HSP90 inhibition sensitizes head and neck cancer to platin-based chemoradiotherapy by modulation of the DNA damage response resulting in chromosomal fragmentation

Cal27 (CRL-2095) and FaDu (HTB-43) cells were obtained from ATCC. LICR-LON-HN5 were a kind gift from Suzanne Eccles (The Institute of Cancer Research, Sutton, London, UK). All three cell lines were HPV negative and p53 mutant. Cells were cultured in DMEM (Invitrogen, Paisley, UK) supplemented with 10% FCS, 2 mM L-glutamine and 1% penicillin/streptomycin in a humidified incubator at 37 °C with 5% CO₂. Cells were tested for mycoplasma using the eMyco PCR kit from IntroBio (Seongnam-Si, South Korea) and authenticated by STR profiling (Bio-Synthesis Inc, Texas, US).

AUY922 was kindly donated by Novartis in the form of the mesylated salt. Cisplatin was from Teva Hospitals (Castleford, UK). In western blot, confocal and FACS analysis 10 nM AUY922 or 10 μM cisplatin was used unless otherwise indicted. AUY922 was added 16 h before cisplatin or irradiation. Irradiation was carried out using an AGO 250 kV X-ray machine (AGO, Reading, UK).

Long-term survival in response to radiation was measured by colony formation assay. Cells were trypsinized, diluted and counted before seeding in 6-well dishes or 10 cm dishes at appropriate seeding densities. Cells were allowed to attach before addition of 5 nM AUY922 or DMSO only control for 16 h. Cells were exposed to 5 μM cisplatin for 3 h with cells subject to concurrent-cisplatin radiotherapy being irradiated immediately after cisplatin addition. After 3 h exposure to cisplatin, both cisplatin and AUY922 were replaced by drug-free medium. Colonies were fixed and stained in 5% gluteraldehyde, 0.5% crystal violet, with colonies containing more than 50 cells counted. Colony counting was performed both manually and by automated quantification using CellProfiler 2.0 (Broad Institute, MA, USA). Surviving fraction was calculated by normalization to untreated controls.

Medium and cells were harvested in PBS-containing 1 mM Na₃VO₄ and 1 mM NaF. Cells were pelleted before lysis in 50 mM Tris.HCl pH 7.5, 150 mM NaCl, 1% NP-40, 0.5% deoxycholate and 0.1% SDS. Samples were thawed on ice, centrifuged at 14,000 rpm for 20 min at 4 °C and supernatants quantified by BCA assay from Pierce (Leicestershire, UK). 30 μg total protein lysate was separated by reducing SDS-PAGE, transferred to PVDF (GE Healthcare, Bucks, UK) and blocked with 5% non-fat dry milk in TBS. The following primary antibodies were used: rabbit anti-HSP72 from Stressgen (Exeter, UK); rabbit anti-GAPDH, rabbit anti-ATR, rabbit anti-phospho-ATR (S428), rabbit anti-CHK1, rabbit anti-phospho-CHK1 (S345), rabbit anti-RAD51, rabbit anti-ATM, rabbit anti-phospho-ATM (S1981), rabbit anti-phospho-BRCA1 (S1524), rabbit anti-phospho-p53 (S15) and rabbit anti-phospho-H2Ax (S139) were purchased from Cell Signaling (MA, USA); rabbit anti-FANCA was purchased from Bethyl Laboratories (TX, USA). Secondary antibodies used were sheep anti-mouse IgG and donkey anti-rabbit IgG HRP from GE Healthcare (Buckinghamshire, UK). Chemiluminescent detection was carried out using immobilon western substrate from Millipore (East Midlands, UK). In vivo samples were processed using a Precellys®24 homogenizer from Bertin Technologies (Montigny, France).

Short hairpin sequences were cloned into the lentiviral shRNA plasmid pHIVSiren [26]. The plasmid pHIVSiren was kindly donated by Professor Greg Towers, University College London and was derived from a parent plasmid, CSGW (Prof Adrian Thrasher, University College London). FANCA and ATM short hairpin target sequences were 5’-GTGGCATCTTCACGTACAA-3’ and 5’-GTGGCATCTTCACGTACAA-3’, respectively. Scrambled short hairpin target sequence was 5’-GTTATAGGCTCGCAAAAGG-3’. Short hairpin containing pHIVSiren was co-transfected with the packaging plasmids psPAX2, pMD2.G into HEK293T cells using lipofectamine 2000 (Life Technologies, Paisley, UK). Viral supernatants were collected and target cells infected in the presence of 1 μg per mL polybrene.

Cells were stained for mitosis or DNA double-stranded breaks with rabbit anti-phospho-histone H3 S10 (DD2C8) AlexaFluor647 or anti-phospho-histone H2Ax S139 (20E3) AlexaFluor488 (Cell Signaling, MA, USA) using the manufacturer’s protocol. Cells were analyzed on an LSR II from BD Biosciences (Oxford, UK).

Cells were plated in 35 mm glass-bottomed dishes (Mattek, MA, USA). Samples were fixed in 4% PFA, permeabilized in 0.2% Triton X-100 and treated with DNaseI (Roche, West Sussex, UK). Cells were blocked in 1% BSA, 2% FCS in PBS before staining with rabbit anti-phospho-H2Ax S139 (γH2Ax), rabbit anti-RAD51, rabbit anti-53BP1, anti-phospho-BRCA1 (S1524), rabbit anti-phospho-p53 (S15) or mouse anti-phospho-ATM (S1981) (Cell Signaling, MA, USA) with goat anti-rabbit Alexafluor488 or goat anti-mouse Alexfluor546 as secondary antibodies (Invitrogen, Paisley, UK). Nuclei were counterstained with DAPI. Samples were imaged using a Zeiss LSM710 inverted confocal microscope (Zeiss, Jena, Germany). Automated quantification of foci in 100-300 nuclei per experiment was carried out using CellProfiler 2.0 (Broad Institute, MA, USA). Formalin-fixed paraffin embedded (FFPE) in vivo blocks were sectioned and antigen retrieved for RAD51 (pH9 Tris-EDTA) or 53BP1 (pH6 citrate buffer). Antigen retrieved slides were blocked, stained, imaged and quantified as outlined for in vitro samples above.

Female 5-6 week-old athymic BALBc nude mice (Charles River, UK) were used with all experiments, complying with NCRI guidelines. 2x10⁶ FaDu cells were injected subcutaneously. Developing tumors were distributed into groups containing a minimum of n = 8 per group, with matching average tumor volumes. AUY922 40 mg/kg in 5% dextrose was administered in three doses by i.p. injection on days one, three and five. Fractionated radiation treatment of the tumor consisted of a total dose of 6 Gy in 2 Gy fractions on day two, four and six. Cisplatin was administered as a single dose of 5 mg/kg on day four immediately before irradiation. Tumor volume was calculated as volume = (width × length × depth)/2 and was plotted as mean tumor volume for each group.

Statistical analysis was carried out using Graphpad prism (version 6.0f). Unpaired two-tailed student t-test was utilized for parametric analysis. Synergy was determined by Bliss independence analysis using the equation E_exp = E_x + E_y – (E_xE_y) [27]. E_exp is the expected effect if two treatments are additive with E_x and E_y corresponding to the effect of each treatment individually. ΔE = E_observed - E_exp. Synergy is represented by ΔE and 95% confidence intervals (CI) from observed data all above zero; addition to values above and below zero; antagonism where all values are below zero.

The ability of AUY922 to sensitize to cisplatin, radiation and CCRT was assessed in a panel of cell lines by clonogenic assay using the scheduling outlined (Fig. 1a). We focused our studies on p53mt since p53 pathway abnormalities exist in 85% of HPV-negative HNSCC (TCGA).

Clonogenic data are presented as surviving fractions, normalised to drug free control wells (Fig. 1b). Qualitative images of colonies for each cell line and condition are also shown with numbers indicating the number of cells plated in each well shown (Fig. 1c). Bliss independence analysis (Fig. 1d) indicated synergy for the combination of AUY922 and cisplatin, except HN5 in which the interaction was additive. Synergy for the combination of AUY922 with both radiation and CCRT was observed in all cell lines tested. Values for synergy between 0 and 0.2 are in keeping with those observed in recent radiosensitization studies for CHK1 and ATR inhibition [28, 29]. FaDu cells were substantially more sensitive to AUY922 and cisplatin monotherapies as well as AUY922 and cisplatin or CCRT combinations. HN5 cells were more resistant to monotherapies and were one to two orders of magnitude more resistant to the combination of AUY922 and CCRT compared to other cell lines.

We investigated the impact of AUY922 on CCRT-induced RAD51 focus formation. CCRT induced an increase in early RAD51 focus formation which was significantly reduced by AUY922 (Fig. 2a, b). Western blots looking at DDR signalling with radiation and cisplatin alone revealed similar levels of S1524 BRCA1 in all cell lines, but diverse phosphorylation of mutant p53 S15. FaDu cells were constitutively high for pS15 p53 with levels in HN5 cells rapidly decreasing after an early radiation induced spike (Fig. 2c).

To more conclusively investigate this difference we looked at nuclear staining of pS1524 BRCA1, pS15 p53 and 53BP1 focus formation. Nuclear intensity of pS1524 BRCA1 signalling increased due to CCRT and was statistically lower due to AUY922 (Fig. 2d). Nuclear intensity of CCRT induced pS15 p53 increased in all cell lines due to the addition of AUY922 (Fig. 2e). CCRT induced 53BP1 foci increased in all cell lines due to the addition of AUY922 (Fig. 2f). Overall it was observed that FaDu cells exhibited the highest basal levels of RAD51, 53BP1 and pS15 p53 with the largest number of CCRT induced RAD51 and 53BP1 foci. HN5 cell displayed the lowest basal DDR signalling pattern and low levels of RAD51 foci persisting at 24 h as well as low levels of pS15 mutant p53. Cal27s fell in between, with this pattern correlating to the results observed in clonogenic assays (Fig. 1b).

Previously reported as a HSP90 client protein, we investigated how FANCA and RAD51 depletion may impact ATM signalling (measured by autophosphorylation on S1981). AUY922 depleted both RAD51 and FANCA to similar levels in CAL27, FaDu and HN5 cells with canonical drug-on-target induction of HSP72 (Fig. 3a). Stable knockdown in CAL27 cells of FANCA and ATM by lentiviral shRNA is shown by western blot (Fig. 3b).

As expected, FANCA knockdown increased sensitivity to cisplatin and CCRT with no statistically significant difference between control and scrambled shRNA conditions (Fig. 3c). FANCA knockdown significantly increased basal RAD51 and pS1981 ATM focus formation in response to CCRT (Fig. 3d, e). CCRT-induced pS1981 ATM foci were further increased by the addition of AUY922 in all cell lines (Fig. 3e, f) coinciding with a reduction in RAD51 foci (Fig. 2b, Fig. 3d).

We then generated ATM knockdown cells to investigate the role of ATM in compensating for loss of RAD51 and FANCA due to AUY922. Knockdown of ATM in CAL27 cells resulted in increased sensitivity to cisplatin, radiation and CCRT (Fig. 3g). AUY922 was able to further sensitize to cisplatin, radiation or CCRT. Overall survival was profoundly decreased in all combinations vs scrambled.

Following on from evidence of increased ATM foci induced by AUY922, we looked at ATR-CHK1 signaling. In all cell lines, moderate decreases in phospho-ATR, total-ATR and total-CHK1 combined to give a substantial reduction in phospho-CHK1 signaling (Fig. 4b). In studying the impact of this inhibition on mitotic entry by phospho-histone H3 staining (Fig. 4c), AUY922 was found to induce a profound increase in the mitotic population. This was much less pronounced in CAL27 cells, while HN5 cells exhibited an increase in the mitotic population from 2.9 to 8.1% due to AUY922 treatment (data not shown).

Co-staining FACS analysis of both the mitotic marker phospho-histone H3 and the double-stranded break marker γH2Ax revealed that this AUY922-induced mitotic population became highly γH2Ax positive immediately after CCRT (Fig. 4d). Confocal microscopy (Fig. 4e) further confirmed the presence of high levels of γH2Ax foci in nuclei displaying a mitotic morphology post-CCRT, as well as chromosome fragments or missegregation. We quantified the presence of micronuclei at 24 h post CCRT (Fig. 4f). AUY922 alone increased micronuclei compared to basal and significantly increased micronuclei when combined with CCRT vs CCRT alone in all cell lines.

FaDu cells showed high levels of chromosomal fragmentation both basally and in response to AUY922 plus CCRT. HN5 cells showed low basal levels and the lowest number of micronuclei in response to CCRT plus AUY922. This pattern of micronuclei formation closely aligned to that observed for RAD51, 53BP1 repair foci formation and signalling into mutant p53 (Fig. 2b, e, f). This DNA repair foci pattern and micronuclei generation correlated to the differences in sensitivity observed in clonogenic assays (Fig. 1b) with FaDus being the most sensitive and HN5s the most resistant.

DNA damage signaling basally and in response to radiation was assessed in FaDus in vivo (Fig. 5a). Cisplatin treatment of FaDu xenografts was confirmed to increase DNA damage signaling 24 h after 5 mg/kg cisplatin injection (Fig. 5b, c). Depletion of HSP90 client proteins by AUY922 in FaDu xenografts was assessed at both 40 mg/kg and 80 mg/kg with three IP injections on days 1, 3 and 5. Tumor lysates were collected 24 h after final injection. HER2 as a known and highly sensitive HSP90 client protein was also assessed. Reductions in RAD51 and HER2 were observed at 40 mg/kg with increased S15 phospho-p53 signaling also detected (Fig. 5b, c).

The lowest dose of 40 mg/kg AUY922 was selected to look for sensitization effects in tumor volume experiments (Fig. 5d). Mice were treated when tumors reached 5-7 mm in width with groups composed of mice with equal average tumor volume. AUY922 40 mg/kg was administered by IP injection on days 1, 3 and 5. Radiation was delivered in 3 × 2 Gy fractions on days 2, 4 and 6 with 5 mg/kg cisplatin administered by IP injection before irradiation on day 4. CCRT combined with AUY922 was tolerable with average weight loss of <10% (data not shown). The addition of AUY922 successfully delayed time to reach 800 mm³ from 27 days for CCRT only to 34 days from CCRT plus AUY922.

To assess DDR signalling at the level of repair foci formation in vivo, staining for RAD51 and 53BP1 was carried out on sections from FFPE tumour blocks which were all fixed 16 h after the final radiation fraction. Cisplatin and radiation combined to increase RAD51 foci in vivo, with AUY922 at the 40 mg/kg dose used in therapy experiments able to reduce RAD51 focus formation (Fig 5e). 53BP1 focus formation as a result of radiation decreased due to the addition of cisplatin. AUY922 addition to CCRT in increased the number of 53BP1 foci detected. These findings are in line with those shown in vitro (Fig. 2b, f).