Over the last decades there has been little improvement in survival statistics for oral cancers, mainly due to the late stage of diagnosis.1, 2 The disease has a high rate of recurrence (up to 20–30%) and of development of secondary cancers.3, 4, 5, 6 This has led to an emphasis on genomic approaches for the identification of genetic alteration in oral premalignant lesions (OPLs), with the goal of developing novel targets for early diagnosis and treatment of early-stage disease.
Oral cancer is believed to progress through a series of histopathologic stages from hyperplasia to dysplasia of varying degrees (mild, moderate, severe) and to carcinoma in situ (CIS) prior to the development of invasive squamous cell carcinoma (SCC). Associated with this histologic progression is the gradual accumulation of critical genetic alterations. A number of such alterations have been identified in oral cancer through comparative genomic hybridization (CGH) and loss of heterozygosity (LOH) analysis.7, 8, 9, 10, 11 However, few of these alterations have been fine mapped to pinpoint a candidate gene, and only a few genes have been associated with histopathologic staging (p16,p53,FHIT,cyclin D1).12, 13, 14, 15, 16, 17, 18
Alterations involving 13q have been previously reported in oral and head and neck cancers and in cancers at other sites, including liver, breast, ovary, larynx, lung, bladder, prostate, and the lymphoreticular system.8, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 The LOH studies suggest at least 3 separate regions of alteration: one centered around Rb and candidate genes associated with CLL and the others centromeric (e.g., around BRCA2) or telomeric to Rb.32, 33, 34, 35
In our study we describe the identification of a novel region of alteration that is frequently observed in both OPLs and tumors. The minimal region contained 3 candidate genes, one of which, AKAP220, showed altered expression in tumors and may play an important role in driving the process of oral carcinogenesis.
AI, allelic imbalance; AKAP220, A-Kinase Anchoring Protein 220; BAC, bacterial artificial chromosome; CGH, comparative genomic hybridization; CIS, carcinoma in situ; DGKH, Diacylglycerol Kinase; LOH, loss of heterozygosity; MRA, minimal region of alteration; OPL, oral premalignant lesion; RANKL, Receptor Activator of NK-kappa-B Ligand; RAPD, randomly amplified polymorphic DNA; SCC, squamous cell carcinoma
Material and methods
Oral dysplasia and tumor specimens for microsatellite and randomly amplified polymorphic DNA (RAPD)-PCR analysis
Formalin-fixed paraffin-embedded samples were retrieved from the archive of the Oral Biopsy Service of British Columbia, and histologic diagnoses were confirmed by 2 oral pathologists. Serial sections from each biopsy were stained with hematoxylin and eosin and microdissected. Dysplastic or malignant cells were digested with SDS/proteinase K at 55°C, and DNA was extracted with a phenol-chloroform mixture as previously described.36 DNA was quantified fluorometrically using Picogreen (Molecular Probes, Eugene, OR). Connective tissue from each specimen was employed as a source of normal DNA.
High-density RAPD-PCR analysis
RAPD-PCR fingerprints were generated from pairs of dysplasia (or tumor) and normal DNA from each patient. A positive control (human genomic DNA) and a negative control (no DNA template) were used for each experiment. Ten picomoles of each primer (5′-GACGCCGCTT-3′ and 5′-CGCCTAATGC-3′) were radiolabeled with 2 μCi of [γ32P]-ATP (6000 Ci/mmol; Amersham Pharmacia Biotech, Quebec, Quebec) using T4 polynucleotide kinase (MBI Fermentas, Burlington, ON) at 37°C for 1 hr, and inactivated at 95°C for 5 min. RAPD-PCRs were performed in a 10 μl volume containing 2 ng of template DNA, 2.5 units of recombinant Taq DNA polymerase, 200 μM of each dNTP, 10 mM TRIS-Cl, pH 8.3, 50 mM KCl, 2 mM MgCl2, and 0.001% gelatin. All reactions were performed in a PTC-100 thermal cycler (MJ Research, Waltham, MA) for 45 cycles of 94°C for 1 min, 35°C for 1 min, and 72°C for 2 min. The PCR products from sets of 20 paired normal and tumor samples were resolved on 8% nondenaturing acrylamide gels (38 acrylamide: 1 N′,N′-methylene-bis-acrylamide) in 100 mM TRIS-borate-EDTA buffer, pH 8, for 3 hr at 800 V. Dried gels were imaged by autoradiography. Gain or loss of signals was identified by pairwise comparison of RAPD-PCR fingerprints. Signal alterations that occurred in multiple tumors were chosen for further analysis.
Cloning and mapping
DNA from excised bands was amplified in 20 μl reactions using 0.25 pmol of each primer (5′-ACAAGCTTCTGCAGGACGCCGCTT-3′, 5′-ACGAATTCGGATCCCGCC TAATGC-3′), Taq DNA polymerase, and buffer (Promega, Madison, WI). Amplification was performed as follows: 2 cycles (94°C, 1 min; 35°C, 1 min; 72°C, 2 min) and then 30 cycles (94°C, 40 sec; 62°C, 40 sec; 72°C, 40 sec). The amplified DNA was digested with BamHI and Pst1, gel purified and ligated to pBlueScript SK vector, and transformed into ElectroMAX DH10B Cells (Invitrogen, Burlington, ON) by electroporation. Multiple clones from each band were sequenced to confirm that the alteration was the same in different cases.
Verification and fine mapping by microsatellite analysis
The sequence of one of the recurrent changes was then localized to bacterial artificial chromosome (BAC) RP11-84N7 (Genbank accession AL139328) by searching Genbank (http://www.ncbi.nlm.nih.gov/Genbank) and the public sequence database at http://genome.cse.ucsc.edu/index.html. Markers in the region close to the RAPD alteration were determined using map information provided by the Ensembl system (http://www.ensembl.org) and University of California Santa Cruz (UCSC) (http://genome.cse.ucsc.edu/index.html). Markers with a high degree of heterozygosity were chosen (Supplemental Table I). The protocol used for microsatellite analysis is described in Zhang et al.36 After PCR amplification, PCR products were separated on denaturing polyacrylamide gels and visualized by autoradiography. For informative cases, allelic imbalance (AI) was inferred when the signal ratio of the 2 tumor alleles differed from the ratio of the normal samples by at least 50%. Samples showing AI were subjected to repeat analysis after a second independent amplification whenever the quantity of DNA was sufficient.
Table I. Primers Used in Microsatellite Analysis and Their Chromosomal Location
BAC contig assembly and candidate gene identification
Fingerprints of ∼400,000 human BACs are currently available at the Washington University Human Genome Project (December 22 freeze; http://genome.wustl.edu.gsc). Using finger printed contigs (FPC) software (http://www.sanger.ac.uk/HGP/), contig maps were constructed for the 13q14 region. Relative positions of all microsatellite markers were verified by BLAST alignment against corresponding BACs. Relative positions of BACs, sequenced tag sites (STS), and known genes were established using the UCSC Human Genome Browser (July 2003 version at http://genome.ucsc.edu). A composite diagram showing the relative position of all markers, BACs, and contigs identified in this region are provided.
Semiquantitative analysis of AKAP220, RANKL, and DGKH expression in oral tissue
Frozen epithelial tissue from the oral mucosa was microdissected from 11 samples from people without cancer or dysplasia, as well as 16 tumors, and placed in denaturing solution (4 M guanidinium isothiocyanate, 0.02 M sodium citrate, 0.5% sarkosyl, 3.45 μl of β-mercaptoethanol). RNA was extracted using the protocol of Chomczynski and Sacchi.37 Extracted RNA was DNase treated prior to cDNA synthesis.
cDNA was synthesized using the Superscript II RNAse H reverse transcriptase system (Invitrogen). Gene-specific primers were used to assay the quantity of RANKL (5′-TCGGCACTTGTGGAAAAACA-3′, 5′-TGGCCACCAGGTGCCTTTCA-3′), AKAP220 (5′-TTGGTTACTTGTGCAGTGTT-3′, 5′-ACTGTTACACAGTACAAGATCT-3′), and DGKH (5′-GGTCTGATGGAATGTTCCT-3′, 5′-TAGGAGTTGGAGGTGATACC-3′) using GAPDH levels (5′-ACCTACCAAATATGATGACATCA-3′, 5′-CGCTGTTGAAGTCAGAGGA-3′) as controls. PCR cycle conditions were as follows: 35 cycles (95°C, 30 sec; 55°C, 30 sec; 72°C, 1 min) and a 10 min extension at 72°C. The same results were obtained with 30 cycles (data not shown). PCR products were resolved by polyacrylamide gel electrophoresis (6% gels), stained with SYBR green (Roche Diagnostics, Mississauga, Canada), imaged on a Storm phosphoimager (Molecular Dynamics, Sunnyvale, CA), and quantified using ImageQuant software (Amersham Biosciences, Piscataway, NJ).
Results and discussion
DNA fingerprint comparison of normal and tumor oral epithelium identifies a ∼400 bp genetic alteration
Tumor epithelial and underlying stromal cell controls were microdissected from 40 SCC cases and DNA was extracted from each sample and subjected to RAPD DNA fingerprinting. A novel ∼400 bp alteration, absent in all 40 control samples, was observed in 4 of these tumors (gained in 385T, 543T, 569T2, 577T) (Fig. 1a). This fragment was cloned from all 4 cases and shown to be identical in sequence. The sequence is available as Genbank accession AY870328.
Chromosomal localization and verification of the novel alteration
The sequence matched that of human BAC RP11-84N7 (AL139328) upon BLAST search against Genbank (http://www.ncbi.nlm.nih.gov/BLAST/). This BAC was localized to 13q14 by fingerprint alignment against the FPC database (http://www.genome.wustl.edu/gsc/) and by sequence alignment against the July 2003 release of the Human Genome Project Working Draft (http://genome.cse.ucsc.edu). The RAPD PCR alteration was confirmed to be at 13q14.11 by microsatellite analysis of the cases in which the alteration was initially identified. A marker, D13S1297, located 400 Kbp telomeric to the alteration was used to verify AI in 543T, 569T, 577T, and 385T. Three of the 4 samples showed AI, with the 4th (385T) being noninformative. Figure 1b shows 2 examples.
Construction of a contig map of BAC clones and definition of the minimal region of alteration (MRA)
To facilitate fine mapping and to demonstrate the relative position of the microsatellite markers used in our study, we used FPC software to construct a BAC contig of this region containing RP11-84N07, which harbors the RAPD sequence (see Material and Methods). Figure 2a shows a 2 Mbp portion of this tiling path that later proved to contain the MRA (described below). The relative positions of microsatellite markers and genes at 13q14 are shown in Figure 2b. Marker positions were verified by BLAST against BAC clones in this tiling path.
To establish the MRA, we first examined a panel of 39 oral SCC cases for AI at D13S1297. Seventeen of the 31 informative cases showed alteration (a high frequency of 54%). We then evaluated D13S263, a neighboring microsatellite marker 1.6 Mbp centromeric to D13S1297. Thirteen of 33 informative cases showed AI (39%) (Fig. 3a, top panel). For 6 of these samples (386T, 469T, 539T, 542T, 569T, 577T), the AI at D13S1297 was flanked by retention at D13S263, suggesting the presence of a centromeric boundary at that locus. Further analysis with an additional marker, D13S1227 (0.3 Mbp telomeric to D13S1297), revealed the other boundary in 3 cases (386T, 539T, 620T). Figure 3b shows 2 examples. These data support a 1.9 Mbp MRA spanning D13S263 and D13S1227 and including BAC clones RP11-57L14, RP11-322A14, RP11-84N7, RP11-215B13, RP11-86N24, and RP11-145J3 (Fig. 2a). This MRA is also observed in oral dysplasia and is discussed below.
This MRA is distinct from regions previously described in the literature for head and neck SCC and excludes Rb, Leu, and BRCA2 (Fig. 2b and c).21, 22, 23 Yoo et al.21 and Gupta et al.23 reported an MRA that included Rb and extended telomerically. This region was further fine mapped to exclude Rb by Ogawara et al.20 Maestro et al.22 described a region that includes the Rb locus but extends centromerically. The location of the 1.9 Mbp MRA defined in this work to the aforementioned studies is indicated in Figure 2c. As shown in this figure, BRCA2 is 9 Mbp centromeric, the Leu locus is ∼6 Mbp telomeric, and Rb is 5 Mbp telomeric to the MRA.
13q alteration in histologic progression
In order to determine if AI was present in premalignant oral lesions, we used the markers that defined the MRA to assess 51 dysplastic lesions. Seventeen of 43 informative cases (40%) showed AI at D13S1297. In 6 of these cases (292D, 332D, 535D, 567D, 585D, 612D), the AI at D13S1297 was associated with a retention at D13S263, again supporting the centromeric boundary (Fig. 3a, bottom panel). Five cases (292D, 332D, 535D, 567D, 612D) showed AI at D13S1297 and retention at D13S1227 supporting the telomeric boundary.
We further determined whether there was any association of AI with histopathologic progression of dysplasia from low-grade dysplasia (mild or moderate, n = 42) to high-grade dysplasia (severe dysplasia or CIS, n = 16) to invasive SCC (n = 36). AI in this region was present in 9 of 32 informative low-grade cases (28%), 9 of 14 high-grade dysplasia (64%), and 22 of 36 SCCs (61%) (p = 0.011). This suggests that although this alteration is present in some low-grade lesions, the frequency increases significantly in high-grade premalignancy before the lesion becomes invasive (Fig. 4).
We also compared the frequency of alteration at 13q14.11 to AI at 2 other regions, 3p14 (containing FHIT) and 9p21 (containing p16), using the microsatellite markers shown in Supplemental Table I. Alterations in these regions are early events, occurring frequently in low-grade dysplasia.14, 25, 38 As shown in Figure 4, there is a higher frequency of AI at these regions in low-grade dysplasia than AI at 13q14.11, suggesting that AIs at 3p14 and 9p21 occur before 13q14.11. This is consistent with the fact that AI is observed at 3p14 and/or 9p21 in the absence of 13q14.11.
We also analyzed AI at 13q32 using the marker D13S170, 40 Mbp telomeric to D13S1297, because this region has also been reported to show AI in oral SCC.14 AI occurred in only 10% (3 of 29 cases) of low-grade dysplasia, with this frequency increasing to 27% (3 of 11 cases) of high-grade dysplasia and 36% (14 of 38) of SCCs (Fig. 4). AI occurred in only 1 of the 2 loci (D13S170 or D13S1297) in 19 of 61 cases that are informative for both loci, supporting their occurrence as separate events.
Identification of candidate genes
Within the 1.9 Mbp minimal region there are 3 known genes, Diacylglycerol Kinase (DGKH), A-Kinase Anchoring Protein 220 (AKAP220), and Receptor Activator of NK-kappa-B Ligand (RANKL) (also known as Osteoclast Differentiation Factor (ODF),Osteoprotegerin Ligand (OPGL), or Tumor Necrosis Factor Ligand Superfamily Member 11 (TNFSF11) or TNF-Related Activation-Induced Cytokine (TRANCE)). We verified the map position of these genes by sequence alignment with BACs as shown in Figure 2a. DGKH maps to RP11- 4K24, AKAP220 maps to RP11-215B13, and RANKL maps to RP11-86N24.
Tissue expression profiles of the candidate genes in normal oral mucosa by semiquantitative RT-PCR
To determine whether the 3 candidate genes are expressed in epithelial cells of normal oral mucosa, the epithelial layer was microdissected from 11 tissue samples from individuals without cancer or dysplasia. RNA was extracted, cDNA produced, and expression measured using semiquantitative RT-PCR. Since there are multiple splice forms for these 2 genes, primers were designed to recognize an exon common to all splice forms. GAPDH expression levels were used to normalize between samples. In all 11 of the normal samples we detected expression of both RANKL and AKAP220. RANKL exhibited a generally higher overall expression in normal mucosa than AKAP220 (Fig. 5). DGKH expression was not detected in normal oral tissue.
Assessment of DGKH, RANKL, and AKAP220 expression in oral tumors by semiquantitative RT-PCR
RNA was extracted from 16 microdissected oral SCCs. There was no elevation in DGKH or RANKL expression in tumors compared to normals (none of the tumors tested for RANKL expression produced a signal >1 SD from the normal average) (Fig. 6a). In contrast, 12 of 16 tumors showed AKAP220 expression levels that were >1 SD from the normal average. Of these 12, 6 expressed AKAP220 levels that were 2–5 times greater than the normal average (Fig. 6b). Figure 6c shows expression levels for AKAP220 for 10 examples.
In this report we have identified a novel genetic alteration in oral cancers that lies within a 1.9 Mbp MRA that contained 3 known genes (Fig. 2). One of the genes, AKAP220, showed altered expression in oral cancers. To our knowledge this is the first report of AKAP220 alteration in tumorigenesis, although the involvement of this AKAP protein has been implicated in cell cycle regulation.
AKAP220 acts as a competitve inhibitor of type 1 protein phophatase (PP1) activity, with this inhibition enhanced by the presence of the RII regulatory unit of protein kinase A (PKA).39, 40 PP1 catalyzes the dephosphorylation of the RB protein, thus acting as a negative regulator of cell cycle progression.41, 42 Since AKAP220 inhibits this activity, its overexpression (as observed in Fig. 6b) could lead to hyperphosphorylation of pRb, release of E2F, and the transcription of genes involved in cell cycle and lead to proliferation (Fig. 7).
Other members of this pathway have been studied in oral and head and neck cancers. Both Cyclin D1 and MTS1/p16/INK4 are frequently dysregulated by multiple mechanisms.13, 16, 18 In contrast, Rb is seldom mutated.19, 21 The involvement of AKAP220 may be yet another mechanism of disrupting this critical pathway. It should be noted that like the p16 region (9p21), the AKAP220 region on 13q14.11 is altered early in development of the oral cancer. It is present in low-grade dysplasia and increases significantly with progression to high-grade dysplasia (Fig. 4). These data strongly support the need for further studies to assess the impact of AKAP220 alterations in early cancers.
The authors thank Ms. P. Calderon, Mr. L. Xie, and Mr. R. Li for their technical assistance and Mr. T. Buys for the careful reading of this article. Tissue DNA is archived using funds from grant 1 R01 DE13124-01 from NIDCR.