JOURNAL OF EXPERIMENTAL ZOOLOGY 284:379–391 (1999)

© 1999 WILEY-LISS, INC.

Distinct Parameters Are Involved in Controlling theNumber of Rounds of Cell Division in Each TissueDuring Ascidian Embryogenesis

ATSUKO YAMADA AND HIROKI NISHIDA*Department of Life Science, Tokyo Institute of Technology, Nagatsuta,Midori-Ku, Yokohama, 226-8501, Japan

ABSTRACT We counted cell numbers during embryogenesis of the ascidian, Halocynthia roretzi,every hour. Cell numbers were determined by counting the numbers of nuclei in squashed em-bryos. The cell number of a larva just after hatching was approximately 3000. Our study ad-dresses the question of what factors control the number of rounds of cell division duringdevelopment. Three kinds of egg fragments were prepared by cutting unfertilized eggs to alter thevolume of cytoplasm and the amount of DNA. After the egg fragments were fertilized, the cellnumbers were estimated at the hatching stage. The cell numbers of the resulting larvae differedfrom those of normal larvae. Precursor blastomeres of various tissues were then isolated fromnormal and manipulated embryos, and cultured as partial embryos. The cell numbers of the re-sulting partial embryos were counted to estimate the number of cell divisions in each larval tis-sue. The results suggested that the number of cell divisions is controlled by a distinct mechanismin each tissue. We propose that the number of rounds of cell division during ascidian embryogen-esis is controlled by three mechanisms: the first depending on the volume of cytoplasm; the secondon the nucleo–cytoplasmic ratio; and the third depending on neither of these parameters. J. Exp.Zool. 284:379–391, 1999. © 1999 Wiley-Liss, Inc.

Multicellular organisms originate from a singlecell, the fertilized egg, which divides a fairly con-sistent, predetermined number of times duringdevelopment. How the developmental fates of em-bryonic cells are determined has been studied in-tensively. The discoveries of factors such as cyclinshave revealed the molecular mechanisms involvedin the cell cycle. However, little is known abouthow the number of rounds of cell division is con-trolled. In Drosophila embryogenesis, string, a ho-mologue of the yeast Schizosaccharomyces pombecdc25 gene, is required for mitosis early in devel-opment and is zygotically transcribed in a dynamicspatiotemporal pattern that is identical to the mi-totic pattern (Edgar and O’Farrell, ’89, ’90; Edgaret al., ’94). This suggests that string is a positiveregulator of cell division. The expression of stringgene may control when cells cease to divide. In alater stage, Drosophila cyclin E is down-regulated,and dacapo, which is a member of the p21/p27family of cdk inhibitors, is up-regulated in epi-dermis cells when they exit from the mitotic cycleafter division 16. Cyclin E overexpression afterthe terminal division results in an additional cellcycle, whereas premature dacapo expression in-duces a precocious cell cycle arrest (Knoblich et

al., ’94; de Nooji et al., ’96; Lane et al., ’96). Theprecise time when cells stop dividing may be de-termined by cdk inhibitors in concert with declin-ing levels of positive regulators.

Ascidians (Urochordata) are primitive chordates.An ascidian larva is composed of a relatively smallnumber of cells. The first report of the cell num-bers of a larval solitary ascidian was presented byMonroy (’79). He estimated that the epidermis, ner-vous system, mesenchyme, muscle, notochord, andendoderm consist of 800, 250, 900, 40, 40, and 500cells, respectively. The cell number totals 2530.Nicol and Meinertzhagen (’91) reported that thecell number of the larval central nervous systemin Ciona intestinalis is about 370. In Halocynthiaroretzi, muscle and notochord consist of just 42 and40 cells, respectively (Nishida and Satoh, ’85;Nishida, ’87). Thus, the cell number differs in each

Grant sponsor: Japan Society for Promotion of Science for YoungScientists; Grant number: 5057; Grant sponsor: Japan Society forthe Promotion of Science; Grant number: 96L00404.

*Correspondence to: H. Nishida, Department of Life Science, To-kyo Institute of Technology, Nagatsuta, Midori-Ku, Yokohama, 226-8501, Japan.

Received 23 September 1998; Accepted 6 January 1999.

380 A. YAMADA AND H. NISHIDA

tissue. These data and cell lineage data indicatethat the number of cell divisions also differs ineach tissue. For example, total number of roundsof cell division from the first cleavage to the larvalstage is nine in muscle and notochord, while inmesenchyme it is 14 or 15.

When unfertilized eggs of the ascidian are cutinto two equal halves without reference to the po-larity of the eggs and then fertilized, both halvesdevelop into normally shaped larvae (Reverberiand Ortolani, ’62; Yamada and Nishida, ’96). Thelarvae resulting from these egg halves of Ascidiamalaca seemed to possess about 20 notochordcells, that is, half the number generally found ina normal larva (Reverberi and Ortolani, ’62). Thisresult suggests that notochord cells stop dividingwhen they have acquired their definitive size. Insea urchins, the giant embryos resulting from fu-sion of two 32-cell embryos develop into pluteithat consist of twice the normal number of cells.The giant embryos consist of cells of normal size.By contrast, giant eggs, which were occasionallyfound naturally in the ovary of Paracentrotuslividus, and were double the size in both cyto-plasm and nucleus, developed into plutei whichconsisted of a normal number of cells of doublesize (Hörstadius, ’70). These results suggest thatsea urchin eggs divide a definitive number oftimes starting from fertilization, unlike the noto-chord cells of ascidians, as pointed out by Taka-hashi and Okazaki (’79). Cell lineages duringascidian embryogenesis are invariant. Thereforeit is interesting to speculate on how the numberof rounds of cell division is regulated in ascidianembryos.

In this study, we counted cell numbers duringthe embryogenesis of H. roretzi. Then, we allowedthe fragments of an unfertilized egg from whichthe nucleus had been removed, or for which theamount of egg cytoplasm had been reduced to half,to develop and counted the cell numbers of theresultant larvae. In the last series of experiments,precursor blastomeres of various tissues were iso-lated from normal and operated embryos and cul-tured as partial embryos until hatching stage. Thecell numbers of the resultant partial embryos werecounted to estimate the number of cells in eachlarval tissue. We propose that the number ofrounds of cell division is controlled by three dis-tinct mechanisms during ascidian embryogenesis.The present study of the control of cell numbersduring ascidian embryogenesis is the first suchdemonstration using animals that exhibit deter-minate development.

MATERIALS AND METHODSHandling of embryos

Adult ascidians, Halocynthia roretzi, were col-lected near Asamushi Marine Biological Station,Aomori, and Otsuchi Marine Research Center,Iwate, Japan. They were kept in laboratoryaquaria supplied with circulating seawater at 8–9°C. Naturally spawned eggs were translucentand about 280 µm in diameter. Artificially fertil-ized eggs were reared in Millipore-filtered (poresize, 0.45 µm) seawater that contained 50 µg/mlstreptomycin sulfate and 50 µg/ml kanamycin sul-fate (MFSW) at 11–12°C. At 12°C, tadpole larvaehatched about 36–39 hr after fertilization.

Egg fragmentsUnfertilized eggs were chemically devitelli-

nated with seawater containing 0.05% actinaseE and 1% sodium thioglycolate at a pH of about10. To identify the polarity of unfertilized eggs,a method previously described by Yamada andNishida (’96) was used. In brief, when de-vitellinated eggs were stained with a solution of0.005% Nile Blue B, the animal pole of the eggwas less intensely stained than the rest of theegg. With the animal pole as a reference, threekinds of fragments, namely, animal-half, vegetal-half, and non-nucleate fragments (see Fig. 3A)were made by cutting eggs into pieces with afine glass needle under a stereomicroscope (SZH-10, Olympus, Tokyo, Japan). In H. roretzi, na-ked eggs without vitelline membrane were rarelyfertilized (Hoshi et al., ’81). To fertilize egg frag-ments without vitelline membrane, sperm wereactivated by treating them with alkaline seawa-ter (about pH 9) before use (Yamada and Nish-ida, ’96). The fertilization of devitellinated eggfragments with activated sperm asynchronously oc-curred. The time of the first cleavage was moni-tored for each egg fragment and the time after thefirst cleavage was used to know the developmentalstage of each fragment. Fertilized egg fragmentswere reared in agar-coated plastic dishes filled withMFSW. When embryos were reared to later stagesthan gastrula, naked embryos were cultured in su-pernatant of a homogenate of cleaving embryos tofacilitate normal formation of a neural tube (Nishidaand Satoh, ’85).

Identified blastomeres were isolated from nor-mal embryos and embryos from egg fragmentswith a fine glass needle at the 64- and 110-cellstages. Isolated blastomeres were cultured sepa-rately as partial embryos.

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 381

Cell countsDevitellinated whole embryos were extracted

and then stained with DAPI (4´, 6-diamidino-2-phenylindole dihydrochloride), a fluorescent dyespecific for DNA, to count the numbers of nuclei.Two kinds of extraction buffers were modified fromSawada and Schatten (’88). Extraction buffer 1,that is, 50 mM MgCl2, 10 mM KCl, 10 mM EGTA,2% Triton X-100, 20% glycerol, and 25 mM imi-dazole, pH 6.9 was used for the extraction of earlyembryos (the 32-cell stage to early tailbud stage).Extraction buffer 2, which was modified from ex-traction buffer 1 by decreasing the concentrationof MgCl2 from 50 mM to 10 mM, was used to ex-tract late embryos (the early tailbud stage to lar-val stage). Extraction buffer 1 had a stronger effecton extraction than extraction buffer 2. Every hourduring embryogenesis, embryos without test cellsand vitelline membranes were transferred to ex-traction buffers. After extraction for 1 hr, embryoswere washed with distilled water and then stainedwith 0.1 µg/ml DAPI in water for 20 min. Theywere rinsed with distilled water again. Stainedembryos were then squashed on glass slides bycompressing them with cover slips until the con-stituent cells of a specimen were spread into amonolayer, and photomicrographs of them weretaken (Fig. 1). The numbers of nuclei in the pho-tographs were counted with each dividing nucleusbeing scored as two.

Partial embryos were fixed for l0 min in metha-nol at –20°C and stained with DAPI. The numbersof nuclei were counted under a fluorescence micro-scope without squashing, or from photographs.

Immunohistochemical staining withmonoclonal antibodies

The monoclonal antibody 5F1D5 recognizes a no-tochord-specific antigen, Not-l (Nishikata and Satoh,’90). At the middle tailbud stage, this antibody ishighly specific for notochord cells (Nakatani and

Nishida, ’94). Therefore, embryos were fixed andimmunostained at the middle tailbud stage. Indi-rect immunohistochemical staining was carried outby standard methods using fluorescein isothio-cyanate-conjugated (FITC-conjugated) second anti-bodies. 5F1D5 antibody was kindly provided by Dr.T. Nishikata (Konan University, Japan).

RESULTSCell numbers during ascidian

embryogenesisThe numbers of embryonic cells were counted at

1-hr intervals at 12°C. Embryos and larvae wereextracted and stained with DAPI, and numbers ofnuclei in the squashed embryos were counted onphotographs (Fig. 1). The results are shown in Fig.2. At 12°C, gastrulation began at about 10 hr afterinsemination (Fig. 2D) and the neural plate stagewas at about 13 hr (Fig. 2F). The number of cellsconstituting a neurula embryo was 500–1000 (Fig.2G–I). An early tailbud at about 19 hr (Fig. 2J)consisted of approximately 1100 cells. Some tad-pole larvae began to hatch at about 36 hr afterinsemination and almost all had completed hat-ching at 39 hr. The number of larval cells wasapproximately 2800 just after hatching. The fre-quency of cell division per hour decreased as theembryos developed (Fig. 2B). Most of the cellsceased dividing by 25 hr. At the hatching stage onlythree percent of constituent cells divided.

Cell numbers of larvae derivedfrom egg fragments

To investigate whether the number of roundsof cell division is affected by the amount of DNAand volume of egg cytoplasm, egg fragments wereprepared and allowed to develop. Three kinds ofegg fragments, namely, animal-half, vegetal-half,and non-nucleate fragments were made by cut-ting unfertilized eggs using the animal pole as areference (Fig. 3A). For convenience, we designate

Fig. 1. A newly hatched ascidian larva stained with DAPI and squashed. (A) A whole larva.The boxed-in region is enlarged in (B). Scale bars, 100 µm.

382 A. YAMADA AND H. NISHIDA

the amount of DNA after fertilization, and the vol-ume of cytoplasm in an intact egg, as 2n and 1.The amount of DNA in animal-half, vegetal-half,and non-nucleate fragments is represented as 2n,n, and n, respectively. The volume of cytoplasm is

represented as 1/2, 1/2, and 1 (Fig. 3A). Thenucleo-cytoplasmic (N/C) ratio is equivalent to theratio of DNA to egg volume, and is representedas 1 in normal egg. When these fragments werefertilized, they divided in normal schedule at least

Fig. 2. Cell numbers during embryogenesis of H. roretzi.(A, B) The number of cells was determined by counting thenumbers of nuclei every hour. The average numbers and thestandard deviations (vertical bars) from observing about 10specimens are plotted linearly (A) and semilogarithmically(B). At 12°C, larvae hatched between 36 and 39 hr after theinsemination. (C–N) Morphology of embryos was photo-graphed at the time indicated at the top of A and B. (C, D)

Vegetal view. (E–J) Dorsal view. Anterior is to the left. (C)Early gastrula at 10 hr after insemination. (D) Gastrula at11 hr. (E) Gastrula at 12 hr. (F) Neural plate stage at 13 hr.(G) Neurula at 14 hr. (H) Neurula at 16 hr. (I) Neurula at 18hr. (J) Early tailbud at 19 hr. (K) Tailbud at 22 hr, lateralview. Anterior is to the left. (L) Tailbud at 26 hr. (M) Tailbudat 29 hr. Otolith was melanized (arrow). (N) Unhatched larvaat 33 hr. Ocellus was also melanized. Scale bar, 100 µm.

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 383

until the 110-cell stage, and developed into dwarflarvae with a normal shape and structure (Fig.3B, C in this paper; Ishida and Satoh, ’98). Werefer to embryos derived from animal-half, veg-etal-half, and non-nucleate fragments as animal-half, vegetal-half, and non-nucleate embryos. Inthe same way, we use the terms, animal-half, veg-

etal-half, and non-nucleate larvae. The cell num-bers in experimental specimens were countedwhen normal larvae hatched for the following rea-sons. The rate of cell division is low at the hatch-ing stage, ensuring that observational error isminimized. In this experiment, eggs and embryoswere cultured at 11°C.

Fig. 3. The egg fragments and the resultant larvae. (A)Various egg fragments that were prepared by cutting unfertil-ized eggs are shown by shading. The small half circles at thetop of the eggs depict the animal pole. Amount of DNA, vol-ume of egg cytoplasm, and nucleo-cytoplasmic ratio for each

fragment are indicated. Those for a normal egg are representedas 2n, 1, and 1, respectively. Relative ratios of each fragmentare shown using normal egg as the standard. Each fragmentdeveloped into normally shaped larvae. (B) Normal larva. (C)Dwarf larva from a vegetal-half fragment. Scale bar, 100 µm.

384 A. YAMADA AND H. NISHIDA

The results are shown in Fig. 4. The total cellnumber of normal larva was about 3000. The dif-ference in cell number between 3000 in this ex-periment and 2800 observed in Fig. 2 may be dueto the difference of the observational stage. Themean cell numbers of animal-half, vegetal-half,and non-nucleate larvae were about 2000, 2500,and 4100, respectively. There are statistically sig-nificant differences between the cell numbers ofnormal and these operated larvae. Relative num-ber of cells was calculated by using the cell num-ber of normal larva as the standard. The relativecell numbers of the animal-half, vegetal-half, andnon-nucleate larvae were represented as 4/6, 5/6,and 8/6, respectively.

The difference in cell number between animal-half and vegetal-half larvae might be attributedto regional differences in cytoplasm in unfertil-ized eggs along the animal-vegetal axis (Yamadaand Nishida, ’96). To test this, the female nucleuswas removed by cutting off a very small animal-pole fragment from animal-half fragments. In theanimal-half fragment that was enucleated, anamount of DNA after fertilization is representedas n. The fragments developed into larvae with2600 cells, almost the same cell number as veg-etal-half larva. Thus, the regional difference alongthe animal-vegetal axis of unfertilized eggs didnot affect the number of cell division. Half eggswere also prepared by cutting an unfertilized eggat the meridional plane. The half eggs possessingthe female nucleus developed into larvae that con-sisted of 4/6 cells compared with the normal larva,

that is the same as animal-half larva. The halfeggs not possessing the egg nucleus developed intolarvae of 5/6 cells, that is the same as vegetal-half larva (data not shown). The result for halfeggs dissected at the meridional plane supportedthe view that the difference in cell numbers be-tween animal-half and vegetal-half larvae ismerely attributable to a difference in DNA amountor number of chromosomes. Numbers of notochordcells were counted at middle tailbud stage, as eachnotochord cell was visible and countable. In ev-ery experiment, notochord consisted of about 40cells as in normal tailbud (Fig. 4, bottom). Thus,the cell numbers of notochord were constantwhereas total cell numbers of larvae were affectedby bisection of eggs. The number of cell divisionsin some larval tissues, but not notochord, was af-fected by altering the volume of cytoplasm andthe amount of DNA.

Ascidian embryogenesis is known to be a mo-saic development. Unlike regulative embryos suchas echinoderms, each blastomere isolated from 2-cell ascidian embryos develop into right- or left-half larvae. We isolated blastomeres of 2-cellembryos and cultured them as partial embryosuntil hatching stage. The resultant embryos had1300 cells including 17 notochord cells. Therefore,blastomeres isolated at the 2-cell stage developedinto half larvae also in terms of cell number.

Cell numbers of partial embryos derivedfrom isolated tissue precursor blastomeresIt is important to know about cell number in

each tissue of larvae derived from various kindsof egg fragments. However, it was hard to countthem in whole larvae except for notochord. By the110-cell stage, the developmental fates of mostblastomeres become tissue restricted in ascidians(Nishida, ’87, ’92a). Furthermore, isolated blas-tomeres autonomously differentiate in a single tis-sue according to their normal fate (Nishida, ’92b).In the present experiments, fertilized egg frag-ments of various kinds developed into 110-cellembryos with normal appearance through normalcleavage pattern. In other words, every blastomereexists in embryos derived from all kinds of eggfragments. Therefore, precursor blastomeres ofvarious tissues in 64- or 110-cell embryos wereisolated and cultured as partial embryos untilhatching stage. The cell numbers of the resultantpartial embryos were counted to estimate cellnumbers in each larval tissue. Precursor blas-tomeres of five kinds of major tissues were iso-lated. These were endoderm, muscle, notochord,

Fig. 4. The cell numbers of larvae developed from eachfragment. Fragments were fertilized and allowed to developuntil hatching stage. The cell numbers of the resultant lar-vae were counted. The means and standard deviations of thecell numbers are shown. Relative number of cells was calcu-lated using the cell number of normal larva as the standard.Numbers of notochord cells in each larva are shown at thebottom. Numbers of specimens are shown in parentheses.N.D., not determined.

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 385

mesenchyme, and epidermis precursors. Severalrepresentative blastomeres for each tissue werechosen to be isolated.

The results are shown in Figs. 5 and 6 as a his-togram with the ranges of mean of populations cal-culated by interval estimation. Figure 6G–I showsthe expected values when the number of cell divi-sions is controlled by distinct parameters. That is,the N/C ratio, the volume of cytoplasm, and theparameters not affected by our manipulation, areshown. A detailed account of these parameters isgiven in the Discussion. As the precursor cells ofthe same tissue generated almost the same num-ber of cells, their results were combined. Irrespec-tive of the kinds of fragments, endoderm, muscle,and notochord precursor cells all divided the samenumber of times as in normal embryos (Figs. 5A–C and 6A–C, I). The number of cell divisions wasnot affected by the change of cytoplasm volume andDNA amount in these tissues. By contrast, isolatedmesenchyme precursors and epidermis precursors,generated different numbers of cells in differentkinds of fragments. When mesenchyme cells, B7.7cells, were isolated from normal embryos, theyformed about 84 cells whereas those from non-nucleate embryos formed about 86 cells. However,

B7.7 cells of animal-half and vegetal-half embryos,generated about half the number of cells, that is,about 40 and 48, respectively (Figs. 5D, 6D). Thus,it seems that the number of cell divisions dependson the volume of egg cytoplasm (Fig. 6H).

Larval epidermis is derived from 50 cells in the110-cell embryos. Twenty-two cells of the a-linegive rise to epidermis of head and trunk region,while the other 28 cells of the b-line develop intothe tail epidermis. The result for a-line epidermisprecursors (Figs. 5E, 6E) was as predicted in thatthe number of cell divisions is controlled by thevolume of cytoplasm (Fig. 6H), as observed in mes-enchyme cells. By contrast, the result of the b-line cells (Figs. 5F, 6F) differed from all of theexpectations shown in Fig. 6G–I.

These results were classified according to stagesof the cell cycle (Fig. 7). An explanation of theclassification is as follows. When the observed cellnumber is X, X can be expressed as 2n-1 ≤ X < 2n.The partial embryo could be assumed to have un-dergone “n - 1” and “n” times of cell division. Forconvenience, the embryo will be described as be-ing in the ‘‘n’’th generation after isolation. As en-doderm and mesenchyme precursor cells havedivided six times at isolation stages, that is the64- and 110-cell stages, the partial embryo willbe described to be in the “n + 6”th generation. Asmuscle, notochord, and epidermis precursor cellshave divided seven times at isolation stages, theembryo is represented as being in the “n + 7”thgeneration. For example, a partial embryo of en-doderm consisting of 26 cells is in the 11th gen-eration. If the partial embryos derived from everyfragment divided the same number of times with-out being affected by change of cytoplasm volumeand DNA amount, they will be in the same gen-eration as shown in Fig. 7I. Given that the num-ber of cell divisions is controlled by the volume ofcytoplasm, the partial embryos derived from ani-mal-half and vegetal-half embryos stay at onegeneration earlier than those derived from nor-mal eggs and non-nucleate fragments (Fig. 7H).The generation was calculated for each specimenand the numbers of specimens of each genera-tion are shown in Fig. 7A–F. The generation ofpartial embryos of endoderm, muscle, and noto-chord derived from various egg fragments weremainly at 11th, 10th, and 10th, respectively (Fig.7A–C). These results are similar to these expectedif cells divide a definitive number of times regard-less of removal of nuclei and reduction of cyto-plasm (Fig. 7I). When mesenchyme blastomereswere isolated from animal-half and vegetal-half

Fig. 5. The number of cells in partial embryos derivedfrom isolated blastomeres. (A–F) The precursor blastomeresof six kinds of tissue were isolated from 64- or 110-cell em-bryos that were derived from intact egg and fragments. Theywere cultured as partial embryos until hatching stage. Thecell numbers of the resultant partial embryos were counted.The average numbers with the standard deviations are shown.Numbers of specimens are indicated in parentheses. On theposition of the blastomeres in the embryo, see Conklin (’05),Satoh (’79, ’94), and Nishida (’92a).

386 A. YAMADA AND H. NISHIDA

Fig. 6. Histograms showing the cell numbers of the par-tial embryo. The upper end of each column shows the meansof population by point estimation. Moreover, the means werecalculated by interval estimation (vertical bars), which is notidentical to the standard deviation. The confidence intervalis 95%. Under the abscissa, the fragments from which iso-lated blastomeres are derived are shown. The thick horizon-tal broken lines represent the cell numbers of partial embryosderived from normal embryo. The thin and medium brokenlines describe half and double numbers. (A) Endoderm pre-cursor cells generated almost the same number of cells inevery fragment. (B) Muscle precursor cells generated almost

the same number of cells in every fragment. (C) Notochordprecursor cells generated almost the same number of cells inevery fragment. (D) Mesenchyme precursor cells were iso-lated. In animal-half and vegetal-half fragments, the cells gen-erated half the number of cells of normal egg. (E) The a-lineepidermis precursor cells were isolated. (F) The b-line epi-dermis precursor cells were isolated. The expected valueswhen the number of cell divisions is controlled by the N/Cratio, the volume of cytoplasm, and parameters not affectedby our manipulation, are shown in G–I, respectively. (G) TheN/C ratio. (H) The volume of cytoplasm. (I) The parametersnot affected by our manipulations.

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 387

embryos, the resultant partial embryos were inthe 12th generation, that is, one generation ear-lier than normal and non-nucleate embryos (Fig.7D). This suggests that the number of cell divi-sions in mesenchyme was affected by the change

in cytoplasm volume as proposed in the previoussection (Fig. 6D).

Because the partial embryos of the a-line epi-dermis cells changed the cell number when theegg cytoplasm was reduced (Figs. 5E, 6E), it would

Fig. 7. Histograms showing the generations of partial em-bryos. Generations of partial embryos after fertilization wereestimated, as described in text. Relative numbers (as per-centages) of specimens are presented on the vertical axis. Thewhite columns (the leftmost columns) indicate the results ofthe normal eggs. The spotted columns (second from the left),the light gray columns (third from the left), and the darkgray columns (the rightmost columns) show the results ofanimal-half, vegetal-half, and non-nucleate egg, respectively.

(A) Endoderm precursor cells. (B) Muscle precursor cells. (C)Notochord precursor cells. (D) Mesenchyme precursor cells.(E) The a-line epidermis precursor cells. (F) The b-line epi-dermis precursor cells. The expected values when the num-ber of cell divisions is controlled by the N/C ratio, the volumeof cytoplasm, and the parameters not affected by our ma-nipulation, are shown in G–I, respectively. (G) The N/C ra-tio. (H) The volume of cytoplasm. (I) The parameters notaffected by our manipulations.

388 A. YAMADA AND H. NISHIDA

have been expected that the histogram showingtheir generations might be like that in Fig. 7H.However, the histogram (Fig. 7E) differed fromFig. 7H. The partial embryos derived from non-nucleate embryos distributed in the 11th, 12th,and 13th generations, uniformly. Those from veg-etal-half embryos were in not only the 11th butalso the 10th and 12th generations. In the b-lineof epidermis, the result did not coincide with anyof Fig. 7G–I.

To sum up, muscle, notochord, and endodermcells divided a constant number of times withoutbeing affected by the volume of cytoplasm and theamount of DNA. Mesenchyme cell divisions variedwith egg volume. The number of cell divisions inepidermis was affected by the operations, howeverwe could not identify the relevant parameters.

DISCUSSIONCell numbers during ascidian

embryogenesisCell numbers of H. roretzi embryos were counted

during embryogenesis. As shown in Fig. 2, neu-rula consisted of about 500–1000 cells, and earlytailbud and larva just after hatching comprisedabout 1100 and 2800 cells, respectively. The fre-quency of cell divisions decreased as developmentproceeded (Fig. 2B). The cell numbers of larvaealmost coincided with the 2530 reported byMonroy (’79). While Monroy (’79) counted cellnumbers of each tissue of normal larvae, we wereunable to do so except for notochord.

A hypothesis for mechanisms controllingthe number of cell division

Various fragments of unfertilized eggs gave riseto larvae with normal shape and structure. How-ever, the cell numbers of the resultant larvae dif-fered from those of normal larvae. The normalnumber of cells may not, therefore, be a prerequi-site for the correct shaping of larvae. This is sup-ported by data from other systems. For example,in echinoderm, isolated blastomeres from the 2-and 4-cell embryos develop into morphologicallynormal larvae, but the resultant larvae containonly a half and a fourth of the cell number ofwhole larva, respectively (Morgan, 1895; Dan-Sohkawa and Satoh, ’78; Takahashi and Okazaki,’79). Similarly, when DNA synthesis is blocked af-ter the vegetal plate of sea urchin embryos hasthickened, development continues normally untilpluteus formation, even though cell division hasceased (Stephens et al., ’86).

There is a possibility that sperm nuclei inenucleated egg fragments, that is, vegetal-half andnon-nucleated fragments, undergo diploidizationafter fertilization. However, enucleated fragmentsdeveloped in normal schedule without additionalDNA-doubling times throughout cleavage stage.Therefore, we assumed that enucleated fragmentsdeveloped as haploid although the chromosomenumbers in the resultant larvae were not checked.

The nucleo-cytoplasmic ratios (the N/C ratios)are proposed as factors governing the timing ofthe transition from synchronous divisions to asyn-chronous divisions (MBT: mid-blastula transition)in amphibian, zebrafish and starfish embryos.Asynchronous cleavages start one division earlierin embryos that have twice the N/C ratio of nor-mal embryos, and later in embryos with half theN/C ratio (Kobayakawa and Kubota, ’81; Newportand Kirschner, ’82a,b; Mita, ’83; Mita and Obata,’84; Kane and Kimmel, ’93). The N/C ratio alsoaffects the duration of mitoses during the blasto-derm divisions in Drosophila (Edgar et al., ’86).In ascidians, cell division becomes asynchronousafter the 16-cell stage (Conklin, ’05). However, thedivision schedule, at least until the 110-cell stage,is not affected by changing the N/C ratio (Rev-erberi and Ortolani, ’62; Ishida and Satoh, ’98;this study). In the present study, it was observedthat the smaller the ratio, the fewer the larvalcell numbers (Figs. 3, 4), although the relationwas not strictly inversely proportional. It is sug-gested that the total number of cells might be con-trolled to some extent by the N/C ratio. On theother hand, the number of mesenchyme cells wasaffected by the volume of cytoplasm, and those ofmuscle, notochord, and endoderm cells were notaffected by the volume of cytoplasm or the N/Cratio. It seems reasonable to assume that larvalcells are composed of three kinds of cells. Thereare cells for which the number of cell divisions isregulated by the N/C ratio, others by the volumeof cytoplasm, and yet others by parameters notaffected by our manipulation. We tried to simu-late our results based on this hypothesis.

Provided that the mechanisms only involve theN/C ratio, the volume of cytoplasm, and parametersnot affected by our manipulation, simulation is notalways successful. The solution can only be obtainedwhen the results satisfy the following equation:

2x – 2y + z = w.x, y, z, and w stand for cell numbers of animal-half, vegetal-half, non-nucleate, and normal lar-vae, respectively.

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 389

The experimental results satisfied this equation.The solution is that, in a normal larva, 1/3 eachof the cells are controlled by the three kinds ofmechanism (Fig. 8). In Fig. 8, each circle repre-sents about 500 cells. The cell number of an ani-mal-half larva is expected to be 4/6 of that ofnormal larva because, in the animal-half frag-ment, the N/C ratio and the volume of cytoplasmare 2 and 1/2 respectively. A vegetal-half larva isexpected to consist of 5/6 cells because of the halfvolume of cytoplasm. A non-nucleate larva is ex-pected to consist of 8/6 cells because of the doubleN/C ratio of normal egg. This simulation coincidedwell with the experimental results (Fig. 4). There-fore, it was suggested that the three distinct kindsof mechanisms are involved in controlling num-ber of cell divisions during ascidian embryogen-esis. The possibility cannot be excluded thatslowdown of the rate of cell division as the resultof overall delay of development is involved in thedifference of cell numbers in the resulting larvaefrom each egg fragment. However, all fragmentsdivided on a normal schedule until the 110-cellstage and developed into larvae at almost sametime as normal larvae. Therefore, we assumedthat overall delay of development of egg fragmentsis not involved in difference in cell number.

We isolated each tissue precursor cell from em-bryos derived from various egg fragments. Mostof the blastomeres that were isolated from nor-mal embryos autonomously developed accordingto their normal fate (Nishida, ’92b). Notochord andprimary muscle precursor cells of 110-cell embryosdivided twice into four cells in normal embryo-

genesis (Nishida, ’87). When they were isolated,the cell numbers of the resultant partial embryoscoincided with those in normal larvae (Nishida,’92b; Fig. 5). There is a possibility that isolatedblastomeres of other tissues do not divide as manytimes as intact embryos. Despite this possibility,it is still significant that cell numbers in the re-sulting partial embryos were compared. Collatingthe experimental results (Fig. 6A–F and Fig. 7A–I) with expectations (Fig. 6G–I and Fig. 7G–I),histograms on endoderm, muscle, and notochordclosely resembled the case in which the numberof cell divisions depends on parameters that werenot affected by our manipulation. The results ofthe blastomere-isolation experiment on notochordand of counting notochord cells in whole larvaewere consistent. The number of cell divisions inmesenchyme seems to be regulated by the mecha-nism that depends on the volume of cytoplasm(Figs. 6H and 7H). The results of the a- and b-lines epidermis did not seem to match any of themechanisms (Figs. 6F, 7F). These results may beexplained by a mixture of two or three mecha-nisms, or by another mechanism. After all, mecha-nisms controlling the number of cell divisions inepidermis are obscure.

The nucleo-cytoplasmic (N/C) ratioAt the MBT in Xenopus, the cells become mo-

tile and transcriptionally active. Newport andKirschner (’82a,b) demonstrated that injection ofDNA, equal to the amount of nuclear DNA pre-sent at MBT, induced premature transcription,and therefore they proposed the titration mod-els. MBT is triggered by the exhaustion of com-ponents present in the unfertilized eggs throughtitration of the components with DNA. In the lar-val development of ascidians, cell division maystop due to the depletion of a factor indispens-able for the division, by titration of the factor byDNA or chromatin. Although the involvement ofthe mechanism dependent on the N/C ratio is sug-gested in cell counts of whole larvae, no candi-dates were found in the blastomere isolationexperiment. Tissues that were not investigatedin this paper, for example, nervous system, mayalso control the number of cell divisions by theN/C ratio.

The volume of cytoplasmIt has been suggested that mesenchyme may

control the number of cell divisions by the vol-ume of cytoplasm. The cell divisions may be lim-ited by cell size and a cell may be unable to

Fig. 8. Illustrations showing the simulation of the cellnumbers. The simulation was performed on the suppositionthat the number of cell divisions was controlled by threemechanisms. When the volume of cytoplasm was half that ofa normal egg, cell divisions decreased, and when the N/C ra-tio was half they increased.

390 A. YAMADA AND H. NISHIDA

become smaller when it reaches a definitive finalsize. This is an attractive idea because cells ofascidian embryos do not grow and mesenchymecells divided most frequently among the tissuecells investigated in this work. However, the fol-lowing observation may rule out this possibility.Mesenchyme cells of larvae give rise to tunic cellsof juveniles (Hirano and Nishida, ’97). Therefore,they would continue dividing to make juvenile tis-sues during metamorphosis, and become smallerthan their size in the larvae. Alternatively, celldivisions may cease because the factors neededto perform the division become depleted. In thiscase, the factors are not titrated by DNA, but areconsumed whenever the cell cycle progresses. Thispossibility can be applied even for the scenarioin which the cells continue dividing. A cell divi-sion may be skipped during embryogenesis afterconsumption of maternal store and until zygoticproducts accumulate. A relevant phenomenon isfound in Drosophila. Maternal String of Droso-phila is exhausted at interphase of cell cycle 14and cells await the zygotical accumulations to re-start subsequent divisions.

Mechanisms not affected by N/C ratio andvolume of cytoplasm

Muscle and notochord cells divided the samenumber of times in all kinds of egg fragments. Ourobservation was different from the results ofReverberi and Ortolani (’62), who reported that halfeggs developed into larvae possessing half the num-ber of notochord cells found in normal larva.

Muscle and notochord cells finish their final celldivisions before the larva hatches (Nishida andSatoh, ’85; Nishida, ’87). There exist mechanismscounting the number of rounds of cell division with-out being affected by our manipulation. The mecha-nisms may involve counting the numbers of DNAreplications, nuclear divisions, or cytokineses. Inmuscle cells, the timing of acetylcholinesterase de-velopment, which is first detected at the neurulastage, has been suggested to be controlled by themechanism counting the cycle of DNA replications(Satoh, ’79; Satoh and Ikegami, ’81a,b). It is pos-sible that the time when a cell stops dividing isalso regulated by the cycle of DNA replications.

Endoderm was classified into this category inthe present study. Because the cells are notpostmitotic and continue to divide after hatching(Nishida and Kumano, ’97), observations at laterstages after hatching may reveal the effect of theN/C ratio and the volume of cytoplasm in endo-derm tissue.

Ascidian as an experimental animalIn ascidians, cell lineages are invariant. An ex-

perimental system in which the number of cell di-visions can be altered by micromanipulations wasdeveloped. From the experiments it appears thatthe mechanisms controlling the number of cell di-visions vary with the type of tissue. These resultswill facilitate the analysis of mechanisms that con-trol cell division and the cell cycle in ascidians.

What controls how many times a cell divides?In many animals, including ascidians, this ques-tion remains to be answered (see review, Raff, ’96).Many regulators of cell cycle progression, for ex-ample the cyclins, Cdc2, and Cdc25, are highlyconserved in a diverse system. In ascidians, thestudy of these regulators may shed light on themolecular mechanisms.

ACKNOWLEDGMENTSThe authors thank Dr. T. Numakunai for sup-

plying live materials, and all other members of theAsamushi Marine Biological Station for facilitat-ing the authors’ work there. Thanks are also dueto the members of the Otsuchi Marine ResearchCenter for supplying live materials. The authorsalso thank Dr. T. Nishikata (Konan University) forproviding monoclonal antibody. A.Y. was supportedby pre-doctoral fellowships from the Japan Societyfor Promotion of Science for Young Scientists witha research grant (5057). This research was also sup-ported by the “Research for the Future Program”from the Japanese Society for the Promotion of Sci-ence (96L00404) to H.N.

LITERATURE CITEDConklin EG. 1905. The organization and cell-lineage of the

ascidian egg. J Acad Sci (Phil) 13:1–119.Dan-Sohkawa M, Satoh N. 1978. Studies on dwarf larvae de-

veloped from isolated blastomeres of the starfish, Asterinapectinifera. J Embryol Exp Morph 46:171–185.

de Nooij JC, Letendre MA, Hariharan IK. 1996. A cyclin-de-pendent kinase inhibitor, dacapo, is necessary for timelyexit from the cell cycle during Drosophila embryogenesis.Cell 87:1237–1247.

Edgar BA, Kiehle CP, Schubiger G. 1986. Cell cycle controlby the nucleo-cytoplasmic ratio in early Drosophila devel-opment. Cell 44:365–372.

Edgar BA, O’Farrell PH. 1989. Genetic control of cell divi-sion patterns in the Drosophila embryo. Cell 57:177–187.

Edgar BA, O’Farrell PH. 1990. The three postblastoderm cellcycles of Drosophila embryogenesis are regulated in G2 bystring. Cell 62:469–480.

Edgar BA, Sprenger F, Duronio RJ, Leopold P, O’Farrell PH.1994. Distinct molecular mechanisms regulate cell cycle tim-ing at successive stages of Drosophila embryogenesis. GenesDev 8:440–452.

Hirano T, Nishida H. 1997. Developmental fates of larval tis-

CONTROL OF CELL DIVISION IN ASCIDIAN EMBRYOS 391

sues after metamorphosis in ascidian Halocynthia roretzi.Dev Biol 192:199–210.

Hörstadius S. 1970. Giant larvae of Paracentrotus lividus.Arkiv Zool 23:417–422.

Hoshi M, Numakunai T, Sawada H. 1981. Evidence for par-ticipation of sperm proteinases in fertilization of the soli-tary ascidian, Halocynthia roretzi: effects of proteaseinhibitors. Dev Biol 86:117–121.

Ishida K, Satoh N. 1998. Quantity of prelocalized maternalfactor is associated with the timing of initiation of an epi-dermis-specific gene expression of the ascidian embryo. DevGenes Evol 208:151–156.

Kane DA, Kimmel CB. 1993. The zebrafish midblastula tran-sition. Development 119:447–456.

Knoblich JA, Sauer K, Jones L, Richardson H, Saint R, LehnerCF. 1994. Cyclin E controls S phase progression and itsdown-regulation during Drosophila embryogenesis is re-quired for the arrest of cell proliferation. Cell 77:107–120.

Kobayakawa Y, Kubota H. 1981. Temporal pattern of cleav-age and the onset of gastrulation in amphibian embryosdeveloped from eggs with the reduced cytoplasm. J EmbryolExp Morph 62:83–94.

Lane ME, Sauer K, Wallace K, Jan YN, Lehner CF, VaessinH. 1996. Dacapo, a cyclin-dependent kinase inhibitor, stopscell proliferation during Drosophila development. Cell87:1225–1235.

Mita I. 1983. Studies on factors affecting the timing of earlymorphogenetic events during starfish embryogenesis. J ExpZool 225:293–299.

Mita I, Obata C. 1984. Timing of early morphogenetic eventsin tetraploid starfish embryos. J Exp Zool 229:215–222.

Monroy A. 1979. Introductory remarks on the segregation ofcell lines in the embryo. In: LeDouarin N, editor. Cell lin-eage, stem cells and cell determination. New York: North-Holland Publishing Company. p 3–13.

Morgan TH. 1895. Studies of the ‘partial’ larvae of Sphae-rechinus. Arch Entw Mech Org 2:81–126.

Nakatani Y, Nishida H. 1994. Induction of notochord duringascidian embryogenesis. Dev Biol 166:289–299.

Newport J, Kirschner M. 1982a. A major developmental tran-sition in early Xenopus embryo: I. Characterization and tim-ing of cellular changes at the midblastula stage. Cell30:675–686.

Newport J, Kirschner M. 1982b. A major developmental tran-sition in early Xenopus embryo: II. Control of the onset oftranscription. Cell 30:687–696.

Nicol D, Meinertzhagen IA. 1991. Cell counts and maps inthe larval central nervous system of the ascidian Cionaintestinalis (L.). J Comp Neurol 309:415–429.

Nishida H. 1987. Cell lineage analysis in ascidian embryosby intracellular injection of a tracer enzyme: III. Up to thetissue restricted stage. Dev Biol 121:526–541.

Nishida H. 1992a. Determination of developmental fates ofblastomeres in ascidian embryos. Dev Growth Differ 34:253–262.

Nishida H. 1992b. Developmental potential for tissue differ-entiation of fully dissociated cells of the ascidian embryo.Roux’s Arch Dev Biol 201:81–87.

Nishida H, Satoh N. 1985. Cell lineage analysis in ascidianembryos by intracellular injection of a tracer enzyme: II.The 16- and 32-cell stages. Dev Biol 110:440–454.

Nishida H, Kumano G. 1997. Analysis of the temporal ex-pression of endoderm-specific alkaline phosphatase duringdevelopment of the ascidian Halocynthia roretzi. DevGrowth Differ 39:199–205.

Nishikata T, Satoh N. 1990. Specification of notochord cellsin the ascidian embryo analysed with a specific monoclonalantibody. Cell Differ Dev 30:43–53.

Raff MC. 1996. Size control: the regulation of cell numbersin animal development. Cell 86:173–175.

Reverberi G, Ortolani G. 1962. Twin larvae from halves ofthe same egg in ascidians. Dev Biol 5:84–100.

Satoh N. 1979. On the ‘clock’ mechanism determining thetime of tissue-specific enzyme development during ascidianembryogenesis: I. Acetylcholinesterase development in cleav-age-arrested embryos. J Embryol Exp Morph 54:131–139.

Satoh N, Ikegami S. 1981a. A definite number of aphidicolin-sensitive cell-cyclic events are required for acetylcholinest-erase development in the presumptive muscle cells of theascidian embryos. J Embryol Exp Morph 61:1–13.

Satoh N, Ikegami S. 1981b. On the ‘clock’ mechanism deter-mining the time of tissue-specific enzyme development dur-ing ascidian embryogenesis: II. Evidence for association ofthe clock with the cycle of DNA replication. J Embryol ExpMorph 64:61–71.

Sawada T, Schatten G. 1988. Microtubules in ascidian eggsduring meiosis, fertilization, and mitosis. Cell Motil Cytoskel9:219–230.

Stephens L, Hardin J, Keller R, Wilt F. 1986. The effects ofaphidicolin on morphogenesis and differentiation n the seaurchin embryo. Dev Biol 118:64–69.

Takahashi M, Okazaki K. 1979. Total cell number and num-ber of the primary mesenchyme cells in whole, 1/2 and1/4 larvae of Clypeaster japonicus. Dev Growth Differ 21:553–566.

Yamada A, Nishida H. 1996. Distribution of cytoplasmic de-terminants in unfertilized eggs of the ascidian Halocynthiaroretzi. Dev Genes Evol 206:297–304.