Content uploaded by Rebecca Kortum
Author content
All content in this area was uploaded by Rebecca Kortum on May 04, 2015
Content may be subject to copyright.
Understanding the Biological Basis of Autofluorescence Imaging
for Oral Cancer Detection: High-Resolution Fluorescence
Microscopy in Viable Tissue
Ina Pavlova
1
, Michelle Williams
2
, Adel El-Naggar
2
, Rebecca Richards-Kortum
4
, and Ann
Gillenwater
3
1
Department of Biomedical Engineering, The University of Texas at Austin, Austin, Texas
2
Pathology, The University of Texas M. D. Anderson Cancer Center
3
Head and Neck Surgery, The University of Texas M. D. Anderson Cancer Center
4
Department of Bioengineering, Rice University, Houston, Texas
Abstract
Purpose—Autofluorescence imaging is increasingly used to noninvasively identify neoplastic oral
cavity lesions. Improving the diagnostic accuracy of these techniques requires a better understanding
of the biological basis for optical changes associated with neoplastic transformation in oral tissue.
Experimental Design—A total of 49 oral biopsies were considered in this study. The
autofluorescence patterns of viable normal, benign, and neoplastic oral tissue were imaged using
high-resolution confocal fluorescence microscopy.
Results—The autofluorescence properties of oral tissue vary significantly based on anatomic site
and pathologic diagnosis. In normal oral tissue, most of the epithelial autofluorescence originates
from the cytoplasm of cells in the basal and intermediate regions, whereas structural fibers are
responsible for most of the stromal fluorescence. A strongly fluorescent superficial layer was
observed in tissues from the palate and the gingiva, which contrasts with the weakly fluorescent
superficial layer found in other oral sites. Upon UV excitation, benign inflammation shows decreased
epithelial fluorescence, whereas dysplasia displays increased epithelial fluorescence compared with
normal oral tissue. Stromal fluorescence in both benign inflammation and dysplasia drops
significantly at UV and 488 nm excitation.
Conclusion—Imaging oral lesions with optical devices/probes that sample mostly stromal
fluorescence may result in a similar loss of fluorescence intensity and may fail to distinguish benign
from precancerous lesions. Improved diagnostic accuracy may be achieved by designing optical
probes/devices that distinguish epithelial fluorescence from stromal fluorescence and by using
excitation wavelengths in the UV range.
Oral cancer is one of the most common malignancies worldwide, and carries one of the lowest
overall survival rates (1,2). Despite the easy accessibility of the oral cavity to examination,
most patients present with advanced disease, when treatment is associated with higher
morbidity, more expense, and less success than earlier interventions. Early detection of oral
cancer can greatly improve treatment outcomes. Unfortunately, there is no method to
adequately screen and diagnose early oral cancers and precancers because detection still relies
Requests for reprints: Ann Gillenwater, Department of Head and Neck Surgery, The University of Texas M. D. Anderson Cancer Center,
Unit 441, Houston, TX, 77030. Phone: 713-792-8841; Fax: 713-794-4662; agillenw@mdanderson.org.
NIH Public Access
Author Manuscript
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
Published in final edited form as:
Clin Cancer Res. 2008 April 15; 14(8): 2396–2404. doi:10.1158/1078-0432.CCR-07-1609.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
on the clinicians' ability to visually identify subtle neoplastic changes, and to distinguish these
changes from more common inflammatory conditions. Technologic advances are needed to
assist clinical diagnosis of oral cancer.
Autofluorescence imaging has been used successfully to rapidly and noninvasively distinguish
malignant oral lesions from surrounding tissue in several pilot studies (3–5). A low-cost device
for visualization of oral autofluorescence was used to identify high-risk precancerous and
cancerous lesions with 98% sensitivity and 100% specificity based on the loss of fluorescence
in abnormal sites compared with normal tissue (6). This device is now commercially available.
5
Autofluorescence spectroscopy has also emerged as a noninvasive technology for diagnosing
precancers and cancers in several organ sites (7–12). In the oral cavity, several groups used
fluorescence spectroscopy to distinguish oral lesions from normal tissue with high specificity
and sensitivity (ranging from 81% to 100%; refs. 13–18). Despite preliminary clinical evidence
indicating the potential role of fluorescence imaging and spectroscopy for improved detection
of early neoplasia in the oral cavity, there is a limited understanding of the biological basis for
optical changes associated with neoplastic transformation of oral tissue.
The diagnostic potential of fluorescence imaging and spectroscopy lies in the ability to
noninvasively probe alterations in tissue morphology and biochemistry that occur during
malignant progression. Fluorescence in epithelial tissue originates from multiple fluorophores
(molecules that, when excited by light, emit energy in the form of fluorescence) and is
influenced by absorption and scattering as light propagates through the epithelium and stroma.
In the cervix, which is histologically similar to oral tissue in many respects, epithelial
fluorescence originates from the cytoplasm of cells and is linked to the metabolic indicators
reduced nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD),
which increase as dysplasia develops (19–21). Neoplastic progression is also associated with
increased nuclear size and chromatin texture, which leads to increased epithelial scattering
(22,23).
Carcinogenesis involves complex biochemical signaling between the epithelial cells and the
surrounding extracellular matrix (24–26). Subepithelial chronic inflammatory
microenvironments express products that induce angiogenesis and degradation of the
extracellular matrix, which in turn, stimulates the promotion of cancer in the epithelium (27).
Because altered stromal properties may precede epithelial changes during carcinogenesis
(28), understanding the autofluorescence patterns in the stroma and the effect of inflammation
on these patterns may help explain the spectral differences in normal oral mucosa and early
dysplasia. Confocal images and spectroscopy analysis indicate that collagen crosslinks are the
major fluorophore in stroma in the cervix (29). Remodeling of the stroma during cervical
carcinogenesis leads to structural changes in the collagen matrix accompanied by loss of
collagen fluorescence (19) and a decrease in stromal scattering (30). Thus, to harness the full
potential of fluorescence-based diagnosis, it is important to clarify how both epithelial and
stromal alterations in oral tissue contribute to the changes in the overall optical properties
during carcinogenesis.
Epithelial and stromal autofluorescence patterns can be directly visualized using fluorescence
microscopy of viable ex vivo oral tissue. High spatial resolution is necessary to assess variability
in the microscopic origin of autofluorescence within the epithelial or stromal layer. Here, we
qualitatively and quantitatively examined the distribution of autofluorescence at the cellular
level in viable oral tissue using laser scanning confocal fluorescence microscopy. Our first
objective was to characterize the origins of autofluorescence in normal oral mucosa and to
assess how anatomic site variations affect these autofluorescence patterns. Second, we
5
http://www.velscope.com
Pavlova et al. Page 2
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
investigated how inflammation and dysplasia alter the autofluorescence properties of oral
tissue. Third, we compared the autofluorescence patterns of oral cancers with different grades
of differentiation. Understanding the biological basis underlying alteration in autofluorescence
in epithelial and stromal layers during oral carcinogenesis will facilitate the development of
accurate diagnostic algorithms to differentiate normal, benign, precancerous, and cancerous
oral tissue; an important step needed to achieve the full diagnostic potential of this technology.
Materials and Methods
Oral tissue collection and preparation of fresh tissue slices
Clinical protocols were approved by the Institutional Review Boards at The University of Texas
M. D. Anderson Cancer Center, The University of Texas at Austin, and Rice University. A
clinically normal and one or more clinically abnormal biopsy were obtained from each
consenting patient at the M. D. Anderson Cancer Center. Biopsies were immediately stored in
iced phenol-free DMEM (Sigma-Aldrich) and kept there until examination. Transverse tissue
slices ∼200 μm thick were obtained from each fresh biopsy using a Krumdieck tissue slicer
(Alabama Research and Development, Munford, AL). Note that prior to fluorescence imaging,
slices were not fixed or processed in any way. Each unstained and unprocessed tissue slice was
imaged with an inverted Leica SP2 AOBS confocal laser scanning fluorescence microscope
(Leica Microsystems) within 12 h after biopsy collection. Detailed procedures for tissue cutting
and preparation for imaging are described elsewhere (19,21).
Confocal microscopy and image collection
Optical sections from each tissue slice were obtained at both UV and 488 nm excitation using
a 40× oil immersion objective. UV excitation was provided by an argon laser at 351 and 364
nm, and an argon/krypton laser was used for 488 nm excitation. Prior analysis on cervical tissue
(29) has indicated that UV excitation (at 351 and 364 nm) targets predominantly NADH in the
epithelium and collagen fibers in the stroma, whereas 488 nm excitation targets FAD in the
epithelium in addition to stromal fibers. The oil immersion objective had a numerical aperture
of 1.25 and a working distance of 80 μm. Single optical sections of each region of interest were
obtained at a fixed depth of 15 μm beneath the coverslip. The fluorescence signal at UV
excitation was collected from 380 to 500 nm, whereas at 488 nm excitation, fluorescence was
collected from 505 to 650 nm. The field of view for each image was 375 × 375 μm. The lateral
resolution was limited by the pixel size of the detector and was 0.73 μm. The axial step size
varied from 1.0 to 1.2 μm. In order to perform quantitative comparisons, all images were taken
with the same detector settings and corrections for laser power variations were done prior to
analysis.
In tissue slices with well-defined layered structures (normal, inflammation, and dysplasia),
adjacent images were acquired to include the whole thickness of the epithelium and the more
superficial stromal regions down to ∼1 mm in depth. In tissue slices without a defined structure
(invasive carcinoma), several images were taken, starting from what seemed to be the surface
of the tissue slice and ending ∼1 mm below this edge. Adjacent confocal images were tiled
together in order to provide a large-scale mosaic view of each region of interest. Some adjacent
images had a considerable area of overlap, which was cropped off prior to assembly of the
mosaic view. Brightness and contrast was readjusted by the same amount for all confocal
images prior to display.
Histopathology
After imaging, tissue slices were fixed in 10% formalin and prepared for pathologic
examination using standard protocols. H&E-stained sections were obtained from each imaged
tissue slice and reviewed by experienced head and neck pathologists. A pathologic diagnosis
Pavlova et al. Page 3
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
of normal, dysplasia, or cancer was rendered; hyperkeratosis and hyperplasia were treated as
normal for further analysis. In addition, the absence or presence (and degree) of inflammation
was determined.
Image analysis including average fluorescence intensity and redox ratio calculations
An initial qualitative assessment of autofluorescence patterns was done, comparing variations
by anatomic site and pathologic diagnosis. Then, the UV and 488 nm images were overlaid
and examined visually. In normal and precancerous tissue slices, the fluorescence patterns were
examined in the superficial, intermediate, and basal epithelium as well as in the superficial and
deep stroma. Tissue slices diagnosed with invasive cancers displayed a loss of layered
morphology. After visually identifying common morphologic and fluorescence patterns, tumor
images were divided into subregions, including regions with tumor cells and surrounding fiber
matrix.
Quantitative image analysis was done in the following manner: after background subtraction,
the UV and 488 nm images for each tissue slice were overlaid. Several regions containing cells
and stroma were outlined based on the appearance of the overlaid images. The average
grayscale values of all pixels within an outlined region were calculated in order to obtain the
mean fluorescence intensity value (FIV) for each region. These calculations were done for both
the UV and 488 nm images. In addition, the average redox ratio value was calculated for the
cellular regions in each tissue slice by dividing the mean FIV at 488 nm excitation for the
region by the sum of the mean FIV at UV and the mean FIV at 488 nm excitation for the same
region. The redox ratio reflects changes in the concentrations and redox potentials of the
metabolic indicators NADH and FAD, and has been used in previous research to monitor
cellular metabolism (20,21).
Finally, all tissue slices were grouped according to their pathologic diagnosis, and the mean
FIV and redox ratio were calculated for each diagnostic category. For normal, inflammatory,
and dysplastic samples, which retain a well-defined layered morphology, tissue slices were
grouped by both pathologic diagnosis and anatomic site. For tissue slices from cancerous
lesions, samples were grouped by diagnosis regardless of anatomic site. The average FIV was
calculated for cellular and stromal regions by averaging the mean FIV for each tissue slice for
that diagnostic group and anatomic site. The average FIVs were used to compare changes in
the fluorescence characteristics of oral tissue that occur in association with neoplastic
development. In a similar manner, the average redox ratio was obtained for each diagnostic
group.
Results
A total of 49 oral biopsies were obtained in this study. The biopsies were subdivided according
to oral anatomic type and pathologic diagnosis. Oral biopsies from five different anatomic
types were imaged including the tongue, palate, gingiva, buccal mucosa, and the floor of the
mouth. Each biopsy was assigned to the following pathologic diagnostic subcategories: normal
without inflammation (eight tongue biopsies, two palate biopsies, two gingival biopsies, two
buccal mucosa biopsies, and four floor of the mouth biopsies), normal with mild to moderate
inflammation (two tongue biopsies, two gingival biopsies, and three floor of the mouth
biopsies), normal with severe inflammation (three tongue biopsies), dysplasia (six tongue
biopsies and two palate biopsies), well-differentiated carcinoma (three tongue biopsies, one
palate biopsy, and two floor of the mouth biopsies), moderately differentiated carcinoma (five
tongue biopsies and one floor of the mouth biopsy), and poorly differentiated carcinoma (one
palate biopsy).
Pavlova et al. Page 4
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Normal oral mucosa autofluorescence patterns
Figure 1 displays confocal images at UV and 488 nm excitation from a representative normal
tongue tissue slice. At UV excitation, most of the epithelial fluorescence originates from the
cytoplasm of cells occupying roughly the lower two-thirds of the epithelial layer (Fig. 1A). At
488 nm excitation, these same cells have less cytoplasmic fluorescence compared with the
upper one-third of the epithelium (Fig. 1B). Strong stromal autofluorescence at both excitation
wavelengths originates from a dense matrix of structural fibers and does not vary significantly
with depth.
The autofluorescence characteristics of normal oral mucosa from different anatomic sites are
compared in Fig. 2. Oral epithelium often retains a superficial keratin-containing layer which
is highly fluorescent. We observed the presence of this highly fluorescent superficial layer in
epithelia from the palate and gingiva (Fig. 2A and B). In contrast, epithelia from the floor of
the mouth, buccal mucosa, and the tongue (Fig. 2C and D, and Fig. 1) display a weakly
fluorescent superficial layer. Deep in this superficial layer, the autofluorescence patterns of
epithelia in all tissue sites except the gingiva are generally similar. Gingival epithelia, in
contrast to other oral sites such as the tongue (Fig. 1), is dominated by cells that have low
fluorescence at UV excitation but high fluorescence at 488 nm excitation. The diminished
cytoplasmic fluorescence at UV excitation was observed in all gingival samples, although the
number of gingival samples examined was limited.
Autofluorescence patterns in inflammatory and dysplastic oral tongue tissue
Representative fluorescence images of oral tongue lesions diagnosed as normal, nondysplastic
epithelium with severe inflammation (Fig. 3A), and mildly dysplastic epithelium with mild to
moderate inflammation (Fig. 3B) were compared. Mild dysplasia and severe inflammation
could be distinguished based on differences in epithelial fluorescence at UV excitation. A large
decrease in fluorescence at UV excitation is observed in the normal basal epithelium overlying
inflammatory stroma compared with normal noninflammatory tongue epithelium (Fig. 1A). In
contrast, dysplastic epithelium displays a small increase in fluorescence at UV excitation,
compared with normal noninflammatory tongue epithelium. Thus, the representative images
in Fig. 3 indicate that dysplastic epithelium is significantly more fluorescent than the normal
basal epithelium overlying severely inflamed stroma. Images from a nondysplastic floor of the
mouth sample with mild to moderate inflammation in the stroma show a similar loss of
epithelial fluorescence at UV excitation as in the tongue (data not shown).
Stromal areas directly beneath the basement membrane in both the inflammatory and the
dysplastic examples exhibit a large loss in fluorescence at UV and 488 nm excitation. In the
loose stromal matrix evident in both samples, autofluorescence signals originate predominantly
from cells rather than from fibers as seen in normal oral tissue. This shift in origin of stromal
fluorescence from fibers to cells, which is more obvious in the severe inflammation case, seems
to be associated with the influx of inflammatory cells in this area, as confirmed in H&E images
(black arrows in Fig. 3). In the dysplastic example, these stromal changes affect only the region
100 to 200 μm below the basement membrane. In particular, stromal fluorescence and fiber
density decrease predominantly in areas underling the epithelium, whereas autofluorescence
patterns of deeper stroma seem to be similar to normal stroma.
Quantitative analysis of fluorescence in normal, benign, and dysplastic oral tongue slices
Figure 4A and B illustrate how images of tissue slices with well-defined layered morphology
were divided into subregions for quantitative analysis. The superficial epithelial region
includes cells that retain keratin (white area in Fig. 4B). Nonkeratinized epithelial cells
displaying a weak cytoplasmic fluorescence signal at UV excitation are defined as type 1 cells
(green area in Fig. 4B). At 488 nm excitation; type 1 cells display an increased cytoplasmic
Pavlova et al. Page 5
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
fluorescence signal compared with the rest of the nonkeratinized epithelium. Epithelial cells
with a strong cytoplasmic fluorescence signal at UV excitation are defined as type 2 cells (blue
area in Fig. 4B). Stromal region 1 occupies an area ∼100 to 200 μm below the basement
membrane. Stromal region 2 is situated below stromal region 1, and occupies an area that is
∼200 to 500 μm deep. For all tissue slices obtained from the tongue, the mean FIV was
calculated for each region by diagnostic category for slices diagnosed as normal, inflammatory
or dysplastic; results are shown in Fig. 4C and D.
In normal tongue epithelium, type 2 cells display the highest average fluorescence intensity at
UV excitation but the lowest value at 488 nm excitation. The average redox values for type 2
cells are 1.8 times lower than the redox values for type 1 cells (data not shown) indicating an
increased metabolic activity in this region of the epithelium. In all normal tongue samples, type
2 cells occupied more than half of the nonkeratinized part of the epithelium. These data suggest
that a majority of the normal tongue epithelium is occupied by cells with bright UV cytoplasmic
fluorescence and weak 488 nm cytoplasmic fluorescence.
Average fluorescence intensities for type 2 cells at UV excitation (Fig. 4C) illustrate the
differences in epithelial fluorescence between normal, inflammatory, and dysplastic tongue
samples. In dysplasias, fluorescence from type 2 cells increases by a factor of 1.3 on average
compared with the normal tongue. In contrast, inflammatory tongue tissue displays a large drop
in the average fluorescence intensity of type 2 cells, which is more pronounced for samples
with severe inflammation. Thus, type 2 cells in dysplastic lesions exhibit an increase in UV-
excited fluorescence by a factor of >4 compared with lesions with severe inflammation. These
modifications in epithelial fluorescence at UV excitation are not accompanied by significant
changes in the 488 nm excited fluorescence.
Average fluorescence values from stromal region 1 (Fig. 4C and D) reveal that both
inflammatory and dysplastic lesions are characterized by a large loss of stromal fluorescence
at both excitations. Mild to moderate inflammation and dysplasia show a similar drop in
fluorescence at UV excitation (by a factor of >2) compared with normal values. Severe
inflammation displays an even more pronounced decrease (by factor of 4 at UV excitation)
compared with mild to moderate inflammation and dysplasia. The degree of inflammation also
determines the depth of the affected stromal areas. Severe inflammation displays a very large
loss of fluorescence in both stromal regions 1 and 2. In mild to moderate inflammation and
dysplasia, fluorescence from stromal region 2 is higher compared with severe inflammation.
Autofluorescence patterns of well, moderately, and poorly differentiated carcinomas
Figure 5 displays representative confocal fluorescence images of well, moderately, and poorly
differentiated carcinomas, whereas Fig. 6A and B show how images of cancers were divided
into three subregions. A common feature present in all cancers is the absence of a layered
morphology and the aggregation of cancer cells in clearly defined regions (Fig. 6A and B).
Type 1 cancer cells are characterized by an absence of cytoplasmic fluorescence at UV
excitation but a strong cytoplasmic signal at 488 nm excitation. Type 2 cancer cells display an
easily visualized cytoplasmic fluorescence at UV excitation. Cancer cells are surrounded by
matrix fibers. Matrix regions with dense, brightly fluorescent fibers, and those without a
significant cellular component are defined here as fibrous stroma (light blue region in Fig. 6B).
Matrix regions with a dominant cellular component, consisting of inflammation and atypical
cells, are very heterogeneous and difficult to outline. These regions were excluded from the
analysis of cancer images and are not shown in Fig. 6. Some tissue slices contained submucosal
tumors. Figure 5A shows an example of a well-differentiated submucosal tumor underneath a
mildly dysplastic surface epithelium. Cells from the surface epithelium overlying submucosal
tumors were excluded from quantitative analysis of carcinomas.
Pavlova et al. Page 6
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Average FIVs for type 1 and type 2 cancer cells are compared in Fig. 6C and D. The poorly
differentiated tumor shows the highest fluorescence intensity at UV excitation for both cell
types and the lowest fluorescence at 488 nm excitation. Average redox values for type 1 and
2 cancer cells are shown in Fig. 6D. The poorly differentiated carcinoma displays the lowest
redox values, especially for type 2 cells. Because redox values are inversely proportional to
metabolic activity, these results provide support for the clinical tenet that poorly differentiated
carcinoma cells are more metabolically active, on average, than cells in more differentiated
tumors. In addition, average FIVs show that fibrous stroma in well-differentiated tumors
generally have higher fluorescence at both excitation wavelengths when compared with less
differentiated carcinomas.
Discussion
Understanding how optical properties are altered during oral carcinogenesis is critical for
optimizing diagnostic technologies for oral cancer detection based on autofluorescence
imaging and spectroscopy. In this study, we used high-resolution microscopy to investigate
patterns of autofluorescence in normal oral mucosa and in benign and neoplastic oral lesions.
Our results show that the autofluorescence properties of oral tissue vary based on the anatomic
site within the oral cavity and the pathologic diagnosis. The fluorescence signals from epithelial
and stromal layers can change independently of other tissue layers. This has important
implications for the clinical diagnosis of oral lesions using fluorescence imaging and
spectroscopy.
When normal oral tissue is illuminated by UV light, most of the epithelial autofluorescence
that is generated originates from the cytoplasm of cells occupying the basal and intermediate
layers. Similar findings were found using confocal images of cervical tissue, in which the
epithelial fluorescence at UV excitation originates partially from the cytoplasm of
metabolically active cells, and the main fluorophore responsible for this signal was shown to
be NADH (19,29). In normal, nondysplastic tongue and floor of the mouth tissue, the presence
of inflammation within the lamina propria is characterized by a significant decrease in
fluorescence from the lower epithelial layers upon UV excitation. Several investigators have
suggested that in benign lesions such as lichen planus, the large influx of inflammatory cells
under the basement membrane triggers apoptosis and changes the proliferation rate of epithelial
cells, which would also affect the autofluorescence of these cells (31,32). In contrast, epithelial
dysplasia in the tongue exhibited a small increase in epithelial fluorescence at UV excitation.
A similar pattern was previously observed in dysplastic cervical tissue and was attributed to
increased cellular metabolism (21).
In normal oral tissue, stromal fluorescence originates from structural fibers such as collagen.
Indeed, collagen crosslinks are believed to be the dominant fluorophores in normal stroma and
the optical signatures of these crosslinks are quite different from that of epithelial NADH and
FAD (29). With the appearance of inflammation, a large loss of stromal fluorescence was
noticed, especially in areas close to the basement membrane. This trend was observed in
inflammatory tissue, regardless of anatomic site and the degree of inflammation correlates well
with the extent of the fluorescence loss. We speculate that the reduction in stromal fluorescence
in the presence of chronic inflammation was linked to the displacement of structural fibers by
the infiltrating lymphocytes which are much less fluorescent, and also promote the expression
of matrix-degrading proteases leading to the breakdown of collagen crosslinks (27). A study
of the distribution of collagen fibers in human gingiva found that collagen types I and III are
lost in stromal tissue with progression of inflammation (33). In vivo multiphoton microscopy
images of hamster cheek pouch tissue showed that the number and length of collagen fibers
was greatly diminished with increasing severity of inflammation (34).
Pavlova et al. Page 7
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Our results showed a marked loss of stromal fluorescence in dysplastic lesions similar to that
observed in normal mucosa with mild to moderate inflammation. Because most of the
dysplastic samples in this study also had mild to moderate inflammatory infiltrate in the lamina
propria, the reduction of stromal fluorescence is most likely due to the presence of chronic
inflammatory cells. Gannot et al. also documented increasing levels of subepithelial
inflammatory infiltrate (sometimes referred to as lichenoid inflammation) in oral tissue during
progression from normal to dysplasia to carcinoma (35). A second investigation found that
stromal T cells increased by roughly a factor of 2 in mild dysplasias, and by ∼5-fold in moderate
and severe dysplasia compared with normal oral tissue (36).
In vivo fluorescence spectroscopy and imaging evaluations consistently find that oral lesions
display a loss of fluorescence intensity when compared with normal oral tissue. Some
investigators have extracted the intrinsic fluorescence spectra from in vivo fluorescence spectra
in the cervix (37) and the oral cavity (15) with a mathematical model. They have found that in
the cervix and the oral cavity, the intrinsic fluorescence spectra could be composed of two
spectra components, NADH and collagen, and that the NADH contribution increases whereas
the collagen contribution decreases as lesions become more malignant. In both the cervix and
oral cavity, the decrease in collagen contribution was larger then the increase in NADH
contribution. These results are in agreement with the autofluorescence patterns from normal
and dysplastic oral tongue tissue summarized in this study. Moreover, the loss of fluorescence
intensity in oral lesions, as observed in both autofluorescence spectra and images, can be
explained mostly by changes in stromal optical and morphologic properties. Lane et al. attribute
the loss of autofluorescence signal in images of oral precancerous and cancerous lesions
primarily to the breakdown of the collagen matrix and increased hemoglobin absorption and
secondarily to epithelial factors, such as increased epithelial scattering and thickness (6).
Previously, Drezek et al. have shown with Monte Carlo simulations of fluorescence spectra
that most of the in vivo fluorescence signal in cervical tissue (∼80% in normal cervix tissue
and 70% in dysplastic tissue) originates from the stroma. She concludes that the decreased
fluorescence in dysplastic cervical tissue is due more to the reduction of stromal collagen
fluorescence than changes in the contribution from epithelial NADH fluorescence (38).
Here, we show that both inflammatory and dysplastic oral tongue tissue display a large decrease
in stromal fluorescence, especially in the superficial stroma, but have very different
fluorescence patterns in the epithelium. Imaging such lesions with optical devices or probes
that measure mostly stromal fluorescence may result in similar findings of loss of fluorescence
intensity and thus fail to distinguish benign inflammation from dysplasia. Our results suggest
that a possible way to distinguish benign inflammation from dysplastic lesions is to probe
differences in epithelial fluorescence in addition to stromal fluorescence. Moreover, whereas
stromal fluorescence decreases with malignant progression at both UV and 488 nm excitations,
significant differences in epithelial fluorescence are observed only at UV excitation. Thus, the
diagnostic ability of fluorescence imaging and spectroscopy for differentiating benign
inflammations from dysplastic lesions could be improved by using excitation wavelengths in
the UV range. Recently, it was shown that a simple imaging device could be used as an aid to
successfully detect and identify high-risk preinvasive lesions with excitation wavelengths in
the 400 to 460 nm range (6). Although the success of this device is quite exciting, our results
suggest that the accuracy of such imaging devices in distinguishing different types of oral
lesions, such as benign inflammation and dysplasia, might be enhanced by using UV excitation
in addition to higher excitation wavelengths.
Bright autofluorescence was noted from the superficial, keratinized epithelial layer, which is
often present in normal oral tissue from specific anatomic sites such as the palate and the
gingiva, as well as in clinically apparent leukoplakia. This superficial layer is composed of
keratinized cells that have different scattering (23) and fluorescence signatures (39) than the
Pavlova et al. Page 8
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
rest of the epithelium. The presence of a thick, keratinized superficial layer could influence
both the intensity and emission peak of in vivo fluorescence spectra. Muller et al. reported that
fluorescence spectra from keratinized oral mucosa exhibit a shift to the red and a lower intensity
compared with nonkeratinized tissue. They explained these differences by a reduction of the
depth of penetration of light excitation due to scattering from the keratin layer, which results
in an increased NADH and decreased collagen contribution to the measured spectra. Thus, in
order to classify dysplastic tissue from cancers with good accuracy, they advocated that
nonkeratinized and keratinized mucosa should be divided into different groups (15). Our results
support this view, and also suggest that this subdivision would be important for distinguishing
nondysplastic from dysplastic oral mucosa.
The results here suggest that the diagnostic potential of fluorescence spectroscopy and imaging
can be improved by designing optical probes or devices that can selectively measure signals
from either the epithelium or the stroma. Excitation wavelengths in the UV range may also
improve the accurate diagnosis of different types of oral lesions.
Acknowledgments
Grant support: NIH grant R01CA095604.
References
1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005;55:74–
108. [PubMed: 15761078]
2. American Cancer Society. Cancer facts and figures 2006. Atlanta: American Cancer Society; 2006.
3. Betz CS, Mehlmann M, Rick K, et al. Autofluorescence imaging and spectroscopy of normal and
malignant mucosa in patients with head and neck cancer. Lasers Surg Med 1999;25:323–34. [PubMed:
10534749]
4. Paczona R, Temam S, Janot F, Marandas P, Luboinski B. Autofluorescence videoendoscopy for
photodiagnosis of head and neck squamous cell carcinoma. Eur Arch Otorhinolaryngol 2003;260:544–
8. [PubMed: 12774235]
5. Kulapaditharom B, Boonkitticharoen V. Performance characteristics of fluorescence endoscope in
detection of head and neck cancers. Ann Otol Rhinol Laryngol 2001;110:45–52. [PubMed: 11201808]
6. Lane PM, Gilhuly T, Whitehead P, et al. Simple device for the direct visualization of oral-cavity tissue
fluorescence. J Biomed Opt 2006;11:024006-1–7. [PubMed: 16674196]
7. Ramanujam N, Mitchell MF, Mahadevan A, et al. Spectroscopic diagnosis of cervical intraepithelial
neoplasia CIN in vivo using laser-induced fluorescence spectra at multiple excitation wavelengths.
Lasers Surg Med 1996;19:63–74. [PubMed: 8836997]
8. Palmer GM, Zhu C, Breslin TM, Xu F, Gilchrist KW, Ramanujam N. Comparison of multi-excitation
fluorescence and diffuse reflectance spectroscopy for the diagnosis of breast cancer. IEEE Trans
Biomed Eng 2003;50:1233–42. [PubMed: 14619993]
9. Nordstrum RJ, Burke L, Niloff JM, Myrtle JF. Identification of cervical intraepithelial neoplasia (CIN)
using UV-excited fluorescence and diffuse-reflectance tissue spectroscopy. Lasers Surg Med
2001;29:118–27. [PubMed: 11553898]
10. Chang SK, Follen M, Malpica A, et al. Optimal excitation wavelengths for discrimination of cervical
neoplasia. IEEE Trans Biomed Eng 2002;49:1102–11. [PubMed: 12374334]
11. Georgakoudi I, Sheets EE, Müller MG, et al. Trimodal spectroscopy for the detection and
characterization of cervical precancers in vivo. Am J Obstet Gynecol 2002;186:374–82. [PubMed:
11904594]
12. Georgakoudi I, Jacobson BC, Van Dam J, et al. Fluorescence, reflectance, and light-scattering
spectroscopy for evaluating dysplasia in patients with Barrett's esophagus. Gastroenterology
2001;120:1620–9. [PubMed: 11375944]
Pavlova et al. Page 9
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
13. Gillenwater A, Jacob R, Ganeshappa R, et al. Noninvasive diagnosis of oral neoplasia based on
fluorescence spectroscopy and native tissue autofluorescence. Arch Otolaryngol Head Neck Surg
1998;124:1251–8. [PubMed: 9821929]
14. Heintzelman DL, Utzinger U, Fuchs H, et al. Optimal excitation wavelengths for in vivo detection of
oral neoplasia using fluorescence spectroscopy. Photochem Photobiol 2000;72:103–13. [PubMed:
10911734]
15. Muller MG, Valdez TA, Georgakoudi I, et al. Spectroscopic detection and evaluation of morphologic
and biochemical changes in early human oral carcinoma. Cancer 2003;97:1681–92. [PubMed:
12655525]
16. De Veld DC, Skurichina M, Witjes MJ, Duin RP, Sterenborg HJ, Roodenburg JL. Autofluorescence
and diffuse reflectance spectroscopy for oral oncology. Lasers Surg Med 2005;36:356–4. [PubMed:
15856507]
17. Wang CY, Tsai T, Chen HM, Chen CT, Chiang CP. PLS-ANN based classification model for oral
submucous fibrosis and oral carcinogenesis. Lasers Surg Med 2003;32:318–26. [PubMed: 12696101]
18. De Veld DC, Skurichina M, Witjes MJ, Duin RP, Sterenborg HJ, Roodenburg JL. Clinical study for
classification of benign, dysplastic and malignant oral lesions using autofluorescence spectroscopy.
J Biomed Opt 2004;9:940–50. [PubMed: 15447015]
19. Pavlova I, Sokolov K, Drezek R, et al. Microanatomical and biochemical origins of normal and
precancerous cervical autofluorescence using laser-scanning fluorescence confocal microscopy.
Photochem Photobiol 2003;77:550–5. [PubMed: 12812299]
20. Chance B, Thorell B. Localization and kinetics of reduced pyridine nucleotide in living cells by
microfluorometry. Biol Chem 1959;234:3044–50.
21. Drezek R, Brookner C, Pavlova I, et al. Autofluorescence microscopy of fresh cervical tissue sections
reveals alterations in tissue biochemistry with dysplasia. Photochem Photobiol 2001;73:636–41.
[PubMed: 11421069]
22. Collier T, Arifler D, Malpica A, Follen M, Richards-Kortum R. Determination of epithelial tissue
scattering coefficient using confocal microscopy. IEEE J Sel Top Quantum Electron 2003;9:307–
13.
23. Clark AL, Gillenwater A, Alizadeh-Naderi R, El-Naggar AK, Richards-Kortum R. Detection and
diagnosis of oral neoplasia with an optical coherence microscope. J Biomed Opt 2004;9:1271–80.
[PubMed: 15568948]
24. Bhowmick NA, Moses HL. Tumor-stroma interactions. Curr Opin Genet Dev 2005;15:97–101.
[PubMed: 15661539]
25. Zigrino P, Löffek S, Mauch C. Tumor-stroma interactions: their role in the control of tumor cell
invasion. Biochimie 2005;87:321–8. [PubMed: 15781319]
26. Hong WK, Sporn MB. Recent advances in chemoprevention of cancer. Science 1997;278:1073–7.
[PubMed: 9353183]
27. Mignogna MD, Fedele S, Lo Russo L, Lo Muzio L, Bucci E. Immune activation and chronic
inflammation as the cause of malignancy in oral lichen planus: is there any evidence? Oral Oncol
2004;40:120–30. [PubMed: 14693234]
28. Ingber DE. Cancer as a disease of epithelial-mesenchymal interactions and extracellular matrix
regulation. Differentiation 2002;70:547–60. [PubMed: 12492496]
29. Sokolov K, Galvan J, Myakov A, Lacy A, Lotan R, Richards-Kortum R. Realistic three-dimensional
epithelial tissue phantoms for biomedical optics. J Biomed Opt 2002;7:148–56. [PubMed: 11818022]
30. Arifler D, Pavlova I, Gillenwater A, Richards-Kortum R. Light scattering from collagen fiber
networks: micro-optical properties of normal and neoplastic stroma. Biophys J 2007;92:3260–74.
[PubMed: 17307834]
31. Sugerman PB, Savage NW, Walsh LJ, et al. The pathogenesis of oral lichen planus. Crit Rev Oral
Biol Med 2002;13:350–65. [PubMed: 12191961]
32. Montebugnoli L, Farnedi A, Marchetti C, Magrini E, Pession A, Foschini MP. High proliferative
activity and chromosomal instability in oral lichen planus. Int J Oral Maxillofac Surg 2006;35:1140–
4. [PubMed: 17095190]
33. Narayanan AS, Clagett JA, Page RC. Effect on Inflammation on the distribution of collagen types I,
III, IV, V, type I. Trimer and fibronectin in human gingivae. J Dent Res 1985;64:1111–6.
Pavlova et al. Page 10
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
34. Wilder-Smith P, Osann K, Hanna N, et al. In vivo multiphoton fluorescence imaging: a novel approach
to oral malignancy. Lasers Surg Med 2004;35:96–103. [PubMed: 15334611]
35. Gannot G, Gannot I, Vered H, Buchner A, Keisari Y. Increase in immune cell infiltration with
progression of oral epithelium from hyperkeratosis to dysplasia and carcinoma. Br J Cancer
2002;86:1444–8. [PubMed: 11986779]
36. Syafriadi M, Cheng J, Jen KY, Ida-Yonemochi H, Suzuki M, Saku T. Two-phase appearance of oral
epithelial dysplasia resulting from focal proliferation of parabasal cells and apoptosis of prickle cells.
J Oral Pathol Med 2005;34:140–9. [PubMed: 15689227]
37. Georgakoudi I, Jacobson BC, Muller MG, et al. NADH and collagen as in vivo quantitative fluorescent
biomarkers of epithelial precancerous changes. Cancer Res 2002;62:682–7. [PubMed: 11830520]
38. Drezek R, Sokolov K, Utzinger U, et al. Understanding the contributions of NADH and collagen to
cervical tissue fluorescence spectra: modeling, measurements, and implications. J Biomed Opt
2001;6:385–96. [PubMed: 11728196]
39. Wu Y, Qu JY. Combined depth- and time-resolved autofluorescence spectroscopy of epithelial tissue.
Opt Lett 2006;31:1833–5. [PubMed: 16729086]
Pavlova et al. Page 11
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 1.
Fluorescence and histologic images of normal tongue without inflammation. Mosaic of
confocal fluorescence images at UV excitation (A), 488 excitation (B), and H&E staining (C).
Bars, 200 μm (in the confocal images); bars, 125 μm (in the H&E image).
Pavlova et al. Page 12
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 2.
Fluorescence and histologic images of four normal oral sites without inflammation. Mosaic of
confocal fluorescence images at both UV (top row) and 488 nm excitation (middle row), and
H&E image (bottom row) from the palate (A), gingiva (B), floor of the mouth (C), and buccal
mucosa (D). White lines, approximate location of the basement membrane. Bars, 200 μm.
Pavlova et al. Page 13
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 3.
Fluorescence patterns of inflammation (A) and mild dysplasia (B) in the tongue. Mosaic of
confocal fluorescence images at UV (left), 488 nm excitation (middle), and H&E (right)
images. The histopathologic diagnosis of (A) is normal, nondysplastic epithelium with severe
inflammation and of (B) is mildly dysplastic epithelium with mild to moderate inflammation.
White lines, the approximate location of the basement membrane. Bars, 200 μm (in the confocal
images); bars, 120 μm (in the H&E images). Arrowheads, lymphocytic infiltration; N, normal
without inflammation; NSI, normal with severe inflammation; DMI, mild dysplasia.
Pavlova et al. Page 14
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 4.
Quantitative analysis of average fluorescent intensities of normal, inflammatory, and dysplastic
tongue tissue. Overlaid UV and 488 nm excited fluorescence images (A) and a simplified
cartoon of normal tongue (B) showing the approximate distribution of the epithelial and stromal
regions. Type 1 cells represent cells with weak cytoplasmic fluorescence at UV excitation.
Type 2 cells represent cells with strong cytoplasmic fluorescence at UV excitation. Stromal
region 1 includes stroma that is 100 to 200 μm below the basement membrane, whereas stromal
region 2 represents deeper stroma. Average fluorescence intensities at UVexcitation (C) and
488 nm excitation (D) for each epithelial and stromal subregion. Bars, 1 SD. N, normal without
inflammation (n = 8); NMI, normal with mild to moderate inflammation (n = 2); NSI, normal
with severe inflammation (n = 3); DMI, dysplasia with mild to moderate inflammation (n =
6).
Pavlova et al. Page 15
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 5.
Fluorescence patterns of invasive tumors in the oral cavity. Mosaic of confocal fluorescence
images at UV (first column) and 488 nm excitation (second column); overlaid UVand 488 nm
images (third column) and H&E images (fourth column). A, mildly dysplastic epithelium
overlaying well-differentiated submucosal carcinoma in the tongue. B, moderately
differentiated carcinoma in the tongue. C, poorly differentiated carcinoma in the palate. Bars,
200 μm.
Pavlova et al. Page 16
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 6.
Quantitative analysis of invasive tumors. Overlaid UV and 488 nm excited fluorescence images
(A) and a simplified cartoon of an invasive tumor (B) showing the distribution of type 1
(green) and type 2 (dark blue) cancer cells surrounded by fibrous stroma (light blue). Type
1cells represent cells with weak cytoplasmic fluorescence at UV excitation. Type 2 cells
represent cells with strong cytoplasmic fluorescence at UV excitation. Fibrous stroma includes
matrix with a dominant fiber component (see Results for details). Average fluorescence
intensities at UV (C) and 488 nm (D) excitation for each subregion and average redox ratio
(E) values for type 1 and type 2 cancer cells. Bars, 1SD.WDC, well-differentiated carcinoma;
MDC, moderately differentiated carcinoma; PDC, poorly differentiated carcinoma.
Pavlova et al. Page 17
Clin Cancer Res. Author manuscript; available in PMC 2009 November 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
