The 5-year survival rate for oral cancer has not improved significantly in the last 3 decades, despite numerous advances in treatment modalities.1, 2 This is largely because oral cancer is usually not diagnosed until later stages. Currently, early detection of oral cancer relies primarily on visual recognition of suspicious lesions and subsequent biopsy examination. Typically, dentists or primary care physicians are the first to examine the mucosal abnormalities of the oral cavity. Because of the diverse training and professional experience in the recognition of suspicious lesions, the detection and treatment of early disease is inconsistent.3 This often results in delayed diagnosis of oral cancers until they reach advanced stages. Even experienced examiners can miss the subtle morphologic changes associated with many early lesions.
Early detection of premalignant lesions and invasive cancer in the oral cavity could be greatly improved through techniques that permit visualization of molecular changes indicative of the neoplastic transformation process. Molecular markers can serve as early indicators of neoplastic transformation or progression of disease,4 and may potentially allow detection of cancer before the development of gross morphologic changes. Targeted contrast agents, which couple an optically active moiety to a probe molecule that binds specifically to tumor-related molecules, have potential to improve early detection in vivo. We have developed a molecular-specific fluorescent contrast agent, consisting of a far-red fluorescent dye coupled to a monoclonal antibody targeted against epidermal growth factor receptor (EGFR) to aid in noninvasive early detection of oral neoplasia in vivo.5 EGFR is overexpressed in a high number of epithelial malignancies, including oral squamous carcinoma. The contrast agent is designed for topical application in vivo, followed by real time optical detection.
Optical techniques can quantitatively assess the presence of bound contrast agent noninvasively in near real-time, in contrast to current diagnostic techniques of invasive biopsy and delayed microscopic assessment. The optical signature of the contrast agent could be easily detected and quantified in vivo using spectroscopy,6, 7, 8, 9, 10, 11, 12 low resolution 2D imaging13, 14, 15, 16 or high resolution 3D confocal microscopy.17, 18, 19, 20, 21, 22, 23, 24, 25 Several studies have previously investigated the use of in vivo fluorescence spectroscopy based on autofluorescence to diagnose malignancy in the oral cavity.6, 7, 8, 9, 10, 11, 12 These spectroscopic techniques could easily be adapted for use with an exogenous contrast agent by modifying the excitation source and emission filters. Further, improved discrimination between normal and abnormal tissue may be possible using the contrast agent for several reasons. The longer excitation wavelength of the fluorescent dye avoids interference from many of the molecules present in tissue that absorb and scatter light at the optimal wavelengths for autofluorescence (300–440 nm).7, 8, 10, 11, 12, 13 Using a dye that fluoresces at a longer wavelength also allows for increased penetration depth. Also, the intensity of fluorescence emitted by the contrast agent is higher than that of endogenous fluorophores, resulting in a stronger optical signal. Technology is rapidly becoming available to image tissue labeled with this type of contrast agent in a clinical setting. In vivo confocal imaging has been achieved using fiber-based probes17, 20, 21, 22, 23, 25 and fluorescence in vivo confocal microscopes are also commercially available from OptiScan Imaging Limited (Victoria, Australia) and Mauna Kea Technologies (Paris, France).
The optically active contrast agent is targeted against EGFR, a promising molecular target for detection of oral cancer. Although EGFR is present in proliferating cells in normal tissue,26 there is marked over-expression during progression to dysplasia and cancer.27 A number of studies have estimated that 50–98% of tumors in the oral cavity overexpress EGFR.28, 29, 30, 31, 32 Investigations in head and neck squamous cell carcinoma (HNSCC) patients revealed that tumor EGFR levels correlate to tumor size30, 31 and that mean values of EGFR expression increase with tumor size and clinical stage.32 In a correlative study, EGFR expression was also a strong prognostic indicator for overall disease-free survival in patients with advanced HNSCC enrolled in a radiotherapy study and was highly predictive for local–regional relapse but not distant metastases.33 Other studies, however, have reported no significant correlation between EGFR levels and tumor staging.28, 32, 33
Research groups have worked to develop targeted fluorescent contrast agents to detect EGFR expression in vivo. Soukos et al. used an anti-EGFR antibody conjugated to a diagnostic fluorescent dye or to a photoactive molecule for imaging and targeting tumors in the Syrian golden hamster cheek pouch model.34 Ke et al. demonstrated the use of a near-infrared dye conjugated to epidermal growth factor (EGF) for in vivo imaging of mammary tumors in mice.35 Previously, we demonstrated the efficacy of a far-red fluorescent contrast agent targeted against EGFR in squamous carcinoma cells of the buccal mucosa in monolayer culture and in 500-μm thick multilayer tissue constructs.5 However, these efforts did not address the efficacy of this contrast agent in intact, living tissue from human patients. In our study, we evaluate the efficacy of this contrast agent for detection of early neoplasia using organ cultures of normal and neoplastic human oral tissue. Fresh tissue sections from paired biopsies obtained from clinically normal and abnormal appearing oral mucosa were exposed to contrast agent, rinsed and imaged using quantitative fluorescence confocal microscopy to determine the diagnostic potential of the agent. Simple cytotoxicity assays and experiments comparing labeling at ambient and physiologic temperature were also performed.
BSA, bovine serum albumin; DMSO, dimethyl sulfoxide; EGFR, epidermal growth factor receptor; H&E, hematoxylin and eosin; HNSCC, head and neck squamous cell carcinoma; IHC, immunohistochemistry; MFI, mean fluorescence intensity; PBS, phosphate buffered saline; PVP, polyvinylpyrrolidone; SCC, squamous cell carcinoma.
Material and methods
Fresh tissue slices
Paired biopsies of clinically abnormal and clinically normal appearing areas of oral mucosa were obtained from consenting patients at the University of Texas M.D. Anderson Cancer Center. The clinical protocols were reviewed and approved by the Institutional Review Boards at the University of Texas M.D. Anderson Cancer Center and the University of Texas at Austin. Biopsies were immediately placed and remained in chilled culture media (phenol-red free, Dulbecco's modified essential media (DMEM)/F12 with high glucose, Invitrogen, Carlsbad, CA) until they were sectioned into 200- to 400-μm thick sections using a Krumdieck tissue slicer (MD 1000-A1, Alabama Research and Development, Munford, AL), which is designed to cut fresh tissue with minimal damage.
The contrast agent is described in detail in Hsu et al.,5 but is briefly reviewed here. The antibody used in the contrast agent was a biotinylated mouse monoclonal anti-human EGFR (clone 111.6, 200 μg/ml, LabVision NeoMarkers, Fremont, CA) and was used at a 1:10 dilution. Alexa Fluor® 660 streptavidin (Molecular Probes, Eugene, OR) was used as the far-red fluorescent dye. For use in labeling, it was diluted from a 1 mg/ml stock to 250 μg/ml in phosphate buffered saline (PBS) and used at a 1:8 dilution.
After slicing, biopsies were labeled using the contrast agent. The tissue slices were first blocked for endogenous biotin using an Avidin/Biotin blocking kit (Vector Laboratories, Burlinghame, CA) in a 5% bovine serum albumin (BSA) solution in PBS. After blocking, the tissue slices were incubated with either the biotinylated primary EGFR antibody or, as a control, a biotinylated normal mouse IgG (Vector Laboratories) for 30 min at room temperature, with constant shaking. Samples were next incubated with the Alexa Fluor® 660 streptavidin for 30 min at room temperature, with constant shaking. All dilutions were prepared in a 1.25% BSA solution in PBS. Between all changes of solution, the tissue slices were washed in PBS for 5 min 2 times with constant shaking. Finally, the fluorescence of each tissue slice was imaged using confocal microscopy.
Confocal fluorescence images of the stained, fresh tissue slices were obtained using an inverted Lecia TCS 4D laser scanning confocal microscope, equipped with a Kr/Ar laser providing excitation at 647 nm and a 665 nm long-pass filter. Images were acquired with a 40× oil-immersion objective with a numerical aperture of 1.0 and working distance of 80 μm. The field of view of the objective is ∼0.4 mm. The laser power for image acquisition varied between 3 and 30 μW. A stack of images was obtained at different depths beneath the sample surface, with a z-step of 1.5–3 μm and 5–30 images per stack. The contrast, brightness, line averaging, integration time and pinhole size remained constant for all acquisitions.
The epithelium of the fresh tissue slice sometimes did not occupy the entire field of view. To facilitate quantitative comparison of fluorescent images, MATLAB (the MathWorks, Natick, MA) code was written to identify the region of the image containing epithelium. The code identified the fluorescently-labeled edges of the epithelium in the images by applying the Laplacian of Gaussian edge detection method to the image, after morphological filtering using a structuring element; a mask was then created, which assigned the value of 1 to each pixel enclosed by the edge and 0 to each pixel outside this area. The fluorescent images were then multiplied by the mask so that the resulting image contained information only from the epithelium. To adjust for variations in the illumination power, each image was divided by the laser power at the time of acquisition. Finally, to correct for small differences in the concentration of the Alexa Fluor® 660 dye used to make the stock solution for labeling samples from the first 3 patients, a scaling factor was determined based on the ratio of the optical density of stock dye solution used for the first 3 patients to that used for the remainder of the patients. The fluorescence images acquired for the first 3 patients were scaled by this correction factor.
The mean fluorescence intensity (MFI) for the epithelial area of each image in each stack was then calculated. In cases where epithelial cells occupied the entire field of view, the MFI was calculated for the entire image in this manner, without using edge detection. The MFI of the entire epithelium was used to determine if there was a correlation between fluorescence and pathological diagnosis. In addition, in each image the epithelium was divided into 3 layers, basal, intermediate and surface, to determine if the gradient of the MFI would provide a more effective means for diagnosis. There is variation in the thickness of normal epithelial tissue, depending on the exact site in the oral cavity, ranging from the thin layer of the floor of the mouth (190 μm ± 40 μm) to the thick layer of the buccal mucosa (580 μm ± 90 μm).36 If an image stack contained the basement membrane or was within ∼50 μm of the basement membrane, it was classified as basal. If it contained the superficial epithelium, epithelial-surface transition or was within 50 μm of the surface, it was classified as surface. If no basement membrane or obvious superficial epithelium was present, it was classified as intermediate. In cases of invasive carcinoma, cells at the surface of the tissue slice or within 50 μm of the surface were classified as surface. Cells that were obviously embedded in the stroma were classified as basal. If the cells were not near the surface or obviously invaded into the stroma, the image stack was classified as intermediate.
Biopsies showing hyperkeratosis and hyperplasia were considered normal, as these are considered normal findings in the oral cavity. Occasionally, a single section from a biopsy had multiple abnormal diagnoses (e.g. mild dysplasia with focal severe dysplasia), and these were classified according to the most severe of the diagnoses. Similarly, different slices from the same biopsy sometimes had different abnormal diagnoses, and the biopsy was again classified according to the most severe diagnosis. However, if a biopsy contained both benign and neoplastic tissue, it was placed into a separate category designated as containing both normal and neoplastic tissue. Two categories for mixed normal and neoplastic tissue were defined: mixed normal and dysplasia and mixed normal and cancer. For example, a biopsy containing invasive carcinoma with overlying normal epithelium was placed into the separate category of mixed normal and cancer.
Following imaging, tissue slices were fixed and submitted for routine histology (hematoxylin and eosin (H&E)) and EGFR immunohistochemistry (IHC). H&E stained sections were examined by 2 board-certified pathologists (AKE and MDW) and a consensus diagnosis was determined using standard histopathological criteria as follows: normal, hyperplasia, hyperkeratosis, dysplasia (mild, moderate, or severe) and in situ or invasive cancer (well-, moderate-, or poorly-differentiated). EGFR IHC stained slides were evaluated using light microscopy and the EGFR staining pattern was compared with fluorescence images.
Statistical analysis was performed using a 2-sided Wilcoxon rank sum test.
Eleven day incubation with Alexa Fluor® 660 streptavidin.
A 2-day cytotoxicity test of the Alexa Fluor® 660 streptavidin was previously performed, as described in Hsu et al.5; after 2 days, the presence of the dye showed very little effect on cell viability at all dye concentrations. To further test the cytotoxicity of Alexa Fluor® 660 streptavidin over a longer period of time, a longer exposure was performed here. A single plate of SqCC/Y1 cells, a squamous cell carcinoma of the buccal mucosa (kindly provided by Dr. Reuben Lotan, at the University of Texas M.D. Anderson Cancer Center), was harvested and used to plate a 6-well tissue culture plate at a 1:8 split, with ∼200,000 cells/well. After 1 day, DMEM/F12 media (Invitrogen) with 5% fetal bovine serum (FBS), penicillin, streptomycin and glutamine (Invitrogen) with dye dilutions of 1:2, 1:5, 1:10, 1:20, 1:50 from a 250 μg/ml stock, or no dye was added. At days 3 and 7, the cells were collected and a viability count performed. The cells were replated at a 1:4 split in media with the same dye concentration. On day 11, the cells were collected and a viability count performed.
Recovery after exposure to Alexa Flour® 660 streptavidin
A final cytotoxicity test was performed to assess the effects of long-term exposure to the dye without harvesting the cells and to assess the recovery of the cells after exposure to the dye. A single plate of SqCC/Y1 cells was harvested and used to plate a 6-well tissue culture plate at a low density of ∼25,000 cells/well. After setting up for 1 day, the media with dye dilutions of 1:2, 1:5, 1:10, 1:20, 1:50 from a 250 μg/ml stock, or no dye was added. On day 7, the cells were collected, a viability count performed, and the cells replated at a 1:4 split in media with no dye. On day 8, after replating in media with no dye, the cells were collected and a viability count performed.
Performance of contrast agent at physiological temperature
MDA-MB-468 (an adenocarcinoma of the mammary gland known to overexpress EGFR at ∼1 × 106 receptors/cell, American Type Culture Collection, Manassas, VA) and SqCC/Y1 cells were grown in monolayer, as described in Hsu et al.5 Labeling was also performed as described in Hsu et al.5 similar to the labeling procedure followed for the fresh tissue slices, at both room temperature and at 37°C, to evaluate whether increased internalization at physiological temperatures would affect targeting of the contrast agent. Cells were imaged similarly to the fresh tissue slices.
Overview of patients
A total of 17 subjects were evaluated in our study. Table I provides the site of biopsy and histological diagnosis of the clinically abnormal and normal appearing biopsies. One of the clinically abnormal biopsies was found to be histologically normal, which reduced the number of overall abnormal samples in the study. The sensitivity and specificity of visual recognition were 76.2% and 91.67%, respectively, in our study. The number of biopsies by histological diagnosis is listed in Table II.
Table I. Patients and Biopsy Characteristics of Patients with HNSCC
FOM, floor of mouth; BM, buccal mucosa; diff, differentiated; mod, moderate. No diagnosis occurs when there is no epithelium in the sample and a pathological diagnosis cannot be made. A site in the center of the “site” column indicates that both the abnormal and normal biopsy were from the same site.
Well diff. cancer
Hyperkeratosis, mild-moderate dysplasia
Mod. dysplasia, moderate-poor diff. cancer
Hyperplaisa, hyperkeratosis, mild dysplasia
Invasive moderate diff. cancer
BM anterior margin
Right lateral tongue
Hyperplasia, hyperkeratosis, mild dysplasia
Left lateral tongue
Right lateral tongue
Hyperkeratosis, focal mild dysplasia
Base of tongue
Invasive moderate diff. cancer
Hyperplasia, hyperkeratosis, moderate dysplasia
Hyperplasia, hyperkeratosis, focal mild dysplasia
Left lateral tongue
Invasive moderate diff. cancer
Focal mild dysplasia
Normal, moderate diff. cancer
Hyperkeratosis, mild dysplasia
Left buccal mucosa
Anterior BM (lower lip)
Hyperkeratosis, mild dysplasia
Right lateral tongue
Hyperplasia, hyperkeratosis, mild dysplasia
Gingiva near buccal mucosa
Normal, mod. dysplasia, invasive mod. diff cancer
Invasive well diff. cancer
Table II. Breakdown of Biopsy Samples by Histological Diagnosis
Number of biopsy samples
Hyperkeratosis and hyperplasia
Moderate dysplasia and moderate to poorly-differentiated cancer
Normal and cancer
Hyperplasia, hyperkeratosis, and dysplasia
Fluorescence confocal imaging
Figure 1 shows fluorescence confocal, H&E and EGFR IHC images from the normal and abnormal biopsies from patient 4. The clinically abnormal biopsy was from the retromolar trigone and diagnosed as moderate dysplasia with underlying foci of moderate to poorly differentiated cancer. The clinically normal biopsy was from the pharyngeal wall and diagnosed as hyperplasia and hyperkeratosis. Fluorescent staining is seen throughout the entire epithelium of the abnormal biopsy (a representative field is shown in Fig. 1(a)). The epithelial-stromal boundary of the normal biopsy is shown in Figure 1(b); fluorescence is evident near the basement membrane of the epithelium, but is not present in the upper layers. The corresponding H&E stained sections are shown in Figures 1(c) and 1(d) and EGFR IHC stained sections are shown in Figures 1(e) and 1(f) for the abnormal and normal biopsies respectively. EGFR staining shows strong reactivity throughout the abnormal squamous epithelium and corresponds to reactivity in the fluorescence confocal image of the same region. In the normal biopsy, strong EGFR staining is limited to the basal layer and progressively decreases in the differentiated surface layers. This also corresponds well with what was observed in confocal images of the normal biopsy.
Images from the surface layer and basal layer of the clinically normal biopsy from patient 6 are shown in Figure 2. The normal biopsy was taken from the mandibular gingiva and was diagnosed as hyperplasia and hyperkeratosis. Figure 2(a) is the fluorescence confocal image from the surface layer, where very little fluorescence is evident. The fluorescence confocal image from the basal layer is shown in Figure 2(b), with the white line indicating the basement membrane. As is evident from these images, fluorescence is strong near the basement membrane and decreased in the surface layer, which correlates well to expected EGFR expression in normal epithelium. The H&E section is shown in Figure 2(c).
Images from the surface and basal layer from the clinically abnormal biopsy from patient 4 are shown in Figure 3 to illustrate the difference in fluorescence throughout the epithelium in abnormal squamous mucosa compared to a normal squamous mucosa as seen in Figure 2. The clinically abnormal biopsy was diagnosed as moderate dysplasia with underlying foci of moderate to poorly-differentiated carcinoma. Figure 3(a) is the fluorescence confocal image from the surface layer and Figure 3(b) is from the basal layer, with the basement membrane indicated by a white line. Fluorescence is strong throughout the entire epithelium, in contrast to the normal tissue shown in Figure 2, where very little fluorescence was evident in the surface layer. The corresponding H&E section is shown in Figure 3(c).
Figure 4 shows fluorescence and stained images from a clinically abnormal biopsy with a mixed normal and neoplastic histological diagnosis; 2 sections were obtained from this biopsy, one was diagnosed as hyperplasia and hyperkeratosis, and the other was diagnosed as moderate dysplasia. Figure 4 also shows fluorescence and stained images from the corresponding clinically normal biopsy, which was also histologically normal. Figures 4(a)–4(c) show the fluorescence confocal, H&E and EGFR IHC images from the clinically abnormal biopsy that was histologically normal. Strong EGFR labeling near the basement membrane is evident in both the IHC and fluorescence images. Figures 4(d)–4(f) show the fluorescence confocal, H&E and EGFR IHC images of the clinically abnormal tissue biopsy that was diagnosed as mild to moderate dysplasia and shows strong fluorescence near the basement membrane (Fig. 4(d)). Thickening of the epithelial layer with dysplastic cells is evident in the H&E image (Fig. 4(e)). These same cells also stain for EGFR by IHC (Fig. 4(f)). Fluorescence in the histologically abnormal sample (Fig. 4(d)) is slightly stronger than in the corresponding histologically normal sample (Fig. 4(a)). Figures 4(g)–4(i) show the fluorescence confocal, H&E and EGFR IHC images for the clinically and histologically normal biopsy. Fluorescence is again evident in the basal layers (Fig. 4(g)). The relative amount of EGFR staining in the IHC image (Fig. 4(i)) is similar to that in Figure 4(c). The histologically abnormal sample (Fig. 4(d)) has slightly brighter fluorescence than both of the 2 histologically normal samples (Figs. 4(a) and 4(g)), which have similar levels of fluorescence.
Diagnosis using mean fluorescence intensity
We observed that the MFI of normal tissues varied substantially from patient to patient, with a difference as great as 8.5 times; however, the MFI for the abnormal tissues were higher than those of the corresponding normal tissues in matched biopsies from a single patient. To minimize intersubject variation in baseline EGFR expression, we calculated the ratio of the MFI of all image stacks from the clinically abnormal and normal biopsies from each patient. The average MFI ratio for each diagnostic category is shown in Figure 5. In cases where both the clinically abnormal and clinically normal biopsy were histologically normal, the ratio was still calculated; the ratio was not calculated for patient 3 because no diagnosis could be obtained for the clinically normal biopsy. There is a general trend for increasing MFI ratio values with more severe diagnosis. We rejected the null hypothesis that the MFI ratio of samples diagnosed as moderate dysplasia or cancer was equal to that of those diagnosed as normal, hyperplasia, hyperkeratosis, mild dysplasia or mixed normal and mild dysplasia (p value = 0.0336). The ratio for samples diagnosed as cancer was typically greater than 2, except for 1 biopsy with well-differentiated cancer (patient 17) with a ratio slightly less than 2. Samples with mixed diagnosis of normal and cancer (patients 12 and 16) also had a ratio less than 2. The inclusion of histologically normal tissue in the abnormal biopsy may have resulted in the lower ratio as the field of view in the fluorescence image may have included predominantly normal tissue.
The data in Figure 5 are based on the MFI calculated from the entire epithelial region imaged. However, from both confocal fluorescence images and EGFR IHC stained sections, it is evident that EGFR expression varies throughout the epithelium and this variation differs for normal and cancerous tissues. For example, basal cells in normal tissue show strong EGFR expression by IHC and fluorescence, while cells in the surface layers show very low EGFR staining. In contrast, EGFR staining was strong throughout the epithelium of samples with dysplasia and cancer. We examined whether the MFI of the surface, intermediate or basal layers provided discriminative capability. The best separation was observed for the MFI of the surface layer as demonstrated in Figure 6, where the average of the MFI for each diagnostic category is plotted. The clinically abnormal biopsies from patients 8, 9 and 12 did not have any images from the surface layer and were not included in the average calculations. Patient 5 is also excluded because neither the clinically abnormal nor normal biopsies had images from the surface layer. Again, there is a general trend of increased surface MFI with increasing severity of diagnosis. We rejected the null hypothesis that the MFI of the surface layer of samples diagnosed as normal, hyperplasia or hyperkeratosis was equal to that of samples of moderate dysplasia or cancer (p value = 3.18 × 10−4). Generally, the MFI of the surface layer was greater than 6 for the samples diagnosed as moderate dysplasia or cancer. The abnormal biopsy from patient 17, which had a histological diagnosis of well-differentiated cancer, had a surface MFI of slightly less than 6. The abnormal biopsies with mixed histological normal mucosa and cancer or dysplasia also had a surface MFI of less than 6. In addition, all the abnormal biopsies with the diagnosis of mild dysplasia had a surface MFI less than 6.
The extended cytotoxicity assay demonstrated a small loss in cell viability associated with exposure to dye. At day 3, all samples had a viability of 94% or higher and a relatively comparable cell count. After replating, however, cells grown with the dye did not reattach as well as cells grown with no dye. At day 7, this was reflected in the overall cell count, with cells grown in the presence of dye having a cell count 2–7 times less than cells grown with no dye. The viability of cells grown in the presence of dye was 80% or higher when compared with 95% for the control cells. Similar results were observed in the experiment where cells were split at day 7 and viability was assessed on day 11.
SqCC/Y1 cells were also grown in the presence of the dye over a 7-day period without splitting to evaluate the effects of the dye without the need for cell reattachment. At day 4, all cells grew well with normal morphology. Between days 5 and 7, however, the cells in the presence of dye began to exhibit spindle cell morphology and less contact rather than confluent growth. In addition, there was marked increase in cell death between days 5 and 7. This was reflected in the cell viability count where cells grown in the presence of dye had a cell count about 2 times lower than that of the control cells. Overall, the viability of cells grown in the presence of dye was generally lower as well. To test the ability of cells to recover after exposure to the dye, these cells were split and grown in their normal media. The cells that had been grown with dye did not initially reattach well, which resulted in a low overall cell count after 8 days of recovery compared with that of the control cells. The overall cell viability increased, however, with the 1:2 dilution having the lowest viability at 88% and all other cells having a viability of 94% or higher.
Performance of the contrast agent at physiologic temperature
The fluorescence of both cell lines labeled with the EGFR contrast agent at room temperature and at 37°C was similar. Cells labeled at 37°C showed slightly more internalization of the contrast agent than those labeled at room temperature as evidenced by small spots of intense fluorescence in the cytoplasm (images not shown). However, the periphery of the cell showed continuous labeling at both temperatures. In addition, the IgG control did not show any increase in fluorescence when labeling was performed at either room temperature or 37°C (images not shown).
The epidermal growth factor receptor has great promise for use as a diagnostic target in epithelial cancer: 50–98% of oral cancers overexpress EGFR. Studies correlating EGFR expression with cancer staging and progression generally use traditional molecular biology or protein biochemistry techniques, such as radiolabeled ligand assays and enzyme-linked immunosorbent assays (ELISA), which involve homogenization of the tissue sample to measure levels of EGFR expression.28, 29, 30, 31, 32 Immunohistochemical assessment of EGFR expression in fixed, paraffin-embedded specimens has also been used with some success.33, 37, 38, 39, 40, 41, 42, 43 However, real time, in vivo evaluation of EGFR expression would yield important clinical advantages for diagnosis and treatment monitoring of epithelial lesions. In our study, we localized EGFR expression in whole 200-μm thick fresh tissue sections, as an initial step toward this goal of in vivo molecular imaging.
Our results show that fluorescence labeling of fresh tissue using a molecular-specific contrast agent correlates well with standard EGFR IHC. The extent and level of EGFR expression differed between normal and abnormal tissue from matched biopsies from individual patients. In normal epithelium, basal cells are constantly dividing and have high EGFR expression; when these cells migrate upward and differentiate, proliferation decreases and EGFR expression is lower, as was seen in our results. During carcinogenesis, basal cells fail to differentiate and continually proliferate throughout the epithelium, rather than in just the basal layer. As a result, EGFR is highly expressed throughout the epithelium in high-grade dysplasia.
Overall, we found that the MFI from abnormal tissue was higher than that from histologically normal tissue within the same subject. Using the ratio of the MFI of the clinically abnormal to clinically normal specimens from a single patient eliminates intersubject variations. This is important because EGFR levels may vary from person to person, depending on other factors such as tobacco and alcohol use. With increasing severity of diagnosis, there was a general increasing trend in the ratio of the MFI of the clinically abnormal specimen to the MFI of the clinically normal specimen.
The MFI from moderate dysplasia and cancer specimens was generally much higher than that from normal, hyperplasia or hyperkeratosis when the MFI was measured from only the surface layer of tissue. This is expected to be due to the decrease in EGFR expression level with keratinocyte maturation and differentiation. Mild dysplasia, where abnormal cells that often overexpress EGFR only occupy the lower one-third of the epithelium, also had a lower surface MFI than moderate dysplasia or cancer. Biopsies with a mixed histological diagnosis of normal and cancer or normal and dysplasia also had a lower surface MFI than moderate dysplasia and cancer. This may be due to the presence of normal tissue overlying invasive cancer cells or partial volume averaging of cancerous and normal microregions, which would reduce the overall surface MFI of the clinically abnormal biopsy.
Neither the ratio of the abnormal or normal MFI or the surface MFI could separate normal, hyperkeratosis and hyperplasia from mild dysplasia with any statistical significance. Clinically, however, mild dysplasia is treated the same as a normal diagnosis because it is not highly indicative of early malignant progression and has the potential to resolve without treatment.3
It may be more clinically relevant for diagnostic purposes to measure EGFR expression within the surface layers of the epithelium only. Mild dysplasia only affects the lower one-third of the epithelium and is generally not considered clinically significant. In contrast, severe dysplasia and carcinoma in situ, in which the entire thickness of the epithelium is affected, are strongly associated with malignant progression and require complete resection.3 To minimize overdiagnosis of early dysplasia, which may result in unnecessary treatment, a confidence level should be established at the moderate to severe dysplasia stages. Because the upper and middle thirds of the epithelium contain abnormal cells in moderate to severe dysplasia, optical targeting of the surface layers should be more specific for advanced dysplasia.
The inability of MFI to distinguish dysplasia and cancer from the normal biopsies in some cases could be due to the small biopsy sizes and presence of mixed normal and abnormal areas, which would lead to a lower overall MFI. In oral neoplasia, often an area of invasive cancer can be found next to normal tissue and dysplasia in a very small area (one high power field), as was seen with 3 of the patients in the study. In these cases, histopathologic review confirmed presence of both normal and malignant tissue within a section and the abnormal biopsy did not classify well. This suggests that a lower overall MFI could have resulted from measuring the normal areas within the biopsy as well as the normal areas.
Clinical diagnosis using MFI could be achieved in vivo in near real-time using topical applications of the contrast agent in conjunction with a number of optical techniques. A simple, inexpensive spectroscopy device could determine the relative levels of EGFR expression by measuring the intensity of the optical signal from the fluorescence contrast agent. A multi-spectral digital microscope could be used for 2D surface imaging with spectral information, allowing for visualization of larger regions.16 In addition, fluorescence confocal imaging could be used to detect the alteration in the pattern of EGFR expression, with strong signal from the contrast agent present throughout the epithelium of abnormal tissue, but limited to the basal layer of normal tissue.
Optical techniques in vivo may be further enhanced by using the fluorescence intensity of the contrast agent in the surface layers of tissue as a diagnostic criterion. This can be achieved with depth-dependent spectroscopy measurements using an angled illumination geometry,44, 45 allowing measurements of fluorescence intensity from only the surface layer. In in vivo microscopy, imaging is performed en face, which may make imaging from the surface layers more straightforward than imaging deeper layers. In en face imaging, the layers closest to the surface would be the first to be imaged and also be the easiest to image. Furthermore, focusing on signal from the surface layer may avoid many problems due to light penetration limitations into the tissue.
To fully utilize the advantages of optical interrogation techniques of the contrast agent, in vivo application of the contrast agent will be necessary. Before the contrast agent can be used in vivo, a number of issues must be considered. To begin addressing these issues, the potential cytotoxicity and immunogenic effects of components of the contrast agent and the performance of the contrast agent under physiologic conditions were investigated.
Our results indicate that over a short period of time, the Alexa Fluor® 660 dye has little effect on cell viability. As the incubation time was increased, however, precipitous decrease in cell growth and viability ensued. These observations provide a general picture of the effects of the Alexa Fluor® 660 dye on cell growth and viability. Recovery experiments demonstrated that cells which did survive exposure to the dye eventually recovered and were able to grow normally after removal of the dye. Further toxicity studies should include animal testing to further assess the effects of the Alexa Fluor® 660 streptavidin dye. While detrimental effects of the Alexa Fluor® 660 streptavidin dye are evident in cells exposed for longer than 4 days, this is considerably longer than the ∼30 min exposure that would be necessary for in vivo labeling.
In vivo labeling will also require the contrast agent be exposed to physiologic temperatures, which could affect the targeting of the contrast agent. Increased temperature will increase the rate of receptor internalization and could result in decreased labeling, if the receptor is internalized before targeting by the contrast agent, or result in increased intracellular fluorescence and decreased membrane-associated fluorescence, if the receptor is internalized after targeting by the contrast agent. Since previous in vitro and ex vivo experiments were performed at room temperature, the effects of labeling at 37°C were evaluated. Very few differences were seen in the labeling at room temperature and at 37°C, though bright spots of fluorescence were more evident in the cells labeled at 37°C than at room temperature. These spots could represent vesicles of receptors that have been endocytosed by the cell or dense areas of receptors that are beginning the endocytosis process. However, there is not an associated decrease in labeling of the cell membrane. The lack of labeling at 37°C with the IgG control indicates that the cell is not non-specifically internalizing the Alexa Fluor® 660 dye, which would result in an increased background. These findings are important because they imply that there would be no decrease in fluorescence intensity or increase in non-specific background fluorescence with in vivo labeling.
Considerations regarding the immunogenic effects of the antibody and in vivo use of the biotin-(strept)avidin system have previously been discussed in greater depth,5 but will be briefly reviewed here. Immunogenic effects of the antibody in the contrast agent can be minimized through the use of low antibody concentrations. In comparison to EGFR antibodies currently in clinical trials for therapeutic use, such as Cetuximab (ImClone Systems, Somerville, NJ), the concentration of EGFR antibodies used here for diagnostic imaging are considerably lesser than the amounts administered for therapeutic reasons with tolerable side effects. Use of the biotion-(strept)avidin system in vivo is not unprecedented, as it has been applied in a number of in vivo radio-applications, including radioimmunodetection and therapy of cancer in humans46 and immunotargeting of cytotoxic molecules and cells to tumor cells.47, 48, 49, 50, 51
An additional consideration for in vivo use of the contrast agent is method of delivery. We previously explored methods to administer the contrast agent topically, rather than through i.v. injection.5 In conjunction with permeability-enhancing agents, such as dimethyl sulfoxide (DMSO) and polyvinylpyrrolidone (PVP), we were able to demonstrate labeling throughout 500-μm thick multilayer tissue constructs formed from oral squamous carcinoma cells. Both DMSO and PVP have been used in drug formulations approved by the FDA. We chose to first demonstrate the efficacy of the contrast agent using fresh tissue slices, which would be easier to evaluate, before moving to topical administration in whole biopsies, which is the next step in our study.
Clearly, there are certain issues that must be addressed before using the contrast agent in vivo. However, it is evident that in vivo use of the contrast agent has potential to yield important clinical advantages for noninvasive, early detection and molecular characterization of oral mucosa through depth-dependent optical interrogation using fluorescence spectroscopy or confocal microscopy. Further, the contrast agent could be used to monitor response to molecularly-targeted therapies against EGFR. Another possible application of the contrast agent would be to identify genetically damaged tissue before there is any demonstrable phenotypic change by light microscopy; however, before beginning genetic analysis of biopsy samples, we felt that the first step to determine applicability was to correlate our experimental findings with the gold standard of histopathology. In addition, although we have concentrated on oral precancer and cancer to assess clinical potential for this agent, it is likely that there will be similar clinical applications for other epithelial malignancies, including cervical and lung carcinoma.
The authors would like to thank Brette Luck, Jennifer Ming and Wendy Kuo for their work on the MATLAB code, Vivian Mack and Saima Ghori for assistance with tissue culture, and Dr. Valen E. Johnson for guidance regarding the statistical analysis.