have sequences that are highly similar or identical to those of PAN in the Ar-Φ loop (Fig. 4a). The conserved tyrosine of the Ar-Φ loop was substituted with glycine in each of the six Rpt proteins. We observed increased levels of ubiquitin-protein conjugates in most of these mutants (rpt1, rpt2, rpt4, rpt5, and possibly rpt6), indicating major in vivo proteolytic defects (Fig. 4c). The rpt4-Y255G mutant had the strongest accumulation of ubiquitin-conjugates, and also exhibited strong sensitivity to stress-inducing treatments typical of partial loss of function of the proteasome: sensitivity to high temperature, and sensitivity to the amino acid analog canavanine (Fig. 4d). rpt4-Y255G also showed the more unusual phenotype of sensitivity to low temperature (Fig. 4d). These data indicate that the Ar-Φ loop is functionally conserved between PAN and the eukaryotic proteasome, and more generally that the power stroke of the eukaryotic proteasome is essentially unchanged from that of PAN. It is thus unlikely that the power stroke is mediated by Rpn1 or Rpn2, two additional subunits of the eukaryotic proteasome, which are not ATPases but appear to be positioned axially²³.

Similarly to the distal ring, mutations of the proximal ring (G285W, G286L, G286W) also led to major defects in the unfolding and degradation of substrate protein (Fig. 4b). These defects could be caused by blocked passage of the axial channel of subcomplex II.

The interface between subcomplexes I and II

Limited proteolysis of the PAN complex led to the generation of two subcomplexes¹⁴, which are connected by the flexible sequences 150–158 (Fig. 5a). Structure-based sequence analysis identified Asp153 as an invariant residue among PAN and the Rpt subunits from S. cerevisiae and humans. In addition, Ala156 in PAN is conservatively replaced by a hydrophobic amino acid in all Rpt subunits. These conserved sequence features are likely to be functionally important.

Figure 5. Role of the interface between subcomplexes I and II.

(a) A schematic diagram of the interface between subcomplexes I and II. (b) A schematic diagram of the distal face of subcomplex II. (c) Impact of mutations of the interface residues on unfolding and degradation activities of the PAN complex.

The observation that mutation in either subcomplex I or II can lead to complete loss of unfolding activity suggests a productive coupling between these two subcomplexes. This notion is supported by the altered ATPase activity of the PAN complex as a consequence of mutation in subcomplex I, as discussed below. We hypothesized that the conserved amino acids in the linker sequences may play an important role in coupling subcomplex I with subcomplex II. Supporting this hypothesis, mutation of Asp153 to Ala or Ala156 to Asp completely abolished substrate unfolding and drastically reduced substrate degradation (Fig. 5c). In addition, insertion of a Gly residue between residues 150 and 151 or between residues 157 and 158 also crippled the function of PAN. Together, these observations point to an essential role of the linker sequences in mechanical coupling between subcomplexes I and II.

Coupling between subcomplexes I and II may involve transient interactions. To pinpoint such interactions, we examined subcomplex II and identified 6 surface-exposed amino acids at its distal face: Ser250, Leu251, Lys253, Asp254, Glu261, and Lys262 (Fig. 5b). All 6 residues are located on helix α5 and conserved in the Rpt subunits. Two double mutations, S250A/L250A and K253A/D254A, led to markedly compromised activities of unfolding and degradation (Fig. 5c). These observations suggest that the surface patch in the ATPase domain of PAN, defined by Ser251-Asp254, may play an important role in crosstalk between subcomplexes I and II.

Mutational effects on PAN ATPase activity

In choosing the 34 new mutants examined in this study, we avoided the ATP-hydrolyzing active site of PAN. However, as a positive control for mutational effects on ATPase activity, we substituted Lys271 for Glu, within the conserved Walker B box, which facilitates ATP hydrolysis. This resulted in a reduction of ATP consumption by PAN to ~7% of WT levels (Table 1). In contrast, none of the test substitutions had such drastic effects on ATP hydrolysis rates in either the presence of a PAN substrate such as GFP-ssrA, or its absence (basal state). Thus, the strong defects in substrate unfolding that result from mutations in the proximal ring of the OB domain, the hydrophobic inner surface of the OB domain, the Ar-Φ loop, the pore-2 loop, and the I-II linker can in no case be attributed to defective functioning of the PAN active site. For all of these mutants, ATP hydrolytic defects were either undetectable or modest in comparison to unfolding defects.

Table 1.

Impact of mutations on ATPase activity of PAN

Mutants	ATP hydrolysis			Unfolding	Degradation	Energy cost of
	PAN alone	with sub	with sub/20S			unfolding	degradation
WT	100.0 ± 1.0	100.0 ± 3.6	100.0 ± 2.6	100.0 ± 2.8	99.9 ±7.4	1.0	1.0
E271K	6.9 ± 5.1	4.2 ± 1.2	2.1 ± 4.2	9.2 ± 3.7	10.7 ±3.1	–	–
Coiled-coil
RR81-82AA	80.5 ± 5.7	144.9 ± 5.0	111.5 ± 5.5	232.5 ±5.3	132.8 ±8.0	0.6	0.8
E83A	49.3 ± 4.9	92.6 ± 4.5	92.2 ± 9.1	71.4 ±6.5	97. 9 ± 6.6	1.3	0.9
E83K	57.1 ± 8.5	91.6 ± 8.2	98.3 ± 11.1	89.9 ±8.1	101.3 ±5.2	1.0	1.0
R86A	78.4 ± 2.8	99.8 ± 3.4	104.7 ± 5.4	137.4 ±5.4	131.1 ±7.4	0.7	0.8
R88A	153.8 ± 4.8	186.0 ± 6.0	173.5 ± 5.1	223.2 ±5.4	122.8 ±0.9	0.8	1.4
VP89-90AA	67.4 ±1.9	102.4 ± 4.7	82.6 ± 11.7	132.7 ± 6.0	137.5 ± 6.7	0.8	0.6
OB Domain, rings
N138A	48.9 ± 4.3	127.9 ± 0.5	105.5 ± 1.3	118.0 ± 7.4	107.4 ±4.2	1.1	1.0
QQ139-140AA	39.8 ± 5.4	100.1 ± 4.8	86.3 ± 1.7	66.0 ± 5.5	112.9 ±2.7	1.5	0.8
G113W	75.5 ± 4.3	97.3 ± 1.9	78.9 ± 0.5	6.9 ±7.8	12.1 ±3.1	14.2	6.5
G113A	69.1 ± 1.9	97.8 ± 0.5	133.5 ± 0.7	26.6 ±5.2	13.3 ±1.1	3.7	10.0
T112A	63.1 ± 0.5	94.0 ± 0.3	92.9 ± 1.6	96.2 ±1.0	84.6 ±8.5	1.0	1.1
OB Domain, hydrophobic surface
I93A	104.4 ± 3.4	156.9 ± 2.5	111.4 ± 2.2	139.6 ±8.8	128.3 ±5.7	1.1	0.9
L142A	52.4 ± 1.4	94.9 ± 3.2	40.4 ± 5.3	65.5 ±0.3	102.3 ±6.5	1.4	0.4
F116A	95.7 ± 3.6	122.8 ± 1.7	79.1 ± 0.8	75.0 ±1.5	83.9 ±7.1	1.6	0.9
P114G	55.5 ± 5.9	77.4 ± 2.6	85.1 ± 0.6	36.6 ±3.7	21.6 ±2.7	2.1	3.9
P114A	57.3 ± 3.1	87.7 ± 2.7	78.0 ± 2.1	64.3 ±1.2	76.6 ±2.4	1.4	1.0
Pore-1 and pore-2 loops
F244A	148.3 ± 9.2	121.9 ± 7.1	150.2 ± 7.8	1.4 ±2.9	9.4 ±3.1	85.9	16.1
F244W	85.0 ± 4.9	83.8 ± 10.4	82.8 ± 5.8	125.2 ±6.9	94.5 ±8.3	0.7	0.9
I245A	122.5 ± 4.8	114.0 ± 3.3	173.2 ± 0.6	3.5 ±1.6	3.9 ±2.5	32.8	44.1
I245W	107.9 ± 1.4	119.0 ± 3.0	157.8 ± 0.8	5.6 ±5.2	4.7 ±2.1	21.1	33.5
G246A	43.0 ± 2.0	34.8 ± 2.1	49.7 ± 1.6	5.4 ±0.5	1.5 ±3.7	6.4	33.8
G286A	120.8 ± 1.1	128.2 ± 3.9	140.8 ± 5.7	99.9 ±0.1	105.1 ±0.6	1.3	1.3
G286L	113.2 ± 1.7	97.2 ± 0.4	188.7 ± 2.2	3.0 ±3.4	5.1 ±1.3	32.7	36.8
G286W	121.1 ± 1.1	129.7 ± 2.0	224.8 ± 2.5	5.4 ±3.6	1.1 ±1.2	24.1	210.6
G285W	146.7 ± 3.9	137.7 ± 0.5	186.4 ± 7.7	1.6 ±2.6	7.7 ±2.1	84.0	24.2
I-II linker
D153A	86.7 ± 3.7	114.7 ± 0.6	131.6 ± 1.7	2.3 ±2.5	34.3 ±0.5	50.2	3.8
A156D	58.0 ± 1.2	82.6 ± 2.5	110.2 ± 1.7	1.5 ±1.6	9.2 ±0.2	56.0	12.0
150G	108.8 ± 1.6	138.3 ± 0.8	141.4 ± 5.9	27.1 ±2.6	39.3 ±0.3	5.1	3.6
150G2	107.7 ± 3.5	164.6 ± 0.2	165.7 ± 7.8	34.6 ±5.1	44.1 ±1.2	4.8	3.8
157G	53.7 ± 7.5	72.0 ± 1.0	103.9 ± 3.2	3.8 ±1.1	3.5 ±2.0	19.2	30.1
157G2	73.7 ± 3.9	96.4 ± 3.3	118.4 ± 2.3	5.5 ±2.5	13.2 ±0.6	17.6	9.0
Subcomplex II distal face
SL250-251 AA	150.7 ± 1.6	147.6 ± 2.0	153.3 ± 2.0	4.2 ±3.9	15.1 ±0.5	34.9	10.2
KD253-254AA	146.4 ± 4.9	142.6 ± 1.6	163.0 ± 0.4	25.0 ±1.0	44.4 ±3.9	5.7	3.7
EK261-262AA	162.4 ± 3.1	181.6 ± 0.1	185.4 ± 0.0	84.3 ±0.6	97.2 ±1.2	2.2	1.9

Among the strongest ATPase defects was that shown by the G246A substitution in the Ar-Φ loop. However, most Ar-Φ loop substitutions had only modest effects on ATPase activity, despite near-complete impairment of unfolding activity. The most surprising aspect of the ATP hydrolysis data is the prevalence of effects observed when residues within subcomplex I were mutated. For example, two-fold or greater reductions in the basal ATPase activity of PAN complexes resulted from substitutions in the distal ring of subcomplex I (N138A, and QQ139-140AA). The same trend was observed across the entire OB domain (Table 1).

Under our experimental conditions, the presence of GFP-ssrA stimulated the ATPase activity of WT PAN by about two-fold. For all such substitutions, the extent to which ATP hydrolysis rates were enhanced by the presence of substrate was increased over that of WT, in some cases by over 2-fold. As a consequence, the ATPase activities of the mutants are, in many cases, restored to a near-WT level in the presence of substrate. Interestingly, with the exception of the VP89-90AA double substitution, both trends held true in the neighboring coiled-coil subdomain of subcomplex I as well, despite the fact that the coiled-coil mutants typically promoted rather than inhibited substrate turnover (Table 1).

In summary, a wide variety of minor surface perturbations of subcomplex I appear to directly affect the ATPase activity of subcomplex II in the absence of substrate, suggesting that subcomplexes I and II are intricately coupled to each other. If so, another class of mutants that would be expected to show comparable phenotypes is that of the I-II linker, a putative coupling element. Indeed, substitutions of the two linker residues that are most strongly conserved between PAN and the eukaryotic RP, Asp153 and Ala156, both showed reduced basal ATPase activity for PAN and a correcting influence of substrate (Table 1). In contrast, significant hyper-stimulation of PAN activity was not seen in any of 13 mutant PAN complexes with substitutions in subcomplex II. Instead, the predominant feature of subcomplex II mutants, seen in 9 of 11 tested, was a higher basal ATPase activity.

Functional perspective

This study was guided by structures of the PAN complex at atomic resolution, which are reported in an accompanying paper¹⁴. For the first time, functional and biochemical investigation of the PAN complex and the archaeal proteasome could be designed on the basis of detailed, three-dimensional structural information. Interpretation of the experimental data can now be based on and correlated with concrete physical evidence.

Subcomplex II of PAN, comprising its ATPase domain, is, as anticipated, largely similar in structure to that of other ATP-dependent proteases. In contrast, subcomplex I is unique to PAN and its homolog, the eukaryotic proteasomal ATPase ring. Also of interest, and unique to these ATPases, is the I-II linker that connects the two domains.

The structure of PAN, taken together with extensive previous work on ATP-dependent proteases^1,6, indicates that the path of substrate through the complex is through the center of the ATPase ring (Fig. 6). As substrates complete their movement through this complex, they pass directly to the axial channel of the 20S CP, eventually to reach the proteolytic sites of the proteases and be hydrolyzed². Based on studies on many ATP-dependent proteases, the translocation of substrates through PAN is generally thought to proceed in multiple steps^1,6,22,24. First, an unfolded segment of the substrate enters the channel, probably by diffusion. Second, the leading segment of the substrate binds to a “motor” element within the channel, thought to be the Ar-Φ loop^6,22. Third, the segment is drawn more deeply into the channel through being pulled by ATP-dependent movement of the Ar-Φ loop. After a certain amount of the substrate has been translocated, further translocation will be hindered by the folded structure of the substrate, which prevents its threading through the narrow axial channel.

Figure 6. A working model for the PAN complex.

In this model, folded substrate protein is bound to the distal face of subcomplex I. The degradation tag is then recognized by surface motifs in subcomplex II, likely by the Ar-Φ loop. Cycles of ATP hydrolysis results in a pulling force that exerts on and unfolds the folded substrate protein, which is unable to pass through the narrow axial channel of subcomplex I in its folded state. Degradation of the unfolded polypeptide by the active 20S CP further facilitates substrate translocation.

Based on the structure of PAN, the key barrier to translocation is expected to be the hexameric ring of OB domains¹⁴. However, the distal face of these domains is partly shielded from contact with the substrate by the three coiled-coil structures at the distal face of PAN. The ability of PAN to unfold GFP, a very tightly folded protein, which remains properly folded at temperatures up to 83°C, indicates that PAN is a very powerful unfoldase. Indeed, ATP-dependent proteases such as HslVU and FtsH appear to be unable to accomplish the unfolding of GFP^25,26. The ability to unfold such proteins requires the application of a strong pulling force via conformational changes deep within the translocation channel, apparently at the Ar-Φ loop. However, the force generated by the ATPase domain, and applied to substrate, must also be resisted by a fixed structure in PAN that restricts the translocation of substrate that has not unfolded. In the accompanying study, we have described features of the OB domain hexamer suggesting that it has the structural stability needed to prevent access to the pore by structural domains that remain folded. For example, the oligomerization of the OB domain is stabilized by the burial of ~9000 Ų of surface area.

It has also been proposed that PAN translocates substrates only after an initial, distinct unfolding event occurring on its surface¹⁵. In the biotin-avidin blocking experiments that suggested this model, GFP-ssrA translocation is blocked by avidin binding to biotin, which is in turn conjugated to the sixth residue from the native C-terminus of GFP. GFP fails to fold if the eighth residue from the C-terminus is absent²⁷, or, presumably, perturbed. The 11-carbon linker¹⁵ leading from GFP to biotin tag is, thus, perhaps too long to prevent GFP unfolding, even with avidin bound. Consequently, we favor the model above in which unfolding and translocation by PAN are inherently coupled, with unfolding driven by translocation. Indeed, the results of detailed mutagenesis and biochemical characterization are consistent with the above-described model of PAN function, and also provide new insights into this complex molecular machine.

Of particular interest was the impaired translocation observed when either the proximal ring or inner surface of the OB domain was mutated. Thus, the critical elements in the translocation channel are not limited to the pore loops of the ATPase domains of subcomplex II. Although not found in other ATP-dependent proteases, the OB domain is a common structural motif in proteins, and is typically used as a ligand-binding site, either for proteins, nucleotides, oligosaccharides, or small molecules. Interestingly, the face of the OB domain that mediates ligand interaction in other OB domains is used to line the translocation channel in PAN. The importance of these residues for PAN function is suggested by the biochemical analysis of mutants at this site (Fig. 2). Further studies will be required in order to determine whether transient binding interactions between the OB domain and substrate may underlie these phenotypes.

In contrast to the loss-of-function phenotypes seen when residues within the translocation channel were mutated, substitutions of residues within the coiled-coil structures at the distal surface of PAN often resulted in enhanced PAN function (Fig 3). This could suggest that the coiled-coil elements negatively regulate PAN activity. Although this model cannot be excluded, the positioning of these residues points instead towards interaction with substrate as the basis for these effects. The outer surface of the coiled-coils is distinctly charged and hydrophilic, as opposed to the translocation channel. These features may minimize interactions between hydrophobic segments of the unfolded translocating polypeptide and the outer face of PAN. In the case of GFP-ssrA, the high positive charge density of the coiled-coil surface may also lead to transient interactions that compete with initial threading of the ssrA segment into the channel. One implication of this hypothetical model is that the effect of coiled-coil mutations on PAN function may depend strongly on the test substrate used.

The strong defects in unfolding and protein degradation seen in the Ar-Φ loop mutants indicates that PAN mediates substrate unfolding via this element, in a manner shared with other ATP-dependent proteases. The same conclusion applies to the eukaryotic proteasome, because the Ar-Φ loop is highly conserved between PAN the eukaryotic proteasome, and because mutations in the Ar-Φ loop in yeast show consistent phenotypes.

It will be interesting to determine whether pore-2 shows functionally significant interactions with the N-termini of the CP. In the well-described complex between the eukaryotic PA26 proteasome activator complex and the CP, such interactions have been reported, in which the N-terminal tails of the CP alpha subunits project outwards to engage the interior of the apposed PA26 channel²⁸. However, PA26 is unrelated to the ATPases of the RP. Similar interactions have been also reported between the bacterial ClpP protease and its partner ClpX, in which the pore-2 loop appears to be contacted²⁹. However, ClpX is not closely related to PAN in general or to its pore-2 loop in particular. Whether CP-RP interactions of this type exist for PAN, they are unlikely to account for the functional defects seen in our pore-2 loop mutants, because their defects in GFP-ssrA degradation are paralleled by loss of function in the unfolding assay, in which the CP is not present (Fig. 4). In addition, in the modeled structure of the subcomplex II hexamer, the pore-2 constriction appears to be very narrow, and the insertion of multiple CP-derived N-terminal peptides within this part of the axial channel of PAN would be likely to obstruct substrate translocation.

Among the most intriguing results of this study were the strong phenotypes resulting from substitutions in the I-II linker segment. This linker shows impressive conservation between PAN and the eukaryotic proteasome (Fig. 5), consistent with its functional importance. The structure of this joining segment between subcomplexes I and II is unknown, as it was not visualized in the crystal structures of either subcomplex I or the ATPase domain of PAN¹⁴. However, the positions of the I-II linker endpoints suggest that the linker is peripheral and thus does not form a constriction in the translocation channel. It is probably not an element of the axial channel at all. Accordingly, we and others have found this segment to be protease-sensitive^14,30, indicating a degree of surface exposure. We hypothesize an important role of the I-II linker in coordinating the function of subcomplexes I and II. It is implicit in this proposal that subcomplex I may undergo ATP-dependent conformational changes. The nature of these structural transitions, and their exact role in substrate processing by PAN, remain to be clarified.

Methods

Protein preparation

PAN and subunits of the 20S proteasome of Methanococcus Jannaschii were cloned from the genomic DNA using standard PCR-based protocol. The N-terminal 6 amino acids of the β subunit of 20S proteasome were deleted (βΔ6) to gain constitutive peptidase activity³¹. All mutants of PAN were generated using Quick-change mutagenesis. The identities of all constructs were verified by double-stranded plasmid sequencing. Wild-type (WT) PAN and all mutants were over-expressed in E. coli strain BL21 (DE3) at 18 °C using vector pET-15b and purified using Ni-NTA (Qiagen) column, followed by anion-exchange (Source-15Q, GE Healthcare) and gel-filtration chromatography (Superdex-200, GE) to homogeneity. 20S proteasome was purified as described¹⁴ and activated by heating at 85 °C for 30 minutes. Protein concentration was measured using Bio-rad Bradford dye (Bio-rad) and verified by SDS-PAGE.

Enzyme assays

GFP-ssrA, the N-terminal His-tagged GFP connecting with an C-terminal 11-residues ssrA peptide²⁰, was used as the substrate in all assays. 20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10 mM MgCl₂ and 2 mM ATP, was used as the reaction buffer. To measure ATP hydrolysis, the reaction mixture containing 0.1 μM PAN hexamer, with or without GFP-ssrA and 20S CP, was incubated at 65 °C for 30 minutes. The produced inorganic phosphate was detected by Malachite green assay kit (Cayman Chemical) and the concentration was calculated using a phosphate standard curve.

The unfolding curve of GFP-ssrA was measured dynamically by Hitachi F-2500 fluorescence spectrophotometer at λ_ex 400 nm and λ_em 509 nm, with the water-circulated temperature control at 65 °C. GFP-ssrA was diluted with the reaction buffer to a final concentration of 0.5 or 2 μM, in the presence or absence of PAN or 20S proteasome. To compare the unfolding ability of wild type and mutant PAN, the total fluorescence decrease of GFP-ssrA after incubating at 65 °C for 30 minutes was measured.

For substrate degradation, 2 μM GFP-ssrA was mixed with 0.1 μM PAN hexamer and 0.1 μM 20S proteasome and incubated at 65 °C for 30 minutes. The degradation product was detected by fluorescamine reaction as described previously³². To inhibit the peptidase activity of 20S proteasome, 0–3.25 mM Calpain inhibitor I (BIOMOL) was added into the reaction mixture.

Yeast strains and growth assays

Standard techniques were used for strain construction, transformation and tetrad dissection³³. For plate assays, 5 OD of cells grown in YPD were resuspended in 125 μl of YPD, serial diluted in five-fold increments, and spotted on YPD plates or synthetic arginine drop-out plates supplemented with canavanine at indicated concentration. The wild-type strain was SUB61 (MATα lys2-801 leu2-3, 2-112 ura3-52 his3-Δ200 trp1-1)³⁴. Mutant strains were identical to SUB61 with the exception of the indicated mutation, which was chromosomally integrated together with a 3′ selectable marker on a PCR fragment. The strains and associated markers were as follows: TG154 (rpt1-Y283G/klTRP1); TG160 (rpt2-Y256G/hphNT1); TG166 (rpt3-Y246G/kanMX6); TG176 (rpt4-Y255G/natNT2); TG181(rpt5-Y255G/kanMX6); TG197 (rpt6-Y222G/HIS3MX6). Insertional PCR templates were as described^35,36.

Preparation of whole-cell extracts of yeast for immunoblotting

Cells were grown in YPD to log phase (OD₆₀₀ ~0.8), then harvested (1.2 × 10⁸ cells) and resuspended in 300μl 1× Laemmli loading buffer. Samples were immediately boiled for 10 min. Samples were then analyzed by SDS-PAGE, followed by imunoblotting. The blots were probed with anti-ubiquitin antibody (Biomol, Plymouth Meeting, PA), then stripped and probed with anti eIF5a antibody (kind gift from R. Zitomer).