Differential transmission of the molecular signature of RBSP3, LIMD1 and CDC25A in basal/ parabasal versus spinous of normal epithelium during head and neck tumorigenesis: A mechanistic study

Head and neck squamous cell carcinoma (HNSCC) is a global disease and mortality burden, necessitating the elucidation of its molecular progression for effective disease management. The study aims to understand the molecular profile of three candidate cell cycle regulatory genes, RBSP3, LIMD1 and CDC25A in the basal/ parabasal versus spinous layer of normal oral epithelium and during head and neck tumorigenesis. Immunohistochemical expression and promoter methylation was used to determine the molecular signature in normal oral epithelium. The mechanism of alteration transmission of this profile during tumorigenesis was then explored through additional deletion and mutation in HPV/ tobacco etiological groups, followed byclinico-pathological correlation. In basal/parabasal layer, the molecular signature of the genes was low protein expression/ high promoter methylation of RBSP3, high expression/ low methylation of LIMD1 and high expression of CDC25A. Dysplastic epithelium maintained the signature of RBSP3 through high methylation/ additional deletion with loss of the signatures of LIMD1 and CDC25A via deletion/ additional methylation. Similarly, maintenance and / or loss of signature in invasive tumors was by recurrent deletion/ methylation. Thus, differential patterns of alteration of the genes might be pre-requisite for the development of dysplastic and invasive lesions. Etiological factors played a key role in promoting genetic alterations and determining prognosis. Tobacco negative HNSCC patients had significantly lower alterations of LIMD1 and CDC25A, along with better survival among tobacco negative/ HPV positive patients. Our data suggests the necessity for perturbation of normal molecular profile of RBSP3, LIMD1 and CDC25A in conjunction with etiological factors for head and neck tumorigenesis, implying their diagnostic and prognostic significance.


Microdissection and genomic DNA isolation
These were performed in two different sets: 1. Separately from basal/parabasal and spinous layers of normal oral epithelium adjacent to tumours. Basal/parabasal and spinous layers were identified in 5μm Hematoxylin and Eosin (H&E) stained serial paraffin/cryosections of normal oral epithelium adjacent to tumours (N = 40). After comparing the stained sections with unstained parallel tissue sections, layers were separated by Laser Capture Microdissection (LCM) procedure using PalmRobo software (Palm Microbeam, Zeiss, Germany) [13]. After demarcating basal/parabasal and spinous layers using a laser line, individual layers were laser-catapulted into separate microfuge tubes (Figs 1A and 2). DNA from each layer was then separately extracted by standard procedure using Proteinase K digestion, followed by phenol-chloroform extraction [10].
2. H&N lesions and blood. The 5μm H&E sections from H&N lesions were identified for dysplasia (N = 41) and tumor-rich regions (N = 143), enriched (>60% lesions) by microdissection under a dissecting microscope (Leica MZ16, Germany) and LCM for removing contaminating normal cells (Fig 1B, Fig 2A and 2B).Genomic DNA from these and blood were isolated by the previously mentioned procedure.

Deletion analysis
Deletion with informative and non-informative microsatellite markers was studied in dysplastic (N = 20) and HNSCC samples (N = 143) in Groups 1-4. In deletion mapping using microsatellites, a standard polymerase chain reaction was performed with [γ-P32] dATP endlabeled forward primer in a 20 μl reaction mixture. For non-informative markers, analysis was performed using SST as a control locus, followed by electrophoresis in a denaturing acrylamide gel and autoradiography, as described previously [6,7] (primers in S2 Table).

Detection of HPV 16 and 18
HPV infection was detected by PCR using primers (MY09/MY11) designed from the consensus L1 region of the virus, followed by typing positive samples for HPV16/18 using type specific primers designed from the E6 region of HPV 16 and LCR region of HPV 18. Results were validated by Southern hybridization using detection and type specific probes [17].

Clinico-pathological correlation
Risk associated with alterations and clinico-pathological factors was calculated using univariate (Fischer's exact test determining Odds ratio) and multivariate (Multivariate Cox proportional hazard regression model determining Hazard Ratio(HR)) analysis. Chi square (trend) was used to associate alterations with increasing tumor stages/grades. Kaplan-Meir method was used to plot survival curves to assess relationship between 5-year disease free survivals with/without alterations. Log rank test evaluated differences in survival between different parameters. All statistical tests were two-sided with probability (p) value <0.05 considered significant. All calculations were by softwares Epi Info 7 (CDC, Atlanta) and IBM SPSS 21 (SPSS, Chicago, IL).

Demography of the patients
Patients presenting with head and neck lesions were predominantly male, with a median age of 54 for dysplastic lesions and 50 for HNSCC and with buccal mucosa as the most common site affected (Table 1). Most patients were habitual users of tobacco.

Differential distribution of patients in tumour groups according to tobacco habit and HPV infection
Presence of HPV increased from dysplastic lesions (52.78%) to malignant tumours (62.43%, Table 1), the most common subtype being HPV 16 (data not shown). Moreover, by simultaneously considering two important etiological factors, tobacco usage (T) and HPV infection (H), patients were segregated into four Etiological Groups (1-4). By doing this, a disparate distribution of patients was observed with a higher fraction of patients being positive for both etiological factors tobacco and HPV (Group 4: dysplasia, 37.5%; HNSCC, 43.75%) ( Table 1).

Candidate genes show differential patterns of expression in basal/ parabasal versus spinous layers of normal head and neck epithelium but show similar low expression in head and neck lesions
RBSP3 showed low cytoplasmic/nuclear expression in the basal/parabasal layer of normal oral epithelium (14%, 31% cases respectively) with significant increased cytoplasmic expression in spinous layer (79%) (Fig 3A and 3B). Like basal/parabasal layer of normal oral epithelium, comparable frequencies of cytoplasmic/nuclear expression were obtained in dysplastic lesions (21%, 44% respectively) and HNSCC (29%, 41% respectively) (Fig 3A and 3B). In contrast, LIMD1 and CDC25A showed high nuclear expressions in both basal/parabasal (80.06%, 82.76% respectively) and spinous layers (80.06%, 74.71%) of normal epithelium with significant loss of their expression during development of dysplasia (63.88%, 52.78%) and HNSCC (43.42%, 47.37%) (Fig 3A and 3B). Comparing the patterns of their expression in etiological Groups 1-4, similar expression of all genes was observed in normal epithelium and during tumorigenesis (Fig 3C), indicating that etiological factors might not play a role in determining expression of the proteins.
Therefore, low expression of RBSP3 and high expressions of LIMD1 and CDC25A may be considered as the molecular signature of the proteins in normal basal/parabasal epithelium. Transmission of the basal-like signature RBSP3 occurred during tumorigenesis, but the

Expression of RBSP3 and LIMD1 in normal head and neck epithelium and during tumorigenesis was due to inverse promoter methylation
Promoter methylation was performed by MSRA, followed by validation using MSP (Fig 2C  and 2D). Promoter methylation frequency of RBSP3 was significantly higher (72.5%, 29/ 40cases) in basal/parabasal layers of normal epithelium compared to spinous layer (17.5%, 7/ 40 cases) (Fig 2E i). Overall methylation frequency in dysplasia was 60%, with comparable frequencies in the mild/ moderate dysplastic lesions (71%) and severe dysplastic lesions (57%) (Fig 2E i). However, early invasive lesions (Stage I) showed low frequency of promoter methylation (31%) with gradual increase with progressing stages II to IV (35-47%) (Fig 2E i).
Significant concordance existed between MSRA and MSP, as determined by Fischer's exact test (S2 Table). However, similar to previous observations, Groups 1-4 showed no significant difference in promoter methylation of genes during tumorigenesis, annulling effect of etiological factors (Fig 2F).
Thus, the expression signature of RBSP3 and LIMD1 in different layers of normal oral epithelium was maintained through inverse promoter methylation. RBSP3 transmitted its basallike promoter methylation signature during tumorigenesis, but LIMD1 gradually lost its expression during tumorigenesis through subsequent promoter methylation.

Deletion of the genes increases from dysplastic lesions to HNSCC
Deletion was performed by microsatellite-based deletion mapping (Fig 4A). Deletion of RBSP3 in dysplastic lesions was high (35%) with comparable frequency in early invasive lesions (stage I/ II) (30-36%), followed by increase in later stages (stage III/ IV) (42-47%) (Fig 4B i).
Intriguing, although frequency of deletion of the genes was similar in Groups 1, 3 and 4, Group 2 (TOB-HPV+) presented significantly lower deletion compared to the other three groups, indicating that presence of HPV and absence of tobacco might play a pivotal role in preventing genomic instability (Fig 4E).

Mutation of the genes might not play an important role in HNSCC
No mutations were observed in the different protein domains in samples screened (data not shown). Moreover, encompassing SNPs presented major alleles in all cases. Therefore, mutation may be an infrequent event in the exons screened.

Etiological factors in conjunction to genetic/ epigenetic alterations were necessary for the development of HNSCC
RBSP3 showed high overall alterations (deletion/ methylation) in dysplasia (55%), which although lower in stage I of HNSCC (38%), gradually increased in successive stages (55-67%) (Fig 5A i). In contrast, overall alterations of LIMD1 progressively increased from dysplasia (60%) to enhancing stages of the disease (69-79%) (Fig 5A ii) with unremarkable differences Transmission/alteration of normal molecular signature of RBSP3, LIMD1 & CDC25A during oral carcinogenesis between etiological groups for both genes (data not shown). There was significant concordance between expression and alterations of the genes during tumorigenesis (Tables 2 and 3).

Etiological (HPV, Tobacco) and demographical factors significantly affected survival of patients with HNSCC
Alterations of RBSP3, LIMD1 and CDC25A were significantly (p = 0.001, <0.001, 0.007 respectively) correlated to overall survival of HNSCC patients by univariate analysis (Fig 5D). Multivariate analysis indicated significant association with HPV negativity, nodal status and stage as important prognostic factors along with alterations of the genes (Fig 5D). Interestingly, analysis of patient prognosis showed that survival of Group 2 patients was superior compared to Groups 1, 3 or 4, irrespective of the presence or absence of alterations (Fig 5E i). Similarly, in patients with alterations of RBSP3 and LIMD1, Group 2 patients showed significantly better survival than those belonging to other groups, indicating yet again the role of etiological factors in influencing patient prognosis (Fig 5E ii-iv).

Discussion
Several studies indicate the importance of stem cells in oral basal/parabasal epithelium for maintaining tissue integrity [3]. Our study thus aims to understand the molecular signature of the cell cycle regulatory proteins RBSP3, LIMD1 and CDC25A in normal basal/parabasal epithelium of H&N followed by analysis of the mode of transmission or alteration of the signature during tumorigenesis.
The molecular signature of the proteins in the normal basal/parabasal epithelium was low nuclear/cytoplasmic expression of RBSP3 and high nuclear with high/ moderate cytoplasmic expressions of LIMD1 and CDC25A.Etiological factors (HPV, tobacco) did not influence the above molecular signature, with the results being comparable to previous reports in H&N epithelium [8,9] and other tissues [18,19]. RBSP3, acts as a tumour suppressor phosphatase, that is important for regulating RNA polymerase II transcriptional machinery, attenuating BMP signaling and dephosphorylating pRB to block G1/ S cell cycle transition [20,21]. RBSP3 shows alteration in different cancers including uveal melanoma, lung cancer, cervical cancer, etc. [20,22,23]. Differential expression of RBSP3 as observed by us, might be necessary for cellular proliferation and differentiation. LIMD1 is a multi-functional protein involved in stabilization of pRB-E2F interaction, miRNA silencing, Hippo signaling, etc. [24,25] and is highly altered in different cancers [26,27,28]. Similarly, CDC25A phosphatase is involved in cellular processes such as response to DNA damage, mitotic entry [29,30] and is also important for tumour growth [31,32,33]. In line with their functions and our previous studies [8,9], the maintenance/ loss of expression signature of the genes during head and neck tumorigenesis indicates their functional importance in this malignancy.
Promoter methylation played an important role in governing the molecular signature of RBSP3 and LIMD1 in normal oral epithelium, as observed by its inverse relationship to protein expression. To the best of our knowledge, reports indicating the reason behind varying expression of the genes in different layers of normal oral epithelium are unavailable. Moreover, the technique of using LCM for separately analyzing promoter methylation in different layers of normal oral epithelium has not been reported, although it was used for RNA expression or epigenetic alterations in other organs [13,34]. Interestingly, no promoter methylation of the genes was observed in the composite normal epithelium, although its low frequency was previously reported [35]. The observed anomaly might be due to the high sensitivity of LCM for separating different layers of normal epithelium, resulting in enrichment of the otherwise sparsely populated basal/ parabasal cells. This might be the reason why methylation in these layers, even in minor frequencies were detected in our study, which was unreported in previous studies where DNA from the entire epithelium was subjected to promoter methylation.
The normal basal-layer like signature of low expression of RBSP3 was transmitted during development of tumours through promoter methylation in addition to deletion. It appears that development of dysplastic lesions required some selection pressure for maintaining low expression of RBSP3 by additional deletion alongside methylation. However, the basal-like signature of LIMD1 and CDC25A was not transmitted during tumour development. Subsequent and significant promoter methylation (for LIMD1 only) and deletion were necessary for their loss of the expression. Separately assaying alterations in the dysplastic/ non-dysplastic      L  +  --+  +  H  H  L  --+  +  +  H  H  M  ---+   1445  M  H  L  ---+  +  H  M  L  --+  +  +  H  M  H  ----4663  L  H  M  +  ----L  L  L  +  -+  +  +  H  M  M  ----2287  M  H  H  +  ----L  L  L  +  +  +  -+  H  H  M  ----1898  L  H  L zones of dysplastic epithelium by using LCM is a novel approach, although the results were comparable with previous reports using composite dysplastic tissues of H&N and cervix [6,7,16,18,35,36]. This could possibly be due to the high frequency of associated alterations. Moreover, till date, no studies have reported the role alterations of these genes in the same set of H&N samples. Interestingly, analysis of the role of etiological factors in influencing genetic deletion indicated that deletion of the genes in tobacco-HPV+ patients (Group 2) was significantly lower than Groups 1/3/4, possibly due to some effect induced by HPV infection. Finally, assessing alterations with respect to etiological factors indicated that tobacco induced significant alterations of the candidate genes in HNSCC, whereas in the case of LIMD1, HPV infection played a positive role in preventing alterations. Patients in Group-2 (HPV+TOB-) had further reduced alterations with better 5-year disease free survival, indicating the possible preservative role of HPV in preventing genomic instability and providing a better prognosis, as previously reported by us and other researchers [17,[37][38][39]. Another alternative might be the evolution of an entirely different subgroup of HPV-infected H&N tumors, which resulted in different frequencies of alterations and prognosis compared to the other groups [40,41], although further studies are warranted for validating the above.
In conclusion, transmission of normal-like signature could help in categorizing RBSP3 as a separate class of TSG, the maintenance TSG, similar to SnoN [42], which alteration of the basal-like signature categorizes LIMD1 and CDC25A in the classical group of TSGs like WWOX, Smad4, PTEN, etc. [43][44][45].Etiological factors, especially HPV appears important in H&N tumorigenesis and for providing better patient outcome. Thus, our data also suggests the  NOR  TUM  NOR  TUM  ALL  NOR  TUM  NOR  TUM  ALL  NOR  TUM  NOR  Transmission/alteration of normal molecular signature of RBSP3, LIMD1 & CDC25A during oral carcinogenesis importance ofLIMD1 and CDC25A in conjunction with HPV for use as diagnostic and prognostic markers of HNSCC, whereas RBSP3 as a prognostic marker only.
Supporting information S1