Human thyroid cancer cells as a source of iso-genic, iso-phenotypic cell lines with or without functional p53

Differentiated thyroid carcinomas (in contrast to the rarer anaplastic form) are unusual among human cancers in displaying a remarkably low frequency of p53 mutation and appear to retain wild-type (wt) p53 function as assessed by the response of derived cell lines to DNA damage. Using one such cell line, K1, we have tested the effect of experimental abrogation of p53 function by generating matched sub-clones stably expressing either a neo control gene, a dominant-negative mutant p53 (143ala) or human papilloma virus protein HPV16 E6. Loss of p53 function in the latter two groups was confirmed by abolition of p53-dependent ‘stress’ responses including induction of the cyclin/CDK inhibitor p21WAF1 and G1/S arrest following DNA-damage. In contrast, no change was detected in the phenotype of ‘unstressed’ clones, with respect to any of the following parameters: proliferation rate in monolayer, serum-dependence for proliferation or survival, tumorigenicity, cellular morphology, or tissue-specific differentiation markers. The K1 line therefore represents a ‘neutral’ background with respect to p53 function, permitting the derivation of functionally p53 + or − clones which are not only iso-genic but also iso-phenotypic. Such a panel should be an ideal tool with which to test the p53-dependence of cellular stress responses, particularly the sensitivity to potential therapeutic agents, free from the confounding additional phenotypic differences which usually accompany loss of p53 function. The results also further support the hypothesis that p53 mutation alone is not sufficient to drive progression of thyroid cancer to the aggressive anaplastic form. © 1999 Cancer Research Campaign

Mutation of the tumour suppressor gene p53 leading to loss of function (with or without dominant-negative activity), is wellestablished as a key event in human tumorigenesis (Greenblatt et al, 1994). The high frequency of mutation observed in most cancers clearly reflects a strong selective advantage, which is supported by both in vitro (Michalovitz et al, 1991) and in vivo transgenic experiments (Lavigeur et al, 1989;Kemp et al, 1993). Indeed, many cancer cells appear to be totally dependent on a loss of p53 function for survival Diller et al, 1990;Johnson et al, 1991).
It is clear, however, that with a few possible exceptions such as ultraviolet (UV)-induced skin cancer (Zeigler et al, 1994), p53 mutation is not an initiating event and only confers a selective growth advantage once a critical stage in the clonal evolution of a tumour has been reached Kemp et al, 1993). This correlates loosely with increasing clinical ÔstageÕ and pathological ÔgradeÕ (Fujimoto et al, 1992;Barnes et al, 1993;Navone et al, 1993), but the exact nature of the step that converts a p53-insensitive to a p53-sensitive tumour clone remains unclear, as indeed does the biochemical basis for selective growth suppression of such tumour cells by wild-type (wt) p53 (Wynford-Thomas, 1996;Wynford-Thomas et al, 1996).
Tumours of the thyroid follicular cell provide a particularly useful model for studying this question. Instead of the usual continuous spectrum, in these tumours there is a remarkably clearcut demarcation between two clinico-pathological entities: i) differentiated cancers (the majority) which usually carry a very favourable prognosis; and ii) undifferentiated (anaplastic) cancers which are among the most malignant of all human cancers (Carcangiu et al, 1985;Williams and Williams, 1989). Furthermore, the rate of transition between these phenotypes appears to be very low, such that most differentiated cancers can metastasize and (in the absence of therapy) generate a potentially fatal tumour burden without ever undergoing anaplastic transformation. This ÔdiscontinuityÕ in progression correlates with an equally striking difference in frequency of p53 mutation, from virtually zero in differentiated cancers (whether primary or metastatic) to well over 50% in anaplastic tumours (Ito et al, 1992;Fagin et al, 1993;Wynford-Thomas, 1993). The magnitude of this contrast, coupled with the stability of the differentiated stage, enabling homogeneous clinical samples, and even cell lines, with wt p53 phenotype to be obtained, makes this a unique model to investigate the nature of the selection pressure driving p53 mutation in human cancer.
The correlation between p53 mutation and the anaplastic phenotype could most readily be explained if loss of p53 function led directly to the increased tumour aggressiveness and loss of differentiation characteristic of this progression step. Evidence from our laboratory, however, indicates that this simple scenario is not correct. Using a cell line derived from a differentiated thyroid cancer (K1), in which wt p53 appeared to be functional in terms of the response to ÔacuteÕ stimuli from DNA-damaging agents such as bleomycin , we previously reported that stable expression of a dominant-negative mutant of p53 (143 ala ), while abrogating the G1/S cell cycle check-point, had no other obvious phenotypic effect Wyllie et al, 1995). This is a potentially important finding which deserves further investigation, since it supports the hypothesis that progression from differentiated to anaplastic thyroid cancer requires events additional to loss of p53 function. Furthermore, we reasoned that if K1 cells were indeed indifferent to the presence or absence of wt p53 function, they would represent an ideal, ÔneutralÕ, background on which to derive matched functionally p53 + or Ð lines for use as a tool to investigate p53-dependency of therapeutic agents.
Our earlier study  examined only a few phenotypic features, however, and employed just a single p53 mutant which has subsequently been found to be an inefficient dominant-negative in some contexts (Williams et al, 1995). We have therefore now re-examined this model in more detail, analysing multiple clones, multiple parameters of proliferation and differentiation state, and using not only mutant p53 but also the human papilloma virus protein HPV16 E6 as a means of abrogating p53 function.

Cell culture
The K1 and FTC human thyroid cancer cell lines were kindly provided by Prof. M Schlumberger (Villejuif, France) and Dr P Goretski (Dusseldorf, Germany), respectively. Both lines and their derivatives were grown as monolayers in a 2:1:1 (by volume) mixture of DulbeccoÕs modified EagleÕs medium, HamÕs F12, and MCDB104 (all from Life Technologies, Paisley, UK) (Bond et al, 1992), supplemented with 10% fetal calf serum (FCS; Imperial Labs, London, UK).

DNA transfection
Cells were plated at 2 × 10 5 per 60-mm dish and transfected 2 days later by the strontium phosphate coprecipitation method (Brash et al, 1987). The plasmids used were pc53-SCX3 which expresses a human mutant p53 (143 ala ) ) and pSV2neo as a negative control (Southern and Berg, 1982). Stable transfectants were selected by growth in 400 µg ml Ð1 G418 (Life Technologies).

Retroviral gene transfer
A high-titre amphotropic retroviral vector (PA317-16E6) (Halbert et al, 1991), expressing the E6 gene from human papilloma virus type 16, was kindly provided by Dr D Galloway (Seattle, WA, USA). As a negative control we used a neo-only vector of similar titre, psi-CRIPneo (Wyllie et al, 1993).

ELISA
Cellular p53 protein content was determined by enzyme-linked immunosorbent assay (ELISA) using a Ôpan-p53Õ kit (Oncogene Science, New York, NY, USA) as described previously .

DNA damage
Bleomycin (Lundbeck, Milton Keynes, UK) was prepared as a 7.5 mg ml Ð1 stock solution in de-ionized water, stored at Ð20°C and used at a final concentration of 250 µg ml Ð1 , which was previously determined as the minimal dose required to obtain maximum growth arrest in these cells . The timing of analysis of p21WAF1 expression following bleomycin treatment was based on previous time-course studies (Bond et al, 1995).

Flow cytofluorimetry
This was performed as described previously .

Northern blot analysis
Total RNA was extracted from cells in log-phase growth by a modified guanidium/phenol method (RNAzol B; AMS Biotechnology, Witney, UK), separated on 1% agarose gels (10 µg per lane) and blotted to Hybond N + (Amersham, Aylesbury, UK). As a positive control for thyroid-specific transcripts, RNA was similarly extracted from thyroid tissue derived from a patient with GraveÕs disease. The following DNA probes were used: for PAX-8, a 0.3 kb HindIII/EcoR1 insert of the human cDNA clone, H26P/S3 (Poleev et al, 1992); for TTF1, a 0.7 kb Sac1 insert of the rat cDNA clone, TTF1-THA (Guazzi et al, 1990); for thyroglobulin (TG), a 0.98 kb Pst1 insert of the human cDNA clone, phTg1 (Brocas et al, 1982); for thyroid peroxidase (TPO), a 3 kb EcoR1 insert of the human cDNA clone, pTPO (Libert et al, 1987); for the thyrotropin receptor (TSHR), a 2.3 kb BamH1/Xbo1 insert of the human clone pSVL (Libert et al, 1989) and for GAPDH (control), a 1.2 kb Pst1/Xba1 insert of the human cDNA clone, pHcGAP (ATCC, Rockville, MD, USA). Filters were washed at high stringency (65°C; 0.1 × SSPE) for all probes except for TTF1, which was washed at low stringency (65°C; 1 × SSPE), and were stripped and rehybridized to the GAPDH probe as a ÔloadingÕ control. Denmark) to the medium for 1 h at a final concentration of 10 µM. After labelling, cells were fixed in 70% ethanol (30 min at 4°C), then pretreated with 4 M HCl (10 min) and 0.1 M borax, pH 8.5 (5 min). Incorporated BrdU was detected by an immunoperoxidase method incubating first for 60 min with a mouse anti-BrdU primary antibody (Dako) in the presence of 25 U ml Ð1 DNase1 (Life Technologies), followed by peroxidase labelled rabbit antimouse immunoglobulin (Dako) and finally diaminobenzidine substrate. After counterstaining with haematoxylin, the proportion of positive nuclei (LI) was determined, using a sample of > 1000 cells per data point.

Colony forming efficiency
A single cell suspension was obtained after trypsinization and 10 2 cells were added to duplicate 60-mm dishes. After 10 days in standard growth conditions Giemsa-stained colonies of 50 cells or more were counted, and colony forming efficiency (CFE) expressed as a % of cell number plated.

Analysis of cell death
Terminal deoxynucleotidyl transferase assay Cells were seeded on coverslips and fixed with 4% paraformaldehyde (30 min). Endogenous peroxidase was blocked with 0.3% hydrogen peroxide in methanol (30 min) and the cells permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate (15 min). Cells were then incubated for 1 h at 37°C with 250 U ml Ð1 terminal deoxynucleotidyl transferase (TdT; Promega, Southampton, UK) and 1 nmole biotinÐ16ÐdUTP (Boehringer Mannheim, Lewes, UK). Sites of biotinÐ16ÐdUTP localization were visualized using the mouse specific avidinÐbiotinÐperoxidase system (Novocastra). Cells were counterstained with haematoxylin and the proportion of apoptotic (brown) cells assessed, in samples of > 1000 cells per data point.
Conventional gel electrophoresis DNA was extracted according to Wyllie et al (1989) from a combination of free-floating and loosely-attached cells, electrophoresed in a 1.5% agarose gel for 1Ð2 h at 90 V and visualized by ethidium bromide staining.
Tumorigenicity 10 6 cells, suspended in 0.2 ml of growth medium were injected subcutaneously into athymic (nude) mice (1 injection per animal) and monitored for appearance of tumours for up to 8 months.

Derivation of sub-clones of the thyroid cancer line K1 expressing mp53
The K1 cell line (Wyllie et al, 1993;Challeton et al, 1997) was derived by spontaneous immortalization from the most common pathological sub-type of differentiated thyroid cancer Ð ÔpapillaryÕ carcinoma . Conventional immunocytochemical (ICC) analysis with PAb421 showed low levels of nuclear p53 and sequencing of the entire coding region (exons 2Ð11) from reverse transcribed mRNA confirmed the presence of only wt sequence .
To directly assess the effect of loss of wt p53 function on the behaviour of K1 cells, sub-clones (designated K1scx) were derived by stable transfection with plasmid pc53-SCX which express the ala 143 mutant of human p53. Control sister clones (K1neo) were obtained using plasmid pSV2neo. Four K1scx clones (K1scx3, 6, 8 and 9) were initially chosen on the basis of their much higher expression of total p53 protein as shown by immunocytochemistry ( Figure 1B), when compared to the four K1neo control clones (3, 4, 5 and 11) ( Figure 1A). (Mutant p53 is known to be stabilized in these cells.) This was confirmed by ELISA assay which showed p53 protein content in clones scx3, 6 and 8 to be 175, 222 and 164 µg mg Ð1 total cellular protein, respectively, whereas in K1neo clones it could not be reliably detected above the detection limit (approximately 5 µg mg Ð1 ). For comparison, FTC133, a transformed thyroid cell line known to contain a homozygous mutant p53 gene (Wright et al, 1991), contained 80 µg p53 per mg total protein.

Functional status of p53 in K1scx subclones
p53 function in K1scx clones was assessed by determining the integrity of p53-dependent DNA damage responses. p53 has been identified as a critical component of a signalling pathway leading to G1 arrest following exposure to ionizing radiation (Kastan et al, 1991). The status of this checkpoint was therefore assessed in K1neo and K1scx clones by analysing the proportions of cells in G1, S and G2/M cell cycle phases following treatment with the radiomimetic agent, bleomycin (Hsu et al, 1989;Wyllie et al, 1995).
In K1neo clones, flow cytometry showed no significant reduction in the G1 fraction up to 31 h after bleomycin treatment, by which time most of the cells that were initially in S phase had progressed into, and remained in, the G2 fraction ( Figure 2). This is consistent with retention of both G1 and G2 cell cycle checkpoints. In contrast, all four K1scx clones showed a major reduction of the G1 fraction by 31 h after bleomycin treatment, with most of the cells having moved into the G2 compartment (Figure 2), indicating retention of the G2/M, but loss of the G1/S checkpoint.
One of the downstream transcriptional targets of p53 responsible for this G1 arrest is the cyclin kinase inhibitor, p21WAF1 and cells that have lost functional p53 fail to induce expression of p21WAF1 in response to some forms of DNA damage (El-Deiry et al, 1994). Intranuclear p21WAF1 protein content in K1neo and K1scx clones was therefore assessed immunocytochemically before and after bleomycin treatment.
Untreated K1neo clones showed marked cellÐcell heterogeneity with an average 4.5 ± 0.9% of nuclei (mean ± SE for the four clones tested) showing detectable immunostaining ( Figure 3A). Bleomycin treatment led to a clear increase in expression, in terms both of the proportion of positive nuclei (which increased to 21.9 ± 1.7%) and the intensity of immunostaining ( Figure 3B and Table 1). Untreated K1scx cells showed a very similar level of immunostaining to K1neo cells (average of 4.4 ± 1.5% for the four clones; Figure 3C). In contrast, however, in K1scx cells (all four clones), no induction of expression could be detected in treated cultures either in proportion positive (3.4 ± 0.4% for the four clones) or intensity ( Figure 3D and Table 1).

Derivation of K1 lines expressing HPVE6
Although the above data supported the effectiveness of our ala143 mutant as a dominant-negative, there remained some concern from other studies (Forrester et al, 1995;Friedlander et al, 1996;Ludwig et al, 1996) that some p53 targets may still be activatable in K1scx cells, due either to their greater sensitivity to any residual wt p53 function, and/or to retention of transcriptional activity by the mutant for a sub-set of p53-dependent promoters.
We therefore also used an independent approach for abrogation of wt p53 function in K1 cells, by expression of the HPV16E6 gene introduced by means of an amphotropic retroviral vector, pLXSNE6 (Halbert er al, 1991). Five G418 resistant clones were chosen at random together with four control lines expressing only a retrovirally-transduced neo gene and p53 functional status was assessed as above by measuring p21WAF1 induction in response to bleomycin treatment.
The neo controls behaved, as expected, like their transfected counterparts, the proportion of p21-positive nuclei rising from 5.0 ± 2.0% to 25.1 ± 5.1% following bleomycin treatment (data not shown). However, a clear difference was observed between K1.E6 clones and their K1scx counterparts, in that in untreated cultures no p21-expressing cells could be detected in three out of five K1.E6 clones and only a very low % in the other two clones (the average for all five clones being < 0.1%; Figure 3E). This is consistent with a more complete abrogation of p53 function by E6 than by the 143 ala p53 mutant. As expected, E6 clones showed no induction of p21 expression following bleomycin treatment ( Figure 3F and Table 1).

Effect of inactivating wt p53 function on growth and survival of K1 cells
The doubling time (mean ± SE) in standard medium during log growth was 22.0 ± 1.8 h for the three K1neo clones and 24.3 ± 0.8 h for the three K1scx clones analysed ( Table 2). The proportion of nuclei in S phase (LI) under the same conditions was 34.0 ± 0.75% and 34.6 ± 1.0%, respectively (Table 2). All clones showed a very high plating efficiency, CFE being 60.0 ± 16.2% for K1neo  Table 2). None of these differences between K1neo and K1scx reached statistical significance.
The effect of inactivation of wt p53 function on response to withdrawal of serum growth factors was tested by plating cells in medium supplemented with 10% serum for 48 h and then refeeding with serum-free media. Assessment of DNA synthesis was performed only after 2 days serum starvation since later analysis proved unreliable due to high cell death rates. Most clones showed only a small decrease in LI, apart from K1neo3 which showed an approximately threefold decrease (Table 2). Overall, there was no significant difference in LI between the four K1neo and the four K1scx clones in serum-starved conditions (23.2 ± 5.0% and 29.4 ± 1.75%, respectively) ( Table 2).
Determination of cell number at 0, 2, 5, 7 and 9 days after serum starvation showed an increase in most neo and scx clones up to day 5 and a decrease thereafter. There was considerable inter-clonal variability in the rate of onset of this decrease in cell number with a trend for the K1scx clones to lose cells more quickly than the K1neo clones ( Figure 4). However, this difference did not reach statistical significance. The basis for this cell loss was investigated in one K1neo clone (No. 3) and one scx clone (No. 6). The TdT assay showed an increase in the number of positive cells from < 1% at days 0 or 2, to 2.5Ð3.0% at day 5 and 6.5Ð14% by day 7 and 9 of serum-starvation, consistent with apoptosis being the mode of cell death. This was confirmed by gel electrophoretic analysis of DNA extracted from floating and loosely attached cells obtained from a pool of day 7 and 9 serum-starved cultures, which showed the laddering pattern characteristic of apoptosis (data not shown).

Tumorigenicity
All clones formed visible tumours in athymic mice within 7 months. The median time taken to form 0.5 cm diameter tumours was 4 weeks in K1scx compared to 5 weeks for K1neo. This difference was entirely attributable to a long latency in two of the animals injected with clone neo3 (Table 2) and, overall, did not reach statistical significance.

Morphology
Under standard growth conditions all K1neo clones grew as a monolayer with well defined cellÐcell boundaries and although at least some cells still continued to cycle at confluence (as indicated by the 24-h BrdU LI; data not shown), this did not result in any appreciable Ôpiling upÕ (see Figure 5A for a representative neo clone). The expression of mutant p53 or HPVE6 did not result in any consistent change in morphology (see Figure 5b for a representative scx clone).

Expression of thyroid-specific differentiation
Previous studies have suggested that altering p53 functional status can lead to an alteration in differentiation, as measured by the expression of the thyroid-specific transcription factor, PAX8   Figure 4 Effect of serum-starvation on growth and survival of control and mp53-expressing K1 sub-clones. Cell number is shown for four K1neo clones (solid lines) and four K1scx clones (broken lines) at 0, 2, 5, 7 and 9 days after serum withdrawal (Battista et al, 1995;Fagin et al, 1996). To investigate this relationship in our cell line, the K1scx, K1E6 and K1neo clones were compared and the ratio of PAX8 to control (GAPDH) mRNA expression was calculated for each clone. No significant difference in PAX8 mRNA expression between the four K1neo and the four K1scx clones was observed (mean ratio of 1.0 ± 0.4 and 0.9 ± 0.3, respectively; Figure 6A). A similar result was seen when the five K1E6 clones were compared with their four control neo clones (mean ratio of 1.6 ± 0.7 for E6 and 1.7 ± 0.8 for neo; Figure 6B). For comparison, a poorly differentiated thyroid carcinoma cell line (FTC133), showed no PAX8 expression ( Figure 6A and B).

DISCUSSION
In this study we have used gene transfer in vitro to directly investigate whether loss of wt p53 function is sufficient to drive progression of the transformed phenotype in thyroid cancer cells. Expression of the 143 ala mutant failed to cause any demonstrable loss of differentiation in a cell line derived from differentiated thyroid (papillary) cancer, nor did it confer any evidence for increased proliferative capacity in vitro or in vivo.
There is now good evidence that expression of such mutants can sometimes fail to eliminate function of the endogenous wt protein and, indeed, the effectiveness of their dominant-negative activity appears to be highly dependent on cell context, mutation site (Forrester et al, 1995;Friedlander et al, 1996;Ludwig et al, 1996), and promoter target. In our study, the 143 ala mutant appeared to be effective, at least as evidenced by abrogation of p21WAF1 induction and G1/S arrest in response to DNA damage. However, to provide a complementary, and potentially more reliable, method of p53 inactivation, we also employed HPVE6 expression, which promotes ubiquitination and degradation of p53 protein (Scheffner et al, 1990). Some indication of a more complete loss of p53 function was indeed seen, in that E6 not only blocked the DNA damage response, but also eliminated the low, basal level of p21 expression seen in untreated cells. As with mp53, however, no evidence for progression of the transformed phenotype by E6 could be observed. Other groups have attempted to reconstruct p53-dependent tumour progression in other tumour models, notably with respect to multi-stage colon carcinogenesis, in which p53 mutation appears to play a role in progression of adenomas to carcinomas. Using an approach similar to our own, Williams et al (1995) also failed, however, to demonstrate any phenotypic effect of mp53 expression in an adenoma-derived cell line. Although, with some of the mutants used, this could be explained by incomplete dominant-negative activity, this could not account for the lack of effect of the his 273 mutant. Likewise, in myeloid malignancies, although there is a clear-cut association between p53 mutation and the progression of chronic myeloid leukaemia (CML) to Ôblast crisisÕ, only minor effects could be demonstrated on introduction of the ala 143 mutant into CML-derived cells (Bi et al, 1994). Clearly, the conclusion from all of these studies is that p53 mutation, while necessary, is not sufficient for tumour progression.
While this conclusion is entirely consistent with the role of p53 in control of cell proliferation (Wynford-Thomas, 1996), it was perhaps less predictable in relation to the now substantial body of data supporting a role for p53 in the control of tissue-specific differentiation (Almog and Rotter, 1997).
In addition to correlative evidence based on changes in p53 protein content, or (more importantly) transactivational activity accompanying differentiation (Aloni-Grinstein et al, 1993;Halevy, 1993;Weinberg et al, 1995) there is good experimental evidence that manipulation of p53 activity can modulate the differentiation process. For example, inhibition of wt p53 by dominantnegative mutants has been shown to block differentiation in B-cell (Aloni-Grinstein et al, 1993Halevy, 1993;Weinberg et al, 1995), myeloid (Soddu et al, 1996), and muscle (Soddu et al, 1996) cell models. Conversely, introduction of wt p53 can restore the differentiation pathway to p53-null cells of B-cell (Shaulsky et al, 1991), myeloid (Soddu et al, 1994), erythroid (Feinstein et al, 1992) or epithelial (Brenner et al, 1994) origin. Indeed, two groups have reported similar data in relation to thyroid. Fagin et al (1996) partially restored differentiation (TPO and PAX8 expression) to a poorly differentiated thyroid cancer line by expression of wt p53, and Moretti et al (1997) were successful in restoring TSH-dependent expression of thyroid differentiation genes (Tg and TPO) to the undifferentiated (anaplastic) cancer cell line ARO using a temperature-sensitive p53 mutant (Val 135). However, it should be noted that the first report was restricted to a single clone derived from a line which was not representative of truly undifferentiated thyroid cancer, and in the second study, the absolute levels of transcription achieved were extremely limited (being detectable only by reverse transcriptionÐpolymerase chain reaction).
While in some of the above examples, it is difficult to exclude an indirect effect of p53 via its action on accompanying changes in cell proliferation, there is evidence in at least some models (Soddu et al, 1996) for a specific effect on differentiation, and indeed some data to demonstrate a direct action of wt p53 on transcription of differentiation genes (Aloni-Grinstein et al, 1993).
The above data therefore suggest that p53 function is necessary for completion of the differentiation programme in many cell types. While at first sight this contradicts the apparently normal differentiation of most lineages in p53-null mice, as pointed out previously (Eizenberg et al, 1996), it is entirely possible that redundant controls exist in normal cells in vivo and that it is only when these are lost following in vitro culture and/or transformation that the role of p53 is revealed.
Even in cell line models, however, there is little evidence to show that loss of p53 function can actually reverse differentiation (as opposed to blocking its execution) and that it can thereby contribute directly to loss of tumour cell differentiation. Interestingly, one of the most clear-cut studies in this respect (Battista et al, 1995) was carried out using a rat thyroid epithelial cell line (PCCL3), in which expression of a dominant-negative p53 (143 ala ) was shown to cause loss of expression of thyroidspecific differentiation genes (Tg, TPO, TSHR), along with the thyroid-specific transcription factor PAX8. Furthermore, forced expression of PAX8 in such cells was enough to restore differentiation.
This finding contrasts strikingly with our failure to observe any diminution of PAX8 expression in our human thyroid cell line following expression of mp53 or E6. We can only speculate that the difference in behaviour reflects differences in co-existing genetic abnormalities or in the species of origin of the two cell lines. With respect to the latter, it should be noted that, in common with nearly all human (as opposed to rodent) thyroid cell lines, K1 lack expression of Tg and TPO. This appears to reflect a spontaneous and unavoidable loss of these differentiation markers in monolayer cultures of thyrocytes. Nevertheless, they retain PAX8, which lies upstream of Tg and TPO in the differentiation pathway and represents one of the fundamental determinants of thyroid differentiation. Since this is a key distinguishing feature between differentiated and undifferentiated cancers (Fabbro et al, 1994), we consider K1 to be a useful, although clearly not ideal, model for this purpose.
The conclusions from this work are entirely consistent with our findings in a different experimental model of thyroid tumour development (Bond et al, 1996), in which we observed that expression of SV40T (which abrogates p53 function) is sufficient only to induce differentiated clones with limited proliferative potential. De-differentiated sub-clones with extended lifespan only arose after a variable period of continuous culture, again pointing to a model in which anaplastic progression requires loss of p53 function together with at least one additional genetic or epi-genetic event.
Quite apart from its direct relevance for thyroid tumour biology, the clones described here represent an ideal panel of matched, isogenic cell lines with which to investigate the role of wt p53 in physiological and pathological cellular responses, an approach whose value has recently been emphasized by Weinstein et al (1997). A major limitation of most resources in this area, including the NCI panel  is that the majority of randomly selected human cancer lines do not tolerate the presence of wt p53, hence by definition it is impossible to obtain an isogenic matched line with wt function.
The starting point therefore must be, as in this study, a tumour cell type that usually retains wt function. Furthermore, it should also be indifferent to loss of this activity in terms of its proliferative phenotype in normal culture conditions, so as to minimize the possibility of differences being merely an indirect consequence of loss of p53 function. Finally, multiple independent sister clones, rather than pools, should be examined. This avoids the pitfall that can arise if an unsuspected minor sub-population of cells in the parent line have acquired an additional abnormality that by itself confers no growth advantage but that synergizes with loss of p53 function, allowing it to become the dominant clone in a pool of transfectants.
The panel of p53 +/Ð sister clones described here meets all these criteria and should be a useful tool to investigate the p53-dependence of a wide variety of cellular responses to external stimuli, particularly the sensitivity to chemotherapeutic agents or other novel therapies. Although other wt p53 human cell lines have been used as the ÔparentÕ in such approaches, notably RKO colon, MCF7 breast (Fan et al, 1995), and A2780 ovarian (McIlwrath et al, 1994), to our knowledge this thyroid-derived panel represents the most well-characterized yet available.