Alpha-ring Independent Assembly of the 20S Proteasome

Archaeal proteasomes share many features with their eukaryotic counterparts and serve as important models for assembly. Proteasomes are also found in certain bacterial lineages yet their assembly mechanism is thought to be fundamentally different. Here we investigate α-ring formation using recombinant proteasomes from the archaeon Methanococcus maripaludis. Through an engineered disulfide cross-linking strategy, we demonstrate that double α-rings are structurally analogous to half-proteasomes and can form independently of single α-rings. More importantly, via targeted mutagenesis, we show that single α-rings are not required for the efficient assembly of 20S proteasomes. Our data support updating the currently held “α-ring first” view of assembly, initially proposed in studies of archaeal proteasomes, and present a way to reconcile the seemingly separate bacterial assembly mechanism with the rest of the proteasome realm. We suggest that a common assembly network underpins the absolutely conserved architecture of proteasomes across all domains of life.

Gel-induced higher order species artifacts.
In order to better visualize some lower-abundance species, native gels in this study were frequently heavily loaded with protein (10-20 µg per lane). When this was done, however, one could sometimes observe higher order species migrating above the band of interest. These are identified by white arrowheads in the main figures and they are gel-induced artifacts. Formation of gel-induced artifacts can be illustrated in the nondenaturing 4-15% gradient gel above, stained with GelCode blue. Lane 1 contains the molecular size standards used throughout this study. In lanes 2-4, some of these standards are heavily loaded and run individually. In addition to the correctly running major band, the heavily loaded samples exhibit higher order species to varying degrees (white arrowheads). Residues selected for mutagenesis.
Space filling models based on the T. acidophilum crystal structure (1PMA) and generated with the program Cn3D. (a,b) Highly conserved charged residues (yellow) in one α subunit (red) are juxtaposed with highly conserved residues of opposite charge (teal) located within 5 angstroms on the neighboring α subunit (pink). The K59E mutant can form some DR without apparently forming SR.
Following the fractionation of the α (K59E) mutant by size exclusion chromatography ( Fig. 2b main text), aliquots of fractions 18, 19, 26, and 27 were analyzed by native PAGE on a nondenaturing 4-15% gradient gel shown here stained with Imperial Stain This analysis confirmed the assignment of the major peak in fractions 25-28 as the nonR peak, and the minor peak in fractions 17-19 as the DR peak (see main text). The data are consistent with the α (K59E) mutation interfering with the assembly of SR, yet still allowing some DR to form.

Supplementary Figure S5.
Assembly of archaeal proteasomes exhibits bacterial-like features.
The experiments described in the panels above use the α (K59E) mutant but are otherwise identical to those carried out with the α (R88D) mutant described in the main text (Fig. 5). Indeed, the figure legend to Figure 5 applies here exactly, with K59E substituted for R88D.
The results of this figure support the same conclusions drawn from the main text ( Fig. 5) namely: (i) that archaeal 20S assembly can proceed along a bacterial-like pathway, independent of SR; (ii) that this SR-independent pathway still proceeds through a halfproteasome; (iii) that the SR-independent pathway does not depend on the β subunit propeptide; and (iv) that the same conclusions could be drawn regardless of the approach, lysate mixing versus coexpression, though the latter approach results in more efficient assembly (see main text pertaining to Fig. 5).
As stated in the main text, there are two minor differences between lysate mixing and coexpression approaches. These differences do not affect the aforementioned conclusions but they are worth noting here. proteasome were similar within wild-type α subunit samples ( Fig. 5a and Supplementary  Fig. S5a, lanes 1, 7,8), and within mutant samples ( Fig. 5a and Supplementary Fig. S5a, lanes 2, 3, 4), regardless of the β subunit propeptide variant employed. By contrast, this uniformity was absent during coexpression (see corresponding lanes in Fig. 5d and Supplementary Fig. S5d). This was likely due to slower assembly in lysate mixing where the rate limiting step might be (i) the association of α subunits into SR and/or (ii) the association of α and β subunits into half-proteasomes. Second, in coexpression experiments, the samples employing β (T1A) exhibited sharp decreases in SR and nonR ( Fig. 5d and Supplementary Fig. S5d, lanes 4 and 7). By contrast, SR and nonR levels were constant in lysate mixing regardless of which β subunit variant was being used. This suggests that under coexpression conditions, where assembly is already more efficient than in lysate mixing, the T1A mutant is even more efficient at interacting with free subunits, and/or SR, than a wildtype β subunit. However, the reason for this is unknown to us. Ring independent assembly of archaeal 20S proteasomes.
This is another iteration of the same experiment as described in Figure 6. The difference is that in this version, the isolated pool 1 and pool 2 samples for each of the α-his proteins (a, wild-type and R88D; b, K59E) were split in half. One half was mixed with an equal volume of lysate from E. coli expressing wild-type archaeal β subunits (+) and the other half was not (−). Following incubation to allow assembly to occur, the proteins were repurified by ICAR and equal volumes of each eluate were electrophoresed on a nondenaturing 5-10% gradient gel. Immediately prior to GelCode staining (top panels), the polyacrylamide gel was overlaid with buffer solution containing the fluorogenic peptide substrate Suc-LLVY-AMC to detect peptidase activity (bottom panel). Black arrowheads denote the positions of assembled 20S core particle (20S), half-proteasome (half), double α-ring (DR) and single α-ring (SR). The position of α subunit species that do not assemble into any ring (nonR), and are mostly free α subunits, is shown with a bracket. The migration of several molecular size standards (in kDa) is indicated.

Supplementary Figure S7.
On the severity of the R88D mutation.
The α (K59E) mutant does not appear capable for forming any detectable SR, though it does form some DR, while the α (R88D) mutant appears to form neither SR nor DR (Figs. 2,3). Both mutants exist primarily as nonR (mostly free α subunits) species. Although the effect of the R88D mutation on ring assembly is profound, it is not absolute. When purified α (R88D) protein is heavily overloaded on a nondenaturing 4-15% gradient gel, one can begin to discern a very faint DR band (left panel). Adjusting the brightness and contrast of this image makes this band a bit easier to visualize (right panel). No SR band is ever seen. Hence, the α (R88D) mutation can be thought of as a more extreme version of the α (K59E) mutation i.e. neither forms SR but α (R88D) is much more severe in its effect on DR.
Since the levels of DR in the α (R88D) mutant sample are almost but not quite zero, this can help explain why pool 1 from this mutant also gave rise to barely-perceptible levels of 20S (Fig.  6a, lane 4 and Supplementary Fig. S6a, lane 4). Although present at much lower levels than the α (K59E) DR, the α (R88D) DR is inherently less stable and will dissociate more readily into assembly competent nonR (mostly free α subunits). Because such a large amount of protein is loaded onto the Sephacryl S-300 column (780 µg), there should be enough DR present in pool 1 that, upon concentration and mixing with β-subunit-containing lysates, vanishingly small but detectable levels of 20S species will form. Regardless, these tiny levels of 20S in the pool 1 samples do not alter the main point ( Fig. 6 and Supplementary Fig. S6) that free α subunits (pool 2) can serve as the starting point for efficient CP assembly along a pathway that does not require SR to form. Rapid assembly of 20S proteasomes following subunit mixing.
To visualize additional assembly intermediates, we sought to carry out mixing experiments using separately expressed wild-type, or mutant, α-his and wild-type β-his subunits that were first purified via ICAR. (a) Purified wild-type, or K59E mutant, α-his (10 µg) was mixed with purified wild-type β-his (10 µg) and incubated for increasing amounts of time (0 to 20 min) prior to loading onto a nondenaturing 5-10% gradient gel. The dead time of the experiment was ~8 min. This is the time it takes to withdraw an aliquot of the assembly mixture, add nondenaturing sample buffer, load the sample, turn on the voltage, and have the sample enter the gel. Assembly can continue during the dead time because the sample is always under nondenaturing conditions and it is only when protein enters the gel (and free α and β subunits begin to separate from each other) that assembly is no longer happening. The data suggest that most of the assembly observed occurred during the dead time. Black arrowheads denote the positions of assembled 20S core particle (20S), half-proteasome (half), double α-ring (DR) and single α-ring (SR). The position of α subunit species that do not assemble into any ring (nonR), and are mostly free α subunits, is shown with a bracket. The migration of several molecular size standards (in kDa) is indicated. As stated in the main text, no new assembly intermediates (i.e. αβ heterodimers) were evident in mixing experiments. This could be because αβ heterodimers are transient and very rapidly assemble into half-proteasomes, or because αβ heterodimers are not stable enough to survive electrophoresis, or both. While purified α-his subunits (wild-type or mutant) resulted in distinct species when separated by nondenaturing PAGE, purified β-his subunits produced a slowly migrating smear (a). We attribute this to the high predicted isoelectric point for β-his (pI = 8.39). Since the pH of the nondenaturing gel is 8.8, β-his may not be appreciably negatively charged at this pH to easily enter the gel. In support of this, we fractionated purified β-his by size exclusion chromatography.
(b) ICAR-purified recombinant β-his subunits were subjected to size exclusion chromatography on a Sephacryl S-300 column. Aliquots of every other fraction were analyzed by 12% SDS-PAGE followed by staining with Imperial Stain. Black arrowheads indicate the column void volume and the elution peaks of molecular size standards (in kDa). M, molecular size standards (size in kDa indicated at left). L, aliquot of the sample load. Purified β-his elutes late as a single peak. No β-his was observed near the void volume or in high molecular weight fractions, consistent with purified β-his behaving as a monomer. The slightly later elution of the β-his protein (predicted M r of 24.6 kDa) relative to the molecular size standards could reflect weak affinity of the β-his subunit for the Sephacryl resin. At the near neutral pH of the size exclusion column, the basic β-his protein (predicted pI = 8.39) may weakly bind to the low levels of carboxylates in the Sephacryl matrix thereby slowing its elution.
SR-independent assembly likely proceeds via αβ heterodimer formation.
(a) Proteasome assembly was initiated by mixing equal volumes of lysates from cells separately expressing the indicated α-his and β subunits. After 30 minutes, proteins were purified by ICAR and loaded onto a Sephacryl S-300 size exclusion column and 3 ml fractions were collected. Aliquots (50 µl) of the indicated fractions were analyzed by three 12% SDS-PAGE gels and stained with Imperial stain. Black lines delineate the position of the 20S peak (also containing half-proteasomes and DR) as well as the α subunit nonR peak and free β subunit peak. We note that both wild-type (Supplementary Fig. S8 Fractions  160  232  18  void  670  440  67 45   13 14  16  18  15  17  19  21  23  25  29  31  33  24  20  22  M  27  26  28  30  32  wild-type α-his α (K59E)-his α (K59E)-his Lysates mixed in each panel in (a)

Supplementary Note
On the function of  subunit propeptides.
In Figure 5, we demonstrate that the archaeal  subunit propeptide is dispensable for CP assembly even when archaeal  subunits cannot form rings; assembly occurs with comparable efficiency with or without the propeptide. The bacterial  subunit propeptide is also not required for CP assembly 2,3 hence the dispensable nature of the propeptide is conserved in archaeal and bacterial CP assembly. However, this is not to argue that the purpose of the propeptide is identical in both. A number of differences are worth discussing.
The cis - subunit interface in bacterial proteasomes is much smaller than in eukaryotic/archaeal counterparts 4,5 . This likely contributes to the inability of bacterial  subunits to form stable  rings on their own 2 , though dimers and trimers are apparently possible 6 . Consequently, bacterial  subunit propeptides are large in part to provide this missing contact surface and help stabilize bacterial  rings 5 . Consistent with this, in the absence of the  subunit propeptide, assembly of CP from Rhodococcus erythropolis is very inefficient 2 . Interestingly, assembly of CP from Mycobacterium tuberculosis is not affected by the absence of its corresponding propeptide 3 . This argues that the "stabilization-of--subunitinterface" function, while important, likely cannot be the only role for a bacterial  subunit propeptide in assembly. The visualization of the Mycobacterium propeptides outside (and below) the -ring in a half proteasome 7 , and their ability to negatively impact assembly 3 , are consistent with this view.
Unlike bacterial propeptides, archaeal  subunit propeptides are much shorter. Clearly, the "stabilization-of--subunit-interface" function is not required by archaeal proteasomes since their  subunits can form stable -rings independently. Also, archaeal  subunit propeptides are completely dispensable for assembly [8][9][10] . Our data suggests that assembly is equally efficient in their presence or absence, even when archaeal  subunits cannot form rings. So what is their role? One possible function is to ensure that the active site threonine is only exposed upon completion of assembly where it is safely enclosed within the central cavity. Numerous observations that propeptide processing is coupled to assembly, for proteasomes from all three domains of life, support this view 6,7,[11][12][13][14] . Related to this function is the finding that the active site threonine can also be inactivated by N-acetylation if exposed to the cytosolic milieu 15 ; a propeptide would again serve in a protective role until assembly is nearly complete. A third possible function is that propeptides can allosterically convey assembly status. We demonstrated the ability of PbaA, the putative archaeal ortholog of the Pba1-Pba2 assembly factor that binds to the outer -ring surface, to preferentially bind to propeptide-containing intermediates 1 . This binding became progressively weaker with increased  subunit processing and let us to propose a "safety" function for this assembly factor: by recognizing propeptidecontaining intermediates, assembly factors like PbaA and Pba1-Pba2 could prevent the premature association of activators (such as PAN in archaea, or RP in eukaryotes) to incompletely assembled CP 1 . This safety function was recently confirmed for Pba1-Pba2 in yeast 16 .
All three of these functions are not mutually exclusive, but all of them can be satisfied by a small propeptide like those found in archaea. Since all three of these functions reduce the incidence of undesirable events, this could help explain the non-essential nature of these propeptides, namely: assembly per se is not affected in their absence, only the incidence of (undesirable) side-reactions would increase. Most pertinent to this study, all three of these functions are compatible with an -ring dependent and an -ring independent assembly mechanism, both of which we show to be possible for the archaeal CP. Other functions beyond those suggested here may also exist, and may come to light in future studies.

On the formation of  heterodimers.
In Supplementary Figure S9, and the accompanying main text, we describe experiments aimed at determining if SR-independent assembly of archaeal proteasomes can proceed through the formation of  heterodimers. Lysate mixing was used to initiate proteasome assembly and, after 30 min, an assembly reaction was subjected to ICAR and the purified proteins loaded onto a size exclusion column. Fractions 15-18 contained coeluting  and  subunits and corresponded to assembled proteasomes (and other large species). Fractions 25-30 also contained coeluting  and  subunits and we present arguments in the main text supporting our claim that these fractions could contain the putative  heterodimers. Here, we put forth additional arguments based on our data to support this claim.
First, the overlap between the nonR peak of -his subunits and the presence of  subunits in fractions 25-30 would be expected given that nonR species (i.e. mostly free  subunits) are the immediate precursor to  heterodimers. Second, the only way  subunits should end up in fractions 25-30 (or any fractions for that matter) is via complex formation with -his subunits because free untagged subunits do not bind to the metal-affinity resin (ICAR purification) that immediately precedes the size exclusion separation. Finally, the elution of  subunits in fractions 25-30 is not observed when  subunits are fractionated on a size exclusion column in the absence of  subunits; free  subunits elute in fractions 32-34 (Supplementary Fig. S8). This shift in elution profile (from fractions 32-34 to fractions 25-30) implies a higher molecular mass and thus complex formation with  subunits.
Taken together, our data support the existence of archaeal  heterodimers but they also do not rule out other interpretations (i.e. heterotrimers). However, it is important to note that this lack of absolute certainty is also the case with bacterial  heterodimers. A massspectrometry analysis has provided the first, and to date the only, physical evidence of bacterial  heterodimer formation 6 . However, all the  heterodimer species identified in that study contained truncated  subunits that lacked no fewer than 25 N-terminal amino acids of the 65 amino acid propeptide 6 . The authors themselves allow that an alternate explanation for these complexes is that they are "trapped" species, not capable of assembly 6 .
The formation of  heterodimers is the most plausible explanation for how proteasome assembly can occur in the absence of SR formation. Data in bacteria 2,5,6 and now in archaea (this study) strongly support this view. However, one must also concede that a bona fide  heterodimer (consisting of a full length  subunit bound to a full length  subunit) functioning as an assembly-competent species remains to be unequivocally demonstrated for any proteasome.