To properly judge the importance of heterozygous mutations in TP53, an appropriate model for its analysis is required. Heterozygosity of TP53 mutations is the basis of the dominant-negative effect. For the most part, the analyses of this phenomenon have been performed in artificial systems and using animal models. However, co-transfection of wild-type and mutated cDNA is artificial and does not allow for the analysis of all aspects of gene function, and animal models do not represent exact human counterparts [7]. To this end, the analysis of human cancer cell lines with single heterozygous TP53 mutations could be an alternative. However, this approach features some caveats as well. Despite the fact that a percentage of surgical and biopsy specimens that are described as harbouring a single heterozygous mutation is very high (35%), the proportion of cultured cancer cell lines supposedly carrying such a mutation is lower (10%) (Figure 1A, 1B). Moreover, based on our analysis, even this number appears to be overestimated. These differences should be investigated in order to estimate the importance of the dominant-negative effect of TP53 mutations. A higher percentage of single heterozygous mutations indicates a greater importance of dominant-negative effect, as has been shown for genes such as IDH1 or EGFR [13, 14].
First, genetic heterogeneity can sometimes be mistaken for the heterozygous status of the TP53 gene. In cancer specimens exhibiting a TP53 mutation, cells without a mutation are present as well. TP53 is mutated at the end of the mutation pathway, and many cancer cells without a mutation in this gene can be observed within the specimen [22–25].
Another explanation for the discussed discrepancies may be the more thorough genetic analysis of TP53 gene sequences in cell lines compared to tumour samples, which implies that cell line studies are more reliable in this respect. For the most part, tumour samples (especially where paraffin blocks are concerned) do not provide a sufficient amount of high-quality DNA to repeatedly analyse all TP53 gene elements. The percentage of single heterozygous mutations is lower than suggested by in vivo analysis, and thus, the importance of the DNE of TP53 mutations is undermined. Database analysis supports such a possibility. Interestingly, cell lines exhibit a lower proportion of wild-type TP53 retention, not merely because of the more frequent 17p LOH, but also due to the more frequent second heterozygous TP53 mutation (Figure 1C). The percentage of samples with at least one mutation outside exons 5-8 is higher in cancer cell lines with at least two mutations than in cancer surgical samples with two mutations (42% in cell lines and 31% in tumour samples) (Figure 1D). Although the difference is not statistically significant, the investigation of this discrepancy leads to the suspicion that some of the mutations outside region 5-8 were not detected due to negligence during the analysis of surgical and biopsy samples. It has been confirmed in many cell line analyses that the first examination is not sufficiently thorough, whereas surgery/biopsy sample analyses are rarely repeated [26–28]. The TE-3 cell line is a good example of this issue. In the first paper, the lack of a mutation was suggested [26]; however, an additional analysis revealed a homozygous splice-site mutation in intron 4 of TP53 [27]. Importantly, TE-3 cells were initially shown as lacking the TP53 protein [26]. Apparently, an undetected homozygous nonsense mutation was the reason for the absence of TP53 protein. This problem can also be demonstrated using the ST-486 cell line. Originally described as exhibiting one heterozygous mutation, ST-486 has currently been redefined as harbouring two mutations [28]. Our analysis confirmed the presence of the two mutations in ST-486 cells (Figure 3B). The data presented here suggest the overestimation of the proportion of heterozygous TP53 mutations, both in vivo and in vitro. Notwithstanding, this information cannot fully account for the discussed (in vivo - in vitro) TP53 discrepancies, which are crucial for estimating the dominant-negative effect of TP53 mutations. Intriguingly, the list of cell lines showing a single heterozygous mutation of TP53 is diminishing. The process of cell line reclassification is currently very vigorous. The Sanger Database features many cell lines that have changed status within the last 2 years, from having a single heterozygous mutation to exhibiting more than one mutation (vide cell lines: CMK, Cha-Go-K-1 and NCI-H661). It is highly unlikely that each pair of double mutations affected the same allele. Moreover, some cell lines are reported differently between databases: MOLT-16, KM-12, SK-LMS-1, J-82, Daudi and Raji. Notably, the LS-123 cell line, which is still catalogued as heterozygous, was provided to us with a hemizygous TP53 mutation (Figure 3C). The reasons for these on-going changes in the databases are not clear because we are unable to determine whether the late discovery of the second mutation is the result of its generation in vitro, selection of an originally small subpopulation of cells with this mutation in vivo, experimental errors or technical inaccuracies. The difficulty in discriminating between the first two potential explanations will be the subject of the following paragraphs.
The in vitro selection of cells exhibiting a hemizygous mutation "hidden" within cells with a heterozygous mutation and the generation of 17p LOH or of a new point mutation in the other allele (initially/in vivo wild-type allele) may be vital in elucidating the lack of a single TP53 heterozygous mutation in cell lines (Figure 6). Our experimental analyses addressed both issues.
Data published by Lozano et al. and their analysis of the H-318 cell line, which showed ROH of 17p, support the latter scenario [21]. Nevertheless, our investigation demonstrated the lack of the wild-type DNA template (Figure 3A). These results suggest the elimination of the wild-type allele during culturing. The conclusions of Lozano et al. are modestly supported by the analysis performed by Boyle et al. The loss of the wild-type allele in vitro was well defined during the examination of fibroblasts derived from patients with Li-Fraumeni syndrome [29]. These cells originally showed only a heterozygous mutation, whereas after 15-20 passages, colonies with 17p LOH were detected. Subsequently, the cells with 17p LOH became dominant [29]. The analysis of Li-Fraumeni patients' fibroblasts supports Lozano's conclusion; however, it is not certain whether the H-318 cell line derives from the selection for cells with a hemizygous mutation or the generation of 17p LOH in vitro.
The phenomenon of the selection for cells with a hemizygous mutation was exemplified by the derivation of the G-16 cell line by our laboratory. This analysis showed the presence of at least two neoplastic compartments, one with the heterozygous and the other with the hemizygous mutation of TP53. This strongly suggests that selection for cells with the hemizygous mutation, as opposed to its de novo generation, was responsible for the discrepancies between the surgical sample and the established cell line. The aforementioned selection of cells without the wild-type TP53 allele indicates that the dominant-negative effect of TP53 mutations is less important than is generally suggested. Clearly, a heterozygous mutation is not as influential as a hemizygous, homozygous or double heterozygous mutation. In some cases, we were not able to precisely define whether the mutation was generated in vitro or selection had occurred. Nevertheless, the cell lines described in the databases as carrying a single heterozygous mutation appeared very unstable in this state or were potentially misclassified. This may be exemplified by the analysis of PF-382 cells. Our study of this cell line revealed two mutations in codon 273 (Figure 5), although the databases indicate only a single mutation. One of these mutations was originally observed only in a subpopulation of cells. We could not exclude that a subpopulation of cells exhibiting two mutations was already present in vivo and was later ignored during the original DSMZ investigation. Alternatively, the second mutation may have been generated during the production of cell culture stocks. SSCP analysis and cell cloning for PF-382 cells support the conclusion that both mutations in codon 273 affected different alleles.
The analysis of another cell line demonstrated similar results. The MOLT-13 cell line was provided to our laboratory by DSMZ, and only one heterozygous mutation in codon 273 of the TP53 gene was detectable in the early passages. After 2 months of culturing under standard conditions, the second mutation was detected during a routine screening (Figure 4B), which prompted us to restart the culture of the original MOLT-13 cells. Culturing these cells for only four weeks produced cells with both mutations, which established that the presence of cells with the two mutations was the result of selection in our laboratory, rather than de novo generation. The data provided by DSMZ did not allow us to exclude that the second mutation was not present in the original patient sample and was generated during the production of vendor stocks. In any case, if generation or selection occurs in vitro, the importance of the dominant-negative effect is undermined by the instability of single heterozygous mutations, as well as the enhanced in vitro survival of cells exhibiting a lack of wild-type TP53.
The data presented above show that although surgical specimens are analysed less precisely than cell lines, single heterozygous mutations are still observed more frequently in vivo than in vitro. The G16, MOLT13-boost and PF-382 cell lines consist of cells that were selected from subpopulations arising in vivo (in the case of the MOLT13-boost minor subpopulation), whereas H-318 cells probably acquired the 17p LOH in vitro (adaptation to in vitro conditions and further stages of tumorigenesis). Thus, it may be presumed that it is possible to detect many biological differences between cells observed in vivo and in vitro in terms of TP53 status. For example, artificial selection forces acting in vitro may change the TP53 status. Alternatively, rapid selection and generation of subpopulations of the more advanced neoplastic cells may be observed under such conditions. The latter interpretation would favour in vitro conditions as selecting the most effective impairment of TP53, i.e., the selection of cells at more advanced stages of carcinogenesis.
Nonetheless, an effective DNE would not be easily replaced by elimination or impairment of the second allele under artificial or in vivo selection pressure. Moreover, it is very unlikely that a mechanism that is sufficiently effective in vivo would become utterly suppressed in vitro. In general, in vitro selection and generation of cells presenting hemi/homozygous mutations undermines the importance of dominant-negative inactivation of TP53. Still, we cannot exclude that in vitro cell culture does not optimally recapitulate all aspects of tumorigenesis involving TP53.
We are aware that DNE has been presented many times for some of the TP53 mutations. Our analysis suggests that DNE TP53 mutations, which are albeit limited in general, differ with respect to specific mutations. First, the mutation R175H seems to exhibit a weaker DNE than mutations in codon 248. Furthermore, our results are very intriguing for DNE mutations in codon 273 (R273L, R273P, R273H) in comparison to R273C, a mutation exhibiting a lack of DNE, as demonstrated by Dearth et al. In part, it could be expected that the percentage of single heterozygous mutations correlates with the magnitude of DNE and the importance of specific dominant-negative mutations defined by Dearth et al. Namely, the percentage of mutations in codon 273 defined by Dearth et al. as exhibiting a DNE was higher in vivo than this number for the R273C mutation (42% vs. 23%). On the other hand, the percentage of single heterozygous R273C mutations was 4 times higher in vitro than other mutations in this codon (20% vs. 5%), undermining the importance of TP53 mutations DNE in vitro. Nevertheless, segregating TP53 mutations into hot spots changes the overall results very little. None of the hot spots exhibit a percentage of single heterozygous mutations higher than 50% (Figure 1F), bearing in mind that these values are overestimated according to our analyses. Still, the data do not imply a complete lack of DNE. It may be inferred that DNE is a mechanism operating solely in vivo, which, however, remains in contradiction with the DNE model. Moreover, other tumour suppressor genes, such as APC and Rb, also show similar differences between the frequency of single heterozygous mutations in vivo and in vitro [17, 18]. Obviously, APC and Rb exhibit nonsense mutations, and DNE is not attributed to these genes, whereas epigenetic silencing is [30–33]. However, considering the latter observation, another analysis of the database is very appealing. The retention of the wild-type allele in cancer cell lines with a missense TP53 mutation is almost the same as the frequency amongst cancer cell lines with a nonsense mutation (Figure 1E). This indicates that a missense mutation does not predispose cells to retain the wild-type TP53 allele any more than does a nonsense mutation, whereas a dominant-negative effect is expected to be the attribute of missense mutations only. In our opinion, this suggests that the dominant-negative effect can be confused with gain of function and even mistaken for an unknown mechanism of wild-type TP53 inhibition if a single heterozygous mutation occurs. Our group has already demonstrated the predominance of mutated mRNA over wild-type mRNA in glioblastoma specimens showing putative single heterozygous mutation of TP53 [20].