Hydrochlorothiazide Use and Risk of Nonmelanoma Skin Cancers: A Biological Plausibility Study

Recent studies reported the association between increased risk of nonmelanoma skin cancers (NMSCs) and the use of hydrochlorothiazide (HCTZ), one of the most commonly prescribed diuretic, antihypertensive drug, over the world. Although HCTZ is known to be photosensitizing, the mechanisms involved in its potential prophotocarcinogenic effects remain unclear. Under acute exposure, therapeutically relevant concentrations of HCTZ (70, 140, and 370 ng/mL) amplified UVA-induced double-strand breaks, oxidative DNA, and protein damage in HaCaT human keratinocytes, and this effect was associated to a defective activity of the DNA repair enzyme, OGG1. Oxidative damage to DNA, but not that to proteins, was reversible within few hours. After chronic, combined exposure to HCTZ (70 ng/mL) and UVA (10 J/cm2), for 9 weeks, keratinocytes acquired a dysplastic-like phenotype characterized by a multilayered morphology and alterations in cell size, shape, and contacts. At the ultrastructural level, several atypical and enlarged nuclei and evident nucleoli were also observed. These transformed keratinocytes were apoptosis resistant, exhibited enhanced clonogenicity capacity, increased DNA damage and inflammation, defective DNA repair ability, and increased expression of the oncogene ΔNp63α and intranuclear β-catenin accumulation (a hallmark of Wnt pathway activation), compared to those treated with UVA alone. None of these molecular, morphological, or functional effects were observed in cells treated with HCTZ alone. All these features resemble in part those of preneoplastic lesions and NMSCs and provide evidence of a biological plausibility for the association among exposure to UVA, use of HCTZ, and increased risk of NMSCs. These results are of translational relevance since we used environmentally relevant UVA doses and tested HCTZ at concentrations that reflect the plasma levels of doses used in clinical practice. This study also highlights that drug safety data should be followed by experimental evaluations to clarify the mechanistic aspects of adverse events.

The most important risk factor for NMSC development is a constant and cumulative UVR exposure [10,11] which is responsible for known harmful cellular effects consisting in a direct UVB-mediated DNA damage and in an indirect UVA-mediated oxidative DNA damage [12][13][14].
On the contrary, the molecular pathways at the basis of the increased risk of NMSCs observed among HCTZ users are unclear, but photosensitizing actions of HCTZ may play a role. Skin photosensitivity reactions, frequently associated to UVA exposure, were indeed reported in patients taking HCTZ [15][16][17]. A combined exposure to UVA and HCTZ induced acute phototoxicity in experimental studies in keratinocytes [18], in cervical cancer cells [19] and in hairless mice [20]. HCTZ also enhanced UVA-induced DNA damage in a mice model defective in nucleotide excision repair [21]. These data contributed to the decision by the International Agency for Research on Cancer Monograph Working Group, to classify HCTZ as possibly carcinogenic to humans (Group 2B) [22].
However, these studies were focused on acute effects and did not investigate the biological pathways linking the photosensitizing/phototoxic properties of HCTZ to its potential prophotocarcinogenic effects.
Since NMSCs (and among them, squamous cell carcinoma) arise from chronic, repeated UV exposures [10,11,23,24], and HCTZ is mostly a long-term treatment, an experimental model involving long-lasting drug exposures and repeated UV irradiations might be a more reliable tool to explore the biological plausibility of the increased risk of NMSC among HCTZ users.
On these basis, we characterized both the acute and the long-term, molecular, morphological, and functional consequences of the combined exposure to HCTZ and UVA in human keratinocytes.
2.2. Short-Term Treatments and Photoirradiation. The experimental flow chart for short-term experiments is depicted in Figure 1(a). Control cells (naive) were maintained under standard culture condition without any treatment. In the HCTZ group, cells were treated with HCTZ alone at concentrations ranging from 70 to 370 ng/mL corresponding to the values of Cmax following the administration in humans of 12.5, 25, and 75 mg doses, respectively [25]. HCTZ (Pub-Chem CID: 3639) (Sigma Aldrich, Milan, Italy) was dissolved in dimethyl sulfoxide (DMSO) as 14 mg/mL stock solutions and used immediately or stored at −20°C. In the UVA group, cells were irradiated with 10 J/cm 2 of UVA light with an emission centered at 365 nm, using a UV Bio-Link (BLX) (Vilber Lourmat, Marne-La-Vallee, France), on cold plates to eliminate UVA-induced thermal effects and without lid. The time of irradiation lasted in 40 minutes. Before irradiating the cells with UVA, the medium was removed, and the cells were covered with a thin layer of phosphate buffer solution (PBS). In the HCTZ+UVA group, cells were preincubated with HCTZ (70-370 ng/mL) for 2 hours (h), exposed to 10 J/cm 2 of UVA, and then treated with HCTZ for additional 1 or 6 h.
2.3. Long-Term Treatments and Photoirradiation. The experimental flow chart summarizing the weekly protocol for long-term experiments is depicted in Figure 1(b). For all the experiments conducted in cells chronically treated, 8 × 10 5 cells, corresponding to 14,000 cells/cm 2 , were seeded in Petri dishes, allowed to grow for 24 h, and then treated as follows: cells maintained in standard culture conditions (naive); cells chronically treated with HCTZ alone (70 ng/mL) for 9 weeks (HCTZ); and cells irradiated with UVA alone (10 J/cm 2 ), twice a week, for 9 weeks (UVA) or in the presence of HCTZ (70 ng/mL) (HCTZ +UVA). Each single exposure lasted for 40 minutes. Fresh HCTZ was added every 2 days. The concentration of 70 ng/mL is similar to that found at the steady state following the administration of the dose of 75 mg of HCTZ in humans [26], and the weekly UVA dose corresponds to a recreational human exposure of approximately 1 hour to midsummer sun [27].
2.6. Oxidative DNA Damage: 8-OHdG Determination. HaCaT cells (5 × 10 5 cells/well corresponding to 55,000 cells/cm 2 ) were seeded and after 24 h exposed to HCTZ (70-370 ng/mL) with or without UVA irradiation (10 J/cm 2 ). Total DNA was extracted using the DNeasy Mini Kit (Qiagen, Hilden, Germany), denatured at 95°C for 5 minutes, and digested with nuclease P1 and alkaline phosphatase at 37°C [31]. 8-OHdG levels were quantified using the 8-OHdG ELISA kit (Jaica, Fukuroi, Japan). The results were calculated from the absorbance at 450 nm of a standard curve of 8-OHdG solutions and expressed as ng of 8-OHdG/mL. 2 Oxidative Medicine and Cellular Longevity 2.7. DNA Repair: OGG1 Activity. HaCaT cells (2 × 10 5 cells/well corresponding to 50,000 cells/cm 2 ) were seeded and after 24 h exposed to HCTZ (70-370 ng/mL) with or without UVA irradiation (10 J/cm 2 ). The activity of OGG1 was deter-mined by a protocol set-up in our laboratory on the basis of Hamann and Schwerdtle [32] with the following modifications: a hairpin-like structured synthetic oligonucleotide with the sequence.   Figure 1: (a) Short-term treatments and photoirradiation: experimental flow chart. Cells were treated with HCTZ alone (70-370 ng/mL) or in combination with UVA (10 J/cm 2 ), left untreated (naive), or irradiated with UVA only. 8-Hydroxy-2′-deoxyguanosine (8-OHdG), DNA repair (OGG1 activity), and carbonyl residues were determined after 1 and 6 h. γ-H2AX (double-strand breaks) was measured after 1 h. Cell viability was measured at 24, 48, and 72 h. Apoptosis was measured at 24 h. (b) Long-term treatments and photoirradiation are as follows: experimental flow chart of the weekly protocol. Cells were treated with HCTZ alone (70 ng/mL) or in combination with UVA (10 J/cm 2 ) for 9 weeks, left untreated (naive), or irradiated with UVA only. 8-OHdG, OGG1 activity, genotoxic damage (γ-H2AX), apoptosis resistance, COX-2 and ΔNp63α expression, PGE2 production, and β-catenin localization were investigated together with morphological characterization and clonogenic capacity. 3 Oxidative Medicine and Cellular Longevity GCAGGACTGGTCGCGCGTGTTATTATTG-3′, X = 8 − OHdG, was designed by us and synthetized by Integrated DNA Technologies (Leuven, Belgium). Total protein (1 μg) or 2 units of recombinant hOGG1, used as positive control (New England Biolabs, MA, USA), were incubated in a reaction mixture containing 1X NEBuffer 2, 40 pmol of synthetic oligonucleotide, and 100 μg/mL BSA, at 37°C for 30 min and then at 95°C for 5 minutes to stop the reaction. A negative control, containing only the reaction mixture was also used. Samples were analyzed by electrophoresis in 3% agarose gel, in TBE 0.5X. The image of residual and intact (negative control) bands was acquired and analyzed through the software Quantity One (Bio-Rad). The percentage of residual oligo was calculated on the basis of the following formula: intact oligonucleotide : 100 = cleaved oligonucleotide : x. The OGG1 glycosylase activity was calculated as follows: 100 -x.
2.11. Apoptosis. For acute experiments, HaCaT cells were seeded on histological slide (2 × 10 5 cells/well corresponding to 50,000 cells/cm 2 ) and treated with HCTZ at the final concentration of 70, 140, and 370 ng/mL. After two hours, cells were irradiated with UVA (10 J/cm 2 ) in 100 μL of PBS. After irradiation, PBS was replaced with standard culture medium and incubated at 37°C for 24 h. Cells were washed in PBS and fixed in Bouin's liquid (acetic acid/formaldehyde/picric acid), for 20 minutes, and then stained with Feulgen's reaction, specific for DNA detection. Briefly, cells were incubated in 1 N HCL for 22 minutes at 60°C, with the Schiff reactive for 60 minutes at room temperature and then washed in 0.05 N HCL and 5% NaHSO 3 . Successively, nuclei were counterstained for 30 seconds with 0.5% fast green in alcoholic solution, dehydrated in ethanol, washed in xylene, and mounted [35].
For long-term experiments, cells chronically exposed to UVA and HCTZ, as described above, were harvested, fixed in Bouin's liquid for 18 h, and then paraffin embedded. Histological sections, 5 μm thick, were obtained and stained with Feulgen's reaction, as described above.
For each sample, 10 images at ×1000 magnification were obtained. Apoptotic cells were recognized by the presence of two morphological prodromal features of apoptotic bodies: nuclear fragmentation and cellular limits evanescence. The percentage of apoptotic cells was carried out on five microscopic fields for each experimental condition with an average number of about 50 cells and was determined by two independent observers in a blind fashion.
2.12. Image Acquisition and Analysis. Microscopic analyses were performed with a fluorescence microscopy (Labophot-2, Nikon, Japan) connected to a CCD camera. Ten photomicrographs were randomly taken for each sample at ×400 magnification. The measurements were made by two independent, blinded investigators, using the ImageJ 1.33 image analysis software (https://rsb.info.nih.gov/ij).
2.13. Cytology. Cells were fixed in cold 4% paraformaldehyde for 15 minutes, dehydrated in alcohol, and then paraffin embedded. Sections (4 μm thick) were stained with hematoxylin-eosin for morphological analysis.
2.16. Anchorage-Independent Growth: The Soft Agar Colony Formation Assay. Anchorage-independent growth assay was performed as described by Borowicz et al. [36]. Briefly, in six-well plates, 2 mL agar medium (0.5% agar-agar in DMEM with 20% FBS) was added to each well and allowed to solidify. HaCaT cells (4 × 10 4 cell/well corresponding to 4500 cells/cm 2 ) were suspended in 1 mL DMEM with 20% FBS and 0.3% agar solution and then laid on top of the hardened agar medium. Cells were then incubated at 37°C in a 5% CO 2 atmosphere for 21 days. Subsequently, the cells were stained with nitroblue tetrazolium chloride solution and incubated overnight at 37°C. Colonies were counted by two blinded independent observers.
2.17. Statistical Analysis. Results are presented as mean ± SEM of at least three independent experiments. Multiple comparisons were performed using 1-way ANOVA followed by Bonferroni's post hoc test. Difference with p < 0:05 was considered significant. Statistical analysis was performed with GraphPad Prism 5 (GraphPad software, San Diego, USA).
The percentage of cells positive for γ-H2AX, a marker for DNA double-strand breaks, was significantly increased in cells treated with UVA compared to naive (p < 0:05) and further enhanced in UVA+HCTZ compared to UVA alone (p < 0:01 at 370 ng/mL vs. UVA) (Figure 2(e)). The clear induction of distinct γ-H2AX foci in cells treated with UVA+HCTZ compared to control cells is shown in Figure 2(d).
UVA caused a time-dependent reduction in cell viability (-20% after 24 h, p < 0:05; -40% after 48 h, p < 0:001; and -50% after 72 h, p < 0:001). HCTZ alone was not cytotoxic nor did it exert additive cytotoxic effects in the presence of UVA (Figure 2(f)). Despite the trend toward increased apoptosis in the function of HCTZ concentration (p < 0:05, posttest for linear trend), no significant differences were observed compared to UVA alone (Figure 2(g)). A representative image of an apoptotic cell is shown in the inset of Figure 2(g).

3.2.
Long-Term Combined Exposure to HCTZ and UVA Induces Dysplastic Features in Human Keratinocytes. By the 7th week, keratinocytes treated with both UVA and HCTZ, but not those treated with HCTZ or UVA alone, started to grow as cellular aggregates and formed colonies visible to the naked eye (Figure 3(e)), exhibiting morphological and ultrastructural alterations suggestive of their transformation. In particular, after 9 weeks of treatment, cells treated with UVA alone or in combination with HCTZ (70 ng/mL) exhibited a pseudoepithelial morphology (Figures 3(c) and 3(d)) but only UVA+HCTZ cells, microscopically appeared as a multilayered pseudoepithelium (Figure 3(d), black dashed line) characterized by many cells with dysplastic features such as altered cytoplasmic to nuclear ratio, vacuolated cytoplasm, atypical and enlarged nuclei, and evident nucleoli (Figure 3(d), black arrows). At the ultrastructural level, UVA cells were polygonal in shape and displayed a number of cytoplasmic bridges without direct cell to cell contacts (Figure 3(h), black arrow heads); on the contrary, UVA +HCTZ cells were completely juxtaposed, mimicking an epithelial-like structure (Figure 3(i), black asterisk) with convoluted nuclei (red arrows) and a "salt and pepper" chromatin pattern (Figure 3(i)). Notably, cells treated with HCTZ alone (Figure 3(b)) displayed a round-shaped morphology identical to that of naive cells (Figure 3(a)) and grew in monolayer, and no ultrastructural abnormalities were noted (Figures 3(f) and 3(g)).

Long-Term Combined Exposure to HCTZ and UVA
Enhances the Clonogenicity Capacity. We used the soft agar colony formation assay, a well-known functional test based on the capability of transformed cells to grow independently of a solid surface and considered a hallmark of in vitro carcinogenesis [36].
Naive and HaCaT cells treated with HCTZ alone did not form any colony in anchorage-independent (soft agar) clonogenicity assay. As expected for not fully transformed cells, on UVA+HCTZ-treated cells, we were able to detect only few colonies, but compared to UVA alone, they exhibited enhanced clonogenic potential, further suggesting their early-stage oncogenic transformation. On average, HaCaT cells treated with UVA alone formed only 1 colony/well, while the number of colonies formed by UVA+HCTZ cells was markedly enhanced compared to UVA alone (p < 0:001 vs. UVA, Figures 6(a) and 6(b)).

Discussion
Our study demonstrates that human keratinocytes chronically coexposed with UVA and HCTZ, but not with HCTZ alone, develop a dysplastic morphology, and acquire molecular characteristics of an oncogenic transformation, thus providing biologically plausible mechanisms (Figure 7) for the increased risk of NMSCs observed in patients taking HCTZ [2,3,[5][6][7][8][9].
HCTZ causes photosensitization by either type I (free radical) or type II (singlet molecular oxygen) mechanisms and undergoes photodehalogenation, yielding a reactive form, which can damage DNA, lipids, and proteins [38][39][40]. Seto et al. [41] also demonstrated that HCTZ is photoreactive and generates singlet oxygen.
As early as 1 h after acute exposure, HCTZ enhances the UVA-induced formation of 8-OHdG, a major mutagenic oxidative DNA lesion [42], and of γ-H2AX, a sensitive marker of double-strand breaks and photogenotoxicity [34,43]. Double-strand breaks may derive from DNA replication or as a result of the repair process [44]; although we cannot completely exclude that some γ-H2AX foci were due to the replication process, they are more likely attributable to a direct DNA damage since we used serum-starved, confluent cells. Moreover, the presence of discrete γ-H2AX foci, rather than a pan-nuclear staining, supports a mechanism independent on repair [33,45].
To ensure genomic integrity, 8-OHdG is quickly removed by OGG1; whether immediate or slightly delayed, the induction of OGG1 repair activity mitigates the accumulation of 8-OHdG, consistently with the high efficiency of base excision repair enzymes [14]. OGG1, as other repair proteins, may itself be susceptible to damage and inactivation by oxidation [46] that may persist longer than DNA damage since protein carbonyls are irreversible modifications that cannot be repaired and require proteasome degradation [47]. Although at the highest concentration of HCTZ tested, OGG1 activity was strongly abated because oxidation, an alternative defense mechanism involving its increased gene transcription, was put in place.
In a mice model defective in nucleotide excision repair, Kunisada et al. [21] demonstrated that a single dose of HCTZ enhanced UVA-induced cyclobutane pyrimidine dimers (CPDs), the most abundant form of DNA damage induced in human skin by UVA [48]. Although we cannot rule out the involvement of other oxidative lesions, the formation of CPDs and 8-OHdG generally exceed that of single-strand breaks and oxidized pyrimidines in a 10 : 3 : 1 : 1 ratio [49,50].
At the molecular level, the presence of double-strand breaks and 8-OHdG lesions coupled with a defective DNA repair activity may lead to genomic mutations; this seems particularly relevant in the basal layer of human epidermis where keratinocytes express less OGG1 compared to the superficial layer [51,52]. Moreover, there is evidence that OGG1 knockout mice are more susceptible to skin carcinogenesis [53], and both the loss of OGG1 and the accumulation of 8-OHdG have been observed in NMSCs [54]. However, there are also other repair proteins whose impairment may have consequences on the removal of damage: the nucleotide excision repair inhibition by UVA-photoactivated fluoroquinolones and vemurafenib has been reported [55] as well as oxidation of MYH and RPA proteins [46].
In addition to genomic insults, cotreatment with UVA and HCTZ activates Wnt, one of the oncogenic pathways involved in skin tumorigenesis [56], and increases inflammation, consistently with the upregulation of a number of proinflammatory cytokines observed in mice treated with HCTZ and UVB [57]. Indeed, human NMSCs and their precursors exhibit increased levels of PGE2 [58][59][60] and nuclear β-catenin localization [61,62].
It is interesting to note that, after chronic treatment, the level of apoptosis was lower than that observed after acute treatment in cells treated with UVA in the presence of 10 Oxidative Medicine and Cellular Longevity HCTZ, suggesting the acquisition of apoptosis resistance, also supported by the enhanced expression of ΔNp63α and Bcl-2. To explore whether chronic treatment with HCTZ and UVA caused the acquisition of resistance to proapoptotic insults, cells were treated with the known proapoptotic drug doxorubicin, observing a significant increase of its IC50 value. He et al. [37] reported that long-term exposure to UVA causes resistance to doxorubicin-induced apoptosis in HaCaT keratinocytes. Moreover, previous evidence demonstrated that the overexpression of ΔNp63α inhibited doxorubicin-induced apoptosis in HaCaT keratinocytes, independently of wild-type p53 [70] and decreased UVBinduced apoptotic pathway in transgenic mice [71] probably as a strategy to evade oxidative stress-induced cell death and to promote long-term cellular survival in cooperation with the antiapoptotic protein Bcl-2 [72].

Limitations and Strengths of the Study
There are some limitations of our study that should be recognized: (i) We did not provide in vivo data in support of our findings in HaCaT cells, and no in vitro model system can perfectly recapitulate the in vivo complexity of a disease in terms of immunological response and microenvironmental interactions

11
Oxidative Medicine and Cellular Longevity (ii) We used a single cell line; however, HaCaT cells are widely used to study keratinocytes transformation upon UVA and UVB radiation [73] and are considered a valuable tool to study skin tumor promotion [33,37]. In fact, despite being nontumorigenic [74] and approximating normal keratinocytes, HaCaT cells harbor UV-type p53 mutations [75] and thus are considered at an early stage of the multistep process of skin carcinogenesis [76].
(iii) We used only UVA rays; despite representing the vast majority of UV received on Earth, the combination of UVA and UVB better simulates the solar radiation. However, no significant increase in the number of UVB-induced skin tumors in mice treated with HCTZ was reported [57].
The strengths of the study, making it translationally relevant, are as follows: (i) We tested HCTZ in a range of concentrations reflecting the plasma levels of doses used in clinical practice [25,26]; despite the lack of data on the accumulation of HCTZ in the skin, it is reasonable that HCTZ reaches the skin in a sufficient amount since patients taking HCTZ often experience cutaneous photosensitivity reactions; the use of therapeutic concentrations is also relevant for mimicking the actual human exposure.
(ii) We applied a cumulative weekly UVA dose of 20 J/cm 2 that mimics a human exposure of approximately 1 hour to midsummer sun in Paris [27], and it is below the UVA minimal erythema dose for I-II skin phototypes [77]; moreover, the long-term treatment resembles the multiple irradiations and chronic HCTZ treatment that may drive the cancerogenic process from cells able to survive upon insults and transform.
(iii) The morphological and molecular features described resemble some aspects of dysplastic lesions described by Coussens and Hanahan [78] in a mouse model of squamous carcinoma and, at least in part, those found in precursor lesions and in nonmelanoma skin cancers in humans.

Conclusions
Our results demonstrate that the combined chronic exposure to HCTZ and UVA radiation is more prophotocarcinogenic than it would be expected from UVA alone and highlight the relevance of associating drug safety data with experimental evaluations to clarify the molecular mechanisms underlining adverse drug effects. In fact, HCTZ alone does not possess any direct damaging effect but rather potentiates that of UVA, in accordance with its photosensitizing properties. This further supports the recommendation made by the regulatory agencies for an accurate photoprotection especially while taking HCTZ.

Data Availability
All data used to support the findings of this study are included within the article.