Subscriber access provided by UNIVERSITY OF CONNECTICUT
Article
Phasor–FLIM as a screening tool for the differential diagnosis of actinic keratosis, Bowen’s disease and basal cell carcinoma Teng Luo, Yuan Lu, Shaoxiong Liu, Danying Lin, and Junle Qu Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b01681 • Publication Date (Web): 29 Jun 2017 Downloaded from http://pubs.acs.org on July 6, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 9
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
Phasor–FLIM as a screening tool for the differential diagnosis of actinic keratosis, Bowen’s disease and basal cell carcinoma Teng Luo†, Yuan Lu‡,*, Shaoxiong Liu§, Danying Lin†, Junle Qu†,* †
Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education and Guangdong Province, College of Optoelectronic Engineering, Shenzhen University, Shenzhen, Guangdong, 518060, China.
‡
Department of Dermatology, The Sixth People's Hospital of Shenzhen, Shenzhen, Guangdong, 518052, China.
§
Department of Pathology, The Sixth People's Hospital of Shenzhen, Shenzhen, Guangdong, 518052, China.
ABSTRACT: The aim of this study was to distinguish basal cell carcinoma (BCC) from actinic keratosis (AK) and Bowen's disease (BD) by fluorescence lifetimes of hematoxylin and eosin (H&E) and phasor analysis. Pseudo-color images of average fluorescence lifetime (τm) exhibited more contrast than conventional bright field and/or fluorescence images of H&E stained sections. The mean values (µ) of τm distribution (τmµ) in three layers of skin were first explored for comparison with the corresponding layers of AK, BD and BCC. Moreover, analysis of the H&E fluorescence lifetimes in the phasor space was performed by observing clusters in specific regions of the phasor plot. Various structures in the skin were distinguished. Comparisons of phase distributions from the corresponding layers of skin resulted in quantitative separation and calculation of distinctive parameters including coordinate values, diagonal slopes and phasor areas. The combination of fluorescence lifetime imaging microscopy (FLIM) and phasor approach (phasor–FLIM) provides a simple method for histopathology analysis, and can significantly improve the accuracy of bright field H&E diagnosis. We therefore believe that phasor–FLIM is an aided tool with the potential to provide rapid confirmation of diagnostic criteria and classification of histological types of skin neoplasms.
Nonmelanoma skin cancers (NMSCs) are the most common tumors in the white population. While the term NMSCs, encompasses a wide variety of cutaneous malignancies, the most frequent subtypes are actinic keratosis (AK), Bowen disease (BD), basal cell carcinoma (BCC) and invasive squamous cell carcinoma (SCC). AK is a pre-cancerous patch of thick, scaly, or crusty skin. BD is a SCC in situ with full-epidermal thickness dysplasia that has the potential for significant lateral spread before invasion. BCC is the most common skin cancer. Infiltrative BCC can present as a skin thickening or scar tissue making diagnosis difficult without using a skin biopsy. AK and SCC can present similarly on physical exam, and many scientists argue that they are in fact simply different stages of the same condition. In addition to BD, AK can be mistaken for other cutaneous lesions including BCC, inflammatory dermatoses, or melanoma. 1 Histopathological feature of biopsies remains the gold standard for the diagnosis of NMSCs. It is based on the morphological interpretation of hematoxylin and eosin (H&E) stained sections. Differential diagnosis among different types of NMSCs often relies on the expertise and skill of pathologists, and can lead to some discrepancies in the interpretation.2 The widely used clinical pathological methods, such as H&E staining, immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) have been gradually developed since being first established. Recently, several emerging techniques, such as Raman spectroscopy,3, 4 mass spectrometry,5, 6 X-ray tomography,7 and Fourier transform infrared spectroscopy (FTIR) 8, 9 have been proposed to replace
pathological biopsy, or realize pathology-assisted imaging. However, because the images obtained from these novel methodologies are different from those obtained from routine diagnostic pathology, complex data sets often require multiple image segmentations in order to reveal their discrepancies, and make their applications in the pathology department challenging. For centuries, histologists have utilized an assortment of bright field microscopy to elucidate functional attributes of tissues through investigating tissue architecture. In comparisons to novel techniques, optical microscopy based on tissue slides is more likely to be accepted by pathologists. Therefore, it is necessary to develop a new method of optical microscopic imaging that could not only incorporate the traditional practice of histopathology to simplify sample preparation, but also aid imaging management to improve the objectivity of pathological diagnosis. In addition, computer processing makes the interpretation of the data automated and less prone to operator errors. Although histotechnologists are probably more familiar with the name H&E, their fluorescence properties are often overlooked. Eosin is usually not regarded as a fluorochrome, but its high fluorescence emission has been described.10 Unlike eosin, hematoxylin is non-fluorescent. Hematein, the oxidation product of hematoxylin, is a weak anionic dye with low fluorescence quantum efficiency.10 H&E stained sections can therefore be viewed directly using fluorescence microscopy.11 In some settings, eosin fluorescence could rapidly provide useful information, without the time delay associated with
1
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
special staining. Some previous studies have utilized the fluorescence intensity of eosin stained tissues for diagnosis of various diseases.12 Analysis of elastic fibers in the skin by fluorescence microscopy is a useful and complementary method that could reveal hidden elastic fibers in H&E stained specimens.13 However, eosin stains proteins nonspecifically,14 so that fluorescence imaging of eosin stained tissues lacks specificity and sensitivity for malignant cells. It is well known that fluorescence analysis, based mainly on measurements of fluorescence intensity, can be influenced by excitation intensity, concentration distribution of dyes, overlapped spectrum signals, and photobleaching, and therefore are rarely used in quantitative investigations.15, 16 Fortunately, the fluorescence signals contain more parameters than just intensity and spectrum. The fluorescence lifetime of the excited state provides an extra dimension of parameter that is specific to the fluorophore and its microenvironment. Fluorescence lifetime imaging microscopy (FLIM) combines the advantages of fluorescence microscopy, by measuring the decay time of fluorescence emission from the local microenvironment of the fluorophores. Compared with conventional fluorescence imaging, FLIM is independent of excitation power, and fluorophore concentration. 17 The excitedstate lifetime of a fluorescent molecule is sensitive to changes in the local microenvironment of the fluorophore and can be used to reveal the spatial distribution of fluorescent molecules and information about their local microenvironment, such as viscosity, pH, hydrophobicity, and the binding of one labeled protein to another. Therefore, FLIM can be potentially correlated with the state of cells and tissues during physiological and/or pathological processes.18 Data analysis of fluorescence lifetime measurements can be performed both in time and frequency domains.17 In time domain, baseline offset as well as optical and electronic noise are often superimposed to the observed decaying wave form. A careful design of the data acquisition and signal averaging instrumentation and a suitable choice of the algorithms used to analyze and process the data are thus required. Frequency domain approach is independent from the signal offset and allows, by means of lock-in techniques, the separation of the fluorescence signal from the noise at the measurement frequency. However, since it requires the measurement of fluorescence during the excitation period, it is highly sensitive to the excitation light “leakage” to the detector, resulting in a measurement error. Whereas the frequency-domain approach can be very sensitive for small lifetime differences and in principle makes it easier to study rapidly decaying compounds, the time-domain approach to FLIM has advantages in the case of a long-lived component and is simple to understand. Usually, the analysis of FLIM data collected in the time domain is performed by fitting the decay at each pixel using multiexponentials. The phasor analysis transforms the histogram of the time delays at each pixel into a pair of sine cosine polar coordinates (phasor) to avoid fitting. Each pixel is then plotted in two-dimensional phasor space (phasor plot). The principal advantages of the phasor analysis, as compared to multiexponential decay fitting, are its requirements of less initial fitting assumptions, iterative calculations, and its provision of a graphical overview of fluorescence decay at each pixel. Moreover, the phasor approach has been well established for the separation of clusters of pixels with distinctly different
Page 2 of 9
lifetimes. Thus, populations having similar lifetimes can be selected in the phasor plot and the fluorescence image can be painted accordingly. The phasor approach to fluorescence lifetime has the potential to simplify the analysis of FLIM data, and offers a fitting free method of separating pixels with different fluorescence lifetimes.19 The fluorescence lifetime and phasor plot have been successfully applied to differentiate cutaneous tumors20 and to quantify melanin distributions in melanoma biopsy slices.21 Most of these studies used unstained tissues, and the detectable laser induced fluorescence was weak due to the low concentration of intrinsic fluorophores. Clinically, diagnosis of NMSCs can be difficult because of the lack of distinctive features. Herein, multiplexed fluorescence lifetimes of nonspecific binding of single fluorescent dyes (H&E) was carried out on a selected set of H&E stained sections to assess the relevance of phasor–FLIM for the differential diagnosis of three types of skin neoplastic lesions. For this purpose, H&E fluorescence intensity and lifetime distributions were applied to pre-processed images to highlight relevant histological structures. The phasor analysis to FLIM was subsequently applied to H&E stained sections for the differential diagnosis of skin neoplastic types based on the clustering of points in the phasor plot.
EXPERIMENT SECTION Sample set. Nine fresh human skin specimens were obtained from nine patients undergoing skin biopsies for routine diagnostic procedures in the Department of Dermatology at the Sixth People's Hospital of Shenzhen. The samples were placed in a standard pathologic transport container covered with ice and then sent to the Department of Pathology. Three consecutive sections were cut from each paraffin block by cryostat microtome with standard histology procedures. A total of 27 H&E stained skin tissue sections from 9 patients were collected from the Sixth People's Hospital of Shenzhen. The histopathological examination of the H&E stained skin sections was performed by a senior pathologist, and the corresponding tissues were identified as follows: AK (3 patients), BD (3 patients), and BCC (3 patients). For each patient, one H&E stained skin section was randomly selected for imaging. This study was performed according to a protocol approved by the Shenzhen Sixth People's Hospital research ethics committee. All patients gave their informed consent for the use of their tissues for medical research. FLIM of H&E stained sections. A time-resolved fluorescence measurement system incorporating a confocal laser scanning microscope (TCS SP2, Leica) and a time-correlated single photon counting (TCSPC) module was used to image the H&E stained sections of AK, BD and BCC. The modelocked Titanium (Ti): Sapphire laser (Coherent Mira 900, 76 MHz, 120 fs) was used for excitation. The laser was tuned to a wavelength of 785 nm for two-photon excitation. The excitation beam was diverted by a dichroic beam splitter to the galvanometer mirrors before being focused on the sample via a scan lens and an objective (63× PL APO CS Leica, NA = 1.4). The fluorescence emission (520–580 nm) was collected using the same objective lens, and detected with a photomultiplier tube (Hamamatsu). For the time resolved setup, a cutting short-pass filter (700 nm, Chroma) was used to block the reflected laser, and the fluorescence signal passed through a
2
ACS Paragon Plus Environment
Page 3 of 9
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
band-pass filter (550±30 nm, Chroma) and was detected with a micro-channel plate photomultiplier tube (MCP-PMT, Hamamatsu) that was mounted to the optional port of the confocal microscope and connected with a single photon counting module (SPC150, Becker & Hickl GmbH) to measure fluorescence lifetime. The average fluorescence lifetime (τm) at each pixel of a 256×256 image was calculated for all stained tissue sections, with bi-exponential components fitting expressed as:
I(t)/I(0) = a1 exp( -t/τ1 ) +a2 exp( -t/τ2 ) (1) where τ1, τ2 and α1, α2 denote the lifetime and amplitude of two different components, respectively. For each image, a distribution of the average lifetime weighted to pixel intensity was determined. The average lifetime was derived according to the following equation: τm= (a1 τ1 +a2 τ2 )/(a1 +a2 ) (2) A pseudo-colored lifetime image for each image was generated by assigning a color to the lifetime value of τm at each pixel. The minimum time channel width of the TCSPC module was 813 fs, and the response time of the whole system was less than 30 ps. The lifetime calculations and fitting were performed using the SPCImage software (Becker & Hickl GmbH, Germany). In fluorescence intensity imaging mode, the HEstained sections were examined under excitation with UV/violet light (340–380 nm). Epifluorescence illumination was focused via an objective (40× PL APO CS Leica, NA = 1.25) and fluorescence passing through a long-pass filter (425 nm, Leica) was collected by the same digital camera (DFC310 FX CCD, Leica) used for bright field imaging. The phasor approach to FLIM data analysis. The fluorescence collected from each pixel of the image was transformed to the Fourier space. The phasor plot, a graphical representation of intensity decays for a FLIM image was constructed. Points in the two-dimensional phasor plot are defined by the values of sine (S) and cosine (G) transforms derived by the following equations: ∞
si,j ሺωሻ=
0 Iሺtሻ sinሺnωtሻ dt ∞
0 Iሺtሻdt
(3)
∞
gi,j ሺωሻ=
0 Iሺtሻ cosሺnωtሻ dt ∞
0 Iሺtሻdt
(4)
where the indices i and j identify a pixel of the image and si,j(ω) and gi,j(ω) are the y and x coordinates of the phasor plot, respectively; ω = 2πf, where f is the laser repetition frequency (i.e.,76 MHz in our experiments); and n is the harmonic frequency. Thus, the fluorescence collects from each pixel of an image was transformed to a point in the phasor plot. The analysis of the phasor distribution is performed by cluster identification. There is a direct relationship between a phasor location and lifetime. Every possible lifetime can be mapped into this universal representation of the decay (phasor plot). All possible single exponential lifetimes lie on the “universal circle”, defined as the semicircle going from point (0, 0) to point (1, 0), with radius 1/2. Point (1, 0) corresponds to τ=0, and point (0, 0) to τ=∞. In the phasor coordinates the single lifetime components add directly because the phasor follows the vector algebra. A mixture of two distinct single lifetime components, each of which lies separately on the single lifetime semicircle, does not lie on the semicircle. Clusters of pixel values are de-
tected in specific regions of the phasor plot. The cluster assignment is performed by taking into account not only the similar fluorescence properties in the phasor plot but also exploiting the spatial distribution and localization in cellular substructures or tissues. Statistics. All data were expressed as mean ± SEM. Statistical analysis was performed using ordinary one-way ANOVA for multiple groups and unpaired t-testing with Welch’s correction for comparison between two groups. The level of significance employed was as follows: significant (*, P< 0.05), very significant (**, P< 0.01), very highly significant (****, P< 0.0001). Data were analyzed with GraphPad software (San Diego, California, USA).
RESULTS AND DISCUSSION Figure 1a (1–3) show the bright field images of H&E stained sections of actinic keratosis (AK), Bowen's disease (BD), and basal cell carcinoma (BCC), respectively. All the same stained areas exhibited fluorescence signal upon UV excitation (Figure 1b), imaged through 425 nm long-pass emission filter. H&E staining is known to be ineffective inquenching of fluorescent background of formalin-fixed, paraffin embedded tissue slides.22 Considering that the low intrinsic fluorophore concentration (paraffin-embedded tissue), unoptimized excitation (785 nm), and unmatched band-pass filter for selective detection of H&E (520–580 nm), the contributors to signals observed in Figure 1b arise from two aspects: autofluorescence (AF) that is produced with lesser contributions from elastin (420–460 nm), collagen (370–440 nm), NADH (450–500 nm, a coenzyme found in all living cells) and keratin (450–550 nm); and fluorochromes produced by H&E dyes. 23 Hematein excited at 785 nm was found to have a less significant effect on the intensity of fluorescence.10 In contrast, eosin appeared to intensify intrinsic fluorescence. In Figure 1b, negligible fluorescence (a region of almost non-fluorescent) was observed in the nucleus of epithelial cells (ECs), whereas the stratum corneum (SC), cytoplasm and dermal connective tissue (CT) showed yellow-green fluorescence with emission maxima at ~550 nm (Figure S1). The absorption and emission maxima of eosin in alcoholic solution are 527 and 550 nm, respectively. We therefore believed that the spectral feature can be mainly attributed to eosin.24 Fluorescence imaging of eosin stained tissues lacks specificity and sensitivity for malignant cells. Thus, it is difficult to separate the data corresponding to the emissions from different sources (see Figure S1). Generally, single photon or two photon induced fluorescence from endogenous fluorophores has been used to distinguish different tissue constituents of skin. The laser source and detection channels have to exchange to get different AF signals.25 Compared to the endogenous signals of skin, the process by which H&E are directly imaged by one or two photon excited fluorescence is simpler than that by which AF and second harmonic generation are acquired from skin.26, 27 Figure 1c and 1d illustrate the fluorescence intensity images of the three types of skin neoplastic lesions and their corresponding FLIM images (encoding average fluorescence lifetime, τm, using a false color scale). In Figure 1c, the images based on H&E fluorescence are sufficient for morphometric evaluation of the tissue in cases of neoplastic growth. Howev-
3
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
er, indiscriminative staining of H&E makes its fluorescence differentiation nonspecific.
Figure 1. Four optical imaging approaches used to study the routine H&E stained sections. From top to bottom, (a1–a3) show bright field histological slides of actinic keratosis (AK), Bowen's disease (BD), and basal cell carcinoma (BCC) stained with H&E; (b1–b3) show photography based on visible fluorescence induced by UV radiation captured at 40× magnification using Leica TCS SP2 fluorescent microscope equipped with a digital camera and acquired directly from the same field of view as that of Figure 1a; (c1–c3) show fluorescence intensity images of H&E upon twophoton excitation; and (d1-d3) show fluorescence lifetime images of H&E, captured at 63× magnification using the same microscope equipped with a photomultiplier tube (PMT). The corresponding continuous color coding scheme ranges from 0 ps (red) to 450 ps (blue). Scale bar, 50 µm; (e) Histogram of average fluorescence life times (τm) in H&E stained sections of AK, BD, and BCC, constructed by fitting fluorescence decay using biexponential components. The black, pink, and purple curves represent the H&E τm of AK, BD, and BCC, respectively.
Alterations in the label's immediate environment provide alternative decay paths for the excited fluorophore, changing its lifetime. Fluorescence lifetime measurements can therefore reveal cellular changes that result from malignant transformation.18, 28 The pseudo-color τm maps (Figure 1d) were taken from the same field of view as that of Figure 1c. The stromal region is quite distinct from cell-rich epithelium. In Figure 1d (1–3), the alterations of dermal CT with various orientations can be clearly identified by the pseudo-color of the lifetime images. In AK, the CT showed a linear structure, with long, straight fibrils, whereas the CT of BD and BCC exhibited a
Page 4 of 9
loss of fine structure and organization. In the epithelium of each lesion type, the ECs were of varying sizes, distributed in multiple layers, and showed signs of cellular atypia and proliferation. Compared to the H&E fluorescence intensity images, the lifetime images (τm) could provide additional contrast that can be extracted to study tissue pathology. Figure 1e is the τm histogram that corresponds to Figure 1d, in which the τm of AK, BD, and BCC exhibits two peaks: the peak of smaller τm in Figure 1e originating from epidermis and the longer one from H&E stained CT. The short fluorescence lifetimes in the range 0–450 ps (Figure 1e) confirm that long fluorescent lifetimes of endogenous fluorophores (from ~1 ns up to ~2 ns),17 as the weak background signal, do not produce an effect or change in the lifetime distribution of each lesion with H&E staining. Moreover, an overlap was observed between τm histograms for pixels representing BCC and those representing AK and BD. This degree of separation among the τm histograms results in the visual contrast on pseudo-color lifetime maps. Previous studies used the pseudo-color lifetime maps to quantitatively distinguish BCC from surrounding uninvolved skin by area under curve (AUC) of a receiver operating characteristic (ROC) curve.20 Other studies have used fluorescence lifetime images of eosin for the quantitative diagnosis of threetiered cervical intraepithelial neoplasia (CIN) classification (CIN 1, 2, and 3).29 As shown in Figure 1e, H&E in the dermis exhibits a longer τm than that in the region of the cell-rich epithelium. The number changes of ECs and CT in the regions of interest could cause τm to change. However, the aforementioned studies used the whole lifetime of ECs and CT in skin, and disregarded their differences. Thus, we first propose that H&E τm among SC, ECs and CT should be compared separately with the respective corresponding layers of skin. Figure 2 shows lifetime maps obtained from three layers of each lesion type and their corresponding τm histograms. Figure 2a (1–4) display the SC that is still attached to the epithelium. Figure 2b (1–4) correspond to ECs from the epithelium and Figure 2c (1–4) show CT in dermis. Each image was processed independently, thus the pseudo-color lifetime maps do not permit the direct comparison of the morphological structures associated with identical colors in different τm histograms. The SC and ECs in the BD sample were each found to have a longer τm in the range from 100 to 400 ps, and 100 to 300 ps, respectively, whereas the CT in BCC had a shorter τm in the range of 120 to 240 ps. On the other hand, the three layers of BCC showed smaller mean value of τm histograms (τmµ) than the corresponding layers of AK and BD. In Figure 2d, the differences between the τmµ of SC in BCC and AK were shown to be very highly significant according to the Welch’s t test (P
