The role of UVA-radiation—the major fraction in sunlight—in human skin carcinogenesis is still elusive. We here report that different UVA exposure regime (4 × 5 J/cm2 per week or 1 × 20 J/cm2 per week) caused tumorigenic conversion (tumors in nude mice) of the HaCaT skin keratinocytes. While tumorigenicity was not associated with general telomere shortening, we found new chromosomal changes characteristic for each recultivated tumor. Since this suggested a nontelomere-dependent relationship between UVA irradiation and chromosomal aberrations, we investigated for alternate mechanisms of UVA-dependent genomic instability. Using the alkaline and neutral comet assay as well as γ-H2AX foci formation on irradiated HaCaT cells (20–60 J/cm2), we show a dose-dependent and long lasting induction of DNA single and double (ds) strand breaks. Extending this to normal human skin keratinocytes, we demonstrate a comparable damage response and, additionally, a significant induction and maintenance of micronuclei (MN) with more acentric fragments (indicative of ds breaks) than entire chromosomes particularly 5 days post irradiation. Thus, physiologically relevant UVA doses cause long-lasting DNA strand breaks, a prerequisite for chromosomal aberration that most likely contribute to tumorigenic conversion of the HaCaT cells. Since normal keratinocytes responded similarly, UVA may likewise contribute to the complex karyotype characteristic for human skin carcinomas.
The incidence of skin cancer has rapidly increased during the last decades, making skin cancer—comprised of basal cell carcinoma (BCC), squamous cell carcinoma (SCC) and malignant melanoma (MM)—the most frequent cancer worldwide (reviewed in Boukamp, 2005b). It is now well accepted that UV radiation is the major risk factor, and that cumulative UV damage is responsible for the development and progression of SCCs, while BCCs and especially MM depend more on intermittent UV overexposure (Armstrong and Kricker, 2001). From the three UV components—UVA (wavelengths 320–400 nm), UVB (280–320 nm) and UVC (200–280 nm)—only UVB and UVA radiation reach the earth while UVC radiation is fully absorbed by the ozone layer. Similarly, most of the UVB radiation is absorbed. Consequently, the UV radiant energy reaching the earth consists of only ∼5% UVB and ∼95% of UVA.
UVB radiation is thought to be the major risk factor for human skin carcinogenesis because its energy is sufficient to enter the nuclei of epidermal cells and to directly damage DNA by causing cyclobutane pyrimidine dimers (CPDs) and pyrimidine(6-4)pyrimidine photoproducts at dipyrimidine sites, thereby giving rise to very specific mutation patterns. Such C to T transitions are particularly frequent in the p53 tumor suppressor gene that is mutated in 50–90% of the human SCCs (summarized in Boukamp, 2005b).
Recently, UVA radiation also gained more interest. Although the photon energy may be insufficient to damage DNA directly, UVA radiation is able to penetrate into the epidermis down to the proliferative basal cells and even further to the dermal compartment where it contributes to photoaging. In addition, evidence is increasing that UVA may also be relevant for skin cancer development. Albino hairless mice exposed to UVA irradiation (330 nm) developed tumors after >400 days and even long-wave UVA irradiation (>340 nm) has been found to induce skin carcinomas (for review see de Gruijl, 2002). Furthermore, the immortal human HaCaT keratinocytes became tumorigenic when irradiated with 24 J/cm2 UVA for 18 weeks (He et al., 2006).
The action of UVA is thought to be indirect by activating cellular chromophores and inducing photo-oxidative cellular stress (Wondrak et al., 2006). The thereby generated reactive oxygen species (ROS) can modify guanine to 8-oxo-7,8-dihydro-7′-deoxyguanosine (8-oxo-guanine) resulting in transversion mutations by changing guanine to thymine during replication (Cadet et al., 2005)—now considered as UVA signature mutations. ROS are also able to damage DNA most preferentially by inducing DNA-protein cross-linking, alkali labile sites as well as CPDs. This said, Ono and co-workers showed that longer wavelength UVA (310–340 nm) contributes significantly to induction of mutations and appearance of hotspots at cytosine phosphate guanine-associated dipyrimidine sites (Ikehata and Ono, 2007). Furthermore, mutations in UVA-exposed epidermis (including wavelength bordering UVB) were dominated by C to T transitions thus making broadband UVA a considerable contributor to UV-dependent mutations (Ikehata et al., 2003, 2004). In addition, ROS are involved in generating DNA single strand (ss) breaks (Kielbassa et al., 1997). If ss breaks are not repaired during the G1 phase of the cell cycle they can generate double strand (ds) breaks during S-phase which then can result in chromosomal aberrations such as amplifications, deletions or translocations (for review see Limoli et al., 2002; Thompson and Limoli, 2003); all hallmarks of skin carcinomas. Interestingly, CPDs can also cause ds breaks in S-phase (Garinis et al., 2005), suggesting that they too may contribute to chromosomal instability.
To determine the functional consequence of UVA radiation on human skin keratinocytes, we utilized the immortal human HaCaT cells (Boukamp et al., 1988). These cells represent an early stage of skin carcinogenesis because they harbor UVB type-specific p53 mutations (Lehman et al., 1993) as well as some chromosomal abnormalities typical for human skin SCCs (Lehman et al., 1993; Boukamp et al., 1997; Popp et al., 2002). While the HaCaT cells remained nontumorigenic through long-term passaging, they became tumorigenic by transduction with the Harvey-ras oncogene, increased temperature or stroma modulating growth factors known to be relevant for tumorigenicity (Boukamp et al., 1990b, 1995, 1997, 1999; Skobe and Fusenig, 1998; Obermueller et al., 2004). Thus, the intrinsic stable nontumorigenic phenotype, together with the sensitivity for tumorigenic conversion upon distinct changes, make HaCaT cells a highly suitable model to study the role of UVA in skin carcinogenesis.
Exposing HaCaT cells to UVA radiation, we now show that UVA is able to induce tumorigenicity upon different treatment regimes. Since tumorigenicity was associated with new chromosomal aberrations, we additionally explored the immediate damage response to UVA radiation and determined its role in inducing DNA ss and ds breaks and the formation of micronuclei as indicators of chromosomal aberrations. We also show that normal human skin keratinocytes exhibit a similar damage response pattern as HaCaT cells, thus providing evidence for a role of UVA-dependent chromosomal damage in human skin carcinogenesis.
UVA-induced tumorigenic conversion of the HaCaT cells
In a first exploratory experiment we irradiated HaCaT cells with a steadily increasing dose (from 7 J/cm2 during the first week up to 42 J/cm2 during the last week) of UVA radiation over a period of 42 days. This treatment caused cell death of almost all cells except one colony that was further expanded. Different from the control HaCaT cells, nodules developed in four of the eight injection sites upon injection into nude mice. Of these, two regressed while the other two continuously expanded and were classified as invasively growing well-differentiated SCCs (Supplementary Figures 1A and B). Recultivation of one of these SCCs and its genetic comparison (multiplex fluorescence in situ hybridization; M-FISH) with the control HaCaT cells clearly demonstrated its HaCaT origin (containing the original HaCaT marker chromosomes) and four new chromosomal aberrations (Supplementary Figure 1C and D). Being present in all metaphases, the new aberrations clearly document the clonal origin of the tumor.
Tumorigenic conversion is induced upon defined UVA treatment regimes
Since tumorigenicity had occurred upon UVA treatment, we next asked whether distinct and physiological relevant doses would be sufficient for tumorigenic conversion. We applied a dose of 20 J/cm2 which equals a 1 h exposure around noon in mid Europe (Federal Office for Radiation Protection, 2006) to two parallel cultures of HaCaT cells (series A and B for each treatment) either as a single dose once per week—allowing a recovery time of 6 days—or as a split dose of 5 J/cm2 four times per week thereby mimicking continuous low-dose UVA exposure. Different from the exploratory experiment, toxicity was negligible; the cells continued to proliferate and required passaging once a week.
All cells were tested for tumorigenicity after 5 and 15 weeks of continuous UVA treatment. With the exception of one rapidly expanding nodule (5-week time point) which by histology was identified as a water- and blood-filled epidermal cyst (data not shown), the control cells remained nontumorigenic (Figure 1a). Series A cells irradiated with 4 × 5 J/cm2 per week also remained nontumorigenic (data not shown). Small nodules occurring shortly after injection regressed within 2 months. However, series B cells irradiated with 4 × 5 J/cm2 (4 × 5 J/cm2 B) developed tumors. After 5 weeks, tumor growth occurred in 2/6 and after 15 weeks of continuous UVA treatment in 5/8 injection sites (data not shown and Figure 1b). Treatment of 20 J/cm2 once a week caused tumor growth in both series. Series A cells formed tumors in 3/6 injection sites already after 5 weeks treatment (data not shown). Unfortunately, these cells were then lost due to infection. Series B cells (1 × 20 J/cm2 B) developed one tumor (1/6 injection sites) after 5 weeks while five tumors formed in the eight injection sites after 15 weeks (data not shown and Figure 1c). As exemplified in Figure 1d, all tumors were histologically classified as well-differentiated invasively growing SCCs.
Tumorigenicity is associated with new chromosomal changes
To determine whether these defined exposure regimes may also be associated with changes in the chromosomal make up, we performed M-FISH analysis on the cells irradiated for 5 weeks either with 4 × 5 or 1 × 20 J/cm2 per week prior to injection into the mice as well as with cells re-established from six tumors formed by the UVA irradiated cells. As shown in Figure 2a, UVA-treated HaCaT cells prior to injection showed the HaCaT-specific aberrations and aberrations that had occurred during passaging. In addition, we found aberrations in individual metaphases (Supplementary Table 1). Since the number was slightly increased in the UV-irradiated versus nonirradiated control cultures, this suggested that some chromosomal changes may have occurred upon UVA treatment.
Of the six tumor-derived lines four originated from series B HaCaT cells irradiated with 4 × 5 J/cm2 per week. Tumor 1 (T1) developed after the 5-week treatment and tumors T2, T3 and T4 after the 15-week treatment. Two cell lines derived from series B HaCaT cells irradiated with 1 × 20 J/cm2 once per week for 15 weeks (tumors T5 and T6). As exemplified in Figures 2b–d, M-FISH analysis clearly identified the original HaCaT markers and a common set of aberrations—representing a subpopulation of the untreated HaCaT cells. In addition and in agreement with the tumor cells established from the exploratory experiment, all tumor-derived lines showed tumor-specific karyotypic changes. First, the number of chromosome 15 was reduced in all metaphases (Table 1). Second, three cell lines (T4, T5 and T6) exhibited 3–5 additional recurrent aberrations, that is, the same aberration present in all metaphases, that were unique for the respective tumor-derived lines (Figures 2b–d and Table 1) demonstrating that in these tumors one subclone with a specific aberration profile dominated during tumor growth. The other three cell lines (T1, T2 and T3) were characterized by different subpopulations, each characterized by distinct individual aberrations (at least be detectable in three metaphases each, see Table 1). Thus, besides selection for a pre-existing HaCaT subpopulation all tumor-derived lines had gained new chromosomal aberrations, which obviously had provided a selective growth advantage in vivo and, accordingly, were crucial for the tumorigenic phenotype of the respective HaCaT populations.
UVA irradiation induces DNA damage in HaCaT cells
Chromosomal aberrations (translocations, deletions, amplifications) require the formation of DNA strand breaks which further progress to chromosomal breakage. To determine whether UVA can induce DNA strand breaks we irradiated HaCaT cells with 60 J/cm2 and performed the alkaline comet assay—detecting ss and ds breaks as well as alkali-labile sites—either directly or after a recovery period of 2 and 4 h. While the nonirradiated mock-treated control cells exhibited almost no DNA damage (average DNA in tail: 1.96% ±0.53), the average DNA in tail of irradiated cells increased significantly (40.35% ±4.68; P<0.0001). After 2 h ∼84% of the DNA damage was repaired leaving 6.45% (±2.88) DNA in the tail (Figure 3a). This decreased further to 3.94% (±2.95) after 4 h demonstrating a repair of 91%. Plotting the DNA in tail for each individual cell additionally clarified that UVA irradiation of 60 J/cm2 had caused significant damage in about every cell (Supplementary Figure 2A) and that although most of the cells had completely repaired their damage within 4 h, a fraction of 15–20% still showed 20% of DNA in tail. Thus, UVA induced DNA strand breaks and a small fraction of cells maintained damage for longer time periods.
UVA induces double strand breaks in HaCaT keratinocytes
Since the alkaline comet assay does not differentiate between ss and ds breaks we performed the neutral comet assay as well as immunofluorescence staining for γ-H2AX, both indicative for DNA ds breaks (Rogakou et al., 1998; Olive, 1999; Sedelnikova et al., 2002; Rothkamm and Lobrich, 2003). HaCaT keratinocytes were irradiated with 20, 40 and 60 J/cm2 UVA and the neutral comet assay was performed directly thereafter. While nonirradiated control cells on average showed ∼23% (±4.3) DNA in tail, cells irradiated with 20 J/cm2 UVA showed 42.93% (±3.7) and this further increased to 48.82% (±4.8) with 40 J/cm2 and 55.67% (±4.2) with 60 J/cm2 (Figure 3b, Supplementary Figure 2B). In all cases the increase was statistically significant (P<0.001). Thus, UVA induced DNA ds breaks in a dose-dependent manner.
We next performed immunofluorescence staining for γ-H2AX. Since it is suggested that also during S-phase ds breaks occur that are highlighted by γ-H2AX staining (Thompson and Limoli, 2003), HaCaT cells were enriched in the G1 phase of the cell cycle (Supplementary Figures 3A and B) before being irradiated with 60 J/cm2 UVA and incubated for 30 min to allow maximal foci induction (Rapp and Greulich, 2004). Furthermore, it was recently shown that UVC radiation induced γ-H2AX phosphorylation mainly as a diffuse, even pan-nuclear staining which was discussed as being dependent on nucleotide excision repair (Marti et al., 2006). Different from this, we here found induction of distinct γ-H2AX foci. While nonirradiated control cells showed on average 2.3 foci per nucleus, the number strongly increased in UVA-irradiated cells (Figures 3c and d). Most importantly, the majority (∼65%) showed on average 6 foci per nucleus while such nuclei were only detected in about 10% of the control population.
Telomere length is not altered upon UVA treatment
Another potential site for damage that can cause genomic instability, are the telomeres. We, therefore, performed telomere length restriction fragment (TRFL) analysis with UVA-treated HaCaT cells prior to injection as well as tumor-derived cell lines. As exemplified in Figure 4a, telomere length of the irradiated pre-injected cells—irrespective of treatment regime and time of treatment—was very similar to that of the nonirradiated control cells (mean TRFL of ∼2.5 kb). Thus, even the 15-week treatment did not cause obvious changes in telomere length.
Different from this and with one exception (T4), all recultivated tumors exhibited an altered telomere length (Figure 4b). Three lines seemed to contain two populations with different telomere length. In two, the original mean TRFL of ∼2.5 kb predominated (TA and T1) over a second peak with a mean TRFL of ∼5 kb while in one cell line the population with longer telomeres (mean TRFL of ∼6 kb, T2) predominated. Two lines showed a mean TRFL of ∼3.5 kb (T3 and T6) and two of ∼6 kb (TB and T5). Thus, instead of potential telomere shortening we rather observed subpopulations with about double the original telomere length. It is noteworthy that the three genetically heterogeneous tumor-derived lines (T1, T2 and T3) were among those showing two different telomere lengths.
UVA irradiation induces DNA damage also in normal human keratinocytes
Since the immortal HaCaT keratinocytes harbor p53 mutations (Lehman et al., 1993) which could influence the UVA-dependent damage response, we next investigated normal human skin keratinocytes. G1-enriched keratinocytes (Supplementary Figures 3C and D) were irradiated with 60 J/cm2 UVA and analysed for γ-H2AX foci formation directly after irradiation (0 h) as well as after 10, 20, 30, 60 and 360 min (Figure 5a). Nonirradiated control cells exhibited on average 1–2 foci per nucleus. While directly after UVA irradiation there was no change, damage increased steadily during the first 30 min post irradiation with an average of 5 foci per nucleus and even after 6 h another slight increase was seen.
Since in HaCaT cells damage and repair was not evenly distributed, we also evaluated each individual nucleus (Figure 5b). The number of nuclei with 1–5 foci decreased from 68% after 1 h to 48% after 6 h and, accordingly, the number of foci-negative nuclei increased from 20 to 28%. Interestingly, the number of nuclei with 6–10 foci also increased from 10 to ∼20%. Thus, while repair took place in some cells, damage persisted in others and here the amount of ds breaks was even increased after 6 h. Although we cannot firmly rule out that an excess of γ-H2AX foci was due to replication processes or repair, it has to be stressed that 83% of the cells were in the G1-phase at the time of irradiation (Supplementary Figures 3C and D).
UVA induces ds breaks already at ‘low’ doses in normal keratinocytes
To determine whether lower doses of UVA radiation would also induce a distinct damage profile, we irradiated normal human skin keratinocytes with 10, 20, 30, 40 and 60 J/cm2 UVA and analysed for γ-H2AX foci after 1 h—identified as the time for maximal foci induction in these cells (Figure 5c). Keratinocytes irradiated with 10 J/cm2 UVA showed 2.26 foci per nucleus, values very similar to that of nonirradiated controls (1–2 foci per nucleus). In contrast, 20 J/cm2 UVA irradiation induced a major increase in foci formation with an average of 5–6 foci/nucleus. Interestingly, there was no further increase with higher doses including 60 J/cm2. Although this suggested that 20 J/cm2 was as effective as 60 J/cm2, we observed some difference when classifying the individual damage response. Eighteen percent of the human keratinocytes were negative (0 foci), 67% showed 1–5 and 15% 6–10 foci after 20 J/cm2 UVA treatment (Figure 5d). After irradiation with 60 J/cm2, 20% of cells were also without foci, while the number of highly damaged keratinocytes was slightly increased with ∼60% of the cells with 1–5 and >20% with 6–10 foci.
Finally, we performed the alkaline and neutral comet assays. Irradiation of normal human keratinocytes with 30 and 60 J/cm2 UVA in the alkaline comet assay resulted in a dose-dependent increase in DNA damage from 9.62% (±2.6%) DNA in tail in nonirradiated control cells to 23.83% (±3.7%; P<0.001) after 30 J/cm2 and 42.54% (±5.5%; P<0.001) after 60 J/cm2 UVA (Figure 5e; Supplementary Figure 2C). Also the neutral comet assay provided evidence for UVA-dependent formation of ds breaks, though the induction was only minor (Supplementary Figure 2D).
UVA irradiation causes micronuclei formation in normal keratinocytes
As a consequence of ds breaks and chromosomal breakage, micronuclei (MN) are formed. Therefore, MN formation represents an established marker for genomic instability (Fenech, 2002). Since it is suggested that the cytokinesis-block MN assay allows a better precision because data obtained are not confound by altered cell division (Fenech, 2000), we treated the keratinocytes with cytochalasin-B for 24 h after irradiation with 20 J/cm2. To distinguish between MN originating from whole chromosomes or acentric fragments, we additionally performed centromere staining and determined the fraction of MN-positive cells as well as the ratio between MNs with and without centromeres in ∼1000 cells.
Investigating keratinocytes from three different donors, the number of MN-positive cells nearly doubled from 1–1.8% in nonirradiated control cells to 2–3.2% after UVA irradiation and this increase was significant for two keratinocytes strains (χ2-test; Figure 6a). Concerning the kind of MNs, we found some inter-personal variation. While one strain showed only induction of centromere-containing MNs, the other two also showed MNs without centromeres (32 and 65%, respectively; Figures 6a and b). This indicates, that UVA is able to induce both, a nondisjunction and thereby loss of entire chromosomes, as well as DNA ds breaks thereby causing chromosome breakage that result in acentric fragments. Whether there is a preference for one or the other may well be donor specific.
To determine whether MN induction by UVA was only a short-term effect or may be longer lasting, we next irradiated keratinocytes with 10, 20 and 40 J/cm2 and cultured them for 5, and in part 10, consecutive days without cytochalsin-B blockade. While only ∼1% of nonirradiated control cells showed MN, we observed a significant dose-dependent increase of MN-positive cells 5 days after irradiation with 1.8% after 10 J/cm2, 2.6% after 20 J/cm2 and 4.4% after 40 J/cm2 UVA (Figure 6c, Cochran–Armitage test, P<0.00001). While both, the fraction of MN with and without centromeres increased, MNs without centromeres dominated with increasing UVA doses (20 and 40 J/cm2). Interestingly, also keratinocytes cultured for 10 days post irradiation (20 J/cm2) still showed twice the amount of MN-positive cells (∼6%) compared to their nonirradiated controls (2–3%; data not shown), demonstrating long-lasting damage.
In the present paper, we report that defined treatment regimes (1 × 20 or 4 × 5 J/cm2 per week) of environmentally highly relevant UVA doses cause tumorigenic conversion of HaCaT skin keratinocytes allowing them to form well-differentiated SCCs in nude mice. Tumorigenicity was characterized by distinct chromosomal aberrations present in the cells of the recultivated tumors. In an approach to assess DNA damage formation by UVA and its mutagenic outcome we show that UVA radiation caused DNA ss and ds breaks in HaCaT cells and that some of this damage was maintained even several hours post irradiation. Furthermore, we demonstrate that this damage profile is not restricted to the immortal HaCaT cells but is similarly seen in normal human keratinocytes. We even provide evidence for a long-lasting increase in micronuclei formation with about half of the MNs containing either entire chromosomes or acentric fragments, that is, chromosomal parts that require chromosomal breakage in order to occur. All this strongly implies that UVA radiation—most likely through oxidative damage (Kielbassa et al., 1997; Cadet et al., 2005)—induces DNA ss and ds breaks. Whether the ds breaks arise directly as or following ss breaks processing, is currently unclear and remains to be addressed. However, the study shows that UVA is able to induce genomic instability by causing DNA strand breaks which presumably induce chromosomal breakage and, upon repair, chromosomal changes as observed in the tumorigenic HaCaT cells. Furthermore, since similar long-lasting damage is seen in normal human keratinocytes and since skin SCCs are characterized by a high frequency of chromosomal changes (reviewed in Boukamp, 2005b), it is tempting to suggest that UVA radiation may contribute to skin cancer development and progression in a similar way.
This assumption is also substantiated by a recent study showing that irradiation of HaCaT cells with as little as 10 J/cm2 UVA reduced long-term plating and colony forming efficiency (Phillipson et al., 2002). Since this was not related to an extended population doubling time but to partial cell death, the authors reasoned that UVA induced delayed cell death. Also we found an increased death rate 14 h (40%) post irradiation while directly after irradiation (up to 4 h) and later on (after 5 days) cell death was negligible. Although we cannot exclude that during this phase all ‘long-lasting’ damaged cells died, we found—again in line with the previous report—that this delayed cell death correlated with increased MN formation. The authors interpreted this as an indication for persistent genomic instability and hypothesized that UVA leads to an increasing frequency of chromosomal rearrangements (Phillipson et al., 2002) something that we now show for the recultivated tumorigenic HaCaT cells. Accordingly, in fibroblasts low doses of UVA induced MNs at a level comparable to 1 Gy ionizing radiation, supporting the view that UVA is a potential inductor of chromosomal aberrations (Emri et al., 2000). Thus, with our study we establish a first association between UVA-dependent damage, chromosomal aberrations and their functional consequences for carcinogenesis.
Having performed Comet-FISH analysis on human lymphocytes, we previously suggested that 50 J/cm2 UVA induced a number of chromatid strand breaks in chromosomes in a nonrandom manner (Rapp et al., 2000). While lacking correlation between strand breaks and chromosome size, an inverse correlation was established between the density of active genes and sensitivity toward UVA radiation. In the HaCaT cells studied here, we found many different chromosomes being involved in the aberrations prior to injection. Furthermore, since all changes were only seen in a single metaphase obviously none of them provided a growth advantage in vitro. In vivo, on the other hand, and most likely due to a massive selection pressure, only few cells survived and gave rise to tumors of genetically defined subpopulations. These were either of monoclonal origin (all carrying the same aberrations) or derived from several clones (different aberrations in different subpopulations). However, in these also the aberration-involved chromosomes differed. Thus, we have to conclude that damage and formation of chromosomal aberrations is a rather random process and that different aberrations are able to contribute to tumorigenicity.
Different from, for example, chronic myeloid leukemia where the formation of specific translocation chromosomes is accepted as being crucial for tumor development (Tefferi et al., 2005), the role of most chromosomal aberrations in solid tumors is still elusive. This is also the case for the aberrations identified for the different tumor-derived cell lines. One exception may be loss of chromosome 15, the only common aberration. We previously showed that loss of chromosome 15 correlated with loss of the angiogenesis inhibitor TSP-1 which in turn correlated with the angiogenic switch required for malignant conversion (Folkman and Hanahan, 1991; Bleuel et al., 1999). We recently also showed that loss of chromosome 15 occurs in human skin carcinomas and that this correlated with progression from benign to malignant tumor phenotype (Burnworth et al., 2006). Accordingly, loss of chromosome 15 seen here, may also have provided a selective advantage and thereby contributed to the high frequency of tumorigenic conversion (three out of the four different HaCaT populations irradiated with 4 × 5 or 1 × 20 J/cm2).
Convincing evidence for the role of UVA in tumorigenic conversion of HaCaT cells is the rapid onset. Already a five-week-UVA-treatment is sufficient to induce tumorigenicity in three of the four populations. We previously showed that mere cultivation even for >6 years and >300 passages did not lead to tumorigenicity (Boukamp et al., 1997). Accordingly, mock-treated control cells remained nontumorigenic. However, increased temperature caused tumorigenic conversion and, similar as shown here for UVA, was correlated with induction of DNA ds breaks and new chromosomal aberrations (Boukamp et al., 1999). Since UVA as well as increased temperature cause oxidative stress, they may well function through a similar pathway. As a major difference, however, tumorigenicity was already evident after 5-week UVA treatment and further increased after 15 weeks, while heat-induced tumorigenic conversion required >15 weeks (Boukamp et al., 1999). Thus, tumorigenicity can be induced by highly relevant UVA doses in a reproducible manner after only few weeks of irradiation thereby making UVA a potent tumor promoter for HaCaT cells.
One potential mechanism of induction of chromosomal aberrations could be critically short telomeres leading to fusion bridge-breakage cycles (Murnane and Sabatier, 2004). ROS, as induced by UVA radiation, are known to cause breaks in G-rich sequences (von Zglinicki et al., 2000). Actually, it was shown that telomeric DNA was 5 to 10-fold more assailable than nontelomeric DNA (Oikawa and Kawanishi, 1999). Thus, telomeres should be optimal substrates for UVA radiation. On the other hand, UVA-induced ROS depend on the up-to-now unknown nuclear sensitizers. If they are not present at telomeres, these genomic regions can be underrepresented in the DNA fragmentation due to the limited diffusion range of the UVA generated ROS. Comparing telomere restriction fragment lengths from HaCaT control cells with UVA-treated cells prior to injection as well as tumor-derived cells, we did not find telomere shortening and we have preliminary evidence that UVA radiation may also not account for rapid telomere erosion when investigating individual cells (D Krunic, K Wischermann and P Boukamp, unpublished results). In fact, some of the tumor-derived lines even had increased mean telomere lengths. Thus, a telomere length-dependent mechanism of genomic instability appears unlikely.
Instead we propose that UVA radiation induces ss and ds breaks (the latter either directly or through clustered and processed ss breaks) in chromosomal sites. These are not repaired to completion immediately but a small fraction of cells maintains the damage for extended periods. Such breaks are prerequisite for chromosomal breakage, which in turn is prerequisite for chromosomal aberrations as identified in the tumor-derived HaCaT cells. This finding in conjunction with an increased formation of micronuclei 10 days post irradiation, therefore, underlines the role of UVA-dependent damage in inducing persistent genomic instability, which finally results in chromosomal aberrations. Furthermore, since we obtained a similar damage profile with HaCaT cells and normal keratinocytes, this may well reflect the general action spectrum of UVA radiation and may suggest a distinct role of UVA in the multi-step process of skin carcinogenesis.
Materials and methods
HaCaT cells were routinely cultured as previously described (Boukamp et al., 1988). Normal human keratinocyte cultures were established from human trunk skin (approved by the Ethics Commission of the Medical Faculty, University of Heidelberg, Germany) and passaged as described (Boukamp et al., 1990a). Second passage keratinocytes (grown in keratinocyte growth medium, Epilife, Cascade Biologics, Tebu-bio GmbH, Offenbach, Germany) were used for irradiation experiments at a density of 70–90%.
G1 phase: enrichment of HaCaT and normal keratinocytes
HaCaT cells (80% confluent) were cultured in RPMI-1640 medium without cystein and methionine (No. R7513, Sigma, Deisenhofen, Germany) for 48 h and for additional 14 h in RPMI-1640 with 400 μM L-Mimosine (Sigma; adopted from Krude, 1999). Normal keratinocytes (80% confluent) were incubated in keratinocyte growth medium without pituitary extract (Promocell, Heidelberg, Germany) in the presence of 10 ng/ml TGF-ß1 (R&D Systems, Wiesbaden, Germany) for 48 h. Cells were stained with 30 μg/ml propidium iodide and the G1 enrichment was controlled using standard fluorescence-activated cell sorting techniques (Supplementary Figure 3).
UVA irradiation of cells
For the exploratory experiment HaCaT cells were irradiated with a daily increasing UVA dose from 7 to 42 J/cm2 (2–12 h per day) over a 6-week period (irradiance of 1.0 mW/cm2 from Philips PL-S 9W/10 blacklight lamps 330–400 nm). Cells were irradiated in phenol-red-free culture medium and closed lid, yielding a transmission of >70% UVA. For comparing single versus split doses, the high intensity UVA source Mutzhas UVASUN 3000, equipped with UVASUN safety filters and a WG305 filter (emitting wavelengths in the 320–450 nm range; Mutzhas et al., 1981), was used. Fluences were determined with a UVA meter (Dr Hönle, Martinsried, Germany) equipped with an UVA-sensor calibrated for emission from 315–400 nm. The emission spectrum was checked with an Optronics OL-754 spectroradiometer (Optronics, Orlando, FL, USA). Cells were irradiated in phosphate-buffered saline (PBS) without lid at a distance of 40 cm for 100 s for 5 J/cm2 UVA and 400 s for 20 J/cm2, respectively (irradiance 50 mW/cm2). For DNA damage and repair analyses the Phillips HB 404 tanning lamp (irradiance 21 mW/cm2) equipped with an UVB blocking filter (Hoya UV34, Galvoptics, Basildon, UK), an infrared filter (M-UG2, Schott, Grünenplan, Germany) and ventilators for cooling were used. The fluence was measured with a Vilbert–Lourmat VLX-3W radiometer. HaCaT cells and normal human keratinocytes plated on glass slides were irradiated on ice in PBS in a distance of 16 cm to the UVA source for 7.03 min (10 J/cm2) to 42.3 min (60 J/cm2). The cells were either analysed immediately or after an incubation time as indicated. For each experiment nonirradiated control cells were treated accordingly.
For all three UVA lamps, spectral characteristics were comparable in emitting UVA with a maximum at 365 nm while the UVB fractions was negligible with less then 0.005% (Supplementary Figure 4).
About 5 × 106 cells in 100 μl were injected subcutaneously into each side of the back of three (5 weeks) and four (15 weeks) nude mice (Swiss/c nu/nu backcrosses). Tumor growth was measured weekly and tumor size calculated in mm3 (Parangi et al., 1996). Tumors were excised and formalin fixed for standard hematoxilin/eosin staining. For recultivation, tumors were shortly rinsed in ethanol (80%) and culture medium, minced into 1 mm3 pieces and plated as explant cultures. After tumor cell outgrowth, the cells were treated according to Boukamp et al. (1988). Tumor cell lines were established from 4 × 5 J/cm2 5 weeks animal 3 right side (T1), 4 × 5 J/cm2 15 weeks animal 1 left side (T2), 4 × 5 J/cm2 15 weeks animal 4 right side (T3), 4 × 5 J/cm2 15 weeks animal 2 left side (T4), 1 × 20 J/cm2 15 weeks animal 3 left side (T5) and 1 × 20 J/cm2 15 weeks animal 4 left side (T6).
Alkaline and neutral comet assay
The alkaline comet assay was performed either directly after irradiation or after a recovery of 2 or 4 h in complete culture medium, according to Wischermann et al. (2007). The neutral comet assay was performed directly after irradiation according to the protocols described by Olive and Banath (Olive et al., 1993). For both assays all steps were carried out on ice or in the cold to minimize repair processes. For each data point two to three areas on parallel slides were scored with 51–60 cells each and DNA in tail (%) was calculated for each cell using the image-analysis software ‘Komet 4’ (Kinetics Imaging Ltd, Liverpool, UK). The median of DNA in tail (%) was calculated for each area and the presented values are the means of the medians of each data point. At least two independent experiments were performed. The Mann–Whitney rank sum test and t-test were used to determine statistically significant differences between control and irradiated cells.
Detection of γ-H2AX
HaCaT cells and normal human keratinocytes (irradiated and nonirradiated control cells) grown on glass slides were fixed in methanol/acetone (1:1) for 10 min at −20 °C. Cells were rehydrated in PBS for 15 min and permeabilized in PBS containing 0.5% Triton X-100 for 20 min. After blocking for 30 min in 3% bovine serum albumin (BSA) in PBS, the cells were incubated with mouse anti-γ-H2AX (Upstate, Charlottesville, VA, USA; 1:100 in blocking solution) for 1 h and after washing (3 × 10 min) with a secondary anti-mouse Alexa 488 antibody (1:800; Molecular Probes/Invitrogen, Karlsruhe, Germany) containing 5 μg/ml bisbenzimid (Aventis Pharma, Frankfurt, Germany) for 30 min. The average number of γ-H2AX foci/nucleus was determined using the ImageJ software (Wayne Rasband, http://rsb.info.nih.gov/ij/) or the cells were classified into four groups (0, 1–5, 6–10 or >10 foci) according to the number of foci per nucleus.
For the cytokinesis-blocked MN assay, normal keratinocytes were irradiated with 20 J/cm2 UVA and cultivated in the presence of 4 μg/ml cytochalasine B (Sigma Aldrich, Munich, Germany) for 24 h. Cells were fixed with methanol and acetone for 5 min each at −20 °C, blocked with 3% BSA and incubated with a human anti-all-centromere antibody (1:100; Antibodies Inc., Davis, CA, USA) for 30 min at 37 °C. After washing in PBS, cells were incubated with a rabbit anti-human-FITC (1:200; Dako, Hamburg, Germany) for 30 min at 37 °C and nuclei were counterstained with 0.1 μg/ml 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; Vectashield, Wertheim, Germany). From 1000 binucleated cells the number of cells with MN (expressed as % MN-positive cells) and the ratio of MNs with and without centromere signals was determined. For long-term maintenance of MNs, the cells were irradiated with 10, 20 and 40 J/cm2 UVA and cultivated for 5 and 10 days without Cytochalasin-B block. MN staining was performed as described.
Statistical significant differences between control and irradiated cells were assessed with the χ2-test (Cohen, 1960) and significance of dose-dependency with the Cochran–Armitage test (Cochran, 1954; Armitage, 1955).
Multiplex fluorescence in situ hybridization analysis
Metaphase spreads were prepared and combinatorial labeling of whole chromosome painting probes and multicolor-hybridization was performed as described (Popp et al., 2000).
Telomere length measurement by Southern blot analysis
Genomic DNA was isolated using DNAzol (Gibco BRL, Eggenstein, Germany). DNA (5 μg) was digested with Rsa I (Roche, Mannheim, Germany) and subjected to electrophoresis. Telomere length measurements by Southern blot was performed as described (Figueroa et al., 2000).
We thank Cees Guikers for the first experiments with daily exposures, Heinrich Steinbauer for performing the tumor experiments, Heidi Holtgreve for her excellent technical assistance in M-FISH hybridization and Stefan Henning for comparing the different UVA-sources. This work was supported in part from the Sander Stiftung, Deutsche Krebshilfe e.V. and EU (LSHC-CT-2004-502943) (both to PB), QLK4-1999-01084 (to PB and KSK) as well as the German Research Foundation (SCHA 411/10-1,-2) and the European Community Marie Curie Program MEIF-CT-2005-023821 (to AR).
About this article