Dysplasia THESIS. Jeffrey M. Hagen. Graduate Program in Dentistry. The Ohio State University. Master's Examination Committee:

Transcription

1 CD31: Invasive Predictive Biomarker for Malignant Transformation of Oral Epithelial Dysplasia THESIS Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University By Jeffrey M. Hagen Graduate Program in Dentistry The Ohio State University 2013 Master's Examination Committee: Susan R. Mallery, D.D.S., Ph.D., Advisor Gregory Ness, D.D.S. Henry Fields, D.D.S., M.S., M.S.D.

2 Copyright by Jeffrey M. Hagen 2013

3 Abstract While not all oral epithelial dysplasia (OED) lesions will transform, approximately one third of OED lesions progress to oral squamous cell carcinoma (OSCC). Provided the poor overall prognosis for OSCC patients, intervention at the premalignant stage of OED is an attractive clinical strategy. We do not, however, currently have biomarkers to identify the most aggressive premalignant lesions. Identification of such biological indicators would allow for targeting of OED lesions with high transformation risk and direct lesional-specific management e.g. anti-angiogenic compounds. A wealth of translational and clinical research supports the key role of angiogenesis during malignant transformation. Past human studies involving preneoplastic lesions of breast and cervical tissue have revealed a positive correlation between angiogenesis and the progression, prognosis and metastasis of ductal carcinoma in situ and cervical intraepithelial neoplasia. In addition to studies involving human breast and cervical tissue, numerous studies have been carried out to examine the association of between angiogenesis and progression of OED. These data present compelling evidence regarding MVD s importance in carcinogenesis. A site-matched longitudinal assessment was conducted to determine if increased microvascular density positively correlates with malignant transformation. The clinical basis of this current study: i) identifiy oral epithelial dysplastic lesions likely to progress ii

4 to OSCC, thus, dictating a more aggressive treatment modality and ii) establish reliable clinically relevant biomarkers for OSCC chemoprevention. iii

5 Vita 1998 Columbia Station High School 2002 B.S. Biology, Bowling Green State University 2006 D.M.D. University of Pittsburgh School of Dental Medicine 2013 Oral Maxillofacial Surgery Certificate, The Ohio State University Major Field: Dentistry Fields of Study iv

6 Table of Contents Abstract... ii Vita... iv List of Tables... vii List of Figures... viii Chapter 1: Introduction... 1 Epidemiology of Oral Squamous Cell Carcinoma...1 Tumor-Associated Angiogenesis...2 Angiogenic Switch in Neoplasia...3 Angiogenesis and Transformation of Oral Epithelial Dysplasia Shortcomings of Previous Oral Cavity Studies...8 Design Novelties Available...10 Inhibition of Angiogenesis as a Therapeutic Strategy...12 Clinical Implications of Current Study...12 Specific Aims...13 Null Hypothesis...13 Chapter 2: Materials and Methods Rational for selection of CD Justification for automated image analysis...14 v

7 Sample Identification and Selection...15 Histological Grading of OED...16 Immunohistochemical Staining...17 Quantification of Vascularity...19 Statistical Analysis...25 Chapter 3: Results...27 Chapter 4: Discussion and Conclusion...42 Conclusion...48 References...50 vi

8 List of Tables Table 1. Patient Profiles Table 2. Correlation between MVD and progression over time...30 Table 3. Correlation between MVD and nonprogression over time...31 Table 4. Nonprogressive association between MVD and histologic grade...35 Table 5. Progressive association between MVD and histologic grade...39 Table 6. MVD analysis between progressive and nonprogressive groups...41 vii

9 List of Figures Figure 1. Western blot analysis of HUVEC Figure 2. CD31 IHC stained specimens Figure 3. Image analysis Figure 4. 40x scanned image Figure 5. Vessel detection Figure 6. Lumen detection Figure 7. Progressive timeline correlation Figure 8. Nonprogressive timeline correlation Figure 9. Nonprogressive association between total MVD and histological grade Figure 10. Nonprogressive association between small vessel MVD and histological grade Figure 11. Nonprogressive association between medium vessel MVD and histological grade Figure 12. Nonprogressive association between large vessel MVD and histological grade Figure 13. Progressive association between total MVD and histological grade Figure 14. Progressive association between small vessel MVD and histological grade viii

10 Figure 15. Progressive association between medium vessel MVD and histological grade Figure 16. Progressive association between large vessel MVD and histological grade Figure 17. MVD analysis...40 ix

11 Chapter 1: Introduction Epidemiology of Oral Squamous Cell Carcinoma Head and Neck squamous cell carcinoma (HNSCC) is the sixth most common human cancer worldwide with a long-term survival rate of less than 50% (1). Similar to other carcinomas, HNSCC s arise from malignant transformation of a recognized precursor lesion known as oral epithelial dysplasia (OED) (2,3). Although up to 36% of these precursor lesions will progress to oral squamous cell carcinoma (OSCC), we are currently unable to predict which lesions will undergo malignant transformation (2,3). Conventional management of oral epithelial dysplasia is based on histologic grading into mild, moderate, severe dysplasia and carcinoma in situ (4). Complete excision with close clinical follow-up is standard for moderate or higher grade OEDs (5). Despite surgically clear margins, local OED recurrences are common (6,7). The economic, physiological and psychosocial impacts of this treatment modality are substantial, especially in cases where post-surgical disfigurement and/or disability result (1). Ideally, patient management would be based on characteristics established to predict lesional behavior. Although tobacco and alcohol are the best-established risk factors, HNSCC tumors also arise in persons that are apparently risk-factor-free. Due to the inability to predict which OED s will progress research has begun to focus on potential biological prognostic and predictive factors, specifically angiogenesis and highly specific endothelial markers. 1

12 Tumor-Associated Angiogenesis Blood vessels are fundamentally composed of endothelial cells, which interconnect to form the tubes that direct and maintain blood flow and tissue perfusion (8). Blood vessels form during embryogenesis via two processes; vasculogenesis, whereby endothelial cells are born from progenitor cell types; and angiogenesis, in which new capillaries sprout from existing vessels (9). New vessels in an adult are formed only by angiogenesis and may be activated by an appropriate stimulus such as wound healing and tumor growth. Angiogenesis is a complex process. The basement membrane surrounding the endothelial cell tube is locally degraded, and the endothelial cells underlying this regional disruption in the barrier change shape and invade into surrounding stroma. This invasion is accompanied by proliferation of the endothelial cells at the leading edge of what becomes the migrating column. A region of differentiation trails behind the advancing front, where the endothelial cells cease proliferating, change shape, and tightly adhere to each other to form a lumen of a new capillary tube. Finally, sprouting tubes fuse and coalesce into loops, circulating the blood into this newly vascularized region (10). The current belief is that both solid and hematological tumors are endowed with angiogenic capability and that their growth, invasion and metastasis are angiogenesisdependent (11). Folkman and colleagues demonstrated that solid tumors couldn t grow any larger than 2-3mm in diameter without being able to induce their own blood supply (12). Neoplastic cell populations can only form a clinically observable tumor if the host 2

13 produces a vascular network sufficient to sustain their growth. Furthermore, new vessels provide them with a gateway through which to enter the circulation and metastasize to distant sites (13). In normal tissues, vascular quiescence is maintained by the dominant influence of endogenous angiogenesis inhibitors over angiogenic stimuli. Tumor angiogenesis, on the other hand, is induced by increased secretion of angiogenic factors and/or down regulation of such inhibitors (13). The balance between the level of angiogenesis inducers and inhibitors may be altered by; increasing activator gene expression, changing the bioavailability or activity of the inducer proteins, or reducing the concentration of angiogenic inhibitors (14,15). The switch to an angiogenic phenotype separates an avascular phase characterized by a dormant tumor and a vascular phase in which exponential tumor growth may ensue (13). Angiogenic Switch in Neoplasia An article by Hanahan and Folkman (14) reviewed three previous studies by them and their colleagues related to the angiogenic switch in transgenic mice that reproducibly elicit tumors of specific cell types (26-30). Specifically, they investigated the pattern of angiogenesis in three diverse transgenic mouse models: islet cell carcinoma, dermal fibrosarcoma, and epidermal squamous cell carcinoma. In all three models an angiogenic switch could be visualized at early stages preceding the appearance of the solid tumors, suggesting that activation of angiogenesis is a discreet event in tumor development (14). Furthermore histological analysis revealed the hallmark of angiogenesis, capillary sprouting and endothelial cell proliferation in all solid tumors of the transgenic mice (26). 3

14 Two significant studies evaluated a transgenic mice multistage model (normal epithelium, hyperplasia, dysplasia, carcinoma) whereas basal keratinocytes were progressively transformed into squamous cell carcinoma by the human papillomavirus type 16 oncogenes (29,30). In the study by Arbeit and colleagues angiogenic activity and increased vessel density could be detected as early as the hyperplastic stage (29). Coussens et al. found angiogenesis became pronounced in the dysplastic stage, with abundant capillaries juxtaposed to the basement membrane separating stroma from dysplastic cells (30). These findings of angiogenesis in pre-neoplastic disease are consistent with the identification of an angiogenic switch involved in tumorogenesis. In addition to the transgenic mice model, human studies involving pre-neoplastic lesions of the breast have correlated angiogenesis with progression, prognosis and metastasis (31-33). Using von Willebrand Factor (vwf) immunohistochemical staining ductal carcinoma in situ (DCIS) lesions and normal surrounding breast tissue, Guidi et al. observed microvessels were clearly more prominent in association with the DCIS than in association with the benign breast tissue. Furthermore, tumors associated with higher grade DCIS demonstrated an increase microvessel density score (32). Similarly, Weidner et al. evaluated DCIS and early stage-breast carcinoma utilizing vwf immunohistochemical staining. Their findings suggest microvascular density is a statistically significant predictor of overall survival and relapse free survival for DCIS and early stage-breast carcinoma (33). Pre-neoplastic lesions of the human cervix also support angiogenesis in correlation with progression (34,35). Cervical intraepithelial dysplasia (CIN) is a 4

15 premalignant precursor to cervical cancer and is graded histologically as mild (CIN I), moderate (CIN II), and severe dysplasia or carcinoma in situ (CIN III) depending on the amount of nuclear abnormality and degree of differentiation (36). In a study by Smith- McCune et al. surgical specimens ranging from CIN I-III were immunohistochemical (IHC) stained for factor VIII antigen. The results demonstrated a region of neovascularization develops along the basement membrane underlying the dysplastic epithelium. Comparison of microvessel counts of CIN I lesions with microvessel counts of CIN III lesions shows a statistically significant increase in the more advanced lesions (34). In a similar study, Guidi et al. evaluated CIN I-III as well as invasive squamous cell carcinoma. Utilizing the same IHC factor VIII antigen, microvessel density counts were significantly increased in invasive carcinoma and CN III lesions compared to CN I lesions (35). These data demonstrates a progression of the switch to an angiogenic phenotype in a naturally occurring pre-invasive lesion and suggests it is an early event in the pathway to invasive squamous cell cancers (14). Angiogenesis and Transformation of Oral Epithelial Dysplasia In addition to studies involving the breast and the cervix, multiple studies have focused on angiogenesis associated with OED and HNSCC. Pazouki et al. examined the vascular changes involved in the transition from normal to dysplastic and neoplastic tissue in the oral mucosa utilizing various vessel markers and methods of assessment. A total of 100 specimens including normal oral mucosa, dysplastic and squamous cell carcinoma tissue specimens were examined. IHC was performed utilizing vwf, CD31 and α v β 3 (38). Vascularity was quantified utilizing microvascular volume (MVV) and 5

16 microvascular density (MVD). MVV was performed by point-counting ten random fields (1000 points), across each section and only vessels coinciding with the points were counted (38). MVD was performed as previously described by Weidner et al. Areas of intense MVD ( hot spot ) were identified by scanning the sections at x40 and x100 magnification. Three of the identified fields were then evaluated at x200 magnification (0.739 mm 2 per field) and the highest value was taken and expressed as vessels per mm 2 (39). vwf and CD31 showed a significant and linear increase in MVV with increasing severity of disease. When vessels were estimated by MVD method, there was a significant difference between normal oral mucosa and dysplastic lesions and between normal oral mucosa and OSCC. However, no significant difference was observed between the various degrees of dysplasia and squamous cell carcinoma. The α v β 3 marker failed to show any significant difference in vascularity between OED and OSCC. The authors felt α v β 3 to have little diagnostic value and it does not appear to reflect the presence of angiogenic vessels in oral tissues. The author concluded vwf and CD31 as biological markers and MVV analysis for quantitation may be useful in prognostic value of vascularity in OSCC (38). In a similar study, Macluskey et al. evaluated angiogenesis in normal, dysplasia classified as mild, moderate and severe, and squamous cell carcinoma tissue samples (37). A total of 47 paraffin-embedded archival oral specimens were IHC stained with vwf and quantification of MVV and MVD was performed as previously described by Pazouki et al and Weidner et al (38,39). MVV values increased significantly with disease 6

17 progression in a step-wise fashion from normal mucosa, through degrees of dysplasia, to carcinoma. In contrast, MVD did not distinguish between carcinoma and dysplasia. Siar et al. evaluated the relationship between angiogenic changes in pre-invasive bronchial lesions and OED (40). Angiogenic Squamous Dysplasia (ASD) is a qualitative distinct form of angiogenesis, first described in pre-invasive bronchial mucosa of highrisk individuals (41). Their study included 20 specimens of mild, moderate and severe OED and 10 normal oral mucosa specimens. IHC was carried out utilizing CD31 and CD34 antigen and MVD assessment was done as previously described by Weidner et al (39). MVD scores were significantly higher in dysplastic than in normal mucosa and a gradual increase in vascularity score was observed as the oral mucosa progressed from normal to increasing dysplasia grade. ASD like foci, CD31 and CD34 positive loops abutting the overlying dysplastic oral epithelium, was found in moderate and severely dysplastic tissue but were not evident in normal mucosa or mild dysplasia (40). This study further reiterates the increase in vascularity as OED lesions progress as well as suggests that an angiogenic stimulus may be associated with epithelial dysplasia in the upper aerodigestive tract. Additional studies evaluated OSCC associated angiogenesis and tumor stage, lymph node involvement and metastasis as a prognostic indicator (42,43). Penfold et al. (42) evaluated 41 formalin fixed wax imbedded specimens of primary OSCC. IHC was performed utilizing CD31 antigen and microvessel density was performed according to the method described by Horak et al (44). Briefly, microvessels were counted on three 400x fields within the areas of maximum vascularity and the mean microvessel count was 7

18 calculated for each area. No significant difference in tumor vascularity was found between tumor stages but the mean vessel count for tumors with node metastasis was significantly higher than the mean vessel count in tumors with no lymph node metastasis (42). Similarly, Ascani et al evaluated 64 specimens of primitive epidermoid carcinoma of the oral cavity. IHC was performed with CD34 antigen, which has a tendency to stain both the lymphatic and blood vessels (43). MVD was performed according to Weidner et al (39). MVD was found to statistically increase with histologically grading as well as was statistically significantly higher in cases with node metastasis (43). These findings may suggest that MVD is useful marker to identify patients with a more aggressive tumor and may benefit from a more aggressive therapeutic approach. Collectively, these findings support that increased neovascularization accompanies the transition from normal oral mucosa to mild/moderate dysplasia to OSCC. Shortcomings of Previous Oral Cavity Studies While the previously cited studies introduced strong data to support a neovascularization-malignant transformation potential in oral neoplasia, aspects of their study design were suboptimal. Collectively, these studies utilized archival tissue specimens from the oral cavity including floor of the mouth (37,42,43), tongue (37,40,42,43), buccal mucosa (37,40,42,43), alveolar mucosa (37,40,42), retromolar pad (37,43), palate (37,43) and others (38,40,43). Within the oral cavity, most tumors arise from the floor of the mouth and ventro-lateral tongue (50). It is possible the difference in angiogenic expression between the sites mentioned above could be due to the inherent 8

19 properties of where the specimens were obtained and their tendency to progress. Because the design was cross-sectional, progression of histological grade was evaluated between different subjects, which could introduce bias due to inter-patient differences. IHC staining utilized antibodies to factor VIII (37,38), CD31 (38,40,42) and CD34 (40,43). Three antibodies currently available to formalin-resistant endothelial antigens include FVIII, CD31 and CD34 (44). Horak et al. compared CD31 and factor VIII vessels stained by the two antibodies in 12 cases. Staining of the identical fields with CD31 consistently revealed higher number of microvessels and was found to be a more sensitive marker for endothelial cells relative to FVIII (44). Ascani et al. indicated CD34 monoclonal antibody stains both lymphatic vessels as well as endothelial cells. Within their study, they found CD31 to more specifically stain endothelial cells (43). While this makes CD34 an optimal marker to assess loco-regional metastasis, this combined staining would affect MVD quantification while assessing OED progression and OSCC. It was determined from the findings of these studies that CD31 is a superior immunohistochemical stain to assess angiogenesis as well as OED and OSCC progression than FVIII and CD34. MVD quantification was performed according to Weidner et al. and Horak et al (39,44). Their methods identified three hot spots or areas of increased vascularity in x200 and x400 fields respectively and the values were recorded. This method has been proven reliable however, represents only a small portion of the tissue in question and may explain varying results between similar studies. 9

20 Design Novelties Available New options exist for studies involving angiogenic progression of neoplasias. As previously mentioned, known studies involving the oral cavity were cross sectional in nature (single biopsy from a patient and did not follow the patient over time) and involved multiple anatomical locations within the oral cavity (37,38,40,42,43). Site matched longitudinal specimens within the same patient can offer advantages. Specifically, when all specimens included are from the same site (i.e. tongue) and each patient included had a minimum of 4 biopsies from this site over a minimum of a fouryear period this is more optimal. This allows both intra-patient and inter-patient comparisons. Intra-patient comparisons allow each participant to serve as their own internal control, which minimizes data skewing effects observed in mean values as a result of large inter-patient differences and enables detection of differences that are significant for the patient. Because inter-patient analyses enable extrapolation to the population at large, these assessments can be conducted. Furthermore, longitudinal intrapatient assessment may provide information regarding early progression as well as a possible specific indicator of OED progression. This methodology description was provided in part by Mr. David Kellough, Polaris Innovation Center, The Ohio State University Department of Pathology. MVD quantification has been calculated by methods developed over 20 years previously as described by Weidner et al. (39) and Horak et al (44). Unique to our study, is the use of Aperio digital image scanner (XT) and Definiens Tissue Studio 3.5 to provide highly specific and sensitive computer based image analysis of CD31 IHC stained specimens. 10

21 Automated and semi-automated workflows allow us to evaluate the number and percentage of small, medium and large vessels as well as MVD within a specific predetermined field. The quantitative measurement of CD31 provides concise data for validation of tissue observations. Furthermore, this will provide accurate and reproducible data that may be used in translational research. Cytokine and Growth Factor Interactions in Tumor Associated Angiogenesis The expression of the angiogenic phenotype in the tumor microenvironment involves the interaction of many cell types and molecules (16). Keratinocytes as well as HNSCC are capable of producing molecules that induce angiogenesis (17). Interleukin-8 (IL-8) has been identified in numerous studies as a major angiogenic factor (18-20) and found to be the primary angiogenic mediator found in bronchogenic carcinoma (21). Vascular endothelial growth factor (VEGF) is involved with every stage of vascular development promoting survival, migration, specialization and proliferation of endothelial cells (22) and its increased expression is seen in HNSCC (23,24). In recent studies Tong et al. demonstrated that in addition to VEGF s paracrine function essential for tumorassociated angiogenesis, VEGF also plays an important intracrine function by directly enhancing HNSCC cell mitogenesis and invasiveness. These data introduce the prospect that VEGF targeted therapy has the potential to fulfill both anti-angiogenic and antitumorogenic functions (25). In addition, macrophages recruited to tumor microenvironment secrete a number of angiogenic factors including Fibroblast Growth Factor (bfgf), VEGF, Tumor Necrosis Factor-alpha (TNF-α), and IL-8 (45). 11

22 Inhibition of Angiogenesis as a Therapeutic Strategy The balance hypothesis proposed by Hanahan et al. suggests that the balance of inhibitors and inducers governs the angiogenic switch (8). α-interferon identified by Brouty-Boye et al. (46) and platelet factor-4 identified by Taylor et al. (47) were among the first identified endogenous angiogenesis inhibitors. They were found to inhibit endothelial cell chemotaxis and proliferation, respectively (46,47). Thrombospondin-1 (TSP-1) was later identified as an inhibitor of angiogenesis utilizing an anti-angiogenic hamster cell line (48). TSP-1 was later isolated by Good et al. and found to block neovascularization in vivo in the rat cornea as well as to inhibit chemotaxis of capillary endothelial cells towards angiogenic factors (49). Furthermore, increased levels of TSP-1 have been shown to be expressed at high levels in a number of normal rodent and human cells and at appreciably lower levels in many tumor cells (8). The prevailing evidence suggests that changes in the relative balance of inducers and inhibitors of angiogenesis can activate an angiogenic switch. Thus, either the reduction of inhibitor concentration or by increasing the activator levels can change the balance and activate the switch, leading to the growth of new blood vessels (8). This evidence has made anti-angiogenic therapy an attractive modality for treating and perhaps preventing the development of malignant neoplasms. Clinical Implications of Current Study The compelling evidence that angiogenesis and neovascularization are critical for the growth, progression, and metastasis of solid tumors has led to the investigation of angiogenic pathways and their inhibition. Hundreds of molecules with anti-angiogenic 12

23 activity in preclinical models have been reported, and many of them have entered clinical testing (54). The efficacy of anticancer therapies, including anti-angiogenic therapies is commonly evaluated by measuring direct effects on the tumor by use of the Response Evaluation Criteria in Solid Tumors scale (51), and by determining the impact of treatment on disease progression (i.e. progression-free survival and/or time to progression, survival or time to death) or quality of life (52). These factors represent an endpoint rather than a prognostic indicator of disease progression. This suggests development of a predictive biomarker targeted strategy such as MVD, particularly OED progression to OSCC. The clinical basis of this current study: i) identify oral epithelial dysplastic lesions likely to progress to OSCC thus, dictating a more aggressive treatment modality and ii) establish reliable, clinically relevant biomarkers for OSCC chemoprevention. Specific Aims The specific aims are to: 1) Determine if increased expression of CD31 MVD is associated with increased malignant transformation of OED and 2) Determine if CD31 is a potential biomarker to aid in prediction and prevention of OED s clinical behavior. Null Hypothesis Levels of IHC stained CD31 MVD will be comparable over time within histologic grade (using appropriate correlative analyses) in the progressed and nonprogressed patient cohorts. 13

24 Chapter 2: Materials and Methods Rational for selection of CD31 Platelet endothelial cell adhesion molecule (PECAM-1/CD31) is a member of the immunoglobin super family of cell adhesion molecules with important roles in angiogenesis and inflammation (60). PECAM-1 is a primary constituent of endothelial cell-cell junctions in confluent vascular beds (61) and their expression pattern is regulated during development and angiogenesis (62). In a study by Park et al. the lack of PECAM- 1 had significant impact on endothelial cell-cell and cell matrix interactions resulting in attenuation of cell migration and capillary morphogenesis. Inversely, re-expression of PECAM-1 restored cell migration and capillary morphogenesis. These finding suggests PECAM-1 expression modulates pro-angionic properties of endothelial cells (60). CD31 is a common antibody utilized to detect vascularity and has been used to identify endothelial cells in multiple tissues including bladder, prostate, ovary, uterus, breast, lung as well as OED and HNSCC. In addition, as previously discussed, CD31 has been identified as one the most specific immunohistochemical markers to detect endothelial cells lining vascular channels (44). Justification for automated image analysis Image analysis has been extensively studied and found to be a highly acceptable adjunct to microscopic evaluation of tissue (55-59). Hilbe et al. compared manual 14

25 microscopy to automated cellular imaging in small cell lung cancer. Seventy-one frozen lung tissue specimens were stained with monoclonal antibodies against p53, ki-67 and p120. The slides were evaluated by standard manual microscopy (MM) and an Automated Cellular Imaging System (ACIS). The results obtained with ACIS correlated significantly with MM examination (p<0.001). However, ACIS showed a higher sensitivity, especially for specimens with low infiltration volume (55). Shaw et al. compared scanned virtual slide images with glass slides in the assessment of morphological characteristics of breast cancer. Thirteen histopathologists reviewed virtual and glass slides and assessed morphological features such as tumor grade and type. A positive inter-rater agreement was found for vascular invasion, necrosis and presence of scar (κ= ) (56). A poor inter-rater agreement, however, was detected for subjective parameters such as pleomorphism (κ=0.1) (56). Lloyd et al. evaluated Human epidermal growth factor-2 (HER2) and the activation of estrogen receptor (ER) pathway in invasive ductal carcinoma utilizing image analysis and manual scoring. They found a 100% inter-rater agreement within the 33 cases evaluated (57). These findings demonstrate that image analysis is highly sensitive and reproducible in identifying vessels. Sample Identification and Selection This section was taken directly from Dr. Jeffrey Hagen s Ohio Cancer Fellowship Application that preceded Dr. Mainville s participation in this project. Prior to sample identification and selection, an Ohio State institutional review board exemption was obtained. Tissues from sequential biopsies obtained from 10 total patients (5 persons 15

26 whose OED lesions remain stable and five whose lesions progressed to OSCC) were identified within The Ohio State University s Oral and Maxillofacial Pathology Biopsy Service reports. Matched parameters include: (i) lesional site (ventral-lateral tongue), (ii) low grade initial biopsy (mild to moderate dysplasia). Mean ages comparable, with patient ages ranging from 39 to 70. The duration of patient follow-up is appreciably longer in the stable group, which provides a more stringent component to these follow up studies. Demographics for the progressive group include: (i) initial histopathologic diagnosis of moderate to severe OED, (ii) mean initial age 50 years old, (iii) same anatomical location (i.e. ventral-lateral tongue) for all specimens, (iv) minimum four same site specimens, (v) progression to squamous cell carcinoma. Demographics for the non-progressive, stable disease group include: (i) initial histopathologic diagnosis of moderate to severe OED, (ii) mean initial age 55 years old, (iii) same anatomical location (i.e. ventral-lateral tongue) for all specimens, (iv) minimum four same site specimens, (v) biopsies obtained over a minimum of four years and (vi) histopathologic grade remains mild or moderate. Following identification and sectioning, a random code was assigned to deidentify the specimens. Histological Grading of OED Hematoxylin and Eosin stained slides were evaluated by two board certified Oral and Maxillofacial Pathologists and graded according to World Health Organization histological classification of tumors. Diagnosis was determined and graded as the 16

27 follows: 0.5 atypia, 1.0 mild OED, 1.5 mild-moderate OED, 2.0 moderate OED, 2.5 moderate-severe OED, 3.0 severe OED, 3.5 carcinoma in-situ, 4.0 OSCC. All lesions were graded prior to being subjected to IHC staining. Immunohistochemical Staining Immunohistochemical (IHC) staining for CD31 was performed on 4µm thick sections from each biopsy tissue block. Sections were heated to 60 C for one hour then de-paraffinized in xylene and rehydrated in graded alcohol. These sections were then immersed in 3% H 2 O 2 in methanol (20ml 30% H 2 O ml methanol) for 10 minutes to block endogenous peroxidase activity then rinsed in PBS (5 minutes, two times). Antigen retrieval was performed with 10mM Citrate buffer (9.4ml Antigen Unmasking Solution ml DH 2 O) (Vector Antigen Unmasking solution. Citrate based. Cat. No. H Vector Laboratories, Inc., Burlingame, CA) heated in a microwave for 2 minutes at maximum intensity, followed by 5 minutes at reduced intensity. Sections were rinsed in PBS twice, for 5 minutes. Excess PBS was removed by blotting and lesional tissue circumscribed with ImmEdge TM pen (Cat. No. H Vector Laboratories, Inc., Burlingame, CA). Sections were incubated for one hour with Blocking Buffer (5ml: 250µl normal horse serum + 500µl 10% BSA + 2.5µl Tween µl PBS). Sections were allowed to incubate overnight at 4 C in a humid chamber with anti human CD31 mouse monoclonal antibody in blocking buffer at 1:100 dilution (10µl CD31 (200µg/ml) + 900µl blocking buffer) (CD31 (PECAM-1) (89C2) Mouse mab #3528, (Cell Signaling Technology, Danvers, MA). Sections were rinsed in PBS 17

28 twice, for 5 minutes. Sections were incubated for one hour at room temperature in 7.5µg/ml biotinylated secondary antibody solution at 1:200 dilution (5µl 1.5mg/ml antimouse IgG biotinylated anti-mouse IgG + 1ml blocking buffer) (Biotinylated anti-mouse IgG (H+L), Vector Laboratories, Burlingame, CA). Sections were rinsed in PBS twice, for 5 minutes. Sections were incubated for 30 minutes with VECTASTAIN ABC (VECTASTAIN Elite ABC PEROXIDASE KITS Cat. No. PK Vector Laboratories, Inc., Burlingame, CA). Sections were rinsed in PBS twice, for 5 minutes. Sections were developed with DAB Peroxidase Substrate Kit (2.5ml DH 2 O + 1 drop Blocking Buffer + 2 drops DAB + 1 drop H 2 O 2 )(Peroxidase Substrate Kit (3,3 - diaminobenzidine) Cat. No. SK Vector Laboratories, Inc., Burlingame, CA) for 2 minutes until a visible brown pigment was produced. The sections were then rinsed and counterstained with Mayer s hematoxylin for 1 minute. Sections were dehydrated in graded alcohol and Clear-Rite 3 and a cover slip was placed with Permount. Negative control, sections were treated with the above protocol, but without the primary antibody. Negative controls were conducted during each IHC processing. Positive control consisted of tissue of known histological diagnosis, pyogenic granuloma, and was treated with the above protocol. Positive controls were conducted during each IHC processing. Specificity and sensitivity were determined per the manufacturer. CD31 is a monoclonal antibody produced by immunizing mice with recombinant CD31 protein. Western blot analysis of human umbilical vein endothelial cells (HUVEC) confirms that 18

29 CD31 (PECAM-1) (89C2) detects endogenous levels of total CD31 protein and does not cross-react with other related proteins (CD31 (PECAM-1) (89C2) Mouse mab #3528 (Cell Signaling Technology, Danvers, MA). Figure 1. Western blot analysis of HUVEC Quantification of Vascularity This methodology description was provided in total by Mr. David Kellough, Polaris Innovation Center, The Ohio State University Department of Pathology. Aperio digital image scanner (XT) and Definiens Tissue Studio 3.5 provided instrument and software support for translational morphometric projects. These technologies were housed in separate areas of the Innovation Centre with each process managed by a technologist. The Definiens software allowed the researcher to quantify bright field H&E and immunohistochemistry stained digital tissue images on a cell-by cell basis using 19

30 automated and semi-automated work flows. Quantitative tissue biomarker measurements provided concise data for validation of tissue observations in translational research. Tissue sample based workflows can be tailored for the investigator s specific needs using simple graphical controls and investigator tissue sample-based training ( Definiens AG. Gernhard-Wicki-Strasse Munchen, Germany). Stained tissue sections are scanned to produce digital slide images using the ScanScope XT by Aperio (Vista, California). Research images may be acquired at 5x, 20x, and 40x magnifications and are completed in batches of up to 120 tissue mounted glass slides and typically take about 6 minutes/slide for a single 20x scan. Digital slides are in SVS (a proprietary version of JPEG2000) files which may be viewed at 1x up through 20x (40x window) and annotated using Aperio s Imagescope viewer software, stored for review on a secure Department of Pathology server, shared with collaborating researchers, and submitted for digital image analysis. Image analysis was undertaken using Tissue Studio 3.5 software by Definiens (Munich, Germany). Tissue Studio uses a context-based, relational analysis of the component pixels in digital slide images to differentiate among tissue morphologies; measure tissue regions of interest (ROI); identify and count brown or red IHC-stained nuclei, cytoplasm, membranes, or cells; count negative and positive nuclei, cytoplasm, membranes, or cells; quantify the area, intensity, and completeness of IHC staining; identify, measure, categorize, and count blood vessels; and score the stained slide. 20

31 Analysis may be conducted on the entire tissue section or limited to just certain tissue morphologies or other designated ROI areas. Tissue Studio 3.5 may be used equally for analysis of whole tissue sections or for cores within a tissue microarray. Arrayed tissue cores were organized for analysis by a facilitated algorithm. Once the software has been trained to recognize desired tissue morphologies, analysis preferences indicated, and marker thresholds set, Tissue Studio 3.5 runs in batch mode, processing large numbers of individual images automatically. Image analysis results using Tissue Studio 3.5 software consist of a color segmentation maps for ROI and cellular analysis that can be checked visually against a detailed numerical data report in the Microsoft Excel spreadsheet reflective of the selected and trained pathways related to the ROI. Included are up to 8 concurrent tissue morphologies, such as specific cell types (epithelial or glands), blood, blood vessels, and necrosis. Specific work paths such as cell counting, cell coloration in nuclear, cytoplasmic and membrane compartments (including percentage of cells with and without coloration) were recorded as numerical data and visually checked against the produced segmentation map. Additional data may be extracted from the analysis by categorizing and comparing measures of size, area, intensity, and relations such as cooccurrence within and between different features. In house studies and industry practice have demonstrated the consistency of data produced by standard work path digital image analysis vs. traditional human visual analysis and estimation of ROI in digitized slide images (55-59). This concludes the methods provided by Mr. Kellough. 21

32 Specific to our study, all tissue samples IHC stained for CD31 were subjected to image analysis. The scanned sections were analyzed for components of blood vessels (i.e. stained epithelial cells, lumens) followed by recognition of complete vessels, and, finally the segmentation of the blood vessel boarders. Parameters were defined to involve the epithelium and 700µm from the lowest portion of the basement membrane for each specimen. The measured distance allowed standardization between specimens for comparison as well as incorporates the area of interest adjacent to the basement membrane. Figure 2. CD31 IHC stained specimens 22

33 Figure 3. Image analysis. Region of interest (dark blue) 700µm from basement membrane. IHC stain level detection was set to most accurately assess stained endothelial cells while minimizing detection of other cell types. Vessel classification was based on size as well as the presence (green) or absence of a lumen (purple). A small vessel was classified as 0-10µm (yellow), medium vessel 11-25µm (orange), and large vessels 25+µm (red). Detection of small vessels and absence of a vessel lumen will most closely correspond to newly forming vessels and the larger vessels will correspond with increase maturity. All settings were consistent throughout the image analysis component of this study. 23

34 Figure 4. 40x scanned image Figure 5. Vessel detection. Small vessel 0-10µm (yellow), medium vessel 11-25µm (orange), and large vessels 25+µm (red). 24

35 Figure 6. Lumen detection. The presence (green) or absence of a lumen (purple) Statistical Analysis Continuous data such as age, number of biopsies and length of follow up were analyzed utilizing a Student T-test. Categorical data such as gender was analyzed utilizing a Fisher s Exact test. Ordinal data such as grade of dysplasia was done with a Mann-Whitney U Test. These patient demographic statistical analyses were conducted by Dr. F. Michael Beck and these data appear in Dr Gisele Mainville s Master s Thesis. Associations between CD31 MVD and time, and CD31MVD and grade of epithelial dysplasia were analyzed by the appropriate correlation analyses as dictated by the data distribution. Both intra and inter-group analyses used these methods. Comparisons of inter-group differences in MVD lumen size (as determined by CD31 25

36 staining of small, medium, large and total vessels) employed an Unpaired T-test. 26

37 Chapter 3: Results Patient profiles data are comparable in the progressed and nonprogressed groups. Histopathogical slides were obtained from 10 patients. Five patients with initial diagnosis of mild, moderate or moderate to severe OED who progressed to OSCC and 5 patients that did not were included in the study. Statistical analyses were conducted to compare the progressed and nonprogressed patient population. A Student T-test was employed to evaluate continuous data. No statistical association was found between age (p=0.62), number of biopsies (p=0.86) and length of follow up (p=0.76). A Fisher s Exact test was employed to evaluate categorical data. No statistical difference was found between gender (p=1). It was concluded that no statistically significant difference was found between the two patient populations. As Drs. Mainville and Hagen s research was conducted on the same pool of patients, these demographic data and resulting statistical values are identical to those presented in Dr. Gisele Mainville s thesis. 27

38 Patient Profiles Progressive Group Nonprogressive Group p Value Number of Patients 5 5 Mean Age Gender Ratio (M/F) 1/4 1/4 1 Mean Number of Biopsies/Patient Mean Follow- Up (months)(sd) 76.6(69.2) 87.0(25.8) 0.76 Table 1. Patient Profiles Preliminary data generated by image analysis software required refinement. Despite inclusion of positive and negative control tissues for every IHC slide set and careful delineation of evaluation criteria for vessel detection, initial data did not detect obvious vascular channels. To rectify this situation, an extensive analytical analysis was conducted on CD31 stained slides. The purpose of these studies was to calibrate or teach the computer recognition to reproducibly recognize CD31 staining associated with vessels. No significant associations were determined between an increase in MVD and progression to OSCC. All MVD intra and inter-patient data were initially analyzed using scatter plots in order to determine if a relationship existed between evaluated parameters. Following plotting MVD data along a scatter plot no clear pattern was identified. Previous literature consistently showed that OED lesions increase in vascularity between normal and dysplastic tissues. Therefore, the assumption was made that as MVD increases, OED will also increase. This assumption was verified as the line of best fit demonstrated a linear, as opposed to a curvilinear, relationship. A Pearson Product 28

39 Moment Correlation was therefore chosen for statistical analysis between MVD and: 1) progression or nonprogression to OSCC over time and 2) increase in histologic grade over time. Similarly, intra-patient analyses used a Pearson product moment correlation test to determine whether or not MVD increased over time. The progressive group and nonprogressive groups mean follow up was 76.6 months and 87.0 months, respectively. Finally, no significant association was found in the progressive or non-progressive groups between increase in MVD for all vessel sizes and total MVD over time. 29

40 Figure 7. Progressive timeline correlation Progressive Group Pearson Linear Correlation Vessel Density n r r 2 p Total Small vessel Medium vessel Large Vessel Small+Medium Table 2. Correlation between MVD and progression over time 30

41 Figure 8. Nonprogressive timeline correlation Nonprogressive Group Pearson Linear Correlation Vessel Density n r r 2 p Total Small vessel Medium vessel Large Vessel Small+Medium Table 3. Correlation between MVD and nonprogression over time 31

42 A significant association was detected between increase in MVD and histologic grade in the nonprogressed patients. Nonprogressive patient group microscopic specimens ranged from a histologic grade of atypia to moderate-severe OED over a mean follow up period of 87.0 months. A statistically significant association was found between increases in histological grade and total MVD (p=0.041), medium vessel MVD (p=0.034) and large vessels MVD (p=0.007). A significant association, however, was not found between small vessel MVD and histological grade (p=0.168). Figure 9. Nonprogressive association between total MVD and histological grade 32

43 Figure 10. Nonprogressive association between small vessel MVD and histological grade 33

44 Figure 11. Nonprogressive association between medium vessel MVD and histological grade 34

45 Figure 12. Nonprogressive association between large vessel MVD and histological grade Nonprogressive Group Pearson Linear Correlation Vessel Density n r r 2 p Total Small vessel Medium vessel Large Vessel Small+Medium Table 4. Nonprogressive association between MVD and histologic grade In contrast, no MVD-histologic grade association was found in the progressed patients. Patients included in the progressive group ranged from a histological diagnosis of atypia to OSCC over a mean follow up period of 76.6 months. No significant 35

46 association was found between histological grade and total MVD (p=0.697), small vessel MVD (p=0.769), medium vessel MVD (p=0.691) and large vessel MVD (p=0.068). Figure 13. Progressive association between total MVD and histological grade 36

47 Figure 14. Progressive association between small vessel MVD and histological grade 37

48 Figure 15. Progressive association between medium vessel MVD and histological grade 38

49 Figure 16. Progressive association between large vessel MVD and histological grade Progressive Group Pearson Linear Correlation Vessel Density n r r 2 p Total Small vessel Medium vessel Large Vessel Small+Medium Table 5. Progressive association between MVD and histologic grade No significant differences were observed in MVD lumen diameter between the progressed and nonprogressed patients. The mean total vessel, small vessel, medium vessel, and large vessel MVD was analyzed for both the progressive and nonprogressive 39

50 groups. Total vessel MVD (p=0.435), small vessel MVD (p=0.328), medium vessel MVD (p=0.519) and large vessel MVD (p=0.889) showed no significant differences between the progressive and nonprogressive groups. Figure 17. MVD analysis between progressive and nonprogressive groups 40

51 Progressive Vs Non Progressive Test P Total Unpaired t test Small vessel Unpaired t test Medium vessel Unpaired t test Large Vessel Unpaired t test Table 6. MVD analysis between progressive and nonprogressive groups Intrapatient comparisons did not reveal any significant associations between MVD relative to histologic grade and MVD relative to time progression. Intrapatient association of total MVD over time in both the progressive group and nonprogressive group was compared utilizing a Pearson product moment correlation test. No significant association was found within all patients. 41

52 Chapter 4: Discussion and Conclusion OSCC is the sixth most common human cancer worldwide (1). Progression of OED to OSCC has been reported up to 36% however we are currently unable to predict which lesions will undergo malignant transformation (2,3). Ideally, patient management would be based on characteristics established to predict lesional behavior. Conventional management of OED is based on histologic grading into mild, moderate, severe dysplasia and carcinoma in situ (5). Complete excision with close clinical follow-up is standard for moderate or higher grades of OEDs (63). Despite surgically clear margins, local OED recurrences are common (7). Ideally, patient management would be based on characterics established to predict lesional behavior. Although tobacco and alcohol are the best-established risk factors, OSCC tumors also arise in patients that are apparently risk factor-free. The difficulty of predicting biologic behavior has prompted much research into potential biological prognostic and predictive factors. The angiogenic switch is a well-documented phenomenon that occurs in the transformation of many solid tumors. Hannahan and Folkman s review of tumors arising in transgenic mice models found induction of angiogenesis is a discrete component of the tumor phenotype, one that is often activated during the early, pre-neoplastic stages in the development of a tumor (14). Similarly, Arbeit et al. (29) and Coussens et al. (30) utilized transgenic mice multistage model (normal epithelium, hyperplasia, dysplasia, 42