found in the yeast Bro1 andmammalian Alix proteins, whichare recruited to a complexinvolved in the sorting of cargointo these inward budding profiles[19] — has a structure thatresembles a ‘boomerang’ witha convex face that containsa highly positively charged regionthat may assist binding to acidicphospholipids [20]. With thecharacterisation of the ability ofthe IM domain to generatenegative membrane curvaturethrough the properties of itsconvex surface, it would certainlybe of interest to establish whetherthe convex surface of the Bro1domain is also able to generateand/or scaffold negativecurvature within the endocyticnetwork.

References1. Zimmerberg, J., and Kozlov, M.M. (2006).

How proteins produce cellular membranecurvature. Nat. Rev. Mol. Cell Biol. 7,9–19.

2. Farsad, K., Ringstad, N., Takei, K.,Floyd, S.R., Rose, K., and De Camilli, P.(2001). Generation of high curvaturemembranes mediated by directendophilin bilayer interactions. J. CellBiol. 155, 193–200.

3. Ford, M.G., Pearse, B.M., Higgins, M.K.,Vallis, Y., Praefcke, G.J., Evans, P.R., andMcMahon, H.T. (2002). Curvature ofclathrin-coated pits driven by epsin.Nature 419, 361–366.

4. Lee, M.C., Orci, L., Hamamoto, S.,Futai, E., Ravazzola, M., andSchekman, R. (2005). Sar1p N-terminalhelix initiates membrane curvature andcompletes the fission of a COPII vesicle.Cell 122, 605–617.

5. Peter, B.J., Kent, H.M., Mills, I.G.,Vallis, Y., Butler, P.J., Evans, P.R., andMcMahon, H.T. (2004). BAR domains as

sensors of membrane curvature: theamphiphysin BAR structure. Science 303,495–499.

6. Gallop, J.L., Jao, C.C., Kent, H.M.,Butler, P.J.G., Evans, P.R., Langen, R.,and McMahon, H.T. (2006). Mechanism ofendophilin N-BAR domain-mediatedmembrane curvature. EMBO J. 25,2898–2910.

7. Itoh, T., and De Camilli, P. (2006). BAR,F-BAR (EFC) and ENTH/ANTH domains inthe regulation of membrane-cytosolinterfaces and membrane curvature.Biochim. Biophys. Acta 1761,897–912.

8. Mattila, P.K., Pykalainen, A.,Saarikangas, J., Paavilainen, V.O.,Vihinen, H., Jokitalo, E., andLappalainen, P. (2007). Missing-in-metastasis and IRSp53 deformPI(4,5)P2-rich membranes by an inverseBAR domain-like mechanism. J. CellBiol. 176, 953–964.

9. Millard, T.H., Bompard, G., Heung, M.Y.,Dafforn, T.R., Scott, D.J., Machesky, L.M.,and Futterer, K. (2005). Structural basisof filopodia formation induced bythe IRSp53/MIM homology domainof human IRSp53. EMBO J. 24,240–250.

10. Lee, S.H., Kerff, F., Chereau, D.,Ferron, F., Klug, A., and Dominguez, R.(2007). Structural basis for theactin-binding function of Missing-in-metastasis. Structure 15, 145–155.

11. Casal, E., Federici, L., Zhang, W.,Fernandez-Recio, J., Priego, E.-M.,Miguel, R.N., DuHadaway, J.B.,Prendergast, G.C., Luisi, B.F., andLaue, E.D. (2006). The crystal structureof the BAR domain from humanBin1/Amphiphysin II and its implicationsfor molecular recognition. Biochemistry45, 12917–12928.

12. Suetsugu, S., Kurisu, S., Oikawa, T.,Yamazaki, D., Oda, A., and Takenawa, T.(2006). Optimization of WAVE2complex-induced actin polymerizationby membrane-bound IRSp53, PIP3,and Rac. J. Cell Biol. 173,571–585.

13. Suetsugu, S., Murayama, K.,Sakamoto, A., Hanawa-Suetsugu, K.,Seto, A., Oikawa, T., Mishima, C.,Shirouzu, M., Takenawa, T., andYokoyama, S. (2006). The RAC bindingdomain/IRSp53-MIM homology domain of

IRSp53 induces RAC-dependentmembrane deformation. J. Biol. Chem.281, 35347–35358.

14. Yamagishi, A., Masuda, M., Ohki, T.,Onishi, H., and Mochizuki, N. (2004). Anovel actin bundling/filopodium-formingdomain conserved in insulin receptortyrosine kinase substrate p53 and missingin metastasis protein. J. Biol. Chem. 279,14929–14936.

15. Dawson, J.C., Legg, J.A., andMachesky, L.M. (2006). BAR domainproteins: a role in tubulation, scission andactin assembly in clathrin-mediatedendocytosis. Trends Cell Biol. 16,493–498.

16. Merrifield, C.J., Qualmann, B.,Kessels, M.M., and Almers, W. (2004).Neural Wiskott Aldrich Syndrome Protein(N-WASP) and the Arp2/3 complex arerecruited to sites of clathrin-mediatedendocytosis in cultured fibroblasts. Eur. J.Cell Biol. 83, 13–18.

17. Habermann, B. (2004). The BAR-domainfamily of protein: a case of bending andbinding? EMBO Rep. 5, 250–255.

18. Disanza, A., Mantoani, S., Hertzog, M.,Ggerboth, S., Frittoli, E., Steffeb, A.,Berhoerster, K., Kreienkamp, H.-J.,Milanesi, F., Di Fiore, P.P., et al. (2006).Regulation of cell shape by Cdc42 ismediated by the synergic actin-bundlingactivity of the Eps8-IRSp53 complex. Nat.Cell Biol. 8, 1337–1347.

19. Slagsvold, T., Pattni, K., Malerod, L., andStenmark, H. (2006). Endosomal andnon-endosomal functions of ESCRTproteins. Trends Cell Biol. 16,317–326.

20. Kim, J., Sitaraman, S., Hierro, A.,Beach, B.M., Odorizzi, G., and Hurley, J.H.(2005). Structural basis for endosomaltargeting by the Bro1 domains. Dev. Cell8, 937–947.

Henry Wellcome Integrated SignallingLaboratories, Department ofBiochemistry, School of MedicalSciences, University of Bristol, BristolBS8 1TD, UK.E-mail: giles.cory@bristol.ac.uk,pete.cullen@bristol.ac.uk

DOI: 10.1016/j.cub.2007.04.015

DispatchR457

Color Vision: Mice See Hue Too

A transgenic mouse has been generated with three cone types, insteadof the normal murine two. Remarkably, some of these mice use the extracone to make trichromatic color discriminations similar to those that arethe basis of human color vision.

Bevil R. Conway

Normal mice have just two conepigments, unlike humans whousually have three. Our extra conecolors the world in a way that micecan only dream of. until now [1].A few years ago, two groups [2,3]independently generateda transgenic mouse that expressesthe human red retinal pigment, in

addition to the native mouse greenand blue pigments, giving themouse a trichromatic retina. Theburning question has been whetherthese mice use the extra pigmentto measure differences in spectraldistribution — do they see inred-green color? Many scientists,including me, would probably haveguessed not, because color visioninvolves all sorts of specialized

neural circuits, both in the retinaand in the cortex [4,5], that wouldseem to require more than a singlegenetic switch to invent. ButJacobs et al. [1] have now shownwith careful psychophysicalexperiments that a fraction of thesetransgenic mice are indeedtrichromatic.

Color vision comes about bya comparison of the relativeactivities of different cone types,a calculation typified by cells withcone-opponent receptive fields[4,5]. In addition to short-wavesensitive cones, mammals typicallyhave one additional cone typesensitive to longer wavelengths.Somewhere around 30–40 million

Current Biology Vol 17 No 12R458

Figure 1. Transgenic mouse performing a red-green color discrimination task.

Photo: Kris Krogh.

years ago, the X-linked gene forthis cone’s pigment was duplicatedin the Old World primates,producing the precursors to ourpresent day red and green pigmentgenes. These genes are 98%identical [6], yielding pigments withpeak sensitivities separated byonly 30 nm. Although this is a smallfraction of the w380–700 nm rangeof the visible spectrum, it issufficient to introduce a newdimension in color, enabling usto see greens, reds andyellows — colors indistinguishableto many color-blind people (thosewho lack the red or green pigmentgene) and to almost all othermammals besides Old Worldprimates.

New World monkeys, such assquirrel monkeys, have color visionthat is intermediate between mostother mammals and Old Worldtrichromats. New World monkeyshave only one X-linked pigmentgene, but because this gene showsallelic variation, females cansometimes carry two differentgenes, one on each of their Xchromosomes. X-inactivationrandomly shuts off one or the otherX chromosome in each cone, so theretinas of female New Worldmonkeys can be trichromatic.Quite remarkably, these specialfemales have red-green colorvision [7]. As with the mice, onewonders how this is possible, ifcolor vision requires elaborateneural circuitry to keep the cone

signals distinct. One popular ideais that the color computationstake advantage of neuralmachinery already present inthe monkeys, machinery thatevolved for a differentpurpose — high-resolution formvision. High acuity vision dependson retinal neurons withcenter-surround antagonisticreceptive fields, called midgetcells. A given midget cell in thefovea derives its high acuity bysampling the activity of just a singlecone cell, and by comparing this tothe activity of surrounding cones.By introducing a new cone type,midget cells could become colorcells, almost by accident,comparing the spectral sensitivityof the pure central cone with theaverage of neighboring cones.

Using knock-in geneticengineering, Jacobs et al. [1] haveeffectively created a mouse witha retina like that of a New Worldmonkey, with one very importantdifference: mice do not havewell-developed form vision, and donot have midget cells. Mice do notalready have the hardware thatmany assume is a critical first stepto the evolution of color vision.Despite this, three of the fivetransgenic heterozygote femalemice tested by the authors coulddiscriminate reds from greens,showing color matches remarkablysimilar to humans (Figure 1).

This result shows how powerfula single genetic mutation can be in

generating a potential behavioraladvantage. But how is color visionmediated in these mice? Bothnormal and transgenic mice arecapable of brightnessdiscriminations. Brightness isa primitive form of color: both colorand brightness are cues tosurfaces; and in primates,brightness and color areprocessed in the same sub-compartment of primary visualcortex, the cytochrome oxidaseblobs. (Cats, incidentally, can makebrightness discriminations andhave cytochrome oxidase blobs,yet lack red-green color vision.)One possibility is that color visionin the transgenic mice piggybackson brightness calculations. But itremains a mystery how this is, orcould be, implemented at theneural level. Do normal mice havededicated brightness-detectingretinal ganglion cells, whichbecome co-opted to handlered-green color? Color (andbrightness) vision is of relativelylow acuity, so the retina wouldnot require many of them.Regardless of the mechanism,the presence of red-green colorvision in an animal that lacksmidget cells resurrects theoff-beat idea that Old Worldprimates do not depend on themidget system for red-greencolor either [8], but rather onsome as yet unidentifiedcoarse-grained non-midgetred-green cell analogous to theprimate blue-yellow retinal cell,which would expose theapparent color-coding ofmidget cells as a red herring.

The receptive fields of mouseretinal ganglion cells are large,having centers that pool responsesof many cones. The cone mosaicof the transgenic mice is patchy,in a pattern that varies fromanimal to animal, just as it doesfrom person to person. Theretinas of color-sighted transgenicmice have not been studied indetail, leaving open thepossibility that the transgenicmice that develop color visionhave cone patches of sizes thatmatch those of ganglion cellreceptive-field centers, producingcolor-responsive retinal neuronsthrough the same mechanismproposed for midget cells.

DispatchR459

Jacobs et al. [1] suggest that it isthe mice with balanced cone ratiosthat develop color. Balanced coneratios presumably sculpt patchsize, although in humans there islittle or no correlation betweencone ratios and color vision ([9], butsee [10]). Moreover, although thepatchiness increases the variabilityin cone ratios that feed differentganglion cells, of the few dozencells tested (in transgenic mice inwhich behavior was not tested),none were cone opponent [2]. Thisis why many believed these micewould not be able to see red-greencolor.

Leaving aside the possibility thatan as yet uncharacterized type ofganglion cell is responsible forred-green color (in mouse and/orprimate), the task of comparingcone signals would seem to betaken up by stages of visualprocessing subsequent to theretina. Primate primary visualcortex contains specializedcone-opponent cells; perhaps thecortex of the transgenic micedevelops them, too, enabled by thesame plasticity that producesvisual receptive fields in auditorycortex when retinal signals areforced to crosswire [11]. Colorreceptive fields in macaque arepush-pull, a complex structure: thecenter of a red-ON cell, forexample, is not only excited by theactivity of red cones andsuppressed by the activity of greencones, but also excited bya decrease in activity of greencones and suppressed bya decrease in activity of red cones[12]. Such a receptive fieldstructure could be constructedfrom midget cells, or hypotheticalnon-midget red-green cells, oreven non-cone-opponent ganglioncells that simply have differentcone ratios, like those found in theretinas of the transgenic mice.

Even if a specific cone ratio isnecessary, it might not be sufficientfor the mice to develop color. Sometransgenic mice might not assignany behavioral relevance to theincoming signals and therefore notdevelop the appropriate neuralarchitecture. As Hubel and Wieselshowed in their ocular dominanceplasticity experiments, the brainwill disregard the input froma perfectly healthy eye if the neural

signals coming from that eye don’tmake sense to the brain (forexample, if the ocular muscles arecut, whacking the eyes out ofalignment). A more subtle examplein which the brain edits incomingsignals is shown by the colorabilities of native speakers oflanguages that do not distinguishcertain colors, like blue and green.Speakers of these languages haveimpoverished color discriminationwithin the relevant part of thespectrum ([13] but see [14]),despite the fact that all humanshave essentially the same geneticsfor color. Yet another example:many women have four differentcone pigments, yet do not seem totake advantage of the additionalcolor information [15]. One couldtest the importance ofdevelopment (and learning) in thecase of the transgenic mice, byraising them in environmentswhere the color cues providesignificant advantages.

The results reported by Jacobset al. [1] demonstrate thetremendous plasticity of the cortexto respond to incoming signals.Surprising at first, such plasticity isprobably the rule rather than theexception. Within some limitations,the brain accommodates novelsignals that are the consequence ofany number of simpleinterventions, so long as theinterventions happen early enoughin development — geneticinterventions, as in this case, orphysical interventions, as in thecase of introducing an extra eye ina frog [16] or forcing sensoryneurons to send projections to thewrong bits of brain [11,17].Synesthesia, brought about bya lack of neural pruning duringdevelopment [18], also points tothe powerful plasticity of the brain,as does the fact that color vision inpeople is largely unaffected byradically skewed cone ratios yetcan be modified by experience [19].These examples show that a basicfeature of brain tissue is its ability tointerpret almost any incomingsignal, raising all sorts of exciting, iffreaky, possibilities. A finaltantalizing possibility: could we usethese mice to uncover other genesimportant in establishing red-greencolor, by looking for differences ingene expression between

color-seeing and non-color-seeingtransgenic mice, or byinterbreeding experimentsfollowed by strong selection ofmice that use the color signals,mimicking the co-evolution offruit-color and primatetrichromacy [20]?

References1. Jacobs, G.H., Williams, G.A., Cahill, H.,

and Nathans, J. (2007). Emergence ofnovel color vision in mice engineered toexpress a human cone photopigment.Science 315, 1723–1725.

2. Smallwood, P.M., Olveczky, B.P.,Williams, G.L., Jacobs, G.H., Reese, B.E.,Meister, M., and Nathans, J. (2003).Genetically engineered mice with anadditional class of cone photoreceptors:implications for the evolution of colorvision. Proc. Natl. Acad. Sci. USA 100,11706–11711.

3. Onishi, A., Hasegawa, J., Imai, H.,Chisaka, O., Ueda, Y., Honda, Y.,Tachibana, M., and Shichida, Y. (2005).Generation of knock-in mice carrying thirdcones with spectral sensitivity differentfrom S and L cones. Zoolog. Sci. 22,1145–1156.

4. Dacey, D.M. (1996). Circuitry for colorcoding in the primate retina. Proc. Natl.Acad. Sci. USA 93, 582–588.

5. Conway, B.R. (2001). Spatial structure ofcone inputs to color cells in alert macaqueprimary visual cortex (V-1). J. Neurosci.21, 2768–2783.

6. Neitz, M., and Neitz, J. (2000). Moleculargenetics of color vision and color visiondefects. Arch. Ophthalmol. 118, 691–700.

7. Jacobs, G.H. (1984). Within-speciesvariations in visual capacity amongsquirrel monkeys (Saimiri sciureus): colorvision. Vision Res. 24, 1267–1277.

8. Hubel, D., and Livingstone, M. (1990).Color puzzles. Cold Spring Harb. Symp.Quant. Biol. 55, 643–649.

9. Brainard, D.H., Roorda, A., Yamauchi, Y.,Calderone, J.B., Metha, A., Neitz, M.,Neitz, J., Williams, D.R., and Jacobs, G.H.(2000). Functional consequences of therelative numbers of L and M cones. J. Opt.Soc. Am. A Opt. Image. Sci. Vis. 17,607–614.

10. Hood, S.M., Mollon, J.D., Purves, L., andJordan, G. (2006). Color discrimination incarriers of color deficiency. Vision Res. 46,2894–2900.

11. Roe, A.W., Pallas, S.L., Kwon, Y.H., andSur, M. (1992). Visual projections routedto the auditory pathway in ferrets:receptive fields of visual neurons inprimary auditory cortex. J. Neurosci. 12,3651–3664.

12. Conway,B.R., andLivingstone,M.S. (2006).Spatial and temporal properties of conesignals in alert macaque primary visualcortex. J. Neurosci. 26, 10826–10846.

13. Davidoff, J., Davies, I., and Roberson, D.(1999). Colour categories in a stone-agetribe. Nature 398, 203–204.

14. Lindsey, D.T., and Brown, A.M. (2002).Color naming and the phototoxic effectsof sunlight on the eye. Psychol. Sci. 13,506–512.

15. Jordan, G., and Mollon, J.D. (1993). Astudy of women heterozygous for colourdeficiencies. Vision Res. 33, 1495–1508.

16. Constantine-Paton, M., and Law, M.I.(1978). Eye-specific termination bands intecta of three-eyed frogs. Science 202,639–641.

17. Kahn, D.M., and Krubitzer, L. (2002).Massive cross-modal cortical plasticityand the emergence of a new cortical area

Current Biology Vol 17 No 12R460

in developmentally blind mammals. Proc.Natl. Acad. Sci. USA 99, 11429–11434.

18. Hubbard, E.M., and Ramachandran, V.S.(2005). Neurocognitive mechanisms ofsynesthesia. Neuron 48, 509–520.

19. Neitz, J., Carroll, J., Yamauchi, Y.,Neitz, M., and Williams, D.R. (2002). Colorperception is mediated by a plastic neural

Mitosis: Springtim

When a eukaryotic cell divides, tensforces pull chromosomes toward opdata show that centromeric chromaforces, revealing a mechanical role

Lawrence C. Myersand Duane A. Compton

DNA replication yields two identicalsister strands, chromatids, whichremain associated throughcohesion until they separate inmitosis and partition to daughtercells. The microtubule-basedmitotic spindle generates force forchromosome segregation. Theaccuracy of chromosomesegregation relies on theattachment of each sisterchromatid to spindle microtubulesfrom opposite poles of thespindle (bi-orientation).Centromere-associated structurescalled kinetochores mechanicallylink spindle microtubules tochromosomes, permitting forcefrom microtubule-dependentmotor proteins — kinesins anddynein — and microtubulepolymer disassembly todisplace chromosomes.Spindles single-mindedlygenerate poleward force onkinetochore-bound microtubulesthroughout all phases of mitosis[1]. That is advantageous inanaphase, where it separateschromatids without equivocation.In metaphase, however, thatsingle-minded behavior causesbi-oriented chromosomes toexperience poleward force towardopposite poles simultaneously.The resulting tug-of-war generatestension on centromeres thatincreases the separationbetween sister kinetochoreson each chromosome.Microtubule-dependent stretchingof sister kinetochores has beenobserved for many years [1];

mechanism that is adjustable in adults.Neuron 35, 783–792.

20. Regan, B.C., Julliot, C., Simmen, B.,Vienot, F., Charles-Dominique, P., andMollon, J.D. (2001). Fruits, foliage andthe evolution of primate colour vision.Phil. Trans. R. Soc. Lond. B 356,229–283.

e for Chromatin

ion builds at centromeres as spindleposite poles during metaphase. Newtin stretches in response to thesefor chromatin packaging in mitosis.

however, the compliant element ofthe chromosome or kinetochorewas not known. Data reportedrecently in Current Biology [2]indicate that centromericchromatin stretches in response tospindle force, suggesting an activerole for chromatin packaging inmitosis.

Centromeres in budding yeastare defined by a unique 125base-pair DNA sequence [3].A nucleosome containing thehistone H3 variant Cse4p(CENP-A in mammals) forms onthis DNA, and works with othercentromere-specific DNA bindingproteins to recruit kinetochorecomponents to create themicrotubule-attachment siteon each chromosome. Thiscentromeric DNA and specializednucleosome are surrounded bya precisely positioned array ofnucleosomes [4]. The strategicplacement of nucleosomessuggests a role for chromatinpackaging in mitosis, and Bouckand Bloom [2] set out to test thatidea by examining mitotic spindlesin cells with reduced histonedensities. They extinguishedhistone H3 or H4 expression in G1phase yeast cells usinga regulatable promoter andexamined cells in the ensuingmitosis. Reducing histone densitydid not inhibit bipolar spindleassembly in most cells andchromosomes established andmaintained bipolar attachments tospindle microtubules. However,both spindle length (pole-to-pole)and the distance between sisterkinetochore clusters increased incells with fewer histones. These

Department of Neurobiology, HarvardMedical School, Boston, Massachusetts02115, USA.E-mail: bconway@hms.harvard.edu

DOI: 10.1016/j.cub.2007.04.017

size increases were not caused byreductions in cohesin recruitment,but appeared to be caused byspindle forces, becauseinactivation of either Cin8p orKip1p kinesin motors led toa significant reduction in bothspindle size and sister kinetochorespacing. Importantly, kinetochoreclusters in histone-depleted cellscontinued to oscillate, indicatingthat spindle and kinetochoredynamics were not adverselyaffected by reductions in histonedensity.

Shortening of sister kinetochoreseparation in the Dcin8 and Dkip1mutant cells suggests that anelastic element in chromatin resiststhese microtubule-based motorswhich provide an outward force.Although an inelastic barrier couldset a maximum distance for sisterkinetochore separation, it wouldnot be expected to provide a forcethat shortens separation upondecreasing the outward force.As a starting point for theinterpretation, chromatin ismodeled as a simple spring thatobeys Hooke’s Law, Fs = –kX,which states that the forceexerted by the spring, Fs, isproportional to the distancestretched, X, and a springconstant k. The distance betweenmetaphase sister kinetochores isproposed to be established whena mechanical equilibrium isreached between outward forcegenerators and inward forcegenerators, such as chromatin.On the basis of this model, onepossibility is that the chromatinbased spring constant decreasesupon histone depletion.A second possibility is thatchromatin rest length — thetotal length of DNA availableto be stretched outwardwithout appreciableresistance — increases uponhistone depletion. Because nosignificant difference in the