Genetic inactivation of the Fanconi anemia gene FANCC identified in the hepatocellular carcinoma cell line HuH-7 confers sensitivity towards DNA-interstrand crosslinking agents

Background Inactivation of the Fanconi anemia (FA) pathway through defects in one of 13 FA genes occurs at low frequency in various solid cancer entities among the general population. As FA pathway inactivation confers a distinct hypersensitivity towards DNA interstrand-crosslinking (ICL)-agents, FA defects represent rational targets for individualized therapeutic strategies. Except for pancreatic cancer, however, the prevalence of FA defects in gastrointestinal (GI) tumors has not yet been systematically explored. Results A panel of GI cancer cell lines was screened for FA pathway inactivation applying FANCD2 monoubiquitination and FANCD2/RAD51 nuclear focus formation and a newly identified FA pathway-deficient cell line was functionally characterized. The hepatocellular carcinoma (HCC) line HuH-7 was defective in FANCD2 monoubiquitination and FANCD2 nuclear focus formation but proficient in RAD51 focus formation. Gene complementation studies revealed that this proximal FA pathway inactivation was attributable to defective FANCC function in HuH-7 cells. Accordingly, a homozygous inactivating FANCC nonsense mutation (c.553C > T, p.R185X) was identified in HuH-7, resulting in partial transcriptional skipping of exon 6 and leading to the classic cellular FA hypersensitivity phenotype; HuH-7 cells exhibited a strongly reduced proliferation rate and a pronounced G2 cell cycle arrest at distinctly lower concentrations of ICL-agents than a panel of non-isogenic, FA pathway-proficient HCC cell lines. Upon retroviral transduction of HuH-7 cells with FANCC cDNA, FA pathway functions were restored and ICL-hypersensitivity abrogated. Analyses of 18 surgical HCC specimens yielded no further examples for genetic or epigenetic inactivation of FANCC, FANCF, or FANCG in HCC, suggesting a low prevalence of proximal FA pathway inactivation in this tumor type. Conclusions As the majority of HCC are chemoresistant, assessment of FA pathway function in HCC could identify small subpopulations of patients expected to predictably benefit from individualized treatment protocols using ICL-agents.


Background
Fanconi anemia (FA) is a rare recessive disorder, characterized by congenital skeletal abnormalities, progressive bone marrow failure and an increased cancer susceptibility [1]. The disease is caused by bi-allelic mutations in one of 13 FA genes, all of which have now been identified [2]. The FA genes appear to act in a common pathway of DNA damage signalling and DNA remodelling, distal parts of which interact with regulators of cell cycle control and DNA repair, especially the repair of DNA interstrand-crosslinks.
There are three consecutive compartments of the FA pathway [3]. The proximal compartment consists of eight FA proteins (A, B, C, E, F, G, L, and M), which form a nuclear FA core complex upon activation. This complex functions as an E3 ligase and mediates the monoubiquitination of FANCD2 [4], which represents the central FA pathway protein. FANCI also becomes monoubiquitinated during this process [5,6] and cooperates with FANCD2 in the ID (FANCI/FANCD2) complex [7,8]. The activated proteins of the ID complex subsequently colocalize with proteins of the distal FA pathway compartment (FANCD1/BRCA2, FANCN, and FANCJ) and with other DNA-repair proteins such as RAD51 at sites of DNA-damage. Cells having a defect in one of the proximal FA core complex genes are deficient in FANCD2/ FANCI monoubiquitination and FANCD2/FANCI nuclear focus formation. Similarly, cells having a defect in one of the distal FA pathway genes FANCD1 or FANCN are deficient in RAD51 focus formation [8][9][10][11][12]. Cells with a defect in one of the ID complex proteins lack the respective protein and are defective in monoubiquitination of the other. Thus, inactivation of the FA pathway can comprehensively be identified at the cellular level by assays detecting FANCD2 monoubiquitination and FANCD2/RAD51 focus formation.
FA pathway inactivation occurs sporadically in a variety of tumor types of non-FA patients, suggesting a role of the FA genes in tumor suppression or maintenance of genomic stability among the general population. Distal FA pathway inactivation via mutations in FANCD1 occurs in familial cases of breast (2-25%) [13] and ovarian cancer (2-6%) [14], in familial cases of pancreatic cancer (17%) [15] and in sporadic cancers of various tumor entities [16][17][18]. In comparison, genetic inactivation of the proximal FA pathway appears to occur infrequently in tumors among the general population and has, in terms of GI cancers, yet only been reported in pancreatic cancer, where it was associated with rare mutations in FANCC or FANCG [19,20]. In addition, germline mutations of FANCC might contribute to the tumorigenesis or tumor progression of pancreatic cancer [20][21][22]. Finally, epigenetic inactivation of the proximal FA pathway via hypermethylation of FANCF has been reported in a variety of tumor entities [23][24][25][26][27], but its significance is not yet well understood [28,29].
Unlike the setting in FA patients, FA pathway-deficient tumors arising in patients of the general population harbor the FA gene defect exclusively in the tumor cells, whereas stroma and all other non-malignant cells lack the defect, thus representing a tumor-specific, absolute biochemical difference [30]. As FA pathway-deficient cells are hypersensitive to ICL-agents and PARP inhibitors, FA pathway inactivation in tumors represents a promising target for rational, genotype-based anticancer therapy [31][32][33][34][35][36][37][38].
The prevalence of FA pathway defects has not yet been systematically investigated in GI cancer. We assessed proximal and distal FA pathway function in 48 cell lines derived from gastric, pancreatic, colorectal, hepatocellular and cholangiocellular carcinomas applying assays for FANCD2 monoubiquitination and FANCD2/RAD51 focus formation [39]. We newly identified a single cell line, HuH-7, derived from a hepatocellular carcinoma (HCC), which exhibited a proximal FA pathway defect, ascribable to genetic FANCC inactivation. When compared to four other HCC cell lines, HuH-7 cells exhibited an increased sensitivity towards ICL-agents, which was reversible in these cells by genetic correction through FANCC overexpression. Our data represent the first evidence for genetic inactivation of the proximal FA pathway in HCC and further support the assumption that genetic inactivation of the proximal FA pathway is a rare event in solid tumors among the general population.

GI cancer cell line-screening identifies proximal FA pathway-deficiency in HuH-7 cells
48 gastrointestinal cancer cell lines (18 colorectal, 15 pancreatic, 8 gastric, 5 hepatocellular and 2 cholangiocellular carcinomas) were screened for proximal and distal FA pathway inactivation using FANCD2 immunoblotting, FANCD2 and RAD51 nuclear focus formation (Fig. 1A, Table 1). Results on FA pathway inactivation in pancreatic cancer cell lines have been previously reported and were confirmed, including defective FANCD2 monoubiquitination due to inactivating mutations in FANCG and FANCC in Hs766T and PL11 [20,22], respectively, and defective RAD51 focus formation due to an inactivating FANCD1 mutation in Capan-1 [17]. Among all samples newly tested in the present study, we identified one cell line without detectable FANCD2 monoubiquitination, the HCC line HuH-7 (Fig. 1A). To stimulate previously undetectable FANCD2 monoubiquitination, HuH-7 cultures were pre-treated with mitomycin C (MMC) at 100 nM for 24 h. Despite such challenge, HuH-7 did not display any trace of monoubiquitinated FANCD2. RKO cells and derived clones harboring an engineered disruption of FANCG or FANCC were used as controls [36] (Fig.  1B). Likewise, no FANCD2 focus formation was observed in HuH-7 after treatment with irradiation (IR) at 15 Gy 6 h prior to workup ( Fig. 2A). In contrast, RAD51 focus formation was readily detectable (Fig. 2B), confirming a proximal FA pathway defect, whereas distal FA/BRCA pathway function remained undisturbed.

Gene complementation studies reveal defective FANCC function in HuH-7 cells
To identify the FA gene responsible for proximal FA pathway inactivation, HuH-7 cells were transduced with retroviral constructs containing full-length cDNAs of FANCA, FANCB, FANCC, FANCE, FANCF, FANCG or FANCL. FANCD2 monoubiquitination was assessed following exposure of the transduced cultures to MMC. Restoration of FANCD2 monoubiquitination was observed only after complementation of HuH-7 cells with the FANCC gene construct, whereas all other constructs had no discernible effect (Fig. 3A). Consequently, exogenous FANCC protein expression was confirmed in FANCC-transduced HuH-7 cultures (Fig. 3B). To validate functional FA pathway restoration upon transduction of FANCC-cDNA, the cell cycle profiles of the transduced cells were analyzed 48 h after treatment with MMC. The pronounced G2 arrest, observed in non-FANCC-complemented HuH-7 cells was restored to normal specifically after transduction with FANCC, but not after transduction with any of the other FA or control constructs (Fig.  3C).

HuH-7 cells harbor a homozygous inactivating nonsense FANCC mutation in exon 6
Direct genomic sequencing of the coding exons and adjacent intron portions of FANCC in HuH-7 revealed a homo-or hemizygous inactivating nonsense mutation c.553C > T, p.R185X in coding exon 6 ( Fig. 3D). No other sequence changes were identified.

Aberrant splicing leads to a truncated FANCC mRNA lacking exon 6 in HuH-7 cells
Using reverse-transcription PCR (RT-PCR) with primers flanking a region that includes exon 6, we identified an additional truncated FANCC gene transcript in HuH-7 cells, which was not detectable in four HCC control cell lines Chang, Hep3B, HepG2 and PLCPRF5 (Fig. 3E). Upon gel-separation of the two RT-PCR products amplified from HuH-7 cDNA, direct sequencing confirmed that the correctly sized band represented the reference FANCC sequence except for the described c.553C > T nonsense mutation, while the smaller band represented the reference FANCC sequence lacking exon 6, leading to a direct exon 5/7 junction (Fig. 3F).

HuH-7 cells are hypersensitive towards ICL-agents as compared to other HCC lines
Five HCC cell lines (Chang, Hep3B, HepG2, HuH-7 and PLCPRF5) were treated either with common ICL-agents (MMC, cisplatin and melphalan) or with common non-ICL chemotherapeutics (5-fluorouracil (5-FU), etoposide and doxorubicin). The sensitivity towards ICL-agents was clearly dissimilar between the FA pathway-proficient cell line panel and FANCC-deficient HuH-7 cells, with the latter displaying a distinct hypersensitivity (inhibitory concentration 50% (IC50) ratios [40] ranging between 3.4 to 18.6 as compared to the other HCC lines) (Fig. 4A, upper panel). In contrast, treatment with non-ICL chemotherapeutics did not distinguish HuH-7 as being most sensitive (Fig. 4A, lower panel). The most consistent difference in ICL-agent sensitivity between the FA-pathway proficient HCC cell lines and FANCC-deficient HuH-7 cells was observed using melphalan. This agent has previously been described to have the strongest effect among 8 ICL-agents tested in an isogenic FA cancer cell model   [36]. Cell cycle distributions of the five HCC cell lines, analyzed upon treatment with MMC for 48 h, revealed a pronounced G2 arrest (defined here by a fraction of >40% of cells in G2) [36] at low MMC concentrations (25 nM) only in HuH-7 cells (Fig. 4B), whereas a similar G2 arrest was achieved in a dose-dependent manner in all HCC cell lines except PLCPRF5 at higher MMC concentrations (50 to 100 nM) (Fig. 4C).

No evidence for inactivation of FANCC, FANCG or FANCF in 18 HCC tissue specimens
To assess the prevalence of proximal FA pathway inactivation in HCC, 18 surgical HCC tissue specimens were analyzed for inactivation of FANCC, FANCF or FANCG through screening for genetic mutations, epigenetic silencing through CpG hypermethylation or lack of mRNA-expression. On the genetic level, direct sequencing of the complete coding sequences of FANCC, FANCF and FANCG in 18 HCC samples identified a single not previously reported, heterozygous, synonymous FANCC sequence variant (c.813G > A) and two heterozygous, non-synonymous FANCG variants (c.20C > T, p.S7F; c.643C > A, p.Q215K) (Fig. 6A). No other sequence variants, especially no inactivating point mutations, small deletions, insertions, large intragenic deletions or complete homozygous gene deletions, were found. On the epigenetic level, HpaII-restriction assays yielded no evidence for FANCC-, FANCF-or FANCG-silencing through CpG hypermethylation (Fig. 6B) and consistently, FANCC-, FANCF-and FANCG-mRNA was detectable in all 18 HCC samples (Fig. 6C).

Discussion
We report here the identification and functional characterization of an inactivating nonsense FANCC mutation in the HCC cell line HuH-7. This cell line was established from a well-differentiated HCC of a 57 year-old Japanese male patient [42], displays an aneuploid phenotype with an average number of 60 chromosomes, and is negative for HBV and HCV [43,44]. To our knowledge, this is the first evidence of genetic inactivation of the proximal FA pathway in a GI tumor entity other than pancreatic cancer.
The identified FANCC nonsense mutation c.553C > T, p.R185X in HuH-7 represents a known FA mutation, first described by Gibson et al. [45]. Interestingly, non-splice site nonsense mutations can cause exon-skipping through aberrant splicing [46], and accordingly, the c.553C > T mutation has been reported to cause partial transcriptional skipping of exon 6 of FANCC in an FA patient [45], a mechanism confirmed for HuH-7 in our study.
Unfortunately, no matched non-malignant tissue is available for HuH-7, precluding definitive genomic copy number analyses in regard to whether the identified FANCC mutation represents a homo-or hemizygous mutation in this cell line. However, according to copy number analyses by the Sanger Institute (Cambridge, UK) using high-density single nucleotide polymorphism (SNP) arrays (SNP 6.0) [47], HuH-7 harbors three nearly identical copies of the chromosomal arm 9q, where FANCC is located at 9q22.3, as evidenced by virtually exclusive homozygosity of all SNPs assessed on 9q. According to proposed evaluation models for the identification of LOH events where no matching normal tissue is available, these data are strongly suggestive (although not definitely evidentiary) of allelic loss of one copy of chromosome 9q including the non-mutated FANCC allele in the original tumor (or its precursor cells), followed by repeated duplication of the remaining chromosome 9q, including the mutated FANCC allele, later on [48][49][50]. Typical recurrent numerical chromosomal aberrations in HCC include losses on 1p, 4q, 8p, 13q, 16q, and 17p and gains on 1q, 8q, 16p and 20q [51]. Although chromosome 9 is rarely clonally altered on the cytogenetic level in HCC, LOH has been reported for several regions on chromosome 9 including the loci of the FANCC (9q22.3) and the FANCG (9p13) genes [52].
FA pathway defects in tumors require bi-allelic inactivation of one of at least 13 FA genes. On the one hand, these bi-allelic mutations could both be inherited, as applies to tumors occurring in FA-patients. On the other hand, mono-allelic germline mutations of distal FA pathway genes, such as FANCD1, FANCN or FANCJ, confer low to medium penetrance susceptibility for breast/ovarian cancer [12,13,53] and, as applies to FANCD1 and FANCN, also for pancreatic cancer [15,[54][55][56][57]. In addition, inherited mono-allelic mutations of proximal FA pathway genes have been associated with the predisposi-

B CULLIN-1 H u H -7 H u H -7 /e v H u H -7 /F A N C C H e L a /e v H e L a /F A N C C
tion for or the accelerated development of certain tumors [21,54,55,58]. In particular, germline mutations of FANCC have been described in pancreatic cancer, associated with LOH in the tumor [21,22]. However, germline and somatic changes in FANCC and FANCG may have comparatively low penetrance for pancreatic cancer [55], which is supported by studies investigating germline mutations of upstream FA pathway genes in sporadic, yet FA-typical tumors among the general population [59]. Nevertheless, as the FANCC mutation in HuH-7 reported in our study represents an established FA mutation and was therefore most likely present in the germline of the patient in mono-allelic form, our data might indicate an increased risk for the development of HCC in individuals of the general population harboring this or other FANCC mutations.
The occurrence of an FA-associated FANCC mutation in HCC could also denote a tissue-specific susceptibility for the development of HCC in FA patients; The majority of solid malignancies seen in FA patients consists of head and neck or gynaecologic carcinomas (5.3%), as reported in a large meta-analysis of 1300 cases [60], but also 2.8% of all FA patients developed liver tumors. These tumors manifested at a significantly younger age than other solid malignancies (median age: 13 years for liver tumors as compared to 26 years for other solid malignancies). In fact, the cumulative probability of liver tumors in FA patients has been estimated to be 46% by age 50 [60]. However, the significance of these data regarding a potential liver-specific cancer susceptibility of FA patients is complicated by the observation that many liver tumors do not proceed to malignancy during the life span of FA patients [60,61]. In addition, there appears to be a strong association between androgen therapy, which is frequently used for the treatment of bone marrow failure in FA, and the occurrence of liver tumors [60][61][62][63]. Nevertheless, HCC represented the majority (58%) of tumors of 36 FA patients with androgen-related liver tumors in one study [61]. Thus, the association of FA with HCC could be attributable both to the primary tumorigenic effects of   We demonstrated that HCC cells having an inactivated FA pathway display the classic cellular FA phenotype, including a specific hypersensitivity towards ICL-agents, illustrated in HuH-7 by a pronounced cell cycle arrest in G2 upon treatment with MMC at low concentrations and a strongly decreased proliferation rate as compared to a panel of non-isogenic HCC lines. Importantly, this ICLhypersensitivity phenotype was reversed using an isogenic HuH-7 model of exogenous FANCC expression, confirming that ICL-hypersensitivity in these cells was attributable specifically to inactivation of FANCC. Of note however, IC50-ratios between FANCC-deficient and FANCC-proficient cells were partly smaller in the isogenic model than could have been expected from our results using the non-isogenic model. This observation could be attributable to FA pathway-independent ICLsensitivity differences among the non-isogenic HCC cell lines, but could also provide further support for our previous hypothesis that constitutive exogenous FANCC over-expression does not completely substitute for physiologically regulated, endogenous FANCC expression [37,38].
It is well established that systemic chemotherapy lacks effectiveness in unselected HCC patients [64,65] and HCC are therefore considered largely chemoresistant, at least partially explaining the poor prognosis of this tumor entity [66]. Accordingly, guidelines are currently lacking also regarding the choice of chemotherapeutic agent to use in transarterial chemoembolization (TACE), one of the major treatment modalities for non-surgical patients at advanced HCC stages [67]. Our data indicate that non-FA patients having a FA-deficient HCC might predictably benefit from treatment using ICL-agents. Consequently, assessment of FA pathway function in HCC could facilitate individualized therapeutic approaches, using genotype-based patient stratification in regard to both systemic chemotherapy and TACE.
To get an estimate of the prevalence of FANCC inactivation in HCC, we sequenced cDNA derived from 18 surgical HCC tissue specimens to screen for genetic FANCC inactivation. We further screened these samples for hypermethylation of the FANCC promoter region and for lack of FANCC mRNA expression, as epigenetic FANCC inactivation has previously been reported in acute leukaemia and breast cancer [68,69]. Additionally, we included FANCG and FANCF in these analyses, as FANCG represents another proximal FA gene that has been described to be mutated in GI cancer, specifically in pancreatic cancer [20,22], while FANCF has been reported to be epigenetically inactivated in various tumor types [23][24][25]27]. On the genetic level, we found no further inactivating alterations, especially no evidence for complete homozygous gene deletions, inactivating point mutations, small deletions or insertions, in FANCC, FANCG or FANCF, respectively. The detected FANCG variant c.20C > T, p.S7F has been reported in an FA patient of the complementation group G in addition to pathogenic FANCG mutations [70]. There is no information available on the nature of the c.643C > A, p.Q215K variant in FANCG. However, LOH or a second sequence variant was not detected in that tumor either. Additionally, since only two to three overlapping PCR reactions were used to amplify the complete coding sequences of FANCC, FANCG and FANCF, respectively, most potential large intragenic deletions would have been detected. However, this mechanism of proximal FA gene inactivation occurs almost exclusively in FANCA, whereas it appears to be extremely rare in FANCC and has not at all been described in FANCG [20,[71][72][73]. On the epigenetic level, we found no evidence for hypermethylation of FANCC, FANCG or FANCF in any of the 18 samples. Consistently, FANCC, FANCG and FANCF were expressed in all samples on the mRNA level.
Our negative screening results for proximal FA pathway inactivation in HCC were not unexpected, as a hypothetical high prevalence should have become evident earlier during clinical trials -manifesting as a selective chemosensitivity of the majority of HCC towards ICL-agents. Nevertheless, the lack of FA mutations in 18 HCC does not rule out rare occurrences of proximal FA pathway inactivation in HCC and is consistent with previous reports on the low prevalence of proximal FA pathway inactivation in various tumor entities among the general population [20][21][22]26,29,68]. Future studies applying a higher sample number are required to definitely determine the prevalence of FA pathway inactivation in HCC.

Conclusions
In summary, we identified a HCC cell line harboring an inactivating mutation of the FANCC gene, specifically causing proximal FA pathway inactivation and the classic cellular ICL-hypersensitivity phenotype. Assessment of FA pathway function in HCC could thus represent a novel approach to identify small subgroups of HCC patients expected to predictably benefit from individualized treatment protocols using ICL-agents. However, as proximal FA pathway inactivation occurs at low frequency in HCC and other tumors among the general population, the development of rapid, economic and clinically applicable screening tests for proximal FA pathway inactivation, particularly in primary tumor tissues, remains an indispensable requirement to facilitate future clinical studies [38].

Nuclear focus formation assays
Experiments were performed as described before [74]. In brief, cells were grown on coverslips until ~80% confluency and were exposed to ionizing γ-radiation (IR) at 15 Gy using a cesium-137 irradiator. After incubation for 6 h, the cells were fixed using 3.7% formaldehyde and -20°C methanol. The cells were permeabilized using Triton X-100 and incubated in blocking buffer (PBS +2% bovine serum albumin +0.5% Triton X-100) for 30 min. Consecutively, the cells were labelled using antibodies against FANCD2 (FI-17) or RAD51 (H-92) (Santa Cruz), respectively, for 2 h at room temperature. After washing, Alexa 488 goat anti-mouse or anti-rabbit antibody (Invitrogen) was applied for 1.5 h. Nuclei were counterstained using Hoechst 33342 (Roche Diagnostics, Mannheim, Germany), mounted and analyzed using a fluorescence microscope and Axiovision Software (Carl Zeiss AG; Oberkochen, Germany). The settings were kept identical for all samples.

gDNA sequencing
High molecular weight genomic DNA was prepared using a salting-out technique. Amplification of the FANCC exons was performed using published primer sets [75]. PCR products were purified using the GFX PCR DNA and Gel Band Purification kit (GE Healthcare). DNA sequencing of PCR products was performed using ABI-PRISM big-dye terminator chemistry on the ABI 310 instrument (Applied Biosystems, Darmstadt, Germany).

Cell proliferation assays
The assays were performed over a broad range of concentrations covering 100% to 0% cell survival. 1,500-2,000 cells/well were plated in 96-well plates, allowed to adhere, and treated. Following incubation for 6 d, the cells were washed, lysed in 100 μl H 2 O, and 0.5% Picogreen (Molecular Probes, Invitrogen) was added. Fluorescence was measured (Cytofluor Series 4000, Applied Biosystems) and growth inhibition calculated as compared to the untreated control samples. Three independent experiments were performed per agent, with each data point reflecting triplicate wells. Error bars represent standard error of the mean (SEM) from three experiments.

Cell cycle analyses
Cells were seeded in 12-well plates and treated in duplicate for 48 h using various MMC concentrations (ranging from 25 to 200 nM) or were left untreated. The cells were fixed, stained with propidium iodide, and the DNA content per cell was measured using flow cytometry (FACS-Calibur, Becton Dickinson, Heidelberg, Germany). The data were analyzed using CELLQuest Pro software (Becton Dickinson). Alternatively, unfixed cells were stained with DAPI at a final concentration of 1 μg/ml in a buffer containing 154 mM NaCl, 1 mM CaCl 2 , 0.5 mM MgCl 2 , 0.1 M TRIS, 0.2% BSA, and 0.1% NP40 for 30 min in the dark. Univariate flow histograms were recorded on an analytical, triple-laser equipped flow cytometer (LSRII, Becton Dickinson) using UV excitation of the DAPI dye. The resulting cell cycle distributions reflecting cellular DNA content were quantified using the MPLUS AV software package (Phoenix Flow Systems, San Diego, CA).