TWO MEMBERS OF THE SIBLING FAMILY OF PROTEINS, DSPP AND BSP, MAY PREDICT THE TRANSITION OF ORAL EPITHELIAL DYSPLASIA TO ORAL SQUAMOUS CELL CARCINOMA

Abstract

Background

Patients with oral premalignant lesions (OPL) present with oral squamous cell carcinomas (OSCC) at a much higher rate than the general population. There are currently no useful markers that indicate specifically which OPLs are most likely to progress. Three SIBLING family proteins, bone sialoprotein (BSP), osteopontin (OPN), and dentin sialophosphoprotein (DSPP), have been shown to be up-regulated in many cancers, including OSCC. The status of SIBLING expression in OPLs and their correlation to transition to oral cancer are unknown.

Methods

Sixty archival surgical biopsies of dysplastic OPLs were evaluated by immunohistochemistry for expression of BSP, DSPP and OPN and correlated with local transformation to OSCC at sites adjacent to surgically removed dysplastic OPL

Results

The OPL patient population was representative of previous studies with 20% progressing to OSCC, and no correlation between degree of dysplasia and progression. 87% were positive for at least one SIBLING protein. OPN expression had no correlation with progression. The BSP+/DSPP− expression pattern however correlated with decreased transformation to OSCC (Point Prevalence=0%, 95%CI=0–20.6%) while the BSP−/DSPP+ pattern was associated with more frequent progression (Point Prevalence=77.8%, 95%CI=47.8–95.4%). Incrementally higher expression scores (0 to 3+) of BSP and DSPP were also associated with increased predictive values (OR=25.53, 95%CI=2.14–304.7 and 10.13, 95%CI=2.0–50.0, respectively, for each increment).

Conclusions

BSP and DSPP are excellent candidate markers for successful OPL surgical intervention and may be predictors of OPL-OSCC progression.

Introduction

Oral cancer is the sixth most common cancer in the world and there is a possible rise within developing countries.¹ In the United States, ~35,000 new cases of oral and pharyngeal cancers are diagnosed annually with ~7,500 reported deaths.¹ The mortality associated with recurrent disease is >50%.¹ Over 95% of oral cancers are classified histologically as oral squamous cell carcinoma (OSCC).^{2, 3} Many OSCC are thought to be associated with, and perhaps derived from, oral premalignant lesions (OPLs) such as clinical leukoplakia, erythroplakia, and speckled leukoplakia.^{4, 5} Leukoplakia, by far the most abundant OPL, is reported to undergo malignant transformation from <1% to 17.5% of the time.⁶ The much rarer erythroplakia presents about half of the time with invasive carcinoma at diagnosis.^{7, 8} The rate of transformation from dysplastic oral epithelium to OSCC is reported to range from 5–36%.^9–14 Based in part on the analogous field of cervical precancerous lesions, many clinicians currently rely on the degree of dysplasia present in OPLs as an indication of their tendency to progress to cancer. While some studies have reported a correlation between more severe dysplasia with increased rate of transformation into cancer, other studies have not found such a clear correlation.^14–16 It is also important to note that not all dysplastic lesions progress to cancer and that some regress or even disappear.^{9, 14–18}

Unlike OSCC, where surgery is the standard and most important treatment option², treatment modalities for OPLs are less standardized and less definitive. The current treatments include: watchful observation; conservative surgery for lesions adjudged clinically and histologically as low risk; radical surgical excision for lesions adjudged high risk; and even radical surgical excision of all OPLs regardless of the clinical and/or histologic estimate of their malignant potentials.^{2, 19} Given the morbidity associated with oral surgery, there is a need to identify simple markers that estimate the transition potential of each OPL to OSCC.

Although a number of markers have been cited as predictive of the transition potential of OPL to OSCC, their usefulness is often hampered by their limited practical clinical applications. Recently for example, Guillaud et al.²⁰ reported a nuclear phenotype score (NPS) comprised of five nuclear morphometric features that “best” discriminate normal nuclei from abnormal nuclei and concluded that within mild/moderate dysplastic OPLs, elevated NPS was strongly associated with high-risk loss of heterozygosity (LOH).²⁰ However, another study concluded that a significant number of dysplatic lesions exhibiting LOH do not progress to OSCC.²¹

Over the past decade the up-regulation of various SIBLING (Small Integrin-Binding LIgand N-Linked Glycoproteins) proteins in a number of cancers has been reported.²² Members of the family include osteopontin (OPN), bone sialoprotein (BSP), dentin matrix protein-1 (DMP1), dentin sialophosphoprotein (DSPP), and matrix extracellular phosphoglycoprotein (MEPE).²³ OPN, BSP, and DSPP have been linked with specific stages/markers of tumor progression including cell growth, adhesion, migration, and/or metastasis.²² Our recent report showed that while all five SIBLINGs are absent in normal oral mucosa, BSP, DSPP, and OPN were variously up-regulated in OSCC. ²⁴ Furthermore, DSPP expression was highly upregulated in both well and poorly differentiated as well as histologically aggressive OSCC while BSP was found associated only with well-differentiated tumors.²⁴

In this retrospective pilot study, we screened for expression of SIBLINGs in surgical biopsies of OPLs from 60 patients where the recorded clinical diagnosis indicated various OPLs, including “leukoplakia” with histopathologic diagnoses of mild, moderate, or severe oral epithelial dysplasia. The purpose was to determine the correlation between expression of any SIBLING or combination of SIBLINGs in the surgically removed dysplastic OPLs and either increased or decreased appearance of OSCC at the primary surgical site.

Material and Methods

Case Selection

Institutional Review Board approvals from the Medical College of Georgia (MCG) and the National Institutes of Health were obtained. A total of 60 cases of surgical biopsies of dysplastic OPLs were randomly selected from a large source based on the following criteria:

Inclusion criteria

1. A medical record indicating an initial clinical diagnosis of OPL.

2. Biopsy (incisional or excisional) carried out at time of initial clinical diagnosis.

3. A histologic diagnosis of dysplasia/carcinoma-in-situ on hematoxylin/eosin sections independently verified by two oral pathologists.

Exclusion criteria

Stipulation on Pathology Request Form of:

1. Presence of neoplasm of other anatomic sites/systems at the time of diagnosis of OPL.

2. Previous diagnosis of and treatment for OSCC.

3. Prior history, diagnosis, or treatment for cancers of other organs/regions.

The cases satisfying the inclusion criteria were randomly selected from accrued oral histopathology diagnostic database of MCG using the search words: mild; moderate; or severe dysplasia. Both board-certified pathologists were blinded to the clinicopathologic details of each case including whether patients presented at later dates with OSCC at previous site of dysplastic OPL. Because of the non-verifiablity of the original clinical diagnosis (leukoplakia etc.), selected cases were stratified only according to the histological degree of dysplasia. Qualifying cases of carcinoma-in-situ were grouped with the severe dysplasias. Cases were randomly picked and scored by both pathologists until there were 20 cases for each of the three categories of dysplasia.

MCG is the major referral hospital in the geographic area and the racial/ethnic/gender make-up of selected cases reflected the population of the Georgia/Central Savannah River area from which the cases were sent. Selected cases also reflected published oral site distribution within the larger affected population.^1,2 Because the majority (over 80%) of the patients were referrals from clinicians in private practice (who as a group are unlikely to perform such advanced surgical procedures), the sample selection is not known to be biased towards more severely affected patients. The absence of subsequent pathology request records within 5 years of initial dysplastic OPL diagnosis and surgical excision was interpreted to mean that such patients did not develop OSCC. Therefore, this retrospective study, conducted under an IRB approval not permitting direct contact with the patients, may somewhat underestimate the occurrence of cancer within this cohort.

Histopathologic Analysis

The presence of dysplasia was determined by a combination of established architectural and cytologic parameters including: nuclear pleomorphism; increased and/or abnormal mitosis; hyperchromatism; basal cell hyperplasia; and abnormal maturation sequence.²⁵ The degree of dysplasia was determined by the thickness of the epithelium from the basement membrane-basal cell layer interface to the keratinized surface layer of the mucosa exhibiting the hallmarks of epithelial dysplasia. Oral epithelium exhibiting normal maturation sequence is represented in Figure 1A. Mild dysplasia affects an estimated one-third of the mucosal epithelial thickness (Figure 1B), moderate dysplasia involves ~two-thirds of the epithelial thickness (Figure 1C), and severe dysplasia/carcinoma-in-situ was the diagnosis for full epithelial thickness dysplasia (Figure 1D).²⁵

Figure 1.

Degree of epithelial dysplasia and SIBLING immunoreactivity. Oral epithelium exhibiting normal maturation sequence (A) without dysplasia. Mild dysplasia shown as dysplasia affecting one-third of the epithelial thickness (B; arrow), moderate dysplasia involves estimated two-third of the epithelial thickness (C; arrow), and severe dysplasia/carcinoma-in-situ involving full epithelial thickness (D; arrow). A through D represents standard hematoxylin and eosin (H&E) staining. Example of DSPP expression scored as 0 (0 to 10% of dysplastic cells with no more than weak immunoreactivity; E) with DSPP monoclonal antibody, LFMb-21. Note the cytoplasmic staining of epithelial dysplastic cells for DSPP (E, G). OPN expression scored as 1 with less than 50% dysplatic cells exhibiting immunoreactivity (F) with OPN monoclonal antibody, LFMb-14. Immunoreactivity is cytoplasmic with occasional reactivity in immune cells and connective tissue (K; arrow). DSPP expression scored as 2 with more than 50% but less than 75% dysplastic cells exhibiting immunoreactivity (G) with DSPP monoclonal antibody, LFMb-21. BSP expression scored as 3 with more than 75% of dysplastic cells being positive with LFMb-2 5 (H). BSP expression is confined to the cytoplasm and perinuclear area of immunoreactive cells (H, I). Representative pre-immune control (L) showed expression scored as 0 for no immunoreactivity. Chromogenic staining (reddish-brown color) was achieved with 3-amino-9-ethylcarbazole (AEC) and counterstained with hematoxylin. All Scale Bar (red): 50µm

Immunohistochemistry

Paraffin block sections (5µm) were immunostained for all five SIBLINGs using the Zymed ST5050 automated system supplied with Super-Picture-Perfect Broad-Spectrum HRP-Polymer and Single-Solution-AEC reagents from Zymed Lab Inc. (San Francisco, CA), and antibodies as previously described (Table 1).^26–28 The SIBLING monoclonal antibodies are available from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Corresponding rabbit polyclonal antisera were used to validate each monoclonal antibody result.^26–28 Negative controls included the substitution of primary antibody with non-immune rabbit serum or mouse IgG control (Zymed, Ibid). Representative results for each case were photographed using the Axioplan 2 Universal microscope equipped with an Axiovision camera and program (Carl Zeiss Gmbh, Jena, Germany).

Table 1.

Antibodies used in this study

Antigen	ID	Concentration	Type	Reference/Source
Human BSP	LFMb-25	10µg/ml	Mouse IgG1	Ogbureke, Fisher 200528
Human DSPP	LFMb-21	2.5µg/ml	Mouse IgG2b	Ogbureke, Fisher 200528
Human DMP1	LFMb-31	5.0µg/ml	Mouse IgG2b	Ogbureke, Fisher 200528
Human MEPE	LFMb-33	1:100	Mouse IgG	Ogbureke, Fisher 200726
Human OPN	LFMb-14	1:100	Mouse IgG	Ogbureke, Fisher 200726
Human BSP	LF-84	1:200	Rabbit Polyclonala	Ogbureke, Fisher 200427
Human DSPP	LF-151	1:200	Rabbit Polyclonal	Ogbureke, Fisher 200427
Human DMP1	LF-148	1:50	Rabbit Polyclonala	Ogbureke, Fisher 200726
Human MEPE	LF-155	1:200	Rabbit Polyclonala	Ogbureke, Fisher 200427
Mouse OPN	LF-175	1:200	Rabbit Polyclonal	Ogbureke, Fisher 200427

Abbreviations: BSP, bone sialoprotein; DSPP, dentin sialophosphoprotein; DMP1, dentin Matrix Protein 1; MEPE, matrix extracellular phosphoglycoprotein; OPN, osteopontin;

^aAffinity purified

Scoring of Immunohistochemistry Results

The area and intensity of immunostaining for each SIBLING were assigned a semi-quantitative score by the two oral pathologists. Scoring was: negative (0, immunoreactive product not detectable or faint staining in less than 10% of dysplastic cells); 1 (10–50% of immunoreactive dysplastic cells); 2 (50–75% of immunoreactive dysplastic cells); and 3 (widely and highly expressed in dysplastic cells).

Statistical Analysis

Odds ratios (OR) and 95% confidence intervals were calculated based on two logistic regression models. In Model I, the parameters of BSP, DSPP, OPN and Dysplasia were coded as binary parameters (0 vs. 1) and in Model II, the semi-quantitative staining parameters for BSP, DSPP, OPN were coded as 0, 1, 2, 3, and Dysplasia was coded as mild=0, moderate=1, and severe=2. Regression coefficients of Model II are based on per unit increase in scores. Percentages are provided with 95% Bayesian confidence intervals. All analyses were performed using SAS Statistical Software (Cary, NC).

Results

Immunolocalization of SIBLINGs in OPLs

As previously reported for frank OSCC²⁴, no significant immunostaining for MEPE or DMP1 was detected in any of the dysplastic lesions (data not shown). Immunostaining for various combinations of BSP, DSPP, and OPN, however, was noted in 52 of 60 (87%) cases (Table 2). All cases of SIBLING-positive dysplastic lesions in this study exhibited similar cytologic and architectural patterns of SIBLING localization regardless of progression status (i.e. progressed versus not-progressed). Representative scores of 0 and 2 are illustrated with DSPP expression in Figures 1E and 1G respectively. A representative score of 1 is shown in Figure 1F for OPN, while a score of 3 is shown for BSP (Figure 1H). Immunoreactivity for BSP was evident within the cytoplasm (which, as in all following results, presumably is within the secretory pathway organelles for these secreted matricellular proteins) and the perinuclear perimeters (presumably rough endoplasmic reticulum and/or Golgi) of dysplastic epithelial cells in the basal and spinous layers (1H & 1I). DSPP immunoreactivity was confined to the cytoplasm and perinuclear perimeter of dysplastic cells (1E & 1G) while OPN (1F, 1J, & 1K) was observed in both the cytoplasm of dysplastic cells as well as in reactive immune cells infiltrating tissue stroma. OPN has previously been reported to be expressed in immune cells.²⁹ Connective tissue stroma was negative except for occasional faint staining by BSP and DSPP antisera in vascular elements and fine collagen fibers. Figure 1L is representative of non--reactive, IgG controls.

Table 2.

Semiquantitative SIBLING Scores (IHC)^**

MILD DYSPLASIA				MODERATE DYSPLASIA				SEVERE DYSPLASIA
Case #  (L)	BSP	DSPP	OPN	Case #  (L)	BSP	DSPP	OPN	Case #  (L)	BSP	DSPP	OPN
1 (4)	2	3	0	21T (3)	1	1	0	41 (1)	3	3	0
2T (2)	0	0	2	22 (2)	1	0	0	42 (6)	2	3	1
3 (3)	2	0	0	23 (1)	0	0	0	43 (4)	1	1	0
4T (3)	0	1	0	24T (5)	0	3	2	44T (1)	0	0	0
5 (4)	1	0	0	25 (3)	0	0	0	45 (4)	0	0	3
6 (1)	0	0	1	26 (4)	0	0	1	46 (4)	0	0	2
7 (2)	1	0	0	27T (1)	1	2	3	47T (1)	0	1	0
8 (1)	0	0	1	28 (1)	1	2	0	48 (1)	0	0	1
9 (2)	3	0	1	29 (1)	2	0	0	49 (4)	2	0	2
10 (1)	1	0	0	30 (4)	1	0	0	50 (1)	0	0	2
11 (1)	0	0	0	31 (4)	0	0	0	51 (4)	3	0	1
12 (2)	0	0	2	32T (1)	0	0	3	52T (1)	0	2	1
13 (6)	0	0	0	33 (4)	0	0	1	53 (1)	2	1	2
14 (4)	0	0	2	34 (1)	0	0	2	54 (1)	3	3	0
15 (1)	2	1	1	35 (1)	3	2	0	55 (4)	2	0	3
16 (6)	0	0	0	36 (1)	0	0	1	56T (2)	0	3	3
17 (4)	0	0	0	37 (4)	3	1	0	57 (6)	3	3	0
18T (3)	0	2	0	38T (1)	2	2	0	58 (4)	3	1	0
19 (3)	3	0	1	39 (2)	0	0	1	59 (4)	0	1	1
20 (6)	0	1	1	40 (6)	0	1	2	60 (4)	1	1	3
Total positive	8	5	9		9	8	9		11	12	13

^**Scoring was as negative (0, not detectable), 1 (detectable but <50% of dysplastic cells), 2 (present in >50% but <75% of dysplastic cells), and 3 (widely and highly expressed).

T = Patient is diagnosed with local OSCC within 5 years of original diagnosis

Abbreviations: OPL, oral premalignant lesions; OSCC, oral squamous cell carcinoma; BSP, bone sialoprotein; DSPP, dentin sialophosphoprotein; OPN, osteopontin; (L)= location of OPL;

⁽¹⁾=Ventrolateral tongue (VLT);

⁽²⁾=Buccal Mucosa/Vestibule (BMV);

⁽³⁾=Alveolar Ridge Mucosa (ARM);

⁽⁴⁾=Floor of Mouth (FOM);

⁽⁵⁾=Retromolar trigone (RMT);

⁽⁶⁾=Others (dorsal tongue, palate, lip, etc).

Relationship between SIBLING expression in dysplastic OPLs and transition to OSCC

Twelve of 60 subjects (20%) in the cohort (Table 2) were found to have developed local OSCC following surgical removal of the OPL, a percentage that is within the broad range (5%–36%) reported within the literature for dysplastic OPL.^9–14 These included three cases with concomitant dysplasia/OSCC. As seen in Table 3a, there was no correlation between the degree of dysplasia and translation to OSCC for either mild vs. moderate + severe (P=0.44) or between each incremental level (P=0.95). There was also no significant relationship between the degree of dysplasia and the expression of BSP, DSPP, or OPN (analysis not shown).

Table 3.

Table 3a: Odds Ratios based on Two Logistic Regression Models (N=60)
Parameter	Logistic regression coefficient ± SE, P-value	Odds Ratio	95% Wald Confidence Limits	Comparison Categories
Model I* :
BSP	3.62±1.22, 0.003	37.16	3.41 – 405.56	>0 vs none
DSPP	3.25±1.02, 0.001	25.79	3.51 – 200.0	none vs >0
OPN	0.35±0.91, 0.70	1.42	0.12 – 4.22	none vs >0
Dysplasia	0.80±1.04, 0.44	2.23	0.29 – 19.61	mild vs >mild
Model II*:
BSP	3.24±1.26, 0.01	25.53	2.14 – 304.7	1 vs 0
DSPP	2.32±0.83, 0.005	10.13	2.0 – 50.0	0 vs 1
OPN	0.38±0.43, 0.38	1.47	0.63 – 3.45	0 vs 1
Dysplasia	0.04±0.56, 0.95	1.04	0.324– 3.12	0 vs 1

Table 3b: Odds Ratios based on Two Logistic Regression Models: Sensitivity Analysis (N=40, No Mild Dysplasia)
Parameter	Logistic regression coefficient ± SE, P-value	Odds Ratio	95% Wald Confidence Limits	Comparison Categories
Model I* :
BSP	3.89±1.59, 0.014	49.12	2.19 – 1000+	>0 vs none
DSPP	4.14±1.57, 0.008	62.5	2.90 – 1000+	none vs >0
OPN	0.06±0.91, 1.14	1.06	0.11 – 9.95	none vs >0
Dysplasia	1.37±1.21, 0.26	3.95	0.37 – 42.67	moderate vs severe
Model II*:
BSP	3.08±1.47, 0.04	21.81	1.23 – 386.6	1 vs 0
DSPP	2.31±1.00, 0.021	10.00	1.41 – 71.4	0 vs 1
OPN	0.26±0.48, 0.59	1.29	0.50 – 3.33	0 vs 1
Dysplasia	0.75±1.10, 0.49	2.12	0.25 – 18.18	moderate vs severe

Abbreviations: Legend BSP: bone sialoprotein, DSPP: dentin sialophosphoprotein, OPN: osteopontin

^*In model I, the parameters of BSP, DSPP, OPN and Dysplasia were coded as binary parameters (0 vs 1). Overall P-value for model I was <0.001. In model II, the parameters of BSP, DSPP, OPN were all coded as 0, 1, 2, 3, and Dysplasia was coded as mild=0, moderate=1, and severe=2. Regression coefficients of model II are based on per unit increase in scores and logistic regression models are geometric not arithmetic. Overall P-value for model II was <0.001. The interaction terms of BSP*dysplasia, DSPP*dysplasia and OPN*dysplasia were tested but were not statistically significant and therefore dropped from each of the models.

Although OPN was expressed in a number of OPL, there was no significant association (P=0.70) between this SIBLING and progression to OSCC within our cohort. DSPP expression on the other hand, showed a strong association with progression to cancer within the binary model (plus/minus DSPP, OR=25.79, 95% CI=3.51–200.0, Table 3a). Furthermore, each increase in DSPP expression score (0 to 1, 1 to 2, and 2 to 3) had a highly significant (P=0.005) corresponding increase in the likelihood of areas adjacent to the original dysplastic OPL progressing to cancer. In contrast, BSP binary expression in the OPL correlated with a decrease in subsequent local progression to OSCC (OR=37.16, 95%CI=3.41–405.56) including a significant decrease in the likelihood of progression with each increment in BSP expression score (P=0.01). (Note that the comparison categories for DSPP and BSP are inverted such that logistic regression coefficients and odds ratios are always expressed as positive values.) A second, sensitivity, analysis carried out by discounting the 20 cases of mild dysplasia (i.e. based on analysis of cases of “moderate” and “severe” dysplasia alone; Table 3b), showed OR results similar to that of all 60 cases (Table 3a) reaffirming that the degree of dysplasia of the surgically removed OPL did not correlate with progression to OSCC.

The more complex relationships of the relative benefits/risks of combined presence or absence of both DSPP and BSP (OPN being excluded for these calculations) is presented in Figure 2 as Point Prevalence (PP) values. Lesions expressing 1) neither SIBLING (BSP=0 & DSPP=0) or 2) both BSP and DSPP together (BSP>0 & DSPP>0), progressed less frequently than the cohort average (dotted line) but more frequently than those expressing BSP alone (BSP>0 & DSPP=0). Interestingly, none of the 12 patients whose surgically removed dysplastic lesion expressing BSP but not DSPP progressed to local OSCC within the experimental time period (PP=0%, 95%CI=0.0–20.6%). In contrast, 7 of the 9 cohort members whose excised OPL expressed DSPP but not BSP (BSP=0 & DSPP>0), had local progression to OSCC within the time period of the study (PP= 77.8%, 95%CI=47.8–95.4%).

Figure 2.

Point Prevalence analysis (+/− 95% confidence interval) of occurrence of OSCC at or near the site of excision for dysplastic OPL expressing BSP and/or DSPP. Expression of only BSP in precancerous lesions was associated with low (0%) OSCC transition while 78% of the patients with OPL expressing only DSPP were later diagnosed with local OSCC. OPL expressing neither or both BSP and DSPP resulted in intermediate transition values. The vertical dashed line indicates the 20% of the entire cohort that progressed.

Discussion

Each member of the SIBLING family has been reported to be up-regulated in at least one type of tumor. In several types of cancer, expression of a SIBLING appeared to predict an increased likelihood of metastasis and/or shortened lifespan although there are rare studies in which the presence of a specific SIBLING suggested a more favorable prognosis (for review see22). None of the SIBLINGs are expressed in normal oral epithelium.²⁴ Three SIBLINGs (BSP, DSPP and OPN), however, have been shown to be up-regulated in primary OSCC.²⁴ In contrast, little information about the prognostic values or even the presence of SIBLINGs in precancerous lesions currently exists. Devoll et al.³⁰ noted OPN expression in OPL and OSCC, but concluded that their study was unable to discern whether or not such expression indicated a propensity for malignant progression. Patients with clinically observed “plakias” are known to have a statistically increased chance of presenting with OSCC in future years, but the probability is low.^9–14 The additional histological diagnosis of dysplasia does indicate an increased likelihood of progression but many patients with even severe dysplasia remain healthy with or without treatment.^9–14 Because OSCC has such high morbidity and mortality, OPL patients need additional prognostic markers that can help their care-givers distinguish those dysplastic OPL that have an increased or decreased likelihood of progressing to cancer and to guide surgeons in how much tissue to remove.

Our current findings support several previous studies noting that the degree of dysplasia of surgically removed dysplastic OPL did not correlate (P>0.40) with later occurrence of OSCC. DSPP expression in the dysplastic lesion, however, was indicative of a large increase in the chance of subsequent local OSCC occurrence (compared to the cohort average), independent of the degree of dysplasia. This was true not only for the bimodal (presence/absence) analysis (P=0.001), but also each level of semi-quantitative expression of DSPP (e.g. +1 to +2, P=0.005). In contrast, the expression of BSP within a dysplasia was found to be associated with patients that later presented with local OSCC much less often than the cohort average (P=0.003), with each increase in the level of BSP expression resulting in a decrease in risk (P=0.01). Furthermore, patients with dysplastic OPL lesions expressing DSPP (but not BSP) were ~4-fold more likely to have been diagnosed with local OSCC than the cohort average and even more likely than those expressing BSP but not DSPP.

At present, we do not know the mechanism underlying the strong statistical association between the pattern of SIBLING expression and the malignant transformation potential of tissue adjacent to sites of original surgical excision of dysplastic OPLs observed in this study. It is possible that the elevated expression of DSPP, regardless of the status of BSP expression, is a necessary pre-requisite for transformation. Alternatively, the perturbations in cell-cell signaling in dysplastic lesions could fortuitously alter SIBLING expression without any direct contribution to the transformation process by the SIBLING itself. In either of these plausible scenarios, the observed pattern of SIBLING expression would be a marker for transformation and underscores the potential utility of determining SIBLING expression in dysplastic OPLs.

Furthermore, because this study was investigating the likelihood of an OSCC developing adjacent to the site of previously removed dysplastic OPL, the occurrence of such a transition would logically be the result of a small population of cells beyond the boundary of histomorphologic hallmarks of dysplasia remaining after surgery that are able to develop into frank malignancy. Therefore, our findings enable us to speculate that surgical protocols allowing for complete excision of all dysplastic OPLs may be frequently successful at removing all cells that can progress within the next few years in patients whose lesions are BSP+/DSPP−. In contrast, such protocols for patients whose lesions are BSP−/DSPP+ apparently may leave cells outside of “clear” histological margins that often progress to primary OSCC.

We are mindful that the number of cases analyzed in this retrospective study is limited and that larger, prospective designs are required to further assess the utility and clinical applicability of DSPP and BSP in the overall clinical management of OPLs. Such prospective designs should incorporate stratification based on accurate clinical designation of lesions as leukoplakia, speckled leukoplakia, or erythroplakia as well as analyses of DSPP/BSP-defined surgical margins against the odds of transition to primary OSCC. Prospective studies comparing long-term outcomes with the presence of DSPP and/or BSP in biopsies of OPL without concurrent surgical excision of the lesion is also encouraged alongside studies analyzing lesions on lesions the basis of defined oral cavity sites. Other parameters of interest such as the roles of tobacco products and alcohol consumption can also be incorporated into these larger prospective studies.

In summary, immunohistochemistry with semiquantitative analysis of the expression of BSP, DSPP, and OPN in surgical biopsies OPL from 60 patients showed that DSPP was correlated with increase risk of the appearance of local primary OSCC while BSP expression was correlated with significantly decreased risk.

Abbreviations

OPL

oral premalignant lesion

OSCC

oral squamous cell carcinoma

SIBLING

Small Integrin Binding Ligands N-linked Glycoprotein

BSP

bone sialoprotein

DSPP

dentin sialophosphoprotein