© 2004 Nature Publishing Group

678 | SEPTEMBER 2004 | VOLUME 5 www.nature.com/reviews/molcellbio

R E S E A R C H H I G H L I G H T S

During early embryonic develop-ment in vertebrates, a round clumpof cells reshapes into an elongatedstructure with the future head at theanterior (A) end and the rump ortail at the opposite, posterior (P)end. This process, which is known asconvergent extension, involves thenarrowing and lengthening of agroup of cells, the chordamesodermcells, parallel to the A–P axis. Areport in Nature now sheds light onhow the cell movements that under-lie convergent extension are alignedwith the A–P axis in Xenopus laevisembryos.

By dissociating chordamesodermcells, mixing them up and lettingthem reaggregate, Ninomiya et al.found that the cells rearranged them-selves according to their original A orP position. In support of this intrinsicchordamesodermal A–P pattern, theX. laevis brachyury (Xbra) andchordin (chd) genes show a typicalreverse expression pattern, with lowA and high P Xbra expression, andthe opposite pattern for chd.

So, what’s the relevance of chor-damesodermal A–P patterning? Toprobe this problem, the authors dis-sociated chordamesoderm cells fromthe A and P regions, reaggregatedthem separately, and then combinedthese aggregates. Combinations ofidentical aggregates remained round,whereas combined A and P aggre-gates became elongated along theA–P axis. A–P polarity thereforeseems to be required for convergentextension.

At certain concentrations, the sig-nalling molecule activin induces chdexpression and reduces that of Xbra.Treating explants in vitro with gradeddoses of activin caused them to elon-gate, whereas the uniform exposureto high or low doses of activin meantthat the explants remained spherical.Expression of chd and Xbra was moreevenly distributed in uniformlyexposed explants when comparedwith the graded explants, whichshowed an Xbra–chd expression gra-dient. So, graded activin signallingcan establish A–P polarity and trigger

Stem-cell research has had more than itsshare of highs and lows. But now, reportingin Nature, Fred Gage and co-workers showthat stem cells that are derived from thenervous system can be converted intoendothelium-like cells in the absence of cellfusion.

Studies indicating that tissue-specificsomatic stem cells can differentiate intoother lineages have caused great excitementamong researchers, as they promise anattractive potential resource for futureregenerative therapies. But, in many cases,these phenotypic changes reflected thedonation of genetic material through theformation of cell hybrids.

In view of these results, Gage and his teamre-evaluated the developmental potential ofstem cells that were derived from adultmouse brain, taking care to exclude fusionevents. The authors harvested neural stemcells (NSCs) from mice that constitutivelyexpressed green fluorescent protein (GFP),isolated single cells, and expanded them inculture to derive purified clonal cell lines.

As NSCs and blood-vessel endothelia areoften anatomical ‘neighbours’, the authorsco-cultured these clones with GFP-negativeprimary human endothelial cells toascertain if the NSCs could be induced todevelop into endothelial cells. After 2–5 daysof co-culture, 6% of the NSCs haddeveloped an endothelial phenotype, asshown by positive staining for both GFP andan endothelium-specific cell-surfaceantigen, CD146. A purified, expandedpopulation of these GFP/CD146-positivecells was then subjected to PCR after reversetranscription of RNA (RT-PCR), and toimmunofluorescence and functional assays(such as capillary-network formation andelectron-microscopic identification ofspecialized secretory vesicles). The results ofthese studies confirmed the acquisition ofendothelium-specific gene expression,protein markers and functional properties.Importantly, these changes wereaccompanied by the decreased expression ofcertain neural-lineage-specific genes.

To exclude the possibility that cell fusionhad occurred, Gage and co-workers adoptedvarious strategies. First, they preparedmetaphase chromosome spreads from NSC-derived endothelial cells. If cell fusion hadoccurred, microscopic examination wouldeasily distinguish human chromosomes —which are metacentric and acrocentric —

from the characteristic telocentric murinechromosomes. In fact, no humanchromosomes were present. Next, theyexcluded initial fusion events by co-culturing NSCs with paraformaldehyde-treated human endothelial cells, which areunable to undergo cell fusion, but canmediate surface-receptor contact. After 5days, 2–4% of these co-cultured GFP-containing NSCs bound to lectin — afeature of mouse endothelium. Similarendothelium-like cells were found insections of embryonic brain after in uterotransplantation of one clonal stem cell line,although in vivo conversion rates were lowerthan those observed in vitro. The authorspropose that these results could reflect anovel mechanism of angiogenesis, in whichcells that line blood vessels are derived fromadjacent nervous tissue.

By showing that ectodermally-derivedNSCs can differentiate into endothelial cells(which originate from mesoderm), thisstudy revives the concept that NSCs have arelatively broad developmental potentialthat enables these versatile cells to adoptalternate germ cell layer affiliations.

Shannon Amoils

References and linksORIGINAL RESEARCH PAPER Wurmser, A. E. et al. Cellfusion-independent differentiation of neural stem cells tothe endothelial lineage. Nature 430, 350–356 (2004)

Separating head from tail

D E V E LO P M E N TA L B I O LO G Y

Altered fates

S T E M C E L L S

© 2004 Nature Publishing Group

R E S E A R C H H I G H L I G H T S

NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 5 | SEPTEMBER 2004 | 679

The selection of substrates for degradation by the26S proteasome is generally thought to lie at thelevel of ubiquitin-chain assembly. But, in Cell,Verma, Deshaies and colleagues now show that theproteins that link polyubiquitylated substrates to theproteasome can add a further layer of substrateselectivity.

The recruitment of polyubiquitylated substratesto the proteasome is essential for ubiquitin-selectivedegradation, but which protein recognizes thepolyubiquitin chains? So far, three different proteins— Rpn10, Rad23 and Rpt5 — have been proposedto be involved in this recognition, and Verma,Deshaies and co-workers addressed the issue of sub-strate recruitment by studying the degradation of apolyubiquitylated yeast protein (UbSic1) in vitro.

They found that, whereas wild-type proteasomesdegraded UbSic1 quickly, rpn10∆ or rad23∆ protea-somes were largely defective in this degradation.These defects could be rescued by the addition of anoptimal amount of recombinant Rpn10 or Rad23 toeither mutant proteasome, although the rescue ofrpn10∆ proteasomes by recombinant Rad23 wasweak. So, what does this mean?

Verma et al. found that the rescue of rpn10∆ pro-teasomes by recombinant Rad23 could be enhancedby the addition of the von Willebrand A (VWA)domain of Rpn10. This Rpn10 construct lacks itsubiquitin-interacting motif (UIM), which highlightstwo points.

First, the ubiquitin-binding domains of Rpn10and Rad23 do not have to function sequentially —in fact, the authors found that Rad23 and the UIMof Rpn10 function redundantly, and in parallel, torecruit UbSic1 to proteasomes. Second, the VWAdomain of Rpn10 seems to facilitate the Rad23-mediated degradation of UbSic1. Furthermore, asthe VWA domain of Rpn10 was not required for theRad23-dependent tethering of UbSic1 to the pro-teasome, the authors propose that it functionsdownstream of Rad23 to enable the productiveengagement of proteasome-bound substrates by thedegradation machinery.

Next, the authors confirmed their results in vivo.Consistent with their proposal that Sic1 can be tar-geted for degradation by Rad23 or the UIM ofRpn10, they showed that Sic1 was significantly stabi-lized in cells that were mutated for both Rad23 andthe UIM of Rpn10 (Sic1 was degraded with normalkinetics in rad23∆ cells). Furthermore, consistentwith the proposed facilitator function of Rpn10, Sic1was more stable in rpn10∆ cells than in cells thatwere only mutated for the UIM of Rpn10 — that is,in the presence of Rad23, Sic1 is more stable in theabsence of the VWA domain of Rpn10.

In the final part of this study, Verma, Deshaiesand co-workers studied the degradation of otherproteasome substrates and found that there was anunexpected degree of specificity in the requirementof substrates for different polyubiquitin-bindingproteins. On the basis of their results, they proposethat Rpn10, Rad23 and Dsk2 (a protein that is simi-lar to Rad23), and possibly Rpt5 and Ufd1–Cdc48 (acomplex that is involved in endoplasmic-reticulum-associated degradation) can function separately torecruit polyubiquitylated substrates to the protea-some. The challenges for the future are to elucidatehow many recruitment pathways exist, to understandhow substrates are targeted to the different pathways,and to determine whether the different pathways aredifferentially regulated.

Rachel SmallridgeReferences and links

ORIGINAL RESEARCH PAPER Verma, R. et al. Multiubiquitin chainreceptors define a layer of substrate selectivity in theubiquitin–proteasome system. Cell 118, 99–110 (2004)FURTHER READING Pickart, C. M. & Cohen, R. E. Proteasomes andtheir kin: proteases in the machine age. Nature Rev. Mol. Cell. Biol. 5,177–187 (2004)WEB SITERay Deshaies’ laboratory: http://www.its.caltech.edu/~rjdlab/

convergent extension. According tothe authors, a different signallingmolecule, Nodal, rather than activin,is likely to be responsible for control-ling A–P patterning in vivo.

In a final set of experiments, theimpairment of the Wnt/planar cellpolarity (PCP) pathway preventedthe elongation of explants, but didnot disturb the graded gene-expres-sion pattern. So, Wnt/PCP signallingseems to control convergent exten-sion independently of A–P polarity,which indicates that the two path-ways probably function in parallel.

Together, these new findings pavethe way for investigating the specificcellular properties that make polar-ized chordamesoderm cells interca-late, and converge and extend alongthe A–P axis, thereby separating headfrom tail.

Arianne HeinrichsReferences and links

ORIGINAL RESEARCH PAPER Ninomiya, H. et al.Antero–posterior tissue polarity links mesodermconvergent extension to axial patterning. Nature430, 364–367 (2004)FURTHER READING Keller, R. Heading awayfrom the rump. Nature 430, 305–306 (2004)

Be more choosy

P R OT E I N D E G R A D AT I O N