news and views

NATURE CELL BIOLOGY VOL 2 OCTOBER 2000 http://cellbio.nature.com E175

Transcriptional regulation: RUPture in the ER

William P. Tansey

Regulated ubiquitin/proteasome-dependent processing controls transcription by triggering the release of dormanttranscriptional activators from the endoplasmic reticulum. This intersection of transcription and proteolysis hasimportant implications for gene control, proteasome–substrate interactions and signal integration.

Ubiquitin-mediated proteolysis seemsto have its finger in just about everybiological pie, from cell-cycle control

to antigen presentation. Recent work byHoppe et al.1 has illustrated yet anotherprocess that can be controlled by this path-way — transcription. By showing thattranscription factors that regulate fatty-acid synthesis are liberated from dormant,ER-bound states by regulated ubiquitin-dependent processing, Hoppe and col-leagues have provided a great example ofhow ubiquitin-mediated proteolysis con-trols gene expression. Importantly, theirwork also affords a rare glimpse of how theproteasome — often thought of as an insa-tiable destruction machine — can cleaveproteins with precision and restraint.

To fully appreciate the significance ofthis story, we must first consider the con-trol of fatty-acid synthesis as a model forgene regulation. Correct ratios of saturatedand unsaturated fatty acids are essential forpreserving the integrity of membraneswithin the cell. For an actively dividingyeast cell, maintaining these ratios is quitea challenge. Fatty-acid synthesis must betied to the growth status of a cell and becapable of responding quickly to changesin membrane composition and fatty-acidpools. Like other regulatory challenges ineukaryotes, fatty-acid synthesis uses tran-scriptional regulation to integrate thesediverse signals and to marshal rapidresponses to changes in the cellular milieu.The study by Hoppe and colleagues pro-vides a framework for understanding howthis happens (Fig. 1).

The central player in this story is OLE1,an enzyme that controls synthesis of theunsaturated palmitoleic and oleic fattyacids. OLE1, its substrates, and its productsreside within the internal membrane net-work of the ER. Importantly for this discus-sion, so too does a precursor of the tran-scription factor SPT23, which is anchored inthe ER by its carboxy-terminal tail (Fig. 1a).When levels of unsaturated fatty acids with-in the ER fall, SPT23 becomes ubiquitinat-ed by a machinery that includes the ubiqui-tin ligase RSP5 (Fig. 1b). What happensnext is quite exceptional. Rather than beingdestroyed by the proteasome, as might beexpected for a ubiquitinated substrate2,

SPT23 is specifically processed in a protea-some-dependent manner to liberate theactive amino-terminal fragment of SPT23(Fig. 1c). After this event, which the authorshave dubbed regulated ubiquitin-depend-ent processing (RUP), the now untetheredSPT23 migrates into the nucleus. There, itinduces the expression of OLE1, which inturn enters the ER and restores appropriatelevels of unsaturated fatty acids (Fig. 1d).Thus, through a combination of physicallinkage of these proteins in the ER and reg-ulation of transcription-factor release byproteolysis, yeast cells are able to sense andrespond appropriately to changes in thecomposition of the ER membrane.

Although the integration of signallingevents at the ER membrane is in itselfintriguing, it seems likely that this discov-ery will receive greater attention because itconsolidates the known involvement of theubiquitin–proteasome pathway in twoaspects of cellular metabolism — regula-tion of transcription and control of pro-tein processing. There have been clearindications of this involvement in the past,but the mechanistic insights provided byHoppe and colleagues are likely to makethis story an important paradigm for bothprocesses.

The idea that proteolysis can regulatetranscription is far from new. In fact, theSPT23 model is very similar to one pro-posed by Wang et al.3 for SREBP, an ER-bound transcription factor that is releasedby proteolysis to control lipid metabolism.An important difference between SREBPand SPT23, however, is that whereasSREBP release depends on the action ofdedicated, site-specific proteases thatreside in the ER membrane, SPT23 is liber-ated by the ubiquitin–proteasome path-way, one of the most widespread and heav-ily used routes of protein destruction.Indeed, it is the very involvement of ubiq-uitin-mediated proteolysis in SPT23release that makes this story so intriguing— if dedicated site-specific proteases canbe used to release transcription factorsfrom the ER, why does SPT23 use ubiqui-tin-mediated proteolysis?

One likely answer to this question is‘coordination.’ Cells frequently use ubiqui-tin-mediated proteolysis to coordinate the

destruction of select groups of proteins —the function of the anaphase-promotingcomplex in mediating the destruction of

SPT23

ER

Cytosol

ER

Cytosol

SPT23

Ub

RSP5

ER

Cytosol

SPT23

Ub

ER

CytosolOLE1

SPT23

a

b

c

d

Nucleus

Nucleus

Ole-1mRNA

Proteasome

Figure 1 RUP links fatty-acid biosynthesis tochanges in ER-membrane composition.Changes in the levels of unsaturated fattyacids in the ER membrane (a), lead toRSP5-mediated ubiquitination of SPT23 (b),followed by its proteasome-mediated cleav-age (c). Free SPT23 migrates into thenucleus to induce expression of OLE1, anER-bound enzyme that controls synthesisof unsaturated fatty acids (d). Ub, ubiquitin.

news and views

proteins at the exit from mitosis is a classicexample of this type of coordination. Byanalogy, it is tempting to speculate thatcleavage of SPT23 through RSP5-mediatedubiquitination serves to link SPT23 releaseto some other function, either of SPT23,RSP5 or another molecule involved in theRUP pathway. Perhaps several transcriptionfactors must be simultaneously releasedfrom the ER to coordinate OLE1 synthesis(there is already a hint of this, as Hoppe andcolleagues have presented genetic evidencethat a second transcription factor, MGA2, isalso controlled by RSP5-mediated process-ing). Alternatively, the coordination mightinvolve events that occur subsequently toSPT23 release, such as the eventual destruc-tion of SPT23 itself. The genetic loop thathas been defined for SPT23 and OLE1would only function if SPT23 were unstable;

otherwise OLE1 synthesis would be consti-tutive. The involvement of a common ubiq-uitin ligase, for example, in both the releaseand subsequent destruction of SPT23would provide an efficient way for the cellto guarantee that the OLE1 regulatory loopis not broken.

A third possibility is that the ubiqui-tin–proteasome pathway does not justcleave SPT23, but is intimately involved inthe activation of OLE1 transcription. Agrowing body of evidence supports a directfunction of the ubiquitin–proteasomepathway in transcriptional activation; tran-scriptional activation domains themselvescan signal ubiquitin-mediated proteolysis4,and proteasome function is required fortranscriptional activation by the oestrogenreceptor5. There is also evidence that ubiq-uitination of the transcription factor Met4by the ubiquitin ligase Met30 can directlyaffect its ability to regulate transcriptionwithout signalling its destruction6. It is con-ceivable that a similar situation existsbetween RSP5, the proteasome and SPT23.Indeed, it is particularly revealing to notethat RSP5 has received quite a bit of atten-tion from those studying transcriptionbecause it is a ubiquitin ligase for RNApolymerase II7, a transcriptional co-activa-tor for the progesterone and glucocorticoidreceptors8 and is related to Tom1, a ubiqui-tin ligase that regulates gene expressionthrough the ADA–SAGA co-activator com-plex9. Perhaps RPT5, after release of SPT23from the ER membrane, migrates withSPT23 into the nucleus where it carries outa co-activator function, mediating interac-tions between SPT23 and components ofthe basal transcriptional machinery. Giventhis intriguing possibility, it will be interest-ing to determine whether ER release is theonly molecular defect that prevents activa-tion of OLE1 in RSP5-defective cells.

Using the proteasome to process SPT23seems a bit like using a chainsaw to prunea bonsai. The proteasome probablydestroys hundreds of different proteins inthe cell, yet it releases SPT23 from the ERwithout a scratch. Although this type ofcleavage is not without precedent — themammalian transcription factor NF-κB issimilarly processed by the proteasomefrom a large precursor molecule10 — littleis known about ubiquitin-dependent pro-cessing, making it difficult to determinewhat regulates the decision betweendestruction and processing. In addition,unlike NF-κB, SPT23 has one end firmlyanchored in a membrane, tightly restrict-ing the ways in which it could access theinterior of the proteasome for cleavage.Given these logistical constraints, and thefact that processing is uncharacteristicbehaviour for the proteasome, how mightSPT23 processing occur?

Two possible mechanisms are illustrat-ed in Fig. 2. The first, proposed by Hoppe

and colleagues, is that part of the SPT23sequence is looped into the interior of theproteasome for cleavage. In this way, theC-terminal tail of SPT23 remains in theER, and the functionally important partsof SPT23 are shielded from the interior ofthe proteasome. Hoppe and colleaguesnote that the interior of the proteasomecould accommodate such a protein loop,and propose that asparagine-richsequences within SPT23 might, by theirlack of complexity, promote loop forma-tion. After cleavage in the loop, the N-ter-minal domain of SPT23 is released fromthe proteasome and migrates to the nucle-us, while the C-terminal domain isextracted from the membrane and degrad-ed. This model is certainly plausible,although how the proteasome cleavesSPT23 in the loop without destroying theentire protein remains unexplained.

A second formal possibility is thatSPT23 is not clipped by the proteasome atall, but rather by a proteasome-associated

protease. Non-proteasomal proteases areknown to associate with the proteasome;a good example is provided by the ubiq-uitin hydrolases that remove ubiquitinfrom substrate proteins. One suchenzyme, DOA4, is essential for NF-κBprocessing in yeast11. Given the structur-al similarities between NF-κB andSPT23, it is a good bet that DOA4 is alsorequired for SPT23 processing. Perhaps,therefore, ubiquitinated SPT23 is pre-sented to the proteasome and is cleavedby one or more ubiquitin hydrolases aspart of the de-ubiquitination process.Cleaved SPT23 would then drift into thenucleus, never having come near to theinterior of the proteasome.

Regardless of the obscurity of themechanisms involved, however, one pointis crystal clear — protein processing bythe ubiquitin–proteasome pathway is hereto stay. The phenomenon has now beenobserved in two diverse eukaryoticspecies, and the similarities between NF-κB and SPT23 processing indicate that aconserved underlying mechanism may bein place. Now that Hoppe and colleagueshave identified the cellular machineryinvolved, it should be only a matter of

NATURE CELL BIOLOGY VOL 2 OCTOBER 2000 http://cellbio.nature.comE176

SPT23Ub?

SPT23Ub

X

UbSPT23

Proteasome

a

b

c

Figure 2 Proteasome-mediated processingof membrane-tethered SPT23.Attachment of SPT23 to the ER mem-brane severely restricts the ways inwhich it can be cleaved by the protea-some (a). One possibility is that SPT23 islooped into the internal cavity of the pro-teasome, where it is endoproteolyticallycleaved (b). A further possibility is that itis not cleaved by the proteasome at all,but by a proteasome-associated protease(c). Ub, ubiquitin.

... Using the proteasome toprocess SPT23 seems a bitlike using a chainsaw toprune a bonsai, yet itreleases SPT23 from the ERwithout a scratch...