A Fluorescence Lifetime Imaging Microscopy (FLIM ...

1 downloads 0 Views 840KB Size Report
In this study, the fluorescence lifetime of haematoxylin and eosin (H&E)-stained tissues were investigated using a wide-field time-domain FLIM system.
A Fluorescence Lifetime Imaging Microscopy (FLIM) System for the Characterization of Haematoxylin and Eosin Stained Sample U. S Dinish a, C.Y Fub*, B.K Ngb , T.H Chowb, V.M Murukeshan a, L.K Seah a, and S.K Tanc a

School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore b School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore c Singapore General Hospital, Singapore ABSTRACT

We present the implementation of a fluorescence lifetime imaging microscopy (FLIM) system for cellular characterisation. FLIM system can be used as an investigative tool to identify minor biochemical changes in cellular abnormalities. These subtle changes could possibly alter cellular fluorescence properties such as emission wavelength and lifetime. In this study, the fluorescence lifetime of haematoxylin and eosin (H&E)-stained tissues were investigated using a wide-field time-domain FLIM system. The transient response of epithelial fluorescence was investigated and the lifetime extracted using a bi-exponential model. It was found that the fluorescence lifetimes of eosin can be correlated to the tissue histology. The preliminary result suggests that tumor-associated molecules are retained in the tissues even after tissue fixation and staining. The developed FLIM system was successfully applied to detect the histological changes in the tissue samples. Optimization of system parameters is also discussed. Keywords: FLIM, fluorescence lifetime, haematoxylin, eosin, tumor, histology

1. INTRODUCTION Fluorescence spectroscopy has been used as one of the most promising investigative tool since long time ago to understand the properties of different organic molecules. As we know, the basic molecules of life, like amino acids, proteins, lipids and other complex bio-organic molecules often emit well defined fluorescence spectra upon excitation with suitable ultraviolet or visible light. These molecules exhibit overlapping spectra, which prevents the discrimination of various fluorophores. The mapping of the fluorescence lifetime allows to discriminate different fluorophores and also to achieve valuable insights into the behaviour of emitting species [1] Fluorescence lifetime imaging microscopy (FLIM) is the functional imaging methodology that can provide the chemical information, not only concerning the localization of specific fluorophores, but also about the alteration of their local environment[2, 3]. Unlike the steady-state measurement, fluorescence lifetime measurement reveals the transient response of the fluorescence process independent of the fluorophore concentration and photobleaching. FLIM technique can provide greater sensitivity and specificity in various biomedical diagnostic applications. In medicine, fluorescence imaging has often been proposed for various applications such as detection of tumours [4] and atherosclerotic plaques [5, 6].

*

Correspondent: Fu Chit Yaw; Email: [email protected] : Dinish U. S ; Email : [email protected]

Imaging, Manipulation, and Analysis of Biomolecules, Cells, and Tissues VI, edited by Daniel L. Farkas, Dan V. Nicolau, Robert C. Leif, Proc. of SPIE Vol. 6859, 68590C, (2008) · 1605-7422/08/$18 · doi: 10.1117/12.764412 Proc. of SPIE Vol. 6859 68590C-1 2008 SPIE Digital Library -- Subscriber Archive Copy

In many cases, the perturbation of cellular environment attributed to cancer development is of great interest. In such cases, FLIM system has been employed to access the metabolism and biochemical interaction within tumor tissues[7-9]. Exogenous fluorophores have also been researched to selectively stain the cellular organelles for fluorescence lifetime imaging [10-14]. Most cancers such as cervical cancer develop in the epithelium due to the frequent exposure to various forms of physical and chemical stimulants. Spectroscopic studies have reported that the neoplastic progression of cervical tissue is accompanied with the variation of epithelial autofluorescence[15, 16]. The drawback is that the diagnosis is limited to the autofluorescence assessment of fresh tissues. Currently, screening for cervical cancer and its precursors is commonly accomplished using the Papanicolaou (Pap) smear, which is unable to achieve a concurrently high sensitivity and high specificity [17]. An abnormal Pap smear is followed by a colposcopy, where biopsies are excised from suspicious areas of the cervix for definitive diagnosis of cancer. Haematoxylin and Eosin (H&E) is a common dye used to enhance the color contrast of biopsies for histopathologic examination. Changes in the nuclei of the epithelial cells are the most important indicators of dysplasia[18]. Pathologists have traditionally used the atypical nuclear features in the epithelium as one of the major diagnostic criteria for dysplasia. Interestingly, it has been reported that strong fluorescence emission was observed from the cells stained with H&E[19, 20] and Pap stain[21]. The aim of this study is to characterize the fluorescence properties of H&E-stained cervical tissues for classifying the stage of cancer. In present study, the FLIM system was developed to detect the abnormalities of tumor that is undetected using a bright-field microscope. This proposed screening method can be potentially applied in conjunction with the conventional screening since it does not require any change in the tissue preparation.

2. METHODOLOGY 2.1

Sample set

Four sets of H&E-stained cervical tissues were collected from Singapore General Hospital. Histological examination shows that there are one case of grade 1 Cervical Intraepithelial Neoplasia (CIN I) and three cases of grade 3 CIN III tissues. Ten regions of normal and cancerous epithelium were selected from each sample for the tissue characterization using the FLIM system. 2.2 Setup of the FLIM System A schematic diagram of a wide-field FLIM system is shown in Fig.1. A pulsed laser diode (LDH 400, PicoQuant, Germany) emitting at a wavelength of 404nm was coupled to an inverted fluorescence microscope (TS100, Nikon, USA). The power of the laser diode was 1.2mW at a repetition rate of 40 MHz. A dichroic mirror (FF409-Di01, Semrock, USA) was used to guide the light pulses to the specimen for fluorescence imaging while transmitting halogen light for bright field imaging. The epithelial layer of specimens was located with bright field observation before a fluorescence image was taken. Upon optical excitation, the laser-induced fluorescence emission was then collected via a 20x objective lens (CFI Plan Fluo 20x, 0.45 NA, Nikon, USA) and transmitted through a 500nm-long pass filter (500FH90-50S, Andover Corporation Optical Filter, USA) that blocks the excitation light. The filtered light is directed to the CCD port of the microscope for fluorescence imaging. Subsequently, the transient response of the specimen was recorded using a cooled time-gated ICCD camera (Picostart HR, La Vision, Germany) with a resolution of 1376 x 1040 pixels. A delay unit (Ps Delay Unit, Kentech Instrments Ltd.) was used to temporally gate the gain of the ICCD camera in an increment step of 50ps with respect to the arrival of laser pulses. Forty-two images were recorded to represent the temporal response of the specimens. In order to increase the signal to noise ratio, the detected signal in each time gate was integrated over a time period of 1.6s. After a measurement of the sample, the long pass filter was removed from the optical path. Finally, Instrumental response function (IRF) of the whole imaging system was measured using attenuated excitation beam to prevent from saturating the ICCD camera.

Proc. of SPIE Vol. 6859 68590C-2

Halogen Lamp

Dichroic mirror Pulsed Laser Diode

Attenuator Specimen 20x Objective lens

Delay unit

Long pass filter

ICCD Controller

ICCD

Figure 1 Schematic diagram of FLIM system

2.3 Optimization of ICCD camera An ultra-fast gated ICCD camera was used to capture fluorescence lifetime images by sampling method. Incident photons are detected using an intensifier coupled to a CCD camera. The camera is cooled to -12°C to minimize the dark noise. The time-gated imaging can be realized by modulating the electron gain of the intensifier. A high rate imager (HRI, Kentech Instruments, Oxfordshire, UK) is operated to control the electron gain by setting a particular voltage across the photocathode. A negative voltage pulse is delivered to trigger time-gated detection with a nominal gate width ranging from 200 to 1000ps. Various gate widths have been calibrated in factory for the use of mode-locked lasers operating at a repetition rate of 80MHz. It is stated that the gate width is sensitive to the repetition rate of the light source. A recalibration of the ICCD camera is therefore necessary if the light source is not pulsed at 80MHz. In this study, a picosecond-diode laser is applied to excite the samples at a maximum frequency of 40MHz. The gate width is recalibrated and optimized to achieve a narrow instrumental response function for the lifetime imaging. The gate width is primarily governed by the width of the voltage pulse applied to the photocathode. The design of the HRI allows the users to tune the gate width by adjusting the average voltage Va and the clamp voltage Vc of the pulse triggering circuit. In this calibration, a blue diode laser is used to deliver light pulses at a repetition rate of 40MHz. The laser beam is attenuated prior to light delivery to the ICCD camera. The IRFs are then measured as a function of input voltages of pulse triggering circuit. Figure 2 shows the representative IRFs measured at different voltage settings. Table 1 presents the FWHM and the peak count of each measured profile. It can be seen that an increase in the average and clamp voltage reduces both the FWHM of IRF and the measured intensity. Apparently, there is a trade off between the temporal resolution and sensitivity of the system. It is known that the autofluroescence from cell is typically weak, and

Proc. of SPIE Vol. 6859 68590C-3

therefore the sensitivity of camera becomes a key issue. A gate width of about 200ps should be adequate for fluorescence lifetime measurement of typical biological fluorophores. Thus, the average and clamp voltage will be set at 21.2V and 6.2V respectively throughout the experiments. With the selected settings, IRF was re-measured with the smallest temporal resolution of 25ps, as depicted in Fig. 3. The FWHM of the time profile is determined to be about 209ps. It should be noted that this profile does not reflect the true gate width. The measured gate width can be approximated by

τ IRF =

(τ camera )2 + (τ pulse )2

(1)

where τIRF is the FWHM of measured pulse, τcamera is the FWHM of camera gate, and τpulse is the FWHM of light pulse. Therefore, the measured FWHM is contributed by both the laser pulse width and camera gate width. The pulse width of laser diode has been determined to be about 57ps. Subsequently, Eq.1 is applied to compute the true width of Comb gate, which is about 201ps.

Vav=20.2V Vcl=5.2V Vav=21.2V Vcl=6.2V Vav=22.2V Vcl=7.2V

250

Intensity (a.u)

200

150

100

50

0 4750

5000

5250

5500

5750

6000

6250

6500

Time (ps)

Figure 2 IRF profiles measured at three different settings of average and clamp voltages.

Table 1: FWMH and peak count of IRFs measured at three different settings Average Voltage (V) 20.2 21.2 22.2

Clamp Voltage (V) 5.2 6.2 7.2

FWHM (ps) 315 231 158

Proc. of SPIE Vol. 6859 68590C-4

Peak Count (a.u) 260 154 28

Vav= 21.2V Vcl = 6.2V ∆ t = 25ps

Intensity (a.u)

150

100

50

0 4800

5000

5200

5400

5600

5800

6000

6200

Time (ps)

Figure 3 IRF profile measured at temporal resolution of 25ps

2.4 Data processing and analysis Two-gate rapid lifetime determination (RLD) algorithm is commonly used to provide a good contrast in fluorescence lifetime images due to its fast computation[22]. Using this method, single fluorescence lifetime is estimated on a pixelby-pixel basis with assumption of negligible narrow IRF. In this study, an accurate estimation of fluorescence lifetimes of sample was implemented using non-linear least square method. While a number of non-linear minimization methods are available, Marquartdt-Levenberg method is used as a benchmark for this study due to its popularity[23, 24]. A multiexponential model is used to describe the intrinsic fluorescence decay F(t) at time t as depicted in Eq. 2. m

F (t ) = ∑ α k e

−t

τk

(2)

k =1

where, m is the number of fluorescence components, and αk and τk are the decay amplitude and fluorescence lifetime of kth component respectively. With consideration of instrumental effect, F(t) is convoluted with the instrumental response function h(t), yielding the observed fluorescence decay R(t) (see Eq.3) .

R(t ) = F (t ) ⊗ h(t )

(3)

In non-linear least square fitting, the estimated parameters, i.e. αk and τk, are iteratively adjusted until a best fit between the estimated decay Rc(ti) and experimental decay R(ti) is obtained. For FLIM analysis, 10 regions of epithelium were selected from each microscopic image. With the fitted result, the mean and standard deviation of τk were computed for differentiation of tissue histology.

Proc. of SPIE Vol. 6859 68590C-5

3. RESULTS The absorption and fluorescence emission spectra of the H&E-stained tissues were first measured, as presented in Fig. 4. The result shows that the specimen exhibits a maximum absorption at 532 nm, whereas its fluorescence emission peaks at 555nm. There is no observable change in spectra among the specimens. The epithelial layer of the specimens was located using a bright field microscope for the fluorescence studies. As shown in Fig.5, the bright field image clearly delineates the cell-rich epithelium and the underlying stroma of the cervical tissues. The wavelength-dependent absorption of the H&E staining provides a color contrast to visualize the cell membrane, nuclei and other intracellular components. On the other hand, similar microstructure of tissue could be identified from the fluorescence images. The stroma was found to emit stronger fluorescence emission than the epithelial layer. However, negligible signal could be measured from the nuclei region of epithelial cells.

1.0

Normalized Intensity

0.8

0.6

0.4

0.2

0.0 400

450

500

550

600

650

700

Wavelength(nm)

Figure 4 Absorption (○) and fluorescence emission (□) spectra of H&E-stained cervical tissue.

S

S

E

E

;

F

___

m

(b)

(a)

Figure 5 (a) Bright field image and (b) fluorescence images showing the epithelium (E) and stroma (S) layer of H&E-stained cervical tissue.

Proc. of SPIE Vol. 6859 68590C-6

In this study, IRF of the imaging system was measured to determine the fluorescence lifetimes of the specimens. The configuration of the whole system was kept constant throughout the experiments. Ten regions of epithelium were sampled to represent the characteristic fluorescence lifetime of each specimen. Figure 6 illustrates a representative data of IRF and fluorescence decay measured from a section of tissue. Iterative deconvolution and model fitting were then performed on the extracted data. The fitted result shows that the fluorescence decay was dominated by the short lifetime component. To assess the lifetime distribution of a given tissue, the mean µ and the standard deviation σ of fitted lifetimes were computed from the selected regions as tabulated in Table 2. The preliminary result indicates that the fluorescence lifetime of fast decay component decreases as the cervical tissue progress from normal to CIN III, but there is no significant change in standard deviation of its lifetime distribution. Furthermore, the cancerous progression is accompanied with an increase in fluorescence lifetime and standard deviation in slow decay component

1.0

Normalized Intensity

0.8

0.6

0.4

0.2

0.0 2400 2600 2800 3000 3200 3400 3600 3800 4000 4200 4400

Time(ps)

Figure 6 Typical data of IRF (-□-) and fluorescence decay profile (-○-) of a H&E-stained tissue.

Table 2: Fitted Fluorescence Lifetime Of H&E-Stained Cervical Tissue Histopathology Normal CIN I CIN III

τ1(ps) µ1 114 86 77

τ2(ps) σ1 17 15 19

Proc. of SPIE Vol. 6859 68590C-7

µ2 543 682 733

σ2 65 127 196

4. DISCUSSION Endogenous fluorescence of tissues has been reported as a metabolic indicator applicable to cancer diagnosis[7-9, 16]. The drawback is that fresh tissues are required to conduct the studies of the intrinsic fluorescence characteristics. In this study, FLIM was applied to evaluate the diagnostic information obtained from the fluorescence emission of H&E stained tissues. H&E stain is a common dye used in animal histology and routine pathology [25]. It is known that hematoxylin stains the nuclei in purplish blue, whereas the eosin shows the extracellular and intracellular protein in pink. Upon laser illumination, the fluorescence images were taken and correlated to the bright field images to understand the sources of fluorescence emission. The observation suggests that the epithelial fluorescence is originated from the extracellular and intracellular components, except for the nuclei. Eosin has been found to be the main fluorescent molecules in the H&E-stained tissues [21]. It has been reported that the spectral properties of eosin are varied upon binding with protein[26]. At increasing concentration of protein, the absorption peak of eosin is shifted from 516nm to 536nm, whereas the emission peak varies from 544nm to 556nm. The measured spectra of the stained slides are therefore in good agreement with the reported findings of protein-bounded eosin. Fluorescence lifetime measurement was subsequently carried out to characterize the protein-bounded eosin resided in the tissues because there is no observable change in spectral properties of the specimens. The deconvoluted fluorescence decay of the specimens shows that the fast decay component dominates the fluorescence emission and exhibits a smaller fluorescence lifetime value as compared to the FWHM of the IRF of the imaging system. It is therefore necessary to implement deconvolution in the fitting so as to obtain the diagnostic information of short lifetime component. Interestingly, there is a pronounced contrast in both short and long fluorescence lifetime of the protein-bounded eosin, thereby providing a mean to classify the tissue histology (refer to Table 2). From these preliminary observations, it is postulated that the tumor-associated molecules are retained in the cancerous tissue even after processes of tissue fixation. FLIM system could be used to detect the differentiated interactions between eosin and the tumor-associated molecules in the specimens. FLIM system is therefore useful to provide diagnostic information extracted from Eosin while white light microscope is applied to observe the cellular morphology for histopathological examination.

5. CONCLUSIONS FLIM was used to investigate the fluorescence characteristics of H&E-stained cervical tissues. Spectral measurement was first implemented to substantiate the distribution of protein-bounded eosin around the intracellular and extracellular region of the cervical epithelium. The temporal response of the fluorescence emission was then recorded using the timegated ICCD camera. The promising result shows that there is a significant discrimination in fluorescence lifetimes between the normal and cancerous epithelium. The preliminary study suggests that the tumor-associated molecules are retained in the cancerous tissue even after the tissue fixation and staining. Hence, the proposed methodology can potentially be applied to supplement the histopathologic screening with the advantage of conserving the protocol of H&E staining.

REFERENCES 1. 2. 3. 4. 5.

Cubeddu, R., et al., Time-resolved fluorescence imaging in biology and medicine. Journal of Physics D: Applied Physics, 2002. 35(9): p. R61-R76. Lakowicz, J.R., Principles of Fluorescence Spectroscopy. 2nd ed. 1999: Plenum Pub Corp. Mycek, M.-A. and B.W. Pogue, Handbook of Biomedical Fluorescence. 2003: Marcel Dekker Inc. 211-236. Lam, S., et al., Detection and localization of early lung cancer by fluorescence bronchoscopy. Cancer, 2000. 89(11 SUPPL.): p. 2468-2473. Baraga, J.J., et al., Characterization of the fluorescent morphological structures in human arterial wall using ultraviolet-excited microspectrofluorimetry. Atherosclerosis, 1991. 88(1): p. 1-14.

Proc. of SPIE Vol. 6859 68590C-8

6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

Verbunt, R.J.A.M., et al., Characterization of ultraviolet laser-induced autofluorescence of ceroid deposits and other structures in atherosclerotic plaques as a potential diagnostic for laser angiosurgery. American Heart Journal, 1992. 123(1): p. 208-216. H.Schneckenburger, et al., Autofluorescence lifetime imaging of cultivated cells using a UV picosecond laser diode. Journal of Fluorescence, 2004. 14(5). C. B. Talbot, et al., Application of hyperspectral fluorescence lifetime imaging to tissue autofluorescence: arthritis. Proc. SPIE, Aug 2005. 5862: p. 182-187. H.Schneckenburger, et al., Fluorescence lifetime imaging of unstained tissues: early results in human breast cancer. J Pathol, 2003(199): p. 309-317. Zhong, W., et al., Sensing cellular function and molecular activity in vivo using fluorescence lifetime imaging microscopy. Proceedings of SPIE, 2005. 5864. H.Schneckenburger, et al., Fluorescence lifetime imaging of rhodamine 123 in living cells. Photochem Photobiol Sci., 2004. 3(1): p. 127-131. P.Taroni, et al., Time-resolved microspectroflurimetry and fluorescence lifetime imaging of hypericin in human retinal pigment epithelial cells. 2005. 81(3): p. 524-528. Kress, M., et al., Time-resolved microspectrofluorometry and fluorescence lifetime imaging of photosensitizers using picosecond pulsed diode lasers in laser scanning microscopes. J Biomed Opt., 2003. 8(1): p. 26-32. Ruck, A., et al., Fluorescence lifetime imaging in PDT.An overview. Medical Laser Application, 2005. 20: p. 125-129. Drezek, R., et al., Understanding the contribution of NADH and collagen to cervical tissue fluorescence spectra: Modeling, measurements, and implications. J Biomed Opt., 2001. 6(4): p. 385-396. Chang, S.K., et al., Analytical model to describe fluorescence spectra of normal and preneoplastic epithelial tissue: comparison with Monte Carlo simulations and clinical measurements. J Biomed Opt., 2004. 9(3): p. 511-22. Mirabal, Y.N., et al., Reflectance spectroscopy for in vivo detection of cervical precancer. J Biomed Opt., 2002. 7(4): p. 587-594. Thiran, J.P. and B. Macq, Morphological feature extraction for the classification of digital images of cancerous tissues. IEEE Trans Biomed Eng. . 1996. 43(10): p. 1011-20. Dirk, C., M.Elston, and F.S. Houston, Medical Pearl: Fluorescence microscopy of hematoxylin-eosin-stained sections. J Am Acad Dermatol., 2002. 47(5). Jakubovsky, J., et al., Fluorescence of hematoxylin and eosin-stained histological sections of human spleen. Acta Histochem., 2002. 104(4): p. 353-6. Steenkeste, K., et al., Cellular autofluorescence imaging for early diagnosis of cancers. Proc. SPIE, 2005. 5862: p. 188-195. Sharman, K.K., Error analysis of the rapid lifetime determination method for double-exponential decays and new windowing schemes. Analytical Chemistry, 1999. 71(5): p. 947-952. Badea, M.G. and L. Brand, Time-resolved fluorescence measurements. Methods in Enzymology, 1979. 61: p. 378-425. Enderlein, J. and R. Erdmann, Fast fitting of multi-exponential decay curves. Optics Communications, 1997. 134(1-6): p. 371-378. Burkitt, H.G., B.Young, and J.W.Heath, Wheater's Functional Histology: A Text and Colour Atlas. 3rd ed. January 2000: Churchill Livingstone. Liliana Birla, A.-M. Cristian, and M. Hillebrand, Absorption and steady state fluorescence study of interaction between eosin and bovine serum albumin. Spectrochim Acta A Mol Biomol Spectrosc., 2004. 60(3): p. 551-6.

Proc. of SPIE Vol. 6859 68590C-9