Retinoic acid and the cyclin dependent kinase inhibitors ...individual.utoronto.ca/paul_hamel/Publications/Dirks1997.pdf · and activity of cell cycle regulatory factors, specifically,

Retinoic acid and the cyclin dependent kinase inhibitors synergistically alterproliferation and morphology of U343 astrocytoma cells

Peter B Dirks1,2, Ketan Patel1, Sherri Lynn Hubbard1, Cameron Ackerley2, Paul A Hamel2

and James T Rutka1,2

1Division of Neurosurgery, Hospital for Sick Children, Toronto, Ontario, M5G 1X8 and 2Department of Pathology, University ofToronto, Toronto, Ontario, M5S 1A8, Canada

We have characterized the expression and activity of thecell cycle regulatory machinery and the organization ofthe cytoskeleton of the p16Ink4a-de®cient astrocytoma cellline, U343 MG-a (U343), following retinoic acid (RA)treatment. RA causes cell cycle arrest at low cell densityand signi®cant morphological changes in U343 cells,re¯ected by reorganization of the intermediate ®lament,GFAP, and actin. RA-induced cell cycle arrest is alsoassociated with induction of p27Kip1 expression, inhibitionof cdk2-associated kinase activity and alteration of thephosphorylation state of the pRB-family proteins. Wenext determined the e�ect of inducing expression of thecyclin dependent kinase inhibitors (CKI's), p16Ink4a,p21Cip1/Waf1 or p27Kip1 on the proliferation and morphologyof these malignant astrocytoma cells in the absence andpresence of RA. Induction of p16, p21 or p27, using thetetracycline repressor system, potently inhibits prolifera-tion of U343 cells. However, rather than resembling RA-treated cells, CKI-induced U343 cells become ¯at withabundant cytoplasm and perinuclear vacuolization. CKI-induced morphological alterations are accompanied by asigni®cant reorganization of glial ®laments within thecytoplasm. Interestingly, when U343 cells are growtharrested by p16, p21 or p27 induction and treatedsimultaneously with RA, a dramatic morphologicalchange occurs, cells acquiring multiple long, taperingprocesses reminiscent of primary astrocytes. Thisrearrangement is accompanied by reorganization ofGFAP, vimentin and actin. Vimentin speci®cally reloca-lizes to the tips of the long processes which form. Thearrangement of intermediate ®laments in these cells is, infact, indistinguishable from their arrangement in primaryhuman astrocytes. These data demonstrate that when astrong proliferative block, produced by CKI expression,occurs in conjunction with the morphogenic signalsgenerated by RA, these p16-de®cient malignant astro-cytoma cells are induced to phenotypically resemblenormal astrocytes.

Keywords: Astrocytoma; cyclins; cyclin-dependentkinases; GFAP; retinoic acid; p16; p21

Introduction

Progression through the G1 phase of the cell cycle iscontrolled, in part, by the regulation of activity ofcyclin-dependent kinases (cdk's) (for review, see

Weinberg, 1995). One important means of regulatingcdk activity occurs through the formation of complexescomprised of the cdk's and the cyclin-dependent kinaseinhibitors (CKI's). The activities of these CKI's includebinding to cyclin/cdk co-complexes and inhibition oftheir associated kinase activity (for review, see Sherrand Roberts, 1995), binding to speci®c cdks andsubsequent inhibition of their association with cyclins(Parry et al., 1995) and, paradoxically, promotion ofthe formation of some cyclin-cdk complexes (LaBaer etal., 1997). So, for example, in response to DNAdamage, p53 induces expression of the CKI protein,p21Cip1/Waf1 (p21), leading to inhibition of activity ofmultiple cyclin/cdk co-complexes (Dulic et al., 1994;El-Deiry et al., 1993; Harper et al., 1993; Waldman etal., 1995). p27Kip1 (p27) contributes to the growthsuppressive signals from TGF-b or cell-cell contact(Hengst et al., 1994; Polyak et al., 1994a,b; Reynisdot-tir et al., 1995; Toyoshima and Hunter, 1994) andinhibition of its expression by antisense oligonucleo-tides prevents G1 arrest in response to mitogendepletion (Coats et al., 1996). The importance of p27in regulating cell growth has recently been appreciatedfrom studies which examined the phenotype of micede®cient for this CKI (Fero et al., 1996; Kiyokawa etal., 1996; Nakayama et al., 1996). Mice nullizygous forp27 exhibit hyperplasia in multiple organ systems dueto failure to exit the cell cycle during terminaldi�erentiation. The loss of p27 is also associated withneoplastic transformation of the pituitary gland,consistent with the notion that p27 functions as atumor suppressor.Loss of another CKI, p16Ink4a (p16), has also been

implicated in malignant transformation (Kamb et al.,1994; Serrano et al., 1993, 1996). Although notassociated with the generalized hypertrophy of thep27 knock-out mice, p16-de®cient mice exhibit anumber of soft tissue sarcomas and are highlysusceptible to radiation and carcinogen-induced tu-mors (Serrano et al., 1996). Loss of p16 expression hasalso been implicated in human cancers due to its highfrequency of deletion or mutation in melanomas,pancreatic adenocarcinomas, leukemias and malignantastrocytomas (Caldas et al., 1994; Hussussian et al.,1994; Ichimura et al., 1996; Jen et al., 1994; Kamb,1995; Kamb et al., 1994; Liu et al., 1995; Schmidt etal., 1994).We have been studying cell cycle regulatory path-

ways in the p16-de®cient, malignant astrocytoma cellline, U343 MG-A (U343). Retinoic acid (RA)treatment of these cells inhibits proliferation, altersmorphology, and increases the expression and organi-zation of a marker of glial di�erentiation, glial

Correspondence: JT RutkaReceived 7 March 1997; revised 30 June 1997; accepted 30 June 1997

Oncogene (1997) 15, 2037 ± 2048 1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

®brillary acidic protein (GFAP) (Rutka et al., 1988). Inthe present study, we sought to determine therelationship between RA-induced di�erentiation andthe expression of several key cell cycle regulators. Weshow that whereas both CKI-induction and RAtreatment can independently cause U343 cells to arrestin G1 phase, the cell morphologies resulting from eachtreatment are markedly di�erent. Our data demon-strate further that when p16-, p21- or p27-induced cellcycle arrest is accompanied by RA treatment, aprofound phenotypic change in U343 astrocytomacells occurs, cells becoming morphologically indistin-guishable from primary astrocytes.

Results

Retinoic acid alters the morphology and growthcharacteristics of U343 cells

The p16-de®cient, human astrocytoma cell line U343responds to all-trans RA treatment (Rutka et al.,1988). Figure 1 shows that RA inhibits U343proliferation (Figure 1a) and induces morphologicalchanges (Figure 1b ± c). So, rather than the tightlypacked `cobble-stone' appearance of untreated cells(Figure 1b), RA causes cells to elongate and acquire asomewhat bipolar appearance (Figure 1c). Occasionallong processes (510% of cells) are also observed in

these RA treated cultures. The altered morphology inRA-treated U343 cells is accompanied by changes incytoskeletal protein assembly (Figure 2). For example,untreated U343 cells exhibit di�use cytoplasmic andprominent perinuclear cytoplasmic staining for GFAP(Figure 2a). This disorganized pattern of expression isreplaced by a ®lamentous arrangement of GFAPstaining in the cytoplasm and processes of RA-treatedcells (Figure 2b). The RA-induced reorganization ofGFAP is also accompanied by changes in its apparentphosphorylation state. Using a phospho-GFAP-speci®cmonoclonal antibody, YC-10, which recognizes phos-phoserine in the context of a ten amino acid peptide ofthe GFAP amino-terminus (Matsuoka et al., 1992;Yano et al., 1991), we show that a prominent punctateor speckled staining pattern is evident both in thecytoplasm and the nucleus of untreated cells (Figure2c ± d). This punctate staining pattern is lost in thecytoplasm but not in the nucleus of U343 cells treatedwith RA. Thus, RA-induced reorganization of theU343 astrocytoma cytoskeleton is accompanied by aloss of the phosphorylated form of GFAP from thecytoplasm.

RA treatment alters the expression and activity of cellcycle regulatory proteins

We next determined if the G1 arrest following RAtreatment is accompanied by changes in the expression

Figure 1 Growth of U343 cells in the presence and absence of retinoic acid. (a) Growth curve analysis of U343 astrocytoma cellstreated in the presence or absence of 561076

M RA-treated astrocytoma cells show a decrease in cellular proliferation beginning onday 3 compared to controls. The di�erence in cell number between control and RA-treated cultures is maximum by day 9. (b, c)Morphological alterations induced by RA treatment of U343 astrocytoma cells. Control U343 cells show colonies with polygonalcells and marked cell crowding (b) In contrast, there is a distinct morphological change brought about by RA treatment. RA-treatedcells are predominantly bipolar with extension of slender, tapering cell processes (c) Cell crowding and colony formation are notobserved. b ± c, phase microscopy, 6300

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2038

and activity of cell cycle regulatory factors, speci®cally,the G1 cyclins, their associated kinases and kinaseinhibitors (Figure 3). The steady state protein levels of

cdk1, cdk2, cdk4, and cdk6 do not change with RAtreatment. In the case of the CKI, p21, inductionoccurs with slightly increased kinetics in RA-treatedrelative to untreated U343 cells. In contrast, expressionof p27 is ®vefold higher in day 5 RA-treated cellscompared to untreated U343 cells. This level of p27 inday 5 RA-treated cells is reached in untreated cells onlyafter day 9 where proliferation of these cells has slowed(data not shown). Following RA treatment, cyclin D1protein shows no change whereas cyclins A and Egradually decrease (data not shown).The altered kinetics of p27 induction following RA

treatment suggested that the kinase activity associatedwith the cdk's would also be inhibited in U343 cells.Thus, cdk2 or cdk4 were immunoprecipitated fromRA-treated or untreated U343 cells. Their associatedkinase activity was then determined in in vitro kinaseassays using histone H1 or puri®ed human recombi-nant full length pRB, respectively, as substrates. As theright panels in Figure 3 demonstrate, the kinaseactivity associated with cdk2 is strongly inhibited byday 5 of RA treatment, the point at which both p21and p27 are strongly induced. In the case of cdk4,however, only a minor but consistent reduction in itsassociated kinase activity is seen in these RA-treatedcells. Thus, the proliferative block induced in U343cells appears to be associated with p21 and p27induction and a commensurate inhibition of cdk2,but not cdk4, associated kinase activity. This block isassociated with a morphological change in U343 cells

Untreated RA Treated

GFAP

Phospho-GFAP

Figure 2 RA-induced rearrangement of the U343 astrocytomacytoskeleton. (a) Control U343 cells show hazy perinuclearstaining for GFAP (nonphospho-speci®c) in polygonal cells.Immuno¯uorescence microscopy, 6780. (b) RA-treated cellshave a prominent network of glial ®laments stained for GFAP.These ®laments bridge the cytoplasmic space from the nucleus tothe plasma membrane and appear far more organized than are®laments in control cells. (c) Staining for phospho-GFAP revealsa punctate cytoplasmic and nuclear pattern of staining (arrows) inuntreated U343 cells. (d) RA-treated cells are characterized by acomplete disappearance of the cytoplasmic pattern of staining butpreservation of nuclear staining. b ± d, immuno¯uorescencemicroscopy, 6630

cdk1

cdk6

cdk4

cdk2

p21

p27

H1

pRB

N RA N RAN RA

N RA N RA N RA

3 5 7

3 5 7

Western

Kinase Activity

Figure 3 Expression and activity of cell cycle regulatory factors in U343 cells. (Left panel) Western analysis of cdk1, cdk2, cdk4,cdk6, p21, p27 and pRB in untreated and RA-treated U343 cells. p21 and p27 are signi®cantly increased in RA-treated cells on days5 and 7 relative to untreated cells while cdk levels are unchanged following RA treatment. Only a minor change in thephosphorylation state of pRB is observed in these cells, both hyper- and hypo-phosphorylated forms persisting in untreated andRA-induced cells. (Right panel) Kinase activity of cdk2 and cdk4 in untreated and RA-treated cells. In vitro kinase assaysdemonstrate that cdk2-associated kinase activity is absent by day 5 in RA-treated cells. In contrast, only a small but consistentdecrease in cdk4-associated kinase activity is seen with RA treatment

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2039

which is distinct from untreated cells and from normalastrocytes.

E�ect of induced CKI expression on U343 astrocytomas

The morphological changes and cell cycle arrest causedby RA were associated with induction of p21 and p27.We wished, therefore, to determine if CKI-induced cellcycle arrest would also alter U343 cell morphology andwould a�ect the outcome of RA treatment. Thus, p16,p21 and p27 under the control of the tetracycline-

repressor (tetR) expression system (Gossen and Bujard,1992; Resnitzky et al., 1994) were individuallyintroduced into U343 cells. Several lines were isolatedwhere tight tetracycline-dependent control of CKIexpression could be demonstrated. All of the lines foreach CKI behaved identically and a representative linefor each is presented in Figure 4.Figure 4a, c and e demonstrate the regulated

expression of p16, p21 and p27, respectively, usingthe tetR system. For all three CKI's little or nodetectable expression is seen under repressed condi-

a

1 3 5 7Days

p16

tet in media + – + – + – + –

b

p21 Induction

Days 1 3 5 7

tet in media + – + – + – + –

p27 Induction

Days 1 3 5 7

tet in media + – + – + – + –

c e

d f

Figure 4 Tetracycline-inducible expression of CKI's in U343 cells. U343 cells with CKI's under the control of the tetO were grownin the presence (+) or absence (7) of tetracycline (tet). (a) Western analysis demonstrates that removal of tet from the media causesa strong induction of p16 protein by 24 h which increases progressively over several days achieving maximal levels by day 5.Uninduced U343 cells express little p16 demonstrating the tight control of expression with this system. Induction of p16 causes apotent proliferative block as demonstrated by growth curve analysis (b). Similarly, induction of p21 or p27 (c, e) causes growtharrest of U343 cells as determined by growth curves analysis (d, f)

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2040

tions (+). In contrast, within 24 h of removingtet(7), expression of all three CKI's can bedetected. Expression reached maximal levels for allthree CKI's near day 5 following induction. Asexpected, expression of p16, p21 or p27 produced apotent proliferative block in U343 cells (Figure 4b, dand f, respectively). In all three cases, no increase incell numbers were observed after day three followingremoval of tet. Table 1 demonstrates further that byday 5, the majority of the cells are blocked in G1 inthe p16 and p27 arrested cells. In the case of p21, we

consistently observe that a signi®cant fraction of thesecells are arrested in the G2/M portion of the cellcycle.We next determined the e�ect of CKI induction on

the expression and activity of factors in¯uenced by theCKI's, speci®cally the cdk's, the pRB- and the E2F-family proteins. Figure 5a illustrates the expressionpattern of cell cycle regulatory proteins during p16induction. The three pRB-family proteins, pRB, p107and p130 are primarily hyperphosphorylated inuninduced cells (day 0) but are found exclusively intheir hypophosphorylated forms by day 3 of p16induction. Additionally, pRB and, more dramatically,p107 are reduced in levels in these cells while p130 isinduced as p16 expression increases. E2F-1, which likep107 and pRB contain E2F-binding sites in theirpromoter regions (Shan et al., 1994), is stronglyrepressed following p16 expression. In contrast, E2F-4, which is the predominant E2F-family memberassociated with p130 in quiescent cells (Cobrinik etal., 1993; Smith et al., 1996; Vairo et al., 1995),increases slightly in these cells.

Table 1 Flow cytometric analysis of tetracycline induced anduninduced U343 cells after 3 days of experiment

%G1 %S %G2M

p16 OFFp16 ONp21 OFFp21 ONp27 OFFp27 ON

719575886993

182123203

113139114

0 1 2 3

a

b

α-pRBp16 Induction

(days) 0 1 2 3

α-p107

0 1 2 3

α-p130

0 1 2 3

α-E2F-1

0 1 2 3

α-E2F-4

0 1 2 3

α-p16

p16 Induction (days)

0 1 2 3

α-cdk4

0 1 2 3

α-cdk6

0 1 2 3

α-cdk2

p16 Induction (days)

cdk2cdk2PO4

0 3 5 P

α-cdk4

0 1 3 5

α-cdk2

0 1 3 5

α-cdk6p16 Induction

(days)

pRB

Figure 5 Expression and activity of factors regulated by CKI's. (a) Coincident with strong induction of p16, there is a decrease inexpression of pRB and p107, with shift to the hypophosphorylated form. p130 is also shifted from four identi®able forms to twohypophosphorylated forms but its levels are increased as p16 is induced. E2F-1 protein is also decreased to almost undetectablelevels while E2F-4 levels remain constant. Western blot analysis for cdk2, cdk4 and cdk6 showed little variation in steady state levelsof these proteins following p16 induction. The Western blot for cdk2 reveals a loss of the active, hyperphosphorylated form of cdk2(lower band) and shift exclusively to the slower migrating, hypophosphorylated form. (b) Kinase activity associated with cdk2, cdk4and cdk6. As expected, cdk2 shows a strong decrease in the associated kinase activity as cells become growth arrested. In contrast,we consistently observe little change in the kinase activity associated with cdk4 and cdk6. The level of non-speci®c kinase activity forcdk4 was con®rmed by addition of a blocking peptide (lane P)

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2041

The bottom row of Western blots in Figure 5ademonstrates that little change in the levels of cdk2,cdk4 and cdk6 occur in these p16-arrested cells. Cdk2does, however, undergo a signi®cant change in mobilityin SDS ±PAGE. Speci®cally, in exponentially growingp16-de®cient cells (day 0), cdk2 migrates as twospecies, a faster migrating, hyperphosphorylated form,which is the active form of cdk2 and a slower, inactivehypophosphorylated form (Poon and Hunter, 1995).When U343 cells become arrested following expressionof p16, the active, hyperphosphorylated form of cdk2is lost. It should be noted further that neither p21 orp27 were induced in the p16-arrested cells (data notshown). The kinase activity associated with cdk2, cdk4and cdk6 were also determined (Figure 5b). Asexpected, cdk2 shows a strong decrease in theassociated kinase activity as cells become growtharrested. In contrast, we consistently observe littlechange in the kinase activity associated with cdk4 andcdk6. The same antibodies for cdk4 and cdk6 detectedchanges (decreased) in associated kinase activity forthese two cdk's when isolated from exponentiallygrowing and di�erentiated myoblasts (C2C12 cells).The persistence of cdk4-associated kinase activity inthese cells was also veri®ed using three independentantibodies (data not shown).The phosphorylation state of the pRB-family

proteins is also altered in the p21 and p27-inducedcells. A representative Western for pRB is seen inFigure 6a. Like the cells where p16 is induced, pRBbecomes primarily hypophosphorylated and its levelsrepressed following p21 or p27 induction. The kineticsof this transition for the p21/p27-induced cells isslightly delayed relative to e�ects induced by p16.p107 levels are also repressed following both p21 andp27 induction. In the case of p130, p21-expressing cells

have increased levels of p130 while little change in p130was observed in p27-arrested cells. As with p16-induced cells, expression of E2F-1 was sharplyreduced and E2F-4 levels were unchanged in p27 andp21-induced cells (data not shown). For p27-arrestedcells, the kinase activity associated with cdk2 and cdk4is similar to that observed for cells arrested followingp16 expression (data not shown).

CKI expression during RA-stimulation induces uniquemorphological and cytoskeletal changes in U343astrocytomas

The cell cycle arrest associated with p16, p21 or p27induction in U343 astrocytomas is accompanied by astrong morphological transition. All three CKI'sinduced identical morphological and cytoskeletalalterations and these changes could be reversed byrepressing CKI expression by adding tet back to themedia. Figure 7 illustrates the gross morphologicalchanges associated with CKI expression, p16 used hereas an example. Figure 7a shows the usual `cobble-stone' appearance of U343 cells when p16 expression isrepressed. Following p16 expression, however, the cells¯atten and contain abundant cytoplasm which has alarge number of vacuoles around the nucleus. These`pancake'shaped cells are in marked contrast to theelongated cells observed following RA-dependent cellcycle arrest (see Figure 7c). We next wished todetermine if the CKI-arrested cells, when treatedsimultaneously with RA, would resemble cells treatedwith RA exclusively. As Figure 7d illustrates, thecombination of a strong proliferative block imposed byp16 in the presence of RA stimulation caused a uniquemorphological transition. Rather than elongate intocells with a bipolar appearance, these p16-induced,

p21 Induction

0 1 3

a

day 1 day 7day 5day 3day 2

– +

pRBPO4

pRB

day 1 day 7day 5day 3

p27 Induction

pRBPO4

pRB

0 3

p21 p27

days of Induction

p130

0 1 3

p21 p27

days of Inductionp107PO4

0 1 3

p107

– +– +– +

– + – +– +– + – +

b c

Figure 6 Expression of the pRB-family proteins in p21 and p27 induced U343 cells. (a) Induction of p21 or p27 induces aquantitative shift of pRB to its active hypophosphorylated form and a reduction in its overall levels. The kinetics of this transitionfor the p21/p27-induced cells is slightly delayed relative to e�ects induced by p16. (b, c) p107 levels are also repressed for followingboth p21 and p27 induction. In the case of p130, p21-expressing cells have increased levels of p130 while little change in p130 wasobserved in p27 arrested cells

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2042

RA-treated cells formed numerous long processes.These processes migrated in many directions andtypically made contacts with other processes and cellbodies in the dish. These cells, in fact, resembled themorphology of primary astrocytes grown in culture.The same morphological changes shown in Figure 7 forp16 were seen in U343 cells expressing p21 and p27,either alone or in conjunction with RA treatment.Under all of these conditions, cells were viable forgreater than 21 days and these e�ects were reversibleafter as long as 14 days following CKI-induction/RA-treatment (data not shown). The return to the usualappearance and proliferative state of U343 cells wasnot accompanied by any obvious increase in cell death.The profound morphological changes associated

with CKI induction and simultaneous RA treatmentindicated that important alterations in the U343cytoskeleton had occurred. We ®rst examined theexpression of the glial/astrocyte-speci®c protein, glial®brillary acid protein (GFAP; Figure 8). In proliferat-ing U343 cells, this intermediate ®lament is highlydisorganized, appearing as a di�use, cytoplasmicfactor. Induction of the CKI's (p21 in this case),clearly results in an altered arrangement of GFAP

(Figure 8, middle panel). The `pancake' appearance ofthese cells can easily be appreciated and a ®ne®lamentous arrangement of GFAP radiating outwardis observed. When the U343 cells are arrested with p21and treated with RA, GFAP acquires a distinctarrangement. Under these conditions, a ®lamentouspattern is still evident in the cell bodies, but GFAP alsoshows very strong localization to the long processesthat have formed under these conditions (Figure 8,lower panel). This ®gure also illustrates the parallelarrangement of processes that are often seen in thesecultures.We next examined the expression of actin and the

intermediate ®lament protein, vimentin (Figure 9). Aswith the pattern of staining for GFAP, both vimentin(green) and actin (red) show di�use, disorganizedstaining patterns in U343 astrocytomas (Figure 9a).p16 induction (Figure 9b) causes a strong rearrange-ment of actin, the radial staining pattern reminiscent ofGFAP in these same cells, while vimentin only showslimited organization. RA treatment (Figure 9c) causesreorganization of both actin and vimentin, well de®ned®laments clearly visible for both of these proteins.Figure 9d reveals, however, that a profound rearrange-

p16 Uninduced/ no RA treatment

tet removed

p16 induced (day 7)

RA added

tet removed

RA treated and p16 induced (day 7)

RA treated (day 5)

RA added

Figure 7 E�ects of p16 expression and retinoic acid treatment on U343 cell morphology. Untreated U343 cells (a) grow tightlypacked in colonies and rarely have any processes. This phenotype is identical to the growth characteristics of the parental U343 cellline. Following 5 days of treatment with all-trans retinoic acid (c), U343 cells undergo a phenotypic change consisting of bipolar cellprocess formation, with short tapering processes and occasional (510%) long process formation. p16-mediated growth arrest (b) isassociated with formation of large ¯at cells with perinuclear vacuolization. Both phenotypes seen in (b) and (c) are reversiblefollowing removal of retinoic acid or suppression of p16, respectively. Panel (d) shows that combining retinoic acid treatment withp16 induction, in any order, results in the formation of cells with long processes which contact other cells, morphologicallyreminiscent of primary astrocytes. p21 and p27 induction with RA treatment had identical e�ects. Phase microscopy, 6300

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2043

ment of both actin and vimentin occur in p16 inducedcells treated simultaneously with RA. Speci®cally, ®neactin ®laments are seen extending into the longprocesses that form in these cells. More dramatically,the vimentin signal, which was originally con®ned tothe cytoplasm in a perinuclear pattern, is now observedat the tips of the processes of these cells. Vimentin isalso localized to a region in the cell body in these RAtreated, p16-arrested astrocytomas, this localizationbeing apparently discontinuous with the vimentin atthe tips of the processes. The pattern of staining forvimentin and actin in p16-induced, RA-treated U343cells is more interesting given their staining pattern innormal, primary human astrocytes (Figure 9e). In theastrocytes, vimentin is also localized to the tips of theprocesses which extend from the cell bodies and,discontinuously, in the cytoplasm in a discrete area.Actin is also highly organized in astrocytes, ®lamentsseen extending into the long processes. These datasuggest that the malignant astrocytoma, U343 can be

induced to phenotypically resemble quiescent primaryhuman astrocytes only when a potent proliferativeblock, provided by CKI expression, and stimulationwith RA occur together.

Discussion

Our study demonstrates that induced expression ofp16, p21 or p27 in the p16-de®cient astrocytoma,U343, in the presence of retinoic acid causes thesemalignant cells to morphologically resemble primaryhuman astrocytes. The morphological changes in U343induced by p16, p21 or p27 alone do not resemblenormal astrocytes nor does induction with retinoic acidalone. Rather, expression of a predominantly astrocyticphenotype with multiple long cytoplasmic processesoccurs only when a strong proliferative block impartedby p16, p21 or p27 expression is accompanied by RAstimulation. Thus, in the absence of a signal from amorphogen (i.e. RA), cell cycle arrest is insu�cient tocause U343 cells to resemble primary astrocytes.It is now well established that cell cycle arrest is

associated with CKI-dependent inhibition of cdkactivity. Cell cycle arrest is thought to be arequirement for terminal di�erentiation. Ectopicexpression of CKI's in undi�erentiated cells can leadto the limited expression of di�erentiation markers orto morphologic changes of the di�erentiated pheno-type. For example, Vitamin D3 treatment of U937leukemia cells causes induction of p21 and p27 mRNAfollowed by expression of mRNA for macrophagemarkers (Liu et al., 1996). Transfer of p21 or p27 intoundi�erentiated U937 cells causes an increasedexpression of CD11b+ and CD14+ cells, suggestingthat CKI expression alone is su�cient to inducedi�erentiation in these cells. However, whereas nearly100% of U937 cells express CD11b+ and CD14+after vitamin D3 treatment, only 20 ± 35% expressthese markers after p21 or p27 transfer, and 30 ± 45%do so after co-transfection with p21 and p27. Similarly,DMSO treatment of mouse neuroblastoma cells causesan increase in p27 expression, a shift of pRB to thehypophosphorylated form, and morphologic differen-tiation (Kranenburg et al., 1995). Transfection of p27into these undi�erentiated cells, however, results inonly 50% of the cells undergoing a similar morpholo-gic change suggesting that induction of CKI expressionalone is ine�cient at causing phenotypic differentia-tion.It is also clear from our study, however, that RA

treatment alone of U343 cells only weakly causes achange in morphology of these cells. We suggest thatRA-treated U343 cells may represent an intermediatephenotype. The transition to a phenotype resemblingprimary astrocytes requires a strong block to cell cycleprogression normally imparted by p16Ink4a. Since U343cells are p16Ink4a-de®cient, only a potent block toproliferation imparted by the induced expression ofp16, p21 or p27 causes these cells to acquire theastrocytic phenotype.We con®rmed that the architecture of the CKI-

induced and/or RA-treated cells resembled primaryastrocytes by examining the arrangement of GFAP,vimentin and actin. The disorganized state of thesestructural proteins is radically altered in response to

untreated

p21 alone

p21+RA

Figure 8 Rearrangement of GFAP in CKI-expressing U343 cells.U343 cells expressing p21 are shown stained for GFAP. Anextensive ®lamentous pattern is seen in these large, ¯at cellsrelative to untreated cells. Following p21 expression and RAtreatment, GFAP staining reveals the dramatic morphologicalchanges of malignant U343 cells. Cells form many long processes,these processes often migrating in the same direction and inbundles. The processes also form contacts with other cells orprocesses. The phenotypic changes described here for p21induction in the presence and absence of RA are identical tothe changes observed following p16 or p27 expression in thepresence or absence of RA. Immuno¯uorescence microscopy,6780 (top), 6630 (middle and bottom)

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2044

Figure 9 Vimentin and actin staining in U343 cells and primary astrocytes. (a) Control U343 astrocytoma cells with typicalpolygonal morphology. Vimentin immunoreactivity is present in a ®ne trabecular pattern throughout the cytoplasm, and actin isdistributed beneath the surface of the cell membrane. Immuno¯uorescence confocal microscopy, 6470. (b) p16-induced U343astrocytoma cells. A typical large ¯at cell is observed. Vimentin immunoreactivity is sparsely dispersed throughout the expandedcytoplasm, but actin is expressed in radial arrays from the cell nucleus reminiscent of stress ®ber formation 6470. (c) RA-treatedU343 astrocytoma cells. The cells have become broadly bipolar, vimentin is found throughout the cytoplasm, and actin is expressedin a ®lamentous pattern in the cell cytoplasm, 6380. (d) RA-treated, p16-induced U343 astrocytoma cells. Speci®cally, ®ne actin®laments are seen extending into the long processes that form in these cells. Vimentin, originally con®ned to the cytoplasm in aperinuclear pattern, is now observed at the tips of the processes of these cells. Vimentin is also localized to a region in the cell bodyin these RA treated, p16-arrested astrocytomas, this localization being apparently discontinuous with the vimentin at the tips of theprocesses. The pattern of staining for vimentin and actin in p16-induced, RA-treated U343 cells resembles their staining pattern innormal, primary human astrocytes, 6600. (e) Primary human astrocytes in culture. In primary cultured astrocytes, vimentin is alsolocalized to the tips of the processes which extend from the cell bodies and, discontinuously, in the cytoplasm in a discrete area.Actin is also highly organized in these astrocytes, ®laments seen extending into the long processes. Multiple long, taperingcytoplasmic processes are seen which stain intensely for vimentin, 6600

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2045

CKI-mediated cell cycle arrest. In fact, in data we havenot shown, the phosphorylated form of GFAP, whichcan be detected in both the nuclear and cytoplasmiccompartments of U343 cells, speci®cally disappearsfrom the nuclear compartment upon CKI expression.These data and those described above clearly demon-strate that the activity of the cell cycle machinery cana�ect cellular architecture and in¯uence the organiza-tion of speci®c cytoskeletal components. The mostdramatic reorganization of an intermediate ®lamentwas observed for vimentin. Slight reorganization ofvimentin occurs in response to RA treatment or CKIexpression alone. However, when both signals areprovided, vimentin localizes to the tips of the newlyformed processes in these cells. This arrangement ofthis intermediate ®lament is indistinguishable from itssubcellular location in primary human astrocytes.Thus, we conclude that an astrocytic phenotype canbe induced in malignant U343 cells when provided withboth a strong proliferative block and a morphogenicsignal. It will be interesting to determine if therelocalization of vimentin occurs in response toreorganization or expression of, as yet, unidenti®edfactors which promote process outgrowth and/orattachment, such as neural adhesion molecules orintegrins.

Materials and methods

Cell culture

The well characterized U343MG-a (U343) malignantastrocytoma cell line was established from a primarymalignant astrocytoma in an adult. This particular cellline is a subclone of the original tumor that stablyexpresses the astrocyte di�erentiation marker, glial®brillary acidic protein (GFAP). U343 grows adherentlyin a-MEM supplemented with 10% fetal bovine serum andpenicillin/streptomycin/fungizone (Gibco BRL, Gaithers-burg, MD, USA). A culture of primary astrocytes wasestablished from a cortical resection of human brainperformed due to a seizure disorder.

For di�erentiation 2.5 ± 56105 cells were plated in 10 cm2

dishes. All trans retinoic acid (RA) (Sigma, St Louis, MO)was added from 561073 M stock solution in 100% ethanol toa ®nal concentration of 561076 M. Control cells werecultured in 0.1% ethanol. The medium was changed everyother day during the course of all experiments.

Cell proliferation assay

Cell growth was assayed by counting cells at de®nedintervals. Brie¯y, cells were trypsinized and resuspended inmedia, and an aliquot of cells was counted using ahemocytometer. Each count represented an average of®ve counts on three separate determinations. Cell prolif-eration assays were repeated in triplicate.

Flow cytometric analysis

To determine the proportion of cells present in a particularcell cycle phase, ¯ow-assisted cell sorting analysis (FACS)of DNA content was performed. Brie¯y, 0.2 ± 16106 cellswere trypsinized, washed in phosphate bu�ered saline(PBS), and resuspended in ice cold 80% ethanol. Cellswere kept at 48C on ice until propidium iodide (PI) DNAstaining was performed. For di�erent samples theconcentration of cells was kept equivalent. For staining,

®xed cells were resuspended in PI and DNAse-inactivatedRNase A (®nal concentration 1 mg/mL) and wereincubated for 30 min at room temperature in the dark.Stained cells were ®ltered through mesh capped tubes andDNA content was analysed on Becton-Dickinson FACS-can. Percent cell cycle phase was determined using Cell Fitsoftware (Becton-Dickinson).

Antibodies

Antibodies to cyclin A, cyclin E, cyclin D1, cdk1, cdk2,cdk4, cdk6, E2F1 and E2F4 were obtained from SantaCruz Biotech. (Santa Cruz, CA) and to pRB and p16 fromPharmingen (San Diego, CA). a-p21 and a-p27 antibodieswere obtained from Upstate Biotechnology Inc. (LakePlacid, NY). Antibodies speci®c for GFAP were obtainedfrom Dako Inc. (Denmark), and vimentin and phalloidin-rhodamine were obtained from Sigma (St Louis, MO,USA), whereas the phosphorylated GFAP antibody (YC-10) was from Medical and Biological Laboratories Co.,Ltd. (Nagoya, Japan).

Western blots

Total cell lysates (10 ± 30 mg) were subjected to SDS ±PAGE. Proteins were transferred to polyvinylidenedi¯uoride (Immobilon P) membranes by semi-dry trans-fer. Blots were rehydrated prior to immunodetection, andthen were blocked in 5% skim milk in PBS/0.1% Tween 20at room temperature for 1 h. Primary and secondaryantibody incubations were performed in blocking solutionat room temperature for 1 h. Goat anti-rabbit or mousehorseradish peroxidase conjugated secondary antibodieswere used at 1 : 5000 ± 1 : 8000 concentrations. Detectionwas performed using the enhanced chemiluminescencesystem (ECL, Amersham, Oakville, ON).

Immuno¯uorescence and confocal microscopy

Glass cover slips with cells from each treatment group werepermeabilized with 0.02% triton6100 for 5 min prior to®xation with 4% paraformaldehyde in 0.1 M phosphatebu�er (PH 7.2) for 30 min. They were then washed severaltimes with PBS and the residual aldehyde blocked withPBS containing 0.5% BSA and 0.15% glycine. Followingseveral rinses in PBS with 0.5% BSA, the cells wereincubated with a monoclonal antibody against vimentin orthe polyclonal antibody for GFAP for 1 h. Rhodaminephalloidin was used for actin staining for 30 min. Speci-mens were then washed thoroughly with PBS prior toincubation with the secondary antibody, a goat anti-murine or anti-rabbit IgG conjugated to FITC orrhodamine (Molecular Probes, Eugene, OR), for 1 h.After a thorough washing, the coverslips were mountedon glass coverslips with mounting medium (InovaDiagnostics, San Diego, CA). Irrelevant antisera or theomission of primary antibody were used as controls.Specimens were then examined under a Leica TCDS 4Dconfocal laser scanning microscope. 0.5 mm opticalsections were obtained and viewed as single planes. Inaddition, data sets from each cell were rendered as astacked perspective of all the optical sections. GFAP stainsshown were examined with a Leica immuno¯uorescencemicroscope.

Histone H1 and pRB kinase assays

Histone H1 and pRB kinase assays were performedaccording to the protocol of Matsushime et al. (1994).For pRB kinase assays, a human recombinant full lengthpRB was used as a substrate (QED Advanced ResearchTechnologies Inc., San Diego, CA, USA). This recombi-

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2046

nant pRB has also been shown to cause cell cycle arrestwhen microinjected into cancer cells (Pagliaro et al., 1995).

Plasmids and transfection

The tetracycline-repressor gene expression system was usedto induce expression of the CKI's, p16, p21 and p27. ThepUHD-15-1neo plasmid contains the E. coli tetracyclinerepressor element fused to the VP16 transactivationdomain of herpes virus. This fusion protein is driven bya CMV promoter and the vector has the neomycinresistance gene for selection. pUHD15-1neo (25 mg) wastransfected into U343 cells using the calcium phosphatemethod. Clones were selected in 900 mg/mL gentecin(Gibco/BRL) in a-MEM and stable expression of thefusion protein was determined by Western blot analysis oftotal cell lysates using a polyclonal antisera to VP16 (kindgift of Dr J Ingles, Univ. Toronto). Twenty clones wereanalysed for VP16 expression and the majority expressedVP16. One clone, which demonstrated high level expressionof VP16 (clone 88) was selected for transfection withpUHD10-3. pUHD10-3 contains a multiple cloning sitedownstream from tandem tetracycline operator sequencesand a CMV promoter. Full length human cDNAs of p16,p21 (gifts of D Beach) and p27 (gift of J Massague) wereinserted into the multiple cloning site of pUHD10-3, andthis plasmid (25 mg) was cotransfected with pgk-puro(0.5 mg) for selection of stable lines. These clones werealso maintained in 4 mg/mL tetracycline added from a

4 mg/mL stock in 70% ethanol. Puromycin was used forselection at 1 mg/mL, and G418 concentration wasmaintained at 500 mg/mL.

To induce expression of CKI's, cells were washed threetimes in PBS, before identical media without tetracyclinewas added. To screen for CKI expression, total cell lysateswere collected and Western blot analysis was performed forp16, p21 or p27. To determine the e�ect of induction ofCKI expression on the growth and morphology of U343cells, 2 ± 56105 cells were plated in 100 mm2 dishes. Thefollowing day, fresh media was placed and cell proliferationassays and ¯ow cytometric analysis were performed asdescribed.

AcknowledgementsThe pUHD15-1neo plasmid was a kind gift of Dr S Reedand pUHD10-3 was a kind gift of Dr H Bujard. Humanp16 and p21 cDNA were gifts from Dr D Beach andhuman p27 was a gift from Dr J Massague. Rabbit antiserato VP-16 was a generous gift from Dr CJ Ingels, Universityof Toronto. This work was supported in part through agrant from the MRC to JTR; through a grant to PAHfrom the National Cancer Institute of Canada with supportfrom the Canadian Cancer Society and through funds fromThe Research Institute, The Hospital for Sick Children toJTR. PBD is supported by an NCIC research fellowshipwith funds provided by the Canadian Cancer Society aswell as a research fellowship from the MRC.

References

Caldas C, Hahn SA, da Costa LT, Redston MS, Schutte M,Seymour AB, Weinstein CL, Hrurban RH, Yeo CJ andKern SE. (1994). Nature Genet., 8, 27 ± 32.

Coats S, Flanagan WM, Nourse J and Roberts JM. (1996).Science, 272, 877 ± 880.

Cobrinik D, Whyte P, Peeper DS, Jacks T and Weinberg RA.(1993). Genes & Dev., 7, 2392 ± 2404.

Dulic V, Kaufmann WK, Wilson SJ, Tlsty TD, Lees E,Harper JW, Elledge SJ and Reed SI. (1994). Cell, 76,1013 ± 1023.

El-Deiry WS, Tokino T, Velculescu VE, Levy DB, ParsonsR, Trent JM, Lin D, Mercer WE, Kinzler KW andVogelstein B. (1993). Cell, 75, 817 ± 825.

Fero ML, Rivkin M, Tasch M, Porter P, Carow CE, Firpo E,Polyak K, Tsai L-H, Broudy V, Perlmutter RM,Kaushansky K and Roberts JM. (1996). Cell, 85, 733 ±744.

Gossen M and Bujard H. (1992). Proc. Natl. Acad. Sci. USA,89, 5547 ± 5551.

Harper JW, Adami GR, Wei N, Keyomarsi K and ElledgeSJ. (1993). Cell, 75, 805 ± 816.

Hengst L, Dulic V, Slingerland JM, Lees E and Reed SI.(1994). Proc. Natl. Acad. Sci. USA, 91, 5291 ± 5295.

Hussussian CJ, Struewing JP, Goldstein AM, Higgins PAT,Ally DS, Sheahan MD, Clark Jr, WH, Tucker MA andDracopoli NC. (1994). Nature Genet., 8, 15 ± 21.

Ichimura K, Schmidt EE, Goike HM and Collins VP. (1996).Oncogene, 13, 1065 ± 1072.

Jen J, Harper JW, Bigner SH, Bigner DD, Papadopoulos N,Markowitz S, Willson JKV, Kinzler KW and VogelsteinB. (1994). Cancer Res., 54, 6353 ± 6358.

Kamb A. (1995). Trends Genet., 11, 136 ± 140.Kamb A, Gruis NA, Weaver-Feldhaus J, Liu Q, HarshmanK, Tavtigian SV, Stockert E, Day III RS, Johnson BE andSkolnick MH. (1994). Science, 264.

Kiyokawa H, Kineman RD, Manova-Todorova KO, SoaresVC, Ho�man ES, Ono M, Khanam D, Hayday AC,Frohman LA and Ko� A. (1996). Cell, 85, 721 ± 732.

Kranenburg O, Scharnhorst V, van der Erb AJ and ZantemaA. (1995). J. Cell Biol., 131, 227 ± 234.

LaBaer J, Garrett MD, Stevenson LF, Slingerland JM,Sandhu C, Chou HS, Fattaey A and Harlow E. (1997).Genes Dev., 11, 847 ± 862.

Liu L, Lassam NJ, Slingerland JM, Bailey D, Cole D,Jenkins R and Hogg D. (1995). Oncogene, 11, 405 ± 412.

Liu M, Lee M-H, Cohen M, Bommakanti M and FreedmanLP. (1996). Genes Dev., 10, 142 ± 153.

Matsuoka Y, Nishizawa K, Yano T, Shibata M, Ando S,Tanabe K, Kikuchi K, Tsuiki S and Nishi Y. (1992).EMBO J., 11, 2895 ± 2902.

Matsushime H, Quelle DE, Shurtle� SA, Shibuya M, SherrCJ and Kato J-A. (1994). Mol. Cell Biol., 14, 2066 ± 2076.

Nakayama K, Ishida N, Shirane M, Inomata A, Inoue T,Shishido N, Horii I, Loh DY and Nakayama K-I. (1996).Cell, 85, 707 ± 720.

Pagliaro LC, Antelman D, Johnson DE, Machemer T,McCulloch EA, Freireich EJ, Stass SA, Shepard HM,Maneval D and Gutterman JU. (1995). Cell Growth &Di�er., 6, 673 ± 680.

Parry D, Bates S, Mann DJ and Peters G. (1995). EMBO J.,14, 503 ± 511.

Polyak K, Kato J, Solomon MJ, Sherr CJ, Massague J,Roberts JM and Ko� A. (1994a). Genes & Dev., 8, 9 ± 22.

Polyak K, Lee M-H, Erdjument-Bromage H, Ko� A,Roberts JM, Tempst P and Massague J. (1994b). Cell,78, 59 ± 66.

Poon RYC and Hunter T. (1995). Science, 270, 90 ± 93.Resnitzky D, Gossen M, Bujard H and Reed SI. (1994).Mol.

Cell Biol., 14, 1669 ± 1679.Reynisdottir I, Polyak K, Iavarone A and Massague J.(1995). Genes Dev., 9, 1831 ± 1845.

Rutka JT, DeArmond SJ, Giblin JR, McCulloch JR, WilsonCB and Rosenblum ML. (1988). Int. J. Cancer, 53, 3634 ±3631.

Schmidt EE, Ichimura K, Reifenberger G and Collins VP.(1994). Cancer Res., 54, 6321 ± 6324.

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2047

Serrano M, Hannon G and Beach D. (1993). Nature, 366,704 ± 707.

Serrano M, Lee HW, Chin L, Cordon-Cardo C, Beach D andDePinho RA. (1996). Cell, 85, 27 ± 37.

Shan B, Chang C-Y, Jones D and Lee W-H. (1994).Mol. CellBiol., 14, 299 ± 309.

Sherr CJ and Roberts JM. (1995). Genes & Dev., 9, 1149 ±1163.

Smith EA, Leone G, DeGregori J, Jakoi L and Nevins JR.(1996). Mol. Cell Biol., 16, 6965 ± 6976.

Toyoshima H and Hunter T. (1994). Cell, 78, 67 ± 74.Vairo G, Livingston DM and Ginsberg D. (1995). Genes &

Dev., 9, 869 ± 881.Waldman T, Kinzler KW and Vogelstein B. (1995). Cancer

Res., 55, 5187 ± 5190.Weinberg RA. (1995). Cell, 81, 323 ± 330.Yano T, Taura C, Shibata M, Hirono Y, Ando S, KusubataM, Takahashi T and Inagaki M. (1991). Biochem. Biophys.Res. Comm., 175, 1144 ± 1151.

Effects of retinoic acid and p16 on astrocytomasPB Dirks et al

2048