Loss of FANCD2 and related proteins may predict malignant transformation in oral epithelial dysplasia

Background: Predicting malignant transformation in oral epithelial dysplasia(OED) is a clinical challenge. The higher rate of malignant transformation in non-smokers supports an endogenous aetiology. Loss of FANCD2 and associated proteins could lead to genomic instability and oncogenesis. Patients & Methods: Longitudinal archival samples from 40 individuals with OED from time of diagnosis to the most recent review in 23 stable OED; or until excision of the SCC in 17 unstable OED undergoing malignant transformation. Histopathological reassessment, immunohistochemistry for FANCD2 and Western blotting for phosphorylation/monubiquitination status of ATR, CHK1, FANCD2 and FANCG were undertaken on each tissue sample. Results:Decreased expression of FANCD2 was observed in the diagnostic biopsy of OED lesions which later underwent malignant transformation. Combining the FANCD2 expression scores with histological grading more accurately predicted malignant transformation (p=0.005) than histology alone and correctly predicted malignant transformation in 10/17 initial biopsies. Significantly reduced expression of total FANCD2, pFANCD2, pATR, pCHK-1 and pFANCG were observed in unstable OED. Discussion: There is good evidence that defects in the DNA damage sensing-signalling-repair cascade are associated with malignant transformation in OED. Loss of post-translational modification in FANCD2 and related proteins, was more predictive of malignant transformation when compared to clinicopathological parameters.


Introduction
In the management of oral epithelial dysplasia (OED), histopathological grading and clinical determinants of malignant transformation to oral squamous cell carcinoma (OSCC) have been the primary influence in the treatment approach adopted (1)(2)(3), despite the numerous studies identifying putative molecular and other predictors of malignant change (4)(5)(6). Most studies aimed at identifying a molecular or pathological marker of malignant change have failed to undertake correlation with longitudinal clinical outcomes, so their translational value has been diminished leading to the lack of clinical application (7)(8)(9). Furthermore, they have mostly not been formally validated in independent series -this approach is difficult as the event rate of malignant transformation is low, and studies are therefore prolonged (3). There is a notable paucity of multi-centre/collaborative protocols, and, where malignant transformation can be predicted, there remains some uncertainty about recommended treatment options. The clinical outcomes of a cohort of patients managed in the Liverpool Multidisciplinary Oral Dysplasia Clinic, identified nonsmoking status and the non-homogenous appearance of OED as the strongest independent predictors of malignant transformation (HR 5.9 and 2.3 respectively) (1). The estimated malignant transformation rate in this study was 22% over 5 years. The more aggressive behaviour of lesions observed in non-smoking patients (or light smokers) with OED supports an endogenous aetiology: while this might seem counter-intuitive, this trend has been seen in other models of carcinogenesis (10,11).
The incidence of head and neck squamous cell carcinoma (HNSCC) in patients with the cancer prone syndrome, Fanconi Anaemia, is 1400 times greater than that of the general population (12,13) and occurs earlier in life. The Fanconi Anaemia pathway (FAP) removes interstrand crosslinks (ICL) lesions and facilitates homologous recombination repair of DNA double-strand breaks and is an integral component of the DNA damage repair mechanism which maintains genomic stability in healthy individuals (14)(15)(16). The FAP consists of 22 proteins (FANCC-FANCW) and has 3 main components (17) (Figure 1.1): (i) The FANCM-MHF1-MHF2 complex senses ICLs and localises to the DNA, acting as a recruitment site for the core complex through FANCM-FANCF interactions. The main role of the core complex is to facilitate ubiquitination of (ii) FANCD2 and FANCI protein dimers. The core complex mediates FANCD2/FANCI monoubiquitination, at residues K561 and K532 through FANCL E3 ubiquitin ligase activity, which then activates (iii) downstream effector proteins which include nucleases, translesion polyermases, homologous recombination proteins, and de-ubiquitinases to complete the DNA repair process and to attenuate FAP signalling. The detection of prognostic markers of DNA damage such as singlestranded DNA and replication stress, leads to the ataxia telangiectasia rad-3 (ATR) and checkpoint 1 (CHK1) kinases mediated cell cycle arrest in G2 or S phase, and activation of key components of the FA pathway via phosphorylation at several functional residues, including; FANCA at S1449, FANCM at S1045, FANCD2 at S717, S222, S331 and T691, FANCG at S387, S383 and S7, and FANCE at S374 and T346. (18)(19)(20)(21).
If left unrepaired, ICL lesions progress through the cell cycle, resulting in the stalling of replication forks and the eventual formation of double-strand breaks, resulting in genomic instability. (22).
We hypothesise that in transforming OED (particularly in non-smokers), an aberration in the DNA damage sensing, signalling or repair pathways leads to accumulation of DNA mutations and malignancy. These aberrations in the tumour suppressor mechanism leads to phenotypical and genotypical changes which occur in the early phase of carcinogenesis. In OED, the potential exists for identification of these aberrations in these lesions with potential for malignant transformation.
The objective of this study was to investigate the status of FANCD2 and related proteins (ATR, Chk1 and FANCG) in the DNA damage repair pathway of OED in patients who presented to the Liverpool Multidisciplinary Oral Dysplasia clinic, specifically to elucidate if this information could be obtained from the initial diagnostic biopsy as a predictive tool which might have been utilised to influence clinical management, and to correlate these findings to clinicopathologic characteristics.

Patients
Forty patients with OED, were identified from the clinical cohort in Chapter 2 and were included in this study after giving informed consent. All specimens for this study additionally met the additional criteria that the diagnostic FFPE block had not been previously utilised for translational research. The tissue size on the diagnostic biopsy FFPE block allowed for four 0.6mm core to be extracted for RNA extraction/Western blotting (23 non-transforming OED -NT and 17 malignant transforming OED). Four micrometre sections of the initial diagnostic incisional biopsy FFPE from each patient were stained with haematoxylin and eosin (H&E) and reported blinded and independently by two oral and maxillofacial pathologists to confirm the presence of OED (Initial attempts to construct a tissue micro-array was unsuccessful as the thickness of the incisional biopsy FFPE specimen was variable and inadequate to allow reproducibility of sections which enabled consistent and meaningful comparison (7)). In this study, the WHO 2017 dysplasia classifications (23) were then grouped as follows: Group 0: severe dysplasia and Group 1: mild or moderate dysplasia FANCD2 Immunohistochemistry FFPE sections from each clinical timepoint, including the initial diagnostic biopsy, for these 40 patients were stained with anti-FANCD2 antibody (F117: sc-20022, Santa Cruz diluted 1/100) and Biogenex Supersensitive Polymer HRP detection kit (Launch Diagnostics QD430-XAKE) as previously described (11). The expression of FANCD2 was scored blinded by two independent observers (Asterios Triantafyllou and Michael Ho), one being an oral and maxillofacial pathologist. The stained sections were identified by their pathology identification number alone, thus the observers were blinded to the clinical details and outcome of OED i.e. either stable OED or OED which underwent malignant transformation. Consequently, staining was classified by localisation, extent and intensity using a descriptive binary scoring system (Table 1). Any discrepancy in scoring was reviewed jointly and a mutually agreed score determined. Control tissue from normal and OED areas of tissue adjacent to OSCC from a different cohort of anonymised patients (n=3) were stained using the same method.

Western blotting
Areas with the highest grade of dysplasia in the incisional biopsies were marked on the H&E stained sections and 0.6mm cores were obtained from the corresponding FFPE blocks (minimum of 2 cores from each block). Cores of the same diameter were obtained from anonymised controls as follows: normal tissue from oesophagus (n=5), areas of OSCC from a different cohort (n=3) and tissue with normal histological architecture located adjacent to OSCC (n=3).
The protein extraction protocol utilised was modified from one previously published (24): FFPE cores were placed in Eppendorf safe-lock tubes (Eppendorf, Hamburg, Germany) and deparaffinised by incubation at room temperature in xylene for 10 min. After each incubation, the tissue was pelleted at 12 000×g for 3 min, and incubation/centrifugation steps were repeated two more times. The deparaffinised tissue pellets were then rehydrated with a graded series of ethanol (100% for 5 minutes, 90% for 5 minutes, 70% for 5 minutes and a further 5 minutes in 70%) and centrifuge at 1300 RPM for 30 secs to remove excess ethanol. Following this, 150µl of Laemmli sample buffer (Sigma-Aldrich) were added. All samples were subjected to high-temperature extraction at 100°C for 20 min, and then cooled in ice for 5 minutes. Extracts were centrifuged for 30 min at 13000 RPM and the supernatant collected and stored at −20°C until needed.
The expression of β-actin, ATR, pATR (s428), CHK1, p-CHK1 (s317), FANCD2 (nonand mono-ubiquitinated isoforms), pFANCD2 (s331), FANCG and pFANCG (s7) were assessed in a blinded fashion (to clinicopathologic details and outcome: stable OED or OED which underwent malignant transformation) by western blotting using antimouse or anti-rabbit fluorescently labelled 680/800 secondary antibodies (1:10 000 dilution) (Invitrogen, Paisley UK) ( Table 2) (25), extracted protein was separated on SDS-polyacrylamide (PAGE) gels: samples were prepared for loading, by the addition of 20 μL of protein extract to 40 μL of Laemmli loading buffer and immediately heated at 100 °C for 5 minutes, prior to loading. The samples were loaded on to a 15% SDS-PAGE gel which were ran overnight at constant 50 V. The protein was then transferred to a nitrocellulose membrane -4 hours at constant 350 mA (26). This was blocked in LI-COR buffer for 1 hour at room temperature prior to the addition of specific antibodies at the specified dilutions.

Patient outcomes
The demographic and clinicopathologic details of the forty patients included in this study are summarised in Table 3.
As previously published (1) patients were progression-free for 18, 52 and 107 months, respectively, following diagnosis. The median time to malignant transformation for T OED in patients with the higher risk binary score at diagnosis was 14.8 months when compared with T OED with a low risk binary score, 44.7 months (p = 0.1).

FANCD2 immunohistochemistry
Nuclear and cytoplasmic staining for FANCD2 was absent in morphologically normal oral epithelium ( Figure 1A), however, FANCD2 staining increased in intensity in noninvasive dysplastic epithelium and was then lost again in micro-invasive OSCC ( Figure 1B).
A decrease in the intensity of both nuclear and cytoplasmic staining were observed in the first diagnostic biopsy of OED lesions that were destined to undergo malignant transformation when compared with non-transforming OED, although the distribution of staining was not statistically different (the quantitative, rather than qualitative, element of FANCD2 staining was predictive of malignant transformation) ( Figure 2; Table 3).
Therefore, only intensity scores were incorporated into a final composite FANCD2-OED score as below. In OSCC sections from transformed OED lesions, both intensity and distribution of FANCD2 staining were reduced compared with the first diagnostic biopsy from the same patient (Table 4).
Combining the FANCD2 nuclear and cytoplasmic intensity scores with the histological grading of OED (mild/moderate dysplasia assigned score of 1 and severe dysplasia/carcinoma-in-situ/SCC assigned score 0) produced a score that more accurately predicted transformation (p = 0.005) ( Table 5), with a cut off score of 1 or less being significantly associated with a higher risk of malignant transformation in OED (p = 0.001). This FANCD2-OED Risk Score in all the retrieved OSCC archival tissue from OED which underwent malignant transformation was <1 (n = 15; 2 archival OSCC specimen not retrievable).
Of the 2 NT OED lesions with low (≤1) FANCD2-OED Risk scores, one patient presented with severe dysplasia which was excised shortly after presentation and has now been progression free for 52 months and one presented with mild dysplasia that has been progression-free for 19 months. Eighty-three percent (  In this analysis, moderate dysplasia was classified together with mild dysplasia as 'low risk' while severe dysplasia was classified as 'high risk'. Other researchers have classified moderate dysplasia as 'high risk', but recent discussion in the literature suggests that both suggested binary classifications are simplistic (33,34).
Alternative suggestions for binary classification of oral dysplastic lesions rely heavily on pathological interpretation, which could be prone to intra and interobserver reliability problems (35,36). It is acknowledged, therefore, that the proposed histopathology/IHC classification in its current format will have skewed the moderately dysplastic lesions in this study towards higher FANCD2-OED scores and is thus prone to false negatives, although it still performed better than histopathology alone. Its strength may be in identifying lesions which will NOT transform, but development into a routine, clinical test requires a more robust definition of reduced immunostaining utilising data from a larger number of transforming and non-transforming lesions. This study is best categorised as proofof-concept as the design and the cohort are not adequate for a robust prognostic biomarker study.
When the NT and T groups were compared, there were significant differences in site, smoking aetiology and appearance between the two cohorts. These are such strong predictors of transformation in our modest cohort that it was impossible to adequately match the two groups. The validity of diagnosis of the grade of OED from an incisional biopsy could potentially be questioned as the issue of heterogeneity of OED especially in a large lesion is a valid consideration. In the Liverpool Oral Dysplasia MDT clinic, the biopsy of OED lesions is carried out by senior surgical members of the team where the area of most clinical concern is sampled. Experience in our practice would suggest that concordance of initial biopsy and definitive histopathology diagnoses were very high and had not adversely impacted on patient outcomes. This was demonstrated in the reported cohort of patients where patients who have their OED excised did not have their grade of OED upgraded but the more often scenario was that the area with the most severe grade of OED has been completely excised by the incisional biopsy.
Therefore, the findings observed for FANCD2 immunohistochemistry and Western blotting were valid from the statistical and clinical perspectives.
Six of the twenty-three NT lesions (that we would have expected to transform) were totally excised-therefore some of the differences observed between NT and T groups might potentially relate to the method of treatment rather than the inherent cancer risk of the sample (Table 6 The hypothesis that malignant change in OED involves alteration of the FA pathway is supported by both our immunohistochemistry and western blotting data. These reinforce the previously described differences observed in the appearance and site of OED in these two groups of patients(1) but are not simply a reflection of these differences as no associations were observed between site or appearance of lesion and PTM of these proteins. Samples from T OED lesions showed significant reduction in the phosphorylation of ATR, CHK1, FANCD2 and FANCG in comparison to nontransforming samples, indicating a lack of ATR-CHK1 activation following DNA damage and/or replicative stress. It may be argued that these observations are due to a lack of stimuli in the transforming group as they have a preponderance of nonsmokers, but analysis of our data does not support this as no significant difference in PTM expression was observed between smokers and non-smokers. The ability of CHK1 to phosphorylate several functionally important sites for optimal function and activation of the FA pathway appears to be compromised in these patients which, it is proposed, will lead to the impairment of the functionality of the FA core complex and lead to a reduction in subsequent HRR activity (20,25,29,38). In contrast, non-transforming samples showed high levels of FANCD2 s331 and FANCG s7 phosphorylation, indicating that these sites were successfully phosphorylated by activated CHK1, and could function effectively in DNA repair, thus reducing the burden of DNA damage in these cells and reducing the risk of malignant transformation.
FANCD2 monoubiquitylation, which is thought to be promoted by ATR-CHK1 mediated FANCD2 phosphorylation (19,30), is a critical step in FA pathway activation(29, 39) and evidence suggests that a reduction in FANCD2 monoubiquitylation has a greater influence on genomic instability than down regulation of FANCD2 expression (40). In our study, it was observed that transforming samples have lower levels of FANCD2 monoubiquitylation (FANCD2 L expression) compared to non-transformers, and interestingly, they consistently displayed lower levels of total FANCD2 expression. These findings agree with our concurrent immunohistochemistry data where we observed a lack of FANCD2 protein expression in T-OED compared with NT-OED lesions.
The direct evaluation of the DNA sensing-signalling-damage repair cascade in OED has not been previously reported, although there is recent evidence that individuals with reduced, systemic, double strand break repair capacity are more prone develop to head and neck cancer (38). It has been suggested that activation of DNA damage response might be protective in the early stages of oral carcinogenesis, but progressive deregulation over time could eventually result in the failure to suppress malignant transformation (41). The results of the current study indicate that additional evaluation of these pathways is worthwhile to understand their capability to predict malignant transformation in OED at the initial diagnostic biopsy, especially as time to transformation may be as long as 7 years (1,3). Loss of heterozygosity (LOH) status at putative tumour suppressor gene loci (3p14, 9p21, 9p22 and 17p13) is currently the most reliable predictor in malignant transformation in OED (42)(43)(44) and there is evidence to suggest that LOH is secondary to homologous recombination deficiency/DNA damage repair deficiencies at 15 cancer sites, including head and neck squamous cell carcinoma (45), indicating a possible link with the current data.
In the process of validating the FANCD2-OED Risk score, a multi-centred setting with larger sample size would be desired, to control for as many variables as possible    Table 3 Demographic and clinic-pathological features of patients (n = 40) (* at initial diagnostic biopsy)    Table 6 Normalised protein expression in initial diagnostic biopsy of NT OED and T OED lesions Figures Figure 1 Representative example of FANCD2 immunohistochemistry during progression to carcinoma. Representative example of Western blot expression analysis of DNA damage sensing and rep