humanities & social sciences - osaka university · duction which initiated the transformation...

10

Humanities & Social Sciences

ANNUAL REPORT OF OSAKA UNIVERSITY—Academic Achievement—2003-2004

The East Asian Path of Economic Development: A Long -Term PerspectivePaper in journals: this is the first page of a paper published in The Resurgence of East Asia: 500, 150 and 50 Year Perspectives.[The Resurgence of East Asia: 500, 150 and 50 Year Perspectives] 78-123 (2003)

� Reprinted from The Resurgence of East Asia: 500, 150 and 50 Year Perspectives, Giovanni Arrighi, Takeshi Hamashita and Mark Selden, eds., The East Asian Path of economic develop-ment: a long-term perspective, 78-123 (2003), Copyright 2004, with permission from Mark Selden.

innovation, a scientific revolution and the industrial revolution.

Labour-intensive industrialisationThe second phase was led by British industrialisation, and dur-

ing the nineteenth century, it spread principally to Europe and theregions of recent European settlement, leading to the rise of theAtlantic economy. The standard account highlights the growth ofthe United States economy, which brought Western technology toa new height, exploiting abundant resources and economies ofscale.

In fact, there were two routes of global industrialisation, onerepresented by the American experience which developed capital-intensive and resource-intensive technology, the other represent-ed by the East Asian experience which developed labour-intensiveand resource-saving technology.

The West European variety of industrialisation did not spreadinto the non-European world in its original form, as the man-landratio was very different there, and the straightforward introductionof Western technology proved to be problematic. Thus Japan pur-sued an alternative pattern of industrialisation, with greater labourinputs relative to capital. This I call labour-intensive industrialisa-tion. Beginning in the 1880s, Japan created a wide range of mod-ern Asian industrial goods, and reactivated traditional Asian localinstitutions, which later emerged as modern corporations com-mitted to raising the quality of labour. During the first half of thetwentieth century other East Asian countries followed suit.

However, during this phase East Asia’s labour productivitylagged behind that of the West, which went through a period offurther technological advance (the second industrial revolution),and the gap in per capita income between the West and East Asiaincreased until about 1930. The region’s share in world GDPdecreased significantly.

The fusion of the two pathsAfter 1945 the trend was reversed, and East Asia’s GDP, as well

as per capita GDP, grew faster than that of the West. The growthof Japan’s per capita GDP from 1955 to 1973 was the most con-spicuous example of this new trend. And the “Japanese miracle”turned out to be the beginning of an “East Asian miracle”.

Geopolitical considerations in the early stages of the Cold Warwere crucial to the changes in the American attitude towards Japan’seconomic future. Japan was now expected to use her economicstrength to counter communist penetration in Asia, and was ableto import all the necessary raw materials and resources, includingoil, from the rest of the world. She also began to export manufac-tured goods to advanced Western countries. This allowed Japan topursue the systematic introduction of capital-intensive and resource-intensive industries to an economy with relatively cheap and dis-ciplined labour.

Thus Japan, and later a number of other Asian countries, acquired

'The argument

In the standard literature on the evolution of the modern worldsystem, industrialisation is understood to have emanated from

Western Europe and spread to the rest of the world, and all indus-trialisation is simply taken as a chain of technological diffusion.In this paper, I argue that in fact there were two paths of econom-ic development, the industrial revolution path, which started in WesternEurope, and the industrious revolution path, which developed inEast Asia. From this perspective, global development since 1500consisted of three phases roughly divided by 1820 and 1945. Thispaper seeks to explain how and why East Asia’s share in worldGDP increased in the first phase, decreased in the second, and increasedrapidly in the third.

The industrious revolution path In the first phase, the two paths developed independently of each

other, with broadly similar results in terms of the standard of liv-ing. However, China’s population increased to nearly 400 millionby the end of the eighteenth century. As a result, East Asia’s econ-omy became much larger than Europe’s. This was a world demo-graphic landmark, and its impact on world production far out-weighed that of post-industrial revolution Britain, whose share ofworld GDP in 1820 was less than 6 per cent. There is an impor-tant, relatively unexplored question of how China managed toescape Malthusian checks, and maintain such a vast populationwithout serious deterioration in the standard of living. Essentiallythe same observation can be made with regard to developments inJapan, which, originally under the influence of the China-centredinternational economy, began to show the slow but steady rise inthe standard of living.

The argument of this paper is that it did this because East Asiasuccessfully responded to natural resource constraints, particular-ly the scarcity of land, by developing a set of technological andinstitutional devices for full absorption of family labour. This path,named here the industrious revolution path, sought to mobilise human,rather than non-human, power wherever possible, to sustain andincrease output. The development of labour-intensive technologycentring on rice cultivation enabled small scale production labour-intensive and efficient, while the identification of the peasant fam-ily as a production unit made labour absorption easier and the incen-tive to work greater. It also encouraged the peasant’s will to developmanagerial and inter-personal skills, and anticipate and prevent poten-tial problems. Commercialisation of agriculture and proto-indus-trialisation reinforced this path.

While this contributed to East Asia’s escape from Malthusianchecks in the form of famine, epidemics and war, the region failedto significantly increase labour productivity, which suggests thatit fell into a Malthusian “trap”, often resulting in resource deple-tion. Under such circumstances there was no chance to foster big

11

Osaka University 100 Papers : 10 Selected Papers


The following is a comment on the published paper shown on the preceding page.

The East Asian Path of Economic DevelopmentSUGIHARA Kaoru(Graduate School of Economics)

12 ANNUAL REPORT OF OSAKA UNIVERSITY—Academic Achievement—2003-2004

'Discussion which followed

The Great DivergenceIn his award-winning book, Pomeranz emphasised the crucial

importance of North American resources in accounting for the glob-al dominance of the West since the nineteenth century. The earli-er Smithian growth path, which had been shared by all the coreregions of the world, suffered inherent resource constraints, espe-cially with regard to land. With the sudden and unexpected avail-ability of vast land and other resources in North America, howev-er, Western Europe escaped such constraints. It was a greatdivergence from the general trend.

Pomeranz commented that the assumption, made in my paper(in the form of Osaka University working paper at that time), thatWestern Europe and East Asia were already following differentpaths obscures this sudden divergence (Pomeranz 2000; 16-17).In the revised version of my paper I replied that “while I substan-tially agree with Pomeranz’s view, I wish to retain my emphasison the differences in the man-land ratio between the core regionsof East Asia and those of Western Europe before 1800. The latternever experienced the type of land scarcity seen in eighteenth cen-tury Japan, and it was in Japan that land productivity rose to theextreme and the perception of work was most systematically mould-ed around labour-intensive technology. It is as crucial to formu-late the concept of the industrious revolution on the basis of thetypical East Asian (Japanese) experience as to formulate the con-cept of the industrial revolution on the basis of the typical European(English) experience. It is surely possible to plot both (experi-ences) in the broadly Smithian-Malthusian comparative perspec-tive, without denying the notable divergences in factor endowmentsin Japan and England.”

In his 2001 article, Pomeranz supported my idea that “there isa distinctive East Asian style of development, in large part based

the highest level of Western technology while retaining the EastAsian institutional framework, which permitted a more thoroughexploitation of human resources than had been possible followingthe American path. It was the fusion between Western technolo-gy and East Asian human resource exploitation that produced thevery high rate of economic growth. This fusion turned out to bemuch more powerful than prewar labour-intensive industrialisa-tion, involving deeper clashes and articulations of technology andinstitutions. It represents the third phase of global development.

Implications for global history The emergence of this third phase has had major implications

for global history. First, it suggests the possibility of a move to endworsening global income inequality (see Figures 1 and 2). Undercolonialism the Western path failed to push up the real wage ofnon-white population. By contrast, the possibility of labour-inten-sive industrialisation is now a real one for the majority of devel-oping countries. If the “European miracle” was a miracle of pro-duction which initiated the transformation of the world economy,the “East Asian miracle” has been a miracle of distribution whichbrought the benefits of global industrialisation to the majority ofthe world’s population.

Second, the resurgence of the East Asian path has contributedto the diffusion of industrialisation by retaining and promoting ener-gy-saving technology. In spite of the rising concern about envi-ronmental destruction as a result of the diffusion of industrialisa-tion and the very high level of energy consumption in advancedcountries, few would argue for a complete halt of this process. Theonly way to make global industrialisation possible is a furtherimprovement in energy efficiency on a global scale. In order toallow the miracle of distribution to continue, the Western path mustconverge with the East Asian path, not the other way round.

00

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

1870

1950

0.9 10

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

1950

0.9 1

1990

Figure 1. Global Lorenz Curve (30 countries) in 1870 and 1950.Source : Maddison 1995, 104-206.

Figure 2. Global Lorenz Curve (199 countries) in 1950 and 1990.Source : Maddison 1995, 104-206 and 217-21.

13



on more intensive use of human resources (and on less capital andland-intensity), that, because of that basis, the East Asian pathcould give rise to impressive growth even without equal access tonatural resources, and that this East Asia pattern of growth has par-tially fused with the Western path since 1945”. He suggested thatthis perspective offers an alternative to both the Eurocentric con-vergence theory and the structuralist position held by those whoargued for the persistence of the North-South divide (Pomeranz2001; 324-26).

If my perspective is taken as a holistic alternative to the stan-dard interpretation of global history in this way, there are a num-ber of issues that should be raised and discussed. For example, mydefinition of the industrious revolution path was based on theconsiderations of supply side factors, and it has not been clear whatkind of industrious revolution path existed in Western Europebefore the industrial revolution. At the first GEHN (see below)workshop held in London in September 2003, I suggested that therewere two types of agricultural development behind two paths, andthat in Western Europe where cultivation was combined with pas-ture and the man-land ratio was much lower, animals as capitalplayed such an important role that, while the hard-working spiritof peasants and agricultural workers was certainly present, theydid not result in the industrious revolution of East Asian varietywhere human capital was the sole centre of attention in techno-logical and institutional innovation (see Sugihara 2004a).

Meanwhile, a further connection between the Great Divergenceand East Asia’s industrialisation had been identified at the secondGEHN workshop held in Irvine in January 2004 (see Sugihara 2004b).The rise of the Atlantic economy pushed up the real wage of allmajor Western economies. This gave room for a non-Westerncountry, deprived of mass migration to the new continents, to pro-duce and export competitive manufactured goods by combiningthe low wage with Western technology. But most of the non-Western world were colonised and industrialisation was often dis-couraged. It was Japan, and later China, that took advantage of theGreat Divergence and developed a resource-saving and labour-inten-sive path of industrialisation.

British imperialism and decolonisationAn important corollary to the argument for the persistence of

the East Asian path during the second phase is that the develop-ment of the City of London as the international financial centrefavoured, rather than hindered, East Asia’s industrialisation. In acomment on British Imperialism (originally published in 1993),one of the most influential works on British imperial history inrecent years, I related this argument to the progressive separationof the City’s economic interests from the British empire (Sugihara2002).

In their preface to the second edition of British Imperialism (2001),Cain and Hopkins took up my argument (published in severalplaces at that time) and commented: although “Sugihara suggeststhat Japan’s economic development capitalised on long-standingcommercial ties within Asia that were independent of Western influ-ences, ....he goes on to emphasise that this process was greatly assist-ed by the complementarity that existed between Japan and the City

of London”. “Japan’s ‘national purpose....was to become an interna-tionally competitive industrial power’”, and “the City ‘came to dependon the global diffusion of industrialisation.’ As important centresof manufacturing developed outside the empire, the City’s com-mitment to what might be termed national imperialism steadilyweakened”.

Cain and Hopkins concluded that “how far this interpretationmodifies or extends our own is a matter that is open to discussion.Irrespective of the outcome, however, Sugihara’s analysis illumi-nates the international ramifications of national expansion and sug-gests how the unraveling of empire helped to produce the glob-alised world of today”. The broadening of the perspective in thisway urges us not only to make a connection between labour-inten-sive industrialisation and the changing role of the City during thesecond phase, but to extend it to link the postwar process of decoloni-sation with the beginning of the East Asian miracle.

The formation of GEHNThe original version of this paper was read and discussed in Hong

Kong in 1998 and in Baltimore in 1999 for an American researchproject. Separately from this group, I organised a workshop in Osakain 2001, which was followed up in Buenos Aires in 2002, in whichan extensive discussion on labour-intensive industrialisation tookplace. Meanwhile, a British group of global historians invited meto speak at several workshops and conferences in Europe.

These contacts were partly responsible for the formation ofGEHN (Global Economic History Network) in September 2003.Headed by Professor Patrick O’Brien at the London School ofEconomics and with Professor Kenneth Pomeranz at Universityof California, Irvine, Professor Peer Vries at Leiden University andmyself at Osaka University as organisers of the other three cen-tres, the network plans to hold ten workshops around the world inthree years, involving at its core about 40 economic historians.Under the sponsorship of the Leverhulme Trust research and teach-ing fellowships, and publication at the website and in the form ofworking papers are in progress. With additional assistance fromthe Suntory Foundation, the fifth GEHN workshop will be held inOsaka in December 2004.

References

1. Cain, P. J. and A. G. Hopkins 2001. British Imperialism, 1688-2000, secondedition, Longman, London.

2. Pomeranz, K. 2000. The Great Divergence; China, Europe, and the Makingof the Modern World Economy, Princeton University Press, Princeton.

3. Pomeranz, K. 2001. “Is There an East Asian Development Path?: along-termComparisons, Constraints and Continuities ” Journal of the Economic andSocial History of the Orient, 44-3.

4. Sugihara, K. 2002. “British Imperialism, the City of London and Global Indus-trialization”, in Shigeru Akita ed., Gentlemanly Capitalism, Imperialism andEast Asia, Palgrave, London.

5. Sugihara, K. 2004a. “The State and the Industrious Revolution in TokugawaJapan”, Working Papers of the Global Economic History Network (GEHN),02-04, Department of Economic History, London School of Economics.

6. Sugihara, K. 2004b. “Japanese Imperialism in Global Resource History”,Working Papers of the Global Economic History Network (GEHN), 07-04,Department of Economic History, London School of Economics.

14

Science


Evidence for a Narrow S = +1 Baryon Resonance in Photoproduction from the NeutronPaper in journals: this is the first page of a paper published in Physical Review Letters.[Physical Review Letters] 91, 012002-1-4 (2003)

� Reprinted with permission from T. Nakano, Physical Review Letters, 91, 012002-1-4 (2003). Copyright 2004 by the American Physical Society.

15




Evidence for a Narrow S = +1 Baryon Resonance inPhotoproduction from the NeutronNAKANO Takashi(Research Center for Nuclear Physics)

Introduction

Quarks are ultimate building blocks of sub-atomic particles althoughan isolated quark has never been observed. According to the

fundamental theory of quarks, quantum chromodynamics (QCD),each quark can have three colors; red, blue, or green. And only thecolorless combination (superposition) of quarks can travel freelyas a particle. An example of such a combination is a proton, whichis made of two ‘u’ quarks and one ‘d’ quark. The relatives of a pro-ton which are made of three quarks are called baryons. A particleconsisting of one quark (q) and one anti-quark (q

_) is called a

meson. Note an anti-quark can have the anti-colors so that a paircan be colorless too. Mesons and hadrons together form a qq

_ par-

ticle type called hadrons.

There were no clear experimental evidence for existence of ahadron with a quark configuration rather than three quarks (threeanti-quarks) or a qq

_pair although QCD does not forbid the exis-

tence of other combination such as qq_qq

_ or qqqqq

_. The absence

of the hadron state with more than three quarks was one of theunsolved mysteries in particle physics for decades.

S =1 BaryonBaryon resonances with the strange quark number S=+1 cannot

be made of three quarks since they must contain one anti-strange(s_

) quark. To make the baron number equal to be 1, the minimal

quark configuration is qqqqs_

, where q stands for a ‘u’ quark or a‘d’ quark. In late ‘60’s and ‘70’s, many kaon-nucleon (KN) scat-tering experiments were carried out. These searches resulted in twopossibilities, the isoscalar Z0(1780) and Z0(1865), for which theevidence of the existence was reviewed to be poor by Particle DataGroup (PDG) [1]. In fact, they dropped the summary of the S=+1baryon searches from the PDG listing after 1986 saying that “Thegeneral prejudice against baryons not made of three quarks andthe lack of any experimental activity in this area make it likely thatit will be another 15 years before the issue is decided.”.

In 1997, Diakonov, Petrov and Polyakov studied anti-decupletbaryons using the chiral soliton model [2]. The mass splittings ofthe established octet and decuplet were reproduced within accu-racy of 1 % in this model, and the lightest member of the anti-decuplet with S=+1, which we now call Θ+, was predicted to havea mass of 1530 MeV/c2 and a total width less than 15 MeV/c2. Thevery narrow predicted width motivated us to search for evidenceof the Θ+ at the LEPS facility (Fig. 1).

LEPS facilityThe Spring-8 facility (Fig. 2) is the most powerful third-gener-

ation synchrotron radiation facility in the world. The energy of theelectrons in the storage ring is 8 GeV and the beam current is 100mA. The laser-electron photon (LEP) beam at the SPring-8 is gen-

Fig. 1: Production of the pentaquark particle studied by theLEPS collaboration.

Fig. 2: The SPring-8, the world largest synchro-tron radiation facility with a 8-GeV electron storagering. (Courtesy of SPring-8)


erated by Backward-Compton scattering of laser photons with the8-GeV electrons [3]. The maximum energy of the beam is cur-rently 2.4 GeV, which is above the threshold for ss

_ productions.

The LEP energy is determined by measuring the energy of a recoilelectron. The energy resolution of the tagged photon is 15 MeV.The Laser-Electron Photon facility at SPring-8 (LEPS) is operat-ed by an international collaboration with about 60 members fromJapan, Taiwan, Korea, US, Canada, and Russia. The LEPS col-laboration constructed a forward angle spectrometer (the LEPS detec-tor) in March, 2000, and started experiments in May, 2000 .

Figure 3 shows a schematic drawing of the LEPS detector. Forthe determination of the momentum of a charged particle, track-ing counters were placed before and after a 0.7 T magnet. A time-of-flight (TOF) scintillator array was positioned 3 m behind thedipole magnet to measure the velocity of a charged paricle. A massof a charged particle was reconstructed from the momentum andvelocity information.

A 0.5-cm thick plastic scintillator (SC) located 9.5 cm down-stream from the 5-cm thick liquid-hydrogen (LH2) target ensuredat least one charged particle produced in the LH2 target. For theΘ+ search at LEPS, the events from the SC was used to study eventsgenerated from neutrons in carbon nuclei at the SC.

Fig. 3: The LEPS detector. It analyzes a momentum and velocity of charged particles in forward angles.

Fig. 4: Vertex position for K+K− events along the photon-beam direction. Cut points to selectthe SC events or the LH2 events are indicated by the arrows. (Ref. [4])

LH2

SC

K+K- vtz position (mm)

Eve

nts/

2.0

mm

0

200

400

600

800

1000

1200

-1050-1025-1000-975 -950 -925 -900 -875 -850 -825 -800

17



Evidence for S=+1 baryon resonance at LEPS

Baryon resonances with strangeness quantum number S=+1were searched for in the K− missing mass spectrum in the γn →K+K−n reaction [4]. Since the LH2 target contained no neutron,we selected K+K− pair events produced in the SC, which account-ed for about half of the K+K−-pair events (Fig. 4).

The main physics background events due to the photo-produc-tion of the φ meson. They were eliminated by removing the eventswith the invariant K+K− mass to be consistent with a mass of φ(1019 MeV/c2).

In case of reactions on nucleons in nuclei, the Fermi motion (motionof a proton or neutron inside a nuclei) has to be taken into accountfor the calculation of a missing mass. The correction was made byusing a fact that the undetected particle in the final state is a nuleon,and its validity was checked with the γn→K+Σ−→ K+π−n sequen-tial process, where the K+ and π− were detected as shown in Fig.7.

Figure 8 shows the Fermi motion corrected K− missing massdistribution for the N(γ, K+K−)X reaction, where N stands for aproton or a neutron. A prominent peak at 1.54 GeV/c2 (1540MeV/c2) is found. The broad background centered around 1.6GeV/c2 is most likely due to non-resonant K+K− production andthe background shape in the region above the peak has been fittedby a distribution of events from the LH2. The upper limit for thewidth was determined to be less than 25 MeV/c2 with a 90 % con-

fidence level. The statistical significance of the peak was 4.6 σ.The observed narrow peak strongly indicates the existence of anS = +1 resonance which may be attributed to the exotic 5-quarkbaryon proposed as the Θ+.

Conclusion

We performed a search for an S = +1 baryon in the K− missingmass spectrum of the γn → K+K−n reaction on 12C, using a newlybuilt photon-beam facility at the SPring-8. A sharp baryon reso-nance peak was found at 1540 MeV/c2. The both measured massand width are in good agreement with a prediction by Diaknov etal. Further experimental efforts to confirm the existence of the stateare in progress all over the world.

Fig. 5: The Fermi-motion corrected K+ missing mass spectra for the K+π− events from theSC (solid histogram) and for Monte Carlo events (dotted curve). The dashed histogramshows the missing mass spectrum without the Fermi-motion correction. Λ(1116) andΣ(1197) peaks are separated only after the Fermi-motion correction. (Ref. [4])

Λ Σ−

MMcγK

+ (GeV/c2)

Eve

nts/

(0.0

1 G

eV/c

2 )

0

5

10

15

20

25

30

35

40

45

1.051.075 1.1 1.1251.151.175 1.2 1.2251.251.275 1.3

MMcγK

− (GeV/c2)

Eve

nts/

(0.0

2 G

eV/c

2 )0

5

10

15

1.45 1.5 1.55 1.6 1.65 1.7 1.75 1.8

Fig. 6: The MMcγk- spectrum for the signal sample (solid histogram) and for events

from the LH2 (dotted histogram) normalized by a fit in the region above 1.59 GeV/c2. (Ref. [4])

References

1. Particle Data Group, Phys. Lett. B170, 289 (1986)

2. D. Diakonov, V. Petrov, and M. Polyakov, Z. Phys. A359, 305 (1997)

3. Nakano, T. et al. [LEPS Collaboration], Nucl. Phys. A684, 71c (2001)

4. Nakano, T. et al. [LEPS Collaboration], Phys. Rev. Lett. 91, 012002 (2003) [arXiv:hep-ex/0301020]

18

Engineering


High-Performance Computing Service Over the Internet for Intraoperative Image ProcessingPaper in journals: this is the first page of a paper published in IEEE Transactions on Information Technology in Biomedicine.[IEEE Transactions on Information Technology in Biomedicine] 8, 36-46 (2004)

� F. Ino, Y. Mizutani, et al., High Performance Computing Service Over the Internet for Intraoperative Image Processing, IEEE Transactions on Information Technology in Biomedicine, 8,36-46 (2004). © 2004 IEEE.

19




High-Performance Computing Service Over the Internet forIntraoperative Image ProcessingINO Fumihiko and HAGIHARA Kenichi(Graduate School of Information Science and Technology)

Introduction

With the rapid advances in information technology, three-dimensional (3-D) image processing plays an increasingly

important role in surgery. 3-D image processing is a key techniquefor preoperative surgical planning and intraoperative guidance sys-tems [1, 2] that assist surgeons in developing surgical plans andperforming surgery according to preoperative plans. However,such techniques require a large amount of computation becauserecent X-ray computed tomography (CT) scans produce high-res-olution 3-D images. Therefore, high-performance computing (HPC)approaches are necessary for intraoperative image processing,which requires real-time processing.

Many HPC approaches [3, 4] have realized medical image pro-cessing on various parallel computers, including the cluster sys-tem [5], or a collection of interconnected computers. However, tothe best of our knowledge, work on developing a framework toutilize HPC resources remote from the operating room is lacking.

The paper presents the design and implementation of a frame-work that is suited for high-resolution image processing over theInternet during surgery. Our testbed, named Medical Image Pro-cessing Cluster (MIP-Cluster), aims at realizing the HPC-assist-

ed surgery and sharing this novel procedure with several hospi-tals. In order to achieve this, the testbed tackles the followingissues: a) physical proximity, b) security environment, c) networkbandwidth, and d) network latency between the operating roomand the remote HPC system, and e) scheduling of high-priorityjobs on the shared system. These issues are resolved in key med-ical applications, including a range of motion (ROM) estimatorfor total hip replacement (THR) surgery. The testbed offers apromising solution to deliver HPC services for performing THRsurgery at Osaka University Hospital.

MIP-Cluster: A Medical Image Processing System for Remote Par-allel Processing

Fig. 1 illustrates hardware overview of the MIP-Cluster system,including a client PC in the operating room at Osaka UniversityHospital, located approximately 5 km from the MIP-Cluster. TheInternet connection from the client to the MIP-cluster solves a)physical proximity issue because it provides HPC resources forany hospital at any time, such as with ubiquitous computing. Ubiq-uitous computing is a key factor for preparing emergency surg-eries. However, because the Internet is a public network, this con-

Fig. 1. Overview of MIP-Cluster. The cluster has 128 CPUs in total. Each node in the cluster consists of off-the-shelf components, such as two Pentium III 1-GHz processors, 2GB of reg-istered main memory with error-correction code (ECC), and 40 GB of local disk storage. Local storage is used for clinical data while network file system (NFS) mounted storage is usedfor application development. All the nodes are interconnected through Myrinet and Fast Ethernet switches, providing full-duplex bandwidth of 2 Gb/s and 100 Mb/s, respectively.


nection raises b) the security issue. To address this issue in termsof hardware architecture, the system is divided into two compo-nents: a cluster of node PCs that executes parallel medical appli-cations and a gateway PC that connects the cluster with the Inter-net. The gateway also serves as a network firewall that protectsthe cluster from malicious accesses and provides the cluster’s HPCresources for remote client PCs in operating rooms.

Each node in MIP-Cluster runs on free software: the Linux oper-ating system and the SCore cluster system software [6]. SCore pro-vides a multiuser environment with fault tolerance based on check-pointing. This environment addresses issue e), because it uses agang scheduling mechanism with priority queues, which enablestime-shared scheduling and provides an interactive parallel pro-gramming environment. In an emergency case, surgeons can sub-mit parallel jobs with the highest priority, and these jobs are alwaysexecuted immediately.

On the top of the above software, we have developed a client-server system, providing an automation framework for remote par-allel processing. This framework has the following facilities in orderto address issues b)-d) in terms of software architecture.�Secure execution environment: The framework has to realize an

environment that 1) protects patients’ data transmitted over theInternet and 2) prevents any unauthorized/unauthenticated/uncer-tificated access to the cluster system. Requirement 1) is ethical-ly obligatory and requirement 2) is required to prevent operat-ing hijacking based on spoofing. Our framework addresses theserequirements by public key cryptography.

�High-speed data transmission: The framework has to provide amechanism that efficiently transmits data over the Internet. Thisefficient mechanism is necessary for real-time remote process-ing, especially when a high-bandwidth network with a low laten-cy is unavailable. Our framework satisfies this requirement bycompressing data as well as avoiding retransmission of the pre-viously transmitted data.

�Automatic parallel execution: The framework should enableusers to submit their jobs in the same way as they do in sequen-tial systems. This is necessary to prevent human error originat-ed from the procedure for remote parallel execution. Thus, theprocedure must be simplified in a mission-critical system, sothat our framework hides it from the users by automation.

Parallelizing Range of Motion (ROM) EstimatorROM estimation aims at assisting surgeons in selecting and

aligning the optimal components of changeable artificial joint: thecup, head, neck, and stem components, as illustrated in Fig. 2(a).This assistance is important for both the surgeon and patient,because either inappropriate or improperly positioned componentsincrease the risk of clinical problems such as dislocation, wear,and loosening.

Existing ROM estimators [1, 2] are useful in developing pre-operative surgical plans, which determine the optimal componentsof an artificial joint, its implant position, and orientation. Howev-er, the preoperative plans may need to be altered in view of unpre-dicted conditions revealed during surgery. For example, if the bone

Fig. 2. Components of artificial joint and representation of hip joint motion. (a) Changeable cup, head, neck, and stem components and (b) hip joint motion defined over the components.© 2004 IEEE

(a) (b)

0

10

20

30

40

50

1360

w/o compression w/ compression

Sequential Remote parallel (128 CPUs)

Turn

arou

nd ti

me

(sec

ond) ROM estimation

Parallel process startup

Data transmission

1354

16.3

16.3

4.8 4.8

13.52.3

22minutes

34.6seconds 23.4

seconds

21



tissue around the preoperatively planned position is proven to befragile, the surgeon has to realign the cup components in a stableposition and to reselect the optimal combination of head and neckcomponents for the new position. Therefore, in order to addressthis problem, a fast estimator based on the actual positions and ori-entations of the implanted components is required.

To describe the details of ROM estimation, we first show a briefrepresentation of hip joint motion. Let Mpf denote the transforma-tion from the pelvis coordinate system (pelvis-CS) to the femurcoordinate system (femur-CS). As illustrated in Fig. 2(b), the hipjoint motion Mpf is given by:

Mpf = Tpc Tcs Rcs Tsf,

where Tpc is a 4×4 transformation matrix representing the ori-entation of the cup in the pelvis-CS, Tcs is a fixed transformationmatrix determined by the selected head and neck components, Rcs

is a variable transformation matrix constrained to the rotationalmotion, and Tsf is a transformation matrix representing the reverseorientation of the stem in the femur-CS. Both Tpc and Tsf are deter-mined by one of the following two methodologies. For preopera-tive assistances, the surgeon determines them by visualization andexperience. For intraoperative assistances, optical position sensorsgive them by measuring implanted components during surgery.

Given Tpc, Tcs, and Tsf, the safe ROM is defined as a set of rota-tion transformation matrices, S, such that for all Rcs ∈ S, Rcs avoidsany implant-implant, bone-implant, and bone-bone impingements.

Our parallel implementation employs a master-worker para-digm, where master nodes manage the safe ROM and assign com-putational tasks to the idle workers, as illustrated in Fig. 3. Here,a task corresponds to collision detections for a set of rotation trans-formation matrices. This paradigm provides a dynamic load-bal-ancing mechanism, enabling us to achieve higher performance onheterogeneous systems.

Experimental ResultsWe applied our ROM estimator to a THR surgery performed at

Osaka University Hospital. The estimator is implemented usingthe C++ language and message passing interface (MPI) routines[7]. Fig. 4 shows the turnaround times of ROM estimations whenprocessing in sequential and remote parallel. Although our systemrequired additional times for remote parallel processing, it reducedthe turnaround time from 22 minutes on a single system to 23.4seconds. This turnaround time of less than one half of a minute isspeedy enough for intraoperative processing. It was also evidentthat the data-compression algorithm successfully reduced the turn-around time by approximately 33%.

ConclusionsWe have presented a remote parallel-processing system for med-

ical image processing, the MIP-Cluster system, developed for theHPC-assisted surgery. Our new system provides an intraoperativeROM estimation by parallel processing based on the master-work-er dynamic load balancing. It also provides an automation frame-work for remote parallel processing, which includes a secure exe-cution environment by public key cryptography and high-speeddata transmission by data compression.

As a result, as compared to a sequential system which takes onehalf of an hour for the estimation, MIP-Cluster accelerates the processto less than one half of a minute, and thereby realizes intraopera-tive surgical planning without degrading the resolution of imagesand limiting the area for estimations. We believe sharing the HPC-assisted surgery will improve the quality and safety of surgery ata low cost.

References

1. DiGioia, A.M.; Jaramaz, B.; Blackwell, M.; Simon, D.A.; Morgan, F.;Moody, J.E.; Nikou, C.; Colgan, B.D.; Astion, C.A.; Labarca, R.S.;Kischell, E.; Kanade, T., Clinical Orthopaedics and Related Research,355, 8-22 (1998)

2. Sato, Y.; Sasama, T.; Sugano, N.; Nakahodo, K.; Nishii, T.; Ohzono, K.;Yonenobu, K.; Ochi, T.; Tamura, S., Proceedings of 3rd InternationalConference on Medical Image Computing and Computer-Assisted Intervention,1114-1125 (2000)

3. Warfield, S.K.; Jolesz, F.A.; Kikinis, R., Parallel Computing, 24, 1345-1368(1998)

4. Rohlfing, T.; Maurer, C.R., IEEE Transactions on Information Technolo-gy in Biomedicine, 7, 16-25 (2003)

5. Anderson, T.E.; Culler, D.E.; Patterson, D.A., IEEE Micro, 15, 54-64(1995)

6. PC Cluster Consortium, http://www.pccluster.org/ (2004)

7. MPI Forum, International Journal of Supercomputing Applications, 8,159-416 (1994)

Fig. 3. Master-worker dynamic load balancing for range of motion estimation.

Fig. 4. Turnaround time of sequential and remote parallel processing.

22

Engineering


FAce MOUSe: A Novel Human-Machine Interface for Controlling the Position of a LaparoscopePaper in journals: this is the first page of a paper published in IEEE Transactions on Robotics and Automation.[IEEE Transactions on Robotics and Automation] 19, 825-841 (2003)

� Reprinted from IEEE Transactions on Robotics and Automation, 19, 825-841(2003), Atushi Nishikawa et al., FAce MOUSe: A Novel Human-Machine Interface for Controlling the Position ofa Laparoscope, with permission from IEEE Transactions on Robotics and Automation. Copyright 2003 IEEE.

23




FAce MOUSe: A Novel Human-Machine Interface for Control-ling the Position of a LaparoscopeNISHIKAWA Atsushi(Graduate School of Engineering Science)

Introduction

In current laparoscopic surgery, the vision of the operating sur-geon usually depends on the camera assistant responsible for

guiding the laparoscope. The assistant holds the laparoscope forthe surgeon and positions the scope according to the surgeon’s instruc-tions. This method of operation is frustrating and inefficient forthe surgeon, because commands are often interpreted and execut-ed erroneously by the assistant. Also, the views may be subopti-mal and unstable because the scope is sometimes aimed incorrectlyand vibrates due to the assistant’s hand tremors (see Fig. 1). Theintroduction of robotic technologies, specifically, the developmentof robotic laparoscope positioning systems to replace the humanassistant, is a major step toward the solution of this problem. Insuch robotic systems, the human-machine (surgeon-robot) inter-face is of paramount importance because it is the means by whichthe surgeon communicates with and controls the robotic cameraassistant.

Several robotic laparoscope positioning systems have beendevised in the last ten years. The human-machine interface of mostlaparoscope positioners proposed in the early years included a joy-stick or foot pedal, and each required the use of the surgeon’s handand/or foot [1-3]. These types of interfaces, however, seem gen-erally difficult to use because surgeons already use their hands and/orfeet to control a variety of surgical tools. To solve this problem,several researchers have introduced a voice-activated system basedon the verbal aspect of human speech [4-6]. However, this systemdoes have some inherent limitations, such as reduced accuracy in

positioning, long reaction times, and erratic movements in a noisyenvironment. We believe that a motion-based laparoscope controller,using the movement of the surgeon’s head, is the best solution becausenonverbal instructions such as face gestures are more intuitive andfaster than verbal instructions. Also, because these gestures havethe potential ability to represent not only the direction of scopemotion but also the degree of motion, such as velocity, laparoscopepositioning accuracy may be improved. Several laparoscope manip-ulators with a head navigation interface have previously been devel-oped [7-9]. Such systems, however, failed to fully utilize the non-verbal features of facial motion. These systems were limited todetecting dominant head gestures, which only served as discrete(verbal) commands, and required not only head movements butalso simultaneous control of an additional footswitch. Furthermore,the surgeon had to wear head-mounted sensing devices, such as aheadband and gyro sensor, which were stressful for the surgeon.

To make the most of the advantages of nonverbal and noncon-tact instructions, we have designed a novel human-machine inter-face, called “FAce MOUSe,” for controlling the laparoscope posi-tioner. This proposed human interface is an image-based systemwhich tracks the surgeon’s facial motion robustly in real time anddoes not require the use of body-contact sensing devices. Usingthe FAce MOUSe interface, we have developed a new robotic laparo-scope positioning system for solo surgery. Our system providesthe surgeon with a means of total hands-off and feet-off laparo-scope operations, while also achieving rapid reaction and highpositioning accuracy.

Figure 1. A scene of laparoscopic surgery. The operating surgeon manipulates a variety of surgical instruments using both hands and one foot, while watching the TV monitor screen dis-playing the image of the patient’s body from the laparoscope held and controlled by the camera assistant. The inconveniences of laparoscopic surgery lie mainly in difficulties with mutu-al understanding between the surgeon and the camera assistant.


the zoom-out and zoom-in movements,respectively, of the laparoscope.

Step 3) Make a Trigger Action to Changethe SHIFT/ZOOM State to the STILL State:All the surgeon has to do is roll the face.As soon as the rolling motion is detected,the laparoscope motion stops and the statereturns to STILL. Note that this action is veryeasy to do.

This control scheme is analogous to thatof a computer mouse device (with two but-tons). Accordingly, this is why we calledour scheme FAce MOUSe. Guiding Action,in which the 2-D translation in the faceplane is dominant, corresponds to the mousebody movement on the mouse pad plane.Trigger Action, in which the 1-D rotationin the face plane is dominant, correspondsto mouse button operations such as click(i.e., press and release the button).

To maintain high levels of “safety” dur-ing surgery, we must note the following twopoints. (1) Without exception, unintentional

movements, which could be misunderstood as Trigger Action (toSHIFT) or Trigger Action (to ZOOM) in the STILL state, shouldbe avoided, and (2) Trigger Action (to STILL) should be definite-ly and immediately recognized in the SHIFT/ZOOM state. We per-formed a successful FAce MOUSe implementation by paying greatattention to these points.

Implementation and ResultsWe designed a novel human-machine interface for controlling

the laparoscope with the above method. The system overview isshown in Fig. 3. Our laparoscope positioning system, the FAceMOUSe system, consists primarily of a CCD camera placed justover the TV monitor, an all-purpose PC with a video-capturingdevice, a robot manipulator that holds the laparoscope, and a scanconverter for superimposing graphics on the scope image. Oncethe camera is set in the appropriate position and pose, the core sys-tem in the PC can detect and track the surgeon’s facial features in

Concept of FAce MOUSeFig. 2 illustrates the FAce MOUSe control scheme. As shown

in this figure, the state of the robotic laparoscope positioning sys-tem can be broken down into the following three states: (1) SHIFTstate: guiding the laparoscope for maintaining the surgical pointof interest in the center of the video frame, (2) ZOOM state: guid-ing the laparoscope for providing the required target magnifica-tion, (3) STILL state: keeping the laparoscope still. All surgery worktime can be classified into any of these states from the viewpointof laparoscope operation.

Let us consider how to control not only the laparoscope motionitself but also the transition between these states by making facegestures only. We refer to the face motion for the transition stateas the Trigger Action, and that for guiding the laparoscope as theGuiding Action (see Fig. 2). The method for positioning the laparo-scope through face motions is summarized as follows.

Step 1) Make a Trigger Action to Change the STILL State to theSHIFT/ZOOM State: To complete the transition, the followingthree consecutive face motions are required: (1) put the positionand pose of the face in the standard position and pose; (2) roll theface counterclockwise (for SHIFT) or clockwise (for ZOOM); and(3) return the face “precisely” to the standard position and pose.Note that the surgeon cannot make this action unconsciously.

Step 2) Make a Guiding Action in the SHIFT/ZOOM State: Oncethe system comes into the SHIFT or ZOOM state, the face trans-lation is represented as a vector from the standard position, and thedirection and magnitude of the vector are, respectively, transformedinto the direction and velocity of the laparoscope motion. As shownin Fig. 2, the face intuitively shifts parallel to the scope image plane,corresponding to the identical pan and tilt movements of the laparo-scope when the state is SHIFT. On the other hand, when the stateis ZOOM, the up and down movements of the face correspond to

Figure 2. FAce MOUSe control scheme. The state of the robotic laparoscope positioning system can be broken downinto STILL, SHIFT, and ZOOM states. Using the proposed scheme, the operating surgeon can control not only thelaparoscope motion itself but also the transition between these states by making face gestures only. The face motionfor state transition is referred to as the Trigger Action, and that for guiding the laparoscope as the Guiding Action.

Figure 3. FAce MOUSe system overview. To evaluate the performance of the proposed sys-tem and its applicability in clinical use, we set up an in vivo experiment, in which the sur-geon used the system to perform a laparoscopic cholecystectomy on a pig.

25



real time (30 Hz) from a sequence of video images captured throughthe CCD camera. Assuming that the surgeon’s face is moving ona virtual plane parallel to the TV monitor screen, the system esti-mates the position and pose of the surgeon’s face in real time fromthe image-processing result and then recognizes the surgeon’s facegestures (i.e., the Trigger and Guiding Actions). According to thestate of the system and the gestural action recognition result, thecontrol command is sent to the laparoscope manipulator. The sys-tem state and image-processing results are also superimposedgraphically on the laparoscope image using the scan converter (asfeedback information for the surgeon), and the resulting image isdisplayed on the TV monitor.

To evaluate the performance of our system, an in vivo experi-ment was carried out, in which a surgeon used our system in a con-ventional laparoscopic operating room environment and a regularset of instruments to perform a laparoscopic cholecystectomy (gall-bladder removal) on a pig. As a result, the entire operative proce-dure was successfully and safely completed with our system. Inthis experiment, the robot never obstructed the surgeon’s work,and worrisome incidents and technical problems did not occur. Fig.4 shows scenes of the surgeon’s facial motions in the experiment.The upper part of Fig. 4 consists of the scope images which thesurgeon looked at; the lower part consists of the images of the sur-geon’s face from the surveillance camera. Each pair of images wastaken at the same time.

The system never mistook any other motion of the operatingsurgeon or any other surgeon who was present during the experi-ment. (e.g., see the second, third, and fourth images of Fig. 4, whichshow a surgeon walking behind the operating surgeon.) Also, theface gestures to stop the laparoscope motion were all correctly rec-ognized. These results demonstrate that our face-gesture recogni-tion system worked well. We received many positive comments,such as fast reaction time, high positioning accuracy, and easy andintuitive camera guidance, from the surgeons who performed or

looked in during the experiment. The operating surgeon, howev-er, did make the negative comment that after the experiment hefelt a little fatigue in his neck from a lot of rolling face motions.In fact, he made rolling gestures 4.2 times/min during the opera-tion.

ConclusionWe have developed a new robotic laparoscope positioning sys-

tem for solo surgery based on a real-time, face-tracking technique.Our system succeeded in freeing the surgeon’s hands and feet fromthe laparoscope guiding task while achieving safety, rapid reac-tion, and high positioning accuracy. Now we are studying animproved method for guiding the laparoscope based on the visu-al tracking of both the surgeon’s face and the surgical instruments,with the goal of reducing not only mental stress but also physicalstress, such as neck fatigue. An exhaustive comparative studybetween our system and other human-machine interfaces is alsoimportant and on-going [10].

References

1.Sackier, J. M.; Wang, Y., Surgical Endoscopy, 8, 63-66 (1994)

2.Begin, E. et al., Surgical Laparoscopy and Endoscopy, 5, 6-11 (1995)

3.Taylor, R. H. et al., IEEE Engineering in Medicine and Biology Magazine,14, 279-288 (1995)

4.Allaf, M. E. et al., Surgical Endoscopy, 12, 1415-1418 (1998)

5.Buess, G. F. et al., Surgical Endoscopy, 14, 395-399 (2000)

6.Vara-Thorbeck, C. et al., Surgical Endoscopy, 15, 924-927 (2001)

7.Kobayashi, E. et al., Computer Aided Surgery, 4, 182-192 (1999)

8.Finlay, P. A.; Ornstein, M. H., IEEE Engineering in Medicine and BiologyMagazine, 14, 289-291 (1995)

9. Dowler, N. J.; Holland, S. R. J., IEEE Robotics and Automation Magazine,3, 38-45 (1996)

10.Nishikawa, A. et al., Advanced Robotics, 17, 503-521 (2003)

Figure 4. Scenes of the surgeon’s face motions in an in vivo laparoscopic cholecystectomy experiment using the FAce MOUSe system. The upper part consists of the scope images whichthe surgeon looked at and the lower part consists of the images of the surgeon’s face from the surveillance CCD camera. Each pair of images was taken at the same time.

26

Engineering


A Synthesis of Sumanene, a Fullerene FragmentPaper in journals: this is the first page of a paper published in Science.[Science] 301, 1878 (2003)

� Reprinted with permission from Science, 301, Sakurai,H. et al., A Synthesis of Sumanene, a Fullerene Fragment, 1878(2003). Copyright 2004, AAAS.

27




First and Practical Synthesis of Sumanene, a C3v SymmetricPartial Structure of FullereneHIRAO Toshikazu and SAKURAI Hidehiro(Graduate School of Engineering)

Fullerene (1) and carbon nano tube, which are key compoundsin carbon chemistry, are recognized as potential materials.

Attention has been also drawn to bowl-shaped π-conjugated com-pounds including a partial carbon skeleton of fullerene, due to theirunique physical and chemical properties[1]. One of such bowl-shapedπ-conjugated compounds with C5v symmetry is corannulene (3),which was first synthesized in 1966[2]. Furthermore, some prac-tical synthetic routes have been developed for corannulene syn-thesis[3]. Another key fullerene fragment is sumanene (2)[4], whichpossesses a C3v symmetric structure. An important advantage fea-ture of sumanene over corannulene depends on the presence ofthree benzylic positions that should permit further functionaliza-tion to create new artificially-designed bowl-shaped species viathe corresponding radicals, cations, anions, carbenes, etc. Generally,functionalization of corannulene requires the electrophilic substi-tution on the aromatic ring. The structure and depth of the bowlscan characterize the π-conjugated compounds. The bowl ofsumanene (1.15 Å) is anticipated to be deeper than that of coran-nulene (0.89 Å).


illustrated below, the key intermediate is hexahydrosumanene 6,which may be prepared from the isomer 4 by transferring the alkenebridges. For this purpose, metathesis reaction provides a useful methodfor the carbon skeleton transformation.

Trimerization of norbornadiene affords syn- and anti-benzotris-(norbornadiene)s, which was achieved by a modified procedurebased on the organocopper-mediated cyclization[8]. Alternatively,stepwise transmetallation via an organotin compound increased thetotal yield of syn-4 from 2% to 11%[9]. The alkene-bridge exchangewas examined by the Ru-catalyzed tandem ring-opening metathe-sis (ROM) and ring-closing metathesis (RCM) reaction. syn-4 wastreated with 10 mol% of Cl2(PCy3)2Ru=CHPh in toluene at 0 ºC,then at room temperature under an atmospheric pressure of ethyl-ene, giving 6 via 5 in 30% yield. On the other hand, tandem ROM-RCM reaction did not proceed from anti-4, probably due to thesteric demanding. This isomeric transformation from syn-4 to 6 is

Conditions: a) butyllithium, potassium tert-butoxide, BrCH2CH2Br, tetrahydrofuran, −78 °C to −45 °C then Cul, room temperature. 7% yield (syn: anti = 1 : 3).b) butyllithium, potassium tert-butoxide, BrCH2CH2Br, tetrahydrofuran, −78 °C to −45 °C then tributylstannyl chloride, room temperature. c) Cu(2−C4H3SCO2), −20 °C to room temperature. 47% yield (2 steps; syn: anti = 1 : 3) d) cat. Cl2[P(C6H11)3]2Ru = CHPh, CH2 = CH2, toluene, −78 °C to room temperature,24 h. 30% yield. e) DDQ, toluene, 110 °C, 3 h. 70% yield.

Although several approaches to sumanene have been attempt-ed[4-6], there is no report of a successful route to synthesize it [7].Previous strategies for the synthesis of sumanene were based onapproaches from planar aromatic compounds as reported in thesynthesis of corannulene and hetero-fullerene. For example, tri-bromomethyltriphenylene has been reported to undergo flash vac-uum pyrolysis to afford mono-cyclized and di-cyclized productswith one or two broken outer-ring bonds, without formation of thedesired sumanene as shown above[4]. This finding suggests thata route from the planar compounds is unlikely to gain the strainenergy effectively, resulting in the termination of single or doublecyclization. Another synthetic strategy should be developed toovercome this problem.

Our synthetic route lies in constructing the three-dimensionalframework using tetrahedral sp3 carbons, which is converted to therequired π-conjugated structure by oxidative aromatization. As

Br

BrBr

FVP

3% 17% 0 %

+ +850 °C, 10 -4 Torr

Synthetic Route to Sumanene

29



considered to be exothermic by 51.4 kcal/mol estimated by DFTcalculation (B3LYP/6-31G*). On the other hand, the reaction fromanti isomer of 4 is calculated to be endothermic by 37.4 kcal/mol,indicating that the transformation from syn-4 to 6 is preferrable.Finally, 6 underwent dehydrogenative aromatization with organ-ic oxidant, DDQ (2,3-dichloro-5,6-dicyano-1,4-benzoquinone) togive sumanene (2) in 70% yield. All of these steps were performedunder mild conditions, so no severe conditions such as flash vac-uum pyrolysis are necessary[7,10,11].

1H-Nuclear magnetic resonance spectra elucidate the symmet-rical structure of sumanene. In the NMR spectrum of 2 in CDCl3,only three signals were observed. One singlet corresponds to thearomatic protons and a pair of doublets is assignable to the ben-zylic protons. 13C-NMR showed four kinds of carbon signals.These results indicate that sumanene is present as a C3 symmetricstructure in solution.

Further dynamic behavior of the bowl-shaped compounds liesin bowl-to-bowl inversion. Corannulene exhibits rapid bowl-to-bowl inversion at room temperature with an energy barrier of 10.2kcal/mol at −64 ºC[10,11]. In contrast, three five-memberred ringsof sumanene are believed to make it more rigid. The first estima-tion by MNDO predicts the inversion energy barrier as 24.2kcal/mol[12], while by recent ab initio calculations the barrierranges from 16.8 to 19.3 kcal/mol[13]. These estimations indicatethat sumanene should not undergo bowl-to-bowl inversion on theshort time like NMR time scale. Variable temperature NMR ofsumanene in p-xylene-d10 solution revealed that a pair of doubletsassignable to the benzylic protons gradually broadened at above60 ºC in the 300-MHz NMR spectra, indicating that the bowl-to-bowl inversion process occurs. Both doublets disappeared at 140ºC, but a new singlet peak was not observed under 150 ºC. Theinversion barrier is estimated as 19.6 kcal/mol at 140 ºC. The find-ings suggest that sumanene appears to be rigid. The inside and out-side moieties of the bowl can be differentiated. In this sense, thedynamic behaivor may permit the design of functional carbonmaterials.

References

1.Mehta, G.; Rao, H. S. P., Advances in Strain in Organic Chemistry, JAI Press,Greenwich, Vol. 6, p 139-187 (1997).

2.Barth, W. E.; Lawton, R. G., Journal of the American Chemical Society, 88, 380(1966).

3.Sygula, A.; Rabideau, P. W., Journal of the American Chemical Society, 122,6323 (2000).

4.Mehta, G.; Shah, S. R.; Ravikumar, K., Journal of the Chemical Society, Chem-ical Communications, 1993, 1006.

5.Gupta, H. K.; Lock, P. E.; McGlinchey, M. J., Organometallics, 16, 3628 (1997).

6.Priyakumar, U. D.; Sastry, G. N., Tetrahedron Letters, 42, 1379 (2001).

7. Imamura, K.; Takimiya, K.; Aso, Y.; Otsubo, T., Chemical Communications,1999, 1859.

8.Durr, R.; Lucchi, O. D.; Cossu, S.; Lucchini, V., Chemical Communications, 1996,2447.

9.Borsato, G.; De Lucchi, O.; Fabris, F.; Groppo, L.; Lucchini, V.; Zambon, A.,The Journal of Organic Chemistry, 67, 7894 (2002).

10.Scott, L. T., Hashemi, M. M., Bratcher, M. S., Journal of the American Chemi-cal Society, 114, 1920 (1992).

11.Borchardt, A.; Fuchicello, A.; Kilway, K. V.; Baldridge, K. K.; Siegel, J. S., Jour-nal of the American Chemical Society, 114, 1921 (1992).

12.Sastry, G. N.; Jemmis, E. D.; Mehta, G.; Shah, S. R., Journal of the ChemicalSociety, Perkin Transactions 2, 1993, 1867.

13.Priyakumar, U. D.; Sastry, G. N., Journal of Physical Chemistry: A, 105, 4488(2001).

In conclusion, we succeeded in the efficient synthesis ofsumanene for the first time. Sumanene is of potentil for the designof a variety of bowl-shaped π-conjugated compounds, which areconsidered to be useful as materials.

30

Engineering


Polymerization within a Molecular-Scale Stereoregular TemplatePaper in journals: this is the first page of a paper published in Nature.[Nature] 429, 52-55 (2004)

� Reprinted by permission from Nature 429, 52-55, 2004/5/6. Copyright: Macmillan Magazines Ltd.

31




Bioinspired Template Polymerization : Polymerization within amolecular-scale stereoregular templateAKASHI Mitsuru(Graduate School of Engineering)

In the field of synthetic polymer chemistry, template polymer-ization is one of the final goals in structure and molecular weight

regulating polymerization. In 1955, Natta used Ziegler organometal-lic catalysts to initiate the stereoregular polymerization of vinylmonomers such as propylene to create polymers superior in somephysical properties to conventional polymers (Nobel Prize in 1963).Nowadays, these commercially available polymers have becomevery important and indispensable materials. After Natta’s epoch-making research, many basic studies on stereoregular polymerizationwere initiated.Concurrently, Watson and Crick clarified the struc-ture and functionalities of double stranded DNA by X-ray analy-sis in 1953 (Nobel Prize in 1962).Even at that time, I think thatchemists noticed the possibility of molecular duplication. As thesebioinspired concepts influenced synthetic chemists, trials on tem-plate polymerization originated and evolved with enthusiasm.Thusfar,several strategies have been developed for the realization oftemplate polymerization. Our research group considers that enzymesefficiently synthesize biopolymers by organizing monomer unitswithin regularly structured molecular-scale spaces and exploitingweak non-covalent interactions, such as hydrogen bonds, to con-trol the polymerization[1] process. This ‘template’ approach isboth attractive and challenging for synthetic polymer synthesis,where structurally regulated molecular-scale spaces could in prin-ciple provide solid-phase reaction sites for precision polymeriza-tion. As shown in Fig. 1, stereoregular poly(methyl methacry-late)(PMMA) is known to form a DNA-like double-helical structure.

Previously, free radical polymerization of methyl methacrylatein solutions containing stereoregular isotactic (it) or syndiotactic(st) poly(methyl methacrylate) (PMMA) has been shown to resultin template synthesis[2,3] of the opposite PMMA based on stere-ocomplex formation[4,5] with van der Waals interactions. However,using the structure of a solid to determine the stereochemical struc-ture of a polymer has not been satisfactorily achieved[6]. Here weshow that macromolecularly porous ultrathin films, fabricated bya single assembly step, can be used for the highly efficient stere-

oregular template polymerization of methacrylates through stere-ocomplex formation. This reaction mould accurately transfers itsstructural properties of stereoregularity, molecular weight and orga-nization within the template to the new polymer.

Structurally regulated molecular-scale spaces (nanospaces) in ultra-thin polymer films are potentially valuable as solid-phase reaction sitesfor precision polymerization. As the macromolecular mobility isrestricted in the solid state, the template effect of a constituent poly-mer is potentially accelerated. It is, however, difficult to prepare suchfilms using conventional methods. We selected a methodology involv-ing the solvent extraction of certain polymer components from ultra-thin polymer films for the construction of porous host films for tem-plate polymerization. The precursor films were fabricated using layer-by-layer assembly[7], which can be demonstrated by the simple alter-nate immersion of solid substrates into interactive polymer solutions.Although layer-by-layer assembly conventionally produces poly-electrolyte multilayers based on polyion complexes, stereocomplex-es between it-PMMA and st-polymers of methacrylates were alsoassembled on solid substrates[8-10], thus forming ultrathin stereo-complex films with a double-stranded helical structure, in which it-PMMA was surrounded by twice the unit-molar amount of st-poly-mers (1:2 stoichiometry)[5]. In the case of the stereocomplex film com-posed of it-PMMA and st-PMAA, st-PMAA was selectively extract-ed from the film in an aqueous alkaline solution, resulting in a macro-molecularly porous it-PMMA film[11]. This porous film subsequentlyincorporated st-PMAA previously extracted, dependent on the st-PMAA concentration, thus demonstrating an unusual macromolecu-lar recognition. However, it-PMMA was not extracted from the filmin organic solvents, indicating that a versatile host composed of it- orst-polymers for template polymerization cannot be prepared by thismethod.

In this study, we focused on the formation of a stereocomplexbetween it-PMMA and st-PMAA with a 1:1 unit-molar stoi-chiometry[12,13]. Layer-by-layer assembly of these polymersunder suitable conditions produced a stereocomplex film with a1:1 stoichiometry, and the subsequent selective extraction of a sin-gle component successfully produced porous st-PMAA or it-PMMA films, of which the macromolecular nanospaces could beused in stereoregular template polymerization of it-PMMA and st-PMAA, respectively. Although it is more difficult to obtain it-poly-mers of methacrylates by free radical polymerization owing to thesteric hindrance of lateral groups, the stereoregularity of polymerssynthesized by this method was absolutely dependent on that ofthe template polymers. Accordingly, the present study demon-strated the novel template synthesis of valuable polymers using

Figure 1. DNA and stereocomplex of PMMA


both weaker van der Waals interactions and artificially construct-ed nanospaces in films. The present template polymerization methodis schematically represented in Fig. 2.

A 9-MHz quartz crystal microbalance (QCM) substrate[8-11],which could determine the amount of assembled polymer from itsfrequency decrease, was used to quantify the assembly of it-PMMAand st-PMAA. To obtain an assembly with 1:1 stoichiometry,dimethylformamide (DMF) and DMF/water (2/3, v/v) were select-ed as solvents for it-PMMA and st-PMAA, respectively. This sol-vent combination was determined by a detailed assembly analy-sis, following methods used in a previous study[13]. Alternateimmersion in both solutions produced a film with the objectivestoichiometry of 1.0±0.2 (st/it-unit ratio). The thickness and themean roughness (Ra) for 6 cycles ((it-PMMA/st-PMAA)6) wereestimated to be 20.1±0.3 nm and 3.2 nm, respectively. Consideringthe assembly amount analysed by the QCM (4.1±0.1 µg cm−2), themean density was estimated to be 1.0 g cm−3. The immersion ofthe film in both chloroform and a 10 mM sodium hydroxide aque-ous solution resulted in desorption of approximately half theamounts of polymers in the film. As the desorption amounts wereconsistent with the amounts of each polymer totally assembled,this suggests that it-PMMA and st-PMAA were selectively extract-ed from the film. Attenuated total reflection (ATR) spectra in thecarbonyl vibration region also demonstrated the selective disap-pearance of peaks at 1,731 cm−1 and 1,717 cm−1, corresponding torespective it-PMMA and st-PMAA. The thickness after the extrac-tion of it-PMMA and st-PMAA was 18.3±2.8 nm and 19.4±2.8nm, respectively, and these values were significantly consistent withthe initial thickness of the stereocomplex film within experimen-tal error. In addition, the Ra was 2.1 nm and 2.8 nm, respectively,and was slightly smoother than before. These observations indi-cated that a porous film composed of st-PMAA or it-PMMA suit-able for template polymerization was successfully prepared by thesolvent extraction. It is difficult to explain reasonably why the filmswere not compressed by the extraction. The terminals of the poly-mers as well as some segments that did not join in the complexmight support the film structure. This observation was similar tothe previously reported case of selective extraction from a double

stranded stereocomplex film[11].For the first step of the analysis of template polymerization, the

QCM substrate coated with the porous st-PMAA film was immersedin a 10 ml DMF solution of methyl methacrylate (MMA; 1.7 mgml−1) for 3 h at 70ºC in the presence of a free radical initiator, 2,2′-azobisisobutyronitrile (5 mg ml−1) under a nitrogen atmosphere.After gentle rinsing with DMF, approximately 90% of the it-PMMA previously extracted was recovered (the percentage wasalmost the same for all template st-PMAAs used). The ATR spec-trum of the obtained film demonstrated the same peaks as thoseobserved for the stereocomplex film, suggesting that it-PMMA wasprepared, and that it formed a stereocomplex with the st-PMAAin the film. In order to analyse the stereoregularity and the mole-cular weight of the PMMA as the second step, the porous film wassimilarly prepared on silica particles (10 g, mean diameter 1.6 µm),following methods described in a previous study[14]. MMA wasthen polymerized, and the PMMA was characterized by extrac-tion using chloroform. Approximately 0.4 g PMMA was obtainedfrom the feeding of 1.5 g MMA per 500 ml DMF, thus resultingin a yield of approximately 25% using these polymerization con-ditions. A typical 1H-NMR spectrum of the PMMA clearly demon-strated the it-specific polymerization of MMA in the porous st-PMAA film (Fig. 3a), thus indicating that the it-specific templatepolymerization was achieved on the basis of stereocomplex for-mation with the st-PMAA. Furthermore, size exclusion chro-matography (SEC) curves potentially demonstrated the control ofmolecular weights with a narrow distribution (Fig. 3b).

The experimental data from the template polymerization ofMMA are summarized in Table 1. The isotacticities were greaterthan 92% in all cases. The number-average molecular weight (Mn)of the host st-PMAA significantly controlled the Mn of the it-PMMA, and was almost the same. These observations are surprisingbecause the polymerization conditions, such as concentrations ofmonomers and initiators, were the same in all cases except for themolecular weight of the template st-PMAA. Although the same it-PMMA (Mn≈20,000) was used for the preparation of host films,the molecular weight of the it-PMMA prepared is absolutely depen-dent on that of the host st-PMAA.

The porous film of it-PMMA was used to prepare st-PMAA.Silica particles coated with the porous it-PMMA film were simi-larly used for free radical polymerization of MAA in water in thepresence of a water-soluble free radical initiator for 2 h at 40ºCunder a nitrogen atmosphere. The PMAA prepared in the film wasextracted in an alkaline solution, methylated to prepare PMMA,and then characterized, as also shown in Table 1. St-PMAA wassuccessfully prepared by using the porous it-PMMA film, and themolecular weight of the host it-PMMA similarly regulated that ofthe st-PMAA. These observations indicate alternative utilizationof the porous films for stereoregular template polymerization ofmethacrylates. We note that conventionally prepared spin- andsolution-casting films did not realize the above template poly-merization, even under the same polymerization conditions.Furthermore, template polymerization of the combination betweenit-PMMA and st-PMAA in the solution system only demonstrat-

Figure 2. A schematic representation of template polymerization using ultrathin porous films.Layer-by -layer assembly prepares the ultrathin film of a stereocomplex comprising it-PMMAand st-PMMA. Two polymer chains are shown as rigid helical rods, but should be partial-ly distorted and entangled in the film. There must be a disordered region of the stereo-complex. A single component is selectively extracted from the film, resulting in thepreparation of the porous film with regular nanospaces. Then, the porous film is used as areaction mould for free radical template polymerization of MMA or MAA, followed by theregeneration of the stereocomplex film.

33



ed the acceleration of polymerization rates, because the associa-tion is relatively weaker than that of stereoregular PMMAs[15].These observations also confirmed the potential of the presentporous films. Preliminary experiments of template polymerizationof MAA in a porous it-PMAA film[11] prepared from the 1:2 (it:st)complex assembly revealed that the molecular weight of the st-PMMA polymerized was approximately twice that of the templateit-PMMA, indicating the potential transferring of the assembly struc-

References

1.Alberts, B. et al., Essential Cell Biology 2nd edn (Garland, New York, 2003).

2.Buter, R., Tan, Y. Y. & Challa, G., J. Polym. Sci. A 10, 1031-1049 (1972).

3.Buter, R., Tan, Y. Y. & Challa, G., J. Polym. Sci. Polym. Chem. Edn 11, 1003-1011 (1973).

4.Spevácek, J. & Schneider, R., Adv. Colloid Interface Sci. 27, 81-150 (1987).

5.Schomaker, E. & Challa, G., Macromolecules 22, 3337-3341 (1989).

6.Buter, R., Tan, Y. Y. & Challa, G., J. Polym. Sci. Polym. Chem. 11, 1013-1024 (1973).

7.Decher, G. in Multilayer Thin Films (eds Decher, G. & Schlenoff, J. B.) 1-46 (Wiley-VCH, Weinheim, 2003).

8.Serizawa, T. et al., J. Am. Chem. Soc. 122, 1891-1899 (2000).

9.Serizawa, T. et al., Langmuir 16, 7112-7115 (2000).

10.Hamada, K.-I. et al., Langmuir 17, 5513-5519 (2001).

11.Serizawa, T., Hamada, K.-I., Kitayama, T. & Akashi, M., Angew. Chem. Int.Edn Engl. 42, 1118-1121 (2003).

12.Lohmeyer, J. H. G. M., Kransen, G., Tan, Y. Y. & Challa, G., Polym. Lett.Edn 13, 725-729 (1975).

13.Lohmeyer, J. H. G. M., Tan, Y. Y., Lako, P. & Challa, G., Polymer 19, 1171-1175 (1978).

14.Caruso, F., Caruso, R. A. & Möhwald, H., Science 282, 1111-1114 (1998).

15.Lohmeyer, J. H. G. M., Tan, Y. Y. & Challa, G., J. Macromol. Sci. A 14, 945-957 (1980).

ture as well as further applications of porous thin films.We now consider the polymerization mechanism in the films.

Two contradictory polymerization processes would be expected:(1) polymerization after sufficient coordination of monomers withtemplates, and (2) stepwise polymerization with gradual incorpo-ration (or slow diffusion) of monomers into the films. The quan-titative QCM analysis did not show the sorption of monomers intothe films, and did not support the former mechanism. In order tounderstand these mechanisms in detail, time-dependent SEC curvesof the it-PMMA prepared by the st-PMAA host were obtained (detaare not shown). The curve with a single and narrow peak widthshifted to a higher molecular weight, thereby supporting the lattermechanism, which is similar to living radical polymerization ratherthan conventional free radical polymerization. As we stopped thepolymerization of MMA after 3 h by opening the reaction vesselto the air, the radical terminals, which had been present in the filmsfor at least 3 h, seemed to react with oxygen. The molecular weightof the st-PMMA prepared was similarly shifted, and a similarmechanism was supported.

This is, to our knowledge, the first report of an artificially con-structed polymerization template and subsequent polymerizationwith the efficient transfer of structural information, similar to thatseen in biosynthetic processes. Porous polymer matrices provid-ed a reaction mould for stereoregular polymerization of methacry-lates. It- and st-polymers with high stereoregularity and with a nar-row molecular weight distribution were successfully prepared.

Figure 3. Characterization of it-PMMAs polymerized in porous st-PMAA films for3 h at 70°C. a : A typical 1H-NMR spectrum of it-PMMA polymerized in aporous st-PMAA (Mn 23,800) film.b : SEC curves of it-PMMAs polymerized in porous films of st-PMAAs with vari-ous molecular weights.

Because particles coated with functional ultrathin films are readi-ly handled and reusable (the porous films were reused at least 3times), we expect large-scale synthesis of the stereoregular poly-mers to be achieved in the near future. Easier synthesis of stereo-regular polymers of methacrylates potentiates their applications intechnological and biomedical fields. Regulated nanospaces pre-pared in synthetic polymer assemblies should open the way to anew field of macromolecular recognition and synthesis.

Table 1. Analytical data on stereoregular template polymerization of methacrylate.

34

Engineering


Localization of Electromagnetic Waves in Three-Diemsional Fractal CavitiesPaper in journals: this is the first page of a paper published in Physical Review Letters.[Physical Review Letters] 92, 093902-1-4 (2004)

� Reprinted with permission from M.Wada Takeda, S. Kirihara, Y. Miyamoto, K. Sakoda, K. Honda, Physical Review Letters, 92, 093902/1-4(2004). Copyright 2004 by the American PhysicalSociety.

35




Localization of Electromagnetic Waves in Photonic FractalsKIRIHARA Soshu and MIYAMOTO Yoshinari(Joining and Welding Research Institute)

Introduction

Propagation properties of electromagnetic waves in periodicand quasiperiodic structures have been of theoretical and prac-

tical interest for the past two decades. Photonic (electromagnetic)crystals consist of a periodic dielectric array with different dielec-tric constants that can totally reflect light or electromagnetic waveswith a wavelength similar to the periodicity due to Bragg diffrac-tion. Formation of photonic band gaps, localization of light, con-trol of spontaneous emission, and other properties associated withphotonic crystals have been studied intensively [1-3]. A photonicband gap is understood as an analogy to the electronic band gapin semiconductors.

On the other hand, interaction of optical, electromagnetic, andacoustical waves with fractal structures has been of interest inrecent years as well. A fractal is defined as a rough and irregularstructure with self-similarity [4]. In other words, the local config-uration or substructure is similar to the whole configuration or theentire structure. The fractal structure is essentially different fromthe periodic structure because it has no periodicity and no transla-tional symmetry.

Though there are many reports concerning the localization oflight or electromagnetic waves in various fractal structures, theyare restricted to one-dimensional or planar fractals [5,6]. A 3D frac-tal structure is necessary to completely localize the light or elec-tromagnetic waves in 3D space. However, 3D fractal structureshave not been investigated due to their difficult construction. Wehave succeeded in fabricating 3D fractal structures called Mengersponge, which are made of dielectric media such as epoxy and ceram-ics. These structures exhibit an electromagnetic wave that is strong-ly localized in the central air cavity region of the 3D fractal with-

out reflection and transmission. Such a material has not been pre-viously reported, thus we defined this localization as “photonic frac-ton” in the “photonic fractal”.

Design and Fabrication of 3D Fractal StructuresWe designed the Menger sponges of size comparable with

microwave range by using a CAD (Computer Aided Design) pro-gram and then automatically fabricated them from epoxy resin usinga rapid prototyping method called stereolithography [8]. Thismethod enables us to form three-dimensional complex objects viaa layer-by-layer process using CAD/CAM (Computer AidedManufacturing) system. Each layer of about 100 µm in thicknessis solidified by scanning the surface of the photosensitive liquidresin with an ultraviolet laser beam. When ceramic particles aredispersed in the liquid resin, we can form the Menger sponges ofceramic-epoxy composite with high dielectric constant. It is verydifficult to fabricate such a complex fractal structure by other meth-ods except CAD/CAM systems.

We constructed the Menger sponge structure from a dielectriccube with a side length a, which is composed of epoxy or TiO2-SiO2 ceramics dispersed epoxy composite. The cube is divided into27 identical cube pieces, and the seven pieces at the body- andface-centers are extracted. By repeating the same extraction processfor the remaining 20 pieces we create the Menger sponge. Mengersponge structures with different stages of the repeating process (0,1, 2) are illustrated in Fig.1. The side length a of the initiator is27mm. The lengths of the longest, middle, and shortest sides ofthe square air rods in the stage 3 Menger sponge are 9mm, 3mm,and 1mm, respectively.

Figure 1. Construction of Menger sponge fractals with self-similar structureThe stage 2 Menger sponge consists of 20 smaller cubes of the stage 1 Menger sponge.


Strong Localization of Electromagnetic WavesNormally incident reflection and transmission spectra of the

Menger sponge of stage1, 2, and 3, and those of simple cubic pho-tonic crystals with square air rods of 3mm and 1mm formed bythe same CAD/CAM stereolithography were measured by usinga network analyzer and two mono-pole antennae. The measure-ment system is schematically illustrated in Fig.2.

The results for the stage 3 Menger sponges made of epoxy andTiO2-SiO2 ceramics dispersed epoxy are shown in Fig.3. In thetransmission spectrum of the Menger sponge made of epoxy asseen at the upper figures, a sharp transmission attenuation as largeas -31dB was observed at 12.8 GHz and a relatively broad atten-uation of reflection as large as -7dB was observed at around 13.0GHz. The dielectric loss of the material was 0.2. In addition, forthe stage 2 Menger sponge, a sharp attenuation and a correspond-ing broad attenuation of reflection were both observed at around12.8GHz as well. In the case of the stage 3 Manger sponge madeof TiO2-SiO2 dispersed epoxy, both highly sharp attenuations forreflection and transmission to -40dB were observed at about 8 GHzas seen in the lower figures. The dielectric loss of the fractal mate-

rial was 0.1.In contrast, such an anomalous attenuation was observed nei-

ther in the transmission and reflection spectra of simple cubic pho-tonic crystals and Menger sponge of stages 0 and 1. Therefore, theorigin of the anomalous attenuations observed both in the reflec-tion and transmission spectra of the Menger sponge of stage 2 and3 should be intrinsic to the fractal structure because they have theself-similar structure.

Figure 4 shows the spatial distribution of the electric field at12.8 GHz in the stage 3 Menger sponge of epoxy, which was mea-sured at positions along the central air-rod by a probe method usingtwo mono-pole antennae. A mono-pole antenna for emission wasplaced at the same position in the case of the transmission mea-surement and another antenna was inserted into the largest air rodat the center. The electrical field intensity shows a typical doublemaximum curve symmetric for the center point. Apparently thismode is strongly localized almost within the central largest cubicair-cage region and the electric field decreases abruptly outside thecentral air cube. The localized mode has a cubic-shell shape whoseedges and corners are rounded. A similar localization mode was

Figure 2. Schematic illustration of experimental setup forthe measurement of microwave transmission andreflection properties of a photonic fractal.

Figure 3. Localization of electromagnetic waves inMenger sponge fractals. The wavelength λloc of the localized mode in thedielectric Menger sponge fractal is calculated usingthe empirical equation λloc = 2/3 .a √ εeff, where aand εeff are the edge length and the spatially averageddielectric constant of the fractal cube, respectively.The calculated and measured frequencies of the localizedmodes in the stage 3 Menger sponges of epoxy (upperfigures) and 10vol.%TiO2-SiO2/epoxy (lower figures)showed good agreements.

37



observed in stage 3 Menger sponge of TiO2-SiO2 dispersed epoxyas well.

Discussion Let us compare these propagation characteristics of the Menger

sponges to those of photonic crystals. In the case of the localizedmode due to a structural defect in a photonic crystal, we must observea sharp transmission peak in a nontransparent region due to thephotonic band gap and a sharp dip of reflectance in the total reflec-tion region. The characteristics observed in Fig.3 do not meet anyof these features.

The optical length of the stage 3 Menger sponge, which is cal-culated as a product of its actual thickness and spatially averagedrefractive index, is 35.6mm. This value is larger than the relevantwavelength by only 50%; that is, the Menger sponge confines theelectromagnetic wave with a wavelength corresponding to 2/3 ofits optical length. It is reported that the optical or electromagneticwave excitations in one dimensional or planar fractals localize insmall subwavelength areas [5,6]. On the other hand, in the case ofphotonic crystals, the localization length is several times of the lat-tice constant of the crystal. So, the optical length necessary to con-fine a localized mode is several times the relevant wavelength infree space. We can, thus, conclude that the confinement of the elec-tromagnetic energy by the Menger sponge is quite strong.

These strong localizations of electromagnetic waves in the 3Dfractal structure can not be explained by simple Bragg diffractionand band gap formation. It is inferred that such a localization mayoccur due to the interference of multiple reflections in the fractalstructure. The theory necessary to account for the localization ofelectromagnetic waves in 3D fractal structures is a very interest-ing challenge.

ConclusionsWe observed localization of microwaves in the Menger sponges

that is intrinsic to their fractal structure, and this is the first reporton the light localization in 3D fractal cavities. This localization isnot caused by the presence of a photonic band gap as in photoniccrystals but should be attributed to a singular photon density ofstates realized in the fractal structure. Confinement of electro-magnetic or radiation energy in the fractals may be useful for var-ious devices in communication, information, energy, medical andother fields.

References

1. Ohtaka, K., Phys. Rev. B19, 5057 (1979)

2. Yablonovitch, E., Phys. Rev. Lett. 58, 2059 (1987)

3. Sakoda, K., Optical Properties of Photonic Crystals (Springer-Verlag, Berlin,2001)

4. Mandelbrot, B.B., The Fractal Geometry of Nature (Freeman, San Francis-co, 1982)

5. Safonov, V. P., Shalaev, V. M., Markel, V. A., Danilova, Yu. E., Lepeshkin,N. N., Kim, W., Rautian, S. G., and Armstrong, R. L., Phys. Rev. Lett. 80,1102 (1994)

6. Wen, W., Zhou, L., Li, J. Ge, W., Chan, C. T., and Sheng, P., Phys. Rev. Lett.89, 223901-1 (2002)

7. Sun, X., and Jaggard, D. L., J. Appl. Phys., 70, 2500 (1991)

8. Kirihara, S., Takeda, M.W., Sakoda, K., Miyamoto, Y., Solid State Commun.124, 135 (2002)

Figure 4. Profile of electric field intensity of the localizedmode at 12.8 GHz in the stage 3 Menger sponge fractalof epoxy.

38

Biology


Involvement of Histone H1.2 in Apoptosis Induced by DNA Double-Strand BreaksPaper in journals: this is the first page of a paper published in CELL.[CELL] 114, 673-688 (2003)

� Reprinted from CELL, 114, Konishi,A. et al., Involvement of Histone H1.2 in Apoptosis Induced by DNA Double-Stand Breaks, 673-688(2003), Copyright 2004, with permission from Elsevier.

39




Involvement of Histone H1.2 in Apoptosis Induced by DNADouble-strand BreaksSHIMIZU Shigeomi and TSUJIMOTO Yoshihide(Graduate School of Medicine)

Introduction

Programmed cell death or apoptosis plays an important role invarious biological events in metazoans, including develop-

ment, maintenance of tissue homeostasis, and elimination of harm-ful cells. The process of apoptosis is driven by a family of pro-teases called caspases, which cleave various cellular proteins tocause apoptotic death. It has been shown that the mitochondria playa crucial role in apoptosis by releasing several apoptogenic mole-cules such as cytochrome c, which activate downstream destruc-tion programs including the caspase cascade [1]. The best-charac-terized regulators of apoptosis are the Bcl-2 family of proteins, whichdirectly modulate outer mitochondrial membrane permeability dur-ing apoptosis. This family of proteins comprises anti-apoptotic mem-bers, such as Bcl-2, as well as pro-apoptotic members [2]. Althoughseveral molecules have been reported to be involved in transmis-sion of DNA damage-induced apoptotic signals to the mitochon-dria, the mechanisms of signal transmission are not fully under-stood.

In the present study, we took a biochemical approach to the iden-tification of a signaling molecule involved in DNA damage-inducedapoptosis by searching for a cytochrome c-releasing factor that appearedin the cytoplasm after X-ray irradiation.

Purification of an X-ray-induced cytochrome c-releasing factoras histone H1.2

To identify DNA damage-activated signaling molecule(s) withcytochrome c releasing activity, we set up an in vitro assay sys-tem using isolated rat liver mitochondria and rat thymus cytosol.The cytosol, obtained from the thymus of rats subjected to whole-body X-ray irradiation, was capable of inducing cytochrome crelease from the mitochondria. Following the purification proto-col described in Fig. 1A, we identified a 34 kD protein withcytochrome c-releasing activity (Fig. 1B). Sequence analysis indi-cated that the 34 kD protein was rat histone H1.2. We also showedthat mouse H1.2 has cytochrome c-releasing activity. Among sev-eral mouse histone H1 isofomrs, only histone H1.2, but not otherisoforms, possesses cytochrome c-releasing activity (Fig. 2).

Figure 1. Purification of a cytochrome c-releasing factor as histone H1.2(A) Diagram of the purification method. (B) Purification of cytochrome c-releasing factor fromrat thymus. Aliquots of the Mini S column fractions were subjected to SDS-PAGE, followedby silver staining (upper panels), and also assessed for cytochrome c-releasing activity withthe mitochondria (lower panels).

Figure 2. Histone H1.2 induces cytochrome c release from isolated mitochondria (A) Histone H1 from mouse thymus was separated by RP-HPLC. (B) Each histone H1 subtype as well as a low molecularweight protein in peak 1were assessed for cytochrome c-releasing activity. Peak 6 corresponds to H1.2


p53-dependent release of nuclear histone H1.2 after X-ray irra-diation

Histone H1 is not detected in the cytosol of healthy mouse thy-mocytes and mouse embryonic fibroblast cells (MEFs), whereasX-ray irradiation induced a cytoplasmic increase of histone H1s(Fig. 3), which was suppressed by an inhibitor of nuclear export,leptomycin B, and depended on p53. Cytoplasmic H1 did not showany X-ray irradiation-specific modifications.

Inhibition of X-ray-induced apoptosis by reducing H1.2 expres-sion

The level of histone H1.2 was reduced in MCF7 cells by theexpression of H1.2 antisense RNA. Two clones (AS#1 and AS#2)with almost complete absence of histone H1.2, but not the otherH1 subtypes were more resistant to X ray-induced cytochrome crelease and cell death than control cells (Fig. 4). AS cells were alsomore resistant to etoposide, but not to UV, paclitaxel and TNF-αthan control cells (Fig. 5). These results indicated that histone H1.2is specifically required for apoptosis induced by etoposide and X-rays, both of which cause DNA double-strand breaks.

Figure 3. Cytoplasmic accumulation ofhistone H1.2 after X-ray irradiationE1A/Ras-transformed MEF were exposed toX-ray irradiation (20 Gy). After 24 hr, the cellswere immunostained with an anti-histone H1monoclonal antibody.

Figure 4. Inhibition of X-ray-induced apoptosis by reducing histone H1.2expressionMCF7 variants transfected with caspase-3 cDNA and GFP DNA were exposed to irradia-tion (20 Gy) (IR) or were not treated (NT). After 24 hr, the cells were stained withHoechst 33342, and apoptotic cells were detected by examining the nuclear morphology.

Concluding remarksBy adopting a biochemical approach to investigate the trans-

mission of X-ray-induced apoptotic signals to the mitochondria,we identified histone H1.2. Histone H1.2 is specifically involvedin apoptosis induced by DNA double-strand breaks by translocat-ing from the nucleus to the mitochondria, and subsequently induc-ing mitochondrial membrane permeabilization. It was particular-ly interesting that histone H1.2 did not play an important role inapoptosis induced by UV irradiation, which also causes DNA dam-ages. DNA double-strand breaks, but probably not UV damages,trigger a megabase range of chromatin remodeling, which likelyleads to release of H1 from the chromatin. Histone H1.2 probablycooperates with other pro-apoptotic proteins in DNA damage-induced apoptosis.

41



Histone H1.2-deficient mice show cellular resistance to X-ray-induced apoptosis

Finally, we examined mice lacking histone H1.2. H1.2-KO thy-mocytes showed superior viability to WT thymocytes over a widerange of X-ray radiation doses. We also examined X-ray-inducedapoptosis of small bowel cells. When WT mice were exposed toX-ray irradiation (10 Gy) and sacrificed after 2 days, the small intes-tine was severely damaged, with denudation of the crypt and vil-lus system (Fig. 6), as well as positive TUNEL-staining of a largenumber of epithelial cells and enterocytes (Fig. 6). In contrast, thesmall bowel looked normal after X-ray irradiation of H1.2 KO mice,and only a few TUNEL-positive cells were observed (Fig. 6).

References

(1)Wolf, B.B., and Green, D.R. (2002). Apoptosis: Let-ting slip the dogs of war. Curr. Biol. 12, R177-R179.

(2)Tsujimoto, Y. (2003). Cell death regulation by theBcl-2 protein family in the mitochondria. J. Cell.Physiol. 195, 158-167.

Figure 5. Requirement of histone H1.2 for apoptosis induced byDNA double-strand breaks, but not other stimuli. MCF7 variants (with caspase-3 cDNA + GFP DNA) were treated as indicated for24 hr. Apoptosis was assessed by nuclear morphology.

Figure 6. Inhibition of X-ray-induced apoptosis in the small bowel by histone H1.2deficiency. Wild-type mice (WT) and H1.2 knockout mice (KO) were irradiated with X-rays (10Gy). After 48 hr, equivalent segments of the jejunum were excised and subjected to hematoxylin andeosin (HE) and TUNEL staining.

42

Biology


Nuclear Cataract Caused by a Lack of DNA Degradation in the Mouse Eye LensPaper in journals: this is the first page of a paper published in Nature.[Nature] 424, 1071-1074 (2003)

� Reprinted by permission from Nature 424, 1071-1074, (2003). Copyright: Macmillan Magazines Ltd.

43




Nuclear Cataract Caused by a Lack of DNA Degradation inthe Mouse Eye LensNAGATA Shigekazu(Graduate School of Frontier Biosciences and Graduate School of Medicine)

Introduction

The eye lens is composed of fiber cells which develop from theepithelial cells on the anterior surface of the lens [1].

Differentiation into a lens fiber cell is accompanied by changes incell shape, the expression of crystallins, and the degradation of cel-lular organelles [2]. The loss of organelles is believed to ensurethe transparency of the lens, but the molecular mechanism behindthis process is completely unknown. Here, we show that DLAD(DNase II-Like Acid DNase) [3], also called DNase IIβ [4] isexpressed in lens cells. Mice deficient in the DLAD gene wereincapable of degrading DNA during lens cell differentiation, andundigested DNA accumulated in the fiber cells. The DLAD-/- micedeveloped cataracts of the nucleus lentis, and their response to lighton electroretinograms was severely reduced. These results indi-cate that DLAD is responsible for the degradation of nuclear DNAduring lens cell differentiation, and that if DNA is left undigestedin the lens, it causes cataracts of the nucleus lentis, blocking thelight path.

Specific Expression of DLAD in the lensTo identify the DNase responsible for the degradation of nuclear

DNA during lens cell differentiation, we first examined by real-time PCR the expression of three DNases (DNase II, DLAD, and

CAD) in various mouse tissues. DNase II is an acid DNase foundin lysosomes that has an essential role in the degradation of nuclearDNA expelled from erythroid precursor cells [5], and in the degra-dation of the DNA of apoptotic cells after macrophages engulf them[6]. DLAD is also an acid DNase that has homology to DNase II[3, 4]. DLAD was reported to be expressed in the mouse liver andhuman salivary glands and lungs, but the physiological role of DLADremained unidentified. CAD, caspase activated DNase, is a neu-tral DNase, and is responsible for the cell-autonomous DNA degra-dation in apoptotic cells [6]. An analysis by real-time PCR indi-cated that the CAD mRNA was expressed in various mouse tissues,but not in the lens (data not shown). The DNase II mRNA was alsoubiquitously expressed, but its expression level in the lens was low.On the other hand, the DLAD mRNA was highly expressed in themouse lens (Fig. 1a). All the lens fiber cells are nucleated at thebeginning of mouse embryogenesis, and begin to lose their nucleiafter embryonal day (E)16. Accordingly, the DLAD mRNA levelin the lens was low at E14.5, but highly increased at the late stageof mouse embryogenesis (Fig. 1b). The DLAD mRNA was detect-ed in human lens cells from patients with senile cataracts. Anepithelial cell line established from human lens also expressed asignificant level of DLAD mRNA. But, little expression of DLADmRNA was found in the other types of cell lines.

Fig. 1. Expression of DLAD mRNA in the lens. a and b, The DLAD mRNA level in the indicated mouse tissues (a), and in the mouse lens at theindicated developmental stages (b) was quantified by real-time PCR, and is expressed as a relative value to β-actin mRNA.


Establishment of DLAD-deficient miceTo assess the physiological role of DLAD, we generated DLAD-/-

mice by gene targeting. The murine DLAD gene is encoded by 6exons within 15.0 kb of genomic DNA. A targeting vector wasconstructed by replacing exons 3 and 4 with the neo gene, and intro-duced into mouse embryonal stem (ES) cells. ES clones carryingthe mutation were identified by PCR, and mice weregenerated from one representative ES clone. The het-erozygous animals were then intercrossed, and themice carrying the wild-type, hetrozyogous, andhomozygous mutations in the DLAD gene wereobtained. To examine the existence of acid DNasesother than DLAD in the mouse lens, and to confirmthe null-mutation of the DLAD allele in DLAD-/-

mice, cell extracts were prepared from the lenses, andthe DNase activity was assayed. As shown in Fig. 2,the lens extracts from wild-type and heterozygous miceshowed a DNase activity at pH 5.9. On the otherhand, little DNase activity was detected under theseacidic conditions in the lens extracts from DLAD-/-

mice. These results indicated that DLAD is the majoracid DNase in the mouse lens, and the DLAD-/- alleleis a null-allele for DLAD. The three genotypes(DLAD+/+, DLAD+/- and DLAD-/-) were representedaccording to Mendelian inheritance. Except for thecataract described below, DLAD-/- mice showed nogross developmental abnormalities, and were fertilewith normal fecundity.

Development of cataract in DLAD-deficient miceThe eyeballs of DLAD-/- mice had normal mor-

phology and were similar in size to the eyes ofDLAD+/+ mice. The lens fiber cells were well differ-entiated into mature cells, and normal levels of crys-tallins were found in the lens of DLAD-/- mice (datanot shown). Histochemical analyses of midsagittal sec-tions of the lens showed a monolayer of cuboidal epithe-

lial cells covering the anterior surface of the lens. Elongating fibercells at the equator and outer cortex of the lens contained nucleithat stained with Feulgen and DAPI (Fig. 3a). They also containedmitochondria because the cells in this area could be stained by anantibody against mitochondrial ATP synthase. In the inner cortexof the DLAD+/+ and DLAD+/- lens, there was an abrupt loss of nuclei,and mitochondria, which formed an “organelle-free zone”. In con-trast, Feulgen and DAPI-positive materials were found even in themature fiber cells located in the nucleus of DLAD-/- lens, althoughthese cells seemed to be well differentiated morphologically.Analyses of the mature lens cells by transmission electronmicroscopy confirmed that the mature DLAD-/- fiber cells did notcontain ER or mitochondria (Fig. 3b). But they contained materi-als that could be condensed DNA. A higher magnification showedthat the condensed materials were associated by loosely-packedsubstances, and some lens fiber cells had amorphous materials thatcould be unpacked DNA. These materials were not surrounded bymembrane-like structures, suggesting that the nuclear membraneswere destroyed.

The eyes of DLAD-/- mice showed a weak but recognizable cataract(Fig. 4a). That is, there was visible opacity in the nucleus of theDLAD-/- lens, and this was enhanced in aged mice. To examinethe effect of the DNA that was left in the fiber cells on the trans-mission of light, electroretinograms (ERGs) were compared between

Fig. 2. No acid DNase activity in the eye extracts of DLAD-/- mice. Plasmid DNA was incu-bated under acid conditions with the cell extracts (15, 45, and 135 µg) from DLAD+/+,DLAD+/-, or DLAD-/- lens, and analyzed by agarose gel electrophoresis. In lane C, DNA wastreated without the cell extracts. In lanes I and II, the DNase activities of bovine pancreasDNase I and porcine spleen DNase II were assayed.

Fig. 3. No breakdown of nuclear DNA in DLAD-/- lens fibers. a, Midsagittal lens slices prepared fromDLAD+/-and DLAD-/- mice were stained with Feulgen (left panels) and the antibody against mitochondrial ATPsynthase (red) followed by DAPI staining (blue)(right panels). Ep, epithelium; OZF, organelle-free zone; AS,anterior suture; PS, posterior suture. b, The lens fiber cells of DLAD-/- mice were analyzed by electron trans-mission microscopy. Unpacked materials spreading from the dense bodies are indicated by arrowheads.

DiscussionIn mammals, erythrocytes and lens fiber cells have

no nuclei; they are removed during the differenti-ation into mature cells. We previously reported thatthe DNA of the nuclei expelled from erythroid pre-cursor cells was cleaved by DNase II inmacrophages after the nuclei were engulfed by themacrophages [5]. Here, we showed that DLAD, aDNase II-like molecule, is responsible for the degra-dation of DNA in the lens fiber cells. As found forDNase II, DLAD is active under acid conditionsand seems to be present in lysosomes [3]. However,unlike the DNase II-null mice in which undigest-ed DNA from erythroid precursor cells accumulatesin the lysosomes of macrophages [5, 6], the undi-gested DNA was found in the fiber cells of theDLAD-/- lens, indicating that DLAD works cell-autonomously to degrade the DNA of fiber cells.This agrees with the fact that the lens is composedonly of fiber cells, and no macrophages are presentin the lens. Lens fiber nuclei often become TUNEL-positive as they degenerate, suggesting that apop-totic process is involved in this process [7, 8]. Thisapparently contradicts with the biochemical prop-erty of DLAD that generates DNA fragments car-rying 3’-phosphate [3]. It is possible that undersome circumstances, phosphatase(s) may remove3’-phosphate from the DLAD-generated DNA frag-ments. Degradation of organelles is often mediat-ed by its own lysosomes through a process calledautophagy [9]. The essential role of DLAD, a lyso-somal DNase, in the degradation of DNA in the lensfiber cells suggest the involvement of autophagicprocess in the lens cell differentiation. Finally, thedevelopment of cataract in the DLAD-/- mice indi-

cates that DNA must be degraded to ensure the light path in thelens. DLAD is expressed in human lens cells as well. It is possi-ble that some human cataract patients carry a deficiency in the DLADgene, and the DLAD-/- mice established here will provide a goodmodel system for human cataract.

45



the DLAD+/- and DLAD-/- mice. As shown in Fig. 4b, when ERGswere recorded from wild-type mice under dark-adapted condi-tions, typical responses composed of a- and b-waves were observed.In DLAD-/- mice, the responses to the light flash were severelyreduced, indicating that the light path was blocked in the DLAD-/-

lenses.

References

1. McAvoy, J. W., Chamberlain, C. G., de Iongh, R. U., Hales, A. M. & Lovicu, F.J. 1999.Lens development. Eye 13, 425-437.

2. Bassnett, S. 2002.Lens organelle degradation. Exp. Eye. Res. 74, 1-6.

3. Shiokawa, D. & Tanuma, S. 1999.DLAD, a novel mammalian divalent cation-independent endonuclease with homology to DNase II. Nucleic Acids Res. 27,4083-4089.

4. Krieser, R. J., MacLea, K. S., Park, J. P. & Eastman, A. 2001.The cloning, genom-ic structure, localization, and expression of human deoxyribonuclease IIbeta. Gene269, 205-216.

5. Kawane, K., Fukuyama, H., Kondoh, G., Takeda, J., Ohsawa, Y., Uchiyama, Y.& Nagata, S. 2001.Requirement of DNase II for definitive erythropoiesis in themouse fetal liver. Science 292, 1546-1549.

6. Kawane, K., Fukuyama, H., Yoshida, H., Nagase, H., Ohsawa, Y., Uchiyama,Y., Iida, T., Okada, K. & Nagata, S. 2003.Impaired thymic development in mouseembryos deficient in apoptotic DNA degradation. Nat. Immunol. 4, 138-144.

7. Wride, M. A. & Sanders, E. J. 1998.Nuclear degeneration in the developing lensand its regulation by TNFalpha. Exp. Eye Res. 66, 371-383.

8. Dahm, R., Gribbon, C., Quinlan, R. A. & Prescott, A. R. 1998.Changes in thenucleolar and coiled body compartments precede lamina and chromatin reorga-nization during fibre cell denucleation in the bovine lens. Eur. J. Cell Biol. 75,237-246.

9. Klionsky, D. J. & Ohsumi, Y. 1999.Vacuolar import of proteins and organellesfrom the cytoplasm. Annu. Rev. Cell Dev. Biol. 15, 1-32.

Fig. 4. Development of cataract in DLAD-/-mice. a, Nuclear cataract in DLAD-/- mice. The eyes of DLAD+/- andDLAD-/- mice at the ages of 6 and 26 weeks were photographed. b, Electroretinograms in response to diffusestimuli of increasing illumination intensities were recorded with 11-week-old DLAD+/- and DLAD-/- mice. The num-bers on the left indicate the values of the neutral density (ND) filter in log units.

46

Biology


Zinc Transporter LIVI Controls Epithelial-Mesenchymal Transition in Zebrafish Gastrula OrganizerPaper in journals: this is the first page of a paper published in Nature.[Nature] 429, 298-302 (2004)

� Reprinted by permission from Nature 429, 227-326, (2004). Copyright: Macmillan Magazines Ltd.

47




Zinc Transporter LIVI Controls Epithelial-Mesenchymal Transitionin Zebrafish Gastrula OrganizerYAMASHITA Susumu and HIRANO Toshio(Graduate School of Frontier Biosciences and Graduate School of Medicine)

Introduction

During embryonic development, organ and tissue regeneration,and cancer progression, epithelial cell subpopulations active-

ly downregulate cell-cell adhesion systems, and leave their “localneighborhood” to move into new microenvironments. This regu-lated phenotypic modulation including cell-cell dissociation andcytoskeleton remodeling, is called epithelial-mesenchymal transi-tion (EMT). The Snail/Slug family zinc-finger transcription fac-tors, master regulators of EMT, are responsible for the transcrip-tional repression of epithelial genes, and for the transcriptional activationof mesenchymal genes [1]. The transcriptional activity of Snail isregulated by its intracellular location [3], suggesting that the reg-ulatory mechanisms for Snail location are present. The signal trans-ducer and activator of transcription (STAT) molecules mediate bio-logical actions in response to cytokines and growth factors andplay critical roles in EMT during gastrulation, organogenesis,wound healing, and cancer progression, however, the molecularmechanisms of STAT’s action in EMT are unknown [3,4]. In thisreport, we identified LIV1, a breast cancer associated zinc trans-porter protein, as a downstream target of STAT3 that is essentialand sufficient for STAT3’s cell-autonomous role in the EMT ofzebrafish gastrula organizer cells. Furthermore, we demonstratedthat LIV1 is essential for the nuclear localization of Snail. Theseresults establish a molecular link among STAT3, LIV1, and Snailin EMT.

STATs in EMTSTATs mediate biological actions, such as cell proliferation, dif-

ferentiation, and survival, in various biological processes. In addi-tion, STATs play a crucial role in cell migration during morpho-genetic processes. Mouse Stat3 is essential for the gastrulation, andfor the keratinocyte migration in skin-wound healing. TheDrosophila JAK/STAT pathway functions in border cell migra-tion during oogenesis, and the establishment of planar polarity dur-ing eye development. Dictyostelium STAT is required for normalchemotaxis during early development, and for the correct move-ment of prestalk cells during terminal differentiation. These obser-vations raised the possibility that the role of STAT signaling inmorphogenetic movements is conserved throughout evolution [4].

In the zebrafish embryo, the activation of STAT3 is observedin the dorsal organizer and is regulated by the maternal Wnt/β-

catenin pathway, which induces the dorsal organizer. A gp130-related molecule and JAKs are involved in this activation, sug-gesting that the targets of the maternal Wnt/β-catenin pathway thatactivates STAT3 may be an IL-6 family cytokine. The downreg-ulation of STAT3 function with antisense morpholino oligonucleotidesimpairs the anterior movement of the organizer cells and the dor-salward convergence of neighboring cells, so that at the end of gas-trulation the head is mispositioned and the embryonic axis is dra-matically shortened. The activation of STAT3 is restricted to theorganizer region during gastrulation; thus, STAT3 is required forthe anterior movement of the organizer cells in a cell-autonomousmanner and the non-cell-autonomous attraction of the neighbor-ing cells, but does not have a significant effect on cell-fate speci-fication (Fig. 1) [5].

Fig. 1 Zinc transporter LIV1 is a target of STAT3 in zebrafish gastrula organizer cells andcontrols cell movement during gastrulation. Activation of STAT3 (pSTAT3) and expressionof LIV1 (LIV1) are detected in gastrula organizer cells. STAT3 is essential for the activemigration of organizer cells and for the recruitment of neighboring cells. LIV1 plays crucialrole in the organizer cells for its active migration.


STAT3 controls EMT through Zn transporter LIV1To understatnd the molecular mechanism of STAT3 function

in EMT, we performed a subtraction screening using normal andSTAT3-depleted zebrafish embryos. The cDNA we isolated waszebrafish LIV1 (LZT-Dr3), which belongs to a subfamily of ZIPzinc transporters (Zrt-, Irt-like Proteins), termed LZT (LIV-1 sub-family of ZIP zinc Transporters), and had been implicated inmetastatic progression of breast cancer [6]. The zebrafish LIV1mRNAwas expressed in the gastrula organizer cells, in which STAT3 isactivated. In STAT3-depleted embryos, the expression of LIV1mRNAin the gastrula organizer was completely abolished, indicating thatLIV1 is a downstream target of STAT3. Loss of function analysisof LIV1 using antisense morpholino highlighted the phenotypicalsimilarity to STAT3-depleted embryos, which showed severe defectin cell movements during gastrualtion. LIV1-depleted embryos dis-played a mispositioned head and shortened anterior-posterior axisat the end of gastrulation, and all marker genes were expressed,but the expression domain of these genes was vegetally misposi-tioned. These results indicated that LIV1 plays essential roles dur-ing gastrulation by affecting cell movements, without significant-ly altering the early cell-fate specification (Fig. 1). To further clarifyhow LIV1 affects cell movements during gastrulation, we performedcell-tracing experiments. In the LIV1-depleted embryos, the ante-rior movement of the marked organizer cells was perturbed, how-ever, marked cells in the lateral blastoderm margin moved nor-mally, clearly showing that LIV1 is essential for the active migrationof organizer cells, but not for the dorsal convergence of non-axialmesodermal cells. These observations suggested that LIV1 is anessential target gene of STAT3 that is required for STAT3’s cell-autonomous, but not non-cell-autonomous, roles. Actually, LIV1mRNA rescued the cell-autonomous defect in the anterior migra-tion of the STAT3-depleted organizer cells. This specific role ofLIV1 was further indicated by the fact that LIV1 mRNA could notrescue the failed non-cell-autonomous role of STAT3 in the gas-

trula organizer. All these data established that LIV1 is essential andsufficient to carry out STAT3’s cell-autonomous role, but not itsnon-cell-autonomous role, and suggested that LIV1 is a key mol-ecule to regulate the mobility of organizer cells (Fig. 2).

LIV1 is essential for the nuclear translocation of Snail, a mas-ter regulator of EMT

To clarify how LIV1 regulate the mobility of organizer cells,we next analyzed the migratory behaviour of organizer cells in LIV1-depleted embryos. During gastrulation, the organizer cells in nor-mal embryos actively downregulate cell-cell adhesion systems, andleave their local neighborhood to migrate individually, resultingin the body axis being fully extended anteriorly at the end of gas-trulation. However, in LIV1-depleted embryos, the organizer cellscould not downregulate their association, resulting in severe per-turbation of the active migration of organizer cells. Similar defectin the migratory behaviour of organizer cells was also seen inSnail-depleted embryos. Although the expression of LIV1 andSnail1 are independently regulated in gastrula organizer cells, bothSnail and LIV1 repressed transcription from the reporter constructunder the control of the human E-cadherin promoter, and LIV1enhanced the repressor activity of Snail in a dose-dependent man-ner. Importantly, the effect of LIV1 was completely abolished inSnail-depleted zebrafish embryos, clearly showing that LIV1 activ-ity is dependent on Snail. Since activity of the Snail transcription-al factor is regulated by its intracellular location, we examined ifLIV1 regulates subcellular location of Snail. In control embryos,ectopically expressed Snail accumulated in the nuclei of organiz-er cells. However, in LIV1-depleted embryos, Snail protein showedextranuclear localization in organizer cells, indicating that LIV1is essential for nuclear localization of Snail protein in organizercells. Together, these results provide evidence that LIV1 regulatesthe nuclear translocation of Snail and involves in the induction ofthe EMT of organizer cells during zebrafish gastrulation (Fig. 3).

Fig. 2 Dependence of cell-autonomous role of STAT3 in gastrulation movements onLIV1 activity. Organizer cells (green or red) from morpholino- (block) and mRNA- (italics)injected donors were transplanted into the organizer region of a host at the early gastrulastage. Panels show the hosts at the end of gastrulation. Arrows indicate leading edge oftransplanted anterior migrating cells, and arrowheads indicate initial transplant position.

49



ConclusionOur results establish that LIV1, a downstream target of STAT3,

plays a critical role in EMT of gastrula organizer cells by affect-ing Snail activity (Fig. 4). Previous work has indicated roles forthe Snail/Slug family of transcription factor in EMT during can-cer progression and gastrulation [1]. Although zebrafish Snails areexpressed in both axial and non-axial mesendoderm, active ante-rior migration is observed only in the anterior axial mesendoderm,indicating the involvement of additional unknown mechanisms inactive migration. The STAT3-dependent expression of LIV1 specif-ically in the anterior axial mesendoderm and the requirement forLIV1 for normal EMT are consistent with the co-operative rolesof Snail and LIV1 in the invasive behavior of organizer cells dur-ing gastrulation. A similar mechanism might contribute to cancerprogression, since LIV1 and Snail are reported to be involved inthe metastatic spread of breast cancer cells [7,8] and STAT3 isconstitutively activated in many tumors, including breast cancer[9], although this possibility remains to be clarified.

References

1. Thiery, J. P. Epithelial-mesenchymal transitions in tumour progression. Nat RevCancer 2, 442-54 (2002).

2. Dominguez, D. et al. Phosphorylation regulates the subcellular location and activ-ity of the snail transcriptional repressor. Mol Cell Biol 23, 5078-89 (2003).

3. Darnell, J. E., Jr. STATs and gene regulation. Science 277, 1630-5 (1997).

4. Yamashita, S. & Hirano, T. STATs in cell mobility and polarity during morpho-genetic movement; IL-6/gp130/JAK/STAT pathway is evolutionally conserved.in Signal Transducers and Activators of Transcription (STATs): Activation andBiology (eds. Sehgal, P. B., Levy, D. E. & Hirano, T.) 595-607 (Kluwer Acade-mic, Dordrecht, 2003).

5. Yamashita, S. et al. Stat3 Controls Cell Movements during Zebrafish Gastrula-tion. Dev Cell 2, 363-75 (2002).

6. Taylor, K. M. & Nicholson, R. I. The LZT proteins; the LIV-1 subfamily of zinctransporters. Biochim Biophys Acta 1611, 16-30 (2003).

7. Blanco, M. J. et al. Correlation of Snail expression with histological grade andlymph node status in breast carcinomas. Oncogene 21, 3241-6 (2002).

8. Manning, D. L. et al. Oestrogen-regulated genes in breast cancer: association ofpLIV1 with lymph node involvement. Eur J Cancer 30A, 675-8 (1994).

9. Bowman, T., Garcia, R., Turkson, J. & Jove, R. STATs in oncogenesis. Onco-gene 19, 2474-88 (2000).

Fig. 3 Zinc transporter LIV1 is essential for the nuclear translocation of zinc finger proteinSnail and affecting Snail transcriptional activity required for EMT. Morphology of organiz-er cells in the anteriormost limit of the hypoblast layer at mid-gastrula (8 hpf) in control, LIV1-depleted embryos. LIV1-dependent nuclear translocation of Snail in zebrafish gastrula orga-nizer cells. Panels show Snail-GFP in control gastrula organizer cells, and in LIV1-depleted gastrula organizer cells.

Fig. 4 Epithelial-Mesenchymal TransitionDuring the EMT, cell-cell adhesion structures such as desmosomes and adherens junctions in epithelial cells are loosened. The cells become individualized, spread, migrate, and expressmesenchymal characteristics. STAT3 and MAPK independently induces the expression of LIV1 and Snail, respectively. Zinc transporter LIV1 is essential for the nuclear translocation ofzinc finger protein Snail and affecting Snail transcriptional activity required for EMT.

humanities & social sciences - osaka university · duction which initiated the transformation...

Documents