Dispatch R551

Giant proteases: Beyond the proteasomeTingting Yao and Robert E. Cohen

Proteasomes and related proteases are thought to bethe principal machinery responsible for intracellularprotein degradation. A new class of giant proteases hasbeen discovered that can augment the catabolicfunctions of proteasomes and, under some conditions,may even substitute for proteasomes altogether.

Address: Department of Biochemistry, University of Iowa, BowenScience Building, 51 Newton Road, Iowa City, Iowa 52242-1109, USA.E-mail: robert-cohen@uiowa.edu

Current Biology 1999, 9:R551–R553http://biomednet.com/elecref/09609822009R0551

© Elsevier Science Ltd ISSN 0960-9822

Our understanding of proteolytic mechanisms of cellularregulation has advanced dramatically in the last fewyears. One important outcome has been the realizationthat intracellular proteolysis is accomplished primarily byonly one or a few proteases. These proteases are quiteunusual: all are large, multisubunit complexes in whichthe proteolytic sites are confined to an internal cavity[1,2]. The proteasome, found in all eukaryotes and alsosome archaebacteria, is perhaps the most familiarexample. As the destination of proteins tagged with ubiq-uitin for subsequent degradation, the proteasome hascome to be regarded as the ultimate proteolytic machine.Unexpectedly, however, some new and even bigger con-tenders have now appeared on the scene. Although thefunctions and mechanisms of these giant proteasesremain obscure, a variety of observations suggest thatthey may cooperate with proteasomes, and possibly evenreplace them.

One hint of the existence of such giant, non-proteasomalproteases came from studies reported two years ago byGlas et al. [3]. With the aim of determining the extent towhich the proteasome is essential in eukaryotic cells, theychallenged EL-4 lymphoma cells with a proteasomeinhibitor — N-blocked tri-leucine vinyl sulfone (NLVS),which covalently modifies the proteasome’s catalytic βsubunits. Surprisingly, in the presence of this inhibitor, asmall proportion of the cells survived and recovered toproliferate. The frequency of survival (0.3%) was wellabove what could be expected from mutations.Immunoprecipitation from the adapted cells andsubsequent biochemical analysis demonstrated that theproteasomes were completely assembled, yet modifiedand inactive. These results suggested that the adaptationinvolved up-regulation of one or more proteases that canfunctionally replace the proteasome in cell-cycle controland other critical processes.

Proteasomes have the remarkable ability to be highlyselective and, at the same time, degrade an enormouslydiverse set of substrates. What other proteases could possi-bly substitute for the proteasome? One clue comes fromthe discovery of other proteases with the potential for‘self-compartmentalization’. The active sites in protea-somes are confined to an internal cavity, and this self-com-partmentalizing architecture provides a unique solution tothe problem of substrate specificity. Once a substrateenters the internal chamber, its degradation is ensured byaccess to multiple endoproteolytic sites.

This strategy of limiting access to a proteolytic chamber toprovide selectivity and processivity has, in fact, beenexploited by several bacterial proteases as well as theproteasome [1]. In Escherichia coli, the ClpP and ClpQ/HslVproteases are each composed of two oligomeric rings thatenclose a central cavity for proteolysis. Subunits that aremembers of the AAA family of ATPases [4] associate withthese proteases — ClpA or ClpX with ClpP, and ClpY/HslUwith ClpQ/HslV — and appear to act as chaperones that cansupply unfolded substrates to the proteolytic core of thecomplex. It is likely that similar concerted unfolding anddegradation reactions also occur when the eukaryotic 20Sproteasome associates with its ATPase-containing 19S regu-latory complex to form the full-size 26S particle.

Archaebacteria offer their own examples of self-compart-mentalizing proteases. In fact, the proteasome in Thermo-plasma acidophilum is the prototype of the eukaryotic core20S proteasome. It is, however, another archaebacterialprotease that may offer some insight into the mystery ofthe NLVS-adapted mammalian cells. Several years ago, ina search for regulatory components of the Thermoplasmaproteasome, Baumeister’s group [5] encountered a bigsurprise, the tricorn protease. They found that, whenexpressed in E. coli, the 120 kDa tricorn protease polypep-tide self-assembled to form a hexameric toroid. Electronmicroscopy and three-dimensional image reconstructionshowed that three tricorn protease dimers enclose achannel that traverses the hexamer. The existence of thischannel, which has 2.6 nm openings into a cavity 10 nmacross and up to 4.3 nm high [6], suggests that tricorn pro-tease, like proteasomes, may be self-compartmentalizing.

The tricorn protease and 20S proteasomes are both ATP-independent peptidases. The tricorn protease hexamer hastrypsin-like and very high chymotrypsin-like activities,whereas the archaebacterial 20S proteasome has only chy-motrypsin-like activity. The surprise came when it was dis-covered that tricorn protease can assemble further into an

unprecedented 55 nm icosahedral capsid composed of 20hexamers (Figure 1a) [6]. This 14.6 MDa homooligomerappears to enclose a cavity approximately 37 nm in diame-ter, large enough to accommodate a ribosome. Because thissuperstructure was only observed in Thermoplasma cellextracts, but not with recombinant protein, it is likely thataccessory factors are required for its assembly in vivo.

Despite its peptidase activity and the beauty of its highly-ordered structure, tricorn protease by itself gives us littleindication as to its physiological role. The identification ofaminopeptidases that act synergistically with tricornprotease has provided evidence that tricorn protease mayserve as one component of a complete proteolyticpathway. Three such factors — F1, F2 and F3 — fromT. acidophilum have been described, each of which canrelease amino acids from the unblocked amino termini ofshort peptides [7,8]. Experiments by Tamura et al. [8]offered insight into how these aminopeptidases andtricorn protease may work together. When tricorn pro-tease, aminopeptidase F2 and fluorogenic substrates weremixed in different orders, it was found that release of thefluorophore was enhanced by F2 when the substrate waspreincubated with tricorn protease. Thus, degradation bytricorn protease generates better substrates for theaminopeptidases. With constant amounts of tricornprotease, activation by the aminopeptidase factors issaturable, a further indication of a sequential mechanism.

A more complete sequential scheme for protein degrada-tion (Figure 2) can be envisioned when the proteasome isbrought into play. It is conceivable that the proteasome, orpossibly other self-compartmentalizing proteases, digestsunfolded proteins into oligopeptides, which in turn aresubstrates for tricorn protease. Digestion by tricorn pro-tease then generates shorter peptides, which are furtherreduced to amino acids by the aminopeptidases. Thishypothesis was supported by an analysis of insulin B-chain

degradation [8]. Firstly, the degradation rate was found tobe limited by the amount of proteasome, but not that oftricorn protease or the tricorn protease activators. Secondly,addition of tricorn protease generated a new set of smallerpeptides, whereas addition of the activating factors gener-ated mostly free amino acids. And thirdly, when the pro-teasome, tricorn protease and tricorn protease activatingfactors were combined, large amounts of free amino acidswith little intermediate-sized peptides were observed.

Yet why should tricorn protease be assembled into a super-molecule? Tamura et al. [8] suggested that, in order to effi-ciently channel reaction intermediates, the capsid structureacts as a scaffold that accommodates the aminopeptidaseactivators. But so far there is no biochemical evidence forchanneling, or for direct interactions between either tricornprotease and the factors or the proteasome and tricorn pro-tease. It also is not known whether the tricorn proteasesupermolecules have any advantage over tricorn proteasehexamers in speeding up this catabolic pathway. An alter-native possibility is that the supermolecule has additionalproteolytic activities which arise only in response to a cellstress, such as loss of proteasomal function.

Returning to eukaryotes, a giant protease assembled fromaminopeptidase monomers was discovered that seems tobe a likely substitute for the proteasome in NLVS-adapted EL-4 cells. In the original study by Glas et al. [3],the adapted cells displayed a remarkable increase in achymotrypsin-like activity detected with the fluorogenictripeptide substrate AAF-AMC. This activity elutedearlier than the proteasome upon gel filtration. Impor-tantly, AAF-chloromethylketone, an inhibitor of tricornprotease, blocked proliferation of NLVS-adapted but notnormal cells. Niedermann’s group [9] later discovered thata large form of a protease known as tripeptidyl peptidaseII (TPPII) could account for the AAF-AMC hydrolyzingactivity in the proteasome-inhibited cells.

TPPII is a serine peptidase that removes amino-terminaltripeptides from unblocked oligopeptides. Its substratesand inhibitors are similar to those of tricorn protease, butcompletely different from those of the proteasome. Aparticularly striking observation is that, like tricorn protease,TPPII can assemble into a higher-order structure that hasan internal channel. Each supermolecule is a rod-like stackof eight 6.5 nm-wide segments, with a central channel alongthe long axis that traverses each segment (Figure 1b). Thisgiant TPPII particle might be another example of a self-compartmentalizing protease, though proof of this awaits ahigher-resolution structure. Interestingly, purified TPPIIalso displayed endoproteolytic activity; compared with theproteasome, TPPII cleaved a 41-residue polypeptide fasterand at different sites. Thus, TPPII may make an essentialcontribution to the proteolytic activities that substitute forthose of the proteasome in proteasome-inhibited cells.

R552 Current Biology, Vol 9 No 15

Figure 1

(a) Views of the tricorn protease of Thermoplasma acidophilum,obtained by electron microscopy and three-dimensional reconstructionof the icosahedral capsid. The three images show views down thethree-fold (left), five-fold (middle) and two-fold (right) symmetry axes.The scale bar represents 50 nm. (Adapted from [6].) (b) The TPPIIprotease from murine EL-4 cells. This image is an average from 1365negatively-stained particles visualized by electron microscopy. Therod-like structure is 50 nm long. (Adapted from [9].)

Permission to reproduce thisfigure electronically has beendenied.

While decidedly provocative, the above observations stilldo not resolve the key questions of whether tricornprotease and TPPII are in fact functional homologues, andwhether the TPPII endoprotease activity is regulated byassembly into a superstructure. At least with regard to thissecond point, the recent characterization of a giant proteasefrom fission yeast suggests that the answer might be yes.Osmulski and Gaczynska [10] observed small(approximately 0.8 MDa) and large (approximately 4 MDa)forms of a protease from the fission yeast Schizosaccha-romyces pombe termed ‘multicorn’ — based on an implied,rather than established, relationship to tricorn protease —that has substrate specificity similar to mammalian TPPII.Examination of the large form of the multicorn protease byatomic force microscopy revealed a 50–55 nm diameterround structure, consisting of six or seven particles, ratherthan the eight-segmented, rod-like TPPII superstructure.The large form of the multicorn protease, which may corre-spond to the giant form of TPPII, was found to have pro-tease activity towards unfolded proteins, such asheat-denatured casein; this activity was not observed withthe small form, however. Intriguingly, the relative amountsof the two multicorn complexes in vivo were found tochange with the growth state of the cells. It is tempting tospeculate that fission yeast cells respond to stress condi-tions, such as proteasomal inhibition, by promoting assem-bly of the large multicorn complex.

The recent studies on giant proteases from diverse speciesseem to converge on a common theme: the assembly ofsmall peptidases into large complexes can regulate enzy-matic activities in response to an environmental change.Higher-order complex assembly probably requires acces-sory factors, which may directly or indirectly sense the needfor protease function. To substitute for the proteasome, thelarge protease assemblies must be specific and processive,requirements that in part might be served by the feature ofself-compartmentalization. Moreover, because the protea-some-inhibitor-adapted EL-4 cells did not accumulate

ubiquitinated proteins, it should be possible for at least asubset of the substrates normally degraded via the ubiqui-tin-proteasome pathway to be diverted to the new protease.It would not be surprising if one or more chaperones areup-regulated to fulfill this job. Of course, all these specula-tions await examination. Nonetheless, from the glimpse wehave had so far, there can be little doubt that the discoveryof these giant proteases has opened a new window onprotein catabolism and its role in cell survival.

AcknowledgementsWe thank Wolfgang Baumeister, Tomohiro Tamura, and Maria Gaczynskafor providing images of proteases, and Lois Weisman and Cecile Pickart forhelpful comments. The authors’ work is supported by a University of IowaCenter for Biocatalysis and Bioprocessing fellowship (T.Y.) and NationalInstitutes of Health grant GM37666 (R.E.C.).

References1. Lupas A, Flanagan JM, Tamura T, Baumeister W: Self-

compartmentalizing proteases. Trends Biochem Sci 1997,22:399-404.

2. Baumeister W, Walz J, Zuhl F, Seemuller E: The proteasome:paradigm of a self-compartmentalizing protease. Cell 1998,92:367-380.

3. Glas R, Bogyo M, McMaster JS, Gaczynska M, Ploegh HL: Aproteolytic system that compensates for loss of proteasomefunction. Nature 1998, 392:618-622

4. Patel S, Latterich M: The AAA team: related ATPases with diversefunctions. Trends Cell Biol 1998, 8:65-71.

5. Tamura T, Tamura N, Cejka Z, Hegerl R, Lottspeich F, Baumeister W:Tricorn protease — the core of a modular proteolytic system.Science 1996, 274:1385-1389.

6. Walz J, Tamura T, Tamura N, Grimm R, Baumeister W, Koster AJ:Tricorn protease exists as an icosahedral supermolecule in vivo.Mol Cell 1997, 1:59-65.

7. Tamura T, Tamura N, Lottspeich F, Baumeister W: Tricorn protease(TRI) interacting factor 1 from Thermoplasma acidophilum is aproline iminopeptidase. FEBS Lett 1996, 398:101-105.

8. Tamura N, Lottspeich F, Baumeister W, Tamura T: The role of tricornprotease and its aminopeptidase-interacting factors in cellularprotein degradation. Cell 1998, 95:637-648.

9. Geier E, Pfeifer G, Wilm M, Lucchiari-Hartz M, Baumeister W,Eichmann K, Niedermann G: A giant protease with potential tosubstitute for some functions of the proteasome. Science 1999,283:978-981.

10. Osmulski PA, Gaczynska M: A new large proteolytic complexdistinct from the proteasome is present in the cytosol of fissionyeast. Curr Biol 1998, 8:1023-1026.

Dispatch R553

Figure 2

Hypothetical pathway for proteolysis inThermoplasma. The sequential action of aproteasome complex (or related ATP-dependent protease), the tricorn protease(shown here in its hexameric form) andaminopeptidases may be needed for completedegradation of a polypeptide to free aminoacids. (Adapted from [8].)

6–12-mers

Proteasome+ AAA ATPase

orother ATP-dependent

proteases

Tricorn protease Aminopeptidases

Current Biology

F3

F2F1

2–4-mersFree amino

acids