BRCA1 mediates ligand-independent transcriptional repression of the estrogen receptor.

Mutational inactivation of BRCA1 confers a cumulative lifetime risk of breast and ovarian cancers. However, the underlying basis for the tissue-restricted tumor-suppressive properties of BRCA1 remains poorly defined. Here we show that BRCA1 mediates ligand-independent transcriptional repression of the estrogen receptor alpha (ERalpha), a principal determinant of the growth, differentiation, and normal functional status of breasts and ovaries. In Brca1-null mouse embryo fibroblasts and BRCA1-deficient human ovarian cancer cells, ERalpha exhibited ligand-independent transcriptional activity that was not observed in Brca1-proficient cells. Ectopic expression in Brca1-deficient cells of wild-type BRCA1, but not clinically validated BRCA1 missense mutants, restored ligand-independent repression of ERalpha in a manner dependent upon apparent histone deacetylase activity. In estrogen-dependent human breast cancer cells, chromatin immunoprecipitation analysis revealed the association of BRCA1 with ERalpha at endogenous estrogen-response elements before, but not after estrogen stimulation. Collectively, these results reveal BRCA1 to be a ligand-reversible barrier to transcriptional activation by unliganded promoter-bound ERalpha and suggest a possible mechanism by which functional inactivation of BRCA1 could promote tumorigenesis through inappropriate hormonal regulation of mammary and ovarian epithelial cell proliferation.

Increased cell survival by inhibition of BRCA1 using an antisense approach in an estrogen responsive ovarian carcinoma cell line.

STATEMENT OF FINDINGS: We tested the hypothesis that BRCA1 may play a role in the regulation of ovarian tumor cell death as well as the inhibition of ovarian cell proliferation. Introduction of BRCA1 antisense retroviral constructs into BG-1 estrogen-dependent ovarian adenocarcinoma cells resulted in reduced BRCA1 expression. BRCA1 antisense pooled populations and derived subclones were able to proliferate in monolayer culture without estrogen, whereas control cells began to die after 10 days of estrogen deprivation. In addition, both populations and subclones of BRCA1 antisense infected cells demonstrated a growth advantage in monolayer culture in the presence of estrogen and were able to proliferate in monolayer culture without estrogen, while control cells did not. Furthermore, clonal studies demonstrated that reduced levels of BRCA1 protein correlated with growth in soft agar and greater tumor formation in nude mice in the absence of estrogen. These data suggest that reduction of BRCA1 protein in BG-1 ovarian adenocarcinoma cells may have an effect on cell survival during estrogen deprivation both in vitro and in vivo.

Isolation of a functional copy of the human BRCA1 gene by transformation-associated recombination in yeast.

The BRCA1 gene, mutations of which contribute significantly to hereditary breast cancer, was not identified in the existing YAC and BAC libraries. The gene is now available only as a set of overlapping fragments that form a contig. In this work we describe direct isolation of a genomic copy of BRCA1 from human DNA by transformation-associated recombination (TAR) cloning. Despite the presence of multiple repeats, most of the primary BRCA1 YAC isolates did not contain detectable deletions and could be stably propagated in a host strain with conditional RAD52. Similar to other circular YACs, approximately 90kb BRCA1 YACs were efficiently and accurately retrofitted into bacterial artificial chromosomes (BACs) with the Neo(R) mammalian selectable marker and transferred as circular BAC/YACs in E. coli cells. The BRCA1 BAC/YAC DNAs were isolated from bacterial cells and were used to transfect mouse cells using the Neo(R) gene as selectable marker. Western blot analysis of transfectants showed that BRCA1 YACs isolated by a TAR cloning contained a functional gene. The advantage of this expression vector is that the expression of BRCA1 is generated from its own regulatory elements and does not require additional promoter elements that may result in overexpression of the protein. In contrast to the results with cDNA expression vectors, the level of BRCA1 expression from this TAR vector is stable, does not induce cell death, maintains serum regulation, and approximates the level of endogenously expressed BRCA1 in human cells. The entire isolation procedure of BRCA1 described in this paper can be accomplished in approximately 10 days and can be applied to isolation of gene from clinical material. We propose that the opportunity to directly isolate normal and mutant forms of BRCA1 will greatly facilitate analysis of the gene and its contribution to breast cancer.

Establishment and characterization of a breast cell strain containing a BRCA1 185delAG mutation.

OBJECTIVE: The aim of this study was to examine whether cells containing the heterozygous form of a BRCA1 185delAG mutation would exhibit abnormal growth or an altered response to DNA damage. METHODS: A primary culture of human mammary epithelial cells (90P) was obtained from the nontumor breast tissue of a 35-year-old patient who had undergone a mastectomy for removal of a breast tumor. These cells were immortalized (90PE6E7) following retroviral infection with HPV-16 viral E6/E7. genes. Both the 90P cell strain and the cell line were characterized for their ability to grow in culture, form colonies in soft agar, and produce tumors in athymic nude mice compared to normal breast epithelial cells containing wild-type BRCA1. 90P cells were also analyzed for cellular response to gamma radiation and H(2)O(2). RESULTS: These cells were confirmed to contain a frameshift mutation, 185delAG, of the BRCA1 gene. Despite being heterozygous for wild-type BRCA1, the 220-kDa full-size BRCA1 protein was abundantly expressed. 90P and 90PE6E7 cells grew at a similar rate and were anchorage dependent. 90PE6E7 also failed to form tumors in athymic nude mice. Finally, 90P cells exhibited a survival response similar to that of normal mammary epithelial cells to radiation damage and exposure to oxidative stress. CONCLUSION: To our knowledge the 90P cells and the 90PE6E7 cells are the first characterized, non-tumor-derived breast epithelial cells that are heterozygous for the BRCA1 germline mutation 185delAG. Our conclusion is that these BRCA1 mutant cells appear to have growth and stress response characteristics similar to those of normal human breast cells, which is consistent with the hypothesis that loss of heterozygosity must occur to impair putative BRCA1 function.

DNA methylation analysis of the promoter region of the human telomerase reverse transcriptase (hTERT) gene.

The promoter of the hTERT gene encoding the catalytic subunit of telomerase was recently cloned and has a dense CG-rich CpG island, suggesting a role for methylation in regulation of hTERT expression. In this study, we have initiated the analysis of the regulation of hTERT expression by examining the methylation status of up to 72 CpG sites extending from 500 bases upstream of the transcriptional start site of the hTERT gene into the first exon in 37 cell lines. These cell lines represent a variety of cell and tissue types, including normal, immortalized, and cancer cell lines from lung, breast, and other tissues. Using bisulfite genomic sequencing, we did not find a generalized pattern of site-specific or region-specific methylation that correlated with expression of the hTERT gene: most of the hTERT-negative normal cells and about one-third of the hTERT-expressing cell lines had the unmethylated/hypomethylated promoter, whereas the other hTERT-expressing cell lines showed partial or total methylation of the promoter. The promoter of one hTERT-negative fibroblast cell line, SUSM-1, was methylated at all sites examined. Treatment of SUSM-1 cells with the demethylating agent 5-aza-2'-deoxycytidine and the histone deacetylase inhibitor trichostatin A induced the cells to express hTERT, suggesting a potential role for DNA methylation and/or histone deacetylation in negative regulation of hTERT. This study indicates that there are multiple levels of regulation of hTERT expression in CpG island methylation-dependent and -independent manners.

Estrogen upregulation of BRCA1 expression with no effect on localization.

Alterations in the expression of the breast and ovarian cancer susceptibility gene BRCA1 may contribute to the development of mammary and ovarian neoplasia. The sex-steroid estrogen modulates cell proliferation of normal and neoplastic breast and ovarian epithelial cells, but the role of estrogen regulation on the expression of BRCA1 remains to be defined. In this study, estrogen-regulated BRCA1 expression was examined in breast and ovarian cancer cells. Estrogen stimulated the proliferation of estrogen receptor (ER)-positive breast MCF-7, C7-MCF-7, and ovarian BG-1 cells as well as the expression of the estrogen-inducible pS2 gene. This was concomitant with upregulation of BRCA1 mRNA (2.5- to 5.0-fold) and a 3- to 10-fold induction of BRCA1 protein (230 kDa). Cell fractionation studies localized the BRCA1 protein to the nucleus in both unstimulated and estrogen-stimulated cells. The antiestrogen ICI-182780 inhibited estrogen-induced cell proliferation, BRCA1 mRNA induction, and BRCA1 protein expression in ER-positive cells. Conversely, estrogen did not influence expression of BRCA1 in HBL-100 cells that lacked the estrogen receptor, although the constitutive levels of BRCA1 mRNA (but not protein) in these cells were 5- to 30-fold higher than in other breast and ovarian cancer cells. Secretion of the BRCA1 protein into the cell medium did not account for the discrepancy between the mRNA and protein levels in HBL-100 cells. Proliferation of HBL-100 cells was not affected by either estrogen or ICI-182780. Taken together, these data support a role for the steroid estrogen and the involvement of the estrogen receptor pathway in the modulation of expression of BRCA1. We therefore propose that stimulation of cell proliferation may be a prerequisite for upregulation of BRCA1 in breast and ovarian cancer cells.

Functional copies of a human gene can be directly isolated by transformation-associated recombination cloning with a small 3' end target sequence.

Unique, small sequences (sequence tag sites) have been identified at the 3' ends of most human genes that serve as landmarks in genome mapping. We investigated whether a single copy gene could be isolated directly from total human DNA by transformation-associated recombination (TAR) cloning in yeast using a short, 3' unique target. A TAR cloning vector was constructed that, when linearized, contained a small amount (381 bp) of 3' hypoxanthine phosphoribosyltransferase (HPRT) sequence at one end and an 189-bp Alu repeat at the other end. Transformation with this vector along with human DNA led to selective isolations of the entire HPRT gene as yeast artificial chromosomes (YACs) that extended from the 3' end sequence to various Alu positions as much as 600 kb upstream. These YACs were retrofitted with a NeoR and a bacterial artificial chromosome (BAC) sequence to transfer the YACs to bacteria and subsequently the BACs to mouse cells by using a Neo selection. Most of the HPRT isolates were functional, demonstrating that TAR cloning retains the functional integrity of the isolated material. Thus, this modified version of TAR cloning, which we refer to as radial TAR cloning, can be used to isolate large segments of the human genome accurately and directly with only a small amount of sequence information.

Functional evidence for a colorectal cancer tumor suppressor gene at chromosome 8p22-23 by monochromosome transfer.

Chromosome 8p is considered, from loss of heterozygosity analysis, to be a strong candidate for the location of a tumor suppressor gene inactivated in colorectal cancer. We have found a 53% (27 of 51) rate of allelic loss at the LPL locus on 8p22, with the smallest region of overlap of deletions including the region D8S258 to D8S277. Using microcell-mediated monochromosome 8 transfer into three colorectal cancer cell lines, SW480, SW620 and HT29, we have demonstrated a reduction of tumorigenicity in SW620 hybrids. Partial deletions of chromosome 8 in some SW620/8 hybrids further delineate the critical region(s) to 8p22-23. Hybrids of the colorectal cancer cell lines SW480 and HT29 containing chromosome 8 did not show suppression of tumorigenesis, but the H29/8 hybrid showed total suppression of soft agar clonicity. This indicates an alternate pathway of mutational progression in these three lines, despite the fact that SW480 was derived from the same patient as SW620.

Evidence for two senescence loci on human chromosome 1.

Microcell-mediated introduction of a neo-tagged human chromosome 1 (HC-1-neo) into several immortal cell lines has previously been shown to induce growth arrest and phenotypic changes indicative of replicative senescence. Somatic cell hybridization studies have localized senescence activity for immortal hamster 10W-2 cells to a cytogenetically defined region between 1q23 and the q terminus. Previous microcell-mediated chromosome transfer experiments showed that a chromosome 1 with an interstitial q-arm deletion (del-1q) lacks senescence inducing activity for several immortal human cell lines that are sensitive to an intact HC-1-neo. In contrast, our studies reveal that the del-1q chromosome retains activity for 10W-2 cells, indicating that there are at least two senescence genes on human chromosome 1. Sequence-tagged site (STS) content analysis revealed that the q arm of the del-1q chromosome has an interstitial deletion of approximately 63 centimorgans (cM), between the proximal STS marker DIS534 and distal marker DIS412, approximately 1q12 to 1q31. This deletion analysis provides a candidate region for one of the senescence genes on 1q. In addition, because this deletion region extends distally beyond 1q23, it localizes the region containing a second senescence gene to approximately 1q31-qter, between DIS422 and the q terminus. STS content analysis of a panel of 11 10W-2 microcell hybrid clones that escaped senescence identified 2 common regions of loss of 1q material below the distal breakpoint of del-1q. One region is flanked by markers DIS459 and ACTN2, and the second lies between markers WI-4683 and DIS1609, indicating that the distal 1q senescence gene(s) localizes within 1q42-43.

Induction of apoptosis by c-Fos protein.

The role of c-Fos in apoptosis was examined in two Syrian hamster embryo cell lines (sup+I and sup-II) and a human colorectal carcinoma cell line (RKO), using the chimeric Fos-estrogen receptor fusion protein c-FosER. As previously reported, contrasting responses were observed when these two cell lines were placed under growth factor deprivation conditions; sup+I cells were highly susceptible to apoptosis, whereas sup-II cells were resistant. In this report, we show that the activated c-FosER protein induces apoptosis in sup-II preneoplastic cells in serum-free medium, indicating that c-Fos protein can induce apoptotic cell death in these cells. c-Fos-induced apoptosis was not blocked by the protein synthesis inhibitor cycloheximide, suggesting that the c-Fos transcriptional activation activity is not involved. This conclusion was further supported by the observation that overexpression of v-Fos, which is highly proficient in transcriptional activation but deficient in the transcriptional repression activity associated with c-Fos, did not induce apoptosis. Constitutively expressed Bcl-2 delayed the onset of low-serum-induced apoptosis in sup+I cells and enhanced survival in sup-II cells. Further, coexpression of Bcl-2 and c-FosER in sup+I or sup-II cells protected the cells from c-FosER-induced apoptosis. The possibility that c-FosER-induced apoptosis requires a p53 function was examined. Colorectal carcinoma RKOp53+/+ cells, which do not normally undergo apoptosis in serum-free medium, showed apoptotic DNA fragmentation upon expression and activation of c-FosER. Further, when the wild-type p53 protein was diminished in the RKO cells by infection with the papillomavirus E6 gene, subsequent c-FosER-induced apoptosis was blocked. The data suggest that c-Fos protein plays a causal role in the activation of apoptosis in a p53-dependent manner. This activity does not require new protein synthesis and is blocked by overexpression of Bcl-2 protein.

Elevated expression of mitochondrial cytochrome b and NADH dehydrogenase subunit 4/4L genes in senescent human cells.

In order to isolate cDNAs expressed differentially in senescent cells, we constructed a cDNA library from poly(A)+ RNA of senescent human fibroblast cells and screened cDNAs by the differential screening method. We isolated five cDNA clones expressed preferentially in the senescent cells. These five cDNA clones fell into two groups by cross-hybridization. Sequence analysis revealed that these cDNA clones were homologous to the human mitochondrial genes for cytochrome b (cytb) and NADH dehydrogenase subunit 4/4L (ND4/4L). Northern blot analysis demonstrated that the expression of these two mitochondrial genes were elevated in senescent cells compared to young cells. However, the expression of the mitochondrial cytochrome oxidase I gene was similar between young and senescent cells, implying that the preferential expression of the mitochondrial genes in the senescent cells was selective for cytb and ND4/4L genes. The results suggest that expressions of mitochondrial cytb and ND4/4L genes are regulated by a senescence-dependent mechanism.

A gene involved in control of human cellular senescence on human chromosome 1q.

Normal cells in culture exhibit limited division potential and have been used as a model for cellular senescence. In contrast, tumor-derived or carcinogen- or virus-transformed cells are capable of indefinite division. Fusion of normal human diploid fibroblasts with immortal human cells yielded hybrids having limited life spans, indicating that cellular senescence was dominant. Fusions of various immortal human cell lines with each other led to the identification of four complementation groups for indefinite division. The purpose of this study was to determine whether human chromosome 1 could complement the recessive immortal defect of human cell lines assigned to one of the four complementation groups. Using microcell fusion, we introduced a single normal human chromosome 1 into immortal human cell lines representing the complementation groups and determined that it caused loss of proliferative potential of an osteosarcoma-derived cell line (TE85), a cytomegalovirus-transformed lung fibroblast cell line (CMV-Mj-HEL-1), and a Ki-ras(+)-transformed derivative of TE85 (143B TK-), all of which were assigned to complementation group C. This chromosome 1 caused no change in proliferative potential of cell lines representing the other complementation groups. A derivative of human chromosome 1 that had lost most of the q arm by spontaneous deletion was unable to induce senescence in any of the immortal cell lines. This finding indicates that the q arm of human chromosome 1 carries a gene or set of genes which is altered in the cell lines assigned to complementation group C and is involved in the control of cellular senescence.

Investigation of the role of G1/S cell cycle mediators in cellular senescence.

Cellular senescence is a state of irreversible cell cycle arrest in which normal cells at the end of their lifespan fail to enter into DNA synthesis upon serum or growth factor stimulation. We examined whether proteins required for G1/S cell cycle progression were irreversibly down-regulated in senescent human fibroblasts. Both the 44- and 42-kDa forms of the MAP-kinase protein were expressed at similar levels in young and senescent cells. In contrast to young cells where both forms were phosphorylated on tyrosine in response to serum, the p42MAP-kinase was not tyrosine phosphorylated upon serum stimulation, whereas p44MAP-kinase was phosphorylated on tyrosine in serum-starved or serum-stimulated senescent cells. Examination of p53 protein in growing, quiescent, and senescent cells revealed no significant differences in levels between the different growth states. In contrast, cdk2 and cyclin A mRNAs were completely down-regulated in stimulated senescent fibroblasts, while the G1 cyclins, C, D1, and E mRNAs, were still expressed in stimulated senescent cells although at reduced levels compared to young cells. The expression of early G1 markers, but not late G1 markers, indicates that senescent cells may be blocked at a point in late G1. We investigated whether transfection of cyclin A, alone or in combination with cdc2, was sufficient for extension of lifespan or escape from senescence. Clones expressing the transfected human cyclin A or cdc2 genes senesced at a population doubling similar to controls, thereby showing that cyclin A or cdc2 expression alone was insufficient for escape from senescence.

Genetic and molecular basis for cellular senescence.

Normal human and rodent cells in culture exhibit a finite life span at the end of which they exhibit morphological changes and cease proliferating, a process termed cellular senescence or cellular aging. Many cancer cells differ from normal cells in that they do not senesce and have an indefinite life span in culture, suggesting that alterations relating to cellular senescence are involved in the neoplastic evolution of tumor cells. Recent experimental results strongly support a genetic basis for cellular senescence. Defects in the senescence program in transformed cells can be corrected by introduction of a specific chromosome from normal cells into the abnormal cells. Using this approach, possible senescence genes have been mapped to specific chromosomes. Cell cycle control genes that regulate entry into the DNA synthetic phase of the cell cycle must be altered in senescent cells. Recent findings suggest that phosphorylation of the retinoblastoma gene is altered in senescent cells. It is possible, but not yet proven, that aging at the cellular level contributes to the aging of the individual. Therefore, an understanding of cellular senescence at the genetic and molecular levels may provide new insights into both the cancer and aging processes.

Growth and transformation suppressor genes for BHK Syrian hamster cells on human chromosomes 1 and 11.

To map putative tumor suppressor genes for the near-diploid baby hamster kidney fibrosarcoma cell line BHK, we transferred five different normal human chromosomes (1, 3, 7, 11, and 12) into these tumor cells by microcell-mediated chromosome transfer. Transfer of human chromosome 1 into BHK cells resulted in suppression of cell growth both on plastic and in soft agar, indicating that chromosome 1 has a generalized effect on cell growth and thereby suppresses anchorage-independent growth. Selection against cells with an intact chromosome 1 was observed. In contrast, the introduction of chromosome 11 into BHK cells resulted in suppression of anchorage independence but not growth on plastic. Most chromosome-11 growth-suppressed BHK hybrids retained intact copies of human chromosome 11. Tumorigenic derivatives of chromosome 11 hybrids had lost this chromosome. Transfer of human chromosome 3, 7, or 12 into BHK cells did not correlate with growth suppression of BHK cells on plastic or in soft agar. Thus, we conclude that genes that suppress BHK-cell growth in general or in agar reside on human chromosomes 1 and 11, respectively.

Down-regulation of cdc2 in senescent human and hamster cells.

Senescent cells fail to respond to serum-induced signals for DNA synthesis. Because a central role for the p34cdc2 protein kinase is postulated in control of the cell cycle, we examined the status of this kinase in senescent cells and other growth-arrested cells. In growing human and Syrian hamster fibroblasts, three 35S-labeled proteins of 34-36 kDa were immunoprecipitated with p34cdc2 antiserum. Only the two slower migrating forms were phosphorylated as determined by 32P labelling. In senescent cells, which failed to incorporate [3H]thymidine, no p34cdc2 protein was synthesized and very little or no cdc2 mRNA was observed. When maintained for 48 h in 0.5% serum, young cells also retained only marginal cdc2 expression. After stimulation of low serum-arrested cells by addition of 10% serum, a time-dependent increase of cdc2 mRNA was observed, whereas serum stimulation of senescent cells did not increase cdc2 mRNA. In contrast to senescent and low serum-arrested cells, cdc2 mRNA was expressed at normal levels in cells partially growth arrested by isoleucine deficiency in G1, by aphidicolin at G1-S, by etoposide in G2, or by Colcemid in the M phase of the cell cycle, indicating that cdc2 down-regulation does not always occur upon growth arrest. Following transfection of a plasmid containing the human CDC2 gene into hamster cells, expression of human cdc2 failed to overcome the block to DNA synthesis in senescent cells. Although p34cdc2 was synthesized in the transfected cells, the multiple phosphorylated forms of the proteins were not observed. Taken together, these data support the concept that a chain of events leads to senescence. While p34cdc2 kinase may be one of the critical elements, other cell cycle controls are also involved.

Senescence of nickel-transformed cells by an X chromosome: possible epigenetic control.

Transfer of a normal Chinese hamster X chromosome (carried in a mouse A9 donor cell line) to a nickel-transformed Chinese hamster cell line with an Xq chromosome deletion resulted in senescense of these previously immortal cells. At early passages of the A9/CX donor cells, the hamster X chromosome was highly active, inducing senescence in 100% of the colonies obtained after its transfer into the nickel-transformed cells. However, senescence was reduced to 50% when Chinese hamster X chromosomes were transferred from later passage A9 cells. Full senescing activity of the intact hamster X chromosome was restored by treatment of the donor mouse cells with 5-azacytidine, which induced demethylation of DNA. These results suggest that a senescence gene or genes, which may be located on the Chinese hamster X chromosome, can be regulated by DNA methylation, and that escape from senescence and possibly loss of tumor suppressor gene activity can occur by epigenetic mechanisms.

Induction of cellular senescence in immortalized cells by human chromosome 1.

The control of cellular senescence by specific human chromosomes was examined in interspecies cell hybrids between diploid human fibroblasts and an immortal, Syrian hamster cell line. Most such hybrids exhibited a limited life span comparable to that of the human fibroblasts, indicating that cellular senescence is dominant in these hybrids. Karyotypic analyses of the hybrid clones that did not senesce revealed that all these clones had lost both copies of human chromosome 1, whereas all other human chromosomes were observed in at least some of the immortal hybrids. The application of selective pressure for retention of human chromosome 1 to the cell hybrids resulted in an increased percentage of hybrids that senesced. Further, the introduction of a single copy of human chromosome 1 to the hamster cells by microcell fusion caused typical signs of cellular senescence. Transfer of chromosome 11 had no effect on the growth of the cells. These findings indicate that human chromosome 1 may participate in the control of cellular senescence and further support a genetic basis for cellular senescence.

Role of a tumor-suppressor gene in the negative control of anchorage-independent growth of Syrian hamster cells.

Tumor-suppressor genes control the neoplastic phenotype of tumor cells, but the function of these genes in normal cells is unknown. In this report we show that the loss of a tumor-suppressor gene function releases negative controls on the growth of cells in agar. This conclusion is based on observations of cell hybrids and studies of cell variants that have retained or lost a tumor-suppressor gene function. Nontumorigenic cell hybrids between normal Syrian hamster embryo cells and a benzo[a]pyrene-transformed tumor-cell line (BP6T) continued to secrete autocrine and/or paracrine growth factors produced by the tumor cells but failed to respond to these factors by growing in agar. Normal diploid cells also failed to grow in agar in response to the growth factors produced by the tumor cells. Clonal variants of nontumorigenic, immortal Syrian hamster cell lines were isolated that either retained (termed supB+) or had lost (termed supB-) the ability to suppress tumorigenicity of BP6T tumor cells after cell hybridization. Neither supB+ nor supB- variants grew in agar under conditions that allowed efficient growth of the tumor cells. However, supB- cells were reversibly induced to grow in agar with high colony-forming efficiencies in the presence of tumor cell-conditioned medium or by supplementation of the medium with a combination of growth factors. Under the same conditions, the supB+ cells failed to grow in agar. This enhanced growth-factor responsiveness in agar was used to select for supB- variants existing at a low frequency in the supB+ population. These two phenotypes, loss of tumor-suppressor function and enhanced growth-factor responsiveness in agar, were seen to cosegregate. These results indicate the tumor-suppressor gene function in these cells negatively regulates the growth response of cells in agar to mitogenic stimuli. This growth regulation may depend on cell shape or adhesion because supB+ and supB- cells grown attached to plastic responded similarly to growth factors.